Rapid purification of human P-glycoprotein mutants expressed transiently in HEK 293 cells by nickel-chelate chromatography and characterization of their drug-stimulated ATPase activities.

P-glycoprotein containing 10 tandem histidine residues at the COOH end of the molecule was transiently expressed in HEK 293 cells and purified by nickel-chelate chromatography. The purified protein had an apparent mass of 170 kDa, and its verapamil-stimulated ATPase activity in the presence of phospholipid was 1.2 μmol/min/mg of P-glycoprotein. We then characterized P-glycoprotein mutants that exhibited altered drug-resistant phenotypes and analyzed the contribution of the two nucleotide binding folds to drug-stimulated ATPase activity. Mutation of residues in either nucleotide binding fold abolished drug-stimulated ATPase activity. The pattern of drug-stimulated ATPase activities of mutants, which conferred increased relative resistance to colchicine (G141V, G185V, G830V) or decreased relative resistance to all drugs (F978A), correlated with their drug-resistant phenotypes. By contrast, the ATPase activity of mutant F335A was significantly higher than that of wild-type enzyme when assayed in the presence of verapamil (3.4-fold), colchicine (9.1-fold), or vinblastine (3.7-fold), even though it conferred little resistance to vinblastine in transfected cells. These results suggest that both nucleotide-binding domains must be intact to couple drug binding to ATPase activity and that the drug-stimulated ATPase activity profile of a mutant does not always correlate with its drug-resistant phenotype.

P-glycoprotein, also known as the multidrug-resistant protein (MDR), 1 is a plasma membrane glycoprotein involved in the ATP-dependent efflux of a broad range of cytotoxic drugs from cells (reviewed by Endicott and Ling (1989), Roninson (1991), and Gottesman and Pastan (1993)). It may be one of the mechanisms responsible for failure of cancer chemotherapy (Bradley and Ling, 1994).
In order to understand the mechanism of drug efflux it is necessary to identify the site(s) of drug binding, determine the role of the two nucleotide-binding domains, and characterize the interactions between the cytoplasmic and transmembrane domains. One approach has been to analyze the effects of mutations on the ability of the enzyme to confer resistance to various cytotoxic agents (Loo and Clarke, 1993a, 1993b, 1994a, 1994bKajiji et al., 1994). This method is time consuming, however, and does not yield enough enzyme to directly measure function. In the absence of a direct assay, it is possible that the response of the cell to a cytotoxic agent could also involve activation of endogenous drug pump(s) or involve other mechanisms of drug resistance. Therefore, structure-function analysis of P-glycoprotein would be greatly enhanced if mutants could be rapidly expressed and purified in a functional state. To date, rapid expression and purification of eukaryotic membrane proteins have not been feasible. An approach that has been successfully applied to soluble proteins is metal-chelate chromatography of proteins containing a histidine tag (Janknecht et al., 1991). In this study, we used nickel chromatography to purify P-glycoprotein containing 10 histidine residues at the COOH end of the protein after expression in HEK 293 cells. We then used this approach to study the contribution of either nucleotide-binding domain of P-glycoprotein to drugstimulated ATPase activity and characterized the drug-stimulated ATPase activities of mutants with altered drug-resistant profiles. We show that both nucleotide-binding domains are essential for coupling of ATPase activity to drug binding and that the drug-stimulated ATPase activities of all the mutants, except for F335A, correlated with their drug-resistant phenotypes.

EXPERIMENTAL PROCEDURES
Construction of Mutants-A full-length MDR1 cDNA was modified to encode for 10 histidine residues the COOH end of the protein. The sequence at the COOH terminus of P-glycoprotein that would normally end as TKRQ now became TKRA(His) 10 LDPRQ.
