An In Vitro Assay to Assess Uveal Melanoma Invasion across Endothelial and Basement Membrane Barriers

Abstract

purpose. To develop a modified in vitro invasion assay to assess uveal melanoma invasion across endothelial and basement membrane barriers.

methods. Using permeable cell culture supports, endothelial cells were grown to confluence on an 8-μM pore polycarbonate membrane precoated with an artificial basement membrane. Primary uveal melanomas were grown as short-term cultures at 37°C and 5% CO₂ and invaded through the endothelial cell layer and basement membrane. Invading cells were counted under ×400 magnification on the lower surface of the membrane. Levels of invasion were correlated with histopathologic markers of prognosis. The relative invasion of individual tumors was established by comparison of invasion through both endothelial and basement membrane barriers with invasion through basement membrane components alone.

results. A series of 13 primary tumors were studied using the modified invasion assay. Tumors varied in their propensity to permeate both barriers. In all cases the endothelial cell layer reduced invasion, but the effect varied between tumors.

conclusions. Some tumors were more adept at overcoming the additional endothelial cell layer, whereas invasion of others was severely inhibited. Tumor invasion through the transendothelial model was found to correlate more closely with clinical characteristics associated with invasion, than was invasion through basement membrane components alone. The transendothelial model may represent a more realistic model for the in vitro study of invasion of uveal melanoma cells, providing a useful in vitro system for the investigation of cellular interactions during the invasion process.

Posterior uveal melanoma is the most common primary intraocular malignancy in adults. Unlike cutaneous melanoma, uveal melanoma disseminates mainly through the bloodstream and preferentially establishes metastases in the liver. ¹ Metastatic liver disease is the leading cause of death in uveal melanoma and can develop after a long disease-free interval. ² To date, little progress has been made in the detection and treatment of metastatic disease, and after detection of liver metastases, median survival is still less than 1 year. ^{3 4}  

Metastasis involves a complex sequence of interrelated steps, and is dependent on both the host responses and intrinsic properties of the tumor cells. The sequential steps include neovascularization and increase in size of the primary tumor, detachment of neoplastic cells and entry into the circulation, adhesion to endothelium of distant organs, passage through the capillary basement membrane, and proliferation to form a secondary tumor. ^{5 6} Interactions of tumor cells with the vascular endothelium and basement membrane are crucial in these processes, involving tumor cell attachment to, and penetration through, the vessel wall and underlying matrix at both the primary and distant sites. ^{7 8} Moreover, endothelial cell cytokines, matrix components and degradation products are all chemotactic for tumor cells. ^{9 10} Degradation of the basement membrane has also been correlated with the metastatic potential of tumor cells, ^{11 12} and, more specifically, type IV collagenase expression has been associated with poor prognosis in uveal melanoma. ¹³  

In many in vitro studies of invasion the Boyden chamber chemotaxis assay has been used, whereby cells invade through a porous membrane coated with reconstituted basement membrane components. ^{14 15} Results from these assays have shown some correlation with invasive potential and have also been useful when assessing stimulation and inhibition of migration and invasion. Such models have been used to study invasion of cutaneous melanoma but have been restricted to the use of cell lines. ^{16 17} However, because invasion in vivo is ultimately more complex, involving both transendothelial cell invasion and degradation and migration through extracellular matrix (ECM) components and basement membrane, a more realistic model system would include each of these barriers. Initial studies established a model using human umbilical vein endothelial cells (HUVECs; large-vessel endothelial cells) and HT1080 fibrosarcoma cells, with migration observed by scanning electron microscopy. ¹⁸ Since this study, there have been few assay systems published mimicking transendothelial cell migration. For uveal melanoma, we have previously found that results of in vitro invasion assays that include only artificial basement membranes have corresponded to some extent with indicators of prognosis, but because long-term follow-up was not undertaken, the clinical relevance is uncertain. ¹⁹  

