Dynamics of α-helical subdomain rotation in the intact maltose ATP-binding cassette transporter

Abstract

ATP-binding cassette (ABC) transporters are powered by a nucleotide-binding domain dimer that opens and closes during cycles of ATP hydrolysis. These domains consist of a RecA-like subdomain and an α-helical subdomain that is specific to the family. Many studies on isolated domains suggest that the helical subdomain rotates toward the RecA-like subdomain in response to ATP binding, moving the family signature motif into a favorable position to interact with the nucleotide across the dimer interface. Moreover, the transmembrane domains are docked into a cleft at the interface between these subdomains, suggesting a putative role of the rotation in interdomain communication. Electron paramagnetic resonance spectroscopy was used to study the dynamics of this rotation in the intact Escherichia coli maltose transporter MalFGK₂. This importer requires a periplasmic maltose-binding protein (MBP) that activates ATP hydrolysis by promoting the closure of the cassette dimer (MalK₂). Whereas this rotation occurred during the transport cycle, it required not only trinucleotide, but also MBP, suggesting it is part of a global conformational change in the transporter. Interaction of AMP-PNP-Mg²⁺ and a MBP that is locked in a closed conformation induced a transition from open MalK₂ to semiopen MalK₂ without significant subdomain rotation. Inward rotation of the helical subdomain and complete closure of MalK₂ therefore appear to be coupled to the reorientation of transmembrane helices and the opening of MBP, events that promote transfer of maltose into the transporter. After ATP hydrolysis, the helical subdomain rotates out as MalK₂ opens, resetting the transporter in an inward-facing conformation.

ATP-binding cassette (ABC) transporters belong to one of the largest protein superfamilies in organisms, and mediate the translocation of a wide range of substrates across the membrane (1). These transporters typically contain two transmembrane domains (TMDs) and are energized by a nucleotide-binding domain (NBD) dimer that closes and opens during cycles of ATP binding and hydrolysis (2, 3). Each NBD consists of a RecA-like subdomain, found in numerous ATPases (4), and an α-helical subdomain that is specific to the ABC family (5). Crystallographic studies (5–8) and molecular dynamic simulations (9, 10), performed on isolated NBDs, suggest that the helical subdomain rotates toward the RecA-like subdomain in response to ATP binding. This rotation positions the ABC family signature motif to interact with nucleotide across the dimer interface so that the NBDs can close to hydrolyze ATP. After ATP hydrolysis, it is suggested that the helical subdomain rotates away (5, 11). In addition, the TMDs contact the NBDs at the interface between these subdomains (12–15), suggesting a putative role of the rotation in interdomain communication (16). Here, we used site-directed spin labeling electron paramagnetic resonance (EPR) spectroscopy (17, 18) to study the dynamics of this rotation in the intact Escherichia coli maltose transporter MalFGK₂ (19, 20). A periplasmic maltose-binding protein (MBP) delivers the substrate to the transporter and stimulates its ATPase activity (21). In the maltose transporter, closure of the NBD dimer requires both ATP and MBP (22) and involves two separate rotational events, a ∼10° rotation of the α-helical subdomains relative to the RecA-like subdomains and a ∼14° rotation of the RecA-like subdomains relative to the C-terminal regulatory domains (23). Analysis of the helical subdomain rotation is crucial for a detailed understanding of how ATP hydrolysis is stimulated by MBP and coupled to transport. We therefore determined the requirements for subdomain rotation and defined the sequence of rotational events in the catalytic cycle.

Results

Positioning Spin Labels to Detect α-Helical Subdomain Rotation in MalK.

