|Home | About | Journals | Submit | Contact Us | Français|
Temperature dependence of dynamic parameters derived from nuclear magnetic resonance (NMR) relaxation data are related to conformational entropy of the system under study. This provides information such as macromolecules stability and thermodynamics of ligand binding. We studied the temperature dependence of NMR order parameter of Glutamine binding protein (GlnBP), a periplasmic binding protein (PBP) highly specific to L-glutamine associated with its ABC transporter, with the goal of elucidating the dynamical differences between the respective ligand bound and free forms. We found that the protein-ligand interaction, which is stabilized at higher temperature, has a striking effect on the stability of the hydrophobic core of the large domain of GlnBP. Moreover, in contrast to what was found for less specific PBPs the decreasing backbone motion of the hinge region at increasing temperature supports the idea that the likelihood that GlnBP can adopt a ligand free closed conformation in solution diminishes at higher temperatures. Our results support the induced-fit model as mode of action for GlnBP. In addition, we found that the backbones of residues involved in a salt bridge do not necessarily become more rigid as the temperature rises as it was previously suggested [Vinther, J.M. et. al. (2011) J. Am. Chem. Soc., 271–278]. Our results show that for this to happen, these residues have to also directly interact with a region of the protein that is becoming more rigid as the temperature increases.
Proteins are dynamic, sometimes adopting a large variety of different conformations, and often changing their interaction partners while exerting their functions. How these dynamics affect function is not totally clear yet, but it is evident that conformational transitions play a key role in events such as allosteric regulation, signal transduction and enzymatic activity. Periplasmic binding proteins (PBPs) undergo large conformational transitions. They are the first component in the membrane import machinery classified as the ABC transporters. PBP role is to recruit the appropriate substrates in the periplasm and bring them to the membrane interface where the ABC transporter resides to initiate the translocation across the cellular membrane. PBPs structure comprises two similar globular domains linked by flexible hinges. Generally, when they are free they assume an open conformation with extended linkers. When binding the substrate the linker bends, bringing the two domains in close contact, thus adopting the closed conformation. In this closed form the active site is embedded in the protein at the interface of the two domains. There are a few exceptions, where the closed-free and open-bound conformations of some PBPs have been observed(1–4). Among PBPs, Glutamine binding protein (GlnBP) is a monomeric 226 residue protein responsible for delivering L-glutamine from the periplasm to the corresponding ABC transporter with very high affinity (Kd 0.1 μM) and selectivity (i.e. it binds only one substrate in contrast to other PBPs)(5). The structure is composed by one large domain (residues 1–84 and 186–226) and one small domain (90–180) that share a common global α/β fold with internal core β-sheets surrounded by αhelices. A flexible hinge, made by two β-strands responsible for a major conformational transition upon binding, separates these two domains. In absence of ligand these two β-strands are extended and when the glutamine is bound these β-strands bend bringing the two domains in direct contact with each other (6). In the closed conformation a cleft is formed by the interface of the two domains where the ligand binding site is located(7).
NMR spin relaxation experiments can provide motional parameters that are helpful in understanding the pico- to nanosecond time scale motions of bond vectors that can report on local structural rearrangements in macromolecules. Conventional backbone 15N T1, T2 and NOE experiments are most commonly analyzed in terms of the Model-Free approach(8–12) with the aim of extracting diffusion parameters useful to describe protein dynamics. Among those, generalized order parameter S2 represents the amplitude of bond vector fluctuations and is related to the thermodynamic properties of the biomolecule. In fact, S2 has been related to conformational entropy(13–16).
Conformational entropy represents the distribution of conformational states of a protein. The main application of measuring conformational entropy changes (i.e. changes in order parameter) between different states of a protein (e.g. free and ligand bound) is to understand binding affinity, and catalysis or cooperativity(17, 18). In addition, one can also measure these changes as a function of temperature, which is equivalent to heat capacity changes (ΔCp). The heat capacity change governs the energy landscape of conformational states with respect to temperature variations, thus defining the stability of the proteins and their complexes at various temperatures(19). To date only a few attempts have been made to analyze the detailed temperature dependence of macromolecule dynamics. Previous studies have exploited both backbone and side chain fast dynamics to explore possible correlated backbone motion in ubiquitin(20), inter-domain behavior in calmodulin(21), folding/unfolding processes of proteins(22) and RNA(23), protein-ligand interactions(24) and thermostability of proteins(25). Even smaller number of studies have analyzed order parameter temperature dependence to specifically unveil heat capacity changes related to protein stability(26, 27) and ligand association(28, 29). The general lesson depicted by these studies is that depending on the protein or protein-complex under study, completely opposite conformational heat capacity contributions can be obtained with the apparent lack of general rules other than internal redistribution of energy.
