Full engagement of liganded maltose-binding protein stabilizes a semi-open ATP-binding cassette dimer in the maltose transporter

Summary

MalFGK₂ is an ATP-binding cassette (ABC) transporter that mediates the uptake of maltose/maltodextrins into Escherichia coli. A periplasmic maltose-binding protein (MBP) delivers maltose to the transmembrane subunits (MalFG) and stimulates the ATPase activity of the cytoplasmic nucleotide-binding subunits (MalK dimer). This MBP-stimulated ATPase activity is independent of maltose for purified transporter in detergent micelles. However, when the transporter is reconstituted in membrane bilayers, only the liganded form of MBP efficiently stimulates its activity. To investigate the mechanism of maltose stimulation, electron paramagnetic resonance (EPR) spectroscopy was used to study the interactions between the transporter and MBP in nanodiscs and in detergent. We found that full engagement of both lobes of maltose-bound MBP unto MalFGK₂ is facilitated by nucleotides and stabilizes a semi-open MalK dimer. Maltose-bound MBP promotes the transition to the semi-open state of MalK when the transporter is in the membrane, whereas such regulation does not require maltose in detergent. We suggest that stabilization of the semi-open MalK₂ conformation by maltose-bound MBP is key to the coupling of maltose transport to ATP hydrolysis in vivo, because it facilitates the progression of the MalK dimer from the open to the semi-open conformation, from which it can proceed to hydrolyze ATP.

Introduction

ATP-Binding Cassette (ABC) transporters compose a superfamily of ubiquitous membrane complexes essential for active translocation of a large variety of molecules across biological membranes (Holland, 2011). They consist of two transmembrane domains (TMDs) that form a translocation pathway and two nucleotide-binding domains (NBDs) that couple ATP hydrolysis to transport (Rees et al., 2009; ter Beek et al., 2014). Most ABC importers require an additional soluble binding protein that recognizes substrate and delivers it to the TMDs (Eitinger et al., 2011; Rice et al., 2014).

The maltose transporter in Escherichia coli, which is responsible for the uptake of linear α-(1→4) maltodextrins up to maltoheptose (Boos and Shuman, 1998), is one of the best characterized ABC transporters (Bordignon et al., 2010; Chen, 2013; Orelle et al., 2014; Shilton, 2008). The membrane-associated transport complex is composed of a heterodimer of TM subunits, MalF and MalG, and a homodimer of NBDs, MalK₂ (Davidson and Nikaido, 1991). The maltose binding protein, MBP, is comprised of 370 amino acids that form two pseudo-symmetric lobe domains separated by a short hinge. The two lobes of MBP close with a twisting motion about the hinge when substrate binds in the cleft between them (Sharff et al., 1992; Spurlino et al., 1991). MalK₂ binds two ATPs along the dimer interface (Chen et al., 2003), and ATP hydrolysis results from the MBP-dependent closure of the two NBDs (Orelle et al., 2008), thereby forming the integral nucleotide-binding sites (Fetsch and Davidson, 2002; Oldham and Chen, 2011b). An alternating access mechanism has been proposed for translocation in which rotation of the TMDs alternates access to a centrally located sugar-binding pocket from one side of the membrane to the other (Chen et al., 2001). Two structures of the maltose transporter, in inward and outward-facing conformations (Khare et al., 2009; Oldham et al., 2007), largely substantiate this model. In the inward-facing conformation (Fig. 1A), obtained in the absence of nucleotide, NBDs are open and the TMDs are oriented such that maltose can pass from the central binding site in MalF to the cytosol, but not to the periplasm (Khare et al., 2009). In the outward-facing state with ATP-bound (Fig. 1D), NBDs are closed and the TMDs are oriented such that the sugar binding site in MalF is cut off from the cytoplasm and instead faces an enclosed cavity surrounded by the outward-facing TMDs and an open MBP (Oldham et al., 2007). A translocation cycle is initiated by binding of both MgATP and a closed-ligand bound MBP to the transporter (Orelle et al., 2010). This promotes a transition to the outward facing conformation in which maltose is transferred from MBP to MalF and ATP is hydrolyzed (Oldham and Chen, 2011b). Very little ATP hydrolysis occurs in the absence of MBP (Davidson et al., 1992) and low activity is displayed in the presence of maltose-free MBP (Davidson et al., 1992; Gould et al., 2009), indicating that the stimulation of ATP hydrolysis by liganded MBP is key to the transport mechanism. However, our understanding of the mechanism by which transport is coupled to ATP hydrolysis is still incomplete. A recent study indicated that both substrate-free and liganded MBP induced an outward-facing conformation of the ATP-bound maltose transporter (Bohm et al., 2013). Although these observations provided insights into the ATPase stimulation by unliganded MBP, it did not explain the far superior capability of liganded MBP to enhance ATP hydrolysis. Other biochemical studies suggested that specific binding of the carbohydrate to the transmembrane site in MalF is not important for activation of its ATPase (Cui et al., 2010; Gould and Shilton, 2010). Therefore, it is likely that the substrate-induced conformational change in MBP is critical for stimulation of MalFGK₂ ATPase. Consistent with this, a pretranslocation state (Fig. 1B) of the transporter was crystallized in which docking of liganded MBP to MalFGK₂ caused a transmembrane re-arrangement and a partial closure of MalK₂ (Oldham and Chen, 2011a).

Figure 1. Crystal structures of MalFGK₂ during the transport cycle.

A, In the absence of ligand, MalFGK₂ was crystallized in an inward-facing conformation with an open MalK dimer (PDB 3FH6) (Khare et al., 2009). MalF is shown in blue, MalG in yellow, and MalK subunits in red or green. B and C, MalFGK₂ was crystallized in an intermediate pretranslocation state in the presence of either B) liganded MBP alone (PDB 3PV0) or C) AMP-PNP-Mg²⁺ in conjunction with a liganded MBP locked in a closed conformation (PDB 3PUZ). The inset shows an enlargement of the two AMP-PNP-Mg²⁺ binding sites (ligands shown in gray and black). Both crystal structures have a similar inward-facing conformation with a semi-open MalK dimer and closed MBP (Oldham and Chen, 2011a). The N- and C-lobes of MBP are shown in magenta and pink, respectively. Maltose sugar is represented as red spheres. D, an outward-facing conformation of MalFGK₂ (PDB 2R6G) was obtained by using a catalytically inactive mutant that was crystallized in the presence of maltose, MBP and ATP (Oldham et al., 2007). In this state, the MalK dimer is closed and the MBP has opened to release maltose into a transmembrane binding site located in MalF.

