Negative Stain Single-particle EM of the Maltose Transporter in Nanodiscs Reveals Asymmetric Closure of MalK2 and Catalytic Roles of ATP, MalE, and Maltose*

The Escherichia coli MalE-MalFGK2 complex is one of the best characterized members of the large and ubiquitous family of ATP-binding cassette (ABC) transporters. It is composed of a membrane-spanning heterodimer, MalF-MalG; a homodimeric ATPase, MalK2; and a periplasmic maltose receptor, MalE. Opening and closure of MalK2 is coupled to conformational changes in MalF-MalG and the alternate exposition of the substrate-binding site to either side of the membrane. To further define this alternate access mechanism and the impact of ATP, MalE, and maltose on the conformation of the transporter during the transport cycle, we have reconstituted MalFGK2 in nanodiscs and analyzed its conformations under 10 different biochemical conditions using negative stain single-particle EM. EM map results (at 15–25 Å resolution) indicate that binding of ATP to MalK2 promotes an asymmetric, semi-closed conformation in accordance with the low ATPase activity of MalFGK2. In the presence of MalE, the MalK dimer becomes fully closed, gaining the ability to hydrolyze ATP. In the presence of ADP or maltose, MalE·MalFGK2 remains essentially in a semi-closed symmetric conformation, indicating that release of these ligands is required for the return to the initial state. Taken together, this structural information provides a rationale for the stimulation of MalK ATPase activity by MalE as well as by maltose.

The Escherichia coli maltose transporter MalFGK 2 is a prototype for ATP-binding cassette (ABC) 3 transporters, an evo-lutionarily conserved family of membrane proteins that use ATP to mediate substrate translocation across membranes (1,2). ABC transporter malfunctions are associated with various diseases including cystic fibrosis, lipid transport defects, and multi-drug resistance (3). ABC transporters have a common architecture with two cytosolic nucleotide-binding domains (NBDs) connected to two transmembrane domains (TMDs) via a pair of coupling helices (4). The binding and subsequent hydrolysis of ATP by the NBDs promote the alternation of the TMDs between the outward facing and inward facing conformations, effectuating substrate transport (4). Although this basic mechanism is conserved between exporters and importers, bacterial importers require additional substrate-binding proteins. The maltose transporter is comprised of the two TMD subunits, MalF and MalG; the NBD MalK 2 ; and the maltosebinding protein MalE (5).
X-ray crystallography structures of MalFGK 2 in different conformations have provided crucial information regarding the chemistry of ATP binding and hydrolysis (6 -11). It is now established that the open, nucleotide-free MalK 2 is associated with the inward facing arrangement of MalF-MalG, whereas the ATP-bound closed MalK 2 is associated with the outward facing conformation. Substrate binding, transport, and release create additional intermediary steps that have not been solved by X-ray crystallography. Moreover, protein crystallization is performed in destabilizing detergent micelles, an environment in which the ATPase activity of MalK 2 is uncoupled from MalE and maltose (5,12).
Biochemical (12)(13)(14)(15)(16)(17)(18) and biophysical (19,20) studies have greatly advanced our understanding of the transport cycle; however, conflicting data concerning the role of ATP, MalE, and maltose remain to be resolved (17,19,20). To do so will require methods to isolate and characterize new intermediate conformations along the transport pathway. In this regard, nanodisc and EM are excellent tools well suited for the biochemical and structural study of membrane proteins. The nanodisc system allows for the stabilization of membrane proteins in a soluble and monodisperse state (21), permitting the determination of the 3D structure using single-particle EM.
Importantly, in nanodiscs (as in proteoliposomes), the ATPase activity of MalFGK 2 is coupled to MalE and maltose (12).
In this study, we have determined the role of nucleotide, MalE, and maltose on the closure and reopening of MalK 2 in the context of the full transporter. In doing so, we have characterized the conformations of MalFGK 2 in nanodiscs under 10 different biochemical conditions using single-particle EM combined with a systematic analysis of the conformational heterogeneity in any given states (for a complete summary, see Table  1). Results show that ATP induces the partial asymmetric closure of MalK 2 . Furthermore, whereas MalK 2 can dynamically sample the closed state, complete stabilization in that conformation requires binding of open state MalE. In the presence of ADP or maltose, most of MalE⅐MalFGK 2 remains in a semiclosed conformation, indicating that release ADP and maltose contributes to the return of the transporter to the resting state.

