Correlation between Steady-state ATP Hydrolysis and Vanadate-induced ADP Trapping in Human P-glycoprotein

P-glycoprotein (Pgp) is a transmembrane protein conferring multidrug resistance to cells by extruding a variety of amphipathic cytotoxic agents using energy from ATP hydrolysis. The objective of this study was to understand how substrates affect the catalytic cycle of ATP hydrolysis by Pgp. The ATPase activity of purified and reconstituted recombinant human Pgp was measured using a continuous cycling assay. Pgp hydrolyzes ATP in the absence of drug at a basal rate of 0.5 μmol·min·mg−1 with aK m for ATP of 0.33 mm. This basal rate can be either increased or decreased depending on the Pgp substrate used, without an effect on the K m for ATP or 8-azidoATP and K i for ADP, suggesting that substrates do not affect nucleotide binding to Pgp. Although inhibitors of Pgp activity, cyclosporin A, its analog PSC833, and rapamycin decrease the rate of ATP hydrolysis with respect to the basal rate, they do not completely inhibit the activity. Therefore, these drugs can be classified as substrates. Vanadate (Vi)-induced trapping of [α-32P]8-azidoADP was used to probe the effect of substrates on the transition state of the ATP hydrolysis reaction. TheK m for [α-32P]8-azidoATP (20 μm) is decreased in the presence of Vi; however, it is not changed by drugs such as verapamil or cyclosporin A. Strikingly, the extent of Vi-induced [α-32P]8-azidoADP trapping correlates directly with the fold stimulation of ATPase activity at steady state. Furthermore, Pi exhibits very low affinity for Pgp (K i ∼30 mm for Vi-induced 8-azidoADP trapping). In aggregate, these data demonstrate that the release of Vi trapped [α-32P]8-azidoADP from Pgp is the rate-limiting step in the steady-state reaction. We suggest that substrates modulate the rate of ATPase activity of Pgp by controlling the rate of dissociation of ADP following ATP hydrolysis and that ADP release is the rate-limiting step in the normal catalytic cycle of Pgp.

often associated with the overexpression of the human multidrug resistance gene (MDR1) (1,2). MDR1 encodes a 170-kDa plasma membrane protein, P-glycoprotein (Pgp), 1 that uses the energy from ATP hydrolysis to expel a variety of anticancer drugs from cells, thus making them ineffective during chemotherapy. The secondary structure of Pgp is predicted to consist of two homologous halves each containing six putative transmembrane helices and a nucleotide-binding domain. These structural elements are common to a large family of membrane transporters called the ATP-binding cassette superfamily (3,4).
The widely accepted hypothesis of Pgp function is that substrate transport is coupled to ATP hydrolysis. However, the mechanism for this reaction is not well understood. In the absence of added substrate, Pgp catalyzes basal ATP hydrolysis (ATPase activity). This basal activity has been suggested to occur because of an endogenous lipid substrate(s) or an uncoupling of ATP hydrolysis and drug extrusion (5,6). The basal rate is stimulated by adding any one of a variety of hydrophobic drug substrates; these drugs bind with unique apparent affinities and affect the ATPase activity of Pgp to varying degrees (7)(8)(9). However, a detailed assessment of the kinetic parameters of ATP hydrolysis in the presence of amphipathic drugs for the identification of the rate-limiting step in the catalytic cycle has not been carried out.
Expression, purification, and reconstitution procedures of endogenous and six histidine-tagged (His 6 ) human Pgp in a heterologous expression system are well described (10 -12). Studies on the ATPase activity of Pgp in crude membrane preparations have indicated various drug-stimulated effects on the ATPase activity of Pgp (5); such experiments with crude protein can be difficult to interpret. Clearly, when purifying Pgp, the lipid environment of this membrane protein affects the ATPase activity and should remain constant for comparison of drug-stimulated activities (13)(14)(15)(16).