Expression and Purification of P-glycoprotein Mutants-Procedures for stable transfection of mouse NIH 3T3 cells, followed by selection in the presence of vinblastine (5 nM) or colchicine (45 nM) have been described previously (Loo and Clarke, 1993a). For purification of Pglycoprotein, twenty (10-cm diameter) culture dishes of subconfluent HEK 293 cells were transfected with MDR1 cDNA, and membranes were prepared 40 h later as described previously (Loo and Clarke, 1994c). The membranes were suspended in 0.3 ml of buffer A (50 mM NaPO 4 , pH 8.0, 500 mM NaCl, 50 mM imidazole, and 20% (v/v) glycerol) and solubilized at 4°C by addition of 1 ml of buffer A containing 1% (w/v) n-dodecyl-␤-D-maltoside (Sigma). Insoluble material was removed by centrifugation at 16,000 ϫ g for 15 min, and the supernatant was applied onto a nickel spin column (Ni-NTA, Qiagen), which had been pre-equilibrated with buffer B (buffer A containing 0.1% (w/v) n-dodecyl-␤-D-maltoside). The column was washed twice with 0.6 ml of buffer B and twice with 0.6 ml of buffer C (10 mM Tris-HCl, pH 7.5, 500 mM NaCl, 80 mM imidazole, pH 7, 0.1% (w/v) n-dodecyl-␤-D-maltoside, and 20% (v/v) glycerol). P-glycoprotein-(His) 10 was then eluted with 0.25 ml of buffer D (buffer C but containing 300 mM imidazole). The eluted material was diluted 6-fold with buffer A lacking imidazole and reapplied onto the same column that had been regenerated by washing with 1 M imidazole, 0.1% (w/v) n-dodecyl-␤-D-maltoside followed by equilibration in buffer B. The column was washed as described above, and P-glycoprotein was eluted with 0.25 ml of buffer D. The yield of Pglycoprotein was determined by measuring the amount of protein as described by Bradford (1976) using bovine serum albumin as a standard and subtracting the amount of protein obtained when the same number of vector-transfected cells was carried through the purification procedure. Yields of purified P-glycoprotein ranged from 6 to 12 g.
Measurement of Mg 2ϩ -ATPase Activity-Purified P-glycoprotein was diluted with an equal volume of 50 mg/ml crude sheep brain phosphatidylethanolamine (Sigma), which had been previously washed with Trisbuffered saline to remove traces of phosphate and then sonicated. 100 ng of purified P-glycoprotein was incubated with the desired drug and ATPase activity initiated by addition of an equal volume of buffer containing 100 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl 2 , and 10 mM ATP. The samples were incubated at 37°C for 40 min, and the amount of inorganic phosphate liberated was determined by the method of Chifflet et al. (1988).

RESULTS
Purification of P-glycoprotein-The cDNA coding for P-glycoprotein was modified to contain 10 histidine residues at the COOH end of the molecule. The presence of the tag did not affect function since transfection of mouse NIH 3T3 cells yielded drug-resistant colonies that had drug-resistant profiles that were similar to those of wild-type enzyme (data not shown).
To purify P-glycoprotein, its cDNA was transiently expressed in HEK 293 cells as they yield a relatively high level of Pglycoprotein. More than 80% of the P-glycoprotein could be extracted from the membranes with the detergent n-dodecyl-␤-D-maltoside. The detergent also did not interfere with the binding of P-glycoprotein to the nickel column (Ͼ90% of Pglycoprotein was bound to the resin). Ionic detergents or inclusion of phospholipids during the purification procedure prevented binding of P-glycoprotein to the column. Other non-ionic detergents (Triton X-100 or C 12 E 8 (octaethylene glycol dodecyl ether)) inactivated P-glycoprotein, while octyl glucoside was less effective (Ͻ20%) in solubilizing P-glycoprotein. SDS-PAGE of the fractions eluted with 300 mM imidazole ( Fig. 1) shows the presence of a single major band of an apparent mass of 170 kDa (lane 6B), which is not present from cells transfected with vector alone (lane 6A). When the purified fractions were subjected to immunoblot analysis with antibody against human P-glycoprotein, the 170-kDa protein was the only immunoreactive band (data not shown). After one round of purification, greater than 50% of the eluted proteins was P-glycoprotein-(His) 10 . A second round of purification resulted in up to 80% of the protein being P-glycoprotein, with yields of 6 -12 g of P-glycoprotein. Occasionally a minor contaminating band of apparent mass 56 kDa was also present when a regenerated rather than a new nickel column was used during the second round of purification.

