Melanoma Cells Homing to the Brain: An In Vitro Model

We developed an in vitro contact through-feet blood brain barrier (BBB) model built using type IV collagen, rat astrocytes, and human umbilical vein endothelial cells (HUVECs) cocultured through Transwell porous polycarbonate membrane. The contact between astrocytes and HUVECs was demonstrated by electron microscopy: astrocytes endfeet pass through the 8.0 μm pores inducing HUVECs to assume a cerebral phenotype. Using this model we evaluated transmigration of melanoma cells from two different patients (M1 and M2) selected among seven melanoma primary cultures. M2 cells showed a statistically significant higher capability to pass across the in vitro BBB model, compared to M1. Expression of adhesion molecules was evaluated by flow cytometry: a statistically significant increased expression of MCAM, αvβ3, and CD49b was detected in M1. PCR array data showed that M2 had a higher expression of several matrix metalloproteinase proteins (MMPs) compared to M1. Specifically, data suggest that MMP2 and MMP9 could be directly involved in BBB permeability and that brain invasion by melanoma cells could be related to the overexpression of many MMPs. Future studies will be necessary to deepen the mechanisms of central nervous system invasion.

Melanoma has one of the highest CNS metastatic capability [5,6] possibly due to the common embryological origin of melanocytes and neuronal subpopulations [7,8]. Melanoma metastases have been shown to have a high mutation rate [9], variable phenotype, a clinically diffuse dissemination pattern [10], and the unique ability to elicit spontaneous antigenspecific host immune responses [11].
Brain is considered a privileged site for the presence of blood brain barrier (BBB) protecting it against the entry of toxic substances and microorganisms [12,13]. Since the CNS also lacks the lymphatic system, the only way of metastatic cells to reach the brain is to cross this barrier [14,15].
BBB is composed by endothelial cells (ECs), sheathing by astrocytic foot processes, and pericytes. Brain endothelial cells are anatomically different from the other endothelial cells and are characterized by the presence of tight junctions (TJ) and by the absence of fenestrations. Nevertheless the presence of BBB is not able to prevent parenchymal brain invasion from different tumor circulating cells and it is surprising that some cancer types give metastases preferentially to the brain [14]. This phenomenon draws the attention to the possibility that the BBB may have a supportive role in 2 BioMed Research International the formation of the metastasis as well: cerebral endothelial cells have been described to actively take part in the tumor cells transmigration process. Besides, giving protection to CNS-invading cells from the immune response and from chemotherapy, BBB may produce substances favorable for metastasis growth [16]. The vasculature of the BBB is unique because of its association with astrocytes: in the brain, astrocytes and endothelial cells are closely apposed, separated only by a thin basal lamina [17]; therefore, model involving physical contact between endothelial cells and astrocytes more closely represent the in vivo situation [18][19][20].
Adhesion and junction molecules involved in tumor cellendothelial cell interaction are poorly known and even less is known about brain endothelial-specific mechanisms. Many authors suggest that tumor cells are able to partly mimic the molecular mechanisms of leukocyte-endothelial interaction occurring during inflammation. Depending on this, a role for integrins and adhesion molecules has been proposed to explain brain invasion [21,22].
Indeed, the crossing of the BBB has been described to involve several mechanisms among which adhesion molecules, extracellular degradation molecules, and growth factors [14,[21][22][23][24]. Using an in vitro heterologous BBB model [14] in the present work we investigated which of these was predominant in melanoma to brain metastasization. Transmigration of melanoma cells from different patients was evaluated to identify and characterize cell cultures with different behaviour. Due to the features of the model, we focused our attention on adhesion [21][22][23] and extracellular matrix degradation processes [24].

Cell Cultures.
HUVECs primary cultures were isolated from healthy donors according to Jaffe method [25] and used in experiments up to the tenth passage. The typical purity of cells cultures was higher than 95% as assessed by flow cytometry after CD31 staining (polyclonal phycoerythrinconjugated goat anti-human CD31, BD Bioscience). HUVECs were cultured in complete EBM (Endothelial Basal Medium, Lonza, Walkersville MD) medium completed with growth factors and antibiotics (EGM SingleQuots, Lonza) in cells culture dishes (Corning, USA) coated with Collagen S type I, 0.3 g/mL (Roche, Mannheim, Germany).
Rat astrocytes cells were commercially available primary cultures (rat brain cortex, Cambrex lot number 151104AT2) and were used in experiments up to the tenth passage. Cells were cultured in 75 cm 2 flasks (Corning) in DMEM (Dulbecco's Modified Eagle's Medium, Gibco) completed with 10% of foetal bovine serum (FBS, Gibco) and antibiotics penicillin/streptomycin 1% (P/S, Gibco). Purity of cells cultures was tested by flow cytometry using an alexa fluorconjugate antiglial fibrillary acid protein (GFAP) antibody; cells were used in experiments when the percentage of GFAP reached at least 95%.
Primary cultures of melanoma metastases were developed from surgical specimens as previously described [26] and were kindly provided by Dr. Licia Rivoltini (National Cancer Institute of Milan). Cells were cultured in 75 cm 2 flasks (Corning) in RPMI (Gibco), in addition to10% FBS (Gibco) and P/S 1% (Gibco).