Here we analyze the residue specific temperature dependence of the pico- to nanosecond dynamics of the backbone amide groups of GlnBP in both the ligand free and bound forms. The relaxation data has been used directly to calculate the residue specific order parameter whose temperature dependence analysis shed light on the thermodynamics of ligand binding, which support the so-called “induced fit model”. According to this model, a protein does not adopt a bound conformation until the ligand triggers the transition from unliganded to liganded conformations. This mechanism was shown to be accompanied by a decrease in conformational entropy due to the active site becoming more rigid and an increase in binding enthalpy due to new bond formations(16). Also, in contrast to what has been proposed in the past(25) our study showed that the presence of a salt bridge is not a satisfactory condition for increased rigidity of the corresponding backbones as a function of increasing temperature.
Wild-type GlnBP, both in the glutamine free and the bound forms, was generated using the pJ133 plasmid(30) as a template. Protein expression and purification were performed as described elsewhere(31). In brief, uniformly 15N-enriched proteins were over expressed in E. coli and cells were allowed to grow in minimal medium containing 15NH4Cl as the sole nitrogen source. The proteins were then extracted by chloroform shock(32) and purified with a two–step purification process, via anion-exchange (DEAE) and size exclusion (Superdex-75) columns (GE Healthcare), respectively(30). In order to get the free form, after the DEAE column, the sample was denatured with 6M GndHCl and size exclusion chromatography run under denaturing conditions. Isolated GlnBP free form was then refolded by extensive dialysis(33). Proteins were subsequently lyophilized and re-dissolved overnight in NMR buffer to a final sample concentration of 2 and 1 mM for the bound and the free form, respectively; when present L-Gln was in 3-fold molar excess. Bound form was dissolved in 20 mM potassium phosphate, 2 mM EDTA at pH 7.2 while free form solubilized in 20 mM potassium phosphate, 2 mM EDTA and 5 mM CHAPS at pH 7.2.
Conventional 15N longitudinal and transverse relaxation times(12) were measured for both bound and free GlnBP at eight different temperatures, 15, 18, 21, 25, 30, 33, 37 and 40 °C. The T1 delays were 8, 160, 520, 720, 960, 1120, 1440 and 1600 ms while the T1ρ delays were 3.28, 11.92, 21.52, 31.12, 40.72, 50.32, 83.92 and 117.52 ms. The T1ρ spin lock frequency was 2.5 kHz(34), and spectra were collected in an interleaved manner to minimize the effects of systematic errors. The relaxation rates were derived from the exponential decay of the peak heights using a home-built program. Finally, the R2 rates were calculated on the basis of R1ρ and R1 rates, applied spin lock strength, and the angular frequency offset of the individual 15N spin. Experiments were repeated in duplicate for GlnBP bound form for estimating the random errors.
R1 and R2 relaxation rates of an amide 15N are expressed in terms of linear combination of spectral densities at different frequencies and it is easy to demonstrate that
where d2 = 0.1 γ2A γ2x h2/(4π2) 1/r3AX and c2 = (2/15) γ2x H20 (σ|| − σ) are constants related to dipolar interaction and chemical shift anisotropy, respectively. In detail, A = 1H, X = 15N, γi is the gyromagnetic ratio of a specific nucleus, h is Plank’s constant, rAX is the internuclear 1H-15N distance, H0 is the magnetic field strength and σ|| and σ are the parallel and perpendicular components of the axially symmetric 15N chemical shift tensor. The spectral density is related to the order parameter by the relation
where the coefficients Ai are related to the cosine directions of the bond vectors in the diffusion tensor frame of the molecule, τi are the relative time constants and τ = (τc τe)/(τc + τe)(35). Both the directions and time constants are detailed elsewhere(36).
from which S2 can be easily calculated.
