Alterations of Ionic Membrane Permeabilities in Multidrug-Resistant Neuroblastoma × Glioma Hybrid Cells

The resistance of tumour cell to cytotoxic drugs is one of the major obstacles to successful chemotherapy. Multidrug resistance (MDR) is the phenomenon observed in tumour cells in vivo and cultured cells in vitro involving cellular resistance to a wide range of chemically unrelated drugs. MDR is commonly associated with overexpression of a membrane protein, P-glycoprotein (P-gp) (for reviews, see Gottesman and Pastan, 1993; Higgins and Gottesman, 1992; Nielsen and Skovsgaard, 1992). This protein is hypothetized to be an active transporter that uses the energy from ATP to pump chemotherapeutic drugs out of cells, reducing their intracellular concentrations and, hence, toxicity. However, in most drug-selected MDR tumour cell lines, the activity of P-gp alone cannot account for the diversity of MDR phenotypes observed. Other cellular mechanisms seem to develop and contribute to resistance. In addition to P-gp overexpression, MDR is accompanied by multiple biochemical and morphological changes, including alterations to membrane ion transport which are thought to play a role in drug resistance. These alterations can be considered to be a direct consequence of P-gp expression when they appear in transfected cells that have not previously been exposed to chemotherapeutic drugs (Hoffman et al. 1996). In contrast, when they appear in drug-selected cells (which was the case in most MDR cells studied until now), they may be due to factors unrelated to P-gp expression, but induced by the drug selection. Whatever the origin of these alterations, it is interesting to know whether they play a role in multidrug resistance. It has been noted that Na⁺, K⁺ and Cl⁻ channel activities differ between MDR and parental cell lines (Gollapudi et al. 1992; Jirsch et al. 1993; Kunzelmann et al. 1994; Yamashita et al. 1987). Since 1992, several groups have focused on the relationship between the volume-sensitive Cl⁻ channel and MDR but conflicting results have been obtained. Although several independent studies confirm a relationship between P-gp expression and the magnitude of a cell’s volume-sensitive current (Gill et al. 1992; Jirsch et al. 1993; Valverde et al. 1992), other studies have found no correlation (De Greef et al. 1995; Dong et al. 1994; Ehring et al. 1994; Kunzelmann et al. 1994; Morin et al. 1995; Rasola et al. 1994; Wang et al. 1994).

As a possible consequence of alterations to ion transport, a decrease in membrane potential has been observed in doxorubicin-resistant Friend leukaemia cells (Hasmann et al. 1989) and in Chinese hamster ovary fibroblasts transfected with hu MDR 1 cDNA and resistant to doxorubicin, vinblastin and colchicine (Hoffman et al. 1996). Since most anticancer drugs are positively charged at physiological pH, their accumulation in cells should be influenced by membrane potential and should decrease as the membrane potential is made less negative. Hoffman et al. (1996) concluded that protein-mediated MDR could be the consequence of a less negative resting membrane potential. Another important cellular parameter that is modified by MDR acquisition is intracellular pH (pHi). Most MDR cells have a higher intracellular pH than their parental cells (Boscoboinik et al. 1990; Marquardt and Center, 1991; Roepe, 1995; Roepe et al. 1996). These latter observations led to recent investigations into the causal relationship between pHi and MDR. Agents that reverse MDR, such as verapamil, can also affect the elevated cytoplasmic pH (Keizer and Joenje, 1989), and an alkaline shift in pHi in sensitive cells is sufficient to prevent the accumulation of chemotherapeutic drugs and to decrease the drug sensitivity (Hoffman et al. 1996; Simon et al. 1994). Moreover, Marquardt and Center (1991) observed that bafilomycin A1, a specific inhibitor of the vacuolar-type proton pump (V-ATPase), induces a major increase in drug accumulation, suggesting a possible involvement of this ATPase in the pathway of drug efflux from MDR cell lines that is independent of P-gp expression. Since we recently advanced arguments in favour of the contribution of a plasma membrane proton pump in determining the resting potential of NG108-15 cells (Gérard et al. 1994), we wanted to know whether this proton pump could play a role in neuroblastoma cell drug resistance. We therefore selected NG108-15 neuroblastoma cells resistant to doxorubicin (NG/DOXR) that exhibited an MDR phenotype. A comparative analysis of electrophysiological properties was carried out on doxorubicin-sensitive and doxorubicin-resistant neuroblastoma cells with particular attention being paid to the V-ATPase and volume-sensitive Cl⁻ channel activities.

