Maria I. Zavodszky

Side-chain flexibility in protein–ligand binding: The minimal rotation hypothesis

The goal of this work is to learn from nature about the magnitudes of side‐chain motions that occur when proteins bind small organic molecules, and model these motions to improve the prediction of protein–ligand complexes. Following analysis of protein side‐chain motions upon ligand binding in 63 complexes, we tested the ability of the docking tool SLIDE to model these motions without being restricted to rotameric transitions or deciding which side chains should be considered as flexible. The model tested is that side‐chain conformational changes involving more atoms or larger rotations are likely to be more costly and less prevalent than small motions due to energy barriers between rotamers and the potential of large motions to cause new steric clashes. Accordingly, SLIDE adjusts the protein and ligand side groups as little as necessary to achieve steric complementarity. We tested the hypothesis that small motions are sufficient to achieve good dockings using 63 ligands and the apo structures of 20 different proteins and compared SLIDE side‐chain rotations to those experimentally observed. None of these proteins undergoes major main‐chain conformational change upon ligand binding, ensuring that side‐chain flexibility modeling is not required to compensate for main‐chain motions. Although more frugal in the number of side‐chain rotations performed, this model substantially mimics the experimentally observed motions. Most side chains do not shift to a new rotamer, and small motions are both necessary and sufficient to predict the correct binding orientation and most protein–ligand interactions for the 20 proteins analyzed.

Modeling correlated main‐chain motions in proteins for flexible molecular recognition

We describe a new method for modeling protein and ligand main‐chain flexibility, and show its ability to model flexible molecular recognition. The goal is to sample the full conformational space, including large‐scale motions that typically cannot be reached in molecular dynamics simulations due to the computational intensity, as well as conformations that have not been observed yet by crystallography or NMR. A secondary goal is to assess the degree of flexibility consistent with protein–ligand recognition. Flexibility analysis of the target protein is performed using the graph‐theoretic algorithm FIRST, which also identifies coupled networks of covalent and noncovalent bonds within the protein. The available conformations of the flexible regions are then explored with ROCK by random‐walk sampling of the rotatable bonds. ROCK explores correlated motions by only sampling dihedral angles that preserve the coupled bond networks in the protein and generates conformers with good stereochemistry, without using a computationally expensive potential function. A representative set of the conformational ensemble generated this way can be used as targets for docking with SLIDE, which handles the flexibility of protein and ligand side‐chains. The realism of this protein main‐chain conformational sampling is assessed by comparison with time‐resolved NMR studies of cyclophilin A motions. ROCK is also effective for modeling the flexibility of large cyclic and polycyclic ligands, as demonstrated for cyclosporin and zearalenol. The use of this combined approach to perform docking with main‐chain flexibility is illustrated for the cyclophilin A–cyclosporin complex and the estrogen receptor in complex with zearalenol, while addressing the question of how much flexibility is allowed without hindering molecular recognition. Proteins 2004.

Distilling the essential features of a protein surface for improving protein-ligand docking, scoring, and virtual screening

For the successful identification and docking of new ligands to a protein target by virtual screening, the essential features of the protein and ligand surfaces must be captured and distilled in an efficient representation. Since the running time for docking increases exponentially with the number of points representing the protein and each ligand candidate, it is important to place these points where the best interactions can be made between the protein and the ligand. This definition of favorable points of interaction can also guide protein structure-based ligand design, which typically focuses on which chemical groups provide the most energetically favorable contacts. In this paper, we present an alternative method of protein template and ligand interaction point design that identifies the most favorable points for making hydrophobic and hydrogen–bond interactions by using a knowledge base. The knowledge-based protein and ligand representations have been incorporated in version 2.0 of SLIDE and resulted in dockings closer to the crystal structure orientations when screening a set of 57 known thrombin and glutathione S–transferase (GST) ligands against the apo structures of these proteins. There was also improved scoring enrichment of the dockings, meaning better differentiation between the chemically diverse known ligands and a ∼15,000-molecule dataset of randomly-chosen small organic molecules. This approach for identifying the most important points of interaction between proteins and their ligands can equally well be used in other docking and design techniques. While much recent effort has focused on improving scoring functions for protein-ligand docking, our results indicate that improving the representation of the chemistry of proteins and their ligands is another avenue that can lead to significant improvements in the identification, docking, and scoring of ligands.

Sampling protein conformations and pathways.

Protein flexibility and rigidity can be analyzed using constraint theory, which views proteins as 3D networks of constraints involving covalent bonds and also including hydrophobic interactions and hydrogen bonds. This article describes an algorithm, ROCK (Rigidity Optimized Conformational Kinetics), which generates new conformations for these complex networks with many interlocked rings while maintaining the constraints. These new conformations are tracked for the flexible regions of a protein, while leaving the rigid regions undisturbed. An application to HIV protease demonstrates how large the flap motion can be. The algorithm is also used to generate conformational pathways between two distinct protein conformations. As an example, directed trajectories between the closed and the occluded conformations of the protein dihydrofolate reductase are determined.