Site-directed spin labeling (SDSL) electron paramagnetic resonance (EPR) spectroscopy has proven to be a powerful tool to study conformational changes in proteins (Hellmich and Glaubitz, 2009; Hubbell et al., 2013; Klug and Feix, 2008). This method can report on the local environment of a small spin label attached to a unique cysteine residue within a large protein complex, allowing the detection of local changes in conformation due to protein-protein or protein-ligand interactions. The introduction of two spin labels within a protein complex allows the measurement of distances between 7 and 20 Å using room temperature continuous wave (CW) EPR spectroscopy (Altenbach et al., 2001). EPR spectroscopy and structural studies suggest that MBP interacts with the transporter throughout the transport cycle, even in the absence of maltose, principally via contact between its N-lobe and a large periplasmic loop (the P2 loop) in MalF (Austermuhle et al., 2004; Bohm et al., 2013; Daus et al., 2007; Daus et al., 2009; Grote et al., 2009; Jacso et al., 2009; Oldham et al., 2007). EPR-based distance measurements (Bohm et al., 2013; Grote et al., 2009; Orelle et al., 2008; Orelle et al., 2010) and the crystal structure of the pretranslocation state (Oldham and Chen, 2011a) reveal that the transporter can be stabilized with NBDs in a semi-open configuration. These findings led us to investigate the interaction of the C-lobe with the transporter in more detail, based on the hypothesis that both N- and C-lobe engagement would be required to stabilize the NBDs in a conformation different from that seen in the resting state. We found that interaction of the C-lobe of MBP occurred only when nucleotide was present and stabilized the NBDs in either a semi-open or a closed state. In addition, our results showed that stabilization of the semi-open state was maltose-dependent when the transporter was reconstituted in lipid bilayer, suggesting a mechanism by which maltose greatly increases MBP-stimulated ATPase activity.

Results

Effect of maltose on the ATPase activity of MalEFGK₂

Maltose greatly enhances the ability of MBP to stimulate the ATPase activity of MalFGK₂ in proteoliposomes (Davidson et al., 1992). Similarly, maltose stimulated activity about 10-fold when the transporter was reconstituted in nanodiscs (Fig. 2B and Fig. S1) (Alvarez et al., 2010). Nanodiscs provide a convenient alternative to liposomes since the membrane protein is incorporated into a patch of phospholipid bilayer contained within a serum lipid carrier protein, thereby rendering both hydrophilic surfaces accessible to the aqueous environment (Denisov et al., 2004; Nath et al., 2007). In detergent, the transporter displayed a basal ATPase activity (~122 nmol/min/mg). Although this activity was stimulated ~10 fold by MBP, this activation did not require maltose. These observations suggest that in lipid bilayers, a higher energy barrier exists between the conformational states of the transporter, thereby resulting in a tighter coupling between substrate transport and ATP hydrolysis. We thus sought to identify what key conformational change(s) are dependent on the presence of maltose in membranes, but not in detergent micelles.

Figure 2. Functional characterization of spin-labeled MBP mutants.

A) Closed MBP (1ANF), viewed from the interface that interacts with MalFGK₂, is shown in a space-filled representation with the residues (cyan) that have been individually mutated to cysteine for labeling. Maltose (red) is visible within the cleft formed by the two lobes of MBP (magenta, N-lobe; pink, C-lobe). Images were prepared using Pymol. B) The ATPase activities of MalFGK₂ in nanodiscs (black) and in detergent micelles (grey) are shown in the absence and presence of 5 μM MBP (wild-type or spin-labeled) as indicated in slanted words, with or without 0.1 mM maltose (as labeled with +/−). The data plotted are an average of two measurements varying ~6% of the mean. MalFGK₂ in nanodiscs did not exhibit any detectable ATPase activity in the absence of MBP, and thus, did not show a visible bar on the graph.

Effect of maltose on the conformation of the nucleotide-binding domains

Although closure of the nucleotide-binding domains MalK is essential for regulating ATPase activity (Chen et al., 2013; Fetsch and Davidson, 2002; Orelle et al., 2010), both unliganded and maltose-bound MBP surprisingly induce the outward-facing conformation of the transporter in the presence of ATP (Bohm et al., 2013; Orelle et al., 2008). By employing the doubly spin-labeled MalK mutant 16C/129C to determine the distance between the two NBD subunits (Orelle et al., 2008), we found, however, that the stability of the semi-open state of MalK was substantially different in detergent or nanodisc systems. When the detergent-stabilized transporter was incubated with MgADP and MBP, both in the presence and absence of maltose, the population of spins was centered around 12 Å (Fig. 3A), representing a semi-open conformation (Orelle et al., 2008). However, when the transporter was reconstituted into nanodiscs (Fig. 3B), most of the spin labels were >20 Å apart in the absence of maltose, suggesting that the NBDs were open. In presence of maltose, ~62% of the spins were in the semi-open configuration. The inability of maltose-free MBP to stabilize the semi-open MalK₂ conformation in nanodiscs with MgADP and the fact that even the liganded MBP does not stabilize the entire population of transporters are consistent with the idea that the semi-open state is much less stable in the membrane environment. We next aimed at studying how liganded MBP stabilizes the nucleotide-binding domains in the semi-open state.

Figure 3. Conformation of MalK dimer.

The 250 G wide EPR spectra of spin-labeled MalFGK₂ (MalK-V16R1/R129R1) in detergent (A) or nanodiscs (B) were collected in the presence of the additions indicated. The spectra were normalized to the same number of spins. Distance distributions for each spectrum were determined using the program shortdistances100 and are shown in the middle. Due to the limit of cw-EPR sensitivity, the overall distribution above 20 Å mostly correspond to a population of non-interacting spins. Although the population of non-interacting spins can be fairly accurately determined, the distance distribution above this limit is inherently inaccurate. On the right, simulated spectra based on the distance distribution (middle) are shown in dashed lines and are overlaid with the experimental spectra (solid lines).