NG108-15 hybridoma cells are a clone of neuroblastoma × glioma hybrid cells formed by Sendai-virus-induced fusion of a subclone N 18TG-2 of mouse C1300 neuroblastoma cells with subclone C6Bu-1 of rat C6 glioma cells (Hamprecht, 1977). Doxorubicin-resistant (NG/DOXR) cells were selected by continuously exposing NG108-15 cells to gradually increasing concentrations (from 2 to 100 nmol l⁻¹) of doxorubicin. After 9 months in culture, NG/DOXR cells were able to grow uninhibited in 100 nmol l⁻¹ doxorubicin and were thereafter cultured in that concentration of drug. The cells were grown at 37°C in a humidified atmosphere (95% air/5% CO₂) in Dulbecco’s modified Eagle’s medium containing 5% foetal calf serum, 100 μmol l⁻¹ hypoxanthine, 0.4 μmol l⁻¹ aminopterine, 16 μmol l⁻¹ thymidine, 2 mmol l⁻¹ glutamine, 3 μmol l⁻¹ glycine, 100 μg ml⁻¹ streptomycin and 50 μg ml⁻¹ penicillin. NG/DOXR cells were continuously grown in the culture medium supplemented with 100 nmol l⁻¹ doxorubicin, except for the last 5–7 days before all experiments.

Sensitive and resistant cells in the exponential phase of growth were used throughout. NG108-15 and NG/DOXR cells were seeded at 1.5×10⁵ cells per 10 ml and were exposed to a range of concentrations of cytotoxic drugs in the absence or presence of chemosensitizers. After 48 h of continuous exposure, the cell number and volume distribution were determined using a Coulter counter (Coultronics). Each counting was made in quintuplicate.

Membrane currents and potentials were measured using the whole-cell configuration of the patch-clamp technique (Hamill et al. 1981). Experiments were performed on non-confluent cells at room temperature. The compositions of the pipette and bathing solutions are given in Table 1. Pipette resistances ranged from 2 to 5 MΩ (external and pipette A media). Capacitance was measured by integrating the uncorrected capacity transient. The ACQUIS1 suite of programs (DIPSI) was used for data acquisition and analysis.

Crude membrane preparations were prepared from the various cultured cells. The human epidermoid carcinoma cells KB3-1 were used as negative control and KBA-1 as a positive control for P-glycoprotein expression (Shen et al. 1986). Confluent cells were washed in 140 mmol l⁻¹ NaCl, 5 mmol l⁻¹ KCl, 20 mmol l⁻¹ Tris, 1 mmol l⁻¹ EDTA, pH 7.1, and then sedimented by centrifugation at 30 g for 5 min. The cells were then lysed and homogenized using a glass–Teflon tissue homogenizer in a hypotonic buffer (100 mmol l⁻¹ sucrose, 10 mmol l⁻¹ Tris, 1 mmol l⁻¹ EGTA, 0.5 mmol l⁻¹ dithiothreitol, 0.5 mmol l⁻¹ phenylmethylsulphonylfluoride, pH 7.1). The undisrupted cells and nuclear debris were removed by centrifugation at 180 g for 10 min. The supernatants were centrifuged for 60 min at 14 000 g, and membrane pellets were resuspended in 100 mmol l⁻¹ KCl, 10 mmol l⁻¹ Tris, 1 mmol l⁻¹ EGTA, 0.5 mmol l⁻¹ dithiothreitol, pH 7.1. Protein contents were estimated according to Bradford (1976). SDS–polyacrylamide gel electrophoresis was performed as described by Laemmli (1970). The samples were treated with solubilization buffer [50 mmol l⁻¹ Tris, 100 mmol l⁻¹ sucrose, 5% SDS (w/v), 2% β-mercaptoethanol (v/v), 0.02% Bromophenol Blue] and were resolved on a 7.5% to 15% acrylamide gel. After SDS–PAGE, proteins were electrophoretically transferred to nitrocellulose membranes. Non-specific binding to the blots was minimized by incubating them in a medium containing 3% milk powder. The nitrocellulose membranes were then incubated overnight at 4°C with C219, an anti-P-glycoprotein monoclonal antibody (0.5 μg ml⁻¹, final concentration in the milk-containing medium). Immunoreactive bands were visualized with peroxidase-conjugated anti-mouse IgG and an enhanced chemiluminescence detection system (Amersham France SA).