Various attempts have been made to develop a suitable animal model to more closely study the metastasis of uveal melanoma. Cases have been reported occurring naturally in cats and dogs, but they are infrequent, and the relationship to human disease is less apparent. ²⁰ Induced models share some similarities with human uveal melanoma metastasis but each has unique advantages and disadvantages. Murine models have been used, but tumors induced using the cutaneous B16 melanoma cell line can result in pulmonary, not hepatic, metastases. ²¹ Rabbit models are possibly more suitable for study because of the larger size of the rabbit eye and the rabbit’s longer life span, but development of hepatic metastases is again dependent on the specific model used. Thus, such models may not accurately represent the human situation in vivo. Because of these limitations, we developed a modified in vitro invasion assay, to assess uveal melanoma invasion across endothelial and basement membrane barriers (transendothelial invasion model) that is perhaps more realistic than previous in vitro invasion models that have included a basement membrane barrier only (standard invasion assays). In the model presented herein, a microvascular endothelial cell monolayer was used with an artificial basement membrane and has been used to mimic invasion for a series of primary uveal melanomas. Levels of invasion were correlated with histopathologic markers of prognosis. Transendothelial cell invasion was also compared with invasion through the basement membrane alone. For one culture, scanning electron microscopy was performed, and for two tumors, both cell populations were prelabeled with fluorescent probes, to confirm tumor cell movement through the invasion assay. 

Materials and Methods

Clinical Material

Fresh uveal melanoma samples were obtained from 13 primary posterior uveal melanomas at enucleation. Histopathologic details are recorded in Table 1 . Ethics committee approval was obtained, and the study’s protocols adhered to the tenets of the Declaration of Helsinki. Informed consent was acquired from the patients before collection of tumor samples. Samples were processed as reported previously. ²² Cultures were maintained by serial passage in RPMI-1640 as short-term cultures (STCs) supplemented with penicillin (100 U/mL), streptomycin (100 μg/mL), amphotericin B (5 μg/mL), epidermal growth factor (0.2 μg/mL), fetal calf serum (20%), and glucose (0.2%) at 37°C in an atmosphere of 5% carbon dioxide-95% air. In all cases apart from Sheffield ocular melanoma (SOM) 196B, STCs were used before passage five. SOM 196B was used initially to develop procedures and as the control. Melanoma status was confirmed by immunohistochemistry for three melanoma-associated markers (Melan-A, S-100, and HMB45). 

Transendothelial Cell Invasion

For each tumor, invasion was assessed through endothelial and basement membrane barriers (transendothelial cell invasion model). The results were compared with invasion through standard invasion assays, in which only the basement membrane was included. The transendothelial cell invasion assay was performed using a modified Boyden chamber system, as adapted from Lewalle et al. ²³ Commercially available dermal endothelial cells (human dermal microvascular endothelial cells [HDMECAs]; TCS Biologicals, Ltd., Buckingham, UK) were used to assess invasion of all 13 tumors. Because uveal melanomas primarily target the liver, for one culture (SOM 196B), the assay was also undertaken using microvascular endothelial cells freshly extracted from human liver resections (HULECs; a gift of Lance Burns, Department of Surgery and Anaesthetic Sciences, University of Sheffield, Sheffield, UK). Levels of transendothelial cell invasion of SOM 196B cells through HDMECAs and HULECs were then compared. 

Briefly, endothelial cells were grown to confluence on an 8-μM pore polycarbonate membrane (Transwell; Costar UK, Ltd., High Wycombe, UK), precoated with an artificial basement membrane (1.5 μg/mL, normal growth factor content; Matrigel; Beckton Dickenson Labware, Bedford, MA; Fig. 1A ). Cultures were dissociated from tissue culture flasks and resuspended in RPMI-1640 medium with 0.1% bovine serum albumin (BSA). Tumor cells were added to the upper chamber (1 × 10⁵/well), and RPMI-1640 medium with 0.1% BSA was added to the lower chamber (Fig. 2) . Cells were incubated at 37°C for 24 hours. Noninvading cells were removed from the upper chamber by gently wiping the upper surface of the membrane with a cotton swab. Membranes were fixed in ethanol and stained with Gill hematoxylin. Levels of invasion were assessed by counting the number of cells present in 10 fields on the lower surface of the membrane, under light microscope (×400; Fig. 1B ). For all tumors, each experiment, including appropriate positive and negative control experiments, was performed in triplicate. Negative control wells were endothelial cells alone. Tumor cell invasion through basement membrane components only (standard invasion levels), acted as the positive control for each culture. Levels of transendothelial cell invasion were compared with those in positive control cultures. 