Characterization of the ATPase Activities of Purified Wildtype P-glycoprotein and Mutants with Altered Drug-resistant
Profiles- Fig. 2 shows that wild-type P-glycoprotein-(His) 10 in the presence of phosphatidylethanolamine had an ATPase activity of 0.11 mol/min/mg of P-glycoprotein. In the absence of lipid, no ATPase activity was detected (Ͻ0.015 mol/min/mg of P-glycoprotein). In the presence of verapamil and lipid, however, the ATPase activity increased about 11-fold (1.20 mol/ min/mg of P-glycoprotein). All ATPase activities were determined in the absence of EGTA and ouabain, as no Ca 2ϩ -ATPase or Na ϩ -K ϩ -ATPase activity was detected (assayed Ϯ 2 mM EGTA or Ϯ 2 mM ouabain, respectively). Lower levels of stimulation were observed in the presence of 800 M colchicine (3.1-fold) and vinblastine (5.8-fold). Maximal stimulation of ATPase activity in the presence of vinblastine occurred at 20 -50 M, and the activity decreased at higher concentrations of vinblastine.
We have previously identified residues in P-glycoprotein, which alter its ability to confer resistance to various cytotoxic drugs in transfected cells (Loo and Clarke, 1993a, 1993b, 1994a, 1994b. For example, mutants G141V or G830V conferred increased resistance to colchicine (about 3-fold) relative to that of wild-type enzyme while mutant F335A conferred decreased resistance to vinblastine. By contrast, mutant F978A conferred decreased resistance to all drugs. In this study, we also included for comparison mutant G185V, which was recently shown by Rao (1995) to have increased verapamil-and colchicine-stimulated ATPase activities (2-and 3.3-fold, respectively). Therefore, to determine the effects of these drugs on the ATPase activities of these mutants, we purified and reconstituted each mutant P-glycoprotein into phospholipid and measured ATPase activity in the presence of various concentrations of vinblastine, colchicine, or verapamil. When analyzed by SDS-PAGE, the major protein in each purified preparation had an apparent mass of 170 kDa (corresponding to the fully glycosylated form of P-glycoprotein) and was present in similar amounts (Fig. 3). The maximal verapamil-stimulated ATPase activities of mutants G141V, G185V, and G830V were all slightly increased (1.4 -1.7-fold) relative to that of wild-type enzyme (Fig. 2). The half-maximal stimulation of the ATPase activities of the glycine mutants was 9 -16 M, compared with 40 M for the wild-type enzyme, suggesting that the mutants had increased affinity for verapamil. Vinblastine-stimulated ATPase activities of all three mutants, however, were similar to that of wild-type enzyme, whereas colchicine-stimulated ATPase activities were markedly increased (2.8 -3.7-fold).
FIG. 1. Purification of P-glycoprotein. HEK 293 cells were transfected with either vector alone (A) or with cDNA coding for P-glycoprotein-(His) 10 (B). P-glycoprotein was purified on a Ni-NTA column (Qiagen, Inc) as described under "Experimental Procedures." Samples from each step of the purification procedure were analyzed by SDS-PAGE on an 8% gel, and the bands were visualized by Coomassie Blue staining. Lanes 1, crude membranes; lanes 2, solubilized extracts after high speed centrifugation; lanes 3, flow-through material from Ni-NTA columns; lanes 4, fractions eluted with 50 mM imidazole; lanes 5, fractions eluted with 80 mM imidazole wash; lanes 6, fractions eluted with 300 mM imidazole (4% of the eluted material and containing 500 ng of protein (lane 6B)). M, molecular weight markers.