The maltose transporter has been crystallized in an inward-facing conformation in the absence of nucleotide (23), and an outward-facing conformational intermediate, stabilized by binding of both ATP and MBP, that may resemble the transition state (22, 24, 25) (Fig. 1A). Alignment of the RecA-like subdomains of these two structures (Fig. 1B) reveals an inward rotation of the helical subdomain toward the RecA-like subdomain, as seen in many isolated NBD structures upon ATP binding (5–8). These two subdomains are connected by the Q-loop (27) that lines part of the cleft surrounding the coupling helix of the transmembrane domains MalF and MalG. The Q-loop is expected to undergo conformational changes upon subdomain rotation that might be detected as a change in mobility of an attached spin label (SL). Residues 83–87 in the Q-loop were individually mutated to cysteine. Residues in the helical subdomain (residues V120, Q122, and A124) and in the RecA-line subdomain (residues A167, L168, and Q171) (Fig. 1B) were also mutated to cysteines, individually and in pairs, to measure changes in the distance between these two subdomains upon rotation. Residues were selected for mutation based on distance constraints (28, 29) and side-chain solvent-accessibility [http://mobyle.rpbs.univ-paris-diderot.fr/cgi-bin/portal.py?form=ASA (30)], to maximize spin-labeling efficiency.

Fig. 1.

Structural views of the maltose ABC transporter. (A) Structures of the maltose transporter in apo (Left, PDB 3FH6) and catalytic intermediate (Right, PDB 2R6G) states. The transmembrane domains MalF and MalG are both shown in orange. MBP is in blue. The RecA-like subdomain (gray), α-helical subdomain (red), and regulatory domain (purple) of MalK are distinguished. ATP (green) and maltose (cyan) molecules are present only in the catalytic intermediate. (B) Closeup of the MalK/MalG interface. Structures of the apo and catalytic intermediate state are superimposed based on the RecA-like subdomain to visualize the helical subdomain rotation and movement of the coupling helix. MalG is rendered in yellow (apo state) and orange (catalytic intermediate state). The coupling helix (or EAA loop) of MalG is shown as a cylinder. The RecA-like and helical subdomains of MalK (colored as in A) from the apo and catalytic intermediate states are shown in lighter and darker hues, respectively (figure modified from ref. 23). Residue Q82 defining the Q-loop (blue) is shown as stick for the catalytic intermediate state. Spheres indicate the positions of residues spin-labeled in this study (Q122 and A167 are indicated). The figure was prepared with Pymol (26).

Plasmids carrying mutation(s) encoding the cysteine substitutions in the malK gene were used to transform an E. coli strain containing a chromosomal deletion of malK, and were tested for function by complementation, as judged by the red color of the colonies on maltose MacConkey agar plates (22). By this criterion, all of the substitutions in the helical and RecA-like subdomains were functional, but only one of the Q-loop substitutions (S83C) was functional. The mutants were purified, spin-labeled, and tested for MBP-stimulated ATPase activity (Table S1). All of the double mutants showed a strong stimulation by MBP, which is characteristic of the maltose transporter (21). However, only four of the double mutants, those that retained reasonably high activity (≥670 nmol/ min /mg), were selected for further characterization.

EPR Analysis of Single Site Mutants.

EPR spectral lineshapes primarily reflect the mobility of the spin label, and are therefore sensitive to conformational changes in the local environment. With spin label at position 83 in the Q-loop, a notable decrease in mobility was detected upon addition of both MBP and the nonhydrolyzable ATP analogue AMP-PNP-Mg²⁺ (Fig. 2), conditions promoting the formation of the outward-facing conformation seen in the crystal structure (Fig. 1A). However, no changes were detected following addition of either ligand alone (Fig. 2). This result was consistent with our previous work demonstrating that closure of the NBD dimer interface requires both nucleotide triphosphate and MBP (22) and suggested that the subdomain may not rotate in response to the addition of either ligand alone because no mobility changes were seen in SL attached to the hinge between the two subdomains. Preincubation with ATP-Mg²⁺ and MBP, to allow ATP hydrolysis and generate a posthydrolysis (ADP-Mg²⁺-bound) state (22), increased the mobility of S83C-SL relative to the catalytic intermediate, though the spectrum was not superimposable with that of the apo state (Fig. 2). Changes in mobility of SL positioned at V120C and Q122C in the helical subdomain were also seen only following the addition of both AMP-PNP-Mg²⁺ and maltose-MBP (Fig. 2). In contrast, changes in mobility of SL positioned at L168C and, to a lesser extent, Q171C in the RecA-like subdomain (Fig. S1) are seen upon addition of AMP-PNP-Mg²⁺ only, in addition to the more substantial changes apparent when both ligands are present. Crystal structures of isolated NBDs indicate that when the NBDs are open, nucleotide binds to the RecA-like subdomain rather than the helical subdomain (31) consistent with the observation that SLs positioned in the RecA-like subdomain, but not the helical subdomain, sensed nucleotide binding. The altered mobilities of SLs in the helical subdomain in the catalytic intermediate may reflect helical subdomain rotation, which brings the SL side chains closer to residues in the RecA-like subdomain.