Chemicals used for cell culture were purchased from GIBCO. Bafilomycin A1 was provided by Professor K. Altendorf (Universität Osnabrück, Germany). Anti-P-glycoprotein monoclonal antibodies were obtained from CIS Biointernational, and secondary antibodies were obtained from Amersham France SA. KB3-1 and KBA-1 cells were kindly provided by M. F. Poupon (Institut Curie, URA CNRS 620, Paris, France). All other products were purchased from Sigma.

All results are expressed as the mean value ± S.E.M. (N), where N represents the number of cells or experiments. Statistical comparisons were performed using Student’s t-tests for paired or unpaired data as appropriate. Differences were considered significant at P<0.05.

Characterization of neuroblastoma cell multidrug resistance The sensitivity of the parental NG108-15 cells and NG/DOXR cells to doxorubicin is presented in Fig. 1A. The doxorubicin concentrations required for 50% inhibition of cell growth (IC₅₀) were 10 and 1300 nmol l⁻¹ for NG108-15 and NG/DOXR cells, respectively. The degree of resistance to doxorubicin (expressed as the ratio of IC₅₀ values for resistant to sensitive cell lines) was 130. One should note that, as already observed in human and murine cells (Vichi and Tritton, 1989), growth of NG108-15 and NG/DOXR cells was stimulated by doxorubicin concentrations below 1 nmol l⁻¹ and 100 nmol l⁻¹, respectively. The doubling time of NG108-15 cells, determined during the exponential growing phase, was 14.5 h (N=3). The rate of proliferation of NG/DOXR cells was not significantly different, with a doubling time of 17.8 h (N=3).

One property of MDR cells is their ability when exposed to a single drug to develop resistance, termed ‘cross-resistance’, to a broad range of cytotoxic compounds. NG/DOXR cells are cross-resistant to two other anticancer drugs: vinblastin and colchicine. Compared with NG108-15 cells, NG/DOXR cells were 87 times more resistant to colchicine (IC₅₀ 1300 nmol l⁻¹versus 15 nmol l⁻¹) and 54 times more resistant to vinblastin (IC₅₀ 540 nmol l⁻¹versus 10 nmol l⁻¹).

Many compounds have been shown to antagonize MDR in a variety of cell lines and in vivo tumour models when co-administered with chemotherapeutic agents to which the cells are resistant (Ford and Hait, 1990). In general, agents used to antagonize MDR, termed ‘chemosensitizers’, increase intracellular drug accumulation. Among these agents are verapamil and quinine. The mechanism by which they antagonize MDR is still uncertain. While there is evidence that they interact directly with P-glycoprotein, they also have other targets. In particular, verapamil is a classical Ca²⁺ channel blocker and to a lesser extent inhibits K⁺ channels; the antimalarial agent quinine is a potent K⁺ channel blocker. In NG/DOXR cells, 1 μmol l⁻¹ verapamil and 5 μmol l⁻¹ quinine produced respectively a 20- and 12-fold decrease in resistance to doxorubicin (Fig. 1B). It should be noted that, in contrast to quinine, verapamil also induced an apparent change in the slope of the dose–response curve.

P-glycoprotein expression was evaluated by Western blotting using the C219 monoclonal antibody, which recognizes a very well-conserved epitope in the C-terminal part of P-glycoprotein (Kartner et al. 1985). Immunoblotting analysis using KBA-1 cells as a positive control demonstrated the presence of P-glycoprotein in NG/DOXR cells (Fig. 2, lanes 1 and 3). A broad, heavily stained band was detected in NG/DOXR cells, but not in NG108-15 cells (Fig. 2, lanes 3 and 2). The labelled protein had an apparent molecular mass of 170 kDa, which is consistent with the known molecular mass of P-glycoprotein (Kartner et al. 1985).

