Acceleration of Glutathione Efflux and Inhibition of γ-Glutamyltranspeptidase Sensitize Metastatic B16 Melanoma Cells to Endothelium-induced Cytotoxicity*

Highly metastatic B16 melanoma (B16M)-F10 cells, as compared with the low metastatic B16M-F1 line, have higher GSH content and preferentially overexpress BCL-2. In addition to its anti-apoptotic properties, BCL-2 inhibits efflux of GSH from B16M-F10 cells and thereby may facilitate metastatic cell resistance against endothelium-induced oxidative/nitrosative stress. Thus, we investigated in B16M-F10 cells which molecular mechanisms channel GSH release and whether their modulation may influence metastatic activity. GSH efflux was abolished in multidrug resistance protein 1 knock-out (MRP-/-1) B16M-F10 transfected with the Bcl-2 gene or in MRP-/-1 B16M-F10 cells incubated with l-methionine, which indicates that GSH release from B16M-F10 cells is channeled through MRP1 and a BCL-2-dependent system (likely related to an l-methionine-sensitive GSH carrier previously detected in hepatocytes). The BCL-2-dependent system was identified as the cystic fibrosis transmembrane conductance regulator, since monoclonal antibodies against this ion channel or H-89 (a protein kinase A-selective inhibitor)-induced inhibition of cystic fibrosis transmembrane conductance regulator gene expression completely blocked the BCL-2-sensitive GSH release. By using a perifusion system that mimics in vivo conditions, we found that GSH depletion in metastatic cells can be achieved by using Bcl-2 antisense oligodeoxynucleotide- and verapamil (an MRP1 activator)-induced acceleration of GSH efflux, in combination with acivicin-induced inhibition of γ-glutamyltranspeptidase (which limits GSH synthesis by preventing cysteine generation from extracellular GSH). When applied under in vivo conditions, this strategy increased tumor cytotoxicity (up to ∼90%) during B16M-F10 cell adhesion to the hepatic sinusoidal endothelium.

B16 melanoma (B16M) 1 cells with high glutathione (GSH; ␥-glutamylcysteinylglycine) content show higher metastatic ac-tivity in the liver than those with low GSH content (1). The liver is a common site for metastasis development, and we demonstrated that GSH protects B16M cells against nitrosative and oxidative stress in the murine hepatic microvasculature (2,3). The concept that high GSH content status is an important factor for metastasis progression was strongly supported by the fact that metastatic B16M cell survival and growth can be enhanced by directly increasing their GSH content with GSH ester, which readily enters the cell and delivers free GSH (4,5). In consequence, the maintenance of high intracellular levels of GSH appears critical for metastatic cells to survive intravascularly and to progress extravascularly.
Multidrug and/or radiation resistance, which are characteristic features of malignant tumors, frequently associate with high GSH content in the cancer cells (6). Efflux of GSH and GSH S-conjugates from different mammalian cells is mediated by multidrug resistance proteins (MRP), among which MRP1 and MRP2 have been characterized at the functional level as ATP-dependent pumps with broad specificity for GSH and glucuronic or sulfate conjugates (7)(8)(9)(10). Multidrug resistance frequently associates with the overexpression of P-glycoprotein and/or MRP1 (11), both functioning as pumps that extrude drugs from tumor cells. GSH depletion induced by L-buthionine-(SR)-sulfoximine (BSO), a specific inhibitor of ␥-glutamylcysteine synthetase (the rate-limiting step in GSH biosynthesis) (12), resulted in a complete reversal of resistance to anticancer drugs of different cell lines overexpressing MRP1 but had no effect on P-glycoprotein-mediated multidrug resistance (13). Most interestingly, cancer cells can release GSH through MRP1 even in the absence of cytotoxic drugs (7).
In a recent study we demonstrated that B16M-F10 cells with a high metastatic potential overexpress BCL-2, show an increase in intracellular GSH content, show no change in the GSH synthesis rate, but show a decrease in GSH efflux (5). This study also provides evidence that BCL-2 can directly inhibit GSH export, thereby accounting for the increase in intracellular GSH. Moreover, it demonstrates that GSH depletion and BCL-2 antisense therapy can sensitize cells to TNF-␣-induced apoptotic death. In consequence, it is plausible that BCL-2, in addition of its anti-apoptotic properties (14), may also increase metastatic cell resistance against oxidative/nitrosative stress by preserving intracellular GSH. In fact, GSH efflux prior or during apoptosis appears an essential regulator of the apoptotic killing mechanism (15).
Because GSH efflux from B16M cells can be increased by using Bcl-2 antisense oligodeoxynucleotides (Bcl-2-AS) (5), and possibly by using different MRP1 regulators, the aim of the present report was to investigate whether the rate of efflux may become an important factor regulating intracellular GSH content and thereby metastatic cell survival.
GSH and GSSG-GSH was measured by the glutathione S-transferase reaction and GSSG by high pressure liquid chromatography (4). Total glutathione (GSH ϩ 2GSSG) was determined by a kinetic assay in which a catalytic amount of GSH or GSSG and glutathione reductase cause the continuous reduction of 5,5Ј-dithiobis-2-nitrobenzoic acid (Sigma) by NADPH (16). GSH monoisopropyl(glycyl) ester was prepared as described previously (17).
Bcl-2 Gene Transfer and BCL-2 Analysis-The Tet-Off gene expression system (Clontech) was used, as in Ref. 5, to insert the mouse Bcl-2 gene and for transfection into B16M cells following the manufacturer's instructions. BCL-2 protein was quantitated in the soluble cytosolic fraction by enzyme immunoassay (18) using a monoclonal antibodybased assay from Sigma (1 unit of BCL-2 was defined as the amount of BCL-2 protein in 1,000 nontransfected B16M cells).
Cellular BCL-2 protein levels were also analyzed, as recently reported (5), by flow cytometry using an EPICS PROFILE II (Coulter Electronics, Hialeah, FL), at 488 nm and 250 milliwatts. Cellular suspensions were diluted to ϳ250,000 cells/ml. Primary monoclonal anti-BCL-2 from mouse followed by biotin-conjugated goat anti-mouse IgG and phycoerythrin-labeled streptavidin (Sigma) were used. BCL-2 protein levels were expressed as arbitrary fluorescence units (⌬FL1). Cell viability in these experiments was determined with propidium iodide (final concentration 10 M, Molecular Probes, Eugene, OR).