Fig. 2.

SLs positioned in the Q-loop and the helical subdomain undergo changes in mobility in the presence of both AMP-PNP-Mg²⁺ and MBP. Single cysteine substitutions were made in the Q-loop (S83C) and helical subdomain (V120C, Q122C, A124C) of MalK and these mutants were spin-labeled with methanethiosulfonate spin label (MTSL) and purified. A series of EPR spectra for each individual mutant, with the indicated additions present at the concentrations specified in Materials and Methods, are shown. The addition of ATP-Mg²⁺ allowed for ATP hydrolysis to occur. No changes in spectral lineshape were seen with SL positioned at A124C.

EPR Analysis of Doubly Labeled Mutants in the Apo State.

To demonstrate the rotation of the helical subdomain relative to the RecA-like subdomain, we tested several different pairs of spin-labels to determine the distance between these subdomains (Table S2). When the interspin distance is > 25 Å, the spectrum of a doubly labeled protein in continuous wave (cw)-EPR strictly corresponds to the sum of the spectra of the noninteracting single mutants. Interspin distances can be determined at room temperature in the range of 7–25 Å although the spectral broadening due to dipolar interaction is small above 18–20 Å generating uncertainty in determining the actual distance and distribution (28, 29). The magnetic dipolar interactions between two spin labels were analyzed in terms of interspin distance populations using the program “shortdistances100” (see Materials and Methods). Distances between the spin-labeled subdomains, calculated in the absence of ligand, were largely consistent with the apo state crystal structure (Tables S2 and S3). In all cases a significant percent of the spins (∼30%) did not interact, most likely reflecting incomplete labeling at both sites, although it is also possible that some fraction of the pairs are outside the measureable range.

Inward Rotation of the α-Helical Subdomain in the Intact Maltose Transporter.

The double mutant Q122C/A167C showed a substantial increase in spin–spin interaction when both AMP-PNP-Mg²⁺ and maltose-MBP were added (Fig. 3). Distance calculations showed a change from ∼15 Å in the apo state to ∼7.5 Å in the catalytic intermediate state for 62% of the population (Fig. 3 and Table S3). No change was seen when maltose-MBP alone was added and only a small change of ∼1 Å was seen when AMP-PNP-Mg²⁺ alone was added. The double mutants A124C/L168C and A124C/Q171C also showed an increase in spin–spin interaction under the same conditions (Fig. S2 and Table S3), although the distance measurements between these positions are likely to be less accurate because, in the apo state, they lie at the limit of detection by cwEPR. These data are consistent with an inward rotation of the helical subdomain toward the RecA-like subdomain occurring only when both AMP-PNP-Mg²⁺ and maltose-MBP are bound to the transporter. The final pair, Q122C-L168C, had a wide distribution of distances making it more difficult to interpret the results, but both the apo and catalytic intermediate states could be fit with a bimodal distribution of spins centered at 8 and 14 Å. The major change upon addition of AMP-PNP-Mg²⁺ and maltose-MBP for this SL pair was a shift in the distribution of the fit, with a large fraction of the spin label shifting from the distribution centered at 14 to 8 Å (Table S3).

Fig. 3.

