ATP Alone Triggers the Outward Facing Conformation of the Maltose ATP-binding Cassette Transporter*

Background: The conformational cycle of MalFGK2 is stimulated by ATP, MalE, and maltose. Results: The transporter outward facing conformation depends on ATP only. Conclusion: MalE and maltose must therefore stimulate the return of the transporter to the inward facing state. Significance: Understanding the dynamics of the transporter is important to interpret the crystal structures. The maltose transporter MalFGK2 is a study prototype for ABC importers. During catalysis, the MalFG membrane domain alternates between inward and outward facing conformations when the MalK dimer closes and hydrolyzes ATP. Because a rapid ATP hydrolysis depends on MalE and maltose, it has been proposed that closed liganded MalE facilitates the transition to the outward facing conformation. Here we find that, in contrast to the expected, ATP is sufficient for the closure of MalK and for the conversion of MalFG to the outward facing state. The outward facing transporter binds MalE with nanomolar affinity, yet neither MalE nor maltose is necessary or facilitates the transition. Thus, the rapid hydrolysis of ATP observed in the presence of MalE and maltose is not because closed liganded MalE accelerates the formation of the outward facing conformation. These findings have fundamental implications for the description of the transport reaction.

ATP-binding cassette (ABC) 3 transporters are found both in prokaryotes and eukaryotes and transport a wide variety of substrates in and out of cells. The understanding of their mode of operation is relevant for medically important processes such as multidrug resistance, chloride conductance, cholesterol transport, and surface-antigen presentation (1). In bacteria, the maltose transporter has been studied for several decades as the prototype of ABC importers (2)(3)(4). The transporter is comprised of a cytosolic homodimeric ABC domain MalK bound to a membrane domain heterodimer, MalFG. On its periplasmic side, the transporter associates with the substrate-binding protein MalE that acts as a receptor for maltose.
Major progress in the description of ABC transporters has been achieved by x-ray crystallography (5)(6)(7). In the structure of MalFGK 2 without ligands, the sites for ATP binding on the MalK dimer are separated, and the MalFG membrane domain forms a cavity that is exposed to the cytosol (inward facing state) (8). In the structure obtained in the presence of MalE and nonhydrolyzable ATP analogs, the ATP-binding sites are bound together, whereas the MalFG cavity opens toward the periplasm (outward facing state) (6,9). Together, these snapshots of the transporter structure have provided the molecular framework to explain how ATP binding and hydrolysis is coupled to membrane transport, the so-called alternating access mechanism. However, what triggers the transport reaction, why MalE stimulates ATP hydrolysis even in the absence of maltose, and how maltose is transferred from MalE to MalFG remain unclear.
MalE consists of two lobes with the maltose-binding site located at the interface. When the sugar binds, the two lobes close and eventually capture maltose with a high affinity (conformation termed closed liganded MalE) (10,11). Because the ATPase activity of the transporter is maximal only in the presence of MalE and maltose, it has been proposed that closed liganded MalE binds the transporter and triggers ATP hydrolysis (12,13). The x-ray structures were interpreted with this assumption in mind, leading to the model in Fig. 1A (2). In that model, the binding of closed liganded MalE to the transporter induces the pairing of MalK. The pairing of MalK triggers the MalFG outward facing conformation, which in turn forces the opening of MalE and the release of maltose. Upon ATP hydrolysis, the transporter returns to the inward facing state, and maltose enters the cytosol. In support of the model, an EPR spectroscopy study reported that the MalK nucleotide-binding interface closes only in the presence of MalE and ATP (14). A partial closure of the MalK interface was also observed when MalFGK 2 was co-crystallized with closed liganded MalE. It was concluded that MalE facilitates the pairing of MalK and therefore hydrolysis of ATP (15). However, despite the evidence and the prevalence of the model, we recently reported that closed liganded MalE does not bind the transporter but dissociates when MalE captures maltose. We also reported that open state MalE binds with nanomolar affinity to the outward facing transporter and creates a low affinity receptor for maltose (16). These data, difficult to reconciliate with the former model, led us to propose a new model ( Fig. 1B and legend for details about the binding affinities).