Electroporation and Selection of B16M-F10 Multidrug Resistance Protein Knock-out Clones-Murine MRP genomic sequences were isolated from a Charon 35 phage library constructed with DNA from mouse strain C57BL/6J. The preparation of the gene-targeting construct, electroporation into tumor cells (Bio-Rad), and selection and screening of the clones, first by PCR analysis and then by DNA blot analysis, were carried out as described previously (19). The multidrug resistance associated protein-1 or -2 knock-out clones (MRPϪ/Ϫ1 and MRPϪ/Ϫ2) were obtained by exposing a single knock-out clone to high concentrations of G418 for 2 weeks.
Quantitative Determination of the Plasma Membrane Potential-Plasma membrane potential (PMP) was measured using a standard technique (21). B16M-F10 cells were cultured for 24 h, as described above, but seeded at 5 ϫ 10 4 cells/cm 2 in glass Petri dishes. Culture dishes were mounted on a tube-focusing microscope (Nikon, Tokyo, Japan). Intracellular measurements were performed, at 25°C, with glass micropipettes filled with 3 M KCl and with a 20-megohm DC resistance. Membrane potentials were measured with a WP Instruments M4-A electrometer amplifier (Sarasota, FL), and the output was displayed using a MacLab System (Castle Hill, Australia). Measurements were made only in cells (from at least four different preparations) that gave a stable membrane potential within 10 s of penetration, which indicates a good seal of the plasma membrane with the recording electrode.
Western Blots-Cultured cells were harvested, as reported previously (1), and then washed twice in ice-cold Krebs-Henseleit bicarbonate medium, pH 7.4. Whole cell extracts were made by freeze-thaw cycles in a buffer containing 150 mM NaCl, 1 mM EDTA, 10 mM Tris-HCl, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml aprotinin, and 1 g/ml pepstatin, pH 7.4. Fifty g of protein (as determined by the Bradford assay) were boiled with Laemmli buffer and resolved in 12.5% SDS-PAGE. Proteins were transferred to a nitrocellulose membrane and subjected to Western blotting with anti-cystic fibrosis transmembrane conductance regulator (anti-CFTR) monoclonal antibody (CFTR (CF3) antibody from Novus Biologicals, Littleton, CO). Immunizing peptide corresponds to amino acid residues 103-1117 found in the first extracellular loop of human CFTR. This sequence is conserved in mice. Blots were developed using horseradish peroxidase-conjugated secondary antibody and enhanced chemiluminescence (ECL system, Amersham Biosciences).
Reverse Transcription-PCR-A 392-bp region corresponding to nucleotides 1340 -1730 of the mouse CFTR gene (22) was amplified with the forward primer 5Ј-GGGAGGAGGGATTTGGGGAA-3Ј and the reverse primer 5Ј-GTGATGTCCTGCTGTAGTTG-3Ј. The CFTR cDNA was obtained using a random hexamer primer and a MultiScribe Reverse Transcriptase kit as described by the manufacturer (TaqMan RT Reagents, Applied Biosystems, Foster City, CA). A PCR master mix containing the specific primers and AmpliTaq Gold DNA polymerase (Applied Biosystems) were then added. Amplifications were performed in a GeneAmp 2400 thermal cycler (PerkinElmer Life Sciences).
Northern Blot Analysis-Total RNA was isolated using the TRIzol kit from Invitrogen and following the manufacturer's instructions. Ten g of total RNA were electrophoresed in 1% agarose gels containing formaldehyde. RNA was electrophoretically transferred to a nylon membrane (GeneScreen; PerkinElmer Life Sciences) and covalently bound to the membrane following UV cross-linking. Murine CFTR and control glyceraldehyde-3-phosphate dehydrogenase (G3PDH, Clontech) cDNA probes were radiolabeled with 32 P by random priming (23). The membranes were hybridized at 65°C with the 32 P-labeled cDNA fragments. After washing at room temperature, the filters were exposed to film, and autoradiography was performed using a PhosphorImager (Bio-Rad).
Perifusion of B16M-F10 Cells-Isolated B16M cells, suspended in DMEM, were incubated in a perifusion system similar to that described previously for rat hepatocytes (24). Briefly, a buffer gassed with 95% O 2 , 5% CO 2 was constantly pumped by an LKB multiperpex roller pump (type 2115; Amersham Biosciences) to a chamber containing a final volume of 10 ml and 3 ϫ 10 6 B16M-F10 cells per ml. The filter (Ultracel Amicon YM100 membrane, Millipore, Billerica, MA) was placed at the top of the chamber. The tumor cell suspension was perifused at 37°C and maintained at a homogenous state by using a magnetic stirrer placed at the bottom of the chamber. The perifusion buffer was Krebs-Henseleit bicarbonate medium, pH 7.4, containing plasma concentrations (aortic blood) of free L-amino acids found in tumor-bearing , sodium pyruvate (10 mg/liter), albumin-free fetal calf serum (1.0%, Invitrogen) and supplemented with 10 units/ml penicillin and 10 g/ml streptomycin. Perfusate flow (2 ml/min) was constant throughout the experiment. Effluent flow was monitored continuously for O 2 and pH with Philips electrodes. Tumor cell viability was always Ͼ97% along the experimental time. A syringe was introduced into the chamber through a rubber septum to take samples (0.5 ml) of the cell suspension without interrupting the flow.
Amino Acid Analysis-Proteins were precipitated by treating 0.1 ml of intracellular compartment or plasma with 0.4 ml of 3.75% (w/v) ice-cold sulfosalicylic acid in 0.3 M lithium citrate buffer, pH 2.8. After centrifugation, 0.25 ml of the supernatant were injected into an LKB 4151 amino acid analyzer (Amersham Biosciences) (25).
Verapamil and Acivicin Cellular Pharmacokinetics-Accumulation of N-methyl[ 3  Cells were centrifuged and washed twice with 1 ml of ice-cold PBS. After centrifugation, cell pellets were solubilized in 1% SDS, and cellassociated radioactivity was determined by liquid scintillation counting.