Spin-labeled Q122C/A167C mutant reports on the α-helical subdomain rotation. Two cysteine substitutions were made in MalK, at position Q122 in the α-helical and A167 in the RecA-like subdomain, and modified by MTSL. (A) Distances between Q122 and A167 in the structures of the apo and catalytic intermediate are indicated with dashed lines. The two structures are superimposed based on the RecA-like subdomain and colored as described in Fig. 1B. (B) EPR spectra are color coded according to the ligands added, overlaid and normalized to maximum amplitude. Enlargements of EPR spectra highlight the increase in dipolar broadening due to spin-spin interaction upon addition of AMP-PNP-Mg²⁺ and maltose-MBP (red). (C) Distance distributions (x-axis values are in Å) in the presence of different ligands were determined using the program “shortdistances100.”

Outward Rotation of the α-Helical Subdomain in the Intact Maltose Transporter.

When studying the distance between the individual NBDs during the catalytic cycle, we observe a posthydrolytic conformation requiring ADP-Mg²⁺ and MBP in which the MalK dimer is semiopen (22). The conformation of the helical subdomain in this posthydrolytic conformation was studied by adding ATP-Mg²⁺ and maltose-MBP to the spin-labeled Q122C/A167C mutant and allowing hydrolysis to occur (Fig. 3). A decrease in spin–spin interaction was observed in the posthydrolysis state (main distance centered at 12.5 Å; Fig. 3 and Table S3) suggesting that the α-helical subdomain of MalK rotated away from the RecA-like subdomain following ATP hydrolysis. Analysis of A124C/Q171C (Fig. S2 and Table S3) also showed a slight distance increase after ATP hydrolysis (main distance ∼20 Å).

Use of a Locked-Closed MBP to Probe Rotation.

To gain a more detailed understanding of subdomain rotation, a mutant of MBP (G69C/S337C) was used that forms a disulfide bond designed to prevent opening of the closed, maltose-bound MBP (32). We tested whether this disulfide cross-linked (CL-)MBP could promote conformational changes in the NBDs. Our mutant MBP preparation is ∼90% cross-linked under nonreducing conditions (see SDS-PAGE gel in Fig. 4). We tested the effect of this MBP mutant first on a spin-labeled double mutant (V16C/R129C in MalK) previously shown to report on the distance between the two NBDs of the MalK dimer (22). In the apo state, SLs in this double mutant were too far apart to detect interactions, indicative of an open MalK dimer interface (ref. 22 and Fig. 4A). In the presence of both AMP-PNP-Mg²⁺ and wild-type MBP, the bulk of the population (65%) was centered at ∼8 Å (Fig. 4A and Table S3), indicative of a closed MalK dimer interface (22). The remainder of the population (∼35% of interacting spins), as fit by “shortdistances100,” was broadly distributed (Fig. 4A and Table S3), although simulations done using an earlier program (28), indicated that 85–90% of interacting spins were centered at 8 Å under these conditions (22). Addition of CL-MBP alone failed to trigger spin-spin interaction, however, when both CL-MBP and AMP-PNP-Mg²⁺ were present, the bulk of the population of interacting spins (82%) were centered at ∼17 Å. This result suggested that the binding of closed MBP in the presence of trinucleotide stabilizes a semiopen configuration at the NBD interface. The distribution of spins seen in the presence of CL-MBP and AMP-PNP-Mg²⁺ (Fig. 4A and Table S3) also included a smaller population (18% of interacting spins) at 8 Å, which most likely resulted from interaction with the small population of uncrosslinked species in the MBP preparation.

Fig. 4.

Locked-closed MBP in the presence of AMP-PNP-Mg²⁺ stabilized a semiopen MalK conformation but did not promote helical subdomain rotation. Two cysteine substitutions were made in MBP (G69C/S337C) that form a disulfide bond locking MBP in a closed, maltose-bound conformation (32). (Inset, Lower), shows an SDS-PAGE gel of the mutant MBP (CL-MBP) preparation, which contains a small fraction of uncrosslinked MBP (unCL-MBP). EPR spectra are color coded according to the ligands added (Lower Left) for both panels (A) and (B). The EPR spectra in absence of ligand were quasi identical to the spectra obtained with maltose and cross-linked MBP (cyan) and are not drawn for clarity. EPR spectra are overlaid and normalized to maximum amplitude. (A) EPR spectra of spin-labeled V16C/R129C. This double mutant was used previously to report on the distance between the two nucleotide-binding domains (22). (B) EPR spectra of spin-labeled Q122C/A167C. The enlargement of the EPR spectrum highlights the difference in dipolar broadening triggered in the presence of mutant MBP as compared to WT-MBP.

