The energetics of structural change in maltose-binding protein

High-resolution structures of proteins and their complexes are now appearing in the Research Collaboratory for Structural Bioinformatics (RCSB) database at a tremendous rate. When analyzing the structures of proteins, it is often tempting to try to extract energetic contributions to stability or binding interactions. However, as many experiments have shown, some specific interactions that look important are not, whereas others that seem minor turn out to have key roles. In evaluating detailed structures, particularly to understand subtle structural differences in mutants, one always wants the highest-resolution structures available, and most of these come from crystallography. However, crystallography suffers from the “tyranny of the lattice” (the ability of the forces that stabilize the crystal lattice to alter the conformation of the molecules making up the crystal). Although this term has been used most often for nucleic acids (see ref. 1 for a discussion of the issues), the idea applies to proteins as well; there are many examples in which the structure of a loop is different in different crystal forms, leaving ambiguity as to the “true” conformation in solution.

The relative positions of domains in multidomain proteins also can be affected by intermolecular interactions. In this issue of PNAS, an NMR approach for determining relative domain orientation in solution has been applied to a series of mutants of a two-domain protein (2). By combining this approach with energetic information derived from careful protein stability measurements, it became possible to provide remarkably detailed insight into the energetic cost of induced structural change and how it couples to ligand binding.

Bending and Binding

The system studied was the periplasmic maltose-binding protein (MBP) from Escherichia coli, a member of a large class of proteins that aid in uptake of small molecules and also act as sensors for signaling through the chemotaxis system. These proteins have two domains of nearly equal size, with a binding cleft between them, termed a “clamshell.” In the ligand-free state the protein is “open,” exposing the binding site, whereas in the presence of ligand it is “closed,” trapping the ligand and providing contacts to it from both domains. Quiocho and colleagues (3) solved high-resolution structures of MBP in both the open and closed states by means of crystallography, demonstrating the binding-induced structural changes. The transition from open to closed state occurs through a rotation of the domains relative to one another, accomplished by changing the backbone angles of a few critical residues in the linkers between domains, with very little change in the structure within either domain and very little relative translation of domains. The change in interdomain angle from open to closed (maltose-bound) state is 35° (Fig. 1).

Fig. 1.

Structures of the open (blue, Protein Database ID code 1OMP) and closed (green, Protein Database ID code 1ANF) forms of MBP (with bound maltose in red) are shown. (Left) The lower domain was superimposed, and a line was drawn through a helix in the open form, with another drawn through the same helix of the closed form, rotating the line by 35°. (Center and Right) The open (Center) and closed (Right) forms are shown separately. Blue arrowheads show general regions of contact with the receptor, which are closely space in the closed form but well separated in the open state. Red arrowheads indicate the region in which packing is altered as the protein closes. The sites at which mutations that affect opening were identified are indicated in yellow.

Protein Mechanics

The chemotactic response signaling by MBP is affected by binding of the ligand–MBP complex to the receptor protein TAR. There are surface residues of MBP on both domains that contact the receptor (4). In the open state, these residues are too far apart to simultaneously contact the receptor, leading to low binding affinity. In the closed state, they are brought together, leading to high affinity for the receptor. In some proteins, hinges are viewed as very flexible tethers, allowing domain orientations to change at will. In the case of MBP (and presumably of other periplasmic binding proteins), this situation does not occur, because closure without ligand would lead to aberrant signaling. Hence, MBP has a “spring-loaded” hinge, the “spring” holding the protein open until the ligand binds and “pays” the energetic cost to close the protein. But what is the spring? How does it behave energetically? Does ligand-binding pull the protein closed, working against the spring, or does it change the conformation and somehow remove the spring's force? All of these questions were answered by Kay and coworkers (2), using a powerful NMR approach for analysis of domain orientations applied to a series of mutant proteins identified by Marvin and Hellinga (5) as mutants with higher maltose-binding affinity. Instead of making changes near the maltose-binding pocket, as might be expected to affect maltose affinity, Marvin and Hellinga made mutations altering residues on the “back” of the protein. This region changes conformation when the domains close, allosterically coupling to the binding event. The two sites at which substitutions were made are close together, and replacing the natural amino acids with larger ones, or adding cysteine and then a large, artificial side chain chemically, destabilizes the open state, making it easier to close the protein, in turn increasing the affinity for ligand.

