Drug-stimulated ATPase Activity of Human P-glycoprotein Requires Movement between Transmembrane Segments 6 and 12*

Transmembrane segments (TM) 6 and 12 are directly connected to the ATP-binding domain in each homologous half of P-glycoprotein and are postulated to be important for drug-protein interactions. Cysteines introduced into TM6 (L332C, F343C, G346C, and P350C) were oxidatively cross-linked to cysteines introduced into TM12 (L975C, M986C, G989C, and S993C, respectively). The pattern of cross-linking was consistent with a left-handed coiled coil arrangement of the two helices. To detect conformational changes between the helices during drug-stimulated ATPase activity, we tested the effects of substrates and ATP on cross-linking. Cyclosporin A, verapamil, vinblastine, and colchicine inhibited cross-linking of mutants F343C/M986C, G346C/G989C, and P350C/S993C. By contrast, ATP promoted cross-linking between only L332C/L975C. Enhanced cross-linking between L332C/L975C was due to ATP hydrolysis, since cross-linked product was not observed in the presence of ATP and vanadate, ADP, ADP and vanadate, or AMP-PNP. Cross-linking between P350C/S993C inhibited verapamil-stimulated ATPase activity by about 75%. Drug-stimulated ATPase activity, however, was fully restored in the presence of dithiothreitol. These results show that TM6 and TM12 undergo different conformational changes upon drug binding or during ATP hydrolysis, and that movement between these two helices is essential for drug-stimulated ATPase activity.

P-glycoprotein, the product of the human multidrug resistance (MDR1) gene, is a plasma membrane glycoprotein that extrudes a broad range of cytotoxic agents from cells (reviewed in Refs. 1 and 2). It belongs to the ABC (ATP-binding cassette) superfamily of membrane proteins. Members of this family of proteins generally have two ATP-binding domains and two hydrophobic domains consisting of six potential transmem-brane segments. In human P-glycoprotein, all four domains are within a single polypeptide of 1280 amino acids organized in two tandem repeats; each repeat consists of a hydrophobic domain followed by an ATP-binding domain (3). This predicted structure is consistent with the results of topology studies (4 -6). The minimum functional unit of the enzyme appears to be a monomer (7).
Many approaches have been used to study the mechanism of P-glycoprotein. Mutational analysis and chemical inhibitors have been used to show that both ATP-binding domains are essential for activity (8 -11). Photolabeling studies with analogs of drug substrates and the results of mutational analysis suggest that the drug-binding sites(s) appear to reside within the transmembrane domains (12)(13)(14)(15)(16)(17)(18)(19)(20). The photolabeling sites are closely associated with TM6 1 and TM12.
TM6 and TM12 may be particularly important for drug transport because they directly connect the two transmembrane domains to their respective ATP-binding domains. Both transmembrane segments may interact and undergo essential conformational changes during drug-binding or during ATP hydrolysis. To test this hypothesis, we replaced the residues in TM6 and TM12 with cysteine and tested for the effect of drug substrates and ATP on the oxidative cross-linking between TM6 and TM12. The pattern of cross-linking was consistent with the TM6 and TM12 helices arranged as a left-handed coiled coil, and cross-linked pairs were affected differently by substrates or ATP. We also show that TM6 and TM12 must undergo conformational changes during drug-stimulated ATPase activity, since formation of a cross-link between TM6 and TM12 inhibited drug-stimulated ATPase activity. Activity could be restored by breaking the disulfide bond with dithiothreitol.

EXPERIMENTAL PROCEDURES
Construction of Mutants-Cysteine residues were introduced into a Cys-less mutant of P-glycoprotein containing a histidine tag at the COOH terminus as described previously (4,9,14). The presence of a histidine tag facilitated purification of the mutant P-glycoprotein by nickel-chelate chromatography (9).