The Q122C/A167C mutant was then used to analyze the helical subdomain movement. Fig. 4B shows that the main distance between spin labels (14–15 Å) was essentially unchanged upon addition of the locked closed CL-MBP, either in the presence or absence of nucleotide, indicating that the helical subdomain does not rotate unless MBP is free to open. Eighteen percent of the population was again seen at a distance suggestive of inward rotation (8 Å) probably resulting from the uncrosslinked species (Table S3). These data indicated that binding of both AMP-PNP-Mg²⁺ and closed-MBP promoted a transition from open MalK to semiopen MalK without substantial helical subdomain rotation.

Helical Subdomain Rotation in the Reconstituted Transporter.

In contrast to a detergent-solubilized sample, the reconstituted transporter does not display basal ATPase activity in the absence of MBP (33), hence the effect of ATP-Mg²⁺ binding on helical subdomain rotation (rather than just a nonhydrolyzable analogue) can be tested in the absence of MBP. Spin-labeled mutants were reconstituted into nanodiscs, a lipid bilayer system recently developed in the laboratory of S. Sligar (34) and demonstrated to be a suitable membrane mimic for study of the maltose transporter (33). Nanodiscs are soluble disk-shaped membrane patches formed by two molecules of membrane scaffold protein wrapped around a bilayer of phospholipids. Advantages of nanodiscs over proteoliposomes include the accessibility of ligands to both sides of the membrane and their ability to be highly concentrated, improving EPR data quality (33). As in detergent, motional changes in SL attached to the RecA-like subdomain were seen upon addition of AMP-PNP-Mg²⁺ in the absence of MBP when the transporter was reconstituted into nanodiscs (Fig. S3) demonstrating that trinucleotides bind to the reconstituted transporter, even though ATP is not hydrolyzed in the absence of MBP (33). Addition of maltose-MBP alone, AMP-PNP-Mg²⁺ alone or, more importantly, ATP-Mg²⁺ alone to the doubly spin labeled Q122C/A167C mutant did not trigger substantial changes in interspin distances with transporter reconstituted in nanodiscs (Fig. 5A and Table S3). Distance changes consistent with inward rotation of the helical domain occurred only when both AMP-PNP-Mg²⁺ and maltose-MBP were added (Fig. 5A), and the helical subdomain was rotated away from the RecA-like subdomain in the posthydrolysis state (Fig. 5B and Table S3). We also tested the effect of CL-MBP plus AMP-PNP-Mg²⁺ on the transporter conformation in nanodiscs. Results were again consistent with those obtained in detergent, although there was a significantly higher proportion of interacting spins (∼36%) corresponding to full rotation of the helical subdomain than seen in detergent (∼18%) (Fig. S4 and Table S3). It is possible that the non-CL-MBP present in the preparation, which will open in the presence of AMP-PNP-Mg²⁺ competed better for binding to the reconstituted MalFGK₂, accounting for the higher percentage. In summary, the EPR data obtained using transporter reconstituted into nanodiscs essentially recapitulated those obtained in detergent.

Fig. 5.