Errors in T1ρ and T1 for the bound form were estimated from the root mean square difference between the two sets of experiments. The error on the measurements of the free form was estimated by comparing the signal-to-noise ratio (S/N) of T1ρ and T1 spectra for each temperature to those of the bound form spectra. The relative ratios of S/N in respective experiments were used to normalize the experimental errors. The errors on S2 were then assessed by error propagation.
In this study a total of 208 and 178 residues were analyzed for GlnBP bound and free forms, respectively. All residues whose resonances overlap, even at a single temperature, were eliminated from the analysis. Diffusion tensors calculation was performed by considering only the residues located in well-defined secondary structure motifs.
GlnBP overall correlation time (τc) (shown in Figure S1), calculated by fitting the relaxation data to an axially symmetric diffusion tensor, linearly decays as a function of η/T (where η is viscosity and T is temperature in Kelvin) for both the ligand free and bound forms following the Stokes-Einstein equation(37). Furthermore, polar angles describing the orientation of the principal diffusion axis within the molecular PDB frame and the diffusion anisotropy do not change (Table S1) over the temperature range in this study. Constant anisotropy and orientation of the diffusion tensor eliminates their possible contribution to the temperature dependence of the relaxation rates ensuring the differences in S2R are solely due to temperature dependent effects on backbone fluctuations.
Order parameters calculated using equation 3 (S2R) was compared with the one obtained with the Lipari-Szabo formalism (S2)(8). The 2R2-R1 approach was first introduced by Habazettl et al. (38, 39) as a faster and convenient way to get order parameters values from R1 and R2 data without resorting to the time consuming and low sensitivity NOE measurements. This approximation was shown to introduce minimal errors in the order parameter values compared to synthetic data. In the present work we quantitatively assessed the maximum error introduced by our approximations (see Materials and Methods). For instance we would expect a maximum of 1% error introduced in our calculated S2R if we assume J(ωH)≈0 and a maximum of 1.4% error for assuming τ′ ≈ 0 within the limit of τe < 200 ps, thus confirming previous qualitative results(38, 39). However, differently from the previous 2R2-R1 approach, we introduced in our calculation the contribution coming from diffusion anisotropy through the directional cosines of the NH bond vectors to S2R. This implementation lead to a reduction in the discrepancies between fitted and calculated order parameters (see Figure S2).
We found an excellent correlation between order parameters that were calculated using equation 3 or extensive fitting of our data to the Model-Free approach as illustrated in Figure 1. For some residues, however, S2R can still be overestimated and in some cases exceeding the non-physical value of 1. The 2R2-R1 values which are higher than 1 have been previously observed and related to conformational exchange(39). In this case the difference between fitted S2 and S2R is not constant but varies with the temperature, meaning that S2R variations do not solely reflect differences in amplitudes of motion, but conformational exchange as well. In this respect, all of the residues with S2R>1 were excluded from the analysis. For most of the residues under study, although in some cases the absolute values of fitted S2 and S2R differ slightly (see Figure 1), the trend of their temperature dependence is the same. This ensures us that the temperature dependence analysis reflects only differences in order parameters and it is unaffected by other factors such as chemical exchange.
Evaluating the differences between the order parameter of free and bound form of GlnBP at single temperatures gives us insights into the thermodynamics of ligand binding. As expected, the average S2R for the whole protein diminishes with increasing temperature (see Figure 2) for both the free and the bound form of GlnBP indicating that the protein becomes globally more flexible as the temperature rises. Also, by comparing the average order parameters of the two GlnBP forms (Figure 2) it is clear that glutamine binding leads to a reduction in flexibility in the GlnBP backbone at all the temperatures under study with an increase in order parameter ranging between 1.8% at 15°C and 3.6% at 40°C. It is interesting to note that most of the residues that in free GlnBP are more flexible, compared to the bound form, are located on the surface of the protein (see Figure 3A). As expected, residues found at the interface of the two domains in the closed conformation (i.e next to the active site) are more flexible in the free form in virtue of their higher surface exposure following the opening of the binding cleft (Figure 3B). This behavior corresponds to an unfavorable decrease in conformational entropy upon ligand binding. To determine the thermodynamic basis for the observed behavior we measured the free energy (ΔG=−9.5 kcal/mol) of binding and its enthalpic (ΔH=−15.9 kcal/mol) and entropic (TΔS=−6.4 kcal/mol) contributions by ITC (see Supplementary Methods). Ligand binding is characterized by a large favorable enthalpy change that overcomes an unfavorable entropy variation (see Figure S3).
