Modulation of extracellular matrix/adhesion molecule expression by BRG1 is associated with increased melanoma invasiveness

To evaluate BRG1 expression during melanoma progression, we examined BRG1 mRNA levels using quantitative (qPCR) arrays (Origene) containing normalized cDNA prepared from patient derived normal skin (3 samples), from stage III (21 samples) and stage IV (19 samples) metastatic melanoma specimen. Although BRG1 mRNA levels were lower in a subset of individual melanoma samples compared to normal skin, the average level of BRG1 was higher in stage III (1.2 fold) and stage IV melanoma (1.7 fold) compared to that in normal skin (Figure. 1A). The higher levels of BRG1 in stage IV melanoma compared to normal skin was statistically significant (p < .05). There was also a statistically significant increase in BRG1 mRNA levels in stage IV melanoma compared to stage III melanoma (p < .01). Although there was also a trend toward increased BRG1 expression in stage III melanoma compared to normal skin, the increase was not statistically significant, possibly due to an insufficient normal skin sample size. Interestingly, microarray profiling of primary melanoma tumors compared to normal skin revealed that BRG1 mRNA levels in primary melanoma is significantly higher than in normal skin [33, 34] (Additional file. 1). In combination, these data suggest that BRG1 mRNA levels are elevated in primary melanoma compared to normal skin and increase during disease progression (from stage III to IV).

We and others determined that SK-MEL5 cells, derived from an axillary node melanoma, are deficient in BRG1 expression [31, 32]. To determine whether BRG1 protein levels are consistently down regulated in other metastatic melanoma cell lines, we compared BRG1 protein levels in SK-MEL5 cells with levels in two highly metastatic melanoma cell lines, A375SM and WM-266-4. The A375SM cell line was established from a lung metastasis formed by injection of parental cells into nude mice [35]. The WM-266-4 cell line was derived from a lymph node metastasis [36]. We found that both A375SM and WM-266-4 express high levels of BRG1 compared to SK-MEL5 cells and to normal human melanocytes (Figure. 1B). We previously reported that re-introduction of BRG1 in SK-MEL5 cells promotes pigmentation as well as increased resistance to cisplatin [31]. As shown in Figure. 1B, BRG1 reconstituted SK-MEL5 cells express BRG1 at similar levels as A375SM and WM-266-4, which we previously estimated to be approximately 2 fold higher than that in normal melanocytes [31].

A previous microarray study showed that re-expression of BRG1 in a BRG1/BRM deficient human adrenal adenocarcinoma cell line (SW13 cells), activated the expression of 80 genes and repressed the expression of 2 genes [28]. Many of the BRG1 regulated genes were cell surface proteins and extracellular matrix remodeling enzymes or secreted proteins such as CD44, E-cadherin, matrix metalloproteinase (MMP) 2, and osteonectin (SPARC) [28‐30]. Thus, re-expression of BRG1 in BRG1/BRM deficient adenocarcinoma cells alters the expression of a subset of genes, and in particular the expression of genes that potentially have important roles in regulating tumor metastasis.

To evaluate how re-expression of BRG1 in the BRG1 deficient melanoma cell line, SK-MEL5, alters the expression of metastasis associated gene expression, we examined BRG1 induced changes in gene expression using quantitative RT² Profiler PCR Arrays (SABiosciences) and assayed the expression of 84 genes related to cell-cell and cell matrix interactions (Additional file 2). We found that the expression of 13 genes on the array was highly up-regulated by BRG1 (greater than 4-fold) (Figure. 2A). The most highly up-regulated genes (greater than 10 fold) were neural cell adhesion molecule (NCAM1), E-cadherin (CDH1), catenin delta 2/neural plakophilin related armadillo protein (CTNND2), MMP2, and laminin β3 (LAMB3) (Figure. 2A). Other highly activated genes (greater than 4 fold) included MMP10, tissue specific inhibitor of metalloproteinase (TIMP) 3, integrins α3 and α7, two collagen genes, and genes encoding components of the basement membrane (Figure. 2A).