Spin-labeled Q122C/A167C mutant reports on the α-helical subdomain rotation in nanodiscs. Mutant transporters, spin labeled at positions Q122C in the helical subdomain and A167C in the RecA-like subdomain, were reconstituted into nanodiscs. EPR spectra in the presence of the indicated ligands are overlaid and normalized to maximum amplitude. (A) Effect of ATP-Mg²⁺ alone or AMP-PNP-Mg²⁺ plus maltose-MBP on dipolar broadening in nanodiscs. ATP-Mg²⁺ can be used in place of AMP-PNP-Mg²⁺ alone because the reconstituted transporter does not hydrolyze ATP in the absence of MBP. The EPR spectrum in the presence of maltose-MBP was quasi identical to the spectrum obtained in absence of ligand (black) and is not drawn for clarity. (B) Effect of ATP-Mg²⁺ plus maltose-MBP on dipolar broadening in nanodiscs. ATP is hydrolyzed to ADP, generating a posthydrolysis state.

Discussion

ABC transporters are powered by NBDs that use the energy of ATP binding and hydrolysis to exert a mechanical task. Determining how the conformational changes in these domains coordinate with those of the transmembrane domains is crucial to understand the molecular mechanism of ABC transporters. Studies of the nucleotide-dependent association/dissociation of the two NBDs during the catalytic cycle (3, 35) indicate a strong coupling between these catalytic events and structural rearrangements in the TMDs (36–38). In addition, structural comparisons of isolated NBDs reveal a putative motion within each NBD, characterized by a rotation of the α-helical subdomain relative to the RecA-like subdomain upon ATP binding and hydrolysis (5–8). This motion is also supported by molecular dynamic studies (10, 11) and the rotation is suggested to transmit conformational changes to the transmembrane domains (16). Interestingly, the helical subdomain (14, 39) is the most variable region of the NBDs (40, 41) and might have evolved to accommodate the structural constraints imposed by the TMDs in ABC transporters. To the best of our knowledge, the only biochemical studies of the subdomain rotation were carried out on isolated NBDs (42, 43). However, one might wonder whether the relevance and role of helical subdomain rotation is overestimated, based on the sole, though established, lack of structural stabilization of this subdomain in absence of trinucleotide.

We carried out site-directed spin labeling and EPR spectroscopy to investigate the movement of the helical subdomain in the context of the intact maltose transporter. Our data provided biochemical evidence that this rotation does occur in an intact ABC transporter. In the absence of ligand, EPR data on the helical subdomain conformation were consistent with the inward-facing structure of the maltose transporter in which the helical subdomain is rotated outward (23) (Tables S2 and S3). In stark contrast with the main conclusion drawn from studies of isolated NBDs, in the intact transporter the binding of trinucleotide to the MalK subunits did not promote an inward rotation of the helical subdomain relative to the RecA-like subdomain. The apo state structure (23) therefore may be representative of the resting state of the transporter in vivo, where ATP is likely to be bound. The transmembrane domains must therefore impose constraints that prevent both the rotation of the helical subdomain (this study) and the closure of the MalK subunits (22) in response to ATP-Mg²⁺ binding alone. In contrast, binding of both AMP-PNP-Mg²⁺ and maltose-MBP triggered the inward motion of the helical subdomain relative to the RecA-like subdomain, the same conditions previously shown to trigger the closure of the NBDs in the intact transporter (22). These results indicated that the rotation played an important role in the transport mechanism, though it appeared to be part of a global conformational change involving MBP and the TMDs rather than a local response to ATP binding.

In the presence of AMP-PNP-Mg²⁺ and maltose-MBP, two different populations of interacting spins were detected in the Q122C-A167C pair. Approximately 65% of the population was centered in a narrow distribution at a distance of ∼7.5 Å, and 35% in a broad distribution centered at 14.5 Å. Whereas it is possible that a fraction of the transporters failed to undergo any conformational change in response to binding, we find this explanation unlikely as we often see evidence of participation of the entire population of transporters, as evidenced by changes in mobility of spin label (see, for example, SL-S83C in Fig. 2). Two populations could also have resulted from the dominance of more than one rotamer of the SL at either position 122 or 167 in the catalytic intermediate state. The length of the spin label side chain is 6–7 Å and the coexistence of two rotamers at a single site facing in different directions could account for two populations even though all of the helical subdomains undergo the same rotation. Finally, given that MalK is a dimer, it is possible that just one of the two subunits has rotated inward, although both are inward in the crystal structure of the catalytic intermediate, which is largely symmetric (24).