Kinetic schemes for the catalytic cycle of Pgp have been proposed based on binding studies and Vi-induced ADP trapping experiments (17)(18)(19)(20)(21). Vi is a proposed transition state analog that replaces P i immediately after its release upon ATP hydrolysis. No phospho-enzyme intermediate of Pgp has been identified in the catalytic cycle of Pgp, indicating that all intermediates in the Pgp reaction are noncovalent (18). Senior and his colleagues (18,19) have extensively characterized the Vi-induced ADP trapping reaction and implied that P i release precedes ADP release and ADP release is the most likely rate-limiting step in catalysis (22). Our objective in this study is to understand how Pgp drug substrates might affect the overall kinetic mechanism of Pgp. Kinetic constants for various drugs that affect the steady-state ATPase activity of Pgp are quantified, and the extent of Vi-induced ADP trapping is compared in the presence and absence of various Pgp drug substrates. The correlation of these values has mechanistic implications. Our results suggest that ADP release is a rate-limiting step in the catalytic cycle of Pgp, and substrates and modulators exert their effect on ATPase activity by modulating this step.
Expression and Purification of Pgp-Human Pgp was expressed and purified as previously described (11) with minor changes. Briefly, His 6tagged Pgp was expressed using Trichoplusia ni (High Five TM ; Invitrogen, San Diego, CA) insect cells grown in monolayer cultures and infected with a recombinant baculovirus BV-MDR1(H 6 ) that contains a human MDR1 that encodes Pgp. Crude Pgp-containing membranes were prepared by Dounce homogenization under hypotonic conditions (23), and membrane proteins were solubilized with octyl ␤-D-glucopyranoside (1.25%) in the presence of 20% glycerol and lipid mixture (0.1%) (see "Routine Procedures"). Solubilized proteins were subjected to metal affinity chromatography (Talon resin from CLONTECH, Palo Alto, CA) in the presence of 0.95% octyl ␤-D-glucopyranoside and 0.04% lipid; 80% purified Pgp was eluted with 100 mM imidazole (11). Pgp in 100 mM imidazole fraction was then concentrated (Centriprep-50, Amicon, Beverly, MA) to ϳ0.5 mg/ml and stored at Ϫ70°C. Pgp was identified by immunoblot analysis (11) and quantitated by Amido Black protein estimation method as previously described (24).
ATPase Assays-ATPase activity of purified, reconstituted Pgp was measured by two methods; the end point P i assay and the continuous cycling assay. In both assays, Pgp-specific activity was recorded as the Vi (0.3 mM)-sensitive ATPase activity. Various test agents were added from 100ϫ stock solutions in Me 2 SO so that the Me 2 SO concentration was no greater than 1%; this concentration of Me 2 SO had no effect on the activity of Pgp or the cycling assay components. For the P i assay, the amount of inorganic phosphate released over 20 min at 37°C was measured. ATPase assay buffer (50 mM MES-Tris, pH 6.8, 50 mM N-methyl-D-glucamine chloride, 5 mM NaN 3 , 1 mM EGTA, 1 mM ouabain, and 2 mM dithiothreitol) was combined with either 5 mM MgCl 2 or CoCl 2 , 0.5-2 g of purified, reconstituted Pgp, and various Pgp drug substrates for a 5-min preincubation at 37°C. The reaction was initiated by the addition of 5-7.5 mM ATP and quenched with SDS (final concentration, 2.5%); the amount of P i released was quantitated using a colorimetric method as previously described (11). For the cycling assay, cycling components (3 mM phosphoenolpyruvate, 0.33 mM NADH, and 10 units/ml both pyruvate kinase and lactate dehydrogenase) were added to the ATPase assay buffer (described above) with 10 -15 mM MgCl 2 to link the hydrolysis of ATP directly with the oxidation of NADH (12). Purified, reconstituted Pgp (1-10 g in 100 l of assay volume) was preincubated with various Pgp drug substrates at 37°C in a temperature-controlled, 96-well plate spectrophotometer (Spectra MAX 250; Molecular Devices, Sunnyvale, CA). The reaction was initiated by the addition of ATP and monitored at OD 340 nm using SoftMax Pro 2.4 software (Molecular Devices) for 5-10 min. The rate of change in absorbance was converted to nmol NADH oxidized per minute using an NADH standard curve; this value is equivalent to nmol ATP hydrolyzed per minute.
Kinetic Analysis-The drug-stimulated ATPase activities in the presence of saturating ATP concentrations (5-7.5 mM) and various Pgp drug substrates were fit to Equation 1.