Electrophysiological techniques allow the estimation of cell size by measurement of membrane capacitance, which is assumed to be proportional to cell membrane area. Whether or not an alteration in cell capacitance reflects changes in membrane properties, it should be determined in order to quantify ionic currents and the number of membrane transporters per unit membrane surface area. The size of NG108-15 and NG/DOXR cells was determined by two methods: measurement of volume distribution using a Coulter counter and measurement of membrane capacitance under the whole-cell patch-clamp configuration. The volume distributions of both NG108-15 and NG/DOXR cells were unimodal, and the mean volume of NG/DOXR cells was significantly smaller (2.89±0.06 pl, N=5) than that of NG108-15 cells (3.48±0.09 pl, N=5). The ratio of resistant to sensitive mean cell volume was 0.89±0.04. The measurement of capacitance also revealed that NG/DOXR cells were smaller than NG108-15 cells. The mean capacitance of NG/DOXR was 43.3±1.2 pF (N=40) and that of NG108-15 cells was 56.5±1.6 pF (N=41) (ratio 0.77±0.04).

While there is some evidence that drug-selected MDR cell lines are more differentiated than parental cells, contrasting evidence have been reported in other studies (Correale et al. 1994). Several chemical agents, including dimethylsulphoxide and dibutyryl-cAMP, induce NG108-15 cells to differentiate to neurones (Kimhi et al. 1976; Krystosek, 1985; Prasad, 1991). When cultured in the presence of 2% dimethylsulphoxide for 72 h and then exposed to 0.5 mmol l⁻¹ dibutyryl-cAMP, NG108-15 cells stop dividing, produce extensive processes, become electrically excitable and, compared with undifferentiated cells, exhibit a 5-to 10-fold increase in voltage-dependent Na⁺, Ca²⁺ and K⁺ currents. Moreover, the voltage-dependent K⁺ current of differentiated cells inactivates faster than that of undifferentiated cells (Rouzaire-Dubois and Dubois, 1991). By these criteria, NG/DOXR cells showed no sign of differentiation when cultured in the absence or in the presence of dimethylsulphoxide or dibutyryl-cAMP.

Resting potentials of NG108-15 and NG/DOXR cells were measured under current-clamp mode with KCl or potassium aspartate in the pipette. The results (Table 2) indicate that the resting potential of NG/DOXR cells is not significantly different (P>0.1) from that of NG108-15 cells. When internal Cl⁻ was replaced by aspartate, the resting potential was significantly reduced in both NG108-15 and NG/DOXR cells and by a similar amount. Since, under these conditions, the equilibrium potential of Cl⁻ was shifted towards negative values, the reduction in membrane potential was unlikely to have been due to changes in the Cl⁻ current.

The K⁺ current of undifferentiated NG108-15 cells is essentially a slowly inactivating delayed rectifier current (Rouzaire-Dubois and Dubois, 1991). No significant differences were observed in the physiological properties of this current recorded in NG108-15 and NG/DOXR cells (Table 3). Moreover, tetraethylammonium (TEA⁺) and verapamil had similar effects on the K⁺ currents of both NG108-15 and NG/DOXR cells. At a depolarizing potential of +50 mV, 10 mmol l⁻¹ TEA⁺ reduced the K⁺ current to 27±2% of the control value in NG108-15 cells (N=5) and to 22±1% (N=5) in NG/DOXR cells, and 10 μmol l⁻¹ verapamil reduced the peak K⁺ current to 50±2% (N=3) in NG108-15 cells and to 59±4% (N=4) in NG/DOXR cells. In contrast to results previously described in multidrug-resistant small cell lung cancer cells expressing MRP (a multidrug-resistance-associated protein distantly related to P-glycoprotein) (Jirsch et al. 1993), an inwardly rectifying K⁺ current could not be detected in NG/DOXR cells.

Current–voltage relationships for the peak Na⁺ current (I_Na) in NG108-15 and NG/DOXR cells are shown in Fig. 3A. Current density estimates revealed that the amplitude of the maximum peak I_Na in NG/DOXR cells (−19.68±2.07 pA pF⁻¹, N=8) was significantly greater than that in NG108-15 cells (−9.96±1.41 pA pF⁻¹, N=7) (P<0.001). Na⁺ conductance was increased in NG/DOXR cells by 1.66-fold without significant changes in its voltage sensitivity, except for a 7 mV shift in the steady-state inactivation curve towards negative potentials (Fig. 3B; Table 3). The addition of 1 μmol l⁻¹ tetrodotoxin to the external medium suppressed the Na⁺ current in NG108-15 and NG/DOXR cells but did not alter the sensitivity of either cell type to doxorubicin (results not shown).

