Determining the dimensions of the drug-binding domain of human P-glycoprotein using thiol cross-linking compounds as molecular rulers.

The human multidrug resistance P-glycoprotein (P-gp) interacts with a broad range of compounds with diverse structures and sizes. There is considerable evidence indicating that residues in transmembrane segments 4-6 and 10-12 form the drug-binding site. We attempted to measure the size of the drug-binding site by using thiol-specific methanethiosulfonate (MTS) cross-linkers containing spacer arms of 2 to 17 atoms. The majority of these cross-linkers were also substrates of P-gp, because they stimulated ATPase activity (2.5- to 10.1-fold). 36 P-gp mutants with pairs of cysteine residues introduced into transmembrane segments 4-6 and 10-12 were analyzed after reaction with 0.2 mm MTS cross-linker at 4 degrees C. The cross-linked product migrated with lower mobility than native P-gp in SDS gels. 13 P-gp mutants were cross-linked by MTS cross-linkers with spacer arms of 9-25 A. Vinblastine and cyclosporin A inhibited cross-linking. The emerging picture from these results and other studies is that the drug-binding domain is large enough to accommodate compounds of different sizes and that the drug-binding domain is "funnel" shaped, narrow at the cytoplasmic side, at least 9-25 A in the middle, and wider still at the extracellular surface.

The human multidrug resistance P-glycoprotein (P-gp) 1 uses ATP to pump out of the cell a wide variety of structurally diverse compounds (recently reviewed in Ref. 1). Many of these compounds are clinically important in cancer and AIDS chemotherapy (2)(3)(4). Therefore, overexpression of P-gp often leads to multidrug resistance. The pattern of expression in tissues and studies on P-gp knock-out mice indicate that its physiological role may be to protect the organism from toxins in the environment and in the diet (5)(6)(7). P-gp is a member of the ABC (ATP-binding cassette) family of transporters (8). Its 1280 amino acids are organized in two repeating units of 610 amino acids that are joined by a linker region of about 60 amino acids (9). Each repeat has six transmembrane (TM) segments and a hydrophilic domain containing an ATP-binding site (10,11).
An important goal in determining the mechanism of P-gp is to understand how P-gp can bind so many different compounds and how ATP hydrolysis causes drug transport. The minimal functional unit in P-gp is a monomer (12). Both nucleotidebinding sites are required for function, because P-gp is inactive when ATP hydrolysis at either site in blocked by mutation or chemical modification (13)(14)(15)(16)(17)(18). The nucleotide-binding domains may function in an alternating mechanism (19,20).
Disulfide cross-linking studies have provided considerable insight into the structure of membrane proteins (33)(34)(35)(36)(37). Recently, we identified residues in TMs 4, 5, 6, 10, 11, and 12 that contribute to the drug-binding domain (38 -41). In this study, we used a series of thiol-specific cross-linkers with spacer arms of various lengths to measure distances between these residues.

EXPERIMENTAL PROCEDURES
Construction of Mutants-There are seven endogenous cysteines at positions 137, 431, 717, 956, 1074, 1125, and 1227 in wild-type P-gp. None of the cysteines are needed for activity, because mutation of all cysteines to alanine (Cys-less P-gp) resulted in an active molecule (10). The Cys-less P-gp cDNA was also modified to code for ten histidine residues at the COOH end of the molecule (Cys-less P-gp(His) 10 ). The histidine tag facilitated purification of the Cys-less P-gp by nickel-chelate chromatography (42). Cysteine residues were then introduced into the Cys-less P-gp(His) 10 as described previously (40). The integrity of the mutated cDNA was confirmed by sequencing the entire cDNA (43).
Expression, Purification, and Measurement of Drug-stimulated ATPase Activity-Expression and purification of histidine-tagged P-gp mutants were described previously (42). Briefly, 50 10-cm diameter culture plates of HEK 293 cells were transfected with the mutant cDNA. After 24 h, the medium was replaced with fresh medium containing 10 M cyclosporin A. Cyclosporin A is a substrate of P-gp and is a powerful chemical chaperone for promoting maturation of P-gp (44). The transfected cells were harvested 24 h later, solubilized with 1% (w/v) n-dodecyl-␤-D-maltoside, and the mutant P-gp was isolated by nickel-chelate chromatography (Ni-NTA columns; Qiagen, Inc., Mississauga, Ontario, Canada).
The P-gp-(His) 10 mutants were eluted from the column and mixed with an equal volume of 10 mg/ml sheep brain phosphatidylethanolamine (Type II-S; Sigma-Aldrich) that was washed and suspended in 10 * This work was supported in part by National Institutes of Health Grant RO1 CA80900 and by grants from the Canadian Institutes for Health Research (CIHR) and the Canadian Cystic Fibrosis Foundation. 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  mM Tris-HCl, pH 7.5, and 150 mM NaCl. The P-gp:lipid mixture was then sonicated for 45 s at 4°C. An aliquot of the mixture was assayed for drug-stimulated ATPase activity by addition of an equal volume of buffer containing 100 mM Tris-HCl, pH 7.5, 100 mM NaCl, 20 mM MgCl 2 , 10 mM ATP, and 2 mM MTS cross-linker. The samples were incubated for 30 min at 37°C, and the amount of inorganic phosphate liberated was determined (45).
Cross-linking Analysis-The mutant P-gps were expressed in HEK 293 cells in the presence of 10 M cyclosporin A. Membranes were prepared from transfected cells and suspended in Tris-buffered saline (10 mM Tris-HCl, pH 7.4, 150 mM NaCl) and treated with 0.2 mM cross-linker (Toronto Research Chemicals, Toronto, Ontario, Canada; see Fig. 1) for 15 min at 4°C. At this concentration, the MTS crosslinkers stimulated the ATPase activity of Cys-less P-gp by about 50%. The reactions were stopped by addition of 2ϫ SDS sample buffer containing 10 mM N-ethylmaleimide. In the protection experiments, the membranes were pretreated for 10 min at 4°C in the presence of 1 mM vinblastine or 1 mM cyclosporin A (saturating conditions). The samples were subjected to SDS-PAGE on 7.5% acrylamide gels and immunoblot analysis with rabbit polyclonal antibody (12).