␥-Glutamylcysteine Synthetase Expression Analysis-Total RNA was isolated by the acid phenol guanidine method (28). cDNA was obtained using a random hexamer primer and a MultiScribe reverse transcriptase kit as described by the manufacturer (TaqMan Reverse Transcription Reagents, Applied Biosystems, Foster City, CA). A PCR master mix containing the specific primers (␥-GCS-heavy subunit (␥-GCS-HS): forward, 5Ј-ATC CTC CAG TTC CTG CAC ATC TAC, and reverse 5Ј-GAT CGA AGG ACA CCA ACA TGT ACTC; ␥-GCS-light subunit (␥-GCS-LS): forward, TGG AGT TGC ACA GCT GGA CTC T, and reverse 5Ј-CCA GTA AGG CTG TAA ATG CTC CAA; G3PDH: forward, 5Ј-CCT GGA GAA ACC TGC CAA GTA TG, and reverse 5Ј-GGT CCT CAG TGT AGC CCA AGA TG) and AmpliTaq Gold DNA polymerase (Applied Biosystems) were then added. Real time quantitation of ␥-GCS-HS and ␥-GCS-LS mRNA relative to G3PDH mRNA was performed with a SYBR Green I assay and evaluated using a iCycler detection system (Bio-Rad). Target cDNAs were amplified in separate tubes using the following procedure: 10 min at 95°C and then 40 cycles of amplification (denaturation at 95°C for 30 s and annealing and extension at 60°C for 1 min per cycle). The increase in fluorescence was measured in real time during the extension step. The threshold cycle (C T ) was determined, and then the relative gene expression was expressed as follows: Measurement of H 2 O 2 -The assay of H 2 O 2 production was based on the H 2 O 2 -dependent oxidation of the homovanillic acid (3-methoxy-4hydroxyphenylacetic acid) to a highly fluorescent dimer (2,2Ј-dihydroxydiphenyl-5,5Ј-diacetic acid) that is mediated by horseradish peroxidase (3).
Cystine Uptake-B16M-F10 cells were plated in 25-cm 2 culture dishes. At the required times, cells were rinsed three times with prewarmed transport medium (10 mM PBS, pH 7.4, with 0.01% CaCl 2 , 0.01% MgCl 2 and 0.1% glucose). Uptake measurement was initiated by addition of 1.0 ml of transport medium containing 1 Ci of L-[ 3 H]cystine (Amersham Biosciences) and nonradioactive L-cystine (0.5 mM). After incubation at 37°C, uptake was finished by rinsing several times with ice-cold PBS until less than 0.001% of the initial radioactivity was present in the supernatant. Cells were then dissolved with 0.5 ml of 0.5 N NaOH, and an aliquot was used for determining radioactivity and another for protein assay. To correct for trapping, transport at 4°C was studied in parallel (29).
Isolation and Culture of Hepatic Sinusoidal Endothelium-Syngenic male C57BL/6J mice (10 -12 weeks old) were from IFFA Credo (L'Arbreole, France) and received care according to the criteria outlined by the National Institutes of Health. Hepatic sinusoidal endothelium (HSE) was separated in a 17.5% (w/v) metrizamide gradient and identified as described previously (3). Cultures of HSE were established and maintained in pyrogen-free DMEM supplemented as described above for the B16M cells. Differential adhesion of endothelial cells to the collagen matrix and washing allows a complete elimination of other sinusoidal cell types (Kupffer, stellate, lymphocytes) from the culture flasks.
B16M-F10-Endothelial Cell Adhesion and Cytotoxicity Assays-B16M-F10 cells were loaded with 2Ј,7Ј-bis(2-carboxyethyl)-5,6-carboxyfluorescein acetoxymethyl ester (BCECF-AM, Molecular Probes) (10 6 cells were incubated in 1 ml of HEPES buffered DMEM, containing 50 g of BCECF-AM and 5 l of Me 2 SO, for 20 min at 37°C). After washing, BCECF-AM-containing cells were resuspended in HEPES buffered DMEM without phenol red at a concentration of 2.5 ϫ 10 6 cells/ml and added (0.2 ml/well) to endothelial cells (plated 24 h before) and also to plastic or collagen pre-coated control wells. The plates were then incubated at 37°C, and 20 min later the wells were washed three times with fresh medium and read for fluorescence using a Fluoroskan Ascent FL (Labsystems, Manchester, UK). The number of adhering tumor cells was quantified by arbitrary fluorescence units based on the percentage of the initial number of B16M-F10 cells added to the HSE culture (3). Damage to B16M-F10 cells during their in vitro adhesion to the HSE was measured, as described previously (2), using tumor cells loaded with calcein-AM (Molecular Probes). The integrity of B16M-F10 cells cultured alone was assessed by trypan blue exclusion and by measuring lactate dehydrogenase activity released to the extracellular medium (1). Other reagents used in experiments of tumor cytotoxicity were from Sigma.
In Vivo Microscopy-C57BL/6J mice were fed ad libitum on a stock laboratory diet (Letica, Barcelona, Spain) and kept on a 12-h light/ 12-h dark cycle with the room temperature maintained at 22°C. Procedures involving animals were in compliance with the national and international laws and policies (EEC Directive 86/609, OJ L 358. 1, December 12, 1987; and National Institutes of Health Guide for the Care and Use of Laboratory Animals Number 85-23). In vivo microscopy to follow metastatic cell dynamics within the liver was performed, as earlier described (2), using calcein-AM (Molecular Probes)labeled B16M-F10 cells. The total number of calcein-AM-labeled cells per hepatic lobule was recorded in 10 different lobules per liver at 15-min intervals for a 6-h period. The microscope was an Eclipse E600FN, providing transillumination or epi-illumination, equipped for video microscopy using a digital DXM 1200 camera (Nikon, Tokyo, Japan).
Apoptosis Detection-The livers used for in vivo microscopy studies were fixed in formalin embedded in paraffin and then processed in 10-m thick sections for routine hematoxylin staining. Immunohistochemical detection of apoptosis at single cell level, based on labeling of DNA strand breaks (TUNEL), was followed using the in situ cell death detection kit POD from Roche Diagnostics. Sections were examined under a Leica Microsystems (Nussloch, Germany) upright microscope with standard ϫ10, ϫ40, and ϫ63 plan apo-objectives.
Expression of Results and Statistical Significance-Data were analyzed by one-or two-way ANOVA or t tests where appropriate (SPSS 9.0 software for Windows; SPSS Inc., Chicago). The homogeneity of the variances was analyzed by the Levene test. The null hypothesis was accepted for all the values of the tests in which the F value was nonsignificant at p Ͼ 0.05. The data for which the F value was significant was examined by Tukey's test at p Ͻ 0.05.