FIG. 2. Drug-stimulated ATPase activities of purified and reconstituted wild-type and mutant P-glycoproteins. Wild-type (E) and mutants G141V (å), G185V (Ⅺ), G830V (q), F335A (f), and F978A (Ç) P-glycoproteins-(His) 10 were purified using Ni-NTA spin columns and reconstituted with sheep brain phosphatidylethanolamine. ATPase activity was determined in the presence of various concentrations of verapamil, vinblastine, and colchicine as described under "Experimental Procedures." Mutant F978A, which confers little resistance to vinblastine, colchicine, doxorubicin, or actinomycin D in transfected cells, also showed little drug-stimulated ATPase activity, except at very high concentrations of verapamil (1.04 mol/min/mg of P-glycoprotein at 800 M verapamil). Similarly, no colchicinestimulated ATPase activity could be detected. Vinblastinestimulated ATPase activity, however, was detected at low concentrations of vinblastine (10 -50 M) but was 3-4-fold lower than that observed with wild-type enzyme. By contrast, mutation of Phe-335, which is found in an equivalent position to Phe-978, when homologous halves of P-glycoprotein are aligned resulted in a large increase in drug-stimulated ATPase activity. Maximal ATPase activities in the presence of verapamil, vinblastine, or colchicine were 4 -9-fold greater than that observed with wild-type enzyme. The basal ATPase activity of mutant F335A was also about 3-fold higher (0.32 mol/min/mg of P-glycoprotein) than that of wild-type enzyme (0.11 mol/ min/mg of P-glycoprotein). Therefore, the major effect of mutation of Phe-335 was to increase the overall ATPase activity of the enzyme.
Effects of Mutations within the Predicted Nucleotide-binding Sites-P-glycoprotein is predicted to contain two nucleotide binding folds. Each fold contains a homology A consensus sequence (Walker et al., 1982) for nucleotide binding (GNS-GCGKS and GSSGCGKS, respectively). We initially tested the ATPase activity of a mutant in which both cysteines in the homology A regions (Cys-431 and Cys-1074) as well as all other cysteines were mutated to alanines. The purified Cys-less Pglycoprotein had a higher amount of the core-glycosylated form (150 kDa) of P-glycoprotein (Fig. 3), which is probably due to the slow maturation of this mutant protein (Loo and Clarke, 1995). Fig. 4 shows that the Cys-less mutant retained approximately 70% of the verapamil-stimulated ATPase activity of wild-type enzyme. To determine the contribution of either nucleotide-binding domain to drug-stimulatable ATPase activity, mutations were made to the core amino acids (G432S, K433M, G1075S, and K1076M, respectively) of the homology A consensus sequences. No drug-resistant colonies were obtained when the cDNAs coding for these mutants were transfected into NIH 3T3 cells and selected in the presence of colchicine (45 nM) or vinblastine (5 nM). These mutants were then transiently expressed in HEK 293 cells and purified using Ni-NTA columns. All four mutants yielded a 170-kDa protein as the major product and in similar amounts to that obtained with wild-type P-glycoprotein-(His) 10 , suggesting that the mutations did not affect processing of the enzyme (Fig. 3). Mutations to the core amino acids (GK) in either nucleotide-binding domain abolished verapamil-stimulated ATPase activity (Fig. 4), suggesting that both nucleotide-binding domains are essential for coupling drug binding to ATPase activity.