Results
ATP Binding Induces Asymmetric Closure of MalK 2 -We previously reported using cysteine cross-linking with variant MalK S83C that when the transporter was reconstituted in nanodiscs, ATP alone was sufficient to trigger the closure of MalK 2 (17). In contrast, spin-labeling EPR studies of MalFGK 2 in detergent micelles showed that maltose-bound MalE and ATP are both required for the closure of MalK 2 (20). The crystal structures of the MalE⅐MalFGK 2 obtained with ATP (6) or different nucleotides such ATP␥S and AMP-PNP (8,9,22) Fig. S1). There was no contact between the two MalK subunits at the level of the nucleotide binding pockets, and the electron densities corresponding to the interface MalK-MalG and MalK-MalF (henceforth called MalK F junction and MalK G junction) were separated from each other by ϳ58 Å. This open conformation of MalK 2 was reported by X-ray crystallography (7).
In the presence of ATP, full closure of MalK 2 was obtained; however, it was only when using the MalK S83C variant and the cysteine-reacting cross-linker BMOE (MalFGK 2 ⅐ATPϩBMOE; Fig. 2B and supplemental Fig. S2) (17). In this closed conformation adopted by 100% of the particles, there was extensive contact at the level of the nucleotide-binding pockets, and the distance between MalK F and MalK G junctions was ϳ16 Å.
In the presence of the non-hydrolyzable ATP analog (ATP␥S) and ADP-phosphate mimics (ADP-VO 4 or ADP-AlF 4 ) MalK 2 was not stabilized in the closed state. Instead, two different semi-closed conformations were obtained, termed asymmetric and symmetric. In the semi-closed asymmetric state (adopted by the majority of the particles), the MalK G junction was closer to the center of the complex, whereas the MalK F junction remained in the open position, similar to the nucleotide-free state ( Table 1, Fig. 2C, and supplemental Fig. S3). Electron densities at the level of the nucleotide pockets were well separated, and the distance between MalK F and MalK G was ϳ40 Å. In the semi-closed symmetric state (31% of the particles incubated with ADP-AlF 4 ), the MalK G and MalK F junctions were both closer to the center of the complex ( Fig. 2D and Table 1). The MalK F and MalK G junctions were separated by only ϳ22 Å, but the contact between the nucleotide pockets in MalK 2 was not as extensive as in the closed conformation (Fig. 3).
Finally, in the presence of ADP, 47% of MalFGK 2 adopted the semi-closed asymmetric conformation (Fig. 2E), whereas the remaining 53% were in the open state (Table 1). ADP was there-fore able to affect the conformation of the MalK G junction, as demonstrated by the presence of some transporters in the semiclosed asymmetric conformation.
Taken together, these results indicate that binding of ATP and non-hydrolyzable ATP analogs stabilize the asymmetric semi-closed conformation. The transporter is able to transiently sample the closed state because this conformation is captured by the cross-linker BMOE. However, without the cross-linker agent, MalK 2 is most stable in the asymmetric semi-closed conformation.
MalE Stabilizes the Closure of MalK 2 -In the crystal structures of MalE⅐MalFGK 2 bound to ATP/AMP-PNP/ADP-P i mimics, MalK 2 was fully closed (6,8,9). In the EM maps of MalFGK 2 ⅐ATP␥S, MalFGK 2 ⅐ADP-VO 4 , and MalFGK 2 ⅐ADP-AlF 4 (Fig. 2), MalK 2 was partially closed. MalE therefore plays a critical role for the closure of MalK 2 . Confirming this, the EM map obtained for MalE⅐MalGFK 2 in the presence of ADP-VO 4 revealed that MalK 2 was fully closed in 100% of the particles analyzed ( Fig. 4A and Table 1). The conformation of our EM map matches that of the equivalent crystal structures (6,8,9) (supplemental Fig. S4). The contact at the level of the nucleotide binding pockets was extensive, and the distance between MalF K and MalF G junctions was similar to that measured in the complex MalFGK 2 ⅐ATPϩBMOE (ϳ18 Å). We conclude that binding of MalE to MalF-MalG stabilizes the ATP-induced closure of MalK 2 .