Visualization of Invasion

To confirm tumor cell invasion in the transendothelial assays, invasion was monitored using fluorescent tagging, due to limitations with available cell numbers; only one cell line (SOM 196B) and one STC (SOM 295) were examined. Tumor and endothelial cells were independently labeled with succinimidyl esters (Molecular Probes, Inc., Eugene, OR) before use, according to the manufacturer’s instructions. In all cases, tumor cells were labeled with carboxy-fluorescein diacetate-succinimidyl ester (20 μM; CFDA-SE), which fluoresced green at 492 to 517 nm, and endothelial cells were labeled with SNARF-1 carboxylic acid, acetate-succinimidyl ester (5 μM; SNARF-1), which fluoresced red at 580 to 640 nm. Labeled cells were used as before in the assay. These reagents passively diffuse into the cells and are colorless and nonfluorescent until their acetate groups are cleaved by intracellular esterases to yield highly fluorescent compounds. After reacting with intracellular amines, the fluorescent conjugates become stable in the cell and can be fixed with aldehyde fixatives. CFDA has been shown to not affect adhesion to endothelial cells. ²⁴ Invading and noninvading cells were visualized with a fluorescence microscope (BX50; Olympus, Tokyo, Japan). 

Additional visualization of transendothelial cell invasion was performed using scanning electron microscopy on one culture (SOM 196B). Ian Palmer (Department of Pathology, Royal Hallamshire Hospital) performed all processing for electron microscopy by standard procedures. Cocultures of SOM 196B and HDMECAs were set up as before. Noninvading tumor cells and endothelial cells were not removed before fixation, to enable visualization of the invasion process. Both sides of the membranes were viewed. 

Statistical Analysis

For each tumor, each assay was performed in triplicate, and the data were analyzed by analysis of variance (ANOVA). Because the data included low counts and hence exhibited heterogeneity of variance, a square-root transformation was applied: √(x + 0.5). P < 0.05 was significant when comparing invasion in the transendothelial model with positive control invasion (basement membrane components alone). These values were used to establish that transendothelial invasion was significantly reduced compared with invasion through the basement membrane alone. Conversely, P > 0.05 was considered nonsignificant, when comparing transendothelial invasion with standard invasion and represents tumors relatively unaffected by the inclusion of the endothelial cell layer. 

Results

Endothelial cells formed confluent monolayers through which tumor cells could invade (Fig. 1A) . Because endothelial cells themselves are migratory, for some tumor cells invasion was monitored with fluorescent tagging and scanning electron microscopy, and the tumor cells could be visualized effectively overcoming the endothelial cell barrier (Figs. 1C 1D 1E) . Dermal endothelial cells were more readily available and were therefore used in all assays with STCs. However, because uveal melanomas are known to target the liver, it is likely that differences may exist between dermal and liver endothelial cells. To validate the use of dermal endothelial cells and to investigate this point, transendothelial cell invasion of one cell line (SOM 196B) was assessed using both human liver-derived endothelial cells (HULECs) and dermal endothelial cells (HDMECAs). HULECs were freshly extracted and not commercially available, making their use in all assays impractical. No difference in levels of invasion through the two types of endothelium was observed (P = 1; n = 3; data not shown). All uveal melanoma cultures expressed at least one marker associated with their nature as melanomas. 

The results of the invasion studies are summarized in Figure 3 . The data represent the mean invasion of the tumor cells through basement membrane components, with and without inclusion of a dermal endothelial barrier and are expressed as the mean number of cells invading per field of view, after counting 10 fields of view for triplicate wells. In wells with endothelial cells alone, maximum endothelial cell invasion was two to three cells per well and was considered negative. According to the ANOVA results, inclusion of an endothelial layer resulted in a significant reduction in invasion (P < 0.05) of SOM 196B, 262, 275, 277, 280, 281, 282, 290, and 295 when compared with invasion through a basement membrane only. Therefore, in relative terms, the invasive tumor cell populations of these cultures were less capable of overcoming the endothelial cell barrier. Conversely, invasion of cultures that was not significantly reduced by the addition of an endothelial cell barrier (P > 0.5; SOM 263 and 296) may infer the presence of invasive tumor cell populations that are more efficient at overcoming such barriers. Two tumors (SOM 269 and 272) were considered noninvasive in either assay. 

