Introduction

Top Abstract Introduction Materials and Methods Results Discussion Appendix References

Decreased drug accumulation is a common mechanism of antineoplastic drug resistance in vitro (Beck and Dalton, 1997). In many cell lines, the development of multiple drug resistance to structurally unrelated antineoplastic agents has been associated with decreased accumulation of drug and increased drug efflux (multidrug resistance phenotype, MDR³). An increasing number of cellular membrane proteins have been reported to be associated with this phenomenon of multidrug resistance (reviewed in van Veen and Konings, 1998). The most extensively studied of the MDR proteins is P-glycoprotein (Pgp), the product of the multidrug resistance gene mdr-1 (Biedler and Riehm, 1970; Beck et al., 1979; Riordan and Ling, 1985; Qian and Beck, 1990; Horwitz et al., 1993; Bhalla et al., 1994; Dumontet et al., 1996). This protein has been shown to mediate the ATP-dependent active efflux of a number of anticancer agents (Sharom et al., 1993, 1996; Eytan et al., 1994, 1996; Ruetz and Gros, 1994; Shapiro and Ling, 1995). However, a number of investigators have noted that the kinetic parameters measured for drug efflux in vesicle studies are much too low to explain the decrease in intracellular drug accumulation (Demant et al., 1990; Bornmann and Roepe, 1994; Roepe, 1995). Thus, an alternative mechanism of action that has been proposed is that Pgp alters the properties of the membrane to increase drug efflux (Demant et al., 1990; Bornmann and Roepe, 1994; Roepe, 1995). Kinetic studies may be able to differentiate between these two models.

Flow cell techniques are an established means for characterizing efflux kinetics of drugs (Spoelstra et al., 1992) and ions (Frank et al., 1977). These techniques are based on monitoring the time-dependent efflux of a drug in effluent buffer from flow cells containing immobilized cells loaded with drug and washed continuously with fresh buffer. Total drug efflux, initial velocity of efflux, and initial drug concentration can be obtained by mathematical analysis of integrated data. The advantages of this technique include the ability to monitor drug efflux continuously and minimization of drug re-entry into cells since excreted drug is removed by the flow of buffer. Furthermore, the continuous flow of fresh buffer obviates the possibility of an apparent prolonged drug retention due to equilibration between the cell and the nominally drug-free buffer. In addition, since only a small number of cells is needed, this technique may have potential for examining drug efflux in clinical samples. However, because cells must be immobilized, this technique has been limited in the past to cell types grown in monolayer culture.

In this article, we report the development of a flow cell technique designed for cells growing in suspension culture. To validate this technique, we examined the effects of Pgp on paclitaxel efflux using a well characterized leukemic cell model for Pgp-mediated drug resistance, the multidrug-resistant CEM/VLB100 cell line, which highly overexpresses Pgp, compared with its parent sensitive CCRF-CEM myelogenous leukemia cell line. We chose paclitaxel because it is a commonly used antineoplastic agent (Rowinsky and Donehower, 1996) that has recently been investigated as a salvage chemotherapy agent in relapsed human acute myelogenous leukemia (Curtis et al., 1996; Munker et al., 1998). Paclitaxel resistance has been linked with decreased intracellular drug accumulation and increased Pgp expression. The initial rate kinetic parameters of efflux were examined under conditions of no external drug in these two well characterized cell lines. Our studies show that the majority of paclitaxel efflux from these cells occurs as a rapid efflux component that increases linearly with paclitaxel concentration, suggesting passive diffusion. Interestingly, this rapid efflux component is increased in Pgp-overexpressing CEM/VLB100, suggesting that Pgp affects the rapid passive efflux of paclitaxel by altering membrane properties to enhance passive efflux. A slower saturable component of efflux was observed in both the parental and Pgp overexpressing cells, and it was found to be distinct in kinetic character in CEM/VLB100 cells, with larger V_max and lower k_m than the parental cells. Verapamil and low temperature, known inhibitors of primary active transport by Pgp, nearly abrogated the slower saturable component in resistant CEM/VLB100 cells, but also partially inhibited the rapid component. These results validated the ability of the present flow cell system for detecting a saturable primary active transport presumably mediated by Pgp. These results also suggest that verapamil can modulate Pgp-mediated drug-accumulation defects both by inhibiting Pgp-mediated transport as well as by interfering with the effects of Pgp on passive paclitaxel efflux.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion Appendix References

Reagents. [³H]Paclitaxel (specific activity: 6.2 Ci/mmol) and [¹⁴C]inulin (1 mCi/ml) were obtained from Moravek Biochemicals (La Brea, CA). Culture supplies were obtained from Life Technologies Inc. (Gaithersburg, MD). Concanavalin A (conA) linked to Sepharose-4B (conA-seph, catalog no. C-9017) was obtained from Sigma Chemicals (St. Louis, MO).