Conformation of MBP during the transport cycle

Maltose directly affects MBP conformation by inducing its closure when it binds between the two lobes of the protein (Spurlino et al., 1991). It is therefore important to distinguish MBP conformation during the transport cycle, because both closed and open conformations can interact with the transporter (see Fig. 1C and 1D). A doubly spin-labeled 41C/211C MBP mutant (Fig. 2A) was used to detect MBP closure or opening during the transport cycle. As shown previously (Austermuhle et al., 2004), when MBP was open (in the absence of maltose), spin labels were > 20 Å apart and no broadening due to spin-spin interaction was seen (Fig. 4). Addition of maltose triggered the appearance of dipolar broadening in the spectra, suggesting that MBP had closed. Although the spin-spin interactions observed for these spectra represented distances from 8 to 16 Å, we were not able to quantify the populations of each distance in the presence of maltose. Spin labeled MBP was stepped through the transport cycle first by adding MalFGK₂ in the absence of nucleotide, which allow MBP binding, then by adding ATP in the presence of EDTA, which prevents ATP hydrolysis, then, finally, by adding excess MgCl₂, which allows ATP hydrolysis. An EPR spectrum was recorded at each step. In a separate experiment, MalFGK₂ and MgADP were added directly to mimic a post-hydrolysis state (Orelle et al., 2008). The series of overlays with or without maltose (Fig. 4) qualitatively indicated, through visible dipolar broadening of the spectral wings (see arrows), that maltose triggered MBP closure under all conditions tested except in the presence of MalFGK₂ and ATP, which provide sufficient energy in the system to open the liganded MBP such that MBP becomes tightly bound to the transporter (Austermuhle et al., 2004). In the absence of maltose, there was no suggestion that interaction with the transporter triggered spin-spin interaction or closure of MBP in any state ((Austermuhle et al., 2004) and Fig. 4). These results show that maltose dictates MBP conformation, unless the transporter is energized with ATP. Taken together, the MalK 16R1/129R1 and MBP 41R1/211R1 data suggested that, in the membrane, only maltose-bound MBP was capable of stabilizing a semi-open configuration of the NBDs following incubation with MgATP, whereas in detergent, both open and closed MBP (Fig. 4) stabilized the semi-open configuration.

Figure 4. Conformation of MBP in complex with transporter in nanodiscs or detergent micelles.

The 200 G wide spectra of MBP-41R1/211R1 are shown with the indicated ligand additions. The spectra are adjusted to the same maximum amplitude and were recorded without (black) and with maltose (red). Arrows indicate broadening due to spin–spin interactions <20 Å.

As extensively described previously (Austermuhle et al., 2004), spectral changes also reveal an immobilization at both 41 and 211 positions, especially in the outward-facing conformation of the transporter, suggesting that engagement of both MBP lobes with the transporter is involved in the progression of the transport cycle.

Spin labeling of MBP C-lobe

Because the N-lobe of MBP was shown to interact with the transporter irrespective of the presence of maltose (Austermuhle et al., 2004; Bohm et al., 2013), we chose to primarily investigate the interaction of the C-lobe of MBP with the maltose transporter and analyze how it correlates with the conformational state of the MalK dimer analyzed earlier.

Eight residues located on the surface of MBP (at positions 202, 206, 215, 219, 341, 345, 352 and 353, Fig. 2A) were singly substituted with cysteine and modified with MTSL to generate the R1 side chain (Table 1). All of the mutants exhibited an in vivo transport positive phenotype because transformants grown overnight at 37 °C on maltose MacConkey agar yielded red colonies. The purified proteins containing R1 were tested for their ability to bind maltose by measuring the quenching of intrinsic fluorescence as a function of ligand concentration (Fig. S2) (Telmer and Shilton, 2003). As shown in Table 1, all the spin-labeled MBP proteins displayed K_d’s for maltose comparable to that of the wild type protein, indicating that the spin label did not interfere with substrate binding and closure of MBP.

Table 1.

Spin-labeling efficiency and affinity of spin labeled MBP mutants for maltose

MBP mutant	Spin-labeling efficiency (%)	Kd (μM)
WT	-	1.06 ± 0.23
202R1	75	0.58 ± 0.18
206R1	74	0.73 ± 0.15
215R1	77	0.96 ± 0.19
219R1	81	0.95 ± 0.28
341R1	68	0.89 ± 0.22
345R1	88	0.97 ± 0.25
352R1	78	1.75 ± 0.42
353R1	80	1.38 ± 0.27

The ability of the MBPs to stimulate the ATPase activity of the transporter stabilized in detergent (DDM) micelles or reconstituted into nanodiscs was also tested (Fig. 2B). The rate of stimulation of ATPase by 5 μM spin labeled maltose-MBPs ranged from 25% to 150% of its activity in the presence of the wild type MBP (~2 μmol/min/mg), both in detergent and in nanodiscs, indicating that the presence of the spin label did not prevent MBP from interacting productively with the transporter. Similar MBP-stimulated ATPase rates were seen when the spin-labeled MBPs were incubated with DTT (data not shown) to remove the spin label on MBP indicating that the mutation itself rather than the spin label was primarily responsible for variation in activity. The variation between rates decreased in the presence of higher MBP concentration (30 μM, data not shown) suggesting the mutations may have moderately decreased the affinity between MBP and the transporter.

Engagement of the C-lobe of MBP with the transporter

Each spin labeled MBP mutant was stepped through the transport cycle as previously described. In separate experiments, MalFGK₂ and MgADP (or MgATP where indicated) were added directly to mimic the post-hydrolysis state (Orelle et al., 2008); MalFGK₂, MgATP and orthovanadate (V_i) were added to trap the system in the catalytic transition state (with MgADP and V_i bound) (Chen et al., 2001; Sharma and Davidson, 2000); Ficoll 400 was added to simulate a decrease in the tumbling rate that MBP would experience following binding to the transporter (Lopez et al., 2009). These experiments were first conducted in detergent micelles to identify spin labels positions on MBP that sense the interaction with the transporter and to determine whether the C-lobe interacts with the transporter at all stages of the catalytic cycle.

