Brg-1 mediates the constitutive and fenretinide-induced expression of SPARC in mammary carcinoma cells via its interaction with transcription factor Sp1

The 5'-flanking region of the mouse SPARC gene has a single major transcriptional start site; no TATA or CAAT boxes but six repeats of a GGAGG sequence within the proximal region of the SPARC promoter (-124/+19) (Figure 1A), as described previously [52]. These characteristics are well conserved in the SPARC gene among various species, including mouse, chicken, cow and human [53]. This GGAGG sequence represents one type of the regulatory sequence GC box (GC-I), which has been shown to bind transcription factor Sp1/3, regulating transcriptional activity of several genes including Adam8, Pgr and Adamts1 [54‐56].

To determine whether Sp1 and Brg-1 bind to the GGAGG repeats sequence present in the mouse SPARC promoter, we employed an immobilized-template assay using a biotin-tagged probe a (spanning from nucleotide -130 to -56 of the SPARC promoter, Figure 1A), which allowed us to isolate the transcription factors that bind to this probe. Using Western blot analysis, we were able to document that both Sp1 and Brg-1 interacted with this region of the SPARC gene (Figure 1B, middle panel). The negative control, p38, was not associated with the probe a. A probe b (spanning from nucleotide -50 to +19 of the SPARC promoter), used as a non-specific binding control, was bound neither by Sp1 nor by Brg-1 (Figure 1B, right panel). The three well characterized mammary tumor cell lines (4T1, 168FARN and 67NR) used in our study are derived from a single, spontaneously arising mouse mammary tumor which differ from each other in their metastatic potential [57]. When injected into the mammary gland of mice, these tumor lines are either non-metastatic (67NR), spontaneously metastatic to lymph node (168FARN) or metastatic to lung, liver, bone and brain via the hematogenous route (4T1) [58]. The nuclear expression of Sp1 and Brg-1 were also determined by Western blot analysis. As shown in Figure 1B (left panel), there was no obvious difference in the protein level of Sp1 present in the three cell lines; whereas the protein level of Brg-1 declined in a concordant manner with the increase of metastatic potential, the latter being highest in non-metastatic 67NR cells, intermediate in partly metastatic 168FARN cells and lowest in highly metastatic 4T1 cells. To investigate whether Sp1 and Brg-1 are associated with the proximal region of the SPARC gene in living cells, we performed ChIP assays using antibodies to Sp1 and Brg-1, and nonspecific IgG as a control. A pair of primers spanning a 179-bp DNA fragment (-201/-23) of the proximal region encompassing the GGAGG repeats was used for PCR. As shown in Figure 1C, both the antibodies recognizing Sp1 and Brg-1 precipitated this promoter fragment, while the nonspecific IgG failed to do so. Further, the occupancy levels of Brg-1 and Sp1 at the SPARC promoter were analyzed by quantitative ChIP analysis. We found that, consistent with the nuclear Brg-1 protein level, Brg-1 occupancy likewise decreased in a concordant manner with the increase of metastatic potential in the three cell lines. The difference of Sp1 occupancy level was not significantly different among the three cell lines (Figure 1D). Our results demonstrate that both Sp1 and Brg-1 bind to the same proximal region of the SPARC promoter and that the level of Brg-1 expression and its binding to the SPARC promoter seem to negatively correlate with the metastatic potential of the three cell lines.

To determine whether Sp1 is specifically responsible for recruitment of Brg-1 to the SPARC promoter, we knocked down Sp1 expression using RNA interference. The effect of Sp1 knockdown on the binding of Brg-1 to the GGAGG repeats sequence of the SPARC promoter was then analyzed by using an immobilized-template assay. As shown in Figure 2A, transfection of specific Sp1 siRNAs into 4T1 cells led to a substantial down-regulation of Sp1 protein levels (decreased by > 76%), without any effect on Brg-1 protein expression. Transfection using a control siRNA had no effect on either Sp1 or Brg-1 levels. Interestingly, binding of Brg-1 to the SPARC promoter was significantly diminished when using nuclear extract from Sp1 siRNA-transfected cells (Figure 2B). To further test the role of Sp1 in mediating Brg-1 binding to the SPARC promoter in living cells, a ChIP-qPCR assay was performed using 4T1 cells transfected with Sp1 siRNA. As shown in Figure 2C, transfection with Sp1 siRNA significantly diminished the amount of Brg-1 bound to the SPARC promoter. These results suggest that the Sp1 transcription factor is essential for efficient binding of Brg-1 to the SPARC promoter.

