Expression of Integrin α5β1 and MMPs Associated with Epithelioid Morphology and Malignancy of Uveal Melanoma

Abstract

purpose. Altered expression of the α5β1 integrin and matrix metalloproteinases (MMPs) is recognized as a hallmark of invasive tumor cells. The purpose of the present study was to investigate the expression of integrin subunit α5, its corresponding ligand fibronectin (FN), and the expression pattern for MMPs in four highly proliferative human choroidal melanomas (TP17, TP31, SP8.0, and SP6.5) to evaluate whether any correlation can be established between these markers and cell tumorigenicity.

methods. Cell tumorigenicity was evaluated by subcutaneous injection of uveal melanoma cell lines in immunodeficient nude mice. Anchorage dependency was evaluated by growth assays in soft agar. The invasive ability of each cell type was also determined using a modified Boyden chamber. Expression of both the α5 integrin subunit and FN was determined at the mRNA level by RT-PCR. The protein level (for α5) was determined by flow cytometry and inhibition of adhesion assays by using an antibody directed against the α5 subunit. Expression of MMPs was determined by standard gelatin zymography.

results. Assays in nude mice provided evidence that the cell lines possess a range of tumorigenic ability of TP17>TP31>SP8.0>SP6.5. Antibody inhibition of cell adhesion and flow cytometry demonstrated that TP17 cells have no detectable membrane-bound α5β1, whereas low levels are found in primary cultured melanocytes, as well as in SP6.5, SP8.0, and TP31 cells. RT-PCR analyses provided evidence that both FN and α5 expression may be regulated at the transcriptional level. Gelatin zymography revealed that all cell lines, as well as normal melanocytes, express MMP-2 at varying levels but that only the highly invasive TP17 cell line secretes a distinctive MMP with a high molecular weight of 117 kDa.

conclusions. Among the four melanoma cell lines selected for the completion of this study, TP17 exhibited the most aggressive phenotype, which also correlated with the mostly epithelioid morphology of these cells. The cell morphology of the TP17 cell line could be related to the loss ofα 5β1, whereas its invasive properties are more likely related to the expression of the 117-kDa MMP.

Uveal melanoma is recognized as the most common intraocular malignant tumor in the adult population. ^{1 2} According to the latest data from the Collaborative Ocular Melanoma Study (COMS), ^{3 4} up to 62% of patients in whom a large primary uveal melanoma is diagnosed die or have metastasis within the next 5 years. Based on the database of 4726 uveal melanoma cases (accumulated by the Ophthalmic Pathology Department of the Armed Forces Institute of Pathology, Washington, DC), patients with uveal melanoma have a median life expectancy of 6.5 years. ⁵  

Prognostic variables commonly used to evaluate survival time are tumor and nucleolar sizes and cellular morphology. The presence of cells with large nucleoli is also related to a poor prognostic outcome. ⁵ However, the Callender classification still provides the most significant prognostic information. Callender has classified uveal melanoma in two morphologic classes: spindle and epithelioid. ^{6 7} Uveal melanoma with an epithelioid-like morphology are usually associated with a poor prognosis. ⁷ In 90% of the cases, metastatic disease progresses to the liver and causes death. ⁵ Because there is no lymphatic circulation in the eye, ⁵ uveal melanoma disseminates to secondary organs through the hematogenous route. Because metastatic disease could appear as late as 12 to 15 years after enucleation of the ocular melanoma, ⁵ it is likely that tumor cells access the bloodstream in the early stages of tumor growth to invade secondary organs where they remain in a dormant state for many years before they progress into metastatic disease. 

The exact molecular events involved in the hematogenous invasion of uveal melanoma still remain obscure. However, similar to other types of cancers, the invasion process must require loss of anchorage-dependent growth, degradation of the various extracellular matrix (ECM) components, and movement of the cell body. Attachment of the cell to the ECM is essentially mediated by membrane-bound receptors that belong to the integrin family, whereas remodeling of the ECM is mostly controlled by a large family of ECM-degrading enzymes, the matrix metalloproteinases (MMPs). MMPs are zinc-dependent endopeptidases involved in ECM modeling related with normal physiological processes such as wound healing, inflammation, and embryogenesis, as well as such pathologic processes as arthritis and cancer metastasis. ⁸ Integrins are membrane-bound heterodimers made up of an α and a β subunit held together through noncovalent interactions. ^{9 10} The action of MMPs coupled with adhesion molecules, such as members of the integrin family, are required to ensure the dissemination of cancer cells into both the bloodstream and the target tissues. 