Disulfide bond effects on protein stability: designed variants of Cucurbita maxima trypsin inhibitor-V.

Attempts to increase protein stability by insertion of novel disulfide bonds have not always been successful. According to the two current models, cross‐links enhance stability mainly through denatured state effects. We have investigated the effects of removal and addition of disulfide cross‐links, protein flexibility in the vicinity of a cross‐link, and disulfide loop size on the stability of Cucurbita maxima trypsin inhibitor‐V (CMTI‐V; 7 kD) by differential scanning calorimetry. CMTI‐V offers the advantage of a large, flexible, and solvent‐exposed loop not involved in extensive intra‐molecular interactions. We have uncovered a negative correlation between retention time in hydrophobic column chromatography, a measure of protein hydrophobicity, and melting temperature (Tm), an indicator of native state stabilization, for CMTI‐V and its variants. In conjunction with the complete set of thermodynamic parameters of denaturation, this has led to the following deductions: (1) In the less stable, disulfide‐removed C3S/C48S (ΔΔGd50°C = −4 kcal/mole; ΔTm = −22°C), the native state is destabilized more than the denatured state; this also applies to the less‐stable CMTI‐V* (ΔΔGd50°C = −3 kcal/mole; ΔTm = −11°C), in which the disulfide‐containing loop is opened by specific hydrolysis of the Lys44‐Asp45 peptide bond; (2) In the less stable, disulfide‐inserted E38C/W54C (ΔΔGd50°C = −1 kcal/mole; ΔTm = +2°C), the denatured state is more stabilized than the native state; and (3) In the more stable, disulfide‐engineered V42C/R52C (ΔΔGd50°C = +1 kcal/mole; ΔTm = +17°C), the native state is more stabilized than the denatured state. These results show that a cross‐link stabilizes both native and denatured states, and differential stabilization of the two states causes either loss or gain in protein stability. Removal of hydrogen bonds in the same flexible region of CMTI‐V resulted in less destabilization despite larger changes in the enthalpy and entropy of denaturation. The effect of a cross‐link on the denatured state of CMTI‐V was estimated directly by means of a four‐state thermodynamic cycle consisting of native and denatured states of CMTI‐V and CMTI‐V*. Overall, the results show that an enthalpy‐entropy compensation accompanies disulfide bond effects and protein stabilization is profoundly modulated by altered hydrophobicity of both native and denatured states, altered flexibility near the cross‐link, and residual structure in the denatured state.

Synthetic curcuminoids modulate the arachidonic acid metabolism of human platelet 12-lipoxygenase and reduce sprout formation of human endothelial cells

Platelet 12-lipoxygenase (P-12-LOX) is overexpressed in different types of cancers, including prostate cancer, and the level of expression is correlated with the grade of this cancer. Arachidonic acid is metabolized by 12-LOX to 12(S)-hydroxyeicosatetraenoic acid [12(S)-HETE], and this biologically active metabolite is involved in prostate cancer progression by modulating cell proliferation in multiple cancer-related pathways inducing angiogenesis and metastasis. Thus, inhibition of P-12-LOX can reduce these two processes. Several lipoxygenase inhibitors are known, including plant and mammalian lipoxygenases, but only a few of them are known inhibitors of P-12-LOX. Curcumin is one of these lipoxygenase inhibitors. Using a homology model of the three-dimensional structure of human P-12-LOX, we did computational docking of synthetic curcuminoids (curcumin derivatives) to identify inhibitors superior to curcumin. Docking of the known inhibitors curcumin and NDGA to P-12-LOX was used to optimize the docking protocol for the system in study. Over 75% of the compounds of interest were successfully docked into the active site of P-12-LOX, many of them sharing similar binding modes. Curcuminoids that did not dock into the active site did not inhibit P-12-LOX. From a set of the curcuminoids that were successfully docked and selected for testing, two were found to inhibit human lipoxygenase better than curcumin. False-positive curcuminoids showed high LogP (theoretical) values, indicating poor water solubility, a possible reason for lack of inhibitory activity or/and nonrealistic binding. Additionally, the curcuminoids inhibiting P-12-LOX were tested for their ability to reduce sprout formation of endothelial cells (in vitro model of angiogenesis). We found that only curcuminoids inhibiting human P-12-LOX and the known inhibitor NDGA reduced sprout formation. Only limited inhibition of sprout formation at ∼IC50 concentrations has been seen. At IC50, a substantial amount of 12-HETE can be produced by lipoxygenase, providing a stimulus for angiogenic sprouting of endothelial cells. Increasing the concentration of lipoxygenase inhibitors above IC50, thus decreasing the concentration of 12(S)-HETE produced, greatly reduced sprout formation for all inhibitors tested. This universal event for all tested lipoxygenase inhibitors suggests that the inhibition of sprout formation was most likely due to the inhibition of human P-12-LOX but not other cancer-related pathways. [Mol Cancer Ther 2006;5(5):1371–82]