The EPR spectrum of MBP-352R1 (black line, Fig. 5A) had the characteristic lineshape of an R1 on a helix surface (Columbus et al., 2001) and the spin label mobility changed only slightly upon addition of maltose (red line, Fig. 5A). Following addition of MalFGK₂ in detergent micelles in the absence of nucleotide, there was only a modest decrease in mobility. A similar decrease in mobility was observed when 20% Ficoll 400 was added to MBP, suggesting that the mobility change resulted from a loss of rotational tumbling upon binding of MBP to the transporter as opposed to any direct contact of 352R1 with the transporter. Under these experimental conditions, in which the transporter is present in 2-fold molar excess relatively to MBP, it was previously determined that ~60% of MBP is bound to the transporter (Austermuhle et al., 2004). Upon addition of ATP, a more marked decrease in the mobility of the spin label was seen, as judged by the broadening of the spectrum. A population of MBP-352R1 did not respond to ligand addition due to the presence of unbound MBP. Similar lineshape alterations were observed in the vanadate-trapped state, suggesting that these changes reflected the engagement of the open MBP to the outward-facing transporter, which is known to involve both lobes (Austermuhle et al., 2004; Oldham et al., 2007). These changes were not maltose-dependent and are thus consistent with the fact that the transporter can be stabilized in the outward-facing conformation both in the presence and absence of maltose (Bohm et al., 2013; Orelle et al., 2008). Addition of MgCl₂ to the sample containing MalFGK₂ and ATP resulted in further changes in the EPR lineshape. A highly immobile component (2A_zz ~ 70–71 G) emerged (see grey line, Fig. 5A), which was more pronounced in the presence than in the absence of maltose. This component was also present when MgCl₂ and ADP were added directly. These data suggested that the C-lobe of MBP remained in contact with the transporter in the presence of MgADP, although the motional changes in the EPR spectra suggested a different mode of interaction from that was seen in the transition state-like conformation (i.e. MgATP/V_i) in which MBP is open. Comparison of the structures of MalFGK₂ in complex with open (Fig. 1D) and closed MBP (Fig. 1B or 1C) reveal differences in the mode of interaction that could generally account for the differences in spin label mobility seen in the overlays with and without maltose when the C-lobe made contact with MalF (Fig. S3).

Figure 5. Interaction of MBP(352R1) with MalFGK₂ during the catalytic cycle.

A spin label was positioned in the C-lobe of MBP at residue 352 and the interaction of MBP with the transporter in detergent (panel A) or nanodiscs (panel B) was studied. The 100 G wide spectra were normalized to the same number of spins and the amplitudes were adjusted by the factor indicated if different from one. MBP was present at 60 μM and MalFGK₂ at 120 μM. Ligands were added at the concentrations specified in Experimental Procedures. The spectra recorded without (black) or with (red) maltose are overlaid. A) EPR spectra of MBP-352R1 incubated with MalFGK₂ in detergent micelles. A vertical line (grey) was drawn to indicate the immobile component of the spectra. B) EPR spectra of MBP-352R1 incubated with MalFGK₂ reconstituted in nanodiscs. Inset spectra on the left are shown to compare the post-hydrolysis states in nanodiscs and detergent micelles.

Changes in the EPR spectra of MBP-345R1 (Fig. 6) during the transport cycle also suggested that the C-lobe of MBP was engaged with the transporter in both the transition state-like conformation and in the presence of MgADP, but not in the absence of nucleotide. The spectrum of MBP-345R1 alone (Fig. 6) had a lineshape typical of a relatively mobile spin label (see arrow) on the surface of a protein (McHaourab et al., 1996). A very slight increase in mobility was observed upon closure of the MBP lobes in the presence of maltose. Addition of the transporter resulted in a new small immobile population, similar to the spectrum obtained for 345R1 in Ficoll 400, suggesting again that the decrease in mobility resulted from decreased tumbling of MBP rather than interaction of 345R1 directly with the transporter. Addition of ATP resulted in a significant increase in the immobile component (2A_zz ~ 70 G, see arrow), also seen in the vanadate-trapped state. Upon addition of MgCl₂ to allow ATP hydrolysis, or direct addition of MgADP, the spectra of MBP-345R1 retained the general features of the ATP-bound state, although the immobile peak (2A_zz ~ 68 G) was somewhat less prominent.

Figure 6. Interaction of the MBP(345R1) with MalFGK₂ during the catalytic cycle.

A spin label was positioned in the C-lobe of MBP at residue 345 and the interaction of MBP with the transporter was studied in detergent. The 100 G wide spectra were normalized to the same number of spins and the amplitudes were adjusted by the factor indicated if different from one. MBP was present at 60 μM and MalFGK₂ at 120 μM. Ligands were added at the concentrations specified in Experimental Procedures. The spectra recorded without (black) or with (red) maltose are overlaid. A vertical line (grey) was drawn to indicate the immobile component of the spectra. A black arrow points at the mobile component in MBP-345R1.

EPR spectra of MBP-202R1, 206R1, 215R1, 219R1, 341R1 and 353R1 are presented in Fig. S4. The minor changes in these spectra upon addition of MalFGK₂ could be explained by decreased tumbling of the molecule, as assessed by addition of Ficoll 400 to MBP alone (not shown), suggesting that no part of the C-lobe studied here interacted with the transporter in the absence of nucleotide. Decreases in spin label mobility were observed at all positions upon the addition of MgATP/V_i (Fig. S4). Upon direct addition of MgADP, the spectra of MBP-202R1 and 206R1 did not change and more closely resembled the MalFGK₂ state, the spectrum of MBP-353R1 overlaid with the MgATP/V_i spectrum, and MBP-215R1, 219R1, and 341R1, displayed a lineshape indicative of a motion between the apo and MgATP/V_i conditions. The three sites 202R1, 206R1 and 353R1 are located >5 Å from the transporter in the available structures of MBP complexed to MalFGK₂ (Oldham and Chen, 2011a) and showed modest changes in mobility, while the remaining residues were closer to the transporter. Based on the analysis of all the spin-labeled MBP mutants, we found no evidence that the C-lobe of MBP engaged with the transporter in the absence of nucleotide. However, a substantial portion of the C-lobe of MBP was in direct contact with the transporter in the presence of both MgADP and MgATP/V_i (i.e. MgADP/V_i, assuming V_i trapping occurred).