The relationship that exists between α5β1, FN, and MMP expression is of particular interest. Indeed, there is experimental evidence that FN, through its interaction with its corresponding integrin α5β1, could regulate both the activity and expression of MMP-2 ¹¹ and MMP-9. ^{12 13 14 15} MMP-2 and MMP-9, also known as type IV collagenases of 72- and 92-kDa, respectively (or gelatinases A and B), are particularly interesting, in that both can degrade collagen type IV and V, which are essential components of the basal membrane. Aberrant expression of these proteinases could then be viewed as a major benefit to cancer cells, because such proteolytic activities are known to participate in the rearrangement of the ECM, accelerate angiogenesis, and participate in the extravasation and intravasation processes. These effects on the ECM have already been shown to occur in monocyte–macrophage cell lineages and are required during the migration and inflammatory events normally mediated by these cells. ^{16 17 18} In a study performed on 15 distinct cultures of ocular melanoma, Cottam et al. ¹⁹ provided evidence that the cells all expressed MMP-2 and that most also expressed MMP-9. More recently, immunohistologic detection of MMP-2 was proposed as a prognostic marker to predict the metastatic potential of uveal melanoma, in as much as 49% of the uveal melanoma tissue examined was positive for MMP-2 and most exhibited a nonspindle morphology. ²⁰  

Although α5β1 was shown to positively influence tumor invasiveness by its role in MMP expression, decreased expression of α5β1 integrin has also been associated with the progression of transformed phenotype. For instance, cell lines transformed with the Rous sarcoma virus or murine sarcoma virus lose their ability to properly expressα 5β1. ²¹ Furthermore, the absence of α5β1 integrin is associated with malignancy of both the HT29 colon carcinoma cell line ²² and Chinese hamster ovary (CHO) cells. ^{23 24} In addition, CHO cells selected for their inability to properly express α5β1 grew more rapidly, whereas restoration of α5β1 expression by stable transfection drastically diminished the malignancy of these cells. ^{23 24}  

In the present study, we cultured four distinct human uveal melanoma cell lines (SP6.5, SP8.0, TP17, and TP31) from the primary choroidal tumor of four different patients. These cells grew in simple culture medium and exhibited doubling times at least two times higher than those measured for existing uveal melanoma cell lines. ^{25 26} The purpose of this study was to compare the expression of α5β1, FN, and gelatinases in uveal melanoma cell lines to determine whether these characteristics are related to the malignancy. 

Materials and Methods

This study was conducted in accordance with our institution’s guidelines and the Declaration of Helsinki, and the protocols were also approved by the institution’s Committee for the Protection of Human Subjects. In addition, all experiments conducted in nude mice strictly adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Tissue Extraction and Cell Culture

The melanoma cell lines were obtained as previously described ^{27 28} from primary tumors extracted from the eyes of four patients with diagnosed uveal melanoma. Samples of tumor tissues were fragmented into small pieces and seeded into T25 flasks in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, Burlington, Ontario, Canada) supplemented with 10% fetal bovine serum (FBS). The melanoma cell lines obtained were designated SP6.5, SP8.0, TP17, and TP31. All these tumor cell lines were maintained in DMEM supplemented with 10% FBS under 5% CO₂ at 37°C. Gentamicin was added to all media at a final concentration of 15 μg/ml. 

Isolation and culture of normal uveal melanocytes were performed according to the procedure described by Hu et al. ²⁹ in eyes of human donors aged 9 to 73 years obtained at death from the National Eye Bank of Quebec, Canada. The cells were maintained in phorbol 12-myristate-13-acetate/isobutylmethylxanthin cholera toxin (PIC) medium containing 10% FBS and 50 μg/ml gentamicin as described. ²⁹ The culture medium was also supplemented with 100 μg/ml geneticin (Gibco) for up to 2 weeks to eliminate contaminating cells, such as fibroblasts and pigment epithelial cells, which, unlike uveal melanocytes, are particularly sensitive to the effects of this cytotoxic agent. ²⁹  

Transmission Electron Microscopy

Cells were fixed in 4% phosphate-buffered glutaraldehyde for 30 minutes, washed twice in 6.8% sucrose in phosphate-buffered saline (PBS), and postfixed in 1% osmium tetroxide. Each cell specimen was dehydrated in a series of graded ethanol and then embedded in Araldite 502 epoxy resin (Taab Laboratory Equipment, Alder Mastron, UK). Ultrathin sections (90 nm) were cut with an ultramicrotome (model 570; Reichert Jung, Vienna, Austria), mounted on a copper grid, and contrasted with uranyl acetate and lead citrate before examination by electron microscopy (model JEM-1010; JEOL, Tokyo, Japan). 

Tumorigenicity Assays

After removal of the medium, cells were isolated from the culture plates with PBS-0.02% EDTA. After centrifugation, 8 × 10⁵ cells were resuspended in 0.1 ml DMEM supplemented with 10% FBS, 100 U/ml penicillin and 100 μg/ml streptomycin (Gibco), and injected subcutaneously into the posterior lateral section of Crl:CD1-nuBR athymic nude mice (Charles River, St. Constant, Québec, Canada). Eight 7-week old female mice were used for each cell line. Mice were examined on a regular basis and tumors, when apparent, were measured on the horizontal and vertical axes with calipers. The tumors were excised when they reached the approximate size of 15 × 15 mm. Fifty days after excision of the primary tumor, the mice were killed. 

