Substrate Transport Activation Is Mediated through Second Periplasmic Loop of Transmembrane Protein MalF in Maltose Transport Complex of Escherichia coli*

Background: ABC transporters are important substrate transport systems in all forms of life. Results: Residual dipolar coupling (RDCs) and solvent-PRE experiments are employed to characterize structural changes of MalE, induced upon addition of MalF-P2. Conclusion: MalF-P2 induces an activated structural conformation of the substrate-binding protein MalE, which is required for substrate transport. Significance: This work gives insight into the activation mechanism of ABC transporters. In a recent study we described the second periplasmic loop P2 of the transmembrane protein MalF (MalF-P2) of the maltose ATP-binding cassette transporter (MalFGK2-E) as an important element in the recognition of substrate by the maltose-binding protein MalE. In this study, we focus on MalE and find that MalE undergoes a structural rearrangement after addition of MalF-P2. Analysis of residual dipolar couplings (RDCs) shows that binding of MalF-P2 induces a semiopen state of MalE in the presence and absence of maltose, whereas maltose is retained in the binding pocket. These data are in agreement with paramagnetic relaxation enhancement experiments. After addition of MalF-P2, an increased solvent accessibility for residues in the vicinity of the maltose-binding site of MalE is observed. MalF-P2 is thus not only responsible for substrate recognition, but also directly involved in activation of substrate transport. The observation that substrate-bound and substrate-free MalE in the presence of MalF-P2 adopts a similar semiopen state hints at the origin of the futile ATP hydrolysis of MalFGK2-E.

In a recent study we described the second periplasmic loop P2 of the transmembrane protein MalF (MalF-P2) of the maltose ATP-binding cassette transporter (MalFGK 2 -E) as an important element in the recognition of substrate by the maltose-binding protein MalE. In this study, we focus on MalE and find that MalE undergoes a structural rearrangement after addition of MalF-P2. Analysis of residual dipolar couplings (RDCs) shows that binding of MalF-P2 induces a semiopen state of MalE in the presence and absence of maltose, whereas maltose is retained in the binding pocket. These data are in agreement with paramagnetic relaxation enhancement experiments. After addition of MalF-P2, an increased solvent accessibility for residues in the vicinity of the maltose-binding site of MalE is observed. MalF-P2 is thus not only responsible for substrate recognition, but also directly involved in activation of substrate transport. The observation that substrate-bound and substrate-free MalE in the presence of MalF-P2 adopts a similar semiopen state hints at the origin of the futile ATP hydrolysis of MalFGK 2 -E.
ATP-binding cassette (ABC) 4 transporters are active transporters of substrate that utilize the chemical energy stored in ATP to translocate the substrate across cellular membranes (1). Most ABC transporters are importers of substrate in bacteria and include a substrate-binding protein, two integral membrane components with low sequential homology, and two conserved nucleotide-binding domains (2). One of the best studied systems is the maltose transporter MalFGK 2 -E of Escherichia coli/Salmonella that consists of the periplasmic maltose-binding protein MalE, the two integral cytosolic membrane proteins MalF and MalG, and two copies of the nucleotide-binding domains MalK (3). A multitude of x-ray structures exists of the maltose transporter in its resting state, a catalytic-intermediate state and most recently in a pretranslocational state (4 -7). It has been shown that the rigid-body rotation of the nucleotide-binding domains is confined to ATP binding and hydrolysis, which coincides with the conformational changes during substrate transport (7)(8)(9). From structural information together with biochemical data, a model for the substrate translocation process has been developed for the ABC import system, the alternatingaccess model (10). An adaptation of this model to the MalFGK 2 -E system is shown in Fig. 1. To date it is not known how the translocation process is initiated. Especially the highly dynamic region of the periplasmic side of the transporter in interaction with the substrate-binding protein MalE has eluded high-resolution information. In a recent study we described the second periplasmic loop P2 of MalF as a key player in the recognition of substrate via MalE (11). In this study we show that MalF-P2 not only is responsible for substrate recognition but also directly involved in activation of the MalFGK 2 -E complex by inducing structural changes in maltose-bound MalE.  (17). Consequently, the alignment tensor was fitted again using a com-mon alignment frame. After rigid-body modulation to obtain the orientations of the N-and C-terminal domains in the common alignment frame, the resulting MalE structure in the MalE/MalF-P2 complex was acquired (17). The procedure was carried out for MalE in the presence and absence of maltose.