MBP interactions with the reconstituted transporter

Nanodiscs have proven to be a superior way to assess protein-protein interaction and conformational change by EPR in a membrane environment (Alvarez et al., 2010; Zou and McHaourab, 2010). Difficulties in achieving highly concentrated preparations of proteoliposomes, as needed for EPR, and of consistently achieving full access to both sides of the membrane have limited the usefulness of proteoliposomes. The characteristics of MBP-stimulated ATPase activity observed in proteoliposomes, which include a low basal ATPase rate in the absence of MBP and a strong stimulation by maltose (Davidson et al., 1992; Gould et al., 2009), were mirrored in nanodiscs but not by transporter in detergent solution. Importantly, the amount and composition of phospholipids in nanodiscs is critical for tight substrate coupling in the maltose transporter (Bao et al., 2013). We reported above that in lipid bilayers, only maltose-bound MBP efficiently stimulated the ATPase activity of MalFGK₂. However, in detergent, such stimulation was insensitive to maltose (Fig. 2B). The MalK 16R1/129R1 and MBP41R1/211R1 data suggested that, in the membrane, only maltose-bound MBP was capable of stabilizing a semi-open configuration of the NBDs following incubation with MgATP whereas, in detergent, both open and closed MBP stabilized the semi-open configuration (Fig. 3 and 4). Because the semi-open conformation of MalK₂ is stabilized by the engagement of both MBP lobes in the crystal structure of the pretranslocation state (Fig. 1B or 1C), we investigated whether the engagement of the C-lobe is maltose-dependent in the nanodisc system. We first chose to analyze the interaction of the mutant MBP-352R1 with MalFGK₂ reconstituted in nanodiscs (Alvarez et al., 2010). This mutant showed similar ATPase stimulation properties as the wild-type MBP (Fig. 2B) and reported on the C-lobe interaction with the transporter in detergent (Fig. 5A). Similar mobility changes were observed as in DDM-stabilized MalFGK₂, except that in nanodiscs, the C-lobe interaction in the presence of MgADP was only apparent if maltose was present (see the immobile component in enlargement, Fig. 5B). We also tested the interaction of MBP-345R1 and MBP-215R1 with the transporter reconstituted in nanodiscs. The MBP-215R1 showed a slight maltose-dependence, whereas MBP-345R1 did not (Fig. S5). This suggested that either these positions failed to sense the interaction with the transporter or the maltose-dependent engagement involves only some regions of the C-lobe. Next, we spin-labeled MalF at position 479, which is in close contact with the C-lobe (Fig. 7). The spin-labeled mutant was active and displayed a stimulated ATPase activity of ~1270 and ~1680 nmol/min/mg in detergent and nanodiscs, respectively. In the absence of ligand or in the presence of liganded MBP solely, the spectra of MalF(479R1) in detergent and in nanodiscs displayed mostly a highly mobile component (solid black arrow). In the presence of ATP/EDTA, spin mobility was strongly reduced in both detergent and nanodisc system. The central peak lineshape broadened and the reduced mobile component was accompanied by an increased immobile component (dashed arrow). However, in the post-hydrolysis state (MgADP), an obvious difference was observed in the detergent and nanodisc systems. In the detergent system, immobilization occurred regardless of the presence of maltose. In the nanodisc system, the decrease in mobility after ATP hydrolysis was much more pronounced when maltose was present (Fig. 7 and inset). Altogether, these results suggested that the engagement of the C-lobe became maltose dependent when the transporter was reconstituted in a lipid bilayer and correlated with the stabilization of the nucleotide-binding domains in the semi-open state.

Figure 7. Interaction of MBP with MalF(479R1)GK₂ in nanodiscs or detergent micelles.

100 G wide spectra were normalized to the same number of spins. Amplitudes were adjusted by the factor indicated in the figure if different from one. MBP was present at 60 μM and MalFGK₂ at 120 μM. Other additions, as indicated, were present at the concentrations specified in Experimental Procedures. The spectra were recorded without (black) and with maltose (red). EPR spectra of MalF(479R1)GK₂ in detergent micelles (left) or nanodiscs (right). Inset shows the overlay of the spectra recorded in the post-hydrolysis state. The black spectrum is not visible because it superimposes with the blue spectrum.

Distance change between MBP C-lobe and MalF

According to pulsed double electron-electron resonance (DEER) distance measurements between the C-lobe of MBP and MalG, maltose binding to MBP causes the C-lobe to come in the vicinity of the transporter (Bohm et al., 2013). In order to further study this interaction and assess the amplitude of the movement involved in the engagement of C-lobe with the transporter, we studied by DEER the interaction between MBP-202R1 and MalF-445R1 reconstituted in nanodiscs (Fig. S6). The stimulated ATPase activity of the reconstituted MalF-445R1 was ~1560 nmol/min/mg. In the absence of maltose, the distance between these positions was centered around 37 Å. In the presence of maltose, however, a substantial population of spin labels became centered around 28 Å. The additional presence of MgATP did not significantly alter these distances, although a higher population transitioned to 28 Å in the presence of maltose. Altogether, these results suggest that maltose binding to MBP causes the C-lobe to come in the immediate vicinity of MalF. Subtle but critical interactions likely occur at the interface when MgATP is present that promote a full engagement of the C-lobe thereby stabilizing the semi-open state of MalK₂.

Discussion

Interactions of MBP with MalFGK₂ are integral to the mechanism for coupling of transport to ATP hydrolysis. In this report, we investigated the mechanism by which maltose-bound MBP stimulates ATP hydrolysis, and showed that liganded MBP contributes to the stability of a semi-open configuration of NBDs. Whereas the N-lobe is known to interact with the transporter in all conformational states, we found that the C-lobe was not engaged in the resting state with NBDs open, consistent with previous EPR studies showing that maltose-bound MBP was unable to bring the NBDs closer in absence of nucleotides (Bohm et al., 2013; Orelle et al., 2008). However, we observed that interactions between the C-lobe of MBP and the transporter are facilitated by nucleotides and coincide with the NBDs being either semi-open or closed. These results suggested that the engagement of MBP C-lobe is critical to stabilizing major conformational states in MalFGK₂.