Although solvent exposed in both free and bound GlnBP, the hinge region (Y85-G89 and E181-Y185) shows a decrease in flexibility in the bound form (see Figure 3C). In fact, in the bound GlnBP the hinge region is stabilized in a bent conformation due to the fact that the two domains are in close contact with each other and form favorable electrostatic interactions. When the ligand is released there are no stabilizing forces acting on the hinge such that it displays increased degrees of freedom and enhanced flexibility. What was not expected was the behavior of some regions that are not directly related with ligand binding (e.g. active site residues and hinges). Remarkably, most of these areas are located in the small domain for which the rms between free and bound form backbones is 1.29 Å as opposite to the large domain where rms is only 0.56 Å. For these regions the increased flexibility corresponds to a change in local structure.
While the analysis of differences in S2R between free and bound GlnBP at a single temperature provides insights into the energetics of ligand binding, the analysis of S2R in a temperature dependent fashion can yield additional and important information on stability of the individual conformations of the protein. In this respect, we conducted a residue specific analysis of the S2R temperature dependence of both the free and the bound forms of GlnBP. Three different classes of residues could be identified based on their temperature dependence pattern. In the first class are residues whose order parameters increased when the temperature increased, the second class contains residues whose order parameters did not change, and the last class has residues with order parameters that decreased with increasing temperature.
In detail, taking S2R value at 15°C as a reference, the percentage differences of order parameters were calculated for the remaining temperatures for each residue. Residues displaying at least three consecutive temperature points over the maximum percentage error (1.2 % for the bound form and 4.2 % for the free form), were considered to behave as “positive” (S2R increasing with the temperature, i.e. decreasing fluctuations), those displaying at least three consecutive temperature points below the maximum percentage error were considered “negative” (decreasing S2R with temperature, i.e. increasing their flexibility) while the residues with at least three temperature points within the maximum percentage error were classified as “constant”. In Figure 4 representative residues from each class are shown while the classification list is provided in the Supporting Information. The temperature dependence behavior of these residues is not related to particular amino acid properties such as hydrophobicity, polarity or charge. On the other hand, the location of the residues belonging to a specific class is more interesting.
Binding site residues (D10, A67, G68, T70, R75, K115, G119 H156 and D157) interact with the glutamine ligand by means of hydrogen bonds (D10, R75 and D157) and salt bridges (D10, R75 and D157) and they form a ligand binding site, which is buried and shared between the two domains. For the bound form we were able to analyze seven out of nine of these residues, with the exception of D10 and K115 due to exchange and spectral overlap, respectively. As shown in Figure 5, T70, G119 and D157 become more rigid as the temperature increases, that is they display increasing S2R that is related to a decrease in conformational entropy. According to Figure 5, G68, R75 and H156 are classified as “constant” residues, which means that temperature variations do not affect their interaction to glutamine ligand and their contribution to complex stability is constant. Residue A67 is the only active site residue showing a negative dependence upon temperature increase.
For the GlnBP free form, however, only two residues from the active site (R75 and D157) were available for analysis due to spectral overlap, making a meaningful comparison impossible between the free and the bound form. The only thing that can be said is that in virtue of its increased solvent exposure D157 backbone in free GlnBP is in absolute value dramatically more flexible (ΔS2R ≈ 0.4) than in the bound form. At the same time however, it displays the same positive order parameter temperature dependence. Also R75 becomes more flexible than in the bound form but its temperature dependence turns from being “constant” in bound GlnBP to “positive” in free GlnBP.
