Identification of the Distance between the Homologous Halves of P-glycoprotein That Triggers the High/Low ATPase Activity Switch*

Background: The mechanism for activation of P-glycoprotein ATPase activity during its reaction cycle is unknown. Results: Cross-linking the homologous halves within 6–19 Å highly activated ATPase activity but cross-linking >20 Å did not. Conclusion: ATPase activation switch is turned on when halves clamped at less than 20 Å. Significance: A key conformational switch in the P-gp reaction cycle is identified. P-glycoprotein (P-gp, ABCB1) is an ATP-binding cassette drug pump that protects us from toxic compounds and confers multidrug resistance. Each homologous half contains a transmembrane domain with six transmembrane segments followed by a nucleotide-binding domain (NBD). The drug- and ATP-binding sites reside at the interface between the transmembrane domain and NBDs, respectively. Drug binding activates ATPase activity by an unknown mechanism. There is no high resolution structure of human P-gp, but homology models based on the crystal structures of bacterial, mouse, and Caenorhabditis elegans ATP-binding cassette drug pumps yield both open (NBDs apart) and closed (NBDs together) conformations. Molecular dynamics simulations predict that the NBDs can be separated over a range of distances (over 20 Å). To determine the distance that show high or low ATPase activity, we cross-linked reporter cysteines L175C (N-half) and N820C (C-half) with cross-linkers of various lengths that separated the halves between 6 and 30 Å (α-carbons). We observed that ATPase activity increased over 10-fold when the cysteines were cross-linked at distances between 6 and 19 Å, although cross-linking at distances greater than 20 Å yielded basal levels of activity. The results suggest that the ATPase activation switch appears to be turned on or off when L175C/N820 are clamped at distances less than or greater than 20 Å, respectively. We predict that the high/low ATPase activity switch may occur at a distance where the NBDs are predicted in molecular dynamic simulations to undergo pronounced twisting as they approach each other (Wise, J. G. (2012) Biochemistry 51, 5125–5141).

The P-glycoprotein drug pump (P-gp, 2 ABCB1) was discovered over 30 years ago during efforts to determine how cancer cells developed multidrug resistance to anticancer drugs (1). Overexpression of P-gp is the most common mechanism of drug resistance when cells in culture are treated with cytotoxic agents (2). P-gp mediates the ATP-dependent efflux of a wide range of hydrophobic compounds (such as anticancer drugs, hydrophobic drugs, steroids, peptides, detergents, and lipids) that enter cells by diffusion through the plasma membrane (3)(4)(5).
The expression pattern of P-gp suggests that its physiological role is to protect us from toxins found in the diet (6). P-gp is clinically important because it can affect the bioavailability of drugs. For example, P-gp expression at the blood-brain barrier can reduce the efficacy of agents to treat epilepsy, infections (such as HIV), and brain tumors (7,8). P-gp expression has also been associated with reduced chemotherapy response rates in cancers such as myelogenous leukemia, myelodysplastic syndrome, and retinoblastoma (9 -12).
Because of its important physiological role and clinical relevance, much effort has focused on understanding the structure and mechanism of this drug pump. The 1280 amino acids of human P-gp (13) are organized as two tandem repeats of 610 amino acids that are joined by a linker region. Each repeat consists of an N-terminal transmembrane (TM) domain (TMD) containing six potential TM segments followed by a hydrophilic domain containing a nucleotide-binding domain (NBD).
Two ATP molecules bind at the interface between the NBDs, and ATP hydrolysis occurs by an alternating site mechanism (14 -18). P-gp shows basal ATPase activity that can be activated over 10-fold upon binding of drug substrates. Drug substrates appear to bind within a cavity located at the interface between the TMDs (19 -23). Drugs appear to bind by an induced-fit mechanism that may involve rotation of the TM segments (24,25).