To our knowledge, no Cl⁻ currents have yet been described in NG108-15 cells. Under different experimental conditions, two types of Cl⁻ current could be observed in both NG108-15 and NG/DOXR cells: a cAMP-activated current and a volume-sensitive current. A cAMP-activated Cl⁻ current could be revealed by the addition of dibutyryl-cAMP, forskolin and isobutylmethylxanthine (IBMX) to the solution bathing the cells. Fig. 4A shows that the cAMP-activated Cl⁻ current is 7–10 times larger in NG108-15 cells than in NG/DOXR cells (see also Table 3). The ionic nature of this current was confirmed by replacement of 80 mmol l⁻¹ external Cl⁻ with isethionate. In these conditions, the cAMP-activated current observed in NG108-15 cells reversed at +18 mV (N=3), a value closed to the Nernst equilibrium potential for Cl⁻ (+20 mV). A volume-sensitive Cl⁻ current could be elicited by exposing the cells to hypo-osmotic external solutions (24% reduction in osmolarity). In NG108-15 cells, this current was sustained for prolonged periods when the cells remained swollen and were dialysed with a pipette solution containing ATP. In contrast, as already observed by Jirsch et al. (1994) in multidrug-resistant lung cancer cells, ATP was not required for the activation and maintenance of the current in NG/DOXR cells. In both cell types, the current was Ca²⁺-independent since the addition of 2 mmol l⁻¹ EGTA to the pipette solution did not affect its activation. When recorded during square voltage steps of various amplitudes from a holding potential of 0 mV, the current showed a pronounced outward rectification; no difference in its density was found between cell types (Fig. 4B).

The resting potential of most eukaryotic cells is generally assumed to be essentially controlled by a selective membrane permeability to K⁺. However, we recently proposed arguments in favour of the idea that the resting potential of differentiated neuroblastoma cells is maintained at negative values by both voltage-dependent K⁺ channels and an electrogenic vacuolar-type H⁺-ATPase (V-ATPase) (Gérard et al. 1994). This conclusion was mainly based on the observations that the V-ATPase inhibitors bafilomycin A1 (BFA) (Bowman et al. 1988), NO₃⁻ and N-ethylmaleimide and the K⁺ channel blocker Cs⁺ induced additive depolarizations. In most cells, V-ATPases function in a variety of intracellular systems where they contribute to decreasing intraorganellar pH. V-ATPases are also functionally expressed in the plasma membrane of several animal cells, including tumour cells (Gluck, 1992; Harvey and Wieczorek, 1997; Nanda and Grinstein, 1991), and evidence has been provided for the involvement of a V-ATPase in MDR (Ma and Center, 1992; Marquardt and Center, 1991). Taking all these factors into consideration, the effects of BFA, NO₃⁻ and Cs⁺ on the resting potential of NG108-15 and NG/DOXR cells were studied. Fig. 5 shows that these three agents depolarized both cell types by 20–30 mV. Although the depolarizations induced by Cs⁺ in sensitive and resistant cells were not significantly different, the depolarizations induced by BFA and NO₃⁻ were significantly (P<0.05) larger in resistant (BFA, 30±2 mV, N=11; NO₃⁻, 30±2 mV, N=7) than in sensitive (BFA, 23±3 mV, N=10; NO₃⁻, 24±1 mV, N=6) cells.

Since cytoplasmic pH and/or V-ATPase activities are thought to play a role in MDR and the resting potential of NG108-15 and NG/DOXR cells seems to be partly determined by a H⁺ pump (see below), we analyzed the effects of the chemosensitizing agents verapamil and quinine on the membrane potential. Before the application of external verapamil (10 μmol l⁻¹) or quinine (100 μmol l⁻¹), the membrane potential was held at different levels between −40 and −90 mV using constant injected currents. Verapamil and quinine induced instantaneous reversible depolarizations in NG/DOXR cells whatever the initial potential. In NG108-15 cells, verapamil and quinine induced depolarizations when the initial potential was more positive than −60 mV, but had no effect for initial potentials near or more negative than the resting potential (−70 mV) (Fig. 6A). In NG/DOXR cells, the depolarization induced by verapamil from an initial potential of −70 mV was inhibited by BFA but was not significantly modified by Cs⁺. When the initial potential was −40 mV, the verapamil-induced depolarization was not significantly modified by BFA but was decreased by Cs⁺ in both NG108-15 and NG/DOXR cells (Fig. 6B).