/E is the ATPase activity at given concentrations of substrates, V is V max Ϫ V b with V max as the maximal activity, S is the drug substrate concentration, K app is the apparent concentration at half-maximal activity, and V b is the basal rate in the absence of added drug. The fold stimulation of ATPase activity by a given substrate is calculated by V max /V b . Michaelis-Menten parameters were determined for ATP and 8-azidoATP in the presence of saturating Pgp drug substrates, verapamil (50 M) and cyclosporin A (10 M); for basal activity (in the absence of added drug) an equivalent volume of Me 2 SO was added. ATPase activities using various nucleotide concentrations were fit to Equation 2.
[NTP] is the concentration of ATP or 8-azidoATP and K m is the concentration of NTP at half-maximal activity. Inhibition constants for ADP were determined using various concentrations of ADP and ATP in the presence and absence of 50 M verapamil. ATPase activities were fit to Equation 3 for competitive inhibition.
K i is the concentration of ADP at half-maximal inhibition. All curve fits in kinetic analyses were performed using Prism 2.0 software for Power-Mac (GraphPad, San Diego, CA).
Determination of the K m for 8-AzidoATP during Vi-induced Trapping-Purified Pgp was reconstituted into proteoliposomes and incubated in the ATPase assay buffer containing 0.3 mM Vi and increasing concentrations (1-75 M) of [␣-32 P]8-azidoATP (3-5 Ci/nmol) in the dark at 37°C for 3 min. The reaction was stopped by addition of 12.5 mM ice-cold ATP and placing the sample on ice. Trapping of Pgp into the Pgp⅐Mg8-azidoADP⅐Vi conformation was carried out under basal conditions and in the presence of either 50 M verapamil or 10 M cyclosporin A. Following SDS-PAGE on an 8% Tris-glycine gel, the radioactivity in the Pgp bands was quantified on a STORM 860 PhosphorImager system. The K m values for 8-azidoATP under basal conditions and in the presence of saturating concentrations of verapamil and cyclosporin A were obtained by fitting the data to Equation 2 using the software Prism as described above.
Inhibition of Vi-induced 8-AzidoADP Trapping by P i -Proteoliposomes containing purified Pgp (5-10 g) were incubated in the ATPase assay buffer containing 0.30 mM Vi. The proteoliposomes were then treated with Me 2 SO (control), 50 M verapamil, or 10 M cyclosporin A either in the presence of 0 -150 mM KH 2 PO 4 (pH 6.8) or KCl. Finally, 50 M [␣-32 P]8-azidoATP (2.5-5 Ci/nmol) was added in the dark and incubated at 37°C for 5 min. The reaction was stopped by quenching with 12.5 mM ice-cold ATP solution and placing on ice. Following SDS-PAGE on an 8% Tris-glycine gel, the extent of trapping of 8-azidoADP was determined as described above.
Binding of [␣-32 P]8-AzidoATP to Pgp-Proteoliposomes (5-10 g of protein) were incubated in the ATPase assay buffer in the presence of Me 2 SO, 50 M verapamil, or 10 M cyclosporin A for 5 min at 37°C and transferred to ice. After 5 min on ice, 10 M [␣-32 P]8-azidoATP (5-10 Ci/nmol) was added to each sample in the dark and incubated at 4°C for 5 min. The samples were then irradiated with UV light (365 nm) on ice (4°C) as described above. Ice-cold ATP (12.5 mM) was added to displace excess noncovalently bound [␣-32 P]8-azidoATP. Excess nucleotides were removed by centrifugation at 300,000 ϫ g at 4°C for 10 min by using S120-AT2 rotor in a RC-M120EX micro-ultracentrifuge (Sorvall, Newtown, CT) and the pellet resuspended in 1ϫ SDS-PAGE sample buffer. Following SDS-PAGE on a 8% Tris-glycine gel at constant voltage, gels were dried and exposed to Bio-Max MR film (Eastman Kodak Co.) at Ϫ70°C for 12-24 h. The radioactivity incorporated into the Pgp band was quantified using the STORM 860 PhosphorImager and the software ImageQuaNT. Routine Procedures-The lipid mixture was made by combining acetone/ether washed E. coli bulk phospholipids, phosphotidylcholine, phosphatidylserine, and cholesterol (Avanti Polar Lipids, Alabaster, AL) in the ratio of 60:17.5:10:12.5 by weight and evaporating to dryness under N 2 gas (26). This dried stock was stored at Ϫ70°C until resuspension at 50 mg/ml in a 2 mM ␤-mercaptoethanol solution. SDS-PAGE was performed using precast 8% Tris-glycine gels (Novex, San Diego, CA). Immunoblot analyses were performed as previously described (27) using the monoclonal antibody, C219 (a gift from Centocor). The Colloidal Blue Staining Kit (Novex, San Diego, CA) was used for total protein staining of SDS-PAGE samples. Sodium orthovanadate (Sigma) was prepared by boiling a 50 mM solution in water for 3 min, and concentration was determined by using molar absorbance ( 268 nm ϭ 3600 M Ϫ1 ).