A central feature of the drug transport mechanism is that the protein appears to undergo major conformational changes. It * This work was supported by a grant from the Canadian Institutes of Health Research (to D. M. C.). 1  alternates between an open conformation with the NBDs far apart and drug-binding cavity facing the cytoplasm and a closed conformation with the NBDs close together and drug-binding cavity facing the outside of the cell (26). Although there is no high resolution crystal structure of human P-gp, homology models have been constructed of the drug pump in the open (27) and closed (28,29) conformations based on the crystal structures of the ABC Sav1866 drug pump (30) and P-gps from mouse (29) or Caenorhabditis elegans (28). Recently, a newly refined crystal structure of mouse P-gp was reported (29). The new structure contains corrections to 669 residues that were located at least 2 Å away from the original structure (31). The refined structure of mouse P-gp was more compatible with the crystal structure of C. elegans P-gp (28). The models are consistent with the biochemical data showing close association of TM segments 2 with 11 (32), 5 with 8 (33), and 1 with 11 (34). Biochemical data have also confirmed the predicted close association of residues in NBD1 and NBD2 with the second intracellular loops of TMD2 and TMD1, respectively (35,36). Amino acids predicted to line the drug-translocation pathway were consistent with biochemical results (22,(37)(38)(39).
Do the human models of P-gp in the open and closed conformations represent snapshots of the protein in conformations with basal and activated levels of ATPase activity, respectively? Is the level of P-gp ATPase activity dependent on the degree of separation between the halves? To address these questions, we used cross-linkers of various lengths to cross-link a P-gp mutant that contained reporter cysteines at positions L175C (N-half) and N820 (C-half) in a Cys-less background. The L175C and N820C reporter cysteines were selected because they are located at positions predicted to undergo large changes in distance between the closed (12 Å between ␣-carbons) and open (about 30 Å between ␣-carbons) conformations. In addition, modifications to L175C and N820C are not expected to have a major impact on activity because they reside outside the drug-and ATP-binding sites. We previously showed that L175C/N820C could be efficiently cross-linked with the short cross-linker 1,4-butanediyl bismethanethiosulfonate (M4M) that can span a distance of 7.5 Å (about 13 Å between ␣-carbons) to activate ATPase activity (40).
Here, we tested the effects of cross-linking mutant L175C/ N820C at 10 different distances between the halves. It was found that linkage of L175C to N820C with cross-linkers predicted to span a distance of less than 20 Å highly activated P-gp ATPase activity (over 10-fold), but cross-linking with longer cross-linkers yielded P-gps with low ATPase activity.

EXPERIMENTAL PROCEDURES
Construction of Mutants-The seven endogenous cysteines at positions 137, 431, 717, 956, 1074, 1125, and 1227 were replaced with alanines to create a Cys-less P-gp (14). The Cysless P-gp cDNA was modified to contain a 10-histidine tag at the C-terminal end to facilitate purification of the expressed protein by nickel-chelate chromatography (41). Mutations L175C and N820C were introduced into Cys-less P-gp as described previously (42).
Purification of P-gp and Measurement of ATPase Activity-Histidine-tagged P-gps were expressed in HEK 293 cells and then isolated by nickel-chelate chromatography as described previously (41). Recovery of P-gp was monitored by immunoblot analysis with rabbit anti-P-gp polyclonal antibody (43). A sample of the isolated histidine-tagged P-gp was mixed with an equal volume of 10 mg/ml sheep brain phosphatidylethanolamine (type II-S, Sigma) that had been washed and suspended in TBS. ATPase activity was measured in the presence of 0.4 mM verapamil.
Cross-linking of L175C/N820C with M1M or oxidant (copper phenanthroline) was performed using P-gp isolated by nickel-chelate chromatography. The isolated P-gp was incubated 20°C for 10 min in the presence of 0.5 mM copper phenanthroline or 0.009 mM M1M. EDTA was then added to a final concentration of 2 mM. Immunoblot analysis and assay of ATPase activity were then performed as described above.
(C-half) in the models of human P-gp based on the crystal structures of Sav1866 (27) (closed structure) or open structures based on the crystal structures of P-gp from C. elegans (28) or mouse (29) are shown in Fig. 1. In the closed conformation ( Fig.  1A), the NBDs and intracellular loops (ICLs) connecting the TM segments are close together, and the predicted drug translocation pathway is closed on the cytoplasmic side. In the open conformation ( Fig. 1, B and C), NBDs and intracellular loops are separated, and the predicted drug translocation pathway is open on the cytoplasmic side. Residue Leu-175 is located in ICL1 ( Fig. 1). ICL1 appears to make relatively minor contributions to the structure and activity of P-gp because mutations to residues in ICL1 had little effect on maturation or activity of the protein (48). Residue Asn-820 is located in a region of ICL3 ( Fig.  1) that is outside an important segment (Asp-800 to Asp-806) that appears to mediate NBD-TMD contacts essential for activity and folding of P-gp (49).