DISCUSSION
Purification of P-glycoprotein using nickel-chelate chromatography following transient expression in HEK 293 cells has several advantages over purification from stable cell lines overexpressing P-glycoprotein (Urbatsch et al., 1994;Shapiro and Ling, 1994;Sharom et al., 1993) or following expression in insect Sf9 cells (Rao, 1995). The main advantage is that the expression, purification, and assay of ATPase activity for any mutant P-glycoprotein can be completed within 2 days, while the other methods often take months. A transient expression system also avoids the problems associated with potential recombination of the mutant P-glycoprotein cDNA with any endogenous P-glycoprotein genes. Although the level of expression of P-glycoprotein is relatively low in HEK 293 cells (0.1-0.3% by weight), the use of nickel-chelate chromatography provided a simple and efficient method to isolate highly purified P-glycoprotein in an active state. This procedure should be applicable to the study of the structure-function relationships of other eukaryotic membrane proteins.
In this study, we addressed two important questions concerning the structure and mechanism of P-glycoprotein: 1) are both ATP-binding sites required for drug stimulation of ATPase activity and 2) is there a correlation between the drugresistant phenotype of a mutant and its pattern of drugstimulated ATPase activity? Our results suggest that both nucleotide-binding sites must be intact for coupling of ATPase activity to drug binding. Mutation of the core (GK) amino acids in either homology A nucleotide-binding consensus sequence abolished basal as well as drug-stimulated ATPase activity. It is possible that both sites must be occupied simultaneously for coupling of drug binding to ATPase activity or that ATP binding occurs sequentially during the reaction cycle. There is, however, no evidence of cooperativity between the nucleotidebinding sites (Sharom et al., 1995).
Mutations that alter the drug-resistant profiles of P-glycoprotein also had profound effects on the pattern of drug-stimulated ATPase activities. For mutants G141V, G185V, G830V, and F978A, the pattern of drug-stimulated ATPase correlated with their relative drug-resistant profiles in transfected cells. In the glycine mutants, there was enhanced stimulation of ATPase activity by colchicine, whereas stimulation by vinblastine resembled that of wild-type enzyme. Similarly, in transfected cells, the relative resistance of these mutants to vinblastine was similar to that of wild-type enzyme, but the relative resistance to colchicine was elevated (about 3-fold). Mutant F978A conferred little resistance to all drug substrates in transfected cells, and the purified protein also showed ex- FIG. 3. SDS-PAGE analysis of purified P-glycoprotein mutants. Mutant P-glycoproteins were expressed in HEK 293 cells and purified using nickel-chelate chromatography as described under "Experimental Procedures." A sample corresponding to 4% of the purified fraction was subjected to SDS-PAGE, and the bands were visualized by Coomassie Blue staining. The positions of the mature (170 kDa) and core-glycosylated (150 kDa) forms of P-glycoprotein are indicated.
FIG. 4. Verapamil-stimulated ATPase activity of P-glycoprotein mutants. Purified P-glycoprotein-(His) 10 proteins were reconstituted with sheep brain phosphatidylethanolamine, and ATPase activity was measured in the presence of various concentrations of verapamil as described under "Experimental Procedures." E, wild type; q, Cys-less; f, mutant G432S. Mutants K433M, G1075S, and K1076M had no detectable ATPase activities and are omitted for clarity. tremely low levels of drug-stimulated ATPase activity. These results suggest that mutation of Phe-978 resulted in either decreased affinity (verapamil) and/or interference in coupling of drug binding to ATPase activity (vinblastine).
Purified mutant F335A P-glycoprotein, however, showed large increases in ATPase activity in the presence of all three drug substrates but conferred decreased relative resistance to vinblastine and only a small increase in resistance to colchicine in transfected cells (Loo and Clarke, 1993b). One explanation for this discrepancy is that mutation F335A alters the dissociation of vinblastine from P-glycoprotein such that the enzyme is slow in effluxing vinblastine. It is also possible that the mutation alters the conformation of the enzyme such that is now in an "uncoupled" state. These possibilities could explain the fact that purified mutant F335A has a higher basal as well as drug-stimulated ATPase activity compared with wild-type enzyme.
The results of this study show that drug-stimulated ATPase activity of a mutant does not always correlate with its drugresistant phenotype.