Boundaries of the N-and C-terminal domains were set accordingly, N: residues 1-110; 260 -313 and C: residues 115-257; 335-370, and the linking domain: residues 314 -334. The linking domain was omitted from the RDC fitting. Resulting PDBs and visualization of alignment tensors from the rigidbody modeling were derived from Module v1.0 (17).
Solvent-PRE Measurements-Paramagnetic relaxation enhancement (PRE) experiments were obtained by quantification of the HSQC peak intensities for maltose-bound MalE in the presence and absence of MalF-P2. Solvent-PREs were measured by the addition of chelated Gd 3ϩ (OmniScan) to the sample buffer. Stock solutions of Gd 3ϩ were prepared in sample buffer, which was titrated into the protein solution to yield a concentration of 0.5-50 mM Gd 3ϩ . Spectra were acquired with 0.5, 1.0, 2.0, 3.0, 5.0, 10.0, 25.0, and 50.0 mM Gd 3ϩ . Concentration-dependent HSQC cross-peak intensities I i were fitted according to Equation 1, and evaluated according to Hilty et al. (18). ⑀ i , T, and c refer to the relative solvation rate, the INEPT, delay and the Gd 3ϩ concentration, respectively. No significant changes in 15 N line widths of MalE due to chemical exchange upon addition of MalF-P2 in the presence and absence of maltose were observed. All NMR spectra were processed with TopSpin v2.0 (Bruker Biospin, Karlsruhe, Germany) and analyzed with CCPN analysis v2.4 (19). Solvent PREs and correlation factors for RDCs were fitted using Prism v5.0b (GraphPad). RDC alignment tensor fitting and rigid-body modulations were accomplished in Module v1.0 (17).  (10). The model contains three distinct steps. First, ATP is loaded into the MalK dimer, and the interface closes upon binding of substrate-loaded MalE at the periplasmic surface (II). Upon ATP hydrolysis substrate is released into the binding pocket of MalG,F and is transported into the cytoplasm (III). After substrate transport, ADP is released, and the transporter goes back to its resting-state (I). Amplitudes of motion for the TMDs are exaggerated in the representation to highlight changes. A TMD (MalF,G) rotation from the inward to the outward conformation was determined to be on the order of 22°(4). Isothermal Titration Calorimetry (ITC) Experiments-MalE and MalF-P2 ITC samples were extensively dialyzed overnight at 7°C (Spectral/Por, Spectrum Laboratories) either separately or mixed at a 1:1.2 (MalE:MalF-P2) molar ratio, with concentrations ranging from 0.5 to 1 mM in 20 mM phosphate buffer, pH 7.4, 100 mM NaCl, with light stirring. Maltose stocks were prepared from the dialysis buffer at concentrations ranging from 0.5 to 2 mM. MalE, MalF-P2, MalE/ MalF-P2, and maltose samples were filtered and degassed before insertion to the injection syringe and calorimeter cell.
High-sensitivity microcalorimetry was performed on a VP-ITC instrument (MicroCal). Experiments of maltose binding to MalE were performed at 5°C with either MalE only or a 1:1.   Experiments were performed with MalF-P2 (50 M) in the calorimetry cell and a 500 M maltose solution in the syringe. No interaction could be detected between maltose and MalF-P2 alone.
Maltose Binding Assay-Binding of [ 14 C]maltose to MalE in the absence and presence of MalF-P2 was monitored by a filtration assay according to Richarme and Kepes (20). Briefly, purified MalE (2.5 M) was incubated with [ 14 C]maltose (12 M) in the absence or presence of MalF-P2 (2.5 and 5 M, respectively) in 20 mM phosphate buffer, pH 7.4, 100 mM NaCl at 25°C for 15 min. The reaction was terminated by adding 2 ml of ice-cold saturated (NH 4 ) 2 SO 4 , and the mixture was immediately passed through a nitrocellulose filter (0.45-m pore size). After two washes with 5 ml of the (NH 4 ) 2 SO 4 solution followed by 5 ml of distilled water, the radioactivity retained on the filters was determined in a liquid scintillation counter.