In the closed structure, residues Leu-175 and Asn-820 are predicted to be 11.6 Å apart (distance between the ␣-carbons) (Fig. 1A). In the open structures, residues Leu-175 and Asn-820 are predicted to be 32.9 Å (C. elegans model) ( P-gp ATPase Activity Is Highly Activated When Residues L175C and N820C Are Cross-linked in Close Proximity-In a previous study, we found that cross-linking the two homologous halves through formation of a direct disulfide bond between A259C in intracellular helix 2 (IH2) and W803C in IH3 inhibited activity (49). It is possible that the A259C/W803C cross-link inhibited activity because these residues are located within IH segments that are predicted to form critical TMD/ NBD contact points. Because cysteines at positions L175C and N820C are located away from the IH TMD/NBD contact points ( Fig. 1), cross-linking L175C and N820C in close proximity might not disrupt TMD/NBD interactions.
The L175C or N820C mutations did not appear to affect activity of P-gp because the drug-stimulated ATPase activity of the L175C/N820C mutant was very similar to the Cys-less parent (42). Activation of P-gp ATPase activity in the presence of drug substrates appears to be a good measure of P-gp/drug interactions because there is a good correlation with drug transport (50,51). Cross-linking of mutant L175C/N820C was low (Ͻ50% cross-linking) when membranes prepared from transfected cells were treated with oxidant (copper phenanthroline) even at 37°C (data not shown). This suggested that these regions may be constrained in the membrane and that increased cross-linking efficiency might be possible after solubilization with detergent. Accordingly, histidine-tagged mutant L175C/N820C was expressed in HEK 293 cells, isolated by nickel chelate chromatography and treated with oxidant. In the presence of dodecyl maltoside detergent, mutant L175C/ N820C could be efficiently cross-linked with copper phenanthroline (Fig. 2, A and B). Cross-linking can readily be detected because cross-linking between cysteines in different domains causes P-gp to migrate slower on SDS-polyacrylamide gels (36).
As a control, we tested whether the isolated histidine-tagged mutant could be cross-linked with the short M1M (4.2 Å) crosslinker. The isolated mutant was also efficiently cross-linked when treated with the M1M cross-linker (Fig. 2, A and B). The lower efficiency of cross-linking with copper phenanthroline or M2M in membranes could be because homology models of human P-gp predict that residue Leu-175 (TM3) is not directly across from Asn-820 (TM9) (40). Solubilization of the mem-  brane may yield a less-constrained environment that allows increased efficiency of cross-linking.
To test for the effects of cross-linking on activity, isolated L175C/N820C was treated with or without 0.5 mM copper phenanthroline or 0.009 mM M1M. P-gp was then mixed with lipid and assayed for ATPase activity in the absence or presence of the drug substrate verapamil. Verapamil was selected because it is transported by P-gp (52), and it is one of the most potent activators of P-gp ATPase activity (Ͼ10-fold activation) (53). The untreated mutant showed a basal ATPase activity of 0.15 mol/min/mg P-gp that increased about 12-fold in the presence of verapamil (Fig. 2C). Cross-linking of the mutant with copper phenanthroline or M1M highly activated its basal ATPase activity by over 15-fold (Fig. 2C). Activation of mutant L175C/N820C basal ATPase activity with M1M was higher than previously observed using membranes (42) suggesting that the efficiency of cross-linking was improved using purified P-gp. The results show that cross-linking to bring the ␣-carbons at positions 175 (ICL1) and 820 (ICL3) within about 5.6 Å (54) (with oxidant) or 9.8 Å (with M1M) highly activates ATPase activity. Cross-linking with oxidant or M1M is predicted to bring the homologous halves of P-gp even closer than predicted when human P-gp was modeled in a closed conformation (11.6 Å) (Fig. 1A).