MalE Modifies the Conformation of MalK 2 in the Presence of ADP-MalE has very low affinity for MalFGK 2 in the absence of nucleotide or in the presence of maltose (16). Thus, to prevent the dissociation of MalE, we tethered the protein via its N-lobe to the periplasmic P2-loop of MalF using disulfide linkage (complex termed MalE-x-MalFGK 2 ; Fig. 5). The EM analysis of this covalently bound complex (Fig. 4, B-G) showed that MalE interacts with MalF and MalG via both lobes (referred to as "bound MalE") or via its N-lobe only because of the disulfide bridge (referred as "detached MalE"). MalE was detached in 19% of the particles in the absence of nucleotide and 14% of the particles in the presence ADP (Table 1). Interestingly, with ADP, 53% of MalFGK 2 had MalK 2 in the open conformation, but this number was decreased to 21% in the presence of MalE (Table 1 and supplemental Figs. S3 and S5). The remaining particles adopted the semi-closed asymmetric (for MalFGK 2 ; Fig. 2E) and symmetric conformations (for MalE-x-MalFGK2; Fig. 4B). In the absence of nucleotide, MalK 2 was always in the open conformation, even when MalE was bound to MalFGK 2 ( Table 1 and Fig. 4C). Thus, the tethering of MalE to MalFGK 2 causes the motion of the MalK F junction but only in the presence of nucleotides.
Maltose Induces Detachment of MalE and Prevents Reopening of MalK 2 -We compared the conformations of MalE-x-MalFGK 2 in the presence and absence of maltose (Fig. 4, C-G). The number of particles with detached MalE increased from 19 to 32% upon the addition of the sugar (Table 1), consistent with the fact that closed liganded MalE has little affinity for the transporter (16). Interestingly, when MalE remained bound to the transporter in the presence of maltose, the number of particles with MalK 2 in the open conformation decreased from 100 to 21%. The majority of the particles (ϳ79%) were in the semi-closed symmetric conformation (Table 1 and Fig. 4E). The dominance of this conformation suggests that release of maltose from the transporter is necessary to permit return of MalK 2 to the open resting state.

Discussion
This single-particle EM analysis has defined the conformational changes of MalK 2 in response to the three factors-nucleotide, MalE, and maltose-individually and cooperatively. It has also identified intermediate conformations not previously captured by X-ray crystallography. The resolution of the EM maps was between 15 and 25 Å, which is sufficient for the detection of the large scale conformational changes of MalK 2 (supplemental Fig. S5). These include 1) the motion of both MalK F and MalK G junction by ϳ20 Å between the open and closed conformations, 2) adoption of a symmetric or asymmetric conformation of MalK 2 , and 3) the degree of contact at the level of the nucleotide binding sites. These are persistently visible at  different contour levels. The differences observed between the EM maps are consistent with the differences between simulated maps of MalK 2 in the open, closed, and semi-closed conformations (Fig. S6). Together, our results combined with those in the literature have allowed us to refine the model of transport (Fig. 6). In the resting apo-state, the MalK 2 dimer exists in an open conformation. The binding of ATP drives the closure of MalK 2 ; however, this conformation is unstable, and MalK 2 reverts to the asym-metrical state (Fig. 6A). The presence of MalE stabilizes the closed conformation of MalK 2 induced by ATP (Fig. 6B), which is in accordance with our previous report that ATP cleavage is stimulated by MalE (5,23). Maltose further increases the rate of ATP hydrolysis by increasing the rate of ADP and P i release from the nucleotide binding pockets (23) (Fig. 6C). This latter point is supported by our results, which show that MalK 2 is primarily in a semi-closed conformation when the transporter is incubated with maltose (Fig. 5, B and E), thus indicating that dissociation of the sugar contributes to the return to the resting state. Additionally, we find that maltose triggers the dissociation of MalE from the transporter, as expected because closed liganded MalE has weak affinity for MalF and MalG (16). Finally, in the absence of nucleotide and maltose, MalK 2 is exclusively in the open conformation, whereas in the presence of either of these factors, it adopts a range of conformations, suggesting that the release of ADP and maltose enhances the return of the transporter to the open resting state.