Most tumors had only recently been excised, and it was therefore not possible to correlate behavior in vitro with the clinical course of the disease. However, when control invasion of cultures through basement membranes alone was compared with histopathologic data (Table 1) , it was apparent that the melanomas that invaded at a relatively high rate, with invasion levels of more than 20 cell counts per field (SOM 196B, 277, 280, 282, 290, and 295) were in general of a mixed or epithelioid cell type: morphologies linked with poor prognosis. When studying the ability of corresponding tumors to invade in the transendothelial cell assay, different patterns of invasion emerged. Only cultures SOM 280, 290, and 295 showed consistently high levels of invasion in both assays. However, a reduction in invasion after the inclusion of an endothelial barrier was still seen for these tumor cells. For other tumor cells, the pattern of invasion varied dependent on the presence or absence of an endothelial cell barrier. In some instances reduction of invasion was highly significant (P < 0.001; SOM 196B, 275, and 280). Conversely, SOM 263 and 296, in invasion assays through basement membranes only, could be considered relatively noninvasive, yet their invasion rate was comparatively unaffected by the inclusion of an endothelial cell barrier and no significance was shown. When comparing transendothelial cell invasion with histopathologic data, we observed that the more invasive tumors, for which invasion levels were more than 50% of the control level (SOM 263, 280, 290, 295, and 296), were on average larger tumors than the most invasive tumors through the control basement membrane only assays. These tumors were also more likely to be of a mixed cell type and more had ciliary body involvement. However, two such tumors (SOM 263 and 290) were very small in volume, and one (SOM 295) was of spindle morphology; factors associated with good prognosis. Thus, the test sample was too small to make reliable correlations. 

Discussion

The invasion process is highly complex, involving a series of adhesive, degradative, and migratory steps, ultimately allowing metastatic tumor cells to detach themselves from the primary tumor and invade neighboring vessels. During these processes, tumor cells interact with a number of extracellular matrix and basement membrane components, together with endothelial cells of the vessel wall. ^{7 8} Adhesive interactions between tumor and endothelial cells themselves and between both cell types and extracellular matrix components are therefore critical at both the primary and secondary sites, whereas degradation is essential for migration through the extracellular matrix and basement membrane. ^{11 12}  

In most in vitro studies of invasion, investigators have used models that include extracellular matrix and basement membrane components only, correlating in vitro invasion of cell lines with relative metastatic potential in vivo through assessment with animal models. ^{14 15} Cutaneous melanoma cell lines have also been shown to behave comparatively in such assays. ^{16 17} However, because cell lines have been used in most work, there have been few published reports of use of STCs of primary tumors, and only limited correlation has been made of invasion with histopathologic details in any tumor type. ¹⁹ For both this study and our previous investigation, ¹⁹ STCs were used within five passages of being established in culture, with the intention of minimizing the effects of long-term culturing. However, it should be taken into consideration that, because resected tumors may be heterogeneous in their composition, clonal populations cultured may not be representative of the most malignant cellular population of the tumor in vivo. 

In the present investigation, to establish whether in vitro invasion could be considered representative of the tumor’s relative metastatic potential in vivo, tumor cell invasion in the assays was compared with predicted invasive potential, by correlating with known prognostic indicators. Results from the control invasion assay (basement membrane alone) confirmed our previous study in which comparable assays were used and in which melanoma cultures invading at a high rate through basement membrane components, in general, were only of epithelioid or mixed cell type. ¹⁹ The transendothelial invasion assay seemed to provide a closer correlation. Specifically, only one spindle cell tumor (SOM 295) was able to invade adequately in the transendothelial invasion assay as opposed to three spindle cell tumors (SOM 275, 281, and 295) in the control assay. In contrast, of the tumors with mixed and epithelioid morphologies, with the exception of one case (SOM 262), all six tumors were able to penetrate both barriers. However, it is also of interest that although the aggressive tumors in the transendothelial model tended to have prognostic features more often associated with a poor prognosis, there were some notable exceptions. In particular, SOM 263 was a small choroidal melanoma. It is known that small tumors can behave highly aggressively, ²⁵ despite indicators to the contrary, and it seems that SOM 263, at least in the transendothelial assay is a case in point. Additional follow-up is needed to clarify this point, and because the number of patients studied was small, the better correlations observed with the transendothelial model could be coincidental. 