We investigated further the mechanisms through which GSH is released from highly metastatic cells. As shown in Table I, GSH efflux from B16M-F10 cells was decreased by L-methionine suggesting that, in part, GSH is released through a system possibly similar to an L-methionine-sensitive sinusoidal GSH transporter detected in hepatocytes (31,32). However, GSH efflux was not inhibited by L-methionine in B16M-F10/ Tet-BCL-2 (Table I), suggesting that likely BCL-2 and L-methionine act on the same transport system (similar results were reported previously in HeLa cells (33)). Further experiments using MRPϪ/Ϫ1 or MRPϪ/Ϫ2 clones showed that GSH efflux was decreased, as compared with controls, in MRPϪ/Ϫ1 B16M-F10 and in B16M-F10/Tet-BCL-2 cells (Table I). GSH efflux in MRPϪ/Ϫ1 B16M-F10/Tet-BCL-2 cells was practically abolished, and identical results were obtained in MRPϪ/Ϫ1 B16M-F10 cells incubated in the presence of methionine (Table I).
Taken together these results indicate that a BCL-2-dependent system and MRP1 are the main mechanisms channeling GSH efflux from B16M-F10 cells.
Molecular Nature of the BCL-2-dependent GSH Channel in B16M-F10 Cells-BCL-2 overexpression has been associated with plasma membrane hyperpolarization (34), and it was suggested that changes in the rate of GSH efflux could be the consequence of changes in membrane polarity (e.g. Ref. 35). We explore first this possibility in our experimental model. The PMP measured in B16M-F10, B16M-F10/Tet-BCL-2, and Bcl-2-AS-treated B16M-F10 cells cultured for 24 h was 62.3 Ϯ 7.1, 75.4 Ϯ 6.2 (p Ͻ 0.05), and 60.3 Ϯ 5.5 mV, respectively (n ϭ 3-4 in each case). Therefore, BCL-2 overexpression (B16M-F10/ Tet-BCL-2) indeed increased significantly the PMP as compared with controls (B16M-F10), which in our model associated with a decrease in GSH efflux (Table I). However, it has been claimed that an increase in GSH efflux is the consequence of hyperpolarization (36). However, we observed that BCL-2 depletion increased GSH efflux from B16M-F10 cells (Table II) without affecting the PMP (see above), and addition to cultured B16M-F10 cells of ouabain (1 mM), a classical inhibitor of the plasma membrane Na ϩ /K ϩ -ATPase, decreased the PMP to ϳ63% of control values (see above) but did not affect significantly rates of GSH efflux measured in controls, in the presence of L-methionine (1 mM) or in MRPϪ/Ϫ1 cells (see e.g. Table  I). Apparently, efflux of GSH is not driven in B16M-F10 cells by a significant change in the PMP. Our results appear in agreement with those reported by Liu et al. (37) for Hep G2 cells, a human-derived hepatoma cell line, where a fall in the membrane potential produced by the replacement of Na ϩ with equivalent K ϩ did not affect GSH efflux significantly. Neither ouabain, vanadate (a Ca 2ϩ -ATPase inhibitor), nor BaCl 2 (a K ϩ channel blocker) significantly affected the GSH efflux; however, L-methionine (1 mM) decreased GSH efflux from Hep G2 cells (37).
The CFTR, a member of the ABC family of membrane transport proteins with structural similarities with the MRP1, forms a phosphorylation-and ATP-dependent channel permeable to Cl Ϫ and to other larger organic anions, including GSH (38). The CFTR is expressed in different cell types (39), and thus we also explored this possibility. GSH efflux increased significantly in Bcl-2-AS-treated B16M-F10 cells (containing ϳ3% of control BCL-2 levels, see the legend to Fig. 1), whereas GSH efflux decreased to ϳ50% of control values in MRPϪ/Ϫ1 B16M-F10 cells (Fig. 1). The difference represents the BCL-2-dependent GSH efflux (see also data in Table I). Monoclonal antibodies anti-CFTR practically abolished the BCL-2-dependent GSH efflux from B16M-F10 cells ( Fig. 1; the presence of CFTR in B16M-F10 cells was confirmed with a Western blot, not shown). In consequence, addition of anti-CFTR antibodies to Bcl-2-AS-treated MRPϪ/Ϫ1 B16M-F10 cells practically abolished GSH efflux (Fig. 1).
CFTR is activated by the binding of ATP to its cytoplasmic nucleotide-binding domain and by phosphorylation of key serine residues in the regulatory domain. Phosphorylation is mediated principally by cAMP-dependent protein kinase A (PKA) and by protein kinase C (although to a less degree than the activation by PKA) (40). Basal expression of the CFTR gene is dependent on PKA activity (41). Treatment of human colon carcinoma T84 cells with the PKA-selective inhibitor N-[2-(pbromocinnamylamine)ethyl]-5-isoquinolinesulfonamide (H-89) (Seikagaku America, Rockville, MD) caused a complete suppression of CFTR gene expression without affecting other constitutively active genes (41). Thus, we also used this approach in our experimental model. As shown in Fig. 2  These results indicate that CFTR is directly involved in channeling GSH from the cytoplasm of B16M-F10 cells to the extracellular space and that this mechanism corresponds to the BCL-2-sensitive channel. Nevertheless we cannot rule out the possibility that other mechanism(s) could also be working in other cell types or that different CFTR gene mutations could be found when comparing different cancer cells. These interesting questions, although far beyond the aim of the present report, deserve further investigation.