Key New Data

The key NMR measurements to determine the degree of opening between the clamshell's domains were made by using residual dipolar couplings (6–8). This technique requires having the protein in a solution with liquid crystal characteristics. This solution is created by addition of large anisotropic (both physically and magnetically) objects such as rods (bacteriophage) or sheets (membrane fragments), which cooperatively and spontaneously align with one another and, when put into a magnet, also align with the field. Such a medium biases the rotation of other nonspherical molecules in the solution, causing them to have a slight preference for some orientations. Direct dipolar couplings between nuclei are usually averaged exactly to zero by molecular tumbling in normal, isotropic solutions. However, when there is an orientational preference, new spin splittings arise in the spectrum from these couplings, the magnitude depending on both the distance between the coupled nuclei and the angle that the internuclear vector makes with the applied field. Because bond lengths are known for backbone amide N-H pairs, measurements directly yield orientational information. When the structures of domains are known, then it is a relatively simple procedure to combine all of the couplings in rigid regions of each domain to give the orientation of that domain relative to the preferred molecular orientation. Of course, from these data the angle of the domains relative to one another can be easily calculated. Because data from many N-H pairs are combined in the fitting, the overall accuracy in determination of the domain orientation is very high: an uncertainty of ≈1° in these studies of MBP was estimated. Because interactions between the protein and the ordering medium (Pf1 phage in this case) are very weak, the relative domain orientation reflects the average degree of opening in solution. The values for open (ligand-free wild type) and closed (maltose-bound wild type) agree very well with the crystallographically determined values.

The next step was to characterize the energetics of the conformational changes occurring. Marvin and Hellinga (5) had shown that there was some correlation between the volume of the variant amino acid and the maltose dissociation constant, albeit with considerable scatter. A related correlation between change in heat capacity between bound and free states, largely due to exposed nonpolar surface (ΔΔC_p), and change in free energy of maltose binding (ΔΔG_b) was seen, again with considerable scatter. However, without any direct structural information, these relationships could not be interpreted further. With the information available from the NMR measurements made with their approach, Millet et al. (2) established a very strong correlation between the degree of closure of the domains and the binding affinity of maltose, the binding free energy (log of the dissociation constant) decreasing linearly with closure angle. This correlation showed that the amino acid substitutions created new steric interactions that held the domains partially closed, and the energy normally provided to accomplish this part of the closure was no longer provided by ligand binding. A double mutant, I329W/A96W, closed the interdomain angle by 28° (with full closure corresponding to 35°), increasing the maltose affinity by well over two orders of magnitude.

There is strong correlation between degree of domain closure and maltose-binding affinity.

Bending and Stability

Further information about the energetics came by measuring the stability of the folded state of each variant. Remarkably, there was again a strong correlation, in this case between the closure angle and the destabilization of the folded state. When corrected for the change in buried hydrophobic surface due to the mutation, the correlation between ΔG_U-F and the closure angle became linear. After examining the structures calculated for the variants, it became clear that, as the protein closed, there was an increase in exposed nonpolar surface area, localized in the hinge region, again correlating in a linear way with the closure angle. The observations give a value of ≈15 cal/Ų (1 cal = 4.184 J) buried, on the low end of estimates of energetics of nonpolar surface burial (9). The burial of this nonpolar surface in the open state provides part of the driving force both for protein folding and to hold the binding cleft open: it is the spring that prevents closure in the absence of ligand. Because the exposure of nonpolar surface continues to increase through the closing process (sampled experimentally for many values and to within 7° of full closure), the spring continues to exert an opening force for any conformation; presumably, evolution selected this as the optimum behavior for this process. When ligand binds, closure leads to new, favorable interactions with maltose that overcome the unfavorable energetics of stretching the spring. This cost of closure is then paid from the binding free energy, a reduction of ≈7 kcal/mol or ×10⁵ in binding constant, but nature seems to have binding energy to spare to keep the signaling response well suppressed in the absence of ligand. The highly accurate NMR data on closure angle were key in this elegant study of the detailed mechanics of a protein undergoing a conformational change.