RESULTS
Production of Pure, Catalytically Active Human Pgp-The previously described method was adopted for the purification of large amounts of pure Pgp (11). From ϳ4 ϫ 10 9 cells of baculovirus-infected High Five insect cells in suspension culture, 500 -600 mg of crude membrane protein was routinely obtained. 60 -70% of this crude membrane protein was recovered after solubilization with octylglucoside, and 4 -6 mg of purified Pgp was collected following metal affinity chromatography. The reconstitution of purified Pgp into an artificial lipid bilayer, which is required to obtain a Pgp preparation totally free of the detergent, typically yielded 65% recovery, a value common among dialysis reconstitution procedures for Pgp (12,26,28).
Measuring the ATPase Activity of Pgp-The ATPase activity of purified, reconstituted Pgp was routinely measured using two techniques: 1) an end point, P i release assay measuring the amount of P i released by the hydrolysis of ATP in a given amount of time and 2) a continuous cycling assay linking the hydrolysis of ATP to the oxidation of NADH using pyruvate kinase, lactate dehydrogenase, and their substrates. The cycling assay was preferred to the end point P i assay because of its ability to monitor the ATPase reaction continuously in real time; in this way, the rates of ATPase activity were easily identified as steady-state, linear rates. Although the cycling assay was used to quantitate the kinetics of ATP and all Pgp drug substrates, some drawbacks were inherent in the method. As the cycling assay measures decrease in the absorbance at 340 nm resulting from the oxidation of NADH, activity measurements using 8-azidoATP, a light-sensitive substrate, were impossible and, therefore, quantitated using the P i assay. Furthermore, the ATPase activity could not be monitored in the presence of ADP because of its stimulatory effect of cycling in the absence of ATP hydrolysis.
A typical ATPase cycling experiment is shown in Fig. 1, where the rate (mOD 340 /sec) of ATPase activity specific to Pgp is determined as the Vi-sensitive activity in the presence or absence of verapamil. The rate of change of absorbance per second is converted to nmol NADH oxidized⅐min Ϫ1 using an NADH standard curve. This value is directly comparable with the ATPase activity of Pgp recorded as nmol ATP hydro-lyzed⅐min Ϫ1 ⅐mg Ϫ1 . Typically, verapamil-stimulated ATPase activity of purified and reconstituted Pgp was 0.6 -1.2 mol ATP hydrolyzed⅐min Ϫ1 ⅐mg Ϫ1 , which is consistent with the values obtained using the P i assay. This specific activity is slightly lower than previously reported for the human Pgp reconstituted by using the rapid dilution procedure (11).