Since depletion of Sp1 diminished Brg1 binding to the SPARC promoter, we postulated that Brg-1 is recruited to the SPARC promoter by forming a complex with Sp1. To test this hypothesis, we performed co-immunoprecipitation experiments using nuclear extracts from 4T1, 168FARN and 67NR cells. As shown in Figure 3A, Sp1 was found in the complex with Brg-1, but not with the IgG control. These co-immunoprecipitation assays were repeated in a reciprocal fashion using an anti-Sp1 antibody. Association of Sp1 and Brg-1 was again detected by immunoblotting of these complexes with anti-Brg-1 antibodies (Figure 3B). We next determined by re-ChIP analyses whether Brg-1 is associated with the Sp1-containing SPARC promoter fragment in living cells. Briefly, chromatin fragments from cross-linked 4T1, 168FARN and 67NR cells were first immunoprecipitated using antibodies against Sp1. The resulting immunocomplex was then eluted and subjected to immunoprecipitation using antibodies against Brg-1. PCR analysis was then used to determine whether the promoter fragment was present in the final precipitate. As shown in Figure 3C, we found that the promoter fragment present in the first immunocomplex generated using anti-Sp1 antibodies was pulled down again by anti- Brg-1 antibodies. These results indicate an association between Brg-1 and Sp-1 at the proximal promoter region containing GGAGG repeats in these cells.

To further confirm the importance of Brg-1 in SPARC gene expression, we used RNA interference to deplete endogenous Brg-1. As shown in Figure 4A, transfection of 4T1 cells with different concentrations of siRNA targeting Brg-1 (0-50nM) resulted in an inhibition of Brg-1 mRNA expression in a dose-dependent manner (upper panel), accompanied by a dose-dependent decrease in SPARC mRNA level (bottom panel). Furthermore, Western blot analysis showed that transfection with 30 nM of Brg-1 siRNA led to a substantial down-regulation of Brg-1 protein levels (>80% decrease), with no effect on transcription factor Sp1 and β-actin expression. Transfection using a control siRNA had no effect on Brg-1, Sp1 or β-actin protein levels. Similarly to what was observed in the Sp1 knockdown experiments also, SPARC protein level was significantly inhibited in Brg-1 knockdown cells (reduced by 69%, Figure 4B). As SPARC is a secreted protein, in order to investigate if Brg-1 knockdown also leads to a decrease of SPARC secretion into the extracellular medium, cell media were collected for Western blot analysis. As shown in Figure 4C, a marked decrease in secreted SPARC was detected in the media harvested from Brg-1 siRNA-transfected cells in comparison to mock-treated cells. The cell-free medium (Med) was used as a control and no signal was detectable using the specific antibody against murine SPARC. To correct for any potential discrepancy in cell number resulting from different experimental conditions, densitometric quantitation was performed and results were adjusted for total protein amounts of cell lysates. After correction, SPARC secretion reduced by 72% in Brg-1 knockdown cells compared with mock-treated cells (P < 0.001). To further study the influence of Brg-1 on SPARC promoter activity, we generated a reporter pREP4-SP-Luc which contains a 220 bp fragment of the SPARC promoter (spanning nucleotides -201 to +19). This particular promoter region of the SPARC gene was previously shown to be necessary and sufficient to maintain constitutive SPARC gene expression, and our present results demonstrated that both Sp1 and Brg-1 are bound to this promoter region. The SPARC promoter-luciferase reporter was then introduced into 4T1 cells with an empty plasmid (as a control) or a plasmid expressing wild-type Brg-1 or mutant Brg-1 (K798R) defective in ATPase activity. Luciferase activity was determined 48 hrs later. As shown in Figure 4D, overexpression of wild-type Brg-1 significantly augmented the luciferase activity driven by the SPARC promoter, whereas mutant Brg-1 (K798R) had a much less pronounced enhancing effect on the luciferase activity. These results document a causal relationship between the level of Brg-1 protein and SPARC gene expression.