P-gp ATPase Activity Is Not Highly Activated When Residues L175C and N820C Are Cross-linked with Long Cross-linkers-
The next step was to test the effects of cross-linking with MTS cross-linkers that could span longer distances such as M2M (5.  (45). The first step was to select the optimum concentration for cross-linking because reaction of the cysteines with the cross-linkers is a competition between cross-linking and reaction of each cysteine with a separate cross-linker molecule. For example, we used 0.009 mM M1M to treat L175C/N820C because we previously showed that cross-linking efficiency was sharply reduced at lower concentrations (0.003 mM) or higher concentrations (0.027 mM) (42). Accordingly, membranes prepared from cells expressing mutant L175C/N820C were treated with various concentrations of MTS cross-linkers with various lengths of spacer arms for 10 min at 20°C, and samples were subjected to immunoblot analysis. Membranes were used instead of isolated P-gp because we previously showed that the longer cross-linkers (M2M to M17M) were substrates of P-gp (45). The presence of unbound cross-linkers in the ATPase assay reactions could make it difficult to detect changes in basal ATPase activity levels caused by cross-linking. It was found that the range of concentrations showing efficient cross-linking increased with increasing length of the cross-linker. For example, the results for the shortest (M2M) and longest (M17M) cross-linkers are shown in Fig. 3. It was found that the majority (Ͼ80%) of the mutant protein was cross-linked with 0.009 to 0.08 mM M2M, but the efficiency of cross-linking started to decrease at concentrations higher than 0.08 mM (Fig. 3, A and B). Little crosslinking was observed with 2.2 mM M2M (Fig. 3, A and B). By contrast, the mutant was efficiently cross-linked at all concentrations of M17M greater than 0.009 mM (Fig. 3, C and D). The other cross-linkers also showed maximal cross-linking at a concentration of 0.027 mM (data no shown).
Membranes prepared from cells expressing mutant L175C/ N820C were first treated for 10 min at 20°C in the absence or presence of 0.027 mM MTS cross-linkers of different lengths. Immunoblot analysis of samples of the treated membranes showed that the mutant was efficiently cross-linked with all the cross-linkers (Fig. 4, A and B).
Because mutant L175C/N820C was efficiently cross-linked with the MTS cross-linkers, we tested whether cross-linking affected activity. Accordingly, histidine-tagged P-gp was isolated by nickel-chelate chromatography after cross-linking membranes and assayed for ATPase activity in the presence or absence of saturating concentrations (0.4 mM) of verapamil. Fig. 5 shows that untreated mutant L175C/N820C isolated from membranes exhibited about a 12-fold increase in verapamil-stimulated ATPase activity. In the absence of verapamil, its basal ATPase activity was 0.14 mol of P i /min/mg of P-gp. In the presence of verapamil, its activity increased to 1.75 mol of P i /min/mg of P-gp. By contrast, treatment of mutant L175C/ N820C with the intermediate size MTS cross-linkers (M2M, M4M, M5M, M6M, and M8M) that span maximum distances of about 5.2-13 Å (distance of about 10.6 -18.6 Å between the ␣-carbons) highly activated ATPase activity in the absence of verapamil. The mutants treated with intermediate size crosslinkers had increased basal ATPase activities (2.30 -2.50 mol of P i /min/mg of P-gp) that were at least 15-fold higher than the basal ATPase activity of the untreated mutant (0.14 mol of P i /min/mg of P-gp) (Fig. 5). Activity was not increased any further in the presence of verapamil. It appeared that cross-linking of mutant L175C/N820C with the intermediate size cross-linkers locked P-gp in an activated state resulting in high levels of ATPase activity in the absence or presence of drug substrates.
A quite different effect was observed when mutant L175C/ N820C was treated with the long cross-linkers M11M, M14M, and M17M that could span maximum distances of about 16.9, 19, and 22 Å in length, respectively (span distances of about 22 to 28 Å between the ␣-carbons) (Fig. 5). Treatment with the long cross-linkers caused only small increases in basal ATPase activity (about 2-3-fold; 0.25-0.40 mol of P i /min/mg of P-gp), but the activity of the cross-linked mutants could be activated 8 -10-fold higher when the samples were assayed in the presence of verapamil (2.65, 3.7, and 3.40 mol of P i /min/mg of P-gp for M11M, M14M, and M17M, respectively).
Cross-linking of mutant L175C/N820C with the intermediate size MTS cross-linkers like M2M increased the basal ATPase activity of the mutant (Fig. 5). It was possible that the increase in activity of the mutant after treatment with MTS cross-linkers could be due to other factors such as activation by nonspecific modification of other amino acids or modification of only one or both of the introduced cysteines by individual cross-linker molecules. To test these possibilities, a series of control experiments were carried out with L175C/N820C using M2M. The M2M cross-linker was selected because it efficiently cross-links at low concentrations (0.009 -0.08 mM) but causes little cross-linking when used at a high concentration (2.2 mM) (Fig. 4A). This may be because a separate cross-linker molecule labels each cysteine.