In the absence of MalE, full closure of MalK 2 is obtained with ATP only when a cross-linking reagent designed to stabilize this conformation is also included during incubation (17). Interestingly, with ATP␥S or the ADP-P i mimics ADP-VO 4 and ADP-AlF 4 , the majority of the particles remain in the asymmetric conformation in which only the MalF G junction has moved toward the center of the complex. This asymmetry indicates that neither ATP␥S nor the ADP-P i mimics have the ability to stabilize MalK 2 in the closed state. In the case of ADP-AlF 4 , ϳ30% of the complexes adopt asemi-closed symmetric conformation in which both MalK F and MalK G junctions have moved toward the center of the complex. This later conformation may be due to the specific or higher number of liaisons that ADP-  AlF 4 establishes at the nucleotide-binding pockets (8). It could also be because the ADP-AlF 4 state is reached by a backward reaction upon addition of ADP and Al 3ϩ /F Ϫ ions, whereas the ADP-VO 4 is reached by a forward reaction that requires ATP binding and hydrolysis.
Importantly, the semi-closed asymmetric conformation of MalK 2 has never been reported by X-ray crystallography. All crystal structures of MalFGK 2 and MalE⅐MalFGK 2 show nearly identical positioning of the MalK F and MalK G junctions, which has led to the proposal that MalK 2 adopts a symmetrical "tweezers-like" motion during the transport cycle (6 -11). Our results now reveal that binding of the nucleotide results solely in the stabilization of the movement of the MalK G junction but not that of the MalK F junction. We therefore propose that closure of MalK 2 occurs in a sequential and asymmetric manner, in agreement with recent molecular dynamics simulations (24) and earlier cross-linking experiments (14,15).
Because only the MalK G junction moves toward the center of the complex during asymmetrical closure, this motion may also drive the asymmetric movement of MalG and MalF during transport.
In the presence MalE and ADP-VO 4 , we find that MalK 2 is converted to the closed conformation. This dependence on MalE for the closure of MalK 2 explains why the cleavage of ATP relies on MalE, resulting in ϳ4-fold increase in the overall ATP turnover rate (5,12). It also explains why binding of vanadate becomes irreversible upon addition of MalE (25) and why EPR studies have reported that ATP alone is insufficient to support the full closure of MalK 2 (20,26). We find that in the presence MalE and ADP, there was a reduction in the number of particles with MalK 2 in the open conformation, as well as a conversion of the semi-closed conformation of MalK 2 from asymmetric to symmetric. In the absence of nucleotide, MalK 2 is in the open conformation independently of the presence of MalE. To-FIGURE 6. Model of the maltose transport cycle. A, ATP induces closure of MalK 2 (EM map MalFGK 2 ⅐ATPϩBMOE), but this conformation is unstable in the absence of MalE (EM maps MalFGK 2 ⅐non-hydrolyzable ATP and ADP-P i analogs) and reverts to a semi-closed symmetric conformation. B, MalE stabilizes the ATP-induced closed conformation (EM map MalE⅐MalFGK 2 ⅐ADP-VO 4 ), and therefore ATP cleavage occurs. C, maltose enhances the return of the transporter to inward facing conformation, thereby increasing the rate of ADP and P i release and consequently increasing the ATP hydrolytic cycle. MalE alone does not modify the open conformation of MalK 2 . The ATPase activity indicated is that previously reported (5,28).
gether, these results imply that MalE allosteric regulation is nucleotide-dependent.
Finally, in the presence of maltose the majority of the particles MalE-x-MalFGK 2 (ϳ79%) switch to the semi-closed symmetric conformation in lieu of remaining in the open state. This semi-closed symmetric conformation is very similar to the conformation of MalE⅐MalFGK 2 ⅐maltose reported by X-ray crystallography (PDB code 3PV0) (9). The X-ray structure shows that one molecule of maltose is trapped in the MalF-MalG transport cavity (9). It is therefore possible that the semi-closed symmetric conformation of MalE-x-MalFGK 2 is caused by the presence of this maltose molecule in the TMDs. This conformation may also be caused by the closure of MalE upon maltose binding. The current resolution of our maps does not allow us to differentiate between these two possibilities. Regardless, the fact that 100% of the particles MalE-x-MalFGK 2 exist in the open conformation without maltose versus 21% with maltose strongly suggests that release of the sugar precedes the return of the transporter to the initial inward facing state.
