Functionally similar vanadate-induced 8-azidoadenosine 5'-[alpha-(32)P]Diphosphate-trapped transition state intermediates of human P-glycoprotin are generated in the absence and presence of ATP hydrolysis.

P-glycoprotein (Pgp) is an ATP-dependent drug efflux pump whose overexpression confers multidrug resistance to cancer cells. Pgp exhibits a robust drug substrate-stimulable ATPase activity, and vanadate (Vi) blocks this activity effectively by trapping Pgp nucleotide in a non-covalent stable transition state conformation. In this study we compare Vi-induced [alpha-(32)P]8-azido-ADP trapping into Pgp in the presence of [alpha-(32)P]8-azido-ATP (with ATP hydrolysis) or [alpha-(32)P]8-azido-ADP (without ATP hydrolysis). Vi mimics P(i) to trap the nucleotide tenaciously in the Pgp.[alpha-(32)P]8-azido-ADP.Vi conformation in either condition. Thus, by using [alpha-(32)P]8-azido-ADP we show that the Vi-induced transition state of Pgp can be generated even in the absence of ATP hydrolysis. Furthermore, half-maximal trapping of nucleotide into Pgp in the presence of Vi occurs at similar concentrations of [alpha-(32)P]8-azido-ATP or [alpha-(32)P]8-azido-ADP. The trapped [alpha-(32)P]8-azido-ADP is almost equally distributed between the N- and the C-terminal ATP sites of Pgp in both conditions. Additionally, point mutations in the Walker B domain of either the N- (D555N) or C (D1200N)-terminal ATP sites that arrest ATP hydrolysis and Vi-induced trapping also show abrogation of [alpha-(32)P]8-azido-ADP trapping into Pgp in the absence of hydrolysis. These data suggest that both ATP sites are dependent on each other for function and that each site exhibits similar affinity for 8-azido-ATP (ATP) or 8-azido-ADP (ADP). Similarly, Pgp in the transition state conformation generated with either ADP or ATP exhibits drastically reduced affinity for the binding of analogues of drug substrate ([(125)I]iodoarylazidoprazosin) as well as nucleotide (2'(3')-O-(2,4,6-trinitrophenyl)adenosine 5'-triphosphate). Analyses of Arrhenius plots show that trapping of Pgp with [alpha-(32)P]8-azido-ADP (in the absence of hydrolysis) displays an approximately 2.5-fold higher energy of activation (152 kJ/mol) compared with that observed when the transition state intermediate is generated through hydrolysis of [alpha-(32)P]8-azido-ATP (62 kJ/mol). In aggregate, these results demonstrate that the Pgp.[alpha-(32)P]8-azido-ADP (or ADP).Vi transition state complexes generated either in the absence of or accompanying [alpha-(32)P]8-azido-ATP hydrolysis are functionally indistinguishable.


P]8-azido-ADP (or ADP)⅐Vi transition state complexes generated either in the absence of or accompanying [␣-32 P]8-azido-ATP hydrolysis are functionally indistinguishable.
Cancer cells resistant to chemically diverse drugs with multiple mechanisms of action are defined as exhibiting the multiple drug resistance (MDR) 1 phenotype. The best defined form of MDR in human cells is due to the overexpression of Pglycoprotein (Pgp). This 170-kDa plasma membrane protein is a member of the ATP-binding cassette (ABC) superfamily of transport proteins, and can extrude a range of hydrophobic anticancer drugs from cells against a concentration gradient and thus render cells resistant to cytotoxic chemotherapeutic agents. Analysis of hydropathy plots suggests that Pgp consists of two homologous halves each containing six transmembrane helices and one nucleotide-binding or ATP site in each half (1,2).
The extrusion of cytotoxic agents is powered by ATP hydrolysis, and the ATPase activity of Pgp has been studied in considerable detail. [␣-32 P]8-azido-ATP, a radiolabeled, photoaffinity analogue of ATP, has proved to be a valuable reagent in understanding interactions between nucleotide and Pgp (3)(4)(5)(6)(7). The use of [␣-32 P]8-azido-ATP along with orthovanadate (Vi) has permitted experimental strategies to elucidate the catalytic cycle (3,5,(7)(8)(9)(10)(11)(12). Vi inhibits Pgp ATPase activity by trapping nucleotide in the catalytic site to generate the transition state conformation, Pgp⅐ADP⅐Vi. It has been established that it is always a nucleoside diphosphate that is the trapped species (3,13). Thus if ATP (or 8-azido-ATP) is used to initiate the reaction, at least one turnover of ATP hydrolysis, converting ATP to ADP, is essential for trapping to occur. This has allowed Vi-trapping experiments to be used to construct a catalytic scheme for ATP hydrolysis by Pgp.