In the GlnBP bound form “positive” residues (i.e. decreasing their backbone flexibility as the temperature increases) are located at different positions depending on the domain that they belong to. In the large domain (residues L5-G84 and G186-K226), they form a buried hydrophobic center that extends from the glutamine binding site to the bottom of the domain (see Figure 6A and 6B). This hydrophobic core mainly includes phenylalanines and tryptophanes that network with each other through parallel and T-stacking interaction. These hydrophobic “positive” residues are then protected from the protein surface by a salt bridge between D30 and K219 that is solvent exposed and also shows a “positive” behavior (see Figure 6B). As pointed out above, general chemical properties of individual residues do not determine their temperature dependent behavior. In fact, hydrophobic residues spread through out the protein are characterized by a different response. In particular, here only those hydrophobic patches rich in residues capable of forming stacking interactions (i.e. phenylalanine and tryptophan) decrease their flexibility at increasing temperature (see Figure 6D and and5E)5E) presumably due to their enhanced ring hydrophobic interactions. Similarly, a possible explanation for the salt bridge-involving residues D30-K219 becoming less flexible relies on the interaction with the above-mentioned hydrophobic core. After careful observation it is clear that not only the backbone of D30 and K219 are in direct contact with the aromatic hydrophobic patch, but also their side chains interact with it (Figure 7).
In the opposite small domain (residues L90–L180) “positive” class mainly comprises charged or polar residues located on the surface of the protein (see Figure 6C). Globally, the location of the regions that become less flexible in going to higher temperatures is very different when comparing the two domains (see Figure 8A and 8B).
In GlnBP free form the distribution of the residue types among the domains is not as definite as for the bound form, being evenly dislocated through out the protein (see Figure 8B). The most interesting feature of the free form is related to the hinge region. As expected, the hinge region S2R value is in average more flexible in free GlnBP than in the bound form. This means that in the open conformation of GlnBP the hinge backbones exhibit wider fluctuations in comparison to the same region found in the closed form of the protein where the two hinges are bent due to domain closure. Their temperature dependences, however, are the exact opposite. When the ligand is bound, these residues behave as “negative” (i.e. become more flexible at increasing temperature), while in the free form they become “positive” (i.e. less flexible).
The overall trend of order parameters for GlnBP in both the presence and absence of a ligand is to increase the fluctuations of most of the backbones at increasing temperature thus globally enhancing the conformational sampling of the protein. “Negative” residues (i.e. S2R decreases and flexibility increases at increasing temperature) represent most of the protein backbones being 60% and 50% of the total analyzed bond vectors for GlnBP bound and free form, respectively (see Figure 8C and 8D). As expected, this category of residues bears a very high number of overlaps between the two forms of GlnBP indicating that the major thermodynamic differences between open and closed structure mainly rely on “positive” and “constant” residues. In both free and bound forms, residues with negative dependence are evenly distributed in the small and the large domain, on the surface and the interior regardless on the type of amino acid. However, it is interesting that most of the salt bridges present in both structures belong to this group. It has been proposed in the past that salt bridges strengthen at increasing temperatures due to the temperature dependence of their pKa and that residues involved in salt bridges reduce their flexibility as the temperature rises(25). Surprisingly, here we observe the opposite trend with the majority of the salt bridge-involved residues whose backbones becoming more flexible at higher temperatures (see Figure 7). A possible explanation for this phenomenon is the fact that the strengthening of such ionic interaction at the level of the side chain does not account for the restriction of motion of the corresponding backbone, but must be accompanied by other factors such as for example an increase in the hydrophobic forces in the neighboring areas.
As previously described, residues are classified as “constant” when their S2R are insensitive to temperature variations, meaning that their mobility remains unperturbed over the range of temperatures under study. Again, it is not possible to group this class of residues based on the chemical properties of the corresponding side chains but it is quite interesting to pay attention to their positions in the structure. Remarkably, “constant residues” are in some cases located between “positive” and “negative” residues such that they are part of “gradient” of fluctuations. Considering that “positive” and “negative” residues respectively correspond to a decrease and an increase in conformational entropy, “constant residues” balance their counteracting forces by smoothing the transition from rigid to flexible regions or segments in the protein. Whether this is the normal behavior common to all the proteins is not clear. There is also no precedence to suggest that this should be a general characteristic and therefore should be expected.