Here, we further investigate the effect of MalE, maltose, and ATP on the conformation of MalFGK 2 . Indeed, an important and central question is the mechanism leading to the outward facing conformation. The transition depends on MalE and maltose in the former model, but not in the newer model. Using cross-linking, co-sedimentation analysis, native gels, and fluorescence assays, we show that MalE and closed liganded MalE do not facilitate the closure of MalK dimer, nor the transition of MalFG to the outward facing state. Instead, we find that ATP alone is sufficient. Because MalE does not facilitate the binding of ATP to the transporter (16,17), nor the outward facing conformation, the results imply that MalE must stimulate a step occurring during the return of the transporter to the inward facing state.
Maltose Transport Assay-Proteoliposomes were reconstituted in R buffer (50 mM sodium phosphate, pH 6.5, 1 mM MgCl 2 , 5 mM ATP). The transport reaction was carried out with 10 M MalE, 0.5 M MalFGK 2 , and 20 M [ 14 C]maltose (57 Ci/mol). At the indicated time, samples (50 l) were removed and diluted into 1 ml of ice-cold 50 mM sodium phosphate buffer containing 10 mM maltose followed by filtration through a 0.22-m nitrocellulose Millipore filter and scintillation counting. The number of cpm were converted into nmol using known amounts of [ 14 C]maltose.
Cysteine Cross-linking Experiments-Cross-linking reactions were performed in 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 containing 0.5 M MalFGK 2 proteoliposomes mixed with 1 mM ATP, 10 M MalE, and 1 mM maltose as indicated. The samples were incubated with 50 M BMOE (10 min at room temperature) and then treated with N-ethylmaleimide (5 mM) before analysis by SDS-PAGE.
Fluorescence Labeling and Spectroscopy-The dye N-(1-pyrene)-maleimide was used to label the MalFGK 2 cysteine mutants. Briefly, the purified MalFGK 2 complex was incubated  with 0.5 mM DTT (30 min on ice) and passed through a G25 column. The labeling reaction was then performed by mixing MalFGK 2 and N-(1-pyrene)-maleimide at a molar ratio of 1:10 in TSG buffer containing 0.01% DDM and 1 mM TCEP for 2 h at room temperature. The unreacted N-(1-pyrene)-maleimide was removed by Superdex 200 10/30 gel filtration. The purified and labeled complex was reconstituted in proteoliposomes as described above. The labeling efficiency was determined by absorbance spectroscopy (343 nm) using an extinction co-efficient of 40,000 cm Ϫ1 M Ϫ1 . A typical ratio was between 1.2 and 1.4. Because there are two cysteine substitutions per MalFGK 2 variant, this is equal to ϳ60 -70% labeling efficiency. The pyrene-labeled MalFGK 2 variants retained ϳ80% ATPase activity and 70% transport activity of wild type MalFGK 2 (see Table 2). The fluorescent measurements were carried out at 25°C in FL buffer (50 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 ) on a Cary Eclipse spectrofluorometer. The fluorescence spectrums were recorded from 365 to 520 nm with excitation wavelength at 345 nm (5-nm slide width). The binding of MalE to MalFGK 2 was measured by fluorescence quenching assay, using MalE-31C labeled with the maleimide derivative ATTO-655 (Atto-Tech), as previously described (16). The fluorescent measurements were carried out at 25°C in FL buffer.
Other Methods-The transporter ATPase activity was determined by measuring the release of inorganic phosphate using photo colorimetric method (16). Linear gradient native gel electrophoresis, sedimentation assay, and protein iodination were performed as described (16,19).