RESULTS
To measure distances between residues in the NH 2 and COOH halves of the drug-binding domain, we constructed P-gp mutants that had a pair of cysteine residues, one in the NH 2 half and the other in the COOH half (Table I). The mutants were then tested for cross-linking with thiol-specific cross-linkers. In attempting to measure distances between residues, it would be most useful to use cross-linkers that had similar spacer arms and reactive groups to minimize differences in chemical reactivity with cysteines. A set of thiol-specific crosslinkers with these properties is the MTS cross-linkers shown in Fig. 1. Alkylthiosulfonates react selectively with cysteines in a protein resulting in a disulfide attachment of the spacer arm and release of a sulfonic acid byproduct (46,47). The MTS compounds are generally more reactive with cysteines than other thiol-specific compounds such as maleimides or iodoacetates (47).
In using cross-linkers to map distances in the drug-binding domain, it is important that the compounds can actually occupy the drug-binding site. It is best to be able to measure P-gpmediated transport of these MTS compounds. This is technically not feasible, because radioactive forms are not available commercially. Another complication is the short half-lives of MTS compounds in aqueous media (46), and they are also not fluorescent. One way around these problems is to measure stimulation of ATPase activity. Binding of most substrates to P-gp stimulates its ATPase activity, and there is good correlation between drug-stimulated ATPase activity and drug transport (48). Accordingly, all the MTS cross-linkers ( Fig. 1) were assayed for their ability to stimulate Cys-less P-gp ATPase activity. Cys-less P-gp has all seven endogenous cysteines replaced with alanine and is nearly as active (drug transport and drug-stimulated ATPase activity) as wild-type P-gp (10, 42).  (Table I). The mutants were first expressed in HEK 293 cells and were found to be processed to the fully mature (170 kDa) form of P-gp (data not shown).
The 36 mutants (Table I) were then subjected to cross-linking by the MTS cross-linkers (Fig. 1). Membranes were prepared from transfected cells and treated with different crosslinkers (0.2 mM) at 4°C. The reactions were done at 4°C to reduce molecular motions in the protein. The reactions were stopped by addition of SDS sample buffer containing N-ethylmaleimide, and the mixture was analyzed by SDS-PAGE. In previous studies we had shown the cross-linking between residues in the NH 2 and COOH halves of P-gp resulted in the cross-linked product migrating with lower mobility in SDS gels (49,50). Similarly   (TM10)  ----G872C (TM10)  5, 6, 17  8, 17  8, 11, 14, 17  -F942C (TM11)  --17  -T945C (TM11)  -8, 11,