Steady-state Kinetics with Various Pgp Drug Substrates-To characterize how human Pgp uses different drugs to stimulate ATPase activity in a concentration-dependent manner, several Pgp substrates were assayed over large concentration ranges. A true K m value cannot be obtained because of the ATPase activity in the absence of added drug substrates (the basal rate, fold stimulation ϭ 1.0). A similar parameter, K app , is used to describe the concentration of progesterone, prazosin, tetraphenylphosphonium chloride ion, valinomycin, verapamil, cyclosporin A, PSC833 (a cyclosporin A analog), and rapamycin at half-maximal ATPase activity (Table I). These drugs either increased (fold stimulation Ͼ 1.0) or decreased (fold stimulation Ͻ 1.0) the ATPase activity of Pgp. The values for K app (often referred to as K m or K 0.5 ) and fold stimulation vary greatly throughout the literature. Much of the published data using either human or Chinese hamster Pgp is consistent with the values obtained in this study (6, 11, 20, 28 -30). However, significant differences in progesterone stimulation (30 -34) and varied data using verapamil stimulation (6, 7, 11, 28, 30 -32,

TABLE I Effect of selected drug substrates on the ATPase activity of Pgp
Pgp hydrolyzes ATP in the absence of drug substrates at a constant basal rate (fold stimulation ϭ 1.0). K app represents the drug concentration at half-maximal stimulation of ATPase activity; fold stimulation represents the ratio of the drug-stimulated V max to the basal V max . All assays were performed at saturating Mg⅐ATP concentrations (5-7.5 mM) using the cycling assay at 37°C to monitor the ATPase activity as described under "Experimental Procedures." NA, not applicable; TPP ϩ , tetraphenylphosphonium chloride.  34 -39) should be noted. In this study, many values for K app are given as upper or lower limits because of concentration limitations. For these experiments, the drug concentrations were at least in 5-fold excess of the amount of Pgp (nmol:nmol) and were considered to be saturated if a 10-fold increase in drug concentration did not affect the ATPase activity. Fig. 2A demonstrates both increased and decreased ATPase activity in the presence of selected agents. In the presence of saturating Mg⅐ATP and increasing concentrations of either prazosin or verapamil, the ATPase activity increases to a saturable V max value. The value of K app for verapamil (4.7 M) is less than for prazosin (16 M), indicating a higher affinity for verapamil, whereas the fold stimulation is greater for prazosin (4.5-fold) with respect to verapamil (2.0-fold); similar values have been previously reported (6, 11, 28 -32, 34 -36, 39). Interestingly, drugs typically considered Pgp inhibitors, like cyclosporin A, its analog PSC833, and rapamycin, are clearly depicted as substrates (Table I and data with cyclosporin A is given in Fig. 2A, inset). Upon saturation, these drugs support ATPase activity by Pgp; however, the rate is less than the basal rate in contrast to a previous report (11). Unfortunately, values of K app for these drugs are not able to be quantitated because of the high concentrations (0.4 M) of Pgp required to assay for activity in either the cycling or P i assay. However, data describing cyclosporin A as an inhibitor report low concentrations (74 -400 nM) for maximal effect (6,33,40,41) so the saturation of these compounds is likely.
Effect of Selected Drug Substrates on the Affinity of Pgp for Nucleotides-Pgp substrates, like verapamil and cyclosporin A, affect the rate at which Pgp hydrolyzes ATP; this effect is often described as fold stimulation with respect to the basal rate of ATPase activity. However, these drugs do not change the ability of Pgp to utilize nucleotides. Fig. 2B depicts the differences in maximal ATPase activity in the absence (basal) and in the presence of verapamil or cyclosporin A and suggests the similarities in values of K m , the concentration required for halfmaximal stimulation. Table II reports the values of V max and K m for ATP (0.22, 0.33, and 0.26 mM) and 8-azidoATP (0.4, 0.6, and 0.5 mM) in the presence of verapamil, no drug, and cyclosporin A, respectively. The K m values for ATP are similar to those reported previously by us (8,11). Clearly, the values of K m for either nucleotide are unchanged in the absence (basal) or presence of verapamil or cyclosporin A, whereas the values of V max remain constant in relation to each other (fold stimulation for verapamil:basal:cyclosporin A ϭ 2:1:0.5). Additionally, the values of K i for ADP are similar in the presence (K i ϭ 0.5 mM) and absence of verapamil (K i ϭ 0.3 mM); these data display competitive inhibition patterns with respect to ATP by intersecting on the y axis of a Lineweaver-Burke plot (data not shown), indicating that ATP and ADP bind to the same site. Other workers also have reported similar inhibition by ADP (11,28,39,42).   Fig. 3A is not due to unequal loading of protein per lane in the gel (see the immunoblot in Fig. 3B; samples in lanes 2-8 in Fig. 3A were used for the immunodetection of Pgp with the monoclonal antibody, C219). Moreover, previous reports (43,44) have also demonstrated drug substrate-dependent variation in the extent of [␣-32 P]8-azido ADP trapping into Pgp. [␣-32 P]8-azidoADP is incorporated into 2-5% of the Pgp in the reaction depending on the Pgp drug substrates present; these values are consistent with those previously reported for Chinese hamster Pgp in the presence of verapamil (45). It should be noted that cross-linking is highly specific using purified, reconstituted Pgp; the total protein profile of the proteoliposomes used in these studies is also shown in Fig. 3 (Fig. 3A). To quantitatively compare the effect of substrates on the ATPase activity and the extent of trapping, it is necessary to obtain the kinetic parameters for 8-azidoADP trapping into Pgp. Generating the Pgp⅐[␣- 32 Table II). How Vi alters the affinity of nucleotides for Pgp is not clear at present.