ATP Controls the Outward Facing Conformation of the
Transporter-Cysteine residues were introduced at diagnostic positions on MalK, MalF, and MalG, so that sulfhydryl-based cross-linking methods could be employed to monitor the conformation of the transporter in the lipid bilayer ( Fig. 2A). The cysteine residues were introduced into a complex mutated at the endogenous Cys-40 of MalK because this position naturally reacts with maleimide reagents (data not shown). According to the crystallographic data (8,9), the distance between the periplasmic positions MalF 446 /MalG 234 increases from ϳ9 to ϳ18 Å during the conformational transition from the inward facing state to the outward facing state. Meanwhile, the distance between the cytosolic positions MalK 83 /MalK 83 and MalF 396 /MalG 185 decreases from ϳ 21 to ϳ8 Å. We thus employed the homobifunctional cross-linker BMOE with a spacer arm of 8 Å to monitor these conformational changes (Fig. 2).
We note that the C-terminal tail of MalG in the crystal structure is located near the MalK dimer interface and may potentially interfere with the formation of the cross-link MalK 83 / MalK 83 . However, as reported by others, an efficient ATPdependent cross-linking occurs with this pair (and also with MalK 85 /MalK 85 ), showing that the C-terminal tail of MalG is sufficiently mobile in solution to allow cross-link formation (20,21). We also considered the possibility that MalFGK 2 becomes inaccessible to MalE when the transporter is reconstituted in liposomes. This was ruled out because (i) the transporter was functional (Table 1), and its activity was dependent on MalE and maltose, (ii) the sedimentation of MalE with the proteoliposome was dependent on MalFGK 2 and nucleotides (Fig. 3, A and C), and (iii) the thiol-reactive membrane-impermeable reagent 5-IAF labeled the periplasmic side of the complex with at least 50% efficiency (Fig. 2B).
We then probed what the transporter conformation is in the membrane. In the absence of ligands (apo state), there was only a little cross-link formation with the pairs MalK 83 /MalK 83 and MalF 396 /MalG 185 (pairs located in the cytosolic side of the membrane), whereas cross-links readily formed with the pair MalF 446 /MalG 234 (located on the periplasmic side) (Fig. 2,  C-E). These results were expected because the transporter resides naturally in the inward facing state in the absence of ligands (8). In the presence of ATP, however, a predominant cross-link appeared with the pairs MalK 83 /MalK 83 and MalF 396 /MalG 185 , whereas the cross-link of MalF 446 /MalG 234 was decreased by ϳ90%, as determined by densitometry analysis (data not shown). The binding of ATP, and not its hydrolysis, was responsible for this conformational change because ATP␥S produced the same effect (Fig. 2F). Strikingly, the addition of MalE with or without maltose did not cause any differences in the cross-linking efficiencies (Fig. 2, C-F). Together, these data provided strong preliminary evidence that ATP on its own controls the transition of the transporter from inward to outward facing state.