After demonstrating the rotation of the helical subdomain in the transport cycle of the maltose transporter, the dynamics of this motion was investigated in more detail. A model is depicted in Fig. 6 to illustrate and summarize these results. Using a MBP locked-closed by cross-linking, we were able to characterize an additional intermediate of the catalytic transport cycle. When the cross-linked MBP interacted with the AMP-PNP-Mg²⁺ bound transporter, MalK transitioned from an open to a semiopen conformation. During this movement, the helical subdomain rotated only slightly. Therefore, the subdomain rotation must occur during the transition from semiopen to closed MalK dimer, an event that coincides with the opening of MBP. Because the crystal structure of the catalytic intermediate (24) indicates that the open conformation of MBP stabilizes an outward-facing configuration of TM helices, we predict that the inward rotation of the subdomain coincides with the reorientation of TM helices to generate the outward-facing conformation as MBP opens. Following ATP hydrolysis within the closed NBDs, results from the Q122C-A167C pair suggested that the helical subdomain rotated almost fully away from the RecA-like subdomain, whereas the MalK subunits adopted a semiopen conformation (22). This result suggests that the outward subdomain rotation, like the inward rotation, occurs during the transition between closed and semiopen conformations of MalK.

Fig. 6.

Model of MalFGK₂ transport cycle. (Upper Left) The ATP-bound transporter is in an inward-facing resting state with open MalK. Helical subdomains (red) are in an outward conformation. When maltose-bound MBP interacts with the transporter (Upper Right), ATP-bound MalK adopts a semiopen conformation, but the helical subdomain is still outward. This conformation can only be observed if the MBP is maintained closed via cross-link. (Lower Right) The helical domain rotates inward and MalK closes as MBP opens, releasing maltose into the binding site of MalF. The transporter is outward-facing. Next, hydrolysis of one or two ATP promotes a presumably inward-facing conformation of the transporter. MalK is semiopen, and the helical subdomain outward. Dissociation of either ADP or MBP returns the transporter to its resting state (Upper Left). Maltose is released inside the cell either when MalK is in the semiopen or open conformations.

These observations strongly support a role of the helical domain in coupling of conformational changes between NBDs and TMDs and the predicted movements are largely consistent with the available crystal structures (23). The transition between the inward- and outward-facing conformations of the transporter involves concerted motions of transmembrane helices and NBDs. The main connection between the TMD and the NBD consists of the coupling helix containing the EAA motif (12, 40) docked into a cleft on the surface of MalK (Fig. 1B). Although the reorientation of TM helices appears to involve rigid body rotations (23), the movement at this interface is likened to a ball and socket joint (23) implying a flexibility that would allow for subtle but potentially crucial changes at the interface without disengaging the subunits. ATP binding induces changes in cross-linking patterns between NBDs and TMDs in the absence of MBP (15). We detected changes in SL mobility at position 168 in the RecA-like subdomain in response to nucleotide-binding even though it is 20 Å from the site of nucleotide-binding. Although the EAA loops move along with MalK during the catalytic cycle, differences in cross-linking between EAA loops and Q-loops are seen during the catalytic cycle of MalFGK₂ (44).

The model depicted in Fig. 6, which bears similarity to the mechanism proposed by Shilton (20), has several mechanistic implications. Closed MBP helps to stabilize an inward-facing intermediate state in which MalK₂ is semiopen. This conformation is visible by EPR when both locked-closed MBP and AMP-PNP-Mg²⁺ are bound to the transporter. Similarly, open MBP plus trinucleotide helps to stabilize an outward-facing intermediate state that is competent for ATP hydrolysis. The transition of bound MBP from closed to open, which coincides with rotation of the helical subdomain and the transition of MalK₂ from semiopen to closed, is associated with a large increase in affinity between MBP and FGK₂ (25, 45). Likewise, ATP binding contributes to the stabilization of these intermediate states, especially the outward-facing conformation that has a closed nucleotide-binding interface. Because neither ATP alone nor MBP alone can promote this transition, it is likely stabilized by both interactions. ATP hydrolysis would lead to an increase in the free energy of the outward-facing conformation by destabilizing the closed MalK dimer, allowing for a spontaneous return to a lower energy inward-facing state.