Antisense Bcl-2 Oligodeoxynucleotides and Verapamil Accelerate GSH Release from B16M-F10 Cells-Parallel to the BCL-2-sensitive CFTR, two mechanisms of transport of GSH by MRP1 have been suggested as follows: passive permeability and a VRP-dependent active transport (42). VRP, an inhibitor of P-glycoprotein-mediated drug efflux, is not transported by MRP1 (43) but may also inhibit MRP1-mediated drug extrusion (44). We tested VRP in combination with Bcl-2-AS treatment to potentiate GSH efflux from B16M-F10 cells. A perifusion chamber, containing a suspension of B16M-F10 cells, was used as an experimental setup that mimics in vivo conditions by providing a constant supply of glucose, amino acids, and TABLE I BCL-2-and MRP-dependent GSH efflux from B16M-F10 cells B16M-F10 cells were cultured for 24 h. Some flasks were used to measure rates of GSH efflux (to prevent degradation of the GSH accumulated in the extracellular space the ␥-glutamyltranspeptidase was blocked by adding 10 M ACV to the culture medium 3 h before measuring efflux; see Ref. 4). Other flasks were used for GSH, GSSG, and BCL-2 determination. GSH efflux corresponded practically to GSH since GSSG was, in all conditions, ϳ1-3% of the total glutathione found in the extracellular space (not shown). BCL-2 levels in control and Tet-BCL-2-treated B16M-F10 cells were 27 Ϯ 5 and 118 Ϯ 20 units/mg protein, respectively (n ϭ 5 in each case, p Ͻ 0.01). Addition of L-methionine (1 mM) or the use of MRPϪ/Ϫ1 or MRPϪ/Ϫ2 clones did not alter significantly the BCL-2 levels (not shown). A one-way ANOVA was performed for comparison among groups. Different superscript letters within a column indicate differences, p Ͻ 0.01. Results are means Ϯ S.D. of 5-6 independent experiments. min (ϳ70 pmol/10 6 cells) after addition to the perfusate flow and then decreased reaching a lower steady-state concentration at 60 min (ϳ40 pmol/10 6 cells) (Fig. 3). These values of VRP accumulation, which are in agreement with those reported previously in HeLa cells (43), were not changed significantly when control and BCL-2-AS-treated cells were compared (Fig. 3). As shown in Table II, Bcl-2-AS and VRP independently increased rates of GSH efflux from perifused B16M-F10 cells. In fact, VRP had no effect on GSH efflux in MRPϪ/Ϫ1 B16M-F10 cells (not shown), and when Bcl-2-AS and VRP were both present their effects were additive and constant along a 12-h perifusion time (GSH efflux was increased by ϳ100%, Table II). Intracellular GSH contents were significantly decreased (ϳ30%), as compared with controls, in Bcl-2-AS-and VRPtreated B16M-F10 cells after 6 h of perifusion (Table III). However, at 12 h of perifusion time, GSH levels were ϳ70% higher in Bcl-2-AS-and VRP-treated B16M-F10 cells than in controls (Table II). Thus, it appeared plausible that loss of GSH accelerates their rate of intracellular synthesis.
Antisense Bcl-2 Oligodeoxynucleotides-and/or Verapamil-induced Acceleration of GSH Efflux Associates with ␥-Glutamylcysteine Synthetase Overexpression in B16M-F10 Cells-As shown in Table III, either both or Bcl-2-AS or VRP treatment TABLE II GSH efflux and content in perifused B16M-F10 cells treated with Bcl-2-AS and VRP B16M-F10 cells were cultured for 3 days in the presence or in the absence of 50 M Bcl-2-AS or a 2-base mismatch oligodeoxynucleotide (see under "Experimental Procedures"). Then the cells were harvested, cultured again for 24 h in the absence of oligodeoxynucleotides, and used for perifusion experiments. BCL-2 levels in control (none under the heading Additions), Bcl-2-AS-, or Bcl-2-mismatch-AS-treated cells were 36 Ϯ 8, 33 Ϯ 6, and 2 Ϯ 1 units/mg protein, respectively (n ϭ 6 -7 in each case). Results obtained by substituting Bcl-2-AS by its 2-base mismatch counterpart were not significantly different from those obtained under no additions or in the presence of VRP alone (not shown). To measure rates of GSH efflux, 2-ml aliquots of effluent buffer were collected each min for 30 min, starting at the indicated perifusion times, and placed on ice. One ml was used to measure total glutathione, whereas the other milliliter was mixed with 1 ml of ice-cold perchloric acid (12%) containing 40 mM N-ethylmaleimide and 2 mM bathophenanthroline disulfonic acid to prevent GSH oxidation and to measure GSSG accurately (4) (see under "Experimental Procedures"). GSH ϭ 2 ϫ (total glutathione Ϫ GSSG). Before determinations were performed, samples were concentrated ϳ40 times by using speed vacuum centrifugation at 4°C. GSSG present in the effluent was, in all conditions, ϳ1-2% of the total glutathione. VRP concentration in the perfusate flow was 1 M (see Fig. 3 and the text for details regarding its cellular pharmacokinetics). VRP was added to the perfusate flow as indicated in the legend to Fig. 3. No significant differences were found when results obtained with control and cells preincubated with the 2-base mismatch oligodeoxynucleotide were compared (not shown). Data are means Ϯ S.D. for 6 -7 independent experiments. A two-way ANOVA was used to make comparisons among different treatments and time points. Perifusion time is given in hours. Different superscript letters indicate differences, p Ͻ 0.05.  significantly increased the rate of GSH synthesis in B16M-F10 cells. Moreover, their effects appeared additive and associated with an increase in ␥-GCS activity (Table III). Bcl-2-AS and/or VRP, as compared with controls, did not change significantly the GSH synthetase activity (see Table III legend). Moreover, as shown in Table IV, the increase in ␥-GCS activity was accompanied by a previous increase in both ␥-GCS-HS and ␥-GCS-LS expression (maximum values were found at 3 h). Therefore, increased GSH efflux associates with ␥-GCS overexpression.

Additions
Changes in ␥-GCS activity can result from transcriptional and post-transcriptional regulation affecting the HS and/or the LS (for review see Ref. 46). Intracellular GSH depletion, e.g. that induced by GSH-conjugating agents such as diethyl maleate, increased transcription of both subunits (46) as it occurs in our experimental conditions. Moreover, oxidative stress, which may arise as a consequence of GSH depletion and increased intracellular levels of H 2 O 2 and OH ⅐ , can also induce increased transcription of both subunits (46). In fact, Bcl-2-ASand VRP-induced GSH depletion (Table II)  Inhibition of ␥-Glutamyltranspeptidase Activity Prevents GSH Content Increase in B16M-F10 Cells Treated with Antisense Bcl-2 Oligodeoxynucleotide and Verapamil-In rapidly growing tumors cyst(e)ine, whose concentration in blood is low, may become limiting for GSH synthesis and cell growth (4,47). Thus, in order to increase the rate of GSH synthesis, malignant cells may require alternative pathways to ensure cyst(e)ine availability.