The preceding observations suggest that the V-ATPase of NG/DOXR cells (but not that of NG108-15 cells) was inhibited by verapamil and quinine. Given that such inhibition should induce a decrease in cytoplasmic pH and, as shown in kidney cells (Mori et al. 1994), that the volume-sensitive Cl⁻ current may be highly sensitive to intracellular pH, we examined the effects of intracellular pH, BFA and verapamil on the activation of this Cl⁻ current. Fig. 7A,B shows that, in both NG108-15 and NG/DOXR cells, the volume-sensitive Cl⁻ current was greatly reduced when the pH of the pipette solution was lowered from 7.3 to 6.8. Similar effects were observed with BFA (Fig. 7C,D). In contrast, verapamil affected the volume-sensitive Cl⁻ current differentially in NG108-15 and NG/DOXR cells. Whereas verapamil had a minor effect on NG108-15 cells, it reduced the volume-sensitive Cl⁻ current of NG/DOXR cells to less than 20% of the control value (Fig. 7E,F).

The electrophysiological studies presented above show that drug resistance is accompanied by several alterations in ionic membrane permeabilities. Among these, modifications to the pharmacological properties of a plasma membrane V-ATPase seem to be particularly interesting. Taking into account the suggested role of intracellular pH in drug resistance (see Introduction), one can ask whether the V-ATPase contributes in some way to doxorubicin resistance. To answer this question, we examined the effects of BFA (0.5 or 1 μmol l⁻¹) on the sensitivity of NG/DOXR cells to doxorubicin (100 or 200 nmol l⁻¹). BFA alone was cytotoxic and decreased the cell number to approximately 40% of the control value. When BFA was applied in the presence of doxorubicin, the effect was not significantly different from that of BFA alone (Table 4).

We have established that NG108-15 neuroblastoma × glioma hybrid cells resistant to doxorubicin exhibit an MDR phenotype. They are cross-resistant to two other anticancer drugs: vinblastin and colchicine, express a 170 kDa membrane protein identified as the P-glycoprotein and their drug resistance is antagonized by the classical chemosensitizers verapamil and quinine. The development of MDR in neuroblastoma cells is accompanied by several alterations in the expression levels or properties of ionic channels or transporters. These alterations include increased expression or activity of Na⁺ channels, decreased expression or activity of cAMP-activated Cl⁻ channels and acquired sensitivity to verapamil and quinine of an electrogenic plasma membrane system which has the pharmacological properties of a vacuolar-type H⁺-ATPase.

It is generally assumed that an increased expression of P-gp is sufficient for the development of an MDR phenotype. However, although overexpression of P-gp is associated with cellular changes that could be directly responsible for MDR, a number of forms of drug resistance not associated with P-gp have been observed. Moreover, P-gp overexpression alone has not been associated with levels of drug resistance above approximately 10-fold, and high levels of drug resistance are related more to drug selection than to P-gp overexpression (Hoffman et al. 1996). The observation that NG/DOXR cells are 130 times more resistant to doxorubicin than NG 108-15 cells suggests that additional changes have been induced in these cells by the doxorubicin selection. Independently of P-gp overexpression, these events may be responsible for at least part of the observed alterations in ionic membrane permeabilities.

The enhancement of Na⁺ current density in NG/DOXR cells is comparable with that observed by Yamashita et al. (1987) in MDR human leukaemia cells. As suggested by these authors, it may be a consequence of either an increased expression of genes encoding Na⁺ channels or an alteration in the lipid structural order of the membrane.

The decrease in cAMP-activated Cl⁻ current density observed in NG/DOXR cells may be due to a decreased expression of genes encoding these channels. Another possibility is that cyclic AMP is pumped out of the cells by P-gp, which would explain the absence of NG/DOXR cell differentiation in the presence of dibutyryl-cAMP.