The NBD–TMD interface in exporters is quite different from that in importers. In importers, NBDs are in close contact with the TMDs, whereas long intracellular loops place the NBDs further from the membrane in exporters (46, 47). Two intracellular loops per dimer subunit form coupling helices, one interacts with residues around the Q-loop of the opposing subunit whereas the second binds a previously unrecognized motif unique to exporters named the “x-loop” (47). The described differences in the NBD/TMD interface of exporters as compared to the maltose importer appear to translate into distinct motions in the TM helices although the net result, alternating access for substrate in the TMD coupled to closure of the NBDs, may be conserved (48, 49). Alignment of the RecA-like subdomains of outward-facing Sav1866 (50) and inward-facing P-glycoprotein (51) revealed a similar inward rotation of the helical subdomain associated with NBD closure and reorientation of TM helices (Fig. S5). It is not yet clear how the substrates of exporters affect this global conformational change, which encompasses the helical subdomain rotation. Multidrug transporters typically exhibit both basal and drug-stimulated ATPase activities (52–54), whereas TAP1/TAP2 is an exporter that displays an ATPase activity tightly regulated by peptide substrates (55). ATP binding alone may permit helical subdomain rotation in a drug efflux pump, whereas TAP may more closely resemble the maltose transporter with respect to the regulation of ATPase activity.

Material and Methods

Site-Directed Mutagenesis.

Mutagenesis was performed using the GeneTailor Kit (Invitrogen), according to the manufacturer’s instructions. Single mutations in malK were introduced on the vector pCO-SSM, which encodes a cysteine-free version of MalK previously described (22). Double mutants were constructed by subcloning fragments from each single mutant plasmid, using restriction enzymes NheI and BsiWI. The sequences of mutant malK genes were verified by DNA sequencing.

Purification and Labeling of MalFGK₂.

Purification was performed as previously described (22), except for two modifications. Five mM imidazole was present in the buffer when the spin-labeled detergent supernatant was incubated with Talon affinity resin, which improved the purity of the samples, as judged by Coomassie-staining of SDS-PAGE gels. The gel filtration chromatography step, previously used to increase the purity of mutants used for spin–spin interaction (22), was then omitted. Mutants were labeled with 1-oxyl-2,2,5,5-tetramethyl-3-pyrroline-3-methyl methanethiosulfonate spin label (MTSL), as described (22).

Expression and Purification of G69C/S337C Mutant of MBP.

Details are provided in SI Text.

ATPase Activity.

Assays were carried out on proteins reconstituted in liposomes, using a coupled assay, as described previously (22).

Cw-EPR Spectroscopy.

X-band EPR was performed at room temperature, as described previously (22). Single and double mutants spectra shown are 100 and 200 G wide, respectively. Protein samples were studied either in detergent (22) or following reconstitution in nanodiscs (see SI Text). Samples (transporter ∼90 μM) were incubated with one or more of the following ligands for 15 min at room temperature before recording spectra: 15 mM ATP, 15 mM ADP, 15 mM AMP-PNP, 10 mM MgCl₂, 600 μM maltose, 250 μM MBP. Spin–spin distances were determined by fitting the experimental EPR data with simulations using a custom program developed by C. Altenbach (“shortdistances100,” written in Labview 2009 and available at http://sites.google.com/site/altenbach/labview-programs/short-distances). Briefly, this program utilizes a convolution method in which the interspin distances are simulated as Gaussian distributions. The simulated spectra consist of an unbroadened EPR lineshape convolved with a sum of Pake functions for randomly oriented spin pairs at a distance r. The distance distributions from dipolar broadening are determined by fitting simulated spectra with experimental data using Tikhonov regularization and nonlinear optimization.