ATP Controls the Binding of MalE to the Transporter-We reported recently that MalE binds with nanomolar affinity to the outward facing transporter (16). Because ATP alone appears sufficient for the transition to the outward facing conformation, we hypothesized that ATP alone should also control the binding of MalE to the transporter. Accordingly, the stabilization of MalFGK 2 in the outward facing state with ATP plus vanadate allowed an efficient co-sedimentation of MalE with the transporter (Fig. 3A, lanes 2). In the absence of vanadate, the binding of MalE was weak because the transporter hydrolyzes  (Fig. 3A, compare lanes 2 with lanes 4). In contrast, when the transporter was locked in the inward facing state with the pair MalF 446 /MalG 234 , the binding of MalE was at background levels, even in the presence of ATP and vanadate (Fig. 3A, right panel, lane 4). The results therefore confirm that ATP controls the transition from inward to outward facing state, which in turn controls the binding of MalE to the transporter. We next employed a fluorescence-based assay to determine the binding affinity of MalE to MalFGK 2 in the inward and outward facing states. We reconstituted the transporter in nanodisc because the system is well suited for spectroscopic analysis (22). First, we confirmed on native gel that MalE binds to the transporter with the same characteristics in nanodiscs as in proteoliposomes (Fig. 3, compare B with A; ATP dependence). We then titrated the fluorescent-labeled MalE with an increasing amount of transporter (Fig. 4). The data showed that MalE binds with a high affinity (K d ϭ ϳ50 -80 nM) only when the transporter is stabilized in the outward facing conformation with ATP-vanadate or ATP-BMOE (binding data and standard deviations reported in Table 1). In contrast, in the absence of nucleotides, the binding affinity of MalE for the transporter was ϳ7 M and increased furthermore to ϳ45 M when MalE was liganded with maltose (Table 1 and Fig. 3C). Because the transporter might spontaneously acquire its outward facing conformation even in the absence of ATP, the affinity of liganded MalE to the inward facing state may in fact be lower than 45 M. Indeed, the interaction of maltose-bound MalE was virtually null when the transporter was stabilized in the inward facing state using the cysteine pair MalF 446 /MalG 234 (Fig. 4D). The reaction was not saturable, and therefore the affinity could not be determined, in agreement with earlier binding assays in proteoliposomes (23). We concluded that ATP controls the transporter outward facing conformation and thereby controls the high affinity binding of MalE to the transporter. MalE Does Not Facilitate the Transition to the Outward Facing Conformation-As an additional way to probe the transporter conformational state and to explore the transition kinetics, we employed the fluorescent dye N-(1-pyrene) maleimide. When attached to a protein thiol, the dye produces two distinct emission peaks at 375 and 395 nm. When two pyrene-thiols interact within 6 -10 Å distance, they form excited state dimers (excimers) that emit at longer wavelengths than the lone excited fluorophore (24,25). The transporter labeled with the pyrene probe at the cysteine positions described above retained ϳ80% ATPase activity and 70 -80% maltose transport activity as compared with the wild type complex ( Table 2). The fluorescence spectra were recorded between 365 and 520 nm to monitor the conformational changes in the presence of ATP, MalE, and maltose (Fig. 5, red traces), compared with without ligands (black traces). With the pairs MalK 83 /MalK 83 and MalF 396 / MalG 185 , which monitor the conformational changes on the cytosolic side of the membrane, ATP decreased the monomeric emissions but increased the excimer emissions (Fig. 5, A and B). The reverse was observed with the pair MalF 446 /MalG 234 . This was expected because the pair monitors the conformational change on the opposite side of the membrane (Fig. 5C). In contrast, MalE and maltose had no effect on the monomeric and excimer emissions, in agreement with the conclusion that ATP alone supports the outward facing conformation of the transporter. Furthermore, we observed that MalE and maltose had no effect on the kinetics of the conformational transition (Fig.  6). These time-resolved experiments showed that liganded MalE is unable to modify the kinetics of MalK closure, even though we employed a cysteine pair that produced the highest fluorescence signal difference (i.e., MalK 83 /MalK 83 ). We therefore conclude that MalE is not required, nor does it facilitate the closure of the nucleotide-binding domain in the intact and functional maltose ABC transporter.

DISCUSSION
We report that ATP alone is sufficient for the closure of the MalK interface and for the conversion of MalFG to the outward facing conformation. This conclusion is based on cysteine cross-links and pyrene-based fluorescence of probes placed at diagnostic positions on the transporter. The cysteine substitutions and the pyrene label affect modestly the ATPase activity of the transporter, and do not affect its capacity to import maltose (i.e., 70 -80% activity of the wild type; Table 2). The conclusion we reach is in fact not surprising but rather in agreement with other studies reporting that ATP drives the dimerization of MalK in the membrane or in solution (26 -28). The conclusion is also consistent with the x-ray structures that showed the rigid body rotation of MalFG upon closure of MalK (8). That ATP alone regulates the conformation of the transporter has also been reported for other type I and type II importers (ModBC and BtuCD, respectively), as well as for ABC exporters (7, 29 -32).
