Combination of molecular modeling, site-directed mutagenesis, and SAR studies to delineate the binding site of pyridopyrimidine antagonists on the human CCK1 receptor

A rational combination of site-directed mutagenesis studies, structure-activity relationships, and dynamic-based docking of pyridopyrimidine-derived CCK1R antagonists into a refined three-dimensional model of the CCK1R allowed us to identify the receptor residues and the ligand functional groups implicated in the molecular recognition process. Our results provided unambiguous evidence that the binding site of these antagonists is overlapping that of the C-terminal tetrapeptide of CCK. In particular, Asn333 and Arg336 residues of the CCK1R are essential for high-affinity binding of these ligands. Moreover, the 2-aryl group in the pyridopyrimidine derivatives shares the same binding pocket as the C-terminal Phe side chain of CCK. Our [pyridopyrimidine.CCK1R] complex model is consistent with previous suggestions concerning the molecular basis that governs functional activity and provides useful considerations about the high CCK1 versus CCK2 selectivity of our derivatives and could contribute to fine-tune the rational design of new molecules with optimized properties.     

Molecular determinants of ligand binding modes in the histamine H(4) receptor: linking ligand-based three-dimensional quantitative structure-activity relationship (3D-QSAR) models to in silico guided receptor mutagenesis studies

The histamine H(4) receptor (H(4)R) is a G protein-coupled receptor (GPCR) that plays an important role in inflammation. Similar to the homologous histamine H(3) receptor (H(3)R), two acidic residues in the H(4)R binding pocket, D(3.32) and E(5.46), act as essential hydrogen bond acceptors of positively ionizable hydrogen bond donors in H(4)R ligands. Given the symmetric distribution of these complementary pharmacophore features in H(4)R and its ligands, different alternative ligand binding mode hypotheses have been proposed. The current study focuses on the elucidation of the molecular determinants of H(4)R-ligand binding modes by combining (3D) quantitative structure-activity relationship (QSAR), protein homology modeling, molecular dynamics simulations, and site-directed mutagenesis studies. We have designed and synthesized a series of clobenpropit (N-(4-chlorobenzyl)-S-[3-(4(5)-imidazolyl)propyl]isothiourea) derivatives to investigate H(4)R-ligand interactions and ligand binding orientations. Interestingly, our studies indicate that clobenpropit (2) itself can bind to H(4)R in two distinct binding modes, while the addition of a cyclohexyl group to the clobenpropit isothiourea moiety allows VUF5228 (5) to adopt only one specific binding mode in the H(4)R binding pocket. Our ligand-steered, experimentally supported protein modeling method gives new insights into ligand recognition by H(4)R and can be used as a general approach to elucidate the structure of protein-ligand complexes.     

Successful virtual screening for a submicromolar antagonist of the neurokinin-1 receptor based on a ligand-supported homology model

The neurokinin-1 (NK1) receptor belongs to the family of G-protein-coupled receptors (GPCRs), which represents one of the most relevant target families in small-molecule drug design. In this paper, we describe a homology modeling of the NK1 receptor based on the high-resolution X-ray structure of rhodopsin and the successful virtual screening based on this protein model. The NK1 receptor model has been generated using our new MOBILE (modeling binding sites including ligand information explicitly) approach. Starting with preliminary homology models, it generates improved models of the protein binding pocket together with bound ligands. Ligand information is used as an integral part in the homology modeling process. For the construction of the NK1 receptor, antagonist CP-96345 was used to restrain the modeling. The quality of the obtained model was validated by probing its ability to accommodate additional known NK1 antagonists from structurally diverse classes. On the basis of the generated model and on the analysis of known NK1 antagonists, a pharmacophore model was deduced, which subsequently guided the 2D and 3D database search with UNITY. As a following step, the remaining hits were docked into the modeled binding pocket of the NK1 receptor. Finally, seven compounds were selected for biochemical testing, from which one showed affinity in the submicromolar range. Our results suggest that ligand-supported homology models of GPCRs may be used as effective platforms for structure-based drug design.     

Structural determinants for high-affinity zolpidem binding to GABA-A receptors

The imidazopyridine zolpidem (Ambien) is one of the most commonly prescribed sleep aids in the United States (Rush, 1998). Similar to classic benzodiazepines (BZDs), zolpidem binds at the extracellular N-terminal alpha/gamma subunit interface of the GABA-A receptor (GABAR). However, zolpidem differs significantly from classic BZDs in chemical structure and neuropharmacological properties. Thus, classic BZDs and zolpidem are likely to have different requirements for high-affinity binding to GABARs. To date, three residues--gamma2Met57, gamma2Phe77, and gamma2Met130--have been identified as necessary for high-affinity zolpidem binding (Proc Natl Acad Sci USA 94:8824-8829, 1997; Mol Pharmacol 52:874-881, 1997). In this study, we used radioligand binding techniques, gamma2/alpha1 chimeric subunits (chi), site-directed mutagenesis, and molecular modeling to identify additional gamma2 subunit residues important for high-affinity zolpidem binding. Whereas alpha1beta2chi receptors containing only the first 161 amino-terminal residues of the gamma2 subunit bind the classic BZD flunitrazepam with wild-type affinity, zolpidem affinity is decreased approximately 8-fold. By incrementally restoring gamma2 subunit sequence, we identified a seven-amino acid stretch in the gamma2 subunit loop F region (amino acids 186-192) that is required to confer high-affinity zolpidem binding to GABARs. When mapped to a homology model, these seven amino acids make up part of loop F located at the alpha/gamma interface. Based on in silico zolpidem docking, three residues within loop F, gamma2Glu189, gamma2Thr193, and gamma2Arg194, emerge as being important for stabilizing zolpidem in the BZD binding pocket and probably interact with other loop F residues to maintain the structural integrity of the BZD binding site.     