Thymidine Incorporation

Thymidine incorporation was determined as previously described. ³⁰ Briefly, approximately 2.5 × 10⁴ cells from each uveal melanoma cell line (SP6.5, SP8.0, and TP31) were plated in triplicate on 60-mm culture dishes (1 × 10⁴ cells per well on a 24-well culture plate for TP17 cells) in DMEM supplemented with 10% FBS and allowed to adhere and spread for 18 hours. Cells from each triplicate were pulse labeled for 5 hours by the addition to the culture medium of 1.0 μCi of tritiated thymidine (³HThd) at 0, 24, 48, and 72 hours. Incorporation was stopped by removing the ³HThd-containing culture medium. The cells were then washed three times each with a solution of 0.9% NaCl and a cold solution of 10% trichloroacetic acid, followed by 70% and 100% ethanol washes. The culture dishes were allowed to dry at room temperature and 1 ml 0.1 N NaOH was then added for 1 hour. The samples were neutralized with 0.4 ml 1 N HCl and counted by liquid scintillation (Aquasol; duPont Canada, Lachine, Québec, Canada). 

Reverse Transcription–Polymerase Chain Reaction

For Reverse transcription–polymerase chain reaction (RT-PCR), total RNA was isolated with Trizol solution (Gibco) and the first cDNA strand synthesized using SuperScript II reverse transcriptase (Gibco), according to the manufacturer’s recommendations. RT was performed using 10 μg total RNA and 3 μg oligo(dT) primer. The DNA sequence of both the 5′ and 3′ template primers for human α5 (5′ primer: nucleotides 3000–3023; 3′ primer: nucleotides 3170–3147; 171-bp PCR product) and human fibronectin (5′ primer: nucleotides 3981–4001; 3′ primer: nucleotides 4346–4325; 365-bp PCR product) were derived from their corresponding human genes (GenBank accession numbers X06256 and X02761, respectively). Human β-actin (5′ primer: nucleotides 2185–2208; 3′ primer: nucleotides 2411–2388; from GenBank accession number M10277; 227-bp PCR product) was used as a control for normalization of FN and α5 PCR amplifications. Taq polymerase (Pharmacia, Baie-d’Urfé, Québec, Canada) was selected for PCR amplification. Cycle parameters were the same for all primers used (denaturation 94°C, 15 seconds; annealing 65°C, 30 seconds; extension 72°C, 30 seconds) with the total number of cycles (30 for β-actin, 35 for α5, and 40 for FN) tailored to the specific primer pair. 

Cell Culture in Soft Agar

Anchorage-dependent growth assays were performed in soft agar exactly as recently described ³¹ except that DMEM was used for the completion of the assays instead of DMEM-F12. 

Inhibition of Adhesion Assays

Inhibition of adhesion was assayed as previously described. ²² Briefly, the wells from a 96-well microtiter plate were filled with a 10-μg/ml solution of FN (Boehringer Mannheim, Laval, Québec, Canada) and incubated 4 hours at room temperature. The FN-containing solution was removed, and the wells blocked with 3% bovine serum albumin (BSA) in PBS for 1 hour. Meanwhile, cells were prepared as follows: each cell line was removed from the culture dish by incubation in PBS-EDTA 0.02%, then washed three times in PBS and resuspended in adhesion medium (serum-free DMEM supplemented with 3% BSA, 2 mM CaCl₂ and 2 mM MgCl₂) at 5 × 10⁵ cells/ml. Twofold serial dilutions (from 250 to 0.5 ng) of a monoclonal antibody (mAb) directed against the α5 integrin subunit (IIA₁) were prepared and incubated for 1 hour with 5 × 10⁴ cells. The cells were then plated in FN-coated culture wells and allowed to adhere for 1 hour at 37°C. After incubation, wells were washed three times with adhesion medium (37°C) and twice with PBS. The cells were fixed 15 minutes with a solution of 0.5% paraformaldehyde in PBS and stained with a solution of 0.5% crystal violet in 20% methanol for 1 hour at room temperature. Stained cells were washed three times with water and bound-dye solubilized in 1% sodium dodecyl sulfate. Absorbance was determined at 570 nm. Data are shown as percentage of adhesion inhibition relative to untreated cells. 

Gelatin Zymography

Each cell line (1.5 × 10⁵) was seeded into 24-well culture dishes in serum-free DMEM for 24 hours. Media were then collected for detection of gelatinase activity by zymography. A 25-μl sample was electrophoresed on a 8% polyacrylamide gel containing 1 mg/ml gelatin. Molecular mass markers were also run alongside the medium samples for precise identification of the MMPs’ molecular masses. After electrophoresis, the gel was washed twice with 2.5% Triton X-100 in 50 mM Tris (pH 7.4) for 30 minutes each, and twice in 50 mM Tris (pH 7.4) for 15 minutes each. The gel was then incubated overnight in a solution of 50 mM Tris (pH 7.4) containing 10 mM CaCl₂ and stained with Coomassie brilliant blue to reveal gelatinase activities. 