Molecular Dynamics Simulation-The MalFGK 2 -E crystal structure (PDB ID code 2R6G) was taken as the starting point to perform a 50-ns molecular dynamics simulation in GROMACS using the GROMOS force field (ffG53a6) (21) of MalE and MalF-P2. The proteins were placed into a triclinic box. The simulations were done under physiological conditions. For this purpose the proteins and the complex structure were calculated in a 0.1 M NaCl (i.e. ϳ15,600 H 2 O molecules, 45 Na ϩ and 33 Cl Ϫ ions, respectively) solution at 310 K and periodic boundary conditions for the box. Short-range nonbonded interactions were calculated with the Cut-off Lennard-Jones potential up to a distance of 1.2 nm between the interacting atoms. For long-range electrostatic interactions the Particle Mesh Ewald (PME) option was used with a grid spacing of 0.12 nm. The bond length of a heavy atom (C, N, O) to an H-atom in protein or water was constrained using a LINCS and SETTLE algorithm, respectively. Every simulation ensemble was energy minimized in a two-step minimization strategy. In a first step steepest descent and in a second a conjugate gradient minimizing routine were used. After equilibration over a period of 500 ps using a positional restraint of 1000 kJmol Ϫ1 nm Ϫ2 on the backbone of the two proteins molecule, a free simulation was performed for a period of 50 ns for the MalE/MalF-P2 complex. The temperature and pressure (1 bar) were controlled using the Berendsen coupling with a relaxation time of 0.1 ps for temperature and 1 ps for pressure. During the course of the simulation, the actual frame was stored every 5 ps. Visualization of trajectories and arrangement of figures was made with VMD (22).

RESULTS
In an earlier report it was shown that the periplasmic loop P2 of MalF in the MalFGK 2 transporter folds independently in solution and can interact with MalE in the presence and absence of maltose (11). Here, the interaction of MalE to MalF-P2 was investigated.
NMR Titration of MalF-P2 to U- 15 (23). When titrating maltose into the measurement cell filled with a 1:1.2 molar ratio of MalE:MalF-P2, the fitted binding curve yields comparable binding affinity. However, the stoichiometry, R, is reduced to a fourth, indicating a lower activity of MalE relative to the reference experiment (Fig. 3) and indicating that release and rebinding of maltose to MalE only occur at negligible rates compared with the uncomplexed state.
To evaluate whether MalF-P2 can stimulate release of maltose from maltose-bound MalE, radioactive maltose precipitation experiments were performed. Maltose-bound MalE was precipitated in the absence or presence of either equimolar or 2-fold excess of MalF-P2 in solution. Radioactive counts were measured before and after precipitation of proteins in solution, and all experiments were performed in duplicate (Fig. 4). No significant difference in maltose counts was found for the maltose-bound MalE in the absence or presence of MalF-P2 in solution, implying that maltose is maintained bound in the substrate-pocket of MalE in the presence of MalF-P2.
RDC Measurements Elucidate Structural Changes of MalE-H N -N RDCs were extracted from spectra of maltose-bound MalE in the absence and presence of MalF-P2, 182 and 131 respectively. RDCs were analyzed employing the substrate-free, the maltose-bound MalE and the MalE structure in a catalytic intermediate state of MalFGK 2 -E (PDB ID codes 1PEB, 1MPD, and 2R6G, respectively). Further, the structures were separated into the N-terminal (residues 1-110; 260 -313) and C-terminal (residues 115-257; 335-370) domains to further evaluate the relative orientation of the two domains to one another. The linker regions between the N-and C-terminal domains were omitted in the analysis. Reference RDCs of MalE/maltose show a high correlation coefficient for the individual N-and C-terminal domains, R ϳ 0.97, and their relative orientation, r ϭ 0.89 (Table 1), with respect to the x-ray structure (PDB ID code 1MPD). H N -N RDCs were then extracted from the MalE/MalF-P2/maltose complex. The extracted RDCs were fitted to the maltose-free and maltose-bound structures of MalE as well as the MalE structure from the MalFGK 2 -E vanadate-trapped intermediate state (PDB ID codes 1PEB, 1MPD, and 2R6G, respectively) ( Table 1). A correlation coefficient of r ϭ 0.65, 0.76, and 0.61 for the N-and C-terminal domain and the fulllength structure, respectively, is observed. The lower correlation coefficient for the individual N-and C-terminal domains can be directly interpreted as a consequence of a conformational rearrangement in the binding interface and the C-terminal domain. Minor structural changes are corroborated by chemical shift changes in the C-terminal domain in the HSQC titration experiments (Fig. 2). The reduced correlation coefficient for the full structure, relative to the individual domains, shows that the relative orientation of the individual domains deviates from all of the previously solved x-ray structures (PDB ID codes 1PEB, 1MPD, and 2R6G).