Our recent work (10 -12) has considerably expanded the original model for the catalytic scheme of Pgp proposed by Senior and co-workers (14). Our data suggest that ATP hydrolysis at one of the two ATP sites results in a dramatic conformational change where the affinities of both the substrate and the nucleotide for Pgp are drastically reduced. The fact that ATP binding to the second site is arrested while the first one is in a catalytic conformation appears to be the basis for alternate catalysis in Pgp (12). Moreover, for Pgp to regain the conformation that binds substrate with high affinity, the hydrolysis of an additional molecule of nucleotide is obligatory. Finally, we showed that release of ADP from the Pgp⅐ADP⅐P i transition state is the rate-limiting step in the catalytic cycle of ATP hydrolysis (11).
Much of the work described above has benefited from the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  availability of [␣-32 P]8-azido-ATP allowing direct visualization and quantification of the nucleotide interaction with Pgp. However, as only the photoaffinity analogue of nucleoside triphosphate was available as a radioisotope, this has precluded directly addressing many questions that require the 32 Plabeled azido derivative of nucleoside diphosphate. In this study, we have characterized for the first time the binding and Vi-induced trapping of [␣-32 P]8-azido-ADP to Pgp. We demonstrate that [␣-32 P]8-azido-ADP binds specifically to Pgp with a K d comparable to that for [␣-32 P]8-azido-ATP. The kinetic scheme for the Vi-induced inhibition of Pgp ATP hydrolysis proposed by Senior's group (14,15) suggests that incorporation of ADP into the Pgp⅐ADP⅐Vi ternary complex may occur either following hydrolysis of ATP to ADP or directly by the addition of ADP in the absence of hydrolysis.
We demonstrate in this study that it is possible to initiate Vi-induced trapping with either [␣-32 P]8-azido-ADP or [␣-32 P]8-azido-ATP, with similar kinetics, and that the trapped [␣-32 P]8-azido-ADP distributed equally between the N-and the C-terminal halves of Pgp under both hydrolysis and nonhydrolysis conditions. Vi-induced trapping under both hydrolysis and non-hydrolysis conditions exhibited the same requirement for divalent cations. Previous work has demonstrated that a point mutation in the Walker B region (D555N and D1200N, respectively) of either the N-or C-terminal ATP sites arrests ATP hydrolysis and Vi-induced trapping (6,9). We find that these mutants also do not show Vi-induced [␣- 32 6 tag at the C-terminal end (BV-MDR1 (His 6 )) as described (5). Crude membranes were prepared as described previously (5,16).
Purification and Reconstitution of Pgp-Human Pgp from crude membranes of High Five insect cells was purified as described (5). The crude membranes were solubilized with octyl ␤-D-glucopyranoside (1.25%) in the presence of 20% glycerol and a lipid mixture (0.1%). Solubilized proteins were subjected to metal affinity chromatography (Talon resin, 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. Pgp in the 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 using the monoclonal antibody C219 (5) and quantified by Amido Black protein estimation method as described previously (18). Purified Pgp was reconstituted into proteoliposomes by dialysis as described (11).
Photoaffinity Labeling with IAAP-The crude membranes (10 -50 g) were incubated at room temperature in 50 mM Tris-HCl, pH 7.5, with IAAP (3-6 nM) for 5 min under subdued light. The samples were then illuminated with a UV lamp assembly (PGC Scientifics, Gaithersburg, MD) fitted with two Black light (self-filtering) UV-A long wave F15T8BLB tubes (365 nm) for 10 min at room temperature (21-23°C). 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 a STORM 860 PhosphorImager system (Molecular Dynamics, Sunnyvale, CA) and the software ImageQuaNT.
ATPase Assays-ATPase activity of Pgp in crude membranes was measured by the end point, P i assay as described previously (5,11), with minor modifications. Pgp-specific activity was recorded as the Vi-sensitive ATPase activity. The assay measured the amount of inorganic phosphate released over 20 -30 min at different temperatures ranging from 22 to 39°C in the ATPase assay buffer. The assay was carried out under basal conditions or in the presence of verapamil, 30 M. The reaction was initiated with 5 mM ATP or 2 mM 8-azido-ATP and quenched with SDS (2.5% final concentration); the amount of P i released was quantitated using a colorimetric method (5). The assay with 8-azido-ATP was performed under subdued lighting.