Synergistic Cooperation between Two ClpB Isoforms in Aggregate Reactivation

Bacterial AAA+ ATPase ClpB cooperates with DnaK during reactivation of aggregated proteins. The ClpB-mediated disaggregation is linked to translocation of polypeptides through the channel in the oligomeric ClpB. Two isoforms of ClpB are produced in vivo: the full-length ClpB95 and ClpB80, which does not contain the substrate-interacting N-terminal domain. The biological role of the truncated isoform ClpB80 is unknown. We found that resolubilization of aggregated proteins in Escherichia coli after heat shock and reactivation of aggregated proteins in vitro and in vivo occurred at higher rates in the presence of ClpB95 with ClpB80 than with ClpB95 or ClpB80 alone. Combined amounts of ClpB95 and ClpB80 bound to aggregated substrates were similar to the amounts of either ClpB95 or ClpB80 bound to the substrates in the absence of another isoform. The ATP hydrolysis rate of ClpB95 with ClpB80, which is linked to the rate of substrate translocation, was not higher than the rates measured for the isolated ClpB95 or ClpB80. We postulate that a reaction step that takes place after substrate binding to ClpB and precedes substrate translocation is rate-limiting during aggregate reactivation, and its efficiency is enhanced in the presence of both ClpB isoforms. Moreover, we found that ClpB95 and ClpB80 form hetero-oligomers, which are similar in size to the homo-oligomers of ClpB95 or ClpB80. Thus, the mechanism of functional cooperation of the two isoforms of ClpB may be linked to their heteroassociation. Our results suggest that the functionality of other AAA+ ATPases may be also optimized by interaction and synergistic cooperation of their isoforms.

Scoring confidence index: statistical evaluation of ligand binding mode predictions

Protein-ligand docking programs can generate a large number of possible binding orientations for each ligand candidate. The challenge is to identify the orientations closest to the native binding mode using a scoring method. Many different scoring functions have been developed for protein-ligand scoring, but their performance on binding mode prediction is often target-dependent. In this study, a statistical approach was employed to provide a confidence measure of scoring performance in finding close to the correct docked ligand orientations. It exploits the fact that the scores provided by an adequately performing scoring function generally improve as the ligand binding modes get closer to the correct native orientation. For such cases, the correlation coefficient of scores versus distances is expected to be highest when the most native-like orientation is used as a reference. This correlation coefficient, called the correlation-based score (CBScore), was used as an indicator of how far the docked pose was from the native orientation. The correlation between the original scores and CBScores as well as the range of CBScores were found to be good measures of scoring performance. They were combined into a single quantity, called the scoring confidence index. High values of the scoring confidence index were indicative of pronounced and relatively smooth binding energy landscapes with easily discernable global minima, resulting in reliable binding mode predictions. Low values of this index reflected rugged energy landscapes making the prediction of the correct binding mode very difficult and often unreliable. The diagnostic ability of the scoring confidence index was tested on a non-redundant set of 50 protein-ligand complexes scored with three commonly employed scoring functions: AffiScore, DrugScore and X-Score. Binding mode predictions were found to be three times more reliable for complexes with scoring confidence indices in the upper half than for cases with values in the lower half of the resulting range of 0–1.6. This new confidence measure of scoring performance is expected to be a valuable tool for virtual screening applications.

Structural basis for the enantiospecificities of R‐ and S‐specific phenoxypropionate/α‐ketoglutarate dioxygenases

(R)‐ and (S)‐dichlorprop/α‐ketoglutarate dioxygenases (RdpA and SdpA) catalyze the oxidative cleavage of 2‐(2,4‐dichlorophenoxy)propanoic acid (dichlorprop) and 2‐(4‐chloro‐2‐methyl‐phenoxy)propanoic acid (mecoprop) to form pyruvate plus the corresponding phenol concurrent with the conversion of α‐ketoglutarate (αKG) to succinate plus CO2. RdpA and SdpA are strictly enantiospecific, converting only the (R) or the (S) enantiomer, respectively. Homology models were generated for both enzymes on the basis of the structure of the related enzyme TauD (PDB code 1OS7). Docking was used to predict the orientation of the appropriate mecoprop enantiomer in each protein, and the predictions were tested by characterizing the activities of site‐directed variants of the enzymes. Mutant proteins that changed at residues predicted to interact with (R)‐ or (S)‐mecoprop exhibited significantly reduced activity, often accompanied by increased Km values, consistent with roles for these residues in substrate binding. Four of the designed SdpA variants were (slightly) active with (R)‐mecoprop. The results of the kinetic investigations are consistent with the identification of key interactions in the structural models and demonstrate that enantiospecificity is coordinated by the interactions of a number of residues in RdpA and SdpA. Most significantly, residues Phe171 in RdpA and Glu69 in SdpA apparently act by hindering the binding of the wrong enantiomer more than the correct one, as judged by the observed decreases in Km when these side chains are replaced by Ala.