To test if the Sp1 is required for the Brg-1-mediated transcriptional activation of the SPARC gene, we introduced Sp1 siRNAs and Brg-1 expression plasmids into the 4T1 cells to inhibit the expression of the Sp1 and enhance the expression of the Brg-1. As shown in Figure 5A, overexpression of Brg-1 increased the SPARC promoter-driven reporter gene expression by 4.7-fold in 4T1 cells without Sp1 knockdown, whereas knocking down Sp1 reduced the induction of SPARC promoter-driven reporter gene expression by Brg-1 from 4.7-fold to 0.7-fold. These results demonstrate that Sp1 plays an important role in Brg-1 mediated transcriptional activation of the SPARC gene.

To exclude the possibility that Sp1 siRNAs may modify Brg-1 protein expression, we examined Brg-1 protein levels in 4T1 cells transfected with control siRNA or siRNA specific for Sp1. As shown in Figure 5B, knockdown of Sp1 did not make a significant impact on the Brg-1 expression level (compare lanes 1 and 2 as well as lanes 3 and 4). The effect of Brg-1 overexpression on the Sp1 protein expression was also analyzed in 4T1 cells transfected with empty or Brg-1-encoding plasmids. Similarly, Sp1 protein expression was not affected significantly by the overexpression of Brg-1 protein (compare lanes 1 and 3).

We found that the endogenous expression of Brg-1 decreased in a concordant manner with the increase of metastatic potential among the three tumor cell lines. We next documented a causal relationship between the level of Brg-1 protein and SPARC gene expression in 4T1 cells. These observations led us to believe that the difference of Brg-1 gene expression could be responsible for the differential expression of SPARC in the three cell lines. To investigate, the SPARC mRNA and protein levels were assessed. As we expected, in parallel with the Brg-1 protein level, the SPARC mRNA and protein levels were highest in 67NR, intermediate in 168 FARN and lowest in 4T1 cells (Figures 6A, B and Figure 1B, left panel). In addition, Western blot analysis of secreted proteins demonstrated that the amount of released SPARC in the conditioned medium is correlated with the intracellular mRNA and protein expression levels of this gene (compare Figure 6C with Figure 6A and 6B). These results suggest that the SPARC expression and secretion are correlated with the Brg-1 expression level. Our results also suggest that higher expression and secretion of SPARC might be correlated with lower metastatic potential of mammary carcinoma cells.

Fenretinide has been shown to have a long-term protective effect against secondary breast cancers [45, 51, 59]. Since we observed an interesting association between the relatively low level of SPARC expression and the high metastatic potential of 4T1 cells, we postulated that treatment with fenretinide might reverse the tumorigenic ability of these cells by affecting the expression of SPARC. To test this hypothesis, the expression of SPARC was detected by RT-qPCR and Western blot analysis in 4T1, 168 FARN and 67NR cells treated with fenretinide at different concentrations (0-5 μM) for 24 hrs. As shown in Figures 7A and B, fenretinide induced the expression of SPARC mRNA and protein in a dose-dependent manner. To investigate if fenretinide affects the secretion of SPARC protein, culture media were collected from the 4T1 and 67NR cells treated with fenretinide for 24 hrs and secreted SPARC was assessed. As expected, fenretinide increased the secretion level of SPARC into the extracellular media in a dose-dependent manner, too (Figure 7C). To test whether the fenretinide-induced expression of SPARC resulted from the transcriptional activation, reporter vectors (pREP4-SP-Luc) were transfected into the three cell lines. The transfected cells were then left untreated or treated with different concentrations of fenretinide for 24 hrs. As shown in Figure 7D, fenretinide increased SPARC promoter-driven luciferase activity in all three cell lines in a dose-dependent manner. The results demonstrate that fenretinide treatment increases the transcriptional activity and expression of the SPARC gene.

To explore in detail fenretinide-induced transcriptional activation of the SPARC gene, the expression and the ability of Brg-1 and Sp1 binding to the SPARC promoter were analyzed in all three cell lines in response to fenretinide treatment. As shown in Figures 8A and 8B, fenretinide induced the expression of Brg-1 in a dose-dependent manner, while it had no significant effect on the expression of Sp1. In addition, treatment of cells with fenretinide increased the interaction between Sp1 and Brg-1 (Figure 8C). To directly examine the effect of fenretinide on the binding of Brg-1 and Sp1 proteins to the SPARC promoter, we performed ChIP-qPCR to assess the occupancy level of Brg-1 and Sp1. As shown in Figure 8D, the occupancies of Brg-1 protein at the selected promoter region increased upon fenretinide treatment in a dose-dependent manner. Fenretinide had slight/no effect on occupancy level of Sp1 at this region. These results suggested that fenretinide induced the formation of the Sp1/Brg-1 complex, resulting in facilitating the access of Brg-1 to the SPARC promoter and enhancing the transactivation of this gene.