Binding of TNP-ATP to Pgp-Binding of the fluorescent ATP analogue TNP-ATP was measured by determining the increase in fluorescence signal of TNP-ATP in the presence of purified Pgp incorporated into proteoliposomes. Proteoliposomes (20 -25 g protein) were incubated with TNP-ATP, 100 M, in ATPase assay buffer in a reaction volume of 200 l for 10 min at room temperature (22-23°C) under subdued light. The samples were then transferred to individual wells in a 96-well flat bottom, UV transparent, disposable microplate (Spectraplate, Molecular Devices, Sunnyvale, CA), and an emission scan (500 -600 nm) was obtained using excitation at 408 nm on a Spectra MAX 250 Microplate Spectrofluorimeter (Molecular Devices, Sunnyvale, CA). The enhancement in fluorescence due to TNP-ATP binding was estimated by running emission scans of TNP-ATP in the assay buffer in the absence of Pgp. Additionally, to obtain signal specific to Pgp, an equivalent emission scan in the presence of 10 mM ATP (i.e. 100-fold excess) was generated and subtracted from the scan obtained in the absence of ATP.

RESULTS
Binding of [␣-32 P]8-Azido-ADP to Pgp-The radiolabeled, photoaffinity analogue of ATP, [␣-32 P]8-azido-ATP, has shown to be extremely useful in understanding the catalytic cycle of ATP hydrolysis by Pgp (3,5,12,19), the ATP binding and hydrolysis of other ABC transporters (20,21), and in mapping the active site of myosin (22). However, the unavailability of the radiolabeled nucleoside diphosphate, [␣-32 P]8-azido-ADP, until recently has precluded many direct experiments to address questions about the catalytic cycle of Pgp. In this study, we characterize the kinetics of [␣-32 P]8-azido-ADP binding to Pgp, and we demonstrate its usefulness as a reagent to study the mechanism of action of Pgp. Fig. 1A shows that [␣-32 P]8azido-ADP binds specifically to the ATP site of Pgp as the binding is completely competed by the nucleotides, ATP and ADP, as well as the nucleotide analogues, 8-azido-ATP and 8-azido-ADP. Additionally, Fig. 1B  To characterize further Pgp in the transition state conformation generated by using [␣-32 P]8-azido-ADP, we compared the distribution of the trapped [␣-32 P]8-azido-ADP in the N-and the C-terminal ATP sites of Pgp. Fig. 2C demonstrates that consistent with previously published reports (9,12,(27)(28)(29), the [␣-32 P]8-azido-ADP distributes approximately equally between the N-and C-terminal halves of Pgp, and the distribution is similar regardless of whether the occluded [␣-32 P]8-azido-ADP is generated through the hydrolysis of [␣-32 P]8-azido-ATP or directly providing the nucleoside diphosphate itself.

Effect of Divalent Cations on the Vi-induced Trapping of [␣-32 P]8-Azido-ADP Under Hydrolysis and Non-hydrolysis
Conditions-The experiments described thus far have used magnesium as a metal cofactor with the nucleotide as it is well established that ATP binds to Pgp as an MgATP complex (30). However, several divalent cations such as manganese and co- balt are known to support ATPase activity (31), although with considerably reduced V max values vis-à -vis magnesium. The V max value for MnATPase is 43% that for MgATPase and only 10% for CoATPase (3). Studies have also demonstrated that replacing magnesium with other cations such as Mn 2ϩ and Co 2ϩ also support the Vi-induced trapping of [␣-32 P]8-azido-ADP (11,27,31). Compared with the extent of Vi-induced [␣-32 P]8-azido-ADP trapping in the presence of Mg 2ϩ , Mn 2ϩ , and Co 2ϩ , the amount of [␣-32 P]8-azido-ADP incorporated in the presence of Ca 2ϩ was negligible (31). In Fig. 3

. [␣-32 P]8-Azido-ADP in the presence of Vi is trapped into a ternary complex with Pgp.
A, catalytic scheme for the Vi-induced trapping of ADP into Pgp. The kinetic scheme presented here is based on published reports (11,14,27) Step 1, MgATP binds to Pgp in the presence of Mg 2ϩ .
Step 2, binding of MgATP is followed by hydrolysis; MgATP is converted to MgADP; P i is released, and MgADP dissociates from Pgp.
Step 3, however, if Vi, an analogue of P i , is present in the reaction, MgADP is trapped to form a stable, ternary, non-covalent complex, Pgp⅐MgADP⅐Vi.