An altered expression or activity of Cl⁻ channels has been thought to play a role in MDR. According to Gill et al. (1992) and Valverde et al. (1992), the appearance of P-gp is associated with the appearance of Cl⁻ currents activated by cell swelling, suggesting a dual role for P-gp as a drug transporter and a Cl⁻ channel. In constrast, several recent studies have indicated a lack of correlation between P-glycoprotein expression and volume-sensitive Cl⁻ channel activity (De Greef et al. 1995; Dong et al. 1994; Ehring et al. 1994; Kunzelmann et al. 1994; Morin et al. 1995; Rasola et al. 1994; Wang et al. 1994). Our results, showing the presence of volume-sensitive Cl⁻ channels with similar activity in both drug-sensitive and drug-resistant cells, led us to conclude that, in neuroblastoma cells, there is no relationship between P-glycoprotein expression and the Cl⁻ channel activated by cell swelling.

Another phenomenon observed in highly MDR cells is an alkaline shift in intracellular pH, suggesting that protons may be extruded from the cells by the action of P-gp. This could be due to a direct effect of P-gp (which would be a proton pump) or to an indirect effect (either the cytosolic substrate of P-gp is protonated or P-gp activity is coupled with proton transport, which provides the motive force for drug transport) (Gottesman and Pastan, 1993; Thiebaut et al. 1990). Another possibility is that P-gp affects pHi by altering the transport of HCO₃⁻ (Hoffman et al. 1996). We have previously presented arguments in favour of the contribution of an electrogenic V-ATPase to the genesis of NG108-15 cell membrane potential (Gérard et al. 1994). This conclusion is confirmed here in both sensitive and resistant cells by the observation that their resting potentials are decreased by BFA and by replacement of internal Cl⁻ with NO₃⁻ or aspartate (Fig. 5; Table 2). The observation that resting potentials of sensitive and resistant cells are not significantly different may indicate that MDR is not associated with an increased expression of V-ATPase. However, we suggest that the properties of the V-ATPase are modified in MDR cells since (1) depolarizations induced by BFA and NO₃⁻ are larger in NG/DOXR than in NG108-15 cells, and (2) verapamil induces membrane depolarizations in NG/DOXR cells that are abolished when the cells are pretreated with BFA. Alterations to the properties of the V-ATPase of MDR cells were confirmed by the observations that the volume-sensitive Cl⁻ current was inhibited by intracellular acidification and BFA in both NG108-15 and NG/DOXR cells, whereas it was inhibited by verapamil in NG/DOXR cells only. We thus conclude that MDR is accompanied by an acquired sensitivity to verapamil of a vacuolar-type H⁺-ATPase. As a consequence, verapamil should induce a decrease in intracellular pH (Keizer and Joenje, 1989), which results in an inhibition of the volume-sensitive Cl⁻ current. The vacuolar H⁺-ATPases of plants have been shown to be inhibited by verapamil (Pfeiffer, 1995). Given the lack of information on the sensitivity to verapamil of V-ATPases in animal cells, the origin of the differential sensitivity to verapamil of drug-sensitive and drug-resistant neuroblastoma cells cannot be precisely identified. It may be due to the existence of two isoforms of one subunit of the ATPase, differentially sensitive to verapamil and differentially expressed in NG108-15 and ND/DOXR cells. Other possibilities are MDR-associated changes in the membrane lipid composition or interactions between V-ATPase and P-glycoprotein. It is important to determine whether this difference is the cause of MDR via an alteration in intracellular pH. Our results showing that BFA did not reverse the resistance of NG/DOXR cells to doxorubicin argue against a role of pHi in multidrug resistance of drug-selected MDR cells (see also Altenberg et al. 1993). However, in addition to its action on pHi, it has been shown in Swiss 3T3 fibroblasts that BFA inhibits DNA synthesis and induces a rounding up of cells (Saurin et al. 1996). Moreover, we have shown that BFA depolarizes neuroblastoma cells. These effects may induce a decrease in chemotherapeutic drug accumulation or efficacy and thus may annul a possible reversing action of BFA mediated by a decrease in pHi.

In conclusion, our results show that, in addition to P-glycoprotein overexpression, MDR in neuroblastoma cells is accompanied by alterations in ionic membrane permeabilities with, in particular, modifications in the properties of a plasma membrane V-ATPase. This latter point may be an important factor which should be examined further.

We are pleased to thank Gérard Sadoc for data processing assistance and Ian Findlay and Ray Kado for critical reading of the manuscript.