Because the metastatic process involves interactions with endothelial cells as well as extracellular matrix and basement membrane components, we have developed an in vitro model, including both a microvascular endothelial cell layer (of dermal or liver origin) and an artificial basement membrane, to mimic transendothelial invasion. Using fluorescent tagging, it was possible to visualize an intact layer of endothelial cells and thus an effective barrier to invasion (Fig. 1A) . Cells therefore had to actively migrate between endothelial cells. It is also probable that confluent endothelial cells on the artificial basement membrane would have secreted additional basement membrane proteins, thus increasing the obstacle for the invading cells. ²⁶ In a recent study using a similar system but with bovine aortic endothelial cells (large-vessel endothelial cells) and a series of seven human malignant and nonmalignant cell lines, the investigators observed transmigration of endothelial cells and the subsequent formation of another confluent endothelial monolayer on the lower surface of the membrane. ²⁷ This phenomenon was not observed in our in vitro model and fluorescent tagging of the tumor and endothelial cell populations confirmed that confluent endothelial monolayers were not established on the lower surface of the membrane, possibly due to the use of endothelial cells derived from a different source. Because the microvasculature is generally thought to be the site of tumor cell extravasation ^{28 29} and because differences have been observed between large- and small-vessel endothelial cells, ³⁰ using microvascular endothelial cells in a transendothelial cell model is possibly more appropriate. Moreover, because uveal melanomas preferentially metastasize to the liver, transendothelial invasion through dermal and liver microvascular endothelial cells may vary. In a preliminary experiment, no differences in the level of invasion were seen between SOM 196B invasion through dermal and liver endothelium. It was therefore hoped that a model using dermal endothelial cells as a barrier to uveal melanoma cells might still be representative of the situation in vivo, and because these cells were more readily available, they were used throughout the study. 

The use of the transendothelial assay could provide an additional experimental system for the investigation of uveal melanoma invasion. For its effective use, the transendothelial model must be able to differentiate reliably between aggressive and nonaggressive melanomas in a manner consistent with their actual invasive capability. Few studies of transendothelial cell invasion cells have been published, but existing evidence suggests invasive abilities in a transendothelial cell invasion model correlates with metastatic potential in vivo, ²⁷ although in such studies, correlation with clinical outcome was not feasible. Because primary tumors have been used in this investigation, it was possible to make some limited comparisons with the outcome of individual patients. Survival data were limited to an average of 12 months, however, and correlation with long-term survival for tumors was not possible. Nonetheless in agreement with published data, it is of interest that one tumor (SOM 196B) had been resected 41 months earlier, and the patient is known to be alive and disease free. Once in culture, this tumor invaded well, without the inclusion of an endothelial cell layer in the standard invasion assay. Yet invasion was significantly reduced (P < 0.05) in the transendothelial cell invasion assay, with only 15% of the invading tumor population capable of overcoming both basement membrane and endothelial cell barriers. This may suggest that for SOM 196B the actual number of tumor cells in the total tumor population capable of overcoming both barriers is relatively small. In common with this tumor, other tumors, such as SOM 277, invaded well through the basement membrane, yet invasion was significantly decreased (P < 0.05) for most melanomas when an endothelial cell layer was included. 

It is widely understood that the metastatic process is highly inefficient. ³¹ Only a small percentage of cells in the primary neoplasm acquire the phenotype necessary to facilitate successful extravasation. Furthermore, once tumor cells have disseminated, only a fraction reach the secondary site and continue growth to become eventual metastatic foci. In vivo studies have shown that only 0.01% of tumor cells that enter the circulation progress to form metastatic colonies. ³² More recently, it has been suggested that growth of metastatic tumor cells into macroscopic tumors is primarily because of the microenvironment in which they are located, in agreement with Paget’s original seed-and-soil hypothesis. ^{33 34} Thus, in the initial stages of metastasis, a metastatic cell differs from a nonmetastatic cell, by its propensity to extravasate. In previous invasion assays, tumor cells have been considered to be invasive if they are capable of overcoming basement membrane barriers alone. In contrast, in the transendothelial model, we have considered cells to be invasive if they are able to overcome both the endothelial cell and basement membrane barriers together. Levels of invasion in the transendothelial model varied between tumors. Most tumors, were less effective at penetrating both the endothelial and basement membrane barriers (P < 0.05; SOM 196B, 262, 275, 277, 280, 281, 282, 290, and 295). For these melanomas, invasion may compare to some extent, with the estimated level of 0.01% of tumor cells forming metastatic colonies. For other tumors (SOM 263 and 296), almost the entire invading cellular population was capable of overcoming both barriers (P > 0.05). These melanomas may therefore equate to highly aggressive tumors in vivo, with more than 0.01% of tumor cells able to penetrate basement membrane and endothelial barriers. 