Step 4, eventually, MgADP and P i dissociate from Pgp. As Step 4 is a reversible process, in principle if MgADP were directly provided to Pgp then the Pgp⅐MgADP formed would be stabilized by Vi into the Pgp⅐MgADP⅐Vi transition state complex. This is depicted in the box at the bottom (Steps 4a and 4b) except that MgADP is replaced with Mg[␣-32 P]8-azido-ADP (see B). Note: although we depict the release of P i from the complex as concurrent with ATP hydrolysis (Step 2), it is very likely that there are several sub-states between the hydrolysis of MgATP and the release of P i . More importantly, each of these could have subtle conformational differences, and agents such as Vi, aluminum fluoride, or beryllium fluoride could trap each of these in a ternary complex. B, [␣- 32  The left and right panels are autoradiograms from the same gel; however, as the signals for lanes 3 and 4 were extremely high the gels were exposed to the x-ray film for different times. The left panel (lanes 1 and 2) were exposed to the x-ray film for 36 h at Ϫ70°C and the right panel (lanes 3 and 4) for 8 h. An equal amount of protein (96 g) was loaded in each lane. C, distribution of trapped [␣-32 P]8-azido-ADP in the N-and the C-terminal ATP sites of Pgp. Crude membranes (2 mg/ml) were incubated in the ATPase assay buffer containing 50 M [␣-32 P]8-azido-ADP (2-4 Ci/nmol) and 250 M Vi in the dark for 10 min at 37°C and cross-linked by UV irradiation (365 nm). In a parallel experiment, the crude membranes were incubated in an identical manner with 50 M [␣-32 P]8-azido-ATP (2-4 Ci/nmol) and 250 M Vi. Samples were then either treated with trypsin (protein/ trypsin, 1:10) for 5 min at 37°C to separate the N-and C-terminal halves of Pgp as described previously (8) or incubated at 37°C in the absence of trypsin. To both untreated and trypsin-treated samples, SDS-PAGE sample buffer containing 5 M urea was added. Following SDS-PAGE on an 8% Tris glycine gel at constant voltage, gels were dried and exposed to Bio-Max MR film at Ϫ70°C for 16 -24 h. Autoradiogram shows the following: The Vi-induced ADP-trapped Conformation of Pgp Exhibits a Marked Decrease in Affinity for the Fluorescent Nucleotide TNP-ATP-We have demonstrated earlier that there is a reduced binding of nucleotide to Pgp in the transition state conformation (12). This can be demonstrated by first generating the Pgp⅐ADP⅐Vi transition state by incubating with ATP and Vi, washing off excess ATP and Vi, and then determining the extent of binding of TNP-ATP, a hydrolyzable, fluorescent analogue of ATP (32), previously used to characterize the ATP sites of Pgp (33,34). Fig. 5 shows that when Pgp is pretreated with Vi and ATP or ADP at 37°C, to generate the Pgp⅐ADP⅐Vi complex, there is a marked decrease in the binding of TNP-ATP to Pgp as evidenced by decreased levels of fluorescence. These results are consistent with our earlier finding that binding of [␣-32 P]8-azido-ATP or TNP-ATP to Pgp is drastically reduced when the transporter is trapped in the transition state conformation (12). The samples were immunoprecipitated, electrophoresed, and exposed to an x-ray film. Fig. 6 demonstrates that the wild-type Pgp shows comparable Vi-induced trapping of [␣- 32  using both ATP and 8-azido-ATP that the Vi-trapped Pgp shows a greatly reduced affinity for the substrate analogue IAAP (5,8,10,12). To demonstrate that trapping the Pgp molecule in the transition state is sufficient to reduce the affinity of substrate analogue IAAP for the transporter and that this relationship is not specific to a particular nucleotide, we used both nucleoside tri-and diphosphates to trap Pgp in the transition state conformation. Crude membranes containing Pgp were incubated either with ATP, ADP, 8-azido-ATP, or 8-azido-ADP (1.25 mM) and 250 M Vi at 37°C in the dark. Aliquots were removed at intervals, incubated with IAAP, and cross-linked by UV irradiation. Following SDS-PAGE, the IAAP incorporated into the Pgp bands was quantified using a PhosphorImager. The results, depicted in Fig. 7, A-D, show that trapping of all the nucleotides tested inhibit IAAP binding in the presence of Vi. We have shown earlier that nucleotides in the absence of Vi or Vi in the absence of nucleotide do not affect IAAP binding (10). These results demonstrate the fact that in the presence of Vi and Mg 2ϩ , it is the nucleoside diphosphate that is trapped at the ATP site, and the resulting conformational changes are sufficient to effect changes in the substratebinding site, resulting in the decreased affinity for substrate. Moreover, these data indicate that the transition state conformation of Pgp generated either in the presence or absence of hydrolysis of nucleotide is similar with respect to its effect on the substrate-binding site(s).