Flow Cytometry

Human uveal melanoma cell lines were harvested in PBS-0.02% EDTA and washed with PBS. Approximately 5 × 10⁵ cells were incubated on ice for 45 minutes with 1 μg of an mAb directed against the human α5 integrin subunit (IIA₁; Pharmingen, Mississauga, Ontario, Canada). A 1:100 dilution of a mouse mAb (C-2-10) ³² raised against the C-terminal end of the DNA binding domain of bovine poly(ADP-ribose)polymerase (PARP) was used as a negative control. After they were washed and labeled for 30 minutes on ice with a 1:100 dilution of a fluorescein isothiocyanate–conjugated secondary antibody (Sigma, St. Louis, MO), cells were resuspended in 500 μl PBS and analyzed by flow cytometer (Epics XL; Coulter, Miami, FL). 

Cell Invasion Assays

Cell invasion assays were conducted in modified Boyden chambers (24-well plates with 8-μm pore-size filter inserts; Becton Dickinson, Franklin Lakes, NJ), essentially as previously described. ^{31 33} Briefly, filters were coated with 200 μl ice-cold basement membrane ECM (Sigma-Aldrich, Oakville, Ontario, Canada) at 250 μg/ml proteins. Approximately 1 × 10⁵ cells were added to the upper chamber in 200μ l of culture medium (serum free DMEM). The bottom chamber was filled with 300 μl NIH 3T3-conditioned DMEM that had been initially supplemented with 50 mg/ml ascorbic acid. Cell adhesion to the ECM-coated membrane was allowed for 18 hours. Cells were then fixed in 2% paraformaldehyde in PBS for 30 minutes at room temperature and stained with 0.5% toluidine blue in 2% Na₂CO₃. The cells remaining on the top side of the filter were removed with a cotton swab. Those cells remaining on the bottom side of the filter were counted under light microscopy. 

Results

Uveal Melanoma Cell Lines with Properties Typical of a Malignant Phenotype

To acquire the invasive and migratory properties typical of the malignant phenotype, a cell must escape the many mechanisms that retain its spatial positioning within a particular tissue. Membrane-bound receptors, such as those that belong to the integrin family, and the specific ligands they interact with, such as fibronectin, strongly contribute to the anchorage dependency of a cell (for a review, see Reference 34). To investigate whether any correlation can be established between the malignant properties of uveal melanoma and altered expression of such components, four distinct uveal melanoma cell lines (SP6.5, SP8.0, TP31, and TP17) were derived by culturing the primary tumor from four individuals with diagnosed uveal melanoma (Table 1) . Unlike nonneoplastic uveal melanocytes, which have a uniformly spindlelike morphology, SP6.5, SP8.0, and TP31 cells appeared as mixed spindle–epithelioid culture types (Fig. 1) and exhibited a malignant phenotype characterized mainly by large and fragmented nucleoli and a disproportionate nucleus-to-cytoplasm ratio (Fig. 2) . Microvilli were also dispersed throughout the cell surface and pigmented granules throughout the cytoplasm (Fig. 2) . In contrast, TP17 cells had a more uniform epithelioid morphology (Fig. 1) . The presence of pigmented melanosomes is further evidence that all these cell lines initially originated from uveal melanocytes. The doubling times of 15, 37, 32, and 27 hours, obtained by TP17, TP31, SP6.5, and SP8.0, respectively, and evaluated by incorporation of tritiated thymidine, are also typical of malignant cells. These rates of proliferation are significantly more rapid than those measured for normal melanocytes, which have doubling times estimated to be more than 78 hours. 

To evaluate the ability of these cells to generate tumors spontaneously in vivo, each of the uveal melanoma cell lines was injected subcutaneously into nude mice, and tumor formation was monitored for approximately 3 months. As indicated in Table 2 , the TP17 cell line was by far the most tumorigenic; all injected mice had detectable tumors within only 18 days. Tumors also developed in 6 of 8 mice injected with the TP31 cells, whereas SP8.0 and SP6.5 yielded detectable tumors in only two of eight and one of eight mice, respectively. 

To further support the findings in the tumorigenicity assays in nude mice, we also evaluated the ability of each uveal cell line to grow in an anchorage-independent manner on soft agar. As expected, normal primary cultured melanocytes failed to grow in soft agar (Figs. 3A 3B ). In contrast, all uveal melanoma cell lines (SP6.5, SP8.0, TP31. and TP17) grew to large colonies after 14 days in soft agar (only the results with the TP31 cells are shown, Figs. 3C 3D ). We therefore conclude that all the melanoma cell lines derived in this study possess the ability to grow in an anchorage-independent manner. 