Our results, however, disagree with the accepted notion that closed liganded MalE facilitates the closure of the MalK dimer. This mechanism was proposed (and perhaps intuitively accepted) because a rapid ATP hydrolysis activity occurs only in the presence of MalE and maltose. To the best of our knowledge, this model has been directly tested by EPR spectroscopy only (14). Why the study concluded that MalK closure strictly depends on MalE is puzzling. The spin label reduced the MalK ATPase activity by 2-fold (from 2,100 to 1,150 nmol/min/mg), yet the transporter was still active. The spin labels were, however, placed at two different positions on the same MalK subunit, giving rise to four different possible ways of spin interactions. Most importantly, the EPR distance measurements were performed in detergent solution, whereas the transporter ATPase is ϳ100-fold higher than in the membrane and poorly dependent on MalE and maltose (16). The sensitivity of EPR in the membrane environment is indeed problematic because of . The nanodiscs were repurified on a desalting G25 column before fluorescence quenching assay using ATTO-655-labeled MalE (20 nM). Three independent measurements were performed, and the data points were fitted to a one-site binding equation (mean and standard deviation results are presented in Table 1)  the large spin decay inherent to intermolecular dipolar interactions (33), and accordingly, the authors could not report the interspin distances with the transporter reconstituted in liposomes. Nevertheless, this pioneering work showed that ATP is essential for the pairing of MalK, and this conclusion is consistent with our results.
Our results also contrast with the interpretation of the crystal structure that captured closed liganded MalE with the transporter in a semi-closed conformation (15). It was concluded that closed liganded MalE binds to MalFG and induces the preclosure of MalK and thus ATP hydrolysis. Our pyrene-based fluorescence experiments rather show that the kinetics of MalK pairing is the same, whether MalE and maltose are present or not (Fig. 5). In addition, we find that the binding of closed liganded MalE to the inward facing transporter is quasi-null in solu-tion (K d Ͼ 45 M; Table 1). In contrast, the binding of MalE to the outward facing transporter occurs with nanomolar affinity (K d ϭ ϳ60 nM; Table 1). Clearly, a far more dominant interaction is taking place between ATP-bound outward facing MalFGK 2 and unliganded open state MalE. Given the high concentration of ATP in the cytosol (ϳ1 mM), we think this mode of interaction represents the resting state of the transporter. This view is supported by an in vivo cross-linking analysis, which reported that MalE is bound to the transporter even in the absence of maltose (34). This view is also consistent with our conclusion that transporter-bound MalE forms a receptor for maltose (16). In contrast, a model involving the low affinity interaction between closed liganded MalE and MalFGK 2 (K d Ͼ 45 M), as a way to deliver the sugar and activate the transporter, seems unsustainable because the concentration of periplasmic MalE decreases further as the environmental maltose concentration decreases. An obvious question raised by this work is how MalE stimulates the MalK ATPase. The co-crystal of closed liganded MalE with MalFGK 2 has led to the proposal that MalE lowers the energy barrier for ATP hydrolysis (15). However, it cannot be excluded that the binding of MalE to the outward facing transporter instead stimulates the release of ADP and/or Pi and thus the ATP turnover rate (and furthermore in the presence of maltose). Earlier studies suggested that ATP product release may coincide with the reorientation of MalFGK 2 to the inward facing conformation (14). For the mammalian P-glycoprotein, it was proposed that phosphate release is coupled to transport (35). Further work should therefore address what is the ratelimiting step during maltose transport. Our current work shows that the affinity of the transporter for ATP is not increased with MalE (16) and that ATP alone is sufficient to convert the transporter to the outward facing conformation. Together, these results imply that MalE must stimulate a step that occurs when the transporter returns to the inward facing conformation. This conclusion is compatible with the autoregulation model presented in Fig. 1B.