We compared the effects of treating mutant L175C/N820C with low (0.027 mM; promotes cross-linking) or high (3 mM; likely labels each cysteine with a separate M2M molecule) concentrations of M2M on ATPase activity. Membranes containing the mutant were treated with low or high concentrations of cross-linker M2M. Histidine-tagged P-gp was then isolated by nickel-chelate chromatography and mixed with lipid, and ATPase activity was determined in the absence or presence of 0.4 mM verapamil. Fig. 6A shows that treatment of the mutant with 0.027 M2M activated ATPase activity, although treatment with 3 mM M2M cross-linker yielded ATPase activities that were similar to that of untreated control. These results suggest that only cross-linking of the mutant caused activation of its basal ATPase activity.
Next, mutant histidine-tagged P-gps were constructed that contained only the L175C or N820C mutation alone to test if modification of only one cysteine had an impact on ATPase activity. Membranes prepared from cells expressing the single cysteine mutants were treated with 0.027 mM M2M. The P-gps were isolated by nickel-chelate chromatography and assayed for ATPase activity in the absence or presence of verapamil. It was found that the activities were similar to those of untreated mutant (Fig. 6B). In addition, no cross-linked product was observed when membranes containing Cys-less P-gp, the single cysteine mutants expressed individually or expressed together  in the same cell, were treated with M2M (data not shown). These results show that activation of mutant L175C/N820C basal ATPase activity by M2M was due to cross-linking between the two cysteines.

DISCUSSION
Double electron-electron resonance spectroscopy studies and molecular dynamics simulations (55) show that the NBDs and cytoplasmic loop NBD/TMD connection regions can adopt a wide range of conformations with separation distances that cover a range of at least 20 Å. The wide range of conformations was observed in both the presence and absence of Mg-ATP and when P-gp was incorporated into lipid. The observations that a wide range of distances were observed in the presence of lipid and ATP were important because there had been concerns that the open conformation may only be observed with solubilized P-gp in the absence of ATP (42,56,57) and that the NBDs may remain in constant contact throughout the reaction cycle (58).
In agreement with the EPR and simulation studies (55), we found that the presence of ATP did not promote the formation of a more closed structure as it did not promote L175C/N820C cross-linking (42). We also found that verapamil did not promote or inhibit cross-linking of L175C/N820C (42). Therefore, it does not appear that verapamil promotes ATPase activity by inducing formation of a closed state. Formation of a closed state alone might not be sufficient to activate ATPase activity because no increase in basal ATPase activity was observed when the C-terminal ends of homologous halves of mouse P-gp were cross-linked together (59). Verapamil might activate ATPase activity because it promotes interactions between the catalytic regions of the NBDs, thereby increasing the probability of generating the ATP-bound sandwich conformation needed for ATP hydrolysis. Conversely, inhibitors that reduce interactions between the NBDs inhibit ATP hydrolysis (40,60).
The experiments presented here show that human L175C/ N820C P-gp remains active when it is trapped in conformations with a wide range of separations between the halves (about 6 -30 Å). An unexpected finding was that the cross-linked mutants did not show a range of basal ATPase activities that was dependent on the degree of separation of the two halves. Instead, we found that the cross-linked mutants could be divided into two categories (Fig. 7). The first category was composed of mutants that were cross-linked at predicted distances of 6 -19 Å (oxidant, M1M, M2M, M4M, M5M, M6M, and M8M) and showed about 10-fold activation in basal ATPase activity after cross-linking. The second category was composed of mutants cross-linked with long (M11M, M14M, and M17M) cross-linkers (distances greater than 20 Å) and did not show highly activated basal ATPase activity.