Culture Conditions. Human leukemia cell lines CCRF-CEM and its multidrug-resistant subline CEM/VLB100 were gifts from William T. Beck (University of Illinois at Chicago) (Beck et al., 1979). The cell subline CEM/VLB100 was originally developed by Dr. Beck for resistance to vinblastine and is 200- to 800-fold resistant compared with the parent line. It has shown to overexpress Pgp, the protein product of the multidrug resistance gene mdr-1 (Beck et al., 1979; Kuttesch et al., 1996). We confirmed that the CEM/VLB100 cells were resistant to paclitaxel compared with the parent cell line (data not shown). Cells were maintained in the log phase of growth by diluting them 1:10 in RPMI 1640 containing 10% fetal bovine serum with penicillin and streptomycin (final concentration: 50 U/ml each) every 3 to 4 days.

Drug Efflux Studies. A Bio-Spin disposable chromatography column (Bio-Rad 732-6008, Hercules, CA) was used for the flow cell (Fig. 1). Concanavalin A linked to conA-seph beads (Sigma Chemicals) was used as a matrix to immobilize the cells. Hanks' balanced salt solution (HBSS) (pH 7.4) containing 1 mg/ml glucose was used as the buffer. The conA-seph was washed thoroughly with methyl--D-mannopyranoside 50 mM to remove any contaminants that could affect drug washout (Goldstein et al., 1965). A bed volume of 0.45 ml (37°C) or 0.6 ml (4°C) of conA-seph was sufficient to retain >99% of cells within the column for the duration of the assay. The column was equilibrated with buffer at a flow rate of 1 ml/min using an LKB (Broma, Sweden) 2132 peristaltic pump. The column was placed inside either a 37°C incubator or a 4°C cold room, and the temperature of the buffer was monitored.

View larger version (10K): [in this window] [in a new window]   Fig. 1.   Diagram of the flow cell. The flow cell consisted of a Bio-Spin disposable chromatography column containing conA bound to Sepharose beads. The stopper is made from a 00 rubber stopper with two holes drilled through it and a separate rubber piece glued on top covering the larger hole. Tubing is passed through the smaller hole to allow buffer to be pumped through the cell. Cells are injected through the larger hole into the flow cell.

(1)

Kinetic Analysis of Paclitaxel Efflux from Cells. Compartment analysis of paclitaxel distribution and efflux from intracellular compartments required corrections for the effects of trapped extracellular paclitaxel, the kinetics of buffer washout, and of conA-seph on the egress of paclitaxel from the flow cell. To determine the amount of trapped buffer, 1 × 10⁷ cells were incubated with [¹⁴C]inulin in an identical manner to that used for loading cells with [³H]paclitaxel, then the supernatant was removed, and an aliquot was counted to determine counts per volume. The cell pellet was then removed and its volume measured using a Pipetman (Gilson, Middleton, WI), and the amount of [¹⁴C]inulin trapped in the cell pellet was determined and converted to extracellular volume. The measured counts per volume of the [¹⁴C]inulin was then used to determine the volume of trapped extracellular buffer. The amount of extracellular volume trapped within the cell pellet was found to be 54 ± 5% (n = 3) of the total volume (25 µl). The kinetic properties of buffer washout from the flow cell were also evaluated by injecting [¹⁴C]inulin alone, or following injection of 1 × 10⁷ cells into the flow cell, and by monitoring the radioactivity in the effluent in the absence or presence of cell; they followed a single exponential decay curve, unaffected by the presence of cells ( = 0.863 ± 0.089 min without cells, n = 3; 0.859 ± 0.033 min with cells, n = 3; and 0.861 ± 0.061 min overall).

View larger version (16K): [in this window] [in a new window]   Fig. 2.   Paclitaxel washout from the flow cell at 37°C as a function of time. [3H]paclitaxel in 0.2 ml of HBSS buffer containing 1 g/l glucose was injected into the flow cell, and aliquots of the effluent were collected every 30 s. To obtain the time-dependent curve of residual paclitaxel in the flow cell, the radioactive counts were added, beginning with the column contents at the end of the experiment. Next, the last collected sample before the end of the experiment was added, followed by the progressive addition of earlier aliquots, ending with the first collected sample and then normalized. Results represent average and standard deviation of two experiments.