The invasion and migration properties of each uveal melanoma cell line were then evaluated in modified Boyden chambers. For this purpose, the membrane separating both the upper and lower compartments was first coated with basement membrane ECM. Then, cells were added to the upper compartment, whereas the bottom compartment was filled with conditioned medium from mouse NIH 3T3 cells, which was used as a source of chemoattractant. ³³ After an 18-hour incubation period, cells were fixed, stained, and counted under light microscopy. Of all the cells examined, TP17 turned out to be the most invasive, followed by the TP31, SP8.0, and SP6.5 melanoma cell lines (Fig. 4) . In contrast to TP17, only a small number of normal melanocytes migrated through the membrane. Therefore, these cells display different rates of invasion and migration, SP6.5<SP8.0<TP31<TP17, which correlate with their observed tumorigenicity. 

Expression Pattern of α5 Integrin in the Uveal Melanoma Cell Lines

It has been postulated that integrin-mediated cell adhesion to a fully organized ECM is a prerequisite for suppression of cell proliferation and thus contribute to contact inhibition of growth. ³⁴ As a consequence, the loss of certain integrins may also contribute to the acquisition of an invasive phenotype by tumor cells by loosening cell adhesion to the ECM. Indeed, in certain types of tumor cells, a correlation could be established between the absence of the FN-binding integrin subunit α5 and the tumorigenic and invasive abilities of such cells. ^{22 24} We therefore investigated the expression of α5 in the uveal melanoma cell lines used in the present study. We initially performed inhibition of adhesion assays to semiquantitatively evaluate the variations in the expression level of the α5 integrin subunit between all these uveal melanoma cell lines. In this assay, an equal number of cells was incubated with increasing concentrations of an antibody directed against the human α5 subunit. Cells were then allowed to adhere to FN-coated culture wells for a predetermined period, as described earlier. The extent of cell adhesion was then determined by densitometric analysis after cell staining with crystal violet. No inhibition of cell adhesion was observed in TP17 cells, which suggests that they express no or only very low amounts of this membrane-bound integrin subunit (Fig. 5) . The adhesion of SP6.5 and TP31 cells was inhibited to 70% by as little as 4 ng α5 antibody, which indicates that their ability to bind FN-coated dishes may be largely dependent on the presence of theα 5 integrin subunit. Further increasing the concentration of the α5 antibody up to 250 ng resulted in an inhibition of nearly 90% and 80% for TP31 and SP6.5 cells, respectively. Adhesion of SP8.0 was only partially inhibited (up to 40% when 250 ng α5 antibody was used) suggesting that at least two distinct membrane-bound FN receptors account for most of the adhesion mediated by this ECM component. 

We then validated the results from the cell adhesion inhibition assays by detection of the membrane-bound α5 integrin subunit through flow cytometric analyses. An unrelated mAb (C-2-10) ³² directed against the DNA binding domain of bovine PARP, a nuclear enzyme involved in many cellular processes including DNA repair, was also used as a negative control to define the background level for each cell-type. As shown on Figure 6 , moderate levels of the α5 integrin subunit were easily detected in primary cultured, nontumoral human melanocytes. Low but clearly detectable levels of α5 were also observed in the SP6.5, SP8.0, and TP31 melanoma cell lines. In agreement with the results of cell adhesion inhibition assays, no expression of this integrin subunit was detected in the highly tumorigenic and invasive TP17 cells. RT-PCR analyses were then conducted to determine whether the disappearance ofα 5 at the cell membrane of the TP17 cells also was caused by a reduction in the transcription of the α5 gene itself. When normalized to the amount of transcript corresponding to endogenous β-actin, the amount of α5 mRNA detected for each cell line ranked in the following order: SP6.5>SP8.0/TP31>TP17. Moderate levels of α5 mRNA could be detected in SP6.5 cells, the least tumorigenic of all the melanoma cell lines we examined (Fig. 7) . SP8.0 and TP31 cells expressed similarly low but detectable levels ofα 5 transcript, whereas the TP17 cells clearly expressed the lowest level of α5 of all four cell lines. We therefore conclude that expression of the α5 integrin subunit is inversely correlated to the tumorigenic ability of the uveal melanoma cell lines examined. The highly tumorigenic TP17 cells expressed low levels of α5 mRNA and no detectable α5 protein at the cell membrane, whereas the least tumorigenic SP6.5 expressed substantial amounts of both α5 mRNA and protein. 

Expression of the Extracellular Matrix Component Fibronectin (FN) in the Uveal Melanoma Cell Lines

Along with altered levels of integrin expression, defects in properly organizing the ECM represent another way through which tumor cells may escape from their surrounding cellular environment. ³⁴ We therefore performed RT-PCR analyses to evaluate whether the melanoma cell lines described in this study express endogenous FN transcripts. As shown on Figure 7 , detection of the appropriate 365-bp RT-PCR product in TP31, SP8.0, and SP6.5 cells indicated that the FN gene was transcribed in these melanoma cell lines; however, this PCR product was not detected in the highly invasive and tumorigenic TP17 cell line. As a control, transcripts corresponding to the β-actin gene could be amplified by RT-PCR in all uveal melanoma cell lines. 