Sulfhydryl Cross-linking with Copper Phenanthroline-For each mutant, ten 10-cm-diameter culture plates of HEK 293 cells were transfected with mutant MDR1 cDNA. After 24 h, the media were replaced with fresh media containing 10 M cyclosporin A. The cells were harvested after another 24 h, and membranes were prepared as described previously (9). The membranes were suspended in 200 l of Trisbuffered saline. A sample of the membrane suspension (12 l) was mixed with 15 l of 2 ϫ ATPase buffer (100 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl 2 ) containing the desired nucleotide or drug substrate for 5 min at room temperature. Cross-linking was initiated by addition of 3 l of 2 mM or 20 mM Cu 2ϩ (phenanthroline) 3 , and the samples were incubated for 10 min at 37°C (22). The reactions were stopped by addition of EDTA to a final concentration of 30 mM in SDS sample buffer containing no disulfide reducing agent. The samples were subjected to SDS-PAGE, transferred onto a sheet of nitrocellulose, and probed with a rabbit polyclonal antibody against P-glycoprotein and enhanced chemiluminescence (22).
Purification of Cross-linked P-glycoprotein Mutant and Measurement of Verapamil-stimulated ATPase Activity-Purification was carried out as described previously (9). Briefly, forty 10-cm diameter culture plates of HEK 293 cells were transfected with the mutant MDR1 cDNA, followed by addition of cyclosporin A as described above. Membranes were then prepared and cross-linked with Cu 2ϩ (phenanthroline) 3 (2 mM final concentration) for 10 min at 37°C. The cross-linked sample was then diluted 100-fold with Tris-buffered saline and centrifuged at 200,000 ϫ g for 2 h at 4°C. The membranes were then solubilized with 1% (w/v) n-dodecyl-␤-D-maltoside (Sigma) and P-glycoprotein-(His) 10 isolated by nickel-chelate chromatography using nickel-spin columns (Ni-NTA, Qiagen).
To measure ATPase activity, the purified P-glycoprotein was diluted with an equal volume of 100 mg/ml crude sheep brain phosphatidylethanolamine (Sigma, Type IIs commercial grade) that had been previously washed with Tris-buffered saline to remove traces of phosphate and then sonicated. 100 ng of purified P-glycoprotein was incubated with 1 mM verapamil, and ATPase activity was initiated by addition of an equal volume of buffer containing 100 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl 2 , 10 mM ATP, with or without 10 mM DTT. The samples were incubated at 37°C, and the amount of inorganic phosphate liberated was determined by the method of Chifflet et al. (23).

Mutagenesis and Oxidative Cross-linking-Predicted TM6
and TM12 directly connect the transmembrane domains to nucleotide-binding fold NBF1 and NBF2, respectively. A relatively high degree of homology exists between TM6 and TM12 since 11 of 21 residues are identical, when both transmembrane segments are aligned. In a previous study, we showed that residues 332 (TM6) and 975 (TM12) are close to each other in the tertiary structure of P-glycoprotein (22). Using this observation as a starting point, we modeled the two helices as left-or right-handed coiled coils (reviewed in Refs. 24 -26) to predict other sites in TM6 and TM12 that may be close to each other. Fig. 1, A and B, shows the arrangement of the residues in TM6 and TM12 as ␣-helical nets, while C shows the imposition of TM6 on TM12 in a left-handed coiled coil. According to Chothia (26), the helices have ridges (consisting of the side chains of residues) and grooves between the side chains. The ridges are formed by the residues that are spaced 1, 3 (i, i ϩ 3), or 4 (i, i ϩ 4) apart. During packing in a right-handed coiled coil, the i ϩ 4 axes in each helix are superimposed, whereas in a left-handed coiled coil, the i ϩ 3 axis of one helix is superimposed on the i ϩ 4 axis of the other. Both left-and right-handed coiled coils are equally common in globular proteins, but lefthanded coiled coils appear to be more common in the membrane proteins whose crystal structures have been determined (27,28). Modeling of TM6 and TM12 in a left-handed coiled coil showed the pairs of residues between TM6 and TM12 (F336/ S979, L339/V982, F343/M986, G346/G989, and P350/S993) that are likely to be close together. To test these predictions, we introduced pairs of cysteines into a Cys-less mutant of P-glycoprotein to create the mutants F336C/S979C, L339C/V982C, F343C/M986C, G346C/G989C, and P350C/S993C. The mutants were expressed in HEK 293 cells in the presence of cyclosporin A. We have previously shown that Cys-less P-glycoprotein retains activity but folds less efficiently during biosynthesis compared with wild-type enzyme (4). Expression of the Cys-less mutant in HEK 293 cells results in about equal amounts of mature (170-kDa) and core-glycosylated (150-kDa) forms of the enzyme, whereas in the wild-type enzyme, the mature form of P-glycoprotein is the major product. We recently showed that P-glycoprotein mutants defective in protein kinesis or that are slow in exiting the endoplasmic reticulum could be rescued by carrying out biosynthesis in the presence of substrates or modulators (29). This approach was also effective for the Cys-less P-glycoprotein. When expressed in the presence of cyclosporin A, the majority of Cys-less P-glycoprotein was in the form of the mature enzyme (for example, see Fig. 1D,  lane 3).