Since the expression of Brg-1 as well as the formation and binding of Sp1/Brg-1 to the SPARC promoter were increased in fenretinide-treated cells, it is likely that Brg-1 is involved in fenretinide-induced transcriptional activation of the SPARC gene. We employed Brg-1 siRNAs to knockdown Brg-1, and its effect on fenretinide-induced transcriptional regulation of the SPARC gene was then assessed by luciferase reporter assays. As shown in Figure 9A, fenretinide-induced promoter activity was significantly inhibited in Brg-1 knockdown cells. Interestingly, fenretinide-induced SPARC gene expression was significantly inhibited both at the protein and mRNA levels in Brg-1 knockdown cells (Figures 9B and 9C, respectively). These results reveal that the Sp1/Brg-1 complex plays an important role in fenretinide-induced SPARC gene expression, demonstrating that the Brg-1 protein is responsible for fenretinide-induced SPARC gene transactivation.

Fenretinide has been shown to inhibit the invasion of several cancer cell types, including breast cancer [60], ovarian cancer [61], prostate cancer [62] and Kaposi's sarcoma [63]. Therefore, we next tested whether fenretinide as well as fenretinide-induced SPARC expression have effects on the migration and invasion of 4T1 cells. Our results demonstrated that fenretinide suppressed the cell motility in a dose-dependent manner and the inhibition of cell motility was not significantly affected by addition of anti-SPARC antibodies (Figure 10A). Similarly, 4T1 cells treated with fenretinide showed a significant decrease in cell invasion compared with untreated cells. However, when cells were incubated with anti-SPARC antibodies, we found that cell invasion inhibited by fenretinide treatment was partly restored. On the contrary, addition of non-specific goat IgG had no effect (Figure 10B). The results suggest that SPARC is one of the mediators involved in the fenretinide-induced suppression of cell invasion.

Mammary tumor cell lines, 4T1, 168FARN and 67NR, were generously provided by Dr. F. Miller (Barbara Ann Karmanos Cancer Institute, MI, USA), and maintained in DMEM supplemented with 7% fetal bovine serum and 1% streptomycin-penicillin, and incubated in a humidified atmosphere containing 5% CO₂ at 37°C.

pREP4-luc was constructed as described previously [77]. pREP-SP-luc was constructed by inserting the PCR-amplified SPARC promoter (spanning nucleotides -201/+19 of SPARC gene), using the forward primer (5'-GAGCTAGCTGTCTGGGT AGCACACAGCCTAC-3') and reverse primer (5'-CAAAGCTTCTGAAGGGCTGC AGGAATGTG-3'), into the Nhe I-Hin dIII sites of pREP4-luc. Expression plasmid encoding human Brg-1 (pBJ5 BRG1) and the ATPase-defective variant of Brg-1 (pBJ5 BRG1 DN, K798R mutant) [78] were obtained from AddGene (Cambridge, MA, USA), and human Brg-1 was shown to be expressed correctly and work properly in mouse cells and in frog oocytes [79, 80]. Fenretinide (2,4,6,8-Nonatetraenamide, N-(4-hydroxyphenyl)-3,7-dimethyl-9-(2,6,6-trimethyl-1-cyclohexen-1-yl-(all-E); 374551) powder was generously provided by Dr. R. Smith (NIH, Bethesda, Maryland, USA).

After appropriate treatments, cells were collected and total cellular extracts or nuclear fractions were prepared. Western blot analysis was then performed as described previously [81]. A monoclonal antibody against β-actin (Sigma, Saint Louis, MO, USA) was used at a 1:5,000 dilution. A monoclonal antibody against SPARC (R&D systems, Minneapolis, MN, USA) was reconstituted at a concentration of 500 μg/ml and used at a 1:2,500 dilution. A rabbit polyclonal antibody against Brg-1 (sc10768×, Santa Cruz Biotechnology, Santa Cruz, CA, USA) or Sp1 (sc14027×, Santa Cruz Biotechnology) was used at a 1:4,000 dilution.