Vanadate-induced [␣-32 P]8-AzidoADP Trapping in the Presence of Various Pgp Drug Substrates-Under
As previously reported (19,25), the Pgp⅐Mg⅐8-azidoADP⅐Vi complex, trapped using Mg⅐[␣-32 P]8-azidoATP, dissociates rather quickly (t1 ⁄2 ϭ 8 min at 37°C). However, replacing MgCl 2 with CoCl 2 displays similar reaction kinetics within the first 2 min, whereas no dissociation of the Pgp⅐Co⅐8-azidoADP⅐Vi complex is seen up to 20 min (data not shown). Although Co 2ϩ supports steady-state Pgp ATPase activity at rates 10 -20-fold slower than Mg 2ϩ (19), 2  strongly correlates with the fold stimulation of ATPase activity by those drugs in the steady-state reaction with saturating Mg 2ϩ ⅐ATP concentrations ( Fig. 5; the data used for this analysis are given in Fig. 3A and Table I, respectively). This correlation is also seen using higher [␣-32 P]8-azidoATP concentrations (up to 500 M) for Vi-induced trapping. In addition when CoCl 2 was replaced with 5 mM MgCl 2 , a similar correlation between Vi-induced [␣-32 P]8-azidoADP trapping, and the fold stimulation of ATPase activity was also observed (data not shown).
The experiments described above clearly demonstrate that drugs, which are substrates of Pgp affect both ATP hydrolysis and the Vi-induced trapping of [␣-32 P]8-azidoATP. This raises the question as to whether the substrates influence nucleotide binding or a subsequent step during hydrolysis. We addressed this issue directly by allowing [␣-32 P]8-azidoATP to bind Pgp under nonhydrolysis conditions (at 4°C) both in the absence and presence of verapamil or cyclosporin A. We observed that even saturating concentrations of verapamil and cyclosporin A do not affect nucleotide binding per se, suggesting that the substrate must act at a step or steps that follow binding of nucleotide (56).

DISCUSSION
For kinetic studies of a membrane transport protein, such as Pgp, large quantities of purified protein that retains complete biological activity are necessary. The baculovirus expression system used in this study produces human Pgp at a high level in the absence of cytotoxic drug selection, and the absence of such selection pressure facilitates the assessment of the intrinsic properties of human Pgp. Because Pgp is an integral membrane protein, lipids play a key role in protein conformation and activity (13)(14)(15)(16). Solubilization and purification in the presence of octyl glucoside and reconstitution with lipid mixture containing bulk E. coli phospholipids, phosphatidylcholine, phosphatidylserine, and cholesterol provides a system for the complete recovery of the ATPase activity of purified Pgp (8,11,26).
The ATPase activity of Pgp in the presence of various Pgp drug substrates can be accurately measured using purified, reconstituted Pgp in the cycling assay. In contrast, crude mem-brane preparations produce high background activities both in the absence of ATP because of NADH oxidases and in the presence of ATP and Vi because of the presence of Vi-insensitive ATPases. Because the cycling assay relies on the catalysis of other enzymes, we demonstrated that compounds in the assay buffer, such as Vi or Pgp drug substrates, do not affect the activity of these enzymes. Furthermore, the slow catalysis of Pgp (ϳ2-10 s Ϫ1 ) with respect to cycling enzymes, pyruvate kinase, and lactate dehydrogenase is crucial to retain the tight coupling from ATP hydrolysis to NADH oxidation.