The membrane fractions containing each mutant P-glycoprotein were treated with 2 mM copper phenanthroline for 10 min at 37°C, followed by immunoblot analysis. Fig. 1D shows that a product with reduced mobility on SDS-PAGE gels was present when mutants F343C/M986C, G346C/G989C, and P350C/ S993C were treated with oxidant. The highest level of crosslinking occurred in mutant P350C/S993C, since most of the protein migrated with reduced mobility after treatment with oxidant (Fig. 1D, lane 14). No cross-linked product was observed for mutants F336C/S979C and L339C/V982C. Similarly, P-glycoproteins containing only a single cysteine mutation did not yield any product with reduced mobility in SDS-PAGE gels (data not shown). We also tested mutants F335C/L976C, L339C/S979C, F343C/F983C, G347C/A987C, and S351C/ V991C for cross-linking since they were predicted to lie on opposing faces of TM6 and TM12 modeled in a right-handed coiled-coil. None of these mutants, however, were cross-linked in the presence of oxidant.
Effect of Nucleotides and Drug Substrates on Cross-linking-An interesting observation was that the amount of crosslinking seen in mutant L332C/L975C in whole cells (9) varied with the metabolic state of the transfected cells. Cross-linking was most efficient with freshly isolated cells and decreased considerably if the cells were starved before cross-linking (data not shown). A possible explanation was that cross-linking of mutant L332C/L975C was promoted by the presence of ATP. To test for this possibility, membranes from HEK 293 cells ex- pressing mutant L332C/L975C were cross-linked in the presence of nucleotides. Fig. 2A (lane 3) shows that cross-linking of the mutant occurred in the presence of ATP. No cross-linked product was observed in the presence of ATP plus vanadate, ADP, ADP plus vanadate, or with the nonhydrolyzable ATP analog, AMP-PNP ( Fig. 2A, lanes 4 -7). These results suggest that cross-linking between L332C and L975C occurred during ATP hydrolysis.
The effect of nucleotides on cross-linking was also tested on mutants F343C/M986C, G346C/G989C, and P350C/S993C. For these mutants, cross-linking was carried out with 10-fold less oxidant (0.2 mM) to detect for any subtle effects of the nucleotides. Fig. 2 (panels B, C, and D) shows that the presence of nucleotides had little detectable effect on cross-linking.
To test the effect of drug substrates, cross-linking of mutants L332C/L975C, F343C/M986C, G346C/G989C, and P350C/ S993C was done in the presence of verapamil, cyclosporin A, vinblastine, or colchicine. No cross-linked product was observed for mutant L332C/L975C (Fig. 3A). By contrast, all the drug substrates were effective in blocking cross-linking of mutants F343C/M986C and G346C/G989C (Fig. 3, B and C), but were less effective in preventing cross-linking of mutant P350C/S993C (Fig. 3D). In mutant P350C/S993C, verapamil, cyclosporin A, and vinblastine were more effective than colchicine in inhibiting cross-linking. Mutants S979C/F336C or L339C/V982C did not yield any cross-linked product even in the presence of ATP or drug substrates (data not shown).