Evidence for a semi-open NBD configuration was first observed by EPR after allowing ATP hydrolysis to occur in the presence of maltose and MBP or upon direct addition of MgADP in the presence of maltose-MBP, suggesting it may represent a post-hydrolysis state (Orelle et al., 2008). Later, a semi-open configuration was observed in the presence of a non-hydrolyzable ATP analogue MgAMP-PNP and an MBP containing a disulfide bond that effectively locked it in a closed, maltose-bound conformation, raising the possibility that stabilization of a semi-open configuration may be critical prior to ATP hydrolysis (Orelle et al., 2010). Here, we found that MBP and MgADP stabilized the semi-open MalK dimer regardless of the presence of maltose when the transporter was in detergent, but only when maltose was present if the transporter was reconstituted into the membrane. These results suggested that maltose-free MBP does not stabilize a semi-open configuration in the native lipid environment.

During a cycle of maltose transport, NBDs would pass through a semi-open configuration twice, both in the progression to the outward-facing, transition state and in the return to the inward-facing, resting state after ATP hydrolysis has occurred. Maltose would be bound to MBP when the transporter first passes through this state, and absent from MBP, having been transferred to MalFGK₂, upon return to the resting state. Hence, in this work, it is logical to relate the maltose-bound intermediates to a pretranslocation state and the maltose-free intermediates to a post-hydrolysis step. In this light, ADP could be viewed either as a product of ATP hydrolysis, or as an analogue of ATP that fails to stabilize the fully closed NBD state. Hence, the semi-open conformation seen in the presence of maltose-MBP and MgADP may represent the reported pretranslocation structure (Oldham and Chen, 2011a). EPR-based measurements indicated that the spin labels in the NBDs were centered at a distance of 17 Å with MgAMP-PNP, maltose and the locked-closed MBP and at 12 Å with MgADP, maltose and the wild-type MBP (Orelle et al., 2008; Orelle et al., 2010). In the former case, the conformation may be dictated by the inability of MBP to open, while in the latter case, it is dictated by the absence of the gamma phosphate, which prevents full closure of the NBDs.

The crystal structure of the transporter in the semi-open NBD configuration with closed, liganded MBP bound offers insight into our observation that the C-lobe of MBP is not docked to the transporter when the NBDs are fully open (Oldham and Chen, 2011a). In the crystal structure, the TM-helices have started to rotate towards the transition state from the resting state. This result suggests that the transporter in the resting state does not present a surface complementary to the shape of either the open MBP or the closed MBP, rather the closed MBP can dock onto the semi-open transporter and the surface of the open MBP is complementary to the outward-facing conformation. Since ligand-bound MBP was required to stabilize the semi-open state in a lipid bilayer, the binding of closed MBP to this intermediate is likely an essential part of the translocation mechanism. As illustrated in Fig. 8, binding of maltose to MBP triggers its closure and closed MBP can bind to and stabilize the semi-open conformation of the transporter, thereby promoting further progression to the transition state, aided by high affinity binding between the outward-facing transporter and the open MBP. When maltose is absent, MBP is open, the semi-open configuration cannot be stabilized, and the transporter returns to the resting state rather than progressing to the transition state, thereby minimizing futile cycles of ATP hydrolysis. It is of note that the transporter crystallized in a semi-open configuration even without nucleotide bound (Oldham and Chen, 2011a), although both maltose-MBP and nucleotide were required to stabilize this intermediate in detergent and membrane (Orelle et al., 2008; Orelle et al., 2010). The reason for this discrepancy is unclear but could be due to the timescale difference between the two types of approaches or to molecular crowding effects (Dong et al., 2010) inherent to crystallography. While these effects may be physiologically relevant (Crowley et al., 2008), our results nevertheless indicate that nucleotide binding accelerates the transition to the semi-open state. Given the prevailing concentrations of ATP in vivo (Buckstein et al., 2008), albeit the concentration of free ATP is less clear (Schneider and Gourse, 2004), the transporter is likely to bind ATP relatively frequently and be primed for a cycle of ATP hydrolysis once MBP binds maltose. As displayed in Fig. 8, we thus suggest that the transport cycle is likely to go through the bottom route shown in Fig. 1.

Figure 8. Model of MBP interactions with MalFGK₂ during the transport cycle.

The transport cycle is drawn with ATP and MBP already bound to MalFGK₂ because the concentrations of ATP and MBP are thought to be saturating inside the cell (Buckstein et al., 2008; Manson et al., 1985), although neither is bound with high affinity to the inward-facing, resting state. MalF (grey) and MalG span the membrane, and MalF has a large periplasmic (P2) loop that binds MBP. The NBDs of the MalK dimer open and close throughout the cycle while the C-terminal regulatory domains remain in contact. In the inward-facing state (A and B), the NBDs are open with ATP bound. MBP interacts with the transporter in this state via the N-lobe both in the absence and presence of maltose. In an intermediate state that occludes the maltose binding site in MalF from both sides of the membrane (C and E), the NBDs are semi-open and the periplasmic interface of the TMDs are complementary to the closed form of MBP (C) but not to the open form of MBP (E). In the absence of maltose, and therefore of a C-lobe interaction, the intermediate state (E) is not stable and returns to the inward-facing state (A). With maltose present and both lobes of MBP engaged (C), the transporter progresses to the outward-facing state (D) in which the periplasmic interface of the TMDs is complementary to the open form of MBP, MBP has opened, NBDs have fully closed and maltose has been transferred from MBP to the sugar binding site in MalF. Hydrolysis of ATP in this state destabilizes the outward-facing state and resets the transporter to the inward-facing state, releasing maltose inside the cell and allowing exchange of ADP for ATP. Maltose can potentially trigger a cycle of ATP hydrolysis either by binding directly to the open MBP-MalFGK₂ complex (A) or by binding first to MBP in solution, which then docks on MalFGK₂.

The failure of maltose-free MBP and MgADP to stabilize a semi-open NBD configuration in nanodiscs suggests that the semi-open configuration has no special significance in the return to the resting state. Rather, it is likely that following ATP hydrolysis and the concomitant destabilization of the closed NBD conformation, the transporter begins to reorient from outward to inward-facing, and at some point along that trajectory, the MBP is no longer tightly bound, one or both lobes disengage, and the transporter then returns directly to the resting state. The energy cost required for lobe closure in the absence of maltose likely outweighs any energetic advantage associated with binding of both lobes to the transporter in the semi-open state. The transporter is likely more flexible in detergent allowing MBP to remain in the energetically favorable open-MBP-transporter interaction even though the NBDs are partially separated.