Gelatin Gel Zymography Detection of MMPs Secreted by the Human Uveal Melanoma Cell Lines

Along with reduced expression of integrins, tumor cells may acquire their invasive properties by remodeling their surrounding ECM through the secretion of enzymes that belong to the family of MMPs (for a review, see Reference 35). Although the normal nonneoplastic tissues that surround the malignant cells may contribute to their invasiveness by producing such matrix-degrading enzymes through a tumor-induced host response, ³⁶ secretion of MMPs by tumor cells has been the subject of many studies. ^{37 38 39 40} Because most ocular melanomas have been reported to secrete both MMP-2 and MMP-9 ¹⁹ (also known as gelatinases A and B), we examined whether these gelatinases are also secreted by our melanoma cell lines and whether the level of MMP secretion also distinguishes the more aggressive TP17 cells. Cells were therefore incubated in serum-free medium for 24-hours before collection of conditioned medium for gelatin gel zymography. As shown on Figure 8a 72-kDa gelatinase (probably corresponding to the 72-kDa inactive form of MMP-2) could be detected in the culture medium from all the cells examined, including nonneoplastic uveal melanocytes. In addition, TP17 cells also expressed an MMP with an apparent molecular mass of 117 kDa that was not detected in the culture medium of the other melanoma cell lines. Under the conditions selected, no MMP-9 activity could be observed in the culture medium of SP6.5, TP17, and TP31 cells, whereas the medium from SP8.0 cells had low but detectable MMP-9 (Fig. 8) . Gelatinase activity in the supernatant from all tumor cells, as well as from primary cultured uveal melanocytes, was inhibited by the addition of 10 mM EDTA to the incubation buffer (results not presented) providing evidence that these enzymes indeed belonged to metalloproteinases. Experiments were also conducted with cell seeding in plates containing increasing amounts of human FN to determine whether this ligand for the α5β1 integrin could induce MMP-2 and MMP-9 expression. However, no MMP induction was observed in the presence of FN under the conditions selected (data not shown). 

Discussion

In an attempt to understand the molecular mechanisms underlying progression to metastasis, we established and characterized four distinct human uveal melanoma cell lines (SP6.5, SP8.0, TP17, and TP31), each derived from distinct primary tumors. When compared with some of their existing counterparts, no cell feeder layer or growth factor–supplemented media ²⁵ was needed to obtain the cell lines, which arose spontaneously and proliferated in DMEM supplemented with 10% FBS. The ease with which they were obtained obviously suggests the growth autonomy of these cells in vitro. This could also mean that, from the various cell phenotypes present in the primary tumor, only the most aggressive clones were selected to grow in vitro. The high proliferative rate of these cells, as determined by tritiated thymidine incorporation, also argues for the selection of robust, growing cells. All cell lines shared the major characteristics of tumor cells such as fragmented nuclei, an elevated nucleus-to-cytoplasm ratio and the absence of anchorage-dependent growth. They also exhibited structures such as mature and immature melanosomes that revealed the melanocytic origin of these cells. 

Despite these similarities, there are features that distinguished TP17 from the remaining three cell lines. For instance, the experiments we conducted in nude mice provided evidence that TP17 cells were, by far, the most malignant of the four cell lines examined. Cell morphology also distinguished TP17 cells from the other cell lines. Indeed, in contrast with the sharp spindle shape of in vitro nonneoplastic uveal melanocytes, TP17 cells showed a clear epithelioid morphology compared with the mixed spindle–epithelioid shape observed in the other cell lines (SP6.5, SP8.0, and TP31). Although it is difficult to establish any direct relationship to the Callender classification, it is noteworthy that the epithelioid morphology, usually associated with a poor prognosis, ^{5 6} was also a characteristic of our most aggressive cell line (TP17). 

There is accumulating evidence that proteins that can alter the cell cytoskeleton could also be involved in tumorigeneses. Therefore, the link between the cell morphology of uveal melanoma and clinical prognosis can be explained in terms of alterations of the cytoskeleton. Although the relationship between the absence of α5β1 at the cell surface of TP17 cells and their epithelioid morphology remains to be demonstrated, the involvement of α5β1 in the formation of an organized actin microfilament bundle has been extensively studied. ⁹ Formation of such a structure depends on the formation of focal contacts that require the engagement and agglomeration clustering of integrins such as α5β1. ⁹ Strong evidence suggests that one major role of focal contacts, the formation of which is mediated by α5β1, is to bring together many protein complexes involved in intracellular signaling, and consequently to determine cellular fate. 