(2)

F(t)		=		the measured amount of paclitaxel in the flow cell at time t,
T		=		the measured time constant of buffer washout from the flow cell in the absence of cells,
		=		the measured fractional size of the rapid compartment for paclitaxel washout,
T1		=		the measured initial time constant of paclitaxel washout,
T2		=		the measured terminal time constant of paclitaxel washout,
A		=		rapid cellular efflux drug amount,
B		=		terminal cellular efflux drug amount,
1		=		rapid cellular efflux time constant, and
2		=		terminal cellular efflux time constant.

(4)

Statistical Analysis. Straight line fits to the kinetic data were performed using linear regression analysis by the SAS statistical package (SAS Institute Inc., Cary, NC), with the lines forced through zero, and their slopes compared using standard F tests. Fitting to the Michaelis-Menten equation was performed using nonlinear regression analysis in the MLAB statistical computer package (Civilized Software Inc., Bethesda, MD) to obtain estimates of the parameters and standard errors for those parameter estimates. In nonlinear regressions, exact p values cannot be calculated; however, the following approach was used to give a reasonable approximation for comparison of the curves. Assuming the null hypothesis, the computed difference of the parameter estimates was compared with the standard error of the difference. Given the number of data points, the T distribution can be well approximated by a standard normal distribution. Thus, for differences in the estimates of parameters greater than 2 times the standard error of the difference, the p value is less than 0.05, and for differences greater than 2.5 times the standard error of the difference, the p value is less than 0.02. A difference in parameter estimates of either the k_m or V_max greater than 2 times the standard error of the difference was accepted as demonstrating that the curves were different.

	Results

Top Abstract Introduction Materials and Methods Results Discussion Appendix References

Paclitaxel Efflux from CEM and CEM/VLB100. Measurements of initial rate kinetics for paclitaxel efflux at 4 and 37°C were performed using cells loaded with varying initial concentrations of paclitaxel between 0.3 and 2 µM. The lowest initial concentration used was limited by the specific activity of the [³H]paclitaxel available, combined with a requirement for cpm per collected sample at least 10-fold above background. Results of a representative experiment fitted to theory are shown; the data are well described by the theoretical curve (Fig. 3). For the experiments described, r² averaged 0.9990 ± 0.0005 and 0.9982 ± 0.0013 for CEM cells and CEM/VLB100 cells, respectively. The amount of drug in CEM cells was greater for both compartments compared with CEM/VLB100 cells as a function of paclitaxel concentration (Fig. 4), with the amount of drug in compartment A increasing linearly with concentration (CEM/VLB100 slope = 166 ± 10 pmol/µM/10⁷ cells versus CEM slope = 232 ± 12 pmol/µM/10⁷ cells, p < 0.001), whereas compartment B appeared to saturate with increasing paclitaxel concentration (B_max = 146 ± 11 pmol/10⁷ cells, k = 21 ± 12 nM for CEM cells versus B_max = 108 ± 11 pmol/10⁷ cells and k = 170 ± 60 nM for CEM/VLB100 cells, p < 0.02) in both cell lines.

View larger version (17K): [in this window] [in a new window]   Fig. 3.   Fit of observed washout to the predicted model. For these studies, 10 × 106 CEM/VLB100 cells were loaded with 1 mM [3H]paclitaxel in the medium for 30 min at 37°C. Cells were centrifuged, and the supernatant was removed and measured for equilibrium drug concentration. The cell pellet was resuspended in 0.2 ml of HBSS at 37°C, injected into the flow cell, and washed with HBSS buffer containing 1 g/l glucose at a flow rate of 1 ml/min. Aliquots were collected every 30 s. Radioactivity data were converted to pmol/107 cells. The observed time-dependent curve of residual paclitaxel in the flow cell was obtained by starting with the column contents at the end of the experiment and progressively adding earlier sample time points, ending with the first sample collected (). The extracellular equilibrium concentration of paclitaxel was used to normalize the paclitaxel washout curve, which was then subtracted from the overall flow cell curve to obtain the theoretical efflux curve given by eq. 3 (see Materials and Methods) (solid line; r2 = 0.9951). See Appendix for details on the calculation of the fitted curve.

View larger version (20K): [in this window] [in a new window]   Fig. 4.   The effect of varying paclitaxel concentration on the amount of drug in each compartment in CEM () and resistant CEM/VLB100 ()cells. The amount of drug in compartment A (A) and compartment B (B) was obtained by fitting the proposed model (eq. 3) to the observed drug washout data as described under Materials and Methods and Appendix. A, solid and dashed lines represent the best linear fit to the data for CEM and CEM/VLB100 cells, respectively, forced through zero. B, lines represent the fit to the observed data using the Michaelis-Menten equation to model the trend. Each point represents the average and standard deviation of at least three experiments.