ATPase activity of Pgp is measured as the Vi-sensitive activity in the presence and absence of various Pgp substrates. The observed variability in the specific activity of Pgp upon reconstitution is likely due to preparation impurities and/or efficiencies of reconstitution; because of these variations, the ATPase activity of Pgp in the presence of drug substrates is typically considered using fold stimulation values that remain constant. The actual concentrations at which Pgp drug substrates stimulate ATP activity are unclear because all Pgp drug substrates are hydrophobic and partition into the membrane lipid surrounding Pgp at unknown local concentrations. However, characterizing drug-stimulated activities in this purified, reconstituted system is useful for comparisons among various drugs.
The determination of Michaelis-Menten parameters for various Pgp drug substrates is challenging. High amounts of Pgp are required to quantify the enzyme activity in the presence of the drug, and this amount of Pgp sets a lower limit on the amount of drug to be used for stimulation. Hence, many values for K app , as shown in Table I The proposed scheme is similar to those previously described (20,22,29) and shows the rate-limiting step (step 4) in the catalytic cycle based on the work reported here.
Step 1, drug substrate and ATP bind to Pgp. There is no evidence that the binding of one is either a prerequisite or inhibitory to the binding of the other.
Step 2, binding of drug and ATP is followed by ATP hydrolysis, and this is accompanied by a conformational change resulting in the translocation of the drug from a high affinity (on) to a low affinity (off) site.
Step 3, following hydrolysis of ATP, both P i and drug are released, although the exact order of release is not known at present.
Step 4, the ADP release, which appears to be the slowest step in the cycle (see below), is essential for regenerating Pgp for the next hydrolysis event (the box is shaded to indicate the rate-limiting nature of this step).
Step 5, if Vi is provided to the system, Vi mimics P i to trap ADP in a stable ternary conformation (Pgp⅐ADP⅐Vi). Moreover, given the chemical analogy between P i and Vi, the general consensus is that the Pgp⅐ADP⅐P i and the Pgp⅐ADP⅐Vi complexes are equivalent and that this transition state represents an intermediate state during the normal reaction pathway (46). Step 6, eventually, Vi and ADP dissociate from Pgp (t1 ⁄2 ϭ 80 -90 min at 37°C (18)) to initiate the next cycle. The observed correlation between the extent of trapped 8-azidoADP in the presence of Vi and fold stimulation of ATP hydrolysis by various substrates (see Fig. 5), and the fact that P i exhibits very low affinity for Pgp (K i ϳ30 mM for Vi-induced 8-azidoADP trapping in human Pgp 2 and K i ϳ200 mM for ATP hydrolysis by Chinese hamster Pgp (19) strongly suggest that 8-azidoADP (ADP) release (step 4) is the rate-limiting step in the catalytic cycle and substrates modulate the Pgp activity by exerting effect on this step. Although not shown, both ATP and ADP are complexed with Mg 2ϩ , which has been omitted for clarity. highly hydrophobic with low solubility limits making saturation at high drug concentrations problematic. For these drugs, both K app and drug-saturated ATPase activities are harder to define. Of particular interest is the ability of cyclosporin A and other Pgp "inhibitors" to support ATPase activity. It is clear from the data presented here that cyclosporin A, its analog PSC833, and rapamycin act as Pgp substrates, albeit at a slower ATPase rate than basal levels ( Fig. 2A and Table II). The idea that these drugs act as substrates promoting ATPase activity is supported by several cyclosporin A transport studies (47)(48)(49)(50). Additionally, human Pgp-specific monoclonal antibody, UIC2 shift assays suggest similar conformational changes with vinblastine, a common Pgp substrate, as well as cyclosporin A (51). These studies also demonstrate that the cycling assay provides a useful tool to assess whether a given modulator is a "true" inhibitor or a substrate of Pgp.