Keeping the extracellular supply of amino acids constant and physiological during the perifusion (see under "Experimental Procedures"), the intracellular availability of amino acid precursors for GSH synthesis was investigated. The concentrations of free L-glutamate and glycine within the B16M-F10 cells were constant through the perifusion time (e.g. 3.3 Ϯ 0.4 and 2.5 Ϯ 0.3 mM, respectively, in controls; n ϭ 6; enough to ensure maximum rates of GSH synthesis (12)) and were not changed by addition of Bcl-2-AS and/or VRP (not shown). However, free intracellular L-glutamine and L-cyst(e)ine were undetectable, which is not surprising because L-glutamine is a major fuel used by cancer cells (48), and L-cyst(e)ine is rapidly used for protein and GSH synthesis (25). L-Cystine is predominant outside the cell because L-cysteine rapidly autoxidizes to L-cystine in the extracellular fluids, but once it enters the cell through the Xc Ϫ system L-cystine is reduced to L-cysteine (Ref. 46 and references therein). Thus, we measured L-cystine uptake by B16M-F10 cells and found that Bcl-2-AS and/or VRP did not significantly affect this rate (e.g. 0.41 Ϯ 0.08 nmol/mg protein in controls, n ϭ 5). Nevertheless, we showed that tumor ␥-GT activity and an intertissue flow of GSH increase GSH content in B16M-F10 cells and work as a tumor growth-promoting mechanism (4). ␥-GT cleaves extracellular GSH-releasing ␥-glutamyl-amino acids and cysteinylglycine, which is further cleaved by membrane-bound dipeptidases into L-cysteine and glycine (12). Free ␥-glutamyl amino acids, L-cysteine and glycine, entering the cell serve as GSH precursors (12). Hence, ␥-GT expression may provide tumor cells with a growth advantage at physiological concentrations of L-cyst(e)ine (4,47). Therefore, if the increase in GSH content (Table II) Table II. For GSH synthesis, cells were taken from the perifusion chamber, washed twice, and resuspended in ice-cold Krebs-Henseleit bicarbonate medium, pH 7.4, and incubated (5 mg of dry weight/ml) (1), in 10-ml-Erlenmeyer flasks (final volume 2 ml) for 60 min, at 37°C, in the presence of the amino acid precursors for GSH synthesis (5 mM L-Gln, 2 mM Gly, 1 mM L-Ser, 1 mM N-acetylcysteine). Glucose (5 mM) and bovine serum albumin (2%) were always present. GSH synthesis was calculated from total GSH content at 0, 20, 40, and 60 min of incubation. No significant differences were found when data displayed in the table for 6 h were compared with those obtained at 1 h of perifusion time (not shown). GSH-S activity in controls (see None under the heading Additions) was 15 Ϯ 3 milliunits/10 6 cells (this value did not change significantly along the perifusion time even in the presence of Bcl-2-AS-and/or VRP, n ϭ 4 -5 in each case, not shown). Results obtained by substituting Bcl-2-AS by its 2-base mismatch counterpart were not significantly different from those obtained under no additions or in the presence of VRP alone (not shown). Data are means Ϯ S.D. for 8 -10 independent experiments. Perifusion time is given in hours. A two-way ANOVA was used to make comparisons among groups. Different superscript letters indicate differences, p Ͻ 0.01.  IV Acceleration of GSH release associates with ␥-GCS overexpression in perifused B16M-F10 cells B16M-F10 cells were cultured and perifused as described in the legend to Table II. ␥-GCS-HS and ␥-GCS-LS expression was determined after 3 and 6 h of perifusion. Results obtained by substituting Bcl-2-AS by its 2-base mismatch counterpart were not significantly different from those obtained under no additions or in the presence of VRP alone (not shown). The figures, expressing fold induction, show mean values Ϯ S.D. from 5 to 6 different experiments. Perifusion time is given in hours. A two-way ANOVA was used to make comparisons among groups. Different superscript letters within a column indicate differences, p Ͻ 0.05.

Additions
the effect of Bcl-2-AS and VRP on GSH efflux depends on L-cyst(e)ine availability, and if this is provided in part by the ␥-GT, then inhibition of this activity could limit GSH synthesis. We tested this possibility by adding to the perifusion system ACV, an irreversible ␥-GT inhibitor (49). ACV was present in the perfusate flow for only 30 min (Fig. 4). ACV accumulation within the B16M-F10 cells peaked 10 -20 min (ϳ20 pmol/10 6 cells) after addition, and then its concentration decreased to a steady-state low level (ϳ4 pmol/10 6 cells) (Fig. 4; these values were not changed when control and Bcl-2-AS-or VRP-treated cells were compared, data not shown). ACV decreased ␥-GT activity to nondetectable levels and decreased GSH synthesis but did not affect the rate of GSH efflux (Table V) or the rate of L-cystine uptake (not shown, see above for control values). However, the rate of GSH synthesis in B16M-F10 cells treated with Bcl-2-AS, VRP, and ACV was found similar to controls (as under "no additions" in Table V) when the concentration of L-cysteine in the perifusion buffer was increased 2-fold (up to 16 M). Hence, intracellular L-cysteine availability indeed appears modulated by its ␥-GT-dependent generation from extracellular GSH. Therefore, we conclude that Bcl-2-AS-and VRP-induced acceleration of GSH efflux combined with inhibition of ␥-GT promote GSH depletion in B16M-F10 cells.
