Auto-FACE: an NMR based binding site mapping program for fast chemical exchange protein-ligand systems

Abstract

Background: Nuclear Magnetic Resonance (NMR) spectroscopy offers a variety of experiments to study protein-ligand interactions at atomic resolution. Among these experiments, 15N Heteronuclear Single Quantum Correlation (HSQC)experiment is simple, less time consuming and highly informative in mapping the binding site of the ligand. The interpretation of 15N HSQC becomes ambiguous when the chemical shift perturbations are caused by non-specific interactions like allosteric changes and local structural rearrangement. Under such cases, detailed chemical exchange analysis based on chemical shift perturbation will assist in locating the binding site accurately.

Methodology/principal findings: We have automated the mapping of binding sites for fast chemical exchange systems using information obtained from 15N HSQC spectra of protein serially titrated with ligand of increasing concentrations. The automated program Auto-FACE (Auto-FAst Chemical Exchange analyzer) determines the parameters, e.g. rate of change of perturbation, binding equilibrium constant and magnitude of chemical shift perturbation to map the binding site residues.Interestingly, the rate of change of perturbation at lower ligand concentration is highly sensitive in differentiating the binding site residues from the non-binding site residues. To validate this program, the interaction between the protein hBcl(XL) and the ligand BH3I-1 was studied. Residues in the hydrophobic BH3 binding groove of hBcl(XL) were easily identified to be crucial for interaction with BH3I-1 from other residues that also exhibited perturbation. The geometrically averaged equilibrium constant (3.0 x 10(4)) calculated for the residues present at the identified binding site is consistent with the values obtained by other techniques like isothermal calorimetry and fluorescence polarization assays (12.8 x 10(4)). Adjacent to the primary site, an additional binding site was identified which had an affinity of 3.8 times weaker than the former one. Further NMR based model fitting for individual residues suggest single site model for residues present at these binding sites and two site model for residues present between these sites. This implies that chemical shift perturbation can represent the local binding event much more accurately than the global binding event.

Conclusion/significance: Detail NMR chemical shift perturbation analysis enabled binding site residues to be distinguished from non-binding site residues for accurate mapping of interaction site in complex fast exchange system between small molecule and protein. The methodology is automated and implemented in a program called "Auto-FACE", which also allowed quantitative information of each interaction site and elucidation of binding mechanism.

Figure 1. Component signals of population and and structural comparison of BH3I-1 and its analogue BH3I-2.

(A) and both contributes to the component peaks at and which are directly correlated with its respective population and . (B) & (C) Structural comparison of BH3I-1 and its analogue BH3I-2.

Figure 2. Simulation of fast, intermediate and slow exchange regimes for two site chemical exchange using Mexico 3.1 .

The offset () are set at 300Hz for site A and B. The relaxation rates are 1Hz each. Assuming forward and reverse rates to be same, the chemical exchange rates are set at 2400 sec, 1200 sec and 100 sec, for fast, intermediate and slow exchange systems, respectively. In all cases, the population of A∶B is fixed at 1∶1. ( : Component of site A, ▪▪▪: Component of site B, – : Sum of both components (A+B)).

Figure 3. Isothermal binding curve for BH3I-1 titrated into .

() : Blank experiment where 1 mM of BH3I-1 was titrated into 20 mM phosphate buffer. (▪) : 1mM of BH3I-1 was titrated into 25 M . In all buffer solutions, concentration of DMSO was adjusted to 2.5%.

Figure 4. Generalized HSQC perturbation observed for all residues.

(A) Overlaid perturbation spectrum for all residues and (B) Selected residues with significant “peak walking” chemical shift perturbation. Reddish-orange contour represents protein alone spectrum and blue contour represents the spectrum of protein with maximum titrated ligand concentration. The overlaid spectra of gradually titrated ligand concentrations are shown in blue, magenta, green, orange, red, grey and pink contours ranging from 0.133 mM to 1.177 mM of BH3I-1. F146, G148 reaches saturation at the protein to ligand ratio of 1∶1, whereas saturation could not be reached for G94.

Figure 5. Comparison of single and double site binding models for different residues.

Comparison of two different models for residues present at the binding site (A–H) and the non-binding site (I–L). : Experimental data, : Single site model, – : Two site sequential model.

Figure 6. ‘3D’ plot to differentiate the binding site residues from bulk residues.

(A) and (B) are plots for and resonances, with no threshold set for slope and magnitude of perturbation. (C) and (D) are plots for and resonances, with threshold set at which corresponds to 0.01 and 0.5 ppm/mM for slope values of and residues and to and ppm for magnitude of perturbation of and residues. For both plots, equilibrium constants falling within 0.15 to 0.7 percentile were used.

Figure 7. Mapping of the unique residues identified from ‘3D’ plot onto the structure of and comparison with J-surface mapping.

(A) Two distinct regions are shown which are colored differently (red, yellow); (B) & (C) are the J-surface mapping of BH3I-1 at lower (P∶L::1∶0.229) and higher (P∶L::1∶0.918) ligand concentrations, respectively. Each red dot represents the possible location of the centroid of the aromatic ring of BH3I-1. The collection of dots suggests that the aromatic ring could be anywhere in that mapped region. The initial map appears diffused covering G94, G196, G148 residues but slowly converges near F143 and F146 as the concentration of ligand increases. J-surface map were calculated using JSURF program considering perturbations ppm. Other parameters like (standard deviation for data spread), (number of random points to fill the sphere) and (an offset in added to radius of sphere) were set at 3, 2000 and 1, respectively. All the figures were made using the software Chimera .

Figure 8. Comparison of the previous and current docked models of BH3I-1 on to .

(A) and (B) compares the published and current docked models of BH3I-1 on , respectively. In the published model, the stoichiometry was constrained to a single site, so the ligand preferred the site in between the two adjacent pockets. The key residues that interact with BH3I-1 within 5 are highlighted in orange. In the current model, two BH3I-1 molecules bind adjacent to each other with distinctive affinities. Site A and B are circled and highlighted in yellow and red color, respectively.