The Inhibition of IAAP Labeling of Pgp during Vi-induced Trapping of 8-Azido-ADP and the Extent of Vi-induced Trapping of [␣-32 P]8-Azido-ADP Are Correlated-To understand
whether there is a cause-effect relationship between Vi-induced trapping of nucleoside diphosphate per se and the inhibition of substrate binding to Pgp, we performed the following two experiments in parallel. Crude membranes were incubated with 1.25 mM 8-azido-ADP and 250 M Vi. Aliquots were removed at different time intervals, treated with IAAP for 5 min, and cross-linked by UV irradiation. In a parallel experiment the crude membranes were incubated with 50 M [␣-32 P]8azido-ADP and 250 M Vi. Aliquots were removed at different time intervals, and 200-fold excess ATP was added followed by cross-linking with UV irradiation. The results of this experiment depicted in Fig. 8 show that the increased [␣-32 P]8-azido-ADP trapping over time is accompanied by decrease in IAAP binding. The inset shows that the two are inversely correlated (r ϭ 0.92) suggesting that Vi-induced trapping of [␣-32 P]8azido-ADP at the ATP site induces conformational changes that reduce the affinity of IAAP for Pgp.

Determination of the Activation Energies for Vi-induced Trapping Using [␣-32 P]8-Azido-ATP and [␣-32 P]8-Azido-ADP-
The results thus far clearly show the following: (a) that nucleoside diphosphates in the presence of Mg 2ϩ and Vi form a stable, ternary, non-covalent complex at the nucleotide-binding site of Pgp. The nucleoside diphosphate can be provided directly or as a nucleoside triphosphate that can be hydrolyzed in situ to a nucleoside diphosphate. (b) When the nucleoside diphosphate is trapped at the ATP site, the ternary complex manifests a profound conformational change at the substratebinding site. This suggests that although the nucleotide-and substrate-binding sites are independent, long range interactions acting via conformational changes result in these two sites being functionally coupled. (c) The functional effect on the substrate-binding site is the same regardless of whether a nucleoside di-or triphosphate is used to initiate the trapping of Pgp into the Pgp⅐ADP⅐Vi transition state intermediate. This raises the question as to whether the two routes for generating the Pgp⅐ADP⅐Vi transition state are thermodynamically comparable. Fig. 9A D555N and D1200N), respectively. It was determined by Western blotting using the monoclonal antibody C219 that wild-type and mutant Pgps were expressed at equivalent levels (data not shown). These crude membranes (60 g of protein) were incubated in the ATPase assay buffer containing 50 M [␣-32 P]8-azido-ATP or [␣-32 P]8-azido-ADP (2-4 Ci/nmol) and 250 M Vi in the dark for 10 min at 37°C. The reaction was stopped by the addition of ice-cold ATP (10 mM), and the samples were cross-linked by UV irradiation and immunoprecipitated using the human Pgp-specific polyclonal antibody PEPG-13 as described previously (7,9). Following SDS-PAGE on a 8% Tris glycine gel at constant voltage, gels were dried and exposed to Bio-Max MR film at Ϫ70°C for 16 -24 h. Autoradiogram shows Vi-induced trapping of [␣-32 P]8-azido-ADP under hydrolysis (lanes 1-3) and non-hydrolysis conditions (lanes 4 -6). Lane 1, wild-type Pgp membranes labeled with [␣- 32  [␣-32 P]8-azido-ADP is observed at 37°C.
To understand this difference, we determined the activation energies for these two processes. Fig. 9B depicts Arrhenius plots for Vi-induced trapping using either [␣-32 P]8-azido-ADP or [␣-32 P]8-azido-ATP in the temperature range 22-39°C. In this temperature range both the plots show no discontinuity, but the slopes are significantly different, and the steeper slope for the formation of the transition state intermediate with [␣-32 P]8-azido-ADP translates to a 2.5-fold higher activation energy (152 kJ/mol). A published report of the activation energy for ATP hydrolysis in crude membranes derived from the Chinese hamster ovary cell line, CH R B30, shows a discontinuity at 21°C (35). We could not perform our experiments at lower temperatures (Ͻ22°C) as the trapping with [␣-32 P]8azido-ADP and Vi even at 23°C was undetectable (cf. Fig. 9A). Thus, if [␣-32 P]8-azido-ADP is trapped into Pgp by incubating [␣-32 P]8-azido-ATP and Vi, allowing the [␣-32 P]8-azido-ATP to be hydrolyzed to [␣-32 P]8-azido-ADP, which is then trapped, the energy barrier is significantly lower than if [␣-32 P]8-azido-ADP is provided directly in the presence of Vi. The possible explanations for this apparently paradoxical result are considered under "Discussion." In addition, Table I lists the activation energies for the basal and substrate (verapamil)-stimulated hydrolysis of ATP and 8-azido-ATP by Pgp, the Vi-induced trapping of [␣-32 P]8-azido-ADP under hydrolysis and non-hydrolysis conditions, and for the binding of [␣-32 P]8-azido-ADP and IAAP to Pgp. The activation energies for the hydrolysis