View larger version (23K): [in this window] [in a new window]   Fig. 5.   Initial rate kinetics of paclitaxel efflux and the effect of verapamil. Plots of equilibrium internal paclitaxel concentration [S] versus initial rate of paclitaxel efflux velocity in pmol/min/107 cells for sensitive CEM (), resistant CEM/VLB100 cells (), and resistant CEM/VLB100 cells in the presence of 10 µM verapamil () are presented. All experiments were performed at 37°C. A, comparison of data for compartment A with best linear fits forced through zero with the data for CEM and CEM/VLB100 cells in the presence or absence of verapamil. B, data for compartment B for CEM and CEM/VLB100 cells in the presence or absence of verapamil, fitted to the Michaelis-Menten equation. Solid and dashed lines, best linear fit to the data for CEM and CEM/VLB100 cells, respectively. Each point represents the average and standard deviation of at least three experiments.

Effects of Verapamil and Temperature on Paclitaxel Efflux from CEM/VLB100. To test the hypothesis that the difference in V_max observed was due to Pgp, the effect of verapamil, a known inhibitor of Pgp function, on paclitaxel efflux in CEM/VLB100 cells was studied. When CEM/VLB100 cells were pretreated with 10 µM verapamil, paclitaxel efflux from compartment B was decreased (V_max = 1.6 ± 0.2 pmol/min/10⁷ cells) below CEMVLB100 cells in the absence of verapamil (p < 0.01) and also below the sensitive CEM cells in the absence of verapamil (p < 0.02), suggesting a large contribution of Pgp to the maximum rate of paclitaxel efflux from compartment B efflux (Fig. 5B, Table 1). These results indicate that compartment B kinetics in CEM/VLB100 cells is a verapamil-sensitive process, consistent with Pgp.

View larger version (21K): [in this window] [in a new window]   Fig. 6.   Temperature dependence of paclitaxel efflux. Plots of equilibrium internal paclitaxel concentration, ([S], µM) versus initial rate of paclitaxel efflux (v, pmol/min/107 cells) for resistant CEM/VLB 100 cells at 37°C () versus 4°C () are presented. A, the results for compartment A, statistical analysis by F test, slope of CEM/VLB100 at 37 versus 4°C, p < 0.0001. B, results for compartment B. Each point represents the average and standard deviation of at least three experiments.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion Appendix References

Resistance of cancer cells to the cytotoxic effects of structurally diverse chemotherapeutic agents (MDR) appears to be mediated in many cases by overexpression of membrane transporters capable of transporting a wide variety of substrates; however, the in situ kinetic properties of these transporters are not completely understood. Drug transport by purified, reconstituted Pgp, the best characterized of these transporters, coupled with ATP hydrolysis has been demonstrated for a number of substrates (Sharom et al., 1993, 1996; Eytan et al., 1994, 1996; Shapiro and Ling, 1995); however, it has been argued that the kinetic constants derived in these studies appear inadequate to account for the differences in drug accumulation observed (Demant et al., 1990; Bornmann and Roepe, 1994; Roepe, 1995).

Flow cell assays (Frank et al., 1977; Spoelstra et al., 1992) use the entire cell aliquot for the duration of the experiment and allow continuous monitoring of drug efflux and measurement of kinetic parameters in an intact cell system. However, such assays have been limited by the requirement for attached cells (Frank et al., 1977; Spoelstra et al., 1992). Thus, we developed a simple, disposable flow cell assay for analyzing drug efflux in malignant cells grown in suspension culture. Although it is possible that binding of cells to conA-seph could alter drug efflux via membrane interactions, a number of investigators using different methodologies (Spoelstra et al., 1992; Wadkins and Houghton, 1993; Bornmann and Roepe, 1994) have noted a two-exponential retention curve for a variety of drugs, suggesting that at least the qualitative aspects of drug efflux are preserved in our system.

CEM/VLB100 cells were highly resistant to paclitaxel compared with the parent CCRF-CEM line (data not shown), consistent with previous observations of cross-resistance to paclitaxel in other Pgp-overexpressing cells (Horwitz et al., 1993; Bhalla et al., 1994; Dumontet et al., 1996). Although our experiments were limited by the specific activity of the radioactive paclitaxel, we were able to examine a range of drug concentrations from below the measured human peak plasma concentrations (Rowinsky and Donehower, 1996) to 10-fold higher, encompassing a clinically relevant range.