Although the exact molecular pathway altered by the absence of α5β1 is not known, downregulation of α5β1 has been observed in cell lines as a result of cell transformation with the oncogenic Rous sarcoma virus. ²¹ CHO cells, ^{23 24} HT29 human colon carcinoma cells, ²² and K562 human erythroleukemia cells ⁴¹ selected for their adhesive properties to FN express detectable α5β1 and have a clear reduction in their proliferative abilities and transformed phenotype when compared with the parental cells from which they derive that do not express α5β1. Furthermore, α5β1-negative HT-29 and CHO cells stably transfected with the α5 subunit cDNA lacked their transformed phenotype, as determined by tumorigenicity assays in nude mice, and exhibited a reduced proliferative potential in vitro. ^{22 24} In HT-29 cells, it has been established that the growth arrest observed in the absence of serum or FN was related to the induction of a protein, growth arrest–specific gene 1 (gas-1), ²² associated with cellular quiescence. ⁴² Furthermore, under these cell culture conditions, transcription of immediate early genes such as c-jun, c-fos, and jun B was prevented. These observations emphasize the role of α5β1 in growth control and highlight the consequences of losing α5β1 expression in the acquisition of a transformed phenotype. Our results with the TP17 cell line are in agreement with these observations, in that the absence ofα 5β1 expression appears to be related to a more aggressive phenotype in nude mice and to a higher proliferative rate in vitro than that of α5β1-positive cell lines of similar origin. 

Gelatin zymography revealed another striking distinction between TP17 and the other three cell lines. In addition to the 72-kDa MMP-2 that has been observed in all cell lines (including melanocytes), a high-molecular-mass MMP of 117 kDa has also been observed in TP17. An MMP with a similar molecular mass has been noted and associated with a very invasive phenotype in the SCp2 murine mammary epithelial cell line. ³¹ In these cells, an MMP with an apparent molecular mass of 120 kDa was associated with the maintenance of an undifferentiated phenotype. Furthermore, human breast cancer cell lines that are less differentiated and particularly aggressive were found to express constitutively high levels of the secreted MMP of 120 kDa. ³¹ Another related study was conducted to investigate MMPs present in urine samples from patients with different types of cancer. ⁴³ The results revealed that all nine patients with breast cancer had detectable amounts of an MMP of 125 kDa in the urine and highlighted the specificity of this MMP to human breast cancer. If there is indeed a relationship between expression of the 120-kDa MMP and the undifferentiated phenotype, as suggested in murine mammary epithelial cells and human breast cancer cells, then we can assume that either changes in cell morphology or expression of 120-kDa MMP (or both) is related to a more undifferentiated state in TP17 cells compared with the other three cell lines. In vitro invasion assays have already shown that TP17 cells are clearly more invasive than the other uveal melanoma cell lines. 

Recently, Vaisanen et al. ²⁰ have proposed MMP-2 as a new prognostic marker of uveal melanoma. Fourteen of 29 primary uveal melanomas displayed positive staining for MMP-2. Among the positive uveal melanomas, nonspindle cells accounted for 63% of the positive staining. Furthermore, all epithelioid cell tumors ³ exhibited MMP-2–positive staining. These results are interesting considering that all the uveal melanoma cell lines tested in vitro for the expression of gelatinases express MMP-2. Furthermore, in our study, the cell’s aggressiveness could not be related to the amount of MMP-2 secreted by the melanoma cell-lines. However, we cannot exclude the possibility that the most aggressive cells selected to grow in vitro are also those that express MMP-2. Because most uveal melanoma cell lines display either a mixed or epithelioid morphology, the argument that aggressive cells express MMP-2 is consistent with the results reported by Vaisanen et al. ²⁰ Although the expression of MMP-2 by nonneoplastic melanocytes does not support this hypothesis, the possibility must be considered that primary cultures of these cells in an environment (such as on plastic culture dishes) that is very different from the choroidal tissue may have triggered expression of an otherwise silent gelatinase. Further studies making use of more sensitive methods of detection such as in vivo PCR or in situ hybridization are needed to clearly establish whether MMP-2 is expressed in uveal melanocyte-containing tissues. That the SP8.0 cell line expresses both MMP-2 and MMP-9 is not surprising; Cottam et al. ¹⁹ also observed coexpression of both these MMPs in some other human uveal melanoma cell lines. However, none of our results could relate this feature to increased tumorigenicity or invasiveness of this cell line. For instance, SP8.0 was one of the least aggressive melanomas when injected into nude mice and was less invasive than TP31 and TP17 in in vitro assays of invasiveness. 

Our attempt to relate FN with MMP-2 or MMP-9 expression, as reported previously for fibrosarcoma, ¹¹ human ovarian cancer cells, ¹⁵ and myeloid cell lines ^{13 14} were unsuccessful. It is likely that FN does not possess the ability to induce MMP-2 or MMP-9 expression in any of our uveal melanoma cell lines. These results suggest that MMP-2 is expressed in a constitutive rather than in an inducible manner in the uveal melanocytic cell lines examined in the present study. 

Many scientists now view cancer invasion as tissue remodeling gone out of control and consider the study of nonneoplastic tissue remodeling processes an interesting option for understanding the molecular mechanisms involved in cancer invasion. Taken individually, the absence of either FN or α5 expression, ⁴⁴ as well as an altered secretion of MMPs, is unlikely to be sufficient to provide the necessary requirements for a cell to become highly invasive or malignant. However, the combination of these conditions within a single cell, as is the case with the TP17 cell line, may prove sufficient to substantially alter organization of the ECM and allow such a cell to escape its tissue environment and access other target tissues. Based on the results presented here, we believe these human uveal melanoma cell lines may prove particularly useful as in vitro cellular models to investigate the role of either α5β1 and fibronectin in the melanocyte transformation and in the specific targeting of uveal melanoma metastasis to the liver. 