of ATP both in the absence (basal) and in the presence of verapamil are comparable (115.5 and 110.4 kJ/mol) and almost identical to that for the hydrolysis of 8-azido-ATP (100.1 kJ/ mol). In addition, the activation energy for ATP hydrolysis by Pgp in crude membranes derived from the Chinese hamster ovary cell line, CH R B30 (98.1 kJ/mol) (35), is comparable to that obtained with human Pgp in this study. These data indicate that in terms of the thermodynamics of the system, ATP hydrolysis by Pgp is independent of the species origin of the protein or the nucleotide (ATP or 8-azido-ATP) and that the drug substrate does not significantly affect the activation energy for nucleotide hydrolysis. Even more intriguing is the fact that the activation energy for the Vi-induced trapping of [␣-32 P]8-azido-ADP under hydrolysis conditions is 62 kJ/mol or approximately half that for the substrate-stimulated ATP hydrolysis. The Vi-induced trapping of [␣-32 P]8-azido-ADP arrests the catalytic cycle after only one hydrolysis event and thus would have an activation energy one-half of that for the entire catalytic cycle. It is also clear from the data in Table I that the activation energies for binding of nucleoside diphosphate or substrate (IAAP) are significantly lower when compared with trapping of nucleotides. Also, whereas either [␣-32 P]8-azido-ADP or [␣-32 P]8-azido-ATP can be used to initiate Vi-induced trapping with very similar kinetics and functional effects, the energy barriers that these two pathways entail are significantly different (152 versus 62 kJ/mol). Aliquots were removed at time intervals indicated on the x axis, treated with IAAP (5 nM) for 5 min, and cross-linked by UV irradiation. In a parallel experiment, the crude membranes were incubated with 50 M [␣-32 P]8-azido-ADP, (2-4 Ci/nmol) and 250 M Vi at 37°C. Aliquots were removed at the same time intervals as indicated above, and 200-fold excess ATP was added followed by cross-linking by UV irradiation. Following SDS-PAGE, the gels were dried, and the radioactivity incorporated into the Pgp band was quantified as described under "Experimental Procedures." The lines represent the best fit for the data by non-linear least squares regression analysis using the software

DISCUSSION
Due to the importance of Pgp in cancer chemotherapy and as a model system for ABC transporters in general, the catalytic cycle of this transporter has been studied in considerable detail (for reviews see Refs. 2, 14, and 30). Building upon the model proposed by Senior's group (14), we have recently elucidated the catalytic cycle of Pgp in considerable detail (10 -12). The essential features of the cycle are as follows. (i) ATP hydrolysis results in a dramatic conformational change where the affinity of both the substrate and the nucleotide for Pgp is reduced Ͼ30-fold. (ii) To transform Pgp from this intermediate state of low affinity for substrate to the next catalytic cycle, i.e. a conformation that binds substrate with high affinity, the hydrolysis of an additional molecule of nucleotide is obligatory. (iii) The release of ADP from the Pgp⅐ADP⅐P i transition state is the rate-limiting step in the catalytic cycle. These studies have relied heavily on the use of the radiolabeled photoaffinity analogue of ATP, [␣-32 P]8-azido-ATP. The Pgp⅐[␣-32 P]8-azido-ADP⅐Vi transition state intermediate has proved to be particularly useful in understanding the catalytic cycle. As illustrated in Fig. 2A, the forward reaction involves the hydrolysis of ATP and the dissociation of P i . Vi, an analogue of P i , then traps the ADP into a stable ternary complex, Pgp⅐ADP⅐Vi. There is published evidence that the trapped moiety is always the nucleoside diphosphate (3,13), which is confirmed by directly providing the nucleoside diphosphate in this work. In addition, we have characterized the trapping of the ADP analogue, [␣-32 P]8-azido-ADP, under both hydrolysis and non-hydrolysis conditions.
The results given in Fig. 1, A and B, indicate that [␣-32 P]8azido-ADP binds specifically to Pgp at the ATP site, similar to ADP. The question as to whether Vi-induced trapping can be effected in the absence of hydrolysis by directly providing the nucleoside diphosphate is addressed in Fig. 2B. We demonstrate that [␣- 32 (9,12,(27)(28)(29). Fig. 2C demonstrates that there is an almost equal distribution of [␣-32 P]8-azido-ADP into the N-and C-terminal halves regardless of whether the [␣-32 P]8-azido-ADP is trapped under hydrolysis or nonhydrolysis conditions. It has been established previously that Mg 2ϩ or another divalent cation such as Mn 2ϩ or Co 2ϩ are necessary for ATP hydrolysis (31) and Vi-induced trapping of [␣-32 P]8-azido-ADP (27,31). Consistent with these findings, the data in Fig. 3 demonstrate that Mg 2ϩ , Mn 2ϩ , and Co 2ϩ permit Vi-induced trapping of [␣-32 P]8-azido-ADP under both hydrolysis and non-hydrolysis conditions.