Effect of Cross-linking on Drug-stimulated ATPase Activity-Mutants L332C/L975C, F343C/M986C, G346C/G989C, and P350C/S993C were still active since they retained about 90, 30, 10, and 70%, respectively, of the verapamil-stimulated ATPase activity of the Cys-less P-glycoprotein. Cross-linking of mutants F343C/M986C, G346C/G989C, and P350C/S993C, but not L332C/L975C, was reversed by treatment with dithiothreitol (Fig. 4A). Therefore, it was of interest to determine whether cross-linking between TM6 and TM12 would inhibit drug-stim-ulated ATPase activity and whether this could be reversed in the presence of dithiothreitol. Mutant P350C/S993C was tested because it exhibited the greatest degree of cross-linking (Fig.  1D). In addition, mutant P350C/S993C had the highest activity of those mutants whose cross-linked products were sensitive to dithiothreitol. Membranes prepared from HEK 293 cells expressing Cys-less or mutant P350C/S993C P-glycoprotein-(His) 10 were treated with or without 2 mM copper phenanthroline for 10 min at 37°C. The reaction was stopped by addition of EDTA to a final concentration of 30 mM, and the membranes were diluted with TBS and recovered by centrifugation. Fig. 4B shows that treatment with oxidant did not affect recovery of P-glycoprotein(His) 10 by nickel-chelate chromatography. Fig.  4B (lane 4) shows that the cross-linked mutant P350C/S993C was also efficiently recovered, indicating that the histidine tag remained accessible after cross-linking.
Equal amounts of purified P-glycoprotein were assayed for verapamil-stimulated ATPase activity. Fig. 4C shows that the activity of the Cys-less mutant was not affected by treatment with oxidant or by including DTT during the ATPase assay. Mutant P350C/S993C, however, showed approximately 75% reduction of the verapamil-stimulated ATPase activity after cross-linking. The activity was fully restored when the crosslinked mutant was assayed in the presence of 5 mM DTT. This result suggested that dithiothreitol broke the disulfide bond between residues P350C and S993C and allowed movement to occur between TM6 and TM12 during drug-stimulated ATPase activity.

DISCUSSION
The results of cross-linking experiments suggest that TM6 and TM12 are close to each other along the entire lengths of the helices and are likely to exist in a left-handed coiled-coil arrangement. In this arrangement four of the six pairs of amino acids predicted to lie close to one another could be cross-linked. Cross-linking was not observed between F336C/S979C or L339C/V982C, even in the presence of ATP or drug substrates (data not shown). One possibility is that these residues are not close. Another possibility is that they are close, but in a nonreactive environment or that they are inaccessible to oxidant.
Cross-linking was significantly influenced by the presence of ATP or drug substrates. ATP hydrolysis rather than nucleotide binding was responsible for cross-linking between L332C/ L975C. In a previous study (22), we observed in intact cells that vinblastine or verapamil, but not colchicine, could inhibit crosslinking between these two residues. The observation that ATP hydrolysis promoted cross-linking between L332C/L975C suggests that inhibition of cross-linking of L332C/L975C in whole cells by verapamil or vinblastine occurred indirectly through depletion of intracellular ATP. Vinblastine and verapamil had a greater effect than colchicine since these two compounds are more efficient in stimulating the ATPase activity of P-glycoprotein, thereby depleting the ATP more rapidly.
Drug substrates inhibited cross-linking of mutants F343C/ M986C, G346C/G989C, and P350C/S993C. These residues either lie close to the binding site(s) for these substrates or drug-binding results in large conformational changes in TM6, TM12, or in both transmembrane segments. Conformational changes occurring globally in P-glycoprotein during substrate binding or during ATP hydrolysis have also been detected indirectly (21,30).
The results of this study show that there is "cross-talk" between the transmembrane domains and the ATP-binding domains of P-glycoprotein. ATP hydrolysis causes conformational changes in the transmembrane domains, while introduction of a cross-link between TM6 and TM12 inhibits drugstimulated ATPase activity. Therefore, conformational changes occurring between TM6 and TM12 appear to be essential for coupling drug binding to stimulation of ATPase activity. A similar mechanism may exist in other ABC transporters.