Establishment of the In Vitro BBB Model
. In vitro BBB model was built with rat astrocytes and HUVECs cocultured through Transwell porous polycarbonate 8 m membrane (inner well), in 24 well plates (outer well) (Corning). The day before the preparation of the BBB model, the polycarbonate membrane was coated with 100 L of collagen IV at a concentration of 33 g/mL (mouse collagen, IV, BD Biosciences) for 1 h at room temperature (RT), then washed three times with phosphate buffered saline (PBS) without CaCl 2 and MgCl 2 (Gibco), and incubated in complete medium (100 L in the inner well and 600 L in the outer well) for 24 hours at 37 ∘ C in humidified atmosphere with 5% CO 2 . Astrocytes (200.000 cells) were seeded on the bottom side of the polycarbonate membrane insert, upside-down in humidified chamber for 1 hour. After that, the Transwell insert was inverted in a 24 well cell culture cluster (Costar, Corning Incorporated) in DMEM complete medium and placed over night at 37 ∘ C in humidified atmosphere with 5% CO 2 . After 24 hours, HUVECs (200.000 cells) were plated on the upper side of polycarbonate membrane and cultured in EBM complete medium. Astrocytes and HUVECs were than cocultured in EBM complete medium for four days before being used in transmigration experiments.

Trans-Endothelial Flux of L-Glucose.
To assess the integrity of the in vitro BBB model, permeability of the barrier to L-glucose was tested as previously described [27]. Briefly, a solution of PBS with HEPES 20 mM containing [ 14 C]-L-glucose 1 M (0.2 Ci/ L, Amersham Biosciences) was prepared. Medium was removed from both inner and outer well of the Transwell and the outer well was filled with 600 L of PBS containing HEPES 20 mM, while the inner well was filled with 100 L of the solution containing [ 14 C]-L-glucose. 10 L aliquots of the inner and outer well solution were removed at 10 minutes intervals for 1 hours, spotted onto glass fibre filters (Filtermat A, Wallac) and air-dried for 2 hours at room temperature. Glass fibre filters were than absorbed in liquid scintillation (Beta Plate Scint 1250-440) and counted (Trilux 1450 MicroBeta, Wallac).

Electron Microscopy (EM).
We checked for the presence of two crucial ultrastructural markers: the presence of HUVECs and astrocytes as a single layer, and the presence of tight junctions (sites of apparent fusion between two adjacent cells formed by the two opposite membranes and characterized by electron-dense material). After four days of coculture, specific polycarbonate membranes were removed from the inserts and specimens were fixed in 2,5% EM grade glutaraldehyde (Fluka Chemie, AG Buchs CH) in 0.05 M phosphate buffer (PB) pH 7.4, postfixed in 1% osmium tetroxide (EMS, Fort Washington, PA, USA) in 0.05 M PB, dehydrated in graded acetone and embedded in Spurr epoxy resin (EMS, Fort Washington PA USA). Thin sections 1 m thick were counterstained with toluidine blue and observed under optic microscope. Ultrathin sections (500Å thick) of selected areas including HUVECs/astrocytes monolayers were harvested on 200 mesh copper grids (EMS), stained with uranyl acetate and lead citrate, and viewed under an electron microscope (Zeiss EM 109, Germany).

Transmigration
Assay. Subconfluent cultured melanoma cells were harvested, washed in PBS, and labelled with the fluorescent tracer 5-6-carboxyfluorescein diacetate, succinimidyl ester (5(6)-CFDA, SE; CFSE) following manufacturer's instruction (Invitrogen, Italy). After CFDA labelling, cells were washed with PBS (with CaCl 2 e MgCl 2 ), centrifuged, and resuspended in complete medium (RPMI 10% FBS) at a concentration of 100.000 cells/100 L. 100 L of this solution were added to the inner well of the Transwell in which the in vitro BBB was established as previously described and incubated at 37 ∘ C. After 24 hours, transmigrated cells were collected in the outer well of the Transwell and counted by fluorescence microscopy using a Fuchs-Rosenthal chamber. Data regarding transmigrated cells are expressed as percentage of the total number of melanoma cells used. As controls, in any transmigration experiment, migration of melanoma cells through the polycarbonate membrane alone was also included. To confirm the presence of the endothelial and astrocytic monolayer, in each experimental session, one polycarbonate insert used in a transmigration experiment was processed for ultrastructural evaluation. Specimens were treated and viewed as described above; melanoma cells were identified by the presence of melanosomes in their cytoplasms.