It is important to determine whether the kinetics of Viinduced trapping differs under hydrolysis and non-hydrolysis conditions. We show that during the trapping of [␣-32 P]8-azido-ADP via the hydrolysis of [␣-32 P]8-azido-ATP into Pgp, the K m for [␣-32 P]8-azido-ATP is nearly identical to the K d of [␣-32 P]8azido-ADP trapping when the nucleoside diphosphate is provided directly to Pgp in the presence of Vi (Fig. 4, A and B). These results provide evidence that [␣-32 P]8-azido-ADP is trapped into Pgp in the presence of Vi in a similar manner under both hydrolysis and non-hydrolysis conditions. In addition, [␣-32 P]8-azido-ADP can be used to delineate the relationship between the two ATP sites in hydrolysis and Vi-induced trapping. Earlier work (6,9) has shown that mutations in the conserved Walker B consensus motif in either the N-or the C-terminal ATP sites of Pgp arrest ATP hydrolysis as well as the Vi-induced trapping of [␣-32 P]8-azido-ADP generated through the hydrolysis of [␣-32 P]8-azido-ATP. The mutations in either the N-terminal ATP site (D555N) or the C-terminal site (D1200N) abolish Vi-induced trapping of [␣-32 P]8-azido-ADP by both the hydrolysis and non-hydrolysis routes (Fig. 6). These findings demonstrate that both ATP sites are required not only for ATP hydrolysis but also for Vi-induced nucleoside diphosphate trapping even under non-hydrolysis conditions. Thus, except for binding of nucleotide, each ATP site does not appear to be able to carry out subsequent steps in the catalytic cycle to 39°C. The reaction was stopped by the addition of 12.5 mM ice-cold ATP and placing the sample immediately on ice and followed by crosslinking by UV irradiation (365 nm) for 10 min on ice. Following SDS-PAGE, the gels were dried, and the radioactivity incorporated into the Pgp band was quantified as described under "Experimental Procedures." A plot of 1/temperature (K) versus log (extent of [␣-32 P]8-azido-ADP incorporated) was used to calculate the activation energy as described in the legend to Table I  without the participation of the other site.
A key aspect of the proposed catalytic cycle of Pgp has been the fact that ATP hydrolysis results in a dramatic conformational change, which results in a drastic decrease in the affinity of both the substrate (5,8,10) and the nucleotide (12) for the Vi-trapped intermediate of Pgp. These parameters thus would be important determinants in characterization of the Vi-trapped intermediate generated in the absence of ATP hydrolysis. The Vi-trapped intermediates of Pgp formed both in the absence of and following ATP hydrolysis show reduced affinity for nucleotide, as measured by the binding of the fluorescent analogue, TNP-ATP (Fig. 5). The second characteristic of the Vi-trapped intermediate, viz. that it exhibits a significant decrease in affinity for the drug substrate IAAP, is a direct and quantitative measure of the long range conformational coupling between the drug-and nucleotidebinding sites. Moreover, from the perspective of this study, the long range effect on the substrate-binding site would be a more stringent evaluation of the Vi-trapped conformations obtained via these two routes. We therefore evaluated this aspect in some detail. Thus, trapping Pgp in the transition state conformation with any of these nucleoside di-or triphosphates such as 8-azido-ADP, 8-azido-ATP, ADP, and ATP inhibits IAAP binding to Pgp (see Fig. 7). The 8-azido-ADP and 8-azido-ATP have lower K i values than ADP and ATP, respectively. Finally, we made simultaneous measurements of [␣-32 P]8-azido-ADP trapping and inhibition of IAAP binding during Vi-induced trapping of [␣-32 P]8-azido-ADP under non-hydrolysis conditions over time. The results show that these two events are inversely correlated, suggesting a cause-effect relationship (Fig. 8). A similar relationship exists when these measurements are made under ATP hydrolysis conditions (data not given).