Immobilized-template assays were performed as described previously [82]. Two hundred micrograms of Dynabeads M280 streptavidin (Dynal) was prepared, concentrated and resuspended in 20 μl of buffer T (10 mM Tris [pH 7.5], 1 mM EDTA, 1 M NaCl) including 10 pmol of biotinylated GGAGG-rich-containing either probe a (spanning nucleotides -130/-56 of the SPARC gene [gene bank accession #M20683]) or probe b (spanning nucleotides -50/+19 of the SPARC gene). The mixture was gently agitated for 1 hr at room temperature (RT) and the beads were then washed 4 times in buffer T to remove unbound probes. Bead-coupled probes were equilibrated in buffer R (10 mM Tris [pH 7.5], 1 mM MgCl₂, 0.1% NP-40, 1 mM EDTA, 10 mM DTT, 5% glycerol, 60 mM KCl, 12 mM HEPES [pH 7.9], 0.03% BSA) for 30 min, centrifuged and resuspended in buffer R containing 200 μg of nuclear extract and 40 ng/μl of poly (dG-dC) (120 μl final volume), and agitated for 30 min at RT. After binding reaction, the beads were washed three times using buffer R containing 10 ng/μl of poly (dG-dC). The bound proteins were eluted by boiling them in SDS sample buffer, and the presence of Brg-1, Sp1 and p38 were detected by Western blot analysis.

The ChIP assay was performed using a chromatin immunoprecipitation assay kit (Upstate, Lake Placid, NY) according to the manufacturer's instructions. Cells were fixed with 1% formaldehyde for 10 min, washed with ice-cold PBS containing protease inhibitors and lysed with SDS lysis buffer. The lysate was sonicated to yield DNA fragments between 300 and 1000 base pairs, and centrifuged at 13,000 rpm for 10 min. The supernatant was diluted and pre-cleared with salmon sperm DNA/protein A agarose. Immunoprecipitation was performed overnight at 4°C using either non-specific IgG or the antibodies against Brg-1 or Sp1. The immunoprecipitates were washed and eluted, and the cross-links were reversed. The precipitated DNA fragments were purified. The 5'-promoter region spanning nucleotide positions -201 to -23 from the transcription start site of the SPARC gene were amplified by PCR using 5'-TGTCTGGGTAGCACACAGCCTAC-3' and 5'-GCAGGAAGCCTCTT GGAGCTCT-3' primers. Re-ChIP assays utilized a similar protocol, except that the primary immunocomplex obtained with the Sp1 antibody was eluted by 10 mM dithiothreitol with agitation at 37°C for 30 min. The eluate was diluted 50 times with buffer (20 mM Tris-HCl, pH 8.1, 150 mM NaCl, 2 mM EDTA, and 1% Triton X-100) and immunoprecipitated using the second antibodies. For ChIP-qPCR assay, the precipitated DNA fragments were purified and quantified with the Quant-iT™ dsDNA Assay Kit (Molecular Probes, Eugene, Oregon) and were then amplified by real-time qPCR using the same primers as for regular PCR.

Total RNA was extracted from cells using TRIzol Reagent (Invitrogen, Burlington, ON, Canada) according to the manufacturer's instructions. One μg of total RNA was reverse-transcribed with the QuantiTect reverse transcription kit (Qiagen, Mississauga, ON, Canada). An equal amount of cDNA or purified DNA fragment from ChIP was then amplified by real-time PCR using the Stratagene Mx-4000 and Brilliant SYBR Green QPCR Master Mix. Gene expression was normalized to a house-keeping gene (GAPDH) and the relative expression values between the samples were calculated based on the threshold cycle (C_T) value using the 2^-ΔΔCT method [83]. The following primers were used for cDNA amplification: Brg-1, 5'-TCTGAGGTGGACGCCCGACACATTA-3' (forward) and 5'-TAAGGACCTGC GTCAACTTGCAGTG-3' (reverse); and SPARC, 5'-AGGTGTGTGAGCTGCACG AGA-3' (forward) and 5'-GAAGTGGCAGGA AGAGTCGAA-3'(reverse).