In Vitro BBB Model.
We developed an in vitro BBB model which is a "contact through-feet" model. It is built with human endothelial cells deriving from HUVECs and rat astrocytes cocultured in Transwell 8.0 m porous polycarbonate membrane (see Materials and Methods for further details). We obtained a model that satisfied some of the in vivo BBB characteristics; by electron microscopy we demonstrated that our model allowed astrocytes endfeet to physically contact HUVECs through membrane pores (Figure 1(a)) inducing the formation of tight junctions (Figure 1(b)) which is a morphological feature indicating the acquisition of functional properties of the cerebral endothelium. To investigate the "tightness" of this model, we evaluated the permeability of our in vitro BBB to [ 14 C]-L-glucose [27]. We measured radioactivity content in the medium of HUVECs cocultured with astrocytes and of HUVECs or polycarbonate inserts alone sampling medium at 10 minutes intervals for 1 hour. The ratio between radioactivity measured in the lower chamber and the radioactivity measured in the upper chamber of the Transwell showed that L-glucose permeability was strongly reduced by the presence of the monolayer of HUVECs alone and even more reduced when HUVECs were cocultured with astrocytes ( Figure 2). The comparison between in vitro BBB and the HUVECs monolayer alone showed a statistically significant reduction in permeability as early as 10 minutes after the addition of [ 14 C]-L-glucose in the upper chamber. In the insert with polycarbonate membrane alone, the equilibrium was almost reached after 60 minutes (Figure 2).
The model was further characterised analysing the expression of zonula occludens-1(ZO-1) and vinculin ( Figure 3). The immunoreactivity for ZO-1 was localized at the boundaries of endothelial cells. Labelling for vinculin showed a slightly different pattern, mainly distributed as a diffuse cytoplasmic band at endothelial cell periphery and around the nucleus.

Transmigration
Assay. The ability of 7 metastatic melanoma primary cultures to transmigrate across our in vitro barrier model was screened in preliminary experiments to estimate if they had different ability to transpass the BBB model (data not shown).    Among the various primary cultures analyzed, two (named M1 and M2) were selected for more detailed analysis based on their different capability to transmigrate across the in vitro BBB model. The M2 cell culture showed a statistically significant higher capability to cross the in vitro BBB model when compared to M1 (Figure 4(a)). As a matter of fact, M1 showed a mean percentage of migrated cells of 0.038% while M2 showed a mean value of 0.13% ( = 0.000016, Student's -test).
To exclude major differences in generic motility of the two primary cultures in our experimental model, transmigration ability across the polycarbonate membrane alone was also measured and no differences were detected (Figure 4(b)). These data suggest that the presence of the in vitro BBB is actually responsible of the observed differences in cells transmigration among the two cell lines.
Since the experimental model forces cells to pass through 8 m pores, we assumed the cell volume as a relevant parameter in our assay. We therefore measured the cellular volume using flow cytometry; forward scatter was measured for both cell lines and results are reported in Figure 5. No statistically significant differences were detected in cellular volume between M1 and M2 cells.

Expression of Metastasis-Related Molecules.
To investigate putative molecular mechanisms responsible for the different transmigration behaviour observed in M1 and M2, we analyzed the expression of some proteins potentially involved in adhesion and transmigration process like integrins and adhesion molecules by cytofluorometry: CD49b, ICAM1, MCAM, CD56, v 3, VCAM1, ICAM2, and CCR7 were analysed and the data obtained are summarized in Figure 5. A statistically different expression was detected for v 3, CD49b, and MCAM, which were more expressed in M1 compared to M2 cells (Figures 6(a) and 6(b); < 0.05, Student's -test). On the other hand, a trend to decreased expression of ICAM-1 and CD56 was detected in M1 cells even if no statistically significant differences were found (Figure 6(a)).
To identify differentially expressed genes possibly contributing to the increased ability of M2 cells to cross the in vitro BBB, mRNA expression of 84 genes involved in tumor

Upregulated Genes.
Genes encoding for molecules involved in cell growth/proliferation (CG/P) were the most represented in the group of upregulated genes (54.8% see supplementary Table 1); however, genes involved in the synthesis of adhesion molecules and matrix degradation molecules are shown in Table 1(b).