 Supported by the Réseau de Recherche en Vision du Fonds de la Recherche en Santé du Québec, the Fondation des Maladies de l’Oeil, and the Cancer Research Society. AB is the recipient of a studentship from the Fondation de l’Université Laval. SG is a Scholar from the Fonds de la Recherche en Santé du Québec.

 Submitted for publication November 24, 1999; revised January 28, 2000; accepted February 9, 2000.

 Commercial relationships policy: N.

 

Table 1.

 

Summary of Cases from Which Cell Lines Were Derived

Cell Line	Age, Sex	Date of Enucleation	Follow-Up (mo)	Histology
TP17	69, M	30/08/89	EX/27	Epithelioid
TP31	62, M	07/11/91	EX/38	Mixed
SP6.5	77, F	24/01/88	142	Mixed
SP8.0	48, M	03/04/88	EX/118	Mixed

 EX, died of metastatic uveal melanoma.

Figure 1.

 

Phase-contrast images of monolayer cultures of the following human uveal melanoma cell lines: nonneoplastic melanocytes (A), SP6.5 (B), SP8.0 (C), TP31 (D), and TP17 (E) cells. Magnification, ×200.

Figure 2.

 

The ultrastructure of SP6.5 (A, D), SP8.0 (B, E), and TP31 (C, F) uveal melanoma cells was examined by transmission electron microscopy at either low (A, B, and C) or high (D, E, and F) magnification. Nu, nucleolus; N, nuclei; C, cytoplasm; Mi, microvilli; Me, melanosome; NM, nuclear membrane. Magnification, (A, B, and C)× 3000; (D, E, and F) ×20,000.

Table 2.

 

Tumorigenic Assays in Nude Mice

Cell Lines	Number of Mice with Primary Tumor (n = 8)	Time between Injection and Appearance of Primary Tumor (days)
TP17	8	<18
TP31	6	∼33
SP8.0	2	∼91
SP6.5	1	∼40

Figure 3.

 

Nonneoplastic human uveal melanocytes (A, B) and TP31 uveal melanoma (C, D) were seeded and grown on soft agar. They were photographed 14 days later under phase-contrast microscopy. Magnification, (A, C) ×40; (B, D) ×100.

Figure 4.

 

To determine the invasive abilities of uveal melanoma cell lines, primary cultured uveal melanocytes or uveal melanoma cells (SP6.5, SP8.0, TP31, and TP17) were plated on ECM-coated filters in modified Boyden chambers and allowed to adhere and migrate for 18 hours. The number of cells that migrated through the membrane (with SD) was then determined for each cell type.

Figure 5.

 

Adhesion of cells was inhibited by FN. Twofold serial dilutions of a mAb (IIA₁) directed against the α5 integrin subunit were added to 1 × 10⁵ cells for 1 hour. Cells were then plated on FN-coated 96-well culture plates (10 μg/ml of FN). The percentage of adherent cells was plotted against the amount of α5 antibody used. The results presented are from one of three representative experiments.

Figure 6.

 

Surface expression of the integrin subunit α5 in the uveal melanoma cell lines was monitored by flow cytometry in SP6.5, SP8.0, TP31, and TP17 cells and in primary cultures of normal melanocytes, by using the mAb IIA₁ (solid line). As a negative control, an mAb directed against bovine PARP was used as the primary antibody (dotted line). Human HeLa cells were also used as a positive control for cell surface expression of α5. Relative fluorescence is shown as a logarithmic scale of 4 log cycles in the x-axis and the cell number as a linear scale in the y-axis. Data from one of three similar experiments are presented.

Figure 7.

 

RT-PCR amplification of both the FN and α5 mRNAs. Primers specific for human FN and human α5 were derived from GenBank sequences. FN andα 5 amplification products were obtained from all four uveal melanoma cell lines (TP17, TP31, SP6.5, and SP8.0) after 40 and 30 PCR cycles, respectively, and normalized to the actin PCR product for semiquantitative evaluation. The position corresponding to the α5 (171 bp), FN (365 bp), and β-actin (227 bp) amplification products is shown, along with that of the most representative DNA markers.

Figure 8.

 

Zymography for gelatinase activity. The gelatinase activities from the culture medium of each cell type (melanocytes, SP6.5, SP8.0, TP31, and TP17) was assessed by gelatin gel zymography. The position of MMP-2 and the 117-kDa MMP (MMP117) secreted by TP17 cells is indicated along with that of the appropriate molecular mass markers: ovalbumin, 45.5 kDa; bovine serum albumin, 74.3 kDa; phosphorylase B, 111.4 kDa; and myosin, 214.2 kDa.

The authors thank Sylvain Picard for technical assistance in transmission electron microscopy and Dean W. Hum and Marc-André Laniel for critically reviewing this manuscript.