Cross-linking of Residues in the Drug-binding Site 36878
cross-linked product was not detected after treatment with 5 mM dithiothreitol (data not shown).
We then tested whether drug substrates could inhibit crosslinking. If cross-linking occurred between residues in the drugbinding domain, then the presence of substrates such as cyclosporin A and vinblastine should inhibit cross-linking. Cyclosporin A is an inhibitor of P-gp-medicated drug resistance (51) whereas vinblastine is a cytotoxic drug substrate (1). The membranes containing P-gp mutant I306C/V982C were preincubated with 1 mM cyclosporin A or 1 mM vinblastine for 10 min at 4°C. The samples were then treated with M17M and subjected to Western blot analysis. Fig. 3 shows that cyclosporin A and vinblastine blocked cross-linking of mutant I306C/V982C by M17M. An example of a blot showing no detectable cross-linking is seen in mutant I306C/A871C (Fig. 3, lower panel).
The cross-linking results of the other mutants (data not shown) are summarized in Table I No cross-linking was detected with the A342C(TM6) mutants. All cross-linking was blocked when the membranes were preincubated with 1 mM vinblastine, 1 mM cyclosporin A. Similarly, no cross-linked product was detected after treatment with dithiothreitol (data not shown).

DISCUSSION
The cross-linking results in Fig. 4A show that residues 222(TM4), 306(TM5), 339(TM6), 868(TM10), 872(TM10), 942C(TM11), 945(TM11), 982(TM12), 984(TM12), and 985(TM12) must line a common drug-binding site. Residues in TMs 5 and 6 were cross-linked to residues in TMs 10, 11, and 12, whereas residue S222C(TM4) was cross-linked with two residues (I868C and G872C) in TM10. In all cases, crosslinking was blocked by substrates such as vinblastine or cyclosporin A. Vinblastine and cyclosporin are quite large molecules. Their crystal structures show that they are about 20 -25 Å in length at their widest point (52,53). The cross-linking results indicate that the drug-binding site would be large enough to accommodate vinblastine or cyclosporin A. For example, mutants L339C(TM6)/F942C(TM11), L339C(TM6)/T945C(TM11), and L339C(TM6)/A985C(TM12) are only cross-linked with MTS compounds with spacer arms of more than 20 Å. The smallest distance between TMs 4, 5, and 6 and TM 10, 11, and 12 as measured by these cross-linkers is about 9 Å and occurs between TMs 4 and 10, because mutants S222C(TM4)/ Cross-linking of Residues in the Drug-binding Site 36879 I868C(TM10) and S222C(TM4)/G872C(TM10) could be crosslinked with M5M. It is interesting to note that none of the 36 mutants (Table I) showed any detectable cross-linking with a zero-length cross-linker (copper phenanthroline) at either 22 or 4°C (data not shown). It was surprising to find mutants such as the S222C mutants that were cross-linked with smaller reagents (e.g. M8M) were also cross-linked with the larger M17M cross-linker. One explanation is that the TMs of P-gp are quite flexible (54) and therefore, can expand or contract to accommodate different sizes of substrate. This mobility of the helices in accommodating different substrates would inherently expose different residues to the drug-binding site and thereby dictate the affinity of P-gp for a particular substrate. It is also possible that the reactive cysteines may be in a sterically favored position to react with cross-linker of a certain size. This may indeed be the case. Table I show that some but not all of the mutants react with the same combination of MTS cross-linkers. Similarly, no cross-linking was detected in 23 of the mutants. Another possibility is that M17M may be a particularly flexible reagent. Recently, Green et al. (55) reported that a potential problem with using homobifunctional protein cross-linking agents as molecular rulers is that they can be quite flexible. These compounds can adopt a range of conformations due to rotation of the bonds that form the linker arm. There are several C-O bonds in the spacer arm of M17M that could adopt gauche conformations, thereby bringing the two reactive ends closer together. Thus, it is possible that some mutants are reacting with M17M when it is in conformations that bring the reactive groups closer together, whereas other mutants would react only with the M17M when it is in the extended conformation.
The residues that have been identified to contribute to the drug-binding site are in the middle of the predicted TM segments and to be 9 -25 Å apart (Fig. 4B). Previous studies have shown that the cytoplasmic side of the TMs (4 -6 and 10 -12) predicted to line the drug-binding site are closer than this, because disulfide cross-linking occurred between residues in these TMs with a zero-length cross-linking agent (copper phenanthroline) (54). The TM segments on the extracellular surface, however, must be quite far apart. A low resolution crystal structure of P-gp shows the presence of a pore of about 50 Å in diameter on the extracellular surface (56).
The emerging picture of P-gp is shown in a cross-sectional view in Fig. 4B. In this model, the drug-binding domain appears to be a "funnel" shape, with the widest point at the extracellular side and the narrowest at the cytoplasmic side. Drug binding would occur close to the center of the pore. ATP hydrolysis would change the affinity of P-gp for substrate through conformational changes in the drug-binding domain by rotation (57) and possibly large movements of helices (50). ATPase activity is inhibited if conformational changes are prevented by cross-linking of the TM segments (50,57). These movements would result in expulsion of the drug substrate and closing of the extracellular side of the pore.