Discussion
CNS metastasis is rather common among patients with melanoma and there is a growing interest in the development of in vitro BBB models to better understand the molecular mechanism of invasion of the CNS by malignant tumor cells.
The pivotal role of astrocytes in maintaining cerebral homeostasis, participating in neural signal transduction and transporting nutrients from the circulation to the neurons and buffering ionic imbalance, has been underlined by a number of authors [41][42][43][44]; in the present study we developed an in vitro contact through-feet BBB model built with human noncerebral endothelial cells and rat astrocytes.
The isolation of human cerebral endothelial cells is still a rare option; therefore in vitro models have been often developed with extracerebral human endothelial cells forced to assume a cerebral phenotype by the contact with astrocytic endfeet or by astrocyte-conditioned culture media [18,20,[45][46][47]. In our experimental conditions, the ability of astrocytic endfeet to take contact with HUVECs is clearly demonstrated by electron microscopy. Indeed, the HUVECs monolayer displayed tight junctions further supporting the idea that in coculture models noncerebral endothelial cells may be induced to assume features typical of the cerebral endothelium [18][19][20][20][21][22]. The tightness of our in vitro BBB model was further supported by the significant decrease in [ 14 C]-L-glucose permeability detected when both HUVECs and astrocytes were in coculture, in agreement with previously reported data about permeability of the BBB to [ 14 C]-L-glucose, sucrose, or drugs [27,[48][49][50]. Transmigration data showed that in our model the percentage of cells crossing the in vitro BBB model was less than 0.2%. This is in line with previously published data in similar models [51,52]. As a further control, we have tested the migration of A375 (a commercially available melanoma cell line) in our model and we found similar transmigration ability if compared to what was previously reported by other authors for the same cell line in a similar in vitro model (data not shown) [53]. However, also in vivo studies have shown that less than 0.1% of the cells survive more than 24 h after injection contributing to the concept that metastasization is an inefficient process [54].
In our experimental conditions, M1 and M2 cell migration was similar when tested through the sole polycarbonate membrane and was statistically different only if HUVECs and astrocytes where present in the model (Figure 3). Taken together, these findings suggest that the different behaviour of the two cell cultures was induced by the presence of the barrier, further supporting that in vitro BBB models might be useful to study CNS-specific migration.
Since the mean volume of M1 and M2 cells was similar, it seems that in our model the cell size was not a limiting event in transmigration. This observation further suggests that the different migratory behaviour depended on the in vitro barrier described rather than intrinsic differences between the two cell lines. However, arrest of tumour cells in vivo was found to take place at the level of capillaries where the diameter of the vessels is similar to those of the metastatic cells suggesting a key role for the cell size in early phases of metastasis formation [55,56]. Although in vitro models might be used in the study of the molecular mechanisms involved in the crossing of the BBB, they can hardly be used in elucidating all the complex phases of the metastatic process [14].
Surprisingly, our findings showed a general decrease in the expression of adhesion molecules in the melanoma primary cell culture displaying the higher in vitro migration rate. In fact, this decrease was significant for CD49b, NCAM, and alfaVbeta3 integrin in flow cytometry experiments and was confirmed by the reduced level of expression detected by real time PCR of these and of other adhesion molecules ( Table 1). The expression of beta3 integrin has been recently analyzed by Berghoff el al. [57] in a study on autopsy specimens which also included melanoma metastasis; beta3 integrin score was reported to be generally low in all metastasis and very low or absent in melanoma metastasis. Our observations are in agreement with these data and would support a negative role for beta3 integrin in CNS metastatization.
On the other hand, in contrast with several evidences supporting a role for MCAM overexpression in metastatic transformation of various types of tumors including melanoma [58] prostate cancer [59] and non-small-cell lung carcinoma [60], our data would suggest that a reduced expression of MCAM promotes migration. This hypothesis is also in contrast with previously published data which clearly showed a role for MCAM in cell motility in melanoma [61]. However, in our experimental conditions, we could not detect a statistically significant difference in the migration of the two cell lines through the sole polycarbonate membrane. Since MCAM was expressed on both cell lines, albeit at much lower level in M2, its role in our model might be marginal.
The increased ability to cross the in vitro BBB detected in M2 cells compared to M1 correlated with the higher expression of some of MMPs in this cell line. MMPs are primarily involved in extra cellular matrix degradation; however, the literature data also showed their complex role in creating and maintaining the appropriate microenvironment for tumor growth at primary and metastatic sites [62]. A role for MMPs in BBB disruption has been previously reported; indeed, Fazakas et al. showed that melanoma cells can metastasize to the brain disrupting the continuity of tight junctions by producing proteases and that protease protease inhibitor might significantly reduce the number of extravasating melanoma cells [63]. Although in a different cellular model, our results are in agreement with previously reported data showing that MMP2 and MMP9 are directly involved in TJs disruption [64][65][66]. Moreover they suggest that other MMPs (MMP7, MMP10, MMP11, and MMP13) might be implicated in CNS invasion by metastatic melanoma cells.
Due to the restricted accessibility to the CNS to chemotherapeutics, the prevention of the formation of a brain metastasis is of crucial importance; therefore the understanding of the mechanism involved in this complex process is a key step in developing new therapeutic options.