Sensitizing of B16M-F10 Cells to Vascular Endothelium-induced Tumor Cytotoxicity-Previously, we demonstrated that B16M-F10 cells with low GSH content (cultured in the presence of BSO), as compared with untreated controls, were more sensitive to HSE-induced metastatic cell cytotoxicity (2). However BSO, when administered under in vivo conditions, decreases GSH in tumor and non-tumor tissues thus placing the normal tissues at a disadvantage (6). Thus, we investigated whether a combination of Bcl-2-AS, VRP, and ACV could be as effective as BSO but without its nonselective effect on tissue GSH. First, we assayed the effect of BSO versus Bcl-2-AS ϩ VRP ϩ ACV on the in vitro interaction between B16M-F10 and HSE cells. As shown in Table VI, both treatments decrease tumor GSH to 35-40% of control values. However, HSE-induced tumor cell death was much higher in the presence of Bcl-2-AS ϩ VRP ϩ ACV (ϳ96%) than in the presence of BSO (ϳ66%) ( Table VI). This is in agreement with our previous report (5) showing that depletion of GSH and BCL-2 levels practically avoids B16M-F10 survival during interaction with the vascular endothelium. Also in agreement with this previous report (5), GSH depletion did not alter the % of tumor cell adhesion to the HSE. To prove that GSH is directly regulating metastatic cell survival, we tested the effect of GSH ester (which enters the cell and delivers free GSH (4)). As shown in Table VI, GSH ester prevented HSE-induced B16M-F10 cytotoxicity in the presence of BSO, but only partially in the presence of Bcl-2-AS, VRP, and ACV (where BCL-2 depletion remains, see Ref. 5). Second, we assayed the effect of Bcl-2-AS, VRP, and ACV on the in vivo B16M-F10 cell arrest and viability within the hepatic microvasculature. Calcein-labeled B16M-F10 cells, which present a green fluorescent cytoplasm, arrested in the liver sinusoids after intraportal inoculation (Table VII). The number of arrested B16M-F10 and Bcl-2-AStreated B16M-F10 cells was similar (Table VII) and constant along the time (no significant differences were found within the 30 -360-min postinjection) (not shown). As reported previously (2), the number of arrested cells was not altered by depleting tumor cell GSH or by treating mice with lipopolysaccharide (LPS) (to pre-activate the endothelium) before tumor cell inoculation (Table VII). In physiological saline-or VRP ϩ ACVtreated mice (Table VII), almost all arrested non-Bcl-2-ASpretreated B16M-F10 cells appeared as round bright fluorescent cells of well delineated profile (nondamaged "intact" cells because no fluorescence diffusion from their cytoplasm to their neighboring tissue was observed). Bcl-2-AS pretreatment decreased the viability of arrested metastatic cells (Table VII), which appeared as irregularly shaped fluorescent cells with a spread of diffuse fluorescence staining the contiguous hepatic tissue, and were considered as damaged with cytoplasmic leakage. The decrease in B16M-F10 viability was

TABLE V Effect of ACV-induced inhibition of ␥-GT on GSH content in perifused B16M-F10 cells treated with Bcl-2-AS and VRP
B16M-F10 cells were cultured and perifused as indicated in the legend to Table II. VRP addition and concentration in the perfusate flow are shown in Table II and Fig. 3. ACV concentration in the perfusate flow was 1 M (see Fig. 4 and the text for details regarding its cellular pharmacokinetics). ACV was added to the perfusate flow as indicated in the legend to Fig. 4. GSH content and ␥-GT activity were measured at 12 h of perifusion time. GSH synthesis and efflux were measured, starting at 12 h of perifusion, as indicated in the legend to Table III but using a complete mixture of amino acids as precursors. This mixture contained plasma concentrations (aortic blood from non-tumor-bearing C57BL/6J mice) ϫ 10 of GSH and free L-amino acids (see under "Perifusion of B16M-F10 Cells" under "Experimental Procedures"). Results obtained by substituting Bcl-2-AS by its 2-base mismatch counterpart were not significantly different from those obtained in the presence of VRP or VRP ϩ ACV (not shown). Data are means Ϯ S.D. for 5-6 different experiments. A one-way ANOVA was performed for comparison among groups. Different superscript letters within a column indicate differences, p Ͻ 0.01.
These in vivo results are in agreement with the in vitro observations displayed in Table VI and indicate that GSH and BCL-2 depletion may be a feasible approach to sensitize metastatic melanoma cells to endothelium-induced cytotoxicity in vivo. Furthermore, we assayed GSH levels in different tissues (brain, lung, heart, liver, kidney, pancreas, skeletal muscle, bone marrow, testis, and erythrocytes, n ϭ 6 in each case) of untreated non-tumor-bearing mice and in LPS-, VRP-and ACV-treated mice inoculated with Bcl-2-AS-pretreated B16M-F10 cells (as in the legend to Table VII), and we found no significant differences as compared with data reported previously for non-tumor-bearing mice and B16M-F10-bearing mice (4) (data not shown), thus indicating that the proposed treatment does not affect GSH levels in normal tissues. DISCUSSION Analysis of a Bcl-2 family of genes revealed that B16M-F10 cells (high metastatic potential), as compared with B16M-F1 cells (low metastatic potential), preferentially overexpressed Bcl-2 (5). BCL-2 overexpression, without changing the rate of GSH synthesis, induces a decrease in GSH efflux and, consequently, an increase of GSH content within B16M-F10 cells (5). Most B16M cells with a high GSH content survive the NO-and H 2 O 2 -mediated tumoricidal activity of endothelial cells (50). However, survival of B16M-F10 during interaction with the vascular endothelium can be challenged by inhibiting their BCL-2 and GSH synthesis in vitro (5). Bcl-2 antisense therapy using G3139, for example, an 18-base phosphorothioate oligonucleotide complementary to the first six codons of the Bcl-2 mRNA, selectively and specifically inhibits BCL-2 expression and promotes apoptosis in different human and murine cancer cell lines (51). Systemic administration of G3139 to Shionogi tumor-bearing mice led to a rapid decrease of tumor size (higher when chemotherapy was simultaneously administered), whereas the oligonucleotide did not affect BCL-2 expression in normal organs (20,52). G3139-induced tumor regression without dose-limiting toxicity was also observed in other tumors, melanoma, lymphoma, or gastric cancers for example (51). Furthermore, synergism of the G3139 and anticancer drugs was also shown in different tumors (53)(54)(55). However, on the other hand, an effective strategy to deplete GSH in metastatic cells has remained elusive.   Table III. The cells were then harvested and cultured again in the absence of oligodeoxynucleotides. Twenty four-hour cultured HSE cells (ϳ2.5 ϫ 10 5 cells/well) were co-cultured with B16M-F10 cells (ϳ5.0 ϫ 10 5 cells/well; precultured for 24 h). Twenty minutes after B16M-F10 addition to the HSE, the plates were washed as described under "Experimental Procedures." The ratio of tumor cells adhering to the HSE was ϳ1:1. TNF-␣ (100 units/ml) and IFN-␥ (50 units/ml), used as a potent activators of NO and H 2 O 2 generation by the HSE (3), were added to the co-cultures when all tumor cells present were attached to the HSE. In endothelium-induced B16M-F10 cytotoxicity assays, tumor cytotoxicity (expressed as the % of tumor cells that lost viability within the 4 -6-h period of incubation, see "Experimental Procedures") was determined after 6 h of incubation. During the 6-h period of incubation, the percentage of HSE cell viability was 98 -99% in all cases. When adding TNF-␣ (100 units/ml) and IFN-␥ (50 units/ml) to cultured B16M-F10 cells alone, no cytostatic or cytotoxic effects were observed within the next 6 h. Where indicated B16M-F10 cells were incubated for 24 h with GSH ester (1 mM Table II. GSH contents in control and Bcl-2-AS-treated B16M-F10 cells before inoculation were similar to those reported in Table II. Results obtained in the presence of the 2-base Bcl-2 mismatch oligodeoxynucleotide were not significantly different from those obtained in the absence of Bcl-2-AS (not shown). Data are means Ϯ S.D. for 4 -5 different experiments. The significant test (Student's unpaired t test) refers, for all groups, to the comparison in the absence or in the presence of Bcl-2-AS (**, p Ͻ 0.01), and also to the difference between LPS-treated mice and untreated mice ( ϩ , p Ͻ 0.05; ϩϩ , p Ͻ 0.01).  (50). However, endothelial NO-mediated partial inactivation of ␥-GCS activity is followed by overexpression of ␥-GCS-heavy and -light subunits, which leads to a rapid recovery of GSH levels within invasive cells (50). Therefore, as suggested by previous achievements in nonmetastatic models (6), a methodology capable of maintaining low GSH levels in metastatic cells could represent a critical advance in cancer therapy.