BRG1 activated the expression of 10 additional genes at least two fold, including CD44, MMP9 and MMP14 (MT1-MMP) (greater than 2 fold) (Figure. 2B). Interestingly, re-expression of BRG1 also significantly inhibited the expression of 8 genes (Figure. 2C), while the remaining 53 genes on the array were not significantly affected by the expression of BRG1 (Additional file 1, Table 1). Thus our data indicate that re-expression of BRG1 in BRG1 deficient melanoma cells affects the expression of a subset of cell surface and extracellular matrix remodeling enzymes, some of which overlap (E-cadherin, CD44, and MMP2) and some which are distinct from those reported to be modulated by reconstitution of BRG1 in BRG1/BRM deficient SW13 adenocarcinoma cells. Many of the genes we found to be modulated by BRG1 (Figure. 2A, B, and 2C) encode proteins that play a role in regulating melanoma invasiveness and metastatic potential [6, 7, 37].

The most highly activated gene in BRG1 reconstituted SK-MEL5 cells was NCAM1 (Figure. 2A). NCAM1 is a cell adhesion molecule (CAM) in the immunoglobulin superfamily that is expressed at the cell surface and mediates cell to cell and cell matrix interactions [38]. High expression of NCAM1 in malignant neoplasms, including melanoma, is associated with an aggressive tumor phenotype [39]. Although high levels of NCAM1 have been associated with metastatic potential, the functional role of NCAM1 in melanoma has not been demonstrated, and high levels of NCAM1 have also been detected in benign nevi [40]. Thus, the role of NCAM1 in melanoma metastasis is unclear. MCAM (MUC18), a related cell adhesion molecule is over-expressed in advanced primary and metastatic melanoma. Its expression in melanoma cell lines enhances metastatic potential in nude mice [41, 42]. We found that in addition to NCAM1, BRG1 significantly increased the expression of MCAM (Figure. 2D). Thus, re-expression of BRG1 in SK-MEL5 cells activated the expression of two related cell adhesion molecules that have been implicated in promoting tumor metastasis. We verified that the changes in NCAM1 and MCAM expression also occurred at the protein level (Figure. 2E). Interestingly, increased levels of the 140KD NCAM1 isoform was detected in BRG1 expressing cells. This isoform is associated with malignant neoplasms and induction of anti-apoptotic programs [39].

Two of the most highly activated genes in BRG1 expressing SK-MEL5 cells were E-cadherin (CDH1) and catenin delta 2/neural plakophilin related armadillo protein (CTNND2) (Figure. 2A). E-cadherin is a calcium dependent transmembrane receptor that localizes to adherens junctions and mediates cell-cell adhesion. In many cancer types, loss of E-cadherin expression coincides with acquisition of an invasive phenotype and development of metastatic disease. In normal melanocytes, E-cadherin mediates melanocyte-keratinocyte interactions and loss of E-cadherin expression or a change in its cellular distribution is associated with early phases of melanoma. Furthermore, over-expression of E-cadherin in melanoma cells reduces melanoma invasiveness [43]. Thus, expression of BRG1 in SK-MEL5 cells could potentially reduce melanoma invasiveness through up-regulation of E-cadherin. Interestingly, BRG1 also promoted expression of δ-catenin/neural plakophilin-related armadillo protein (CTNND2) (Figure. 2A), but had no effect on the expression of β-catenin or α-catenin (data not shown), two other members of armadillo/β-catenin superfamily of cell adhesion molecules. Increased expression of CTNND2 in prostate cancer has been associated with redistribution and loss of E-cadherin at the adherens junction [44].

We verified that reconstitution of BRG1 in SK-MEL5 cells resulted in increased E-cadherin and CTNND2 expression at the protein level (Figure. 3A). To determine the status of E-cadherin at the cell surface in control SK-MEL5 cells and SK-MEL5 cells expressing BRG1, we performed flow cytometry. We found that although total E-cadherin expression increased (Figure. 3A), the localization of E-cadherin to the cell surface was reduced in cells expressing BRG1 compared to control cells (Figure. 3B). Furthermore, immunofluorescence revealed that E-cadherin was mostly cytoplasmic in BRG1 expressing SK-MEL5 cells (Figure. 3C). Reduced localization of E-cadherin to the cell surface suggested that in SK-MEL5 cells, re-expression of BRG1 may further compromise E-cadherin function.

Re-expression of BRG1 in SK-MEL5 cells resulted in an altered pattern of integrin expression (Figures. 2A and 2C). Integrins are transmembrane glycoproteins that mediate specific interactions between cells and the ECM and regulate migration [45]. Hetero-dimers composed of α and β subunits serve as receptors with specificity for different ligands. Integrin expression is a key determinant of a cell's ability to attach to different ECM components and to migrate on these substrates. Aberrant integrin expression has been associated with melanoma progression [45].