Here we present the results of the temperature dependence study of GlnBP dynamics by NMR. Very often protein dynamics have been studied in the past in terms of Model-Free approach(8, 9), which is a powerful strategy for extracting dynamic parameters such as global and internal correlation times and order parameter. Also the Extended Model Free approach proved to be helpful in separating the diverse contributions to backbone dynamics coming from coupled fast and slow motions(11). However, due to relatively long experimental time of NOE measurements and their characteristic larger errors compared to T1 and T1ρ, a more practical approach to get residue specific S2 values would be preferable. In this respect, to calculate the order parameter we make use of the 2R2-R1 approximation as a simplified method that takes into account only the high quality raw conventional relaxation data without any fitting to analytical functions. Comparison between S2R of free and bound GlnBP at single temperatures showed a global decrease in flexibility of backbones upon ligand binding. In particular, motional restriction was mainly found at binding interface and in the hinge region. Similarly, temperature dependence analysis of S2R showed that in GlnBP bound form active site residues experience a decrease in their conformational entropy at increasing temperature indicating an entropically unfavorable interaction between protein and ligand. As confirmed by ITC measurements, this overall reduced flexibility upon ligand binding is a classical example of enthalpy-entropy compensation deriving from unfavorable decrease in entropy (loss of conformational degrees of freedom) and favorable increase in enthalpy (formation of new interactions). As it was also previously shown in protein-ligand interaction studies at a single temperature(40–42), this finding well correlates to the induced-fit mechanism(43). Importantly, to our knowledge this is the first time that experimental results clearly support a specific model for interaction for PBPs.
What is more striking is the effect that ligand interaction can have on the stability of the core of the GlnBP large domain. The packing of F13 and F50 around the glutamine ligand is translated through a chain of hydrophobic interactions resulting in the overall increased in stability of the large domain as a function of increasing temperature. Although it has been predicted and demonstrated in simple model compounds that hydrophobic interactions increase their strength at higher temperatures(44, 45), to our knowledge only Vinther et al.(25) were able to experimentally observe positive temperature effects on hydrophobic patches by NMR and relate the phenomenon to a functional role.
This is a good example where the dynamics or stability of a protein can be a vehicle to transfer information from a ligand binding site to other parts of a protein. It is very suggestive that the link between ligand binding and the change in large domain dynamics will play a role in its recognition of the trans-membrane components of the ABC transporter.
In addition, we were able to get insights into the conformational equilibrium of GlnBP free form by observing the behavior of the flexible hinges. GlnBP binds its substrate in a very selective fashion, being able to accommodate only the glutamine ligand in its binding site. Indeed, we show that within the experimental conditions in this study diffusion anisotropy of the GlnBP free form does not change and the linker region becomes more rigid. These findings support the idea that the two domains will become less likely to get close to each other even at increasing temperature (i.e. do not adopt the closed conformation) when the ligand is absent. This result well correlates with a previous study made at 41°C where the authors used a paramagnetic relaxation enhancement (PRE) approach to investigate the existence of a ligand free-closed conformation of GlnBP(46). At least in the PRE time scale, Bermejo et al. did not have any evidence of such a structure in solution. These considerations altogether also support the induced fit mechanism. Previous structural evidence for the induced fit model comes from crystallographic studies, which suggest that interaction of glutamine ligand with the hinge region via indirect hydrogen bond triggers the interdomain closure and formation of a stable close conformation(7).
Last we show that temperature alone is not a sufficient condition for increased rigidity of residues involved in forming a salt bridge. In general, the backbones of all residues involved in the salt bridges analyzed in this study become more flexible at increasing temperature. With one exception, the salt bridge (D30-K219) that is in direct contact with a core that becomes less flexible because of its increased hydrophobic interactions. As it was proposed previously(25), side chain ion pairs forming salt bridge interactions will stabilize as temperature increases, but for the backbones of the corresponding amino acids to become less flexible at higher temperature other forces must play a role.
We would like to thank Dr. Junhe Ma and Dr. Marie-Paule Strub for preparing the plasmid and purification of the GlnBP. We acknowledge Dr. Motoshi Suzuki for assistance in resonance assignment of the protein and Dr. Kang Chen for helping with the NMR experiments. We would like to acknowledge the professional skills and advice of Dr. Grzegorz Piszczek (Biophysics Core Facility, National Heart, Lung and Blood Institute, National Institutes of Health) regarding ITC measurements.
†This work was supported by the Intramural Research Program of the NIH, NHLBI.
Supporting Information Available: A figure of correlation time dependence on temperature, a figure to compare order parameters determined by various methods for two representative residues, and a figure of ITC traces and binding isotherm. A table of diffusion parameters for GlnBP at different temperatures, as well as a table listing order parameters at 15°C for the free and bound forms of GlnBP and a table listing percent changes in order parameters at different temperatures. This material is available free of charge via the Internet at http://pubs.acs.org.