It is clear that the ATPase activity of Pgp is stimulated by a myriad of drug substrates. The number and interactions of the drug binding site(s) of Pgp has been discussed extensively in the literature (29, 31-34, 52, 53). Although these drugs have profound effects on the overall stimulation of the ATPase activity of Pgp, they have no effect on the values of K m for nucleotide substrates, ATP and 8-azidoATP ( Fig. 4 and Table  II). Furthermore, the value of K i for ADP, a competitive inhibitor of ATP, remains constant in the presence of verapamil (Table II). Taken together, these data suggest that drug substrates, like verapamil and cyclosporin A, have no effect on ATP binding to Pgp; these conclusions are comparable with those suggested using mouse Pgp (mdr1b) (20).
The proposed scheme for the catalytic cycle of Pgp and the rate-limiting step is given in Fig. 6 (see the figure legend for the detailed description). To investigate other steps of the Pgp catalytic cycle where drug might have an effect, we used Viinduced ADP trapping experiments, a method commonly used to investigate the transition state(s) of ATP-hydrolyzing proteins without covalent phosphoenzyme intermediates. Ortho-Vi can mimic P i and bind in its place immediately after ATP hydrolysis and P i release to trap ADP and inhibit the steady-state turnover of the enzyme (Fig. 6, step 5) (18). Therefore, the amount of ADP trapped on Pgp is indicative of the Vi-inhibited conformation, which is comparable with the transition state conformation Pgp⅐ADP⅐P i of the ATP hydrolysis reaction. In this study, the substitution of CoCl 2 for MgCl 2 avoided complications with dissociation of the Pgp⅐8-azidoADP⅐Vi trapped complex while supporting apparently the same rate of reaction for the first 2 min of trapping (data not shown).
Although Vi inhibits all Pgp activity in steady-state experiments independent of the presence of drug substrates, different amounts of 8-azidoADP are trapped on the protein when the extent of Vi-induced trapping is complete. A striking correlation exists between the extent of Vi trapping and the steadystate fold stimulation of ATPase activity in the presence of various drug substrates. Drugs that support a higher fold stimulation of steady-state ATPase activity also demonstrate a higher extent of 8-azidoADP (or ADP) trapping; the opposite holds true for less active Pgp drug substrates (Fig. 5). A previous report (44) also suggested a close relationship between the extent of Vi-induced nucleotide trapping and the drug substrate-stimulated ATPase activity of Pgp. Directly correlating effects on the Vi-induced ADP trapped conformation and the steady-state reaction rates in the presence of Pgp drug substrates indicate that these experiments are measuring the same step in the Pgp catalytic cycle. The Vi-induced ADP trapping experiment measures the amount of ADP released (the inverse of the amount of ADP trapped) from the transition state conformation, which, by its correlation, suggests that the rate-limiting step measured in the steady-state reaction is the release of ADP (step 4 or 6 in Fig. 6).
These experiments further indicate that the release of P i is not likely to be the rate-limiting step of the overall catalytic cycle. The low affinity of P i for Pgp makes this step an unlikely candidate to act as the rate-limiting step of the Pgp reaction cycle (K i ϳ30 mM for Vi-induced 8-azidoADP trapping in Pgp, 2 and K i ϳ200 mM for ATP hydrolysis (19)). Additionally, the correlation between the steady-state reaction (fold stimulation) and the amount of ADP trapping in the presence of drug substrates like verapamil and cyclosporin A indicates that the rate-limiting step is a step after Vi binds and traps Pgp, and P i release is a prerequisite for Vi binding (Fig. 6, steps 3 and 5).
The ADP release being the rate-limiting step in the catalytic cycle is further supported by our recent observation that there is an inverse relationship between ADP release from the Pgp⅐MgADP⅐Vi complex and the recovery of the substrate binding to the transporter following the transition state step (25). In addition, the rate of the release of 8-azidoADP (or ADP) from the Vi-trapped Pgp is not affected by the addition of excess nucleotides such as ATP, ADP, or AMPPNP (56). It is now clear that most ATP-binding cassette transporters catalyze Vi-sensitive ATP hydrolysis, which is stimulated by substrates. Viinduced 8-azidoADP trapping has been demonstrated in other transporters such as MRP1 and ATP-binding cassette R (54,55). It is perhaps most likely that the ADP release is a ratelimiting step in the catalytic cycle of other members of the super family of ATP-binding cassette transporters.