Since the Vi-trapped intermediates generated under hydrolysis or non-hydrolysis conditions do not show functional differences, we compared the thermodynamics of these two pathways. The data in Fig. 9A (Fig. 9B). When Vi-induced trapping is effected by incubating [␣-32 P]8-azido-ATP in the presence of Vi under ATP hydrolysis conditions, the activation energy is 62 kJ/mol (mean value of three experiments). Conversely, when [␣-32 P]8-azido-ADP is directly trapped in the presence of Vi without ATP hydrolysis, the activation energy is 152 kJ/mol (mean value of three experiments). Thus, the latter route of generating the transition state conformation has an energy barrier ϳ2.5 times higher than the hydrolysis route. Moreover, the trapping of [␣-32 P]8-azido-ADP under non-hydrolysis conditions has an activation energy 1.5 times higher than that required for basal or verapamil-stimulated hydrolysis of ATP or 8-azido-ATP, which represents the complete catalytic cycle (see Table I and Fig. 10). This result is consistent with the hypothesis that the hydrolysis of [␣-32 P]8azido-ATP provides energy to facilitate the conformational changes that accompany Vi-induced trapping. On the other hand, when [␣-32 P]8-azido-ADP is directly trapped into Pgp, there is no accompanying hydrolysis, and thus this reaction would necessarily have a much greater energy barrier to overcome; for this reason, in the normal catalytic cycle of ATP hydrolysis, this reaction would be highly unfavorable (see Fig.  10 for the comparison of the activation energies required for various steps in the catalytic cycle of ATP hydrolysis). These   Fig. 9B. E act ϭ Ϫ(slope)2.3R; R ϭ 1.98. b Vi-sensitive basal (in the absence of drug substrate). ATP hydrolysis was measured as described below. c Verapamil (30 M)-stimulated ATP-and 8-azido-ATP hydrolysis by Pgp in crude membranes was carried out as described under "Experimental Procedures" at different temperatures ranging from 22 to 39°C.
d Experiment was carried out as described in the legend to Fig. 9B (values represent mean Ϯ S.D.; n ϭ 3, and r is given for a representative experiment).
e Crude membranes (1 mg/ml) were incubated in the ATPase assay buffer with 10 M [␣-32 P]8-azido-ADP (10 Ci/nmol) in the dark for 6 min at different temperatures ranging from 22 to 39°C. Following UV irradiation and SDS-PAGE, the radioactivity incorporated in the Pgp band was determined using the PhosphorImager as described under "Experimental Procedures." Values represent average of two experiments, and r is given for a representative experiment.
f Crude membranes (10 -50 g of protein) were incubated in 50 mM Tris-HCl, pH 7.5, with IAAP (5-7 nM) for 5 min under subdued light at different temperatures ranging from 22 to 39°C. The samples were photocross-linked and, following SDS-PAGE radioactivity in the Pgp band, determined as described under "Experimental Procedures." Values represent average of two experiments, and r is given for a representative experiment. data also provide an explanation for previous studies (3), which show that the inhibition of ATPase activity at 37°C was much more rapid with ATP and Vi than with ADP and Vi. Moreover, as shown in Table I, the activation energy for the binding of [␣-32 P]8-azido-ADP in the absence of Vi is only 19.5 kJ/mol. This is about 7.5-fold lower than the activation energy for the trapping of [␣-32 P]8-azido-ADP in the presence of Vi. This suggests that it is not the binding step but the subsequent conformational changes that generate Vi-trapped intermediate(s) that are energetically intensive. Similarly, the activation energy for the binding of the hydrophobic drug-substrate IAAP to Pgp is extremely low (7.97 kJ/mol).
In recent years, the crystal structures of the ATP subunits of several ABC and analogous transporters have been resolved. These include the following: HisP, the ATP subunit of the bacterial histidine permease (36); MutS, a protein that recognizes mispaired and unpaired bases in duplex DNA and initiates mismatch repair (37,38); ArsA, the soluble ATPase component of the bacterial arsenite pump, ArsAB (39); and MalK, the ATPase subunit of the trehalose/maltose transporter (40). In most of these studies (viz. MutS, ArsA, and MalK), the nucleoside diphosphate was directly incorporated by including ADP (and in some cases Vi or aluminum fluoride) during crystallization. Our study suggests that these structures where the ADP has not been incorporated in situ by the hydrolysis of ATP are nonetheless representative of the native conformation. Also, published reports postulate that the nucleoside diphosphate may have interesting regulatory roles to play in the catalytic cycles of several ABC transporters such as cystic fibrosis transmembrane conductance regulator and the sulfonyl urea receptor, SUR1 (41)(42)(43). In many of these instances [␣-32 P]8-azido-ADP would prove very useful in designing experimental strategies to address these hypotheses directly.
Taken together, our results provide compelling evidence that although, there is a 2.5-fold difference in the activation energies required to generate the Pgp⅐[␣-32 P]8-azido-ADP⅐Vi complex using [␣-32 P]8-azido-ADP and Vi compared with [␣-32 P]8-azido-ATP and Vi, the transition state complex generated by either route is functionally indistinguishable. Our preliminary observations with another ABC transporter, the MRP1 (44), suggest that MRP1 also exhibits Vi-induced trapping of [␣-32 P]8-azido-ADP under both hydrolysis and nonhydrolysis conditions. 2 These findings are consistent with the results reported in this paper for Pgp. This work, however, does not address the effect of drug substrates on Vi-induced trapping of [␣-32 P]8-azido-ADP under non-hydrolysis conditions. These experiments are currently in progress.