It was noted that both the paclitaxel washout curve and the cellular drug efflux curve exhibited an initial shoulder (Figs. 2 and 3), which does not occur when directly observing drug efflux from cells (Wadkins and Houghton, 1993; Bornmann and Roepe, 1994). However, in a flow cell the drug is not collected immediately upon its exit from the cell membrane but must first be washed out of the flow cell. Our mathematical model for this physical process automatically produces the shoulder observed in the data, without requiring any manipulation of the equations. Thus, we believe this shoulder occurs due to the physical properties of the flow cell. Since the measured time constants for buffer washout, the rapid phase of drug washout, and the rapid phase of drug efflux from the cells were all relatively close to each other, none of these physical factors could be ignored (see Appendix). The experimental data demonstrated a very satisfactory fit to the theoretical model.

The flow cell was clearly able to differentiate between the kinetics of wild-type and Pgp-overexpressing CEM cells. In both the sensitive and resistant cell lines, there was an initial rapid efflux proportional to the calculated equilibrium internal drug concentration, compatible with a passive efflux process. Notably, this initial rapid efflux was increased in the resistant cells as compared with the sensitive cells, and was partially inhibited by the addition of verapamil. This could represent an alteration in passive diffusion, consistent with the membrane model of Pgp action (Demant et al., 1990; Wadkins and Houghton, 1993; Pawagi et al., 1994; Seydel et al., 1994; Drori et al., 1995; Ayesh et al., 1996), or passive diffusion plus the tail end of a very high capacity, relatively low-affinity transport mechanism. Further studies with the use of energy poisons such as sodium azide or 2-deoxyglucose may help resolve this, although such poisons can also affect intracellular pH, which has also been suggested as a mechanism of Pgp action (Roepe, 1995).

In CEM/VLB100 cells, terminal efflux (compartment B) was also increased compared with wild-type CEM cells, saturable, preferentially inhibited with the Pgp inhibitor verapamil, and almost completely abolished at 4°C, consistent with a carrier-mediated transport model of Pgp (Horio et al., 1988; Jusa and Tsuruo, 1989; Higgins and Gottesman, 1992; Sharom et al., 1993). Since the current studies were performed in zero-trans conditions (no external drug), we cannot address the question of whether such transport is active, which would require demonstrating transport against a gradient.

Since CEM/VLB100 cells were selected for vinca alkaloid resistance, an alternative explanation for these findings is that there could be alterations in tubulin as well as transport, resulting in decreased binding of drug to its target and a more rapid terminal efflux phase in the resistant cell subline (Sirotnak et al., 1986; Pain et al., 1988). However, CEM/VLB100 cells have not been reported to have altered tubulins. Furthermore, if the observed changes in the terminal phase in our studies are caused by decreased binding of paclitaxel to tubulin in CEM/VLB100 cells, it is not obvious why this binding should be affected by verapamil, which clearly decreased paclitaxel efflux in our studies.

Our results show alterations in both efflux compartments in the resistant cells, suggesting that both models of Pgp action may be operating in parallel. Interestingly, the alteration in the passive efflux component appeared to be significantly larger than the saturable component. One critique of the standard transport model of Pgp action is that the measured kinetic constants of carrier-mediated transport of Pgp (albeit with other drugs) appeared to be too small to counteract passive influx. It appears, at least in the CEM/VLB100 cell line, that alterations in passive efflux are a significant contributor to decreased intracellular concentrations. Our observations are consistent with those of Wielinga and colleagues (2000), who observed a significant contribution of passive efflux to P-glycoprotein-mediated anthracycline efflux. However, because CEM/VLB100 cells were derived by exposure to drug in vitro, we cannot rule out the possibility that some of the effects we observed may be due to additional resistance mechanisms aside from overexpression of Pgp, such as alterations in membrane lipids (reviewed in Ferte, 2000).

Our flow cell method produced results that are at least qualitatively similar to other reports and could represent a relatively simple and rapid method to obtain initial kinetic rate constants in intact suspension cells. This method has the advantage of not requiring that cells be subjected to the potentially deleterious effects of azide, MDR reversal agents, or other treatments, or to the rigors of vesicle preparation. In addition, it has the advantage of displaying all components of drug efflux quantitatively and simultaneously. In contrast to fluorescent microscopy or flow cytometry methods, it does not require specialized equipment and uses widely available radioactive compounds. However, unlike those methods, this approach cannot examine single-cell drug kinetics. Furthermore, since it requires a relatively small sample per assay, this technique may be useful in studying drug efflux in clinical samples.

To validate this promising methodology, it will be necessary to perform flow cell studies in cells transfected with, and overexpressing transporters using, drugs whose kinetic parameters have been previously reported in the literature to confirm that similar results can be obtained, and such studies are in progress.