The low ATPase activity triggered by interaction with ligand-free MBP in the membrane environment has been studied. While it could result from a failure to remove all maltose from MBP (Silhavy et al., 1975), or from the ability of MBP to close in the absence of maltose (Tang et al., 2007), careful studies showed that the apparent K_m for maltose-free MBP in stimulation of ATP hydrolysis is much smaller than that seen for maltose-bound MBP (Gould et al., 2009). This result suggests that open MBP interacts directly with a different transporter conformation that is rarely sampled. One possibility is that the reaction that occurred in detergent, which allowed open MBP to dock to the semi-open NBD state, still occurs in membrane-reconstituted systems, although infrequently.

The stability of the open MBP/semi-open NBD complex in detergent may be responsible for the observation that MBP stimulates ATPase equally well in the presence or absence of maltose in detergent, but not in a membrane environment (see Fig. 2A). If our model for transport is correct, then the interaction of closed MBP with the semi-open intermediate is a key step in the activation of ATPase that ensures coupling of transport to hydrolysis. This state likely has a relatively high energy compared with the inward-facing and outward-facing conformations, and likely represent an energy barrier for their interconversion. Its presence limits the likelihood for uncoupled ATP hydrolysis because, if open MBP were able to stabilize this conformation, then the transporter could proceed to hydrolyze ATP without maltose being transported. In excellent agreement with our observations, Shilton recently proposed that, if maltose-bound MBP were to preferentially interact with MalFGK₂ in its resting state, it would probably stabilize a low-energy conformation that is incompetent for ATP hydrolysis (Shilton, 2015). However, binding to a higher-energy conformation, intermediate between resting state and outward-facing state, would offer an energetically rational pathway for liganded MBP to promote conformational change in the system (Shilton, 2015). Nucleotide binding is likely to lower the energy of this state since, according to our results, the engagement of the C-lobe was clearly nucleotide-dependent. Thus, closure of MBP plays two critical roles in the transport process, first to enable high affinity transport, and second, to ensure that cycles of ATP hydrolysis are accompanied by maltose transport. While this realization was made some time ago (Ames, 1990; Davidson et al., 1992; Quiocho, 1990), the details of the mechanism are only just coming to light.

Recently, Bao and Duong proposed a completely different transport model based on cross-linking and fluorescence studies (Bao and Duong, 2013). In their model, ATP binding alone is sufficient to promote the outward-facing conformation of MalFGK₂ and open, unliganded, MBP binds tightly to this conformation. When maltose is present, it diffuses to the substrate-binding site of MBP in complex with the outward-facing MalFGK₂, with a K_d estimated at ~120 μM. This process stimulates ATPase activity by facilitating the release of MBP and the return to the inward-facing conformation. This model is not supported by the available crystal structures wherein the cavity formed by MBP in complex with the outward-facing MalFGK₂ is not accessible to periplasmic maltose (Oldham et al., 2007; Oldham and Chen, 2011b). In addition, a periplasmic MalG loop inserts into the cleft between the lobes of MBP, thereby occupies the maltose-binding site in the open MBP. Consistent with the crystal structure, deletion of this “scoop” loop uncouples maltose transport from ATPase activity (Cui et al., 2010). Furthermore, this model is difficult to reconcile with transport rates observed in vivo and in vitro, since the K_m for maltose transport was found very close to the K_d of MBP for its substrate (~1 μM) (Manson et al., 1985; Szmelcman et al., 1976). Mathematical analysis of maltose transport suggested that both liganded and unliganded MBP interact with the transporter but the liganded form binds with higher affinity, and K_m approaches K_d when MBP concentration becomes large (Bohl and Boos, 1997; Merino et al., 1995). It is also problematic to envision a productive transport in the model proposed by Bao and Duong, considering that MBP is present at concentrations as high as 1 mM in the periplasm (Dietzel et al., 1978), and thus, most of the maltose would preferentially bind the free periplasmic MBP over the MBP that is bound to the transporter.

In contrast, the transport model we propose here is supported by decades of functional and structural data. The key concept of our model is that MBP is required to stabilize the outward-facing conformation (Bohm et al., 2013; Orelle et al., 2008). According to our studies, maltose-bound MBP stimulates the ATPase activity of MalFGK₂ by at least two mechanisms. When MBP is liganded, interaction of its C-lobe with the transporter is facilitated in presence of ATP and induces the pretranslocation state (Oldham and Chen, 2011a), thus explaining the well-known enhancement of ATP hydrolysis by maltose. Because the NBDs are closer together in this state, ATP can complete the progression to the outward-facing state (Oldham and Chen, 2011a; Orelle et al., 2010), in which MBP opens to deliver the substrate into the MalF binding site, aided by MalG scoop loop (Oldham et al., 2007). In this conformation, the open MBP also participates in the stimulation of ATPase activity by stabilizing the ATP hydrolysis transition state (Chen et al., 2001; Oldham and Chen, 2011b).

Experimental procedures

Mutagenesis

The Quikchange II XL site-directed mutagenesis kit (Stratagene) was used to generate single mutations in a variant of the malE gene that encodes MBP with a C-terminal polyhistidine tag (Austermuhle et al., 2004) or of the malF gene that encodes a cysteine-free version of MalF (Samanta et al., 2003). The sequences of the mutant genes were confirmed by DNA sequencing. In vivo function of the mutant proteins was verified by complementation of a chromosomal deletion of malE (strain BAR1000 (Fikes and Bassford, 1987)), or of a transposon insertion in malF (strain AD110, Δ(lac pro) supE thi malF::Tn10, F’traD36 proAB lacIq ZΔM15), as judged on maltose MacConkey agar (Difco) (Miller, 1972).

Expression, purification and spin labeling of MBP

E. coli strain BAR1000 was used for overexpression of MBP. The protein was purified essentially as described previously (Austermuhle et al., 2004) and dialyzed against buffer A (20 mM Tris-HCl pH 8, 150 mM NaCl). For spin labeling of cysteine mutants, 1 ml of a 200 μM solution of purified MBP was pretreated with 200 μM dithiothreitol (DTT), then incubated overnight with rotation at 4°C in the presence of 2 mM MTSL (1-oxyl-2,2,5,5-tetramethyl-Δ3-pyrroline-3-methyl methanethiosulfonate spin label, Toronto Research Chemicals) and 500 μL nickel-Sepharose 6 Fast Flow resin (GE Healthcare). The resin was collected in a column, washed with buffer A to remove excess spin label, eluted with buffer A plus 150 mM imidazole and dialyzed to remove imidazole. Protein concentrations were determined spectrophotometrically using an extinction coefficient of 1.7 (ε^0.1%_1cm) for MBP (Gehring et al., 1991).