The transendothelial model considers only invasion through basement membrane and endothelial layers, and metastasis itself is a highly complex procedure. Many factors influence the establishment of metastases and thus although tumor cells may be capable of overcoming both barriers, aspects such as survival in the bloodstream and proliferation in the target organ cannot be assessed by such a model. Of additional note, it has now been reported that mosaic vessels may exist that are lined with both tumor and endothelial cells. ³⁵ Aggressive uveal melanomas have also demonstrated the formation of vascular channels in vitro and in vivo, composed of tumor cells only. ³⁶ Therefore, in uveal melanoma, metastatic progression may be far more complicated than previously thought and may use a number of mechanisms of extravasation. 

This study has shown a closer association of transendothelial cell invasion in vitro with known prognostic markers and, as such, may more reliably distinguish between melanomas invading at high and low levels than previous invasion assays using basement membrane alone. (Most spindle cell tumors were incapable of penetrating both endothelial and basement membrane barriers, whereas in comparison, only one of the mixed and epithelioid tumors was unable to penetrate both of these barriers.) The transendothelial assay is too impractical to be of prognostic value but was a useful experimental system in which to study differences in invasion and the mechanisms involved in both aggressive and nonaggressive tumors. Long-term clinical follow-up is necessary to further correlate in vitro transendothelial invasion in this model with clinical outcome, but it is possible that this system is a more representative model of invasion than previous in vitro models. 

 Presented in part at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2001.

 Supported by The Humane Research Trust.

 Submitted for publication July 16, 2001; revised January 18, 2002; accepted January 25, 2002.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Histopathologic Details of Patients Treated for Posterior Uveal Melanoma by Enucleation

SOM	Age	Sex	Location	Volume (mm3)	Cell Type	Metastatic Disease	Postsurgical Survival (mo)
196B	80	M	Choroid	885	Mixed	None	41
262	73	M	Choroid	Too large to scale	Epithelioid (50%)	None	15
263	71	M	Choroid	725	Mixed	None	13
269	72	M	Ciliary body	2273	Spindle B	None	12
272	71	F	Choroid	1795	Spindle B	None	10
275	71	M	Choroid	1070	Spindle B	None	8
277	74	M	Choroid	1476	—	None	7
280	85	M	Choroid	2621	Mixed	None	7
281	54	M	Choroid	1273	Spindle B	None	7
282	64	M	Ciliary body	1141	Mixed	None	6
290	53	M	Choroid	754	Mixed	None	4
295	13	F	Choroid/ciliary body	Too large to scale	Spindle B	None	2
296	55	M	Ciliary body	1270	—	None	2

 All patients were alive at this writing. SOM, Sheffield ocular melanoma.

Figure 1.

 

(A) Human dermal microvascular endothelial cells, labeled with SNARF-1 and grown to confluence on an artificial basement membrane–coated filter. Cells were observed under fluorescence at 580 to 640 nm. (B) Invading tumor cells (SOM 277) on the underside of the permeable membrane. Nuclei are stained with Gill hematoxylin. (C) Scanning electron micrograph, showing the underside of the membrane, with cellular projections through the porous membrane. (D, E) Photographic image of invading uveal melanoma cells in the transendothelial cell invasion assay. Tumor cells were labeled with CFDA-SE (green) and endothelial cells with SNARF-1 (red), and cells were observed under fluorescence at 492 to 517 nm and 580 to 640 nm, respectively. (D) Tumor cell interactions (SOM 295) with the endothelium on the noninvaded side of the membrane. (E) Invading tumor cells (SOM 295) on the under side of the membrane. Magnification: (A, D) ×100; (B) ×200; (C) ×1250; (E) ×1000.

Figure 2.

 

Schematic diagram of the permeable cell culture membrane invasion assay.

Figure 3.

 

Mean (bars, SE) number of invading cells of primary uveal melanoma cultures through the modified invasion assay and through a basement membrane barrier only (control invasion). *P < 0.05 when compared with a positive control (invasion through the basement membrane alone). Tumors with significantly reduced relative invasion were SOM 196B, 262, 275, 277, 280, 281, 282, 290, and 295. **P < 0.001 when compared with the positive control. Tumors with highly significant reduction in relative invasion were SOM 196B, 275, and 280.

The authors thank Robin Farr for assistance with photographic and poster work.