BRG1 enhanced the expression of integrins α7 and α3 and inhibited the expression of integrins α4 and β3 (Figures. 2A and 2C). Modulation of integrin expression by BRG1 suggested that reconstitution of BRG1 in BRG1 deficient melanoma cells might alter melanoma cell interactions with specific ECM components. We compared the ability of the control (empty vector) SK-MEL5 cells with that of the SK-MEL5 cells expressing BRG1 to adhere to laminin, collagen, and fibronectin. We found that BRG1 expressing cells demonstrated increased adhesion to laminin and collagen and decreased adhesion to fibronectin (Figure. 4). The observed increase in adhesion to laminin is consistent with increased expression of integrin α7, which is a component of α7β1, a complex that has high affinity for laminin [46]. Increased adhesion to collagen is consistent with increased expression of α3, which is a component of the α3β1 complex that has high affinity for several ECM components, including collagen [45]. Reduced adhesion to fibronectin is consistent with decreased expression of α4, which forms the α4β1 complex and β3 which forms the αVβ3 complex, two integrins with high affinity for fibronectin [45]. The expression of these integrins is elevated in primary or metastatic melanomas [47‐50], however it is not possible to designate specific integrins as "pro-neoplastic" because their effect on tumor progression is dependent on the cellular context and the specific step in tumor progression [51]. Our data indicate that BRG1 may regulate metastatic potential by modulating the integrin profile and altering adhesiveness to different ECM components.

In addition to changes in adhesion, metastasis also requires extensive ECM remodeling. The matrix metalloproteinases (MMPs) are the main proteases that remodel the ECM [52]. Re-expression of BRG1 in SK-MEL5 cells resulted in a dramatic increase in MMP2 and MMP10 expression and a smaller but significant increase in MMP9 and MMP14 (MT-MMP1) expression (Figures. 2A and 2B) and a decrease in MMP1 and MMP16 expression (Figure. 2C). We verified that the observed changes in the mRNA profile resulted in consistent changes in protein expression for MMP1, MMP2, MMP9, and MMP14 (Figure. 5A).

Expression of MMPs is controlled at the transcriptional and post-translational levels [53]. Our data indicated that BRG1 promotes expression of MMP2, MMP9, and MMP14 at the protein level (Figure. 5A). MMP2 (gelatinase A, 72-kDa type IV collagenase) and MMP9 (gelatinase B, 92-kDa type IV collagenase are secreted as inactive pro-zymogens that are subsequently processed and activated. MMP14 (MT1-MMP) is a membrane bound MMP that activates MMP2 at the cell surface [54]. Furthermore, naturally occurring tissue inhibitor of metalloproteinases (TIMPs) down-regulate MMP activity [55]. The balance between TIMP and MMP expression is critically important in determining overall MMP activity. We found that in addition to MMPs, BRG1 also activated expression of TIMP2 and TIMP3, which would be expected to down-modulate MMP activity (Figures. 2A, 2B, and 5A).

In order to determine if re-expression of BRG1 in SK-MEL5 cells resulted in increased secretion of active MMP2 and MMP9, we performed gelatin zymography on supernatants derived from control and BRG1 expressing SK-MEL5 cells. We determined that although TIMP levels were increased, there was still a substantial increase in active MMP2 and MMP9 secreted by SK-MEL5 cells expressing BRG1 compared to BRG1 deficient SK-MEL5 cells (Figure. 5B).

The observed increase in MMP2 and MMP9 activity as well as other alterations in extracellular matrix and adhesion molecule expression suggested that BRG1 plays an important role in regulating melanoma invasiveness. To determine the overall biological consequence of BRG1 re-expression in SK-MEL5 cells, we investigated whether BRG1 promotes changes in the ability of melanoma cells to be invasive in vitro. We found that SK-MEL5 cells that express BRG1 had significantly increased ability to invade through Matrigel coated Boyden chambers (Figure. 5C).