Maltose-binding assay

Binding of maltose to spin-labeled MBP was assessed by measuring quenching of intrinsic fluorescence (Martineau et al., 1990; Miller et al., 1983). Changes in the fluorescence intensity as a function of maltose concentration were recorded with an excitation wavelength of 280 nm and emission spectra monitored from 300–400 nm using a Varian Cary Eclipse spectrofluorometer. All measurements were done at room temperature using 0.25 μM MBP in 3 mL of 20 mM Tris-HCl pH 8, varying maltose from 0 μM to 8 μM. The K_d values ± standard deviation were calculated by fitting the data with GraFit version 7 (Erithacus software) to the equation F = F_max − {(F_max − F_min) [(E_t + L + K_d) − ((E_t + L + K_d)² − 4EtL)^1/2]}/2E_t (Divita et al., 1993), where F_max is the fluorescence intensity at the start of the titration, F_min is the fluorescence at saturating concentration of maltose, Et is the total concentration of MBP and L is the maltose concentration.

Purification and spin-labeling of MalFGK₂

MalFGK₂ containing a polyhistidine tag at the C-terminus of MalK was overexpressed and purified as described previously (Orelle et al., 2008) except that the binding buffer contained 150 mM NaCl and 5 mM imidazole and protein was allowed to bind to the cobalt resin in batch during an overnight incubation at 4°C. For purification and spin labeling of the MalF(445)GK₂ or MalF(479)GK₂ transporters that contain a single cysteine at position 445 or 479 of MalF, the following modifications to the published procedure (Orelle et al., 2008) were incorporated: the proteins were incubated with TALON resin (Clontech) for two hours in the presence of 10 mM 2-mercaptoethanol, and the resin was placed into a column and washed to remove unbound protein and reducing agent. While still bound to the resin, MalFGK₂ was incubated with a ten-fold molar excess of spin label for two hours with gentle mixing by inversion. Finally, the resin was washed extensively to remove unbound spin label prior to elution of the transporter complex. Protein concentrations were determined using a modified Bradford assay (BioRad Protein Assay Kit I).

ATPase assay

ATPase activities were monitored at 25 °C using a coupled assay that measures the rate of conversion of NADH to NAD⁺ at 340 nm and regenerates the hydrolyzed ATP. Typically, 90 nM of transporter was present in a final volume of 350 μL of an ATPase reaction mixture containing 50 mM HEPES/KOH pH 8, 10 mM MgCl₂, 4 mM phosphoenolpyruvate, 60 μg/mL pyruvate kinase, 32 μg/mL lactate dehydrogenase, 0.3 mM NADH and 1.5 mM ATP, in the presence or absence of 5 μM MBP, 100 μM maltose and 0.01% n-dodecyl-β-D-maltoside (DDM). The transporter was either stabilized in detergent (DDM) micelles or reconstituted in nanodiscs. Nanodiscs were prepared as described previously using MSP1E3D1 (membrane scaffold protein) (Alvarez et al., 2010). Briefly, the nanodiscs were prepared using a lipid:MSP:MalFGK₂ molar ratio of 80:5:1. Cholate was added to the lipid stock (~50 mM of soybean lipids, Sigma P5638, in 50 mM Tris-HCl pH 7.5) at 25 mM final concentration. The sonicated, solubilized lipids, DDM-stabilized MalFGK₂ and purified MSP1E3D1 were mixed while maintaining cholate at 25 mM final concentration. The assembly mixture was allowed to equilibrate for 1 h at room temperature with gentle rocking. Biobeads SM-2 (BioRad) were added to the assembly mixture (amount of Biobeads ~2/3 of mixture volume) and incubated for 3 h at room temperature with gentle rocking to remove detergent and trigger self-assembly of the nanodiscs. The Biobeads were removed and the sample was then centrifuged at 10,000 × g to remove any precipitates.

EPR spectroscopy

CW-EPR spectroscopy was carried out on a EMX-plus spectrometer (Bruker) fitted with an ER4119HS cavity using WinEPR v4.40 as the computer interface. Thirty μL samples were contained in a glass capillary plugged at one end and inserted in a quartz tube. Ligands were used at the following concentrations: 10 mM ATP/1 mM EDTA, 10 mM ADP, 10 mM MgCl₂, 1 mM maltose, 1 mM orthovanadate. Experiments with MBP alone or in the presence of nanodiscs were performed in 20 mM Tris-HCl pH 8 or as indicated, while those with the transporter in micelles were performed in 20 mM Tris-HCl pH 8, 50 mM NaCl, 3 % glycerol and 0.01% DDM. The spectra were recorded at room temperature using 10 mW microwave power and 100 kHz modulation amplitude over 1.0 Gauss (G). The spectra were generally signal-averaged nine times over a 200 G scan width. Where indicated, the spectra were normalized to the same number of spins, calculated using Bruker WinEPR Processing, as the normalized double integral of the signal. The degree of mobility/immobility is often signified by 2A_zz values, the splitting between the outer lines, which are measured in G (e.g. (Hubbell et al., 1996; Klug and Feix, 2008)). Spin–spin distances were determined as previously reported (Orelle et al., 2010) by fitting the experimental EPR data with simulations using shortdistances100, a custom program written in Labview 2009 by Dr. Christian Altenbach (University of California-Los Angeles, CA).

Double electron electron resonance (DEER) spectroscopy experiments were carried out at X-band on a Bruker ELEXSYS 580 spectrometer equipped with a 3 mm split-ring resonator (Bruker Biospin) using a four-pulse sequence (Jeschke et al., 2006). Thirty μL samples of MalF445R1 in nanodiscs and MBP202R1 with ligands as noted were contained in a sealed quartz capillary with 20% deuterated glycerol as a cryoprotectant. Each sample was flash frozen in an acetone/dry ice slurry before insertion into the resonator and run at 80 K. Data were phased, background corrected and analyzed for distance distributions using the LongDistances software program (Toledo Warshaviak et al., 2013) provided by C. Altenbach.