Cells were washed once with ice-cold PBS and lysed in EBMK/0.1% NP-40 buffer (25 mM HEPES, pH 7.6, 5 mM MgCl₂, 1.5 mM KCl, 75 mM NaCl, 175 mM sucrose, 0.1% NP-40 and protease inhibitors) on ice for 10 min. The nuclear pellet was collected by centrifugation at 500 × g for 4 min. and washed three times with EBMK buffer (no NP-40). The nuclei were then lysed in 1 ml of RIPA buffer containing protease inhibitors, passed repeatedly through a 22-gauge needle and centrifuged at 10,000 × g for 30 min. The supernatants were pre-cleared with protein A/G agarose for 30 min. Immunoprecipitation was performed overnight at 4°C using the antibody against Brg-1 or Sp1. To precipitate the antigen-antibody complex, protein A/G agarose was added and incubated for 1 hr at 4°C. After washing with RIPA buffer, the precipitated proteins were eluted by boiling in 2× SDS sample buffer and analyzed by immunoblotting using antibodies to Brg-1 or Sp1.

Transient transfections of 4T1, 168FARN or 67NR cells were performed using the Lipofectamine™ 2000 and Plus reagent (Invitrogen, Burlington, ON) according to manufacturer's instructions. Briefly, cells were seeded into 12-well plates one day before transfection at a density of 5 × 10⁴ cells (4T1 or 168FARN) or 1 × 10⁵ cells (67NR) per well. Cells were transfected with 1.8 μg luciferase reporter constructs and 100 ng of Renilla luciferase control vector (pRL-CMV). 24 hrs after transfection, cells were treated with fenretinide or left untreated for 24 hrs before harvest. Luciferase reporter assays were performed using the Dual-Luciferase^® Reporter Assay System (Promega, Madison, WI) and the luminescence measurements were done with a Turner Designs model TD-20/20 luminometer. Firefly luciferase activity was normalized to Renilla luciferase activity. Each transfection was done in triplicate and repeated three times.

Following the treatment of cells with siRNA or fenretinide, cell media were collected and centrifuged to remove cell debris. Equal volumes (15 μl) of supernatant were used for Western blot analysis. Densitometric quantitation was performed, and to correct for any potential discrepancy in cell number resulting from different experimental conditions, the densitometric results were adjusted for total protein contents of cell lysates.

Cells were seeded in 24-well culture plates at 1.5 × 10⁵ cells/well. After 18 hrs, the cells were left untreated or pretreated with 2.5 μM or 5.0 μM fenretinide for 6 hrs before wound formation. The in vitro 'scratch' wounds were created by scraping the confluent cell monolayer with a 200 μl pipette tip and cultures were then washed twice with PBS to remove floating cells. Cells were then cultured in fresh medium or medium containing 2.5 μM or 5.0 μM fenretinide alone or combined with 8 μg/ml goat nonspecific IgG or SPARC antibody (R&D systems). The plates were photographed at 0 hr and 18 hrs after treatment. The wound width was measured using the program Image J http://​rsbweb.​nih.​gov/​ij/​ between two certain points on either side of the gap. For proper statistical evaluation, at least three measurements at different points were performed at each image. The wound width at the 18-hr time point was subtracted from that at the 0-hr time point. The distance was normalized to the wound width at 0 hr. The values were then expressed as relative motility, setting the cell motility of untreated cells as 100%. Three independent experiments were done in triplicates.

Cell invasion ability was assessed using a cell invasion assay kit (Chemicon International, Temecula, CA, USA) according to the manufacturer's instructions. The assay was performed in an invasion chamber, which consists of a 24-well tissue culture plate containing 12 cell culture inserts. The inserts contain an 8-μm pore size polycarbonate membrane coated on the upper side with a thin layer of ECMatrix™. Cells were pretreated with 2.5 μM or 5.0 μM fenretinide or left untreated for 6 hrs and were then collected and counted. 5 × 10⁴ untreated cells were resuspended in 0.3 ml of serum-free medium and pretreated cells were resuspended in 0.3 ml of serum-free medium containing 2.5 μM or 5.0 μM fenretinide only or combined with goat nonspecific IgG (5 μg/ml) or SPARC antibody (5 μg/ml). The lower chamber of the plate was filled with 0.5 ml medium containing 10% FBS with or without fenretinide. The cell suspension was then placed in the upper chamber and incubated at 37°C for 24 hrs. The noninvasive cells on the upper side of the membrane were removed. The invasive cells on the lower surface of the membrane were stained and then lysed. Absorbance was measured with a microplate reader at 560 nm. Each experiment was repeated three times, and the data represent the mean ± SEM of three determinations.

All data are presented as means ± SEM of three or four experiments. Analysis was performed using unpaired Student's t test. P < 0.05 was considered significant.