To elucidate the mechanisms by which BRG1 promotes invasion, we treated cells with an inhibitor of MMP2/MMP9 and performed invasion assays. We found that inhibition of MMP2 and MMP9 activity partially abrogated the BRG1 mediated increase in invasive ability (Figure. 5D). Consistently, siRNA mediated down-regulation of MMP2 (Figure. 5E) also reduced the BRG1 medicated increase in invasiveness (Figure. 5F). Thus, activation of MMP2 and possibly MMP9 expression contributes to the BRG1 induced increase in SK-MEL5 invasive ability.

Most established melanoma cell lines express high levels of BRG1 [31], including two metastatic melanoma cell lines, A375SM and WM-266-4 (Figure. 1B). This raised the possibility that BRG1 is required for these cells to be invasive. To determine if loss of BRG1 compromises invasive ability in one of these highly invasive cell lines, we down-regulated BRG1 expression in WM-266-4 cells using a pool of siRNAs that target BRG1 but not the alternative ATPase, BRM (Figure. 6A and 6B). We performed a timecourse after siRNA transfection and determined that BRG1 down-regulation was effective 120 hours after transfection (Figure. 6B). Interestingly, BRM expression was slightly lower in cells transfected with control siRNA compared to untreated but then increased in BRG1 down-regulated cells. However, expression of the BRG1/BRM associated factor, INI1, did not change as a result of siRNA transfection. Previous studies have suggested that BRM expression is highly sensitive to growth conditions [56]. We found that in WM-266-4 cells, BRM expression but not BRG1 or INI1 expression is sensitive to changes in WM-266-2 confluency (Additional file 3). Thus, we speculate that siRNA transfection may have an inhibitory effect on BRM expression because of non-specific effects on proliferation. Nevertheless, BRM expression was elevated in BRG1 knockdown cells compared to both untreated cells and cells that expressed control siRNA (Figure. 6B).

We found that down-regulation of BRG1 resulted in decreased MMP2 and MCAM expression (Figure. 6C) and reduced invasion through Matrigel-coated Boyden chambers (Figure. 6D). Furthermore, although BRM levels increased in BRG1 down-regulated cells, our data suggest that BRM can not compensate for these BRG1 specific functions. Thus, both a gain of function and loss of function approach show that high levels of BRG1 promote melanoma invasive ability in vitro.

Our data suggested that activation of MMP2 is an important mechanism by which BRG1 promotes melanoma cell invasive ability. To determine the mechanism by which BRG1 activates MMP2 expression in SK-MEL5 melanoma cells, we investigated whether BRG1 intereacts with a transcriptional regulator of MMP2. BRG1 was previously shown to directly activate the MMP2 promoter through interactions with the transcription factor, SP1 in SW13 cells [30]. Similarly, we found that siRNA knockdown of SP1 (Figure. 7A) reduced the level of MMP2 that was secreted by SK-MEL5+BRG1 cells (Figure. 7B). Furthermore, we detected a physical interaction between BRG1 and SP1 (Figure. 7C) and found that BRG1 was recruited to the MMP2 promoter (Figure. 7D). As was previously demonstrated in SW13 cells, BRG1 significantly increased the binding of SP1 to the MMP2 promoter [57] (Figure. 7E). This data suggests that BRG1 directly regulates MMP2 expression in melanoma cells through interactions with SP1 and by facilitating SP1 association with the MMP2 promoter. Interestingly, SP1 has been shown to preferentially interact with the BRG1 catalytic subunit in vitro [57]. Thus, a specific role for BRG1 in the activation of MMP2 and melanoma invasiveness may result from selective interactions with the SP1 transcriptional regulator.

SK-MEL5 and WM-2664 melanoma cells were from the ATCC. A375SM melanoma cells were a kind gift from Dr. Menashe Bar-Eli (M.D. Anderson Cancer Center). SK-MEL5 cells expressing an empty vector or BRG1 were described in [31]. Human melanocytes were from Cascade Biologics (Portland, Oregon, USA) or Yale Cell Culture Core Facility (New Haven, Connecticut, USA). With the exception of melanocytes, all cells were grown in DMEM supplemented with 10% FBS. Human melanocytes were grown in Media 254 with added growth supplements (Cascade Biologics). The MMP2/MMP9 inhibitor, 4-Biphenylylsulfonyl)amino-N-hydroxy-3-phenylpropionamide (BiPS) was from Calbiochem (San Diego, CA, USA) and was used at 10 μM.