In conclusion, negative stain single-particle EM combined with 3D classification negative stain single-particle EM allows to assess protein conformations and their equilibrium in many different biochemical conditions. Here, the EM study of MalFGK 2 in nanodiscs, by assessing how three factors modified the conformations of MalK 2 , has enabled us to refine the model of maltose transport.
The MalFGK 2 transporter reconstituted in nanodiscs with MalE were incubated with 2 mM ADP or ATP␥S for 20 min at room temperature. The ADP-VO 4 state was obtained by incubating the complex with 1 mM ATP and 10 M vanadate (sodium orthovanadate) in reaction buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 5 mM MgCl 2 ) at room temperature for 20 min, time enough for trapping to occur (supplemental Fig. S2). The complexes were then purified by size exclusion chromatography in reaction buffer plus 1 mM ATP and 10 M vanadate. For the ADP-AlFx state, the protein in the ADP buffer was incubated for ϳ4 h at 4°C with 2 mM ADP, 8 mM NaF, and 2 mM AlCl 3 in TEM buffer.
EM Sample Preparation and Image Acquisition-MalFGK 2 and derivate samples (each 5 l diluted to 50 g/ml) were applied onto negatively glow-discharged carbon-coated grids (400 mesh, copper grid) for 1 min, and excess liquid was removed by blotting with filter paper. Freshly prepared 1.5% uranyl formate (pH 5) was added (5 l) for 1 min and then blotted. Digital micrographs were collected using a FEI Tecnai G2 F20 microscope operated at 200 kV and equipped with a Gatan Ultrascan 4k ϫ 4k Digital CCD Camera. The images were recorded at defocus between 0.7 and 1.4 m at a magnification of 134,010ϫ at the camera and a pixel size of 1.12 Å.
EM Data Processing and Image Analysis-Contrast transfer function parameters were determined using CTFFIND3 (30), and micrographs were phase flipped using XMIPP 2.3 software (31). Protein particles were boxed using e2boxer from the EMAN2 software suite (32). After extraction, images of the particles were binned twice for a final box size of 128 ϫ 128 pixels and a resulting pixel size of 2.24 Å at specimen level. False positives (images that do not contain particles), images of empty nanodiscs, and images that show particles too closed to one another, were eliminated after 2D classification using the Kmeans ascendant classification (with 2, 4, 8, 16, 32, and 64 classes successively) of the SPARX software suite (33). The reference volume was obtained by EMAN2 common line algorithm using the resting state particles. This same reference volume was used for 3D reconstructions of all the data sets 3D analysis and was low pass filtered to 60 Å at the start of the analysis procedure of each data set. The 3D analysis was conducted using the RELION-1.2 software suite (34) was used for analyzing each data set. The first step consisted of a maximum likelihood 3D classification with 4 -7 seeds. This classification allowed for the identification of stable dominant conformations and the elimination of poorly defined particles. The second step was a refinement of each stable conformation. When similar classes were obtained as for the MalFGK 2 -only particles (supplemental Fig. S1) belonging to the best defined class were further analyzed in a refinement conducted using Relion (34). When different classes were obtained, all were refined separately. For the MalFGK 2 ⅐ATPϩBMOE and MalE⅐MalFGK 2 ⅐ADP-VO 4 data sets, an additional refinement step using SPARX was performed to further improve the maps. For the refinement, the initial volume was low pass filtered to 50 Å, and refinement was performed for 10 iterations with an angular sampling of 10 degrees, followed by 10 iterations with an angular sampling of 5°degrees. Resolution was estimated using forward scatter and gold standard forward scatter with the criteria of 0.5 and 0.143 (supplemental Table S1 and Fig. S5). The 3D-EM density maps of the complex without MalE were visualized using Chimera (35). All docking shown in this study was performed using Chimera rigid body docking tools.