The ATPase activities of mutants in the second category resembled the untreated mutant because they could still be activated by about 10-fold using verapamil. Molecular dynamic simulation studies have shown that the longer MTS cross-linkers are flexible and can span a range of distances such as 4.9 -7.8 Å for M4M, 5.7-10.4 Å for M6M, 3.9 -13 Å for M8M, 3.8 -16.9 Å for M11M, 3-19 Å for M14M, and 3-22 Å for M17M (61,62). Because the longer cross-linkers are flexible, it might be expected that cross-linked mutants would resemble the uncross-linked parent if they could still undergo a similar range of motions. For example, similar cross-linking studies were performed on cysteine mutants of lactose permease (LacY) (62). Pairs of cysteines were introduced to span the predicted transport cavity, and the mutants were treated with MTS cross-linkers of various lengths. It was found that cross-linking with cross-linkers that spanned 10 Å or less (M3M and M6M) inhibited activity, whereas the mutants remained fully active when cross-linked with cross-linkers that could span distances greater than 17 Å (M14M and M17M). Partial activity was observed when the LacY mutants were cross-linked with crosslinkers predicted to span 11-15 Å (M8M and M11M). Their results suggested that the maximum opening of the cavity was   (10)) was compared with that treated with copper phenanthroline (1). 17 Å, a distance close to that determined by double electronelectron resonance (63). The LacY studies suggest that distance measurements from cross-linking experiments are compatible with other biochemical approaches.
Although the lac permease studies showed that cross-linking with cross-linkers of various lengths showed a gradation of effects on activity depending on the length of the cross-linker, the results on P-gp showed a sharp change at a distance of about 20 Å (Fig. 7). A distance of 20 Å would be about midway between the open and closed states. A molecular dynamics simulation study of human P-gp by Wise (64) suggests that one important conformational change that may occur about halfway between the transition between the open and closed conformations is twisting of the NBDs. When the distances between the catalytic carboxylates (556 and 1201) were monitored, it was found that the two residues remained about 30 Å apart during the transition from wide open to about midway between the open and closed conformations. As the protein then transitioned to a more closed conformation, there was a sharp reduction in the distance separating the two residues corresponding to a pronounced twisting of the NBDs as they approached each other. Residues 556 and 1201 are separated by about 15 Å in the closed conformation (64). Perhaps the 20-Å trigger length identified in the cross-linking experiment corresponds to this sharp conformational shift where residues 556 and 1201 begin to approach each other due to twisting of the NBDs. When the halves are linked at distances less than 20 Å, the NBDs may come together more frequently to align the ATP ␥-phosphate and hydrolytic water within the Walker B, LSGGQ, Q-loop, D-loop, and H-loop network for ATP hydrolysis (65).
The P-gp molecular dynamics simulation study (64) may also provide an explanation for the inhibition of ATPase activity previously observed when cysteines A295C (in IH2 that connects TM segments 4 and 5) and W803C (in IH3 that connects TM segments 9 and 10) were directly cross-linked (49). It was found that straightening of TM segments 4 and 5 occurs during the transition from the open to closed conformational change. It is possible that direct cross-linking of A295C (in IH2 that connects TM segments 4 and 5) to W803C in IH3 would interfere with straightening of TM segments 4 and 5 to trap P-gp in an inactive conformation. By contrast, most movement between the intracellular ends of TM segments 3 and 9 is predicted to only occur during the very early stages of the open to closed conformational change (64).
Could P-gp still transport drug substrates when L175C/ N820C is cross-linked with short (M2M to M8M) cross-linkers? Unfortunately, it was not possible to use whole cell transport or drug resistance assays because cross-linking could not be detected when cells expressing the L175C/N820C mutant were treated with the MTS cross-linkers. 3 The lack of crosslinking in whole cells could be due to the presence of intracellular thiol reductants such as glutathione that would not only react with the MTS cross-linkers but also reduce any disulfide bond. Similarly, when purified P-gp was reconstituted into lipid and treated with MTS cross-linkers, the amount of cross-linked product was very low (Ͻ5%; data not shown).
We predict, however, that P-gp cross-linked at L175C/ N820C with short cross-linkers should be able to bind substrates with relatively high affinity because it was reported that P-gp trapped in the closed vanadate-trapped transition state had similar affinity for the drug substrates verapamil, Hoechst 33342, cyclosporin A, vinblastine, colchicine, rhodamine B, and doxorubicin as in the relaxed (open) state (66). It was suggested that release of drug substrate from the transporter occurred before formation of the transition state (66). Cross-linked P-gp might have altered drug transport because studies on the bacterial ABC transporters MJ0796 (67) and MsbA (68) suggest that the NBDs completely dissociate during the hydrolysis cycle.
In summary, activation of ATPase activity in human P-gp appears to involve a high/low switch that occurs when cytoplasmic ends of ICL1 and ICL3 are separated by about 20 Å.