Fitted alignment tensors from the MalE/MalF-P2/maltose complex to the individual domains used a 100-step Monte Carlo analysis in Module v1.0 (17) yielding a good convergence. The obtained alignment tensors for the individual Nand C-terminal domains are on the same order of magnitude, which allows for rigid-body modulation (Fig. 5) (17). The alignment tensors from the N-and C-terminal domains of MalE/MalF-P2/maltose are then set to a common alignment frame and allowed to undergo rigid-body rotations to accommodate the common alignment frame. MalE/MalF-P2/maltose-derived RDCs were then fitted to the resulting full-length structure after rigid-body modulation. This yields the relative orientation of the two domains in the MalE/MalF-P2/maltose complex (Fig. 5). An increase in the correlation factor is now observed (r ϭ 0.71), indicating that an improved relative domain orientation is found ( Table 1). The reorientation of the relative domains to each other

Ratio of solvent-PRE relaxation rates for substrate-bound MalE in presence and absence of MalF-P2
Relaxation rates of residues lining the binding pocket and/or directly associated with substrate binding were extracted and compared relatively with each other. A Ͼ5-fold increase in relaxation rates of residues located in the vicinity of the maltosebinding pocket is observed after addition of MalF-P2.

Residue
⑀   (Table 2), solvent-PRE measurements with chelated Gd 3ϩ (OmniScan) in solution were performed. Gd 3ϩ is a paramagnetic ion that drastically increases the relaxation rates of hydrogen atoms in its vicinity. Relaxation rates were measured via 1 H-15 N HSQC experiments recorded with increasing concentrations of Gd 3ϩ (0.5-50 mM) (Fig. 6). MalE amide 1 H, 15 N cross-peak intensities were fitted with respect to the Gd 3ϩ concentration. In principle, a reduced solvent accessibility is expected for residues in MalE that directly interact with MalF-P2. Residues not interacting with MalF-P2 should not experience any changes. However, the opposite effect was observed in our experiments. We found strongly increased solvation rates for residues Lys 15 , Glu 61 , Trp 62 , Ala 63 , Arg 66 , Val 110 , Glu 111 , Ala 112 , and Glu 153 , located directly in the substrate-binding pocket. As these residues are not directly involved in MalF-P2 binding, a change in the solvent accessibility of the binding cleft must occur. This is corroborated by the conformational changes in the N-and C-terminal domain of MalE, as seen in the changes of the RDC values.
Surface representations of the MalE/maltose structures, in the absence (PDB ID code 1PEB) and presence of MalF-P2 (Module-derived), highlight the increased solvent exposure of the binding pocket of the MalE/MalF-P2 complex. The increased solvent-accessibility of specific amino acids in the substrate-binding pocket is readily appreciated (Fig. 6).
Molecular Dynamics Simulation of MalE/MalF-P2 Complex-To further assess the MalE and MalF-P2 binding interface a GROMACS molecular dynamics simulation (21) was performed. As starting points for the simulated annealing the crystal structures of MalE and MalF-P2 in the MalFGK 2 complex were used. The full-length sequence of MalE was calculated for each simulation ensemble. The C-terminal part of MalF-P2, residues 266 -275, was discarded as earlier data show no interaction between MalE and MalF-P2 in this region (11). Further, the C-terminal ␣-helix ␣3 cannot fold up independently in solution (24). Root mean square deviation (r.m.s.d.) trajectories were plotted for MalE, MalF-P2, and the MalF-P2/MalE complex (Fig. 7). The molecular dynamics trajectory shows that MalE maintains its initial structure during the course of the simulation. Dynamics is only observed for surface-exposed loops in the C-terminal lobe of MalE. MalE thus behaves as a rigid body with a r.m.s.d. amplitude of 3 Å throughout the simulation. In contrast to MalE, MalF-P2 shows large dynamic variations throughout the simulation. The backbone r.m.s.d. initially rises to accommodate the binding surface to MalE properly. After 25 ns, MalF-P2 is properly positioned on the binding surface of MalE (Fig. 7). The second half of the simulation yields a higher r.m.s.d. trace, as the two domains of MalF-P2 are dynamic in their interaction with MalE. From NMR interaction studies it is known that the interaction between MalF-P2 and MalE particularly involves the two ␣-helices ␣1 and ␣2 (11) (Fig. 8). These ␣-helices are located in each of the two domains of MalF-P2. During the simulated annealing both domains of MalF-P2 are reorienting to accommodate MalE. After 40 -50 ns of the simulated annealing ␣-helix ␣1 in domain 1 reorients and moves toward domain 2 of MalF-P2 on the binding surface of MalE (Fig. 7). Whereas the angle between the ␣-helices at the beginning of the simulation is on the order of 30°, the two domains are oriented almost perpendicular to each other at the end of the trajectory.