Here we present evidence showing that GSH is released from highly metastatic B16M-F10 cells through MRP1 and CFTR (Table I and Fig. 1). By using a perifusion system that mimics in vivo conditions, we show that GSH efflux from B16M-F10 cells can be accelerated be using Bcl-2-AS (which prevents BCL-2-induced inhibition of GSH release through CFTR) and VRP (which activates GSH release through MRP1 (42)) ( Tables  I and II). However, Bcl-2-AS and/or VRP treatment associated with overexpression of ␥-GCS (Table IV) increased rates of GSH synthesis (Table III) and higher GSH content (Table II) within the B16M-F10 cells.
Recently, we showed that tumor ␥-GT activity, by providing an extra supply of L-cysteine from extracellular GSH, supports GSH synthesis in B16M-F10 cells and promotes their metastatic growth (4). Hence, we tested whether inhibition of this activity could prevent the increase of GSH content within BCL-2-AS-and VRP-treated B16M-F10 cells. Indeed, as shown in Table III, ACV limited GSH synthesis and allowed GSH efflux to deplete tumor cell GSH. This strategy, which combines induction of BCL-2 and GSH depletion, sensitized perifused B16M-F10 cells to the cytotoxic effects of the vascular endothelium (Tables VI and VII).
Can possible clinical applications be derived from our study? The proto-oncogene Bcl-2 and its anti-apoptotic homologs are mitochondrial membrane permeabilization inhibitors (56) and participate in the development of chemoresistance (57), whereas expression of pro-death genes, e.g. Bax or Bak, is often reduced in cancer cells (58). In agreement with this idea, Takaoka et al. (59) observed that Bcl-2 overexpression in B16M cells enhanced pulmonary metastasis. In fact, a major form of multidrug resistance in human tumors is caused by overexpression of the MRP1 gene (7). In vivo Bcl-2-AS therapy, as explained above, is feasible. In addition, VRP (at doses that promote a similar plasma concentration than that used in our experiments) has already been used in cancer patients with myeloma or acute lymphocytic leukemia, for example, where it increased accumulation of daunorubicin or vincristine within the tumor cells (60). ACV, the L-glutamine analog anti-metabolite, has followed phase I and II clinical trials in different tumors (61). Although its use is limited by severe central nervous system toxicity, a maximum tolerated dose of 50 mg of acivicin/m 2 /day has been proposed in combination with the amino acid solution aminosyn (which decreases drug uptake in the central nervous system) (61). In our studies, ACV concentration in the perfusate flow was 1 M (present only during 30 min) (Fig. 4 and Table III). By taking into account the circulating blood volume and the in vivo pharmacokinetics in humans (61), this means that ACV doses required to block tumor ␥-GT activity will remain within nontoxic levels. This is important because an increased expression of ␥-GT has been found in melanoma as well as in other cancers (including human tumors of the liver, lung, breast, and ovary) (62).
In vitro HSE-induced B16M-F10 cytotoxicity was very high (ϳ90%) when Bcl-2-AS-treated metastatic cells were treated with VRP and ACV (Table VI). These results appear in agreement with a recent report (63) showing that GSH depletion enforces the mitochondrial permeability transition and causes cell death in HL60 cells that overexpress BCL-2. Furthermore, when BSO-and Bcl-2-AS-pretreated B16M-F10 cells were inoculated intravascularly into mice, the number of intact arrested cells on the HSE decreased by ϳ98%, and the very small number of metastatic cell survivors (probably bearing molecular damages) did not form detectable colonies (5). Nevertheless, as indicated by the data displayed in Table VII, after applying in vivo the methodology proposed in this report, some metastatic cells may still survive. Invasive cells may prolong survival under dormancy conditions (64) or may benefit from oxidative stresspromoting metastatic mechanisms, e.g. increasing cell adhesion molecule expression (65), activating early growth response-1 transcription factor gene (66), activating metalloproteinases (67), or increasing resistance to oxidative stress (68). In fact, in mammalian cells at least 40 different gene products are involved in adaptive responses to oxidative stress (68). However, invasive B16M-F10 cells that survive after in vitro interaction with the HSE show a transient impairment of the mitochondrial system for GSH uptake (50). The mitochondrial GSH is one of the endogenous effectors that regulates the mitochondrial permeability transition pore complex (69), and we observed that B16M-F10 cells with low mitochondrial GSH levels were highly susceptible to TNF-␣-induced oxidative stress and death (50). Most interestingly, this effect can be potentiated by Bcl-2-AS therapy (5). Moreover, therapy could be improved by combining nontoxic TNF-␣ doses with IFN-␥ (70), or with a L-glutamine-enriched diet to facilitate an L-glutamate-induced inhibition of GSH transport into tumor mitochondria (71). Furthermore, our methodology can be combined with cytotoxic drugs and/or ionizing radiation. Thus the mechanisms described in this report may have useful applications to improve therapy against metastatic melanoma and, possibly, against other malignant tumor types.