Total RNA was isolated with the Qiagen RNeasy mini kit and reverse transcribed as described [31]. Quantitative real-time PCR was performed in SYBR Green Master Mix (Qiagen, Germantown, Maryland) with an Applied Biosystems Prism 7500 PCR system and analyzed with the SDS software as described [31]. MCAM and GAPDH primers were purchased from SABiosciences (Frederick, MD, USA).

The Tissue Scan Melanoma qPCR Arrays (MERT501) containing cDNAs from normal skin, stage III, and stage IV melanomas were obtained from Origene Technologies (Rockville, MD, USA). The primers used to detect BRG1 (SMARCA4) were from SABiosciences (Frederick, MD, USA). BRG1 levels were normalized by amplifying with primers to Human β-actin (Forward: CAGCCATGTACGTTGCTATCCAGG) and (Reverse: AGGTCCAGACGCAGGATGGCATG). The results were averaged from values obtained by running three PCR arrays. Statistical significance was determined by utilizing a Mann-Whitney Wilcoxon test.

Extracellular Matrix and Adhesion molecule RT² Profiler PCR Arrays were purchased from SABiosciences (Fedrick, MD). The primer sets in this array are described in http://​www.​sabiosciences.​com/​rt_​pcr_​product/​HTML/​PAHS-013A.​html. CT values obtained for 84 extracellular matrix and adhesion molecule gene expression were normalized to a value obtained by averaging the CT values of four different housekeeping genes. For each primer set, the fold change in SK-MEL5+BRG1 cells was determined relative to values obtained in control SK-MEL5 cells +empty vector. Average values were obtained from four PCR arrays with cDNA from control cells (from three different samples) and an additional four PCR arrays with cDNA from SK-MEL5 cells +BRG1 (from three different samples). Statistical significance was calculated using the student's t test.

The Tubulin antibody was from Sigma (St. Louis, Missouri, USA). FLAG M2 antibody and FLAG M2-Agarose were from Sigma. The E-cadherin, CTNND2, and MCAM antibodies were from BD Biosciences (San Jose, CA, USA), The BRG1, NCAM1 and SP1 antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA, USA). The MMP2 and MMP9 antibodies were from Cell Signaling (Beverly, MA, USA). The MMP14 antibody was from Millipore (Temecula, CA, USA). The TIMP3 antibody was from Abcam (Cambridge, MA, USA). Control IgG antibody used for ChIPs was from Millipore (Billerica, MA, USA).

Cells were lysed in 20mMTris (pH 7.4),150 mM NaCl, 2 mM EDTA, 1% Triton X, 10% glycerol, supplemented with a protease inhibitor cocktail (Sigma). SDS-PAGE and Western blotting were carried out as described [73].

Cells were incubated in fetal bovine calf serum (Invitrogen, Carlsbad, CA, USA) for 10 minutes at room temperature to block nonspecific antibody binding and then with the primary antibody or an isotype matched IgG control for 20 minutes at 4°C. After one wash with FACS buffer (PBS+0.5%BSA, 5% fetal calf serum, 0.1% sodium azide, cells were incubated with secondary antibody for 20 minutes at 4°C, then washed twice with FACs buffer. Cells were re-suspended in 0.1% paraformaldehyde then loaded onto a FACS-Calibur (BD Biosciences, San Jose, CA, USA). Data was analyzed using Cell Quest Pro (BD Biosciences). Statistical significance was calculated using the student's t test.

Immunocytochemistry was performed as described [17] using an E-cadherin antibody (BD Biosciences) and goat anti-mouse-Alexa Fluor568 (Molecular Probes). Images were taken with a Nikon Eclipse TE2000-U fluorescence microscope at 60× magnification.

Zymography was performed as previously described [30]. Control SK-MEL5 and SK-MEL5+BRG1 cells were cultured in serum free medium for 36 hours. Conditioned medium was collected, normalized to cell number, and subjected to electrophoresis in a polyacrylamide gel containing 1 mg/ml gelatin. After electrophoresis, the gel was washed in 2.5% Triton X-100 for 1 hour at room temperature to remove the SDS and then incubated for 24 hours at 37°C in a buffer consisting of 5 mM CaCl₂ and 1 μM ZnCl₂. The gel was stained in 0.25% Coomasie Blue for 30 minutes, de-stained in methanol/acetic acid solution and photographed on a light box. Proteolytic activity was detected as white bands against a blue background.