A second 50-ns molecular dynamics simulation was performed to monitor the r.m.s.d. traces from the respective calculations. The two calculations converged and showed similar dependence on the dynamic behavior of the MalF-P2/MalE interaction.

DISCUSSION
In an earlier report we have demonstrated that the periplasmic loop P2 of MalF acts as a receptor that recognizes MalE (11). Here we extend this study and show that MalF-P2 induces an activated conformational state of MalE that precedes substrate transport in MalFGK 2 . Although x-ray structures have been able to give valuable snapshots of the maltose transporter in three different conformational states (4 -6), none could so far explain how the initial activation of MalFGK 2 at the periplasmic side is accomplished. A recent molecular dynamics study hints at the dynamic behavior of the association process, but experimental data are so far missing (25).
We found that substrate recognition is mediated by MalF-P2, which causes an alternative active conformation of the two domains of MalE in the presence of MalF-P2. From these new data we propose an adapted activation mechanism for the existing transport model (26) which is represented in (Fig. 9). Here, MalF-P2 first acts as a receptor for MalE (Fig. 9, Ia). In the presence of MalF-P2, a conformational change of MalE is induced. The substrate becomes solvent-exposed in the binding pocket, and concomitantly space is provided for insertion of Structures are shown at every 10 ns of the simulation with colors shifting from blue to white to red. The two ␣-helices of MalF-P2, ␣1 and ␣2, make up the main interaction with MalE (11). During the course of the simulation they undergo large rearrangements while still maintaining the interaction to MalE. the periplasmic loop P3 of MalG (Fig. 9, Ib). In the x-ray structure (6), the loop MalG-P3 is fully inserted into the binding pocket, effectively "scooping out" the substrate. So far, however, it is not understood how the MalG-P3 loop gets inserted into the binding cleft of MalE. MalE binds maltose with micromolar affinity (23) and creates a highly stable complex burying the substrate between the two domains (20). In the substratebound conformation, solvent does not have access to the binding cleft (23). Conformational changes of MalE triggered by MalF-P2 allow MalG-P3 either to interact directly with the substrate or to insert into the binding pocket, stabilizing the complex during the substrate import cycle (Fig. 9, Ic). ATP hydrolysis drives substrate transport, and much effort has been made to identify the interactions that cause hydrolysis (26). In an earlier study by Grote et al. (27) it was demonstrated that MalF-P2 signals across the lipid bilayer to the cytosolic part of MalFGK 2 , effectively communicating reciprocally over the cellular membrane during the substrate transport cycle (Fig. 9, Ic).
The MalE/MalF-P2 interaction is mediated through complementary electrostatic forces (Fig. 8). MalE has two large negatively charged patches at the binding interface to MalF-P2. Complementarily, MalF-P2 has one positively charged patch at each of its two domains around ␣-helices ␣1 and ␣2. In the molecular dynamics simulation, the two domains of MalF-P2 turn out to be highly dynamic. We speculate that the complementary electrostatic patches stabilize the complex, inducing structural changes in MalE, which are required to dock to the MalFGK 2 -E complex. The molecular dynamics simulations suggest that the interaction of MalF-P2 and MalE is highly dynamic with MalF-P2 "gliding" on the interaction surface of MalE. The dynamics is a necessity to facilitate the binding of MalE when docking to the transmembrane proteins MalF/G. Dynamics enables different kinds of binding modes of MalF-P2 with respect to MalE, which reflect the different states during substrate transport in the MalFGK 2 -E complex.
As shown here, MalF-P2 also affects the substrate-free MalE, inducing the activated conformation of MalE. This activated state would enable the insertion of MalG-P3 and the stabilization of the MalFGK 2 -E complex in the absence of substrate. In turn, this would cause ATP hydrolysis and explain the observed futile ATP hydrolysis of the transporter in the absence of substrate (28,29).