Virtual screening for aryl hydrocarbon receptor binding prediction

The overall goal of this study has been to validate computational models for predicting aryl hydrocarbon receptor (AhR) binding. Due to the unavailability of the AhR X-ray crystal structure we have decided to use QSARs models for the binding prediction virtual screening. We have built up CoMFA, Volsurf, and HQSAR models using as a training set 84 AhR ligands. Additionally, we have built a hybrid model combining two of the final selected models in order to give a single operational system. The results show that CoMFA, VolSurf, HQSAR, and the hybrid models gives good results (R(2) equal to 0.91, 0.79, 0.85, and 0.82 and q(2) 0.62, 0.58, 0.62, and 0.70, respectively). Since the techniques analyzed show a good correlation and good prediction also for an external test set, particularly the HQSAR and the hybrid model, we can conclude that these models can be used for predicting AhR binding in virtual screening.     

Thermal denaturation: a method to rank slow binding, high-affinity P38alpha MAP kinase inhibitors

It has been reported that the diaryl urea class of p38alpha inhibitors binds to p38 map kinase with both high affinity and slow binding kinetics (Pargellis et al. Nat. Struct. Biol. 2002, 9, 268-272). The slow binding kinetics of this class of inhibitors is believed to be the result of binding to an allosteric pocket adjacent to the p38alpha active site. The use of traditional kinetic and equilibrium methods to measure the binding affinity of this class of compounds has created many challenges for determination of structure-activity relationships (SAR). The thermal denaturation method provides a means of measuring high-affinity interactions. In this paper, the method of thermal denaturation will be described as it has been applied to the diaryl urea class of p38 map kinase inhibitors.     

The 4-anilinoquinazoline class of inhibitors of the erbB family of receptor tyrosine kinases

The erbB family of receptor tyrosine kinase enzymes, and particularly EGFR and HER2/neu, have become important targets for potential anticancer drugs. The substrate protein binding site theoretically is the more attractive intracellular target on these enzymes, possessing lower homology than the ATP site between different receptor kinases. However, a major breakthrough in this field was the discovery that 4-anilinoquinazolines are potent and selective inhibitors, despite binding at the ATP site. The very tight structure-activity relationships shown by these compounds suggested a clearly-defined binding mode, where the quinazoline ring binds in the adenine pocket and the anilino ring binds in an adjacent, unique lipophilic pocket. A unique cysteine (Cys-773) adjacent to the quinazoline binding site has prompted the development of irreversible inhibitors that target this residue. Three 4-anilinoquinazoline analogues (two reversible and one irreversible inhibitor) have been evaluated clinically as anticancer drugs. Data from the most advanced, the reversible inhibitor Iressa, suggest that this class of compounds may be of value in cancer chemotherapy.     

Nonapeptide analogues containing (R)-3-hydroxybutanoate and beta-homoalanine oligomers: synthesis and binding affinity to a class I major histocompatibility complex protein.

Crystal structures of antigenic peptides bound to class I MHC proteins suggest that chemical modifications of the central part of the bound peptide should not alter binding affinity to the MHC restriction protein but could perturb the T-cell response to the parent epitope. In our effort in designing nonpeptidic high-affinity ligands for class I MHC proteins, oligomers of (R)-3-hydroxybutanoate and(or) beta-homoalanine have been substituted for the central part of a HLA-B27-restricted T-cell epitope of viral origin. The affinity of six modified peptides to the B2705 allele was determined by an in vitro stabilization assay. Four out of the six designed analogues presented an affinity similar to that of the parent peptide. Two compounds, sharing the same stereochemistry (R,R,S,S) at the four stereogenic centers of the nonpeptidic spacer, bound to B2705 with a 5-6-fold decreased affinity. Although the chiral spacers do not strongly interact with the protein active site, there are configurations which are not accepted by the MHC binding groove, probably because of improper orientation of some lateral substituents in the bound state and different conformational behavior in the free state. However we demonstrate that beta-amino acids can be incorporated in the sequence of viral T-cell epitopes without impairing MHC binding. The presented structure-activity relationships open the door to the rational design of peptide-based vaccines and of nonnatural T-cell receptor antagonists aimed at blocking peptide-specific T-cell responses in MHC-associated autoimmune diseases.     

Investigating protein-ligand interactions with a mutant FKBP possessing a designed specificity pocket

Using structure-based design and protein mutagenesis we have remodeled the FKBP12 ligand binding site to include a sizable, hydrophobic specificity pocket. This mutant (F36V-FKBP) is capable of binding, with low or subnanomolar affinities, novel synthetic ligands possessing designed substituents that sterically prevent binding to the wild-type protein. Using binding and structural analysis of bumped compounds, we show here that the pocket is highly promiscuous-capable of binding a range of hydrophobic alkyl and aryl moieties with comparable affinity. Ligand affinity therefore appears largely insensitive to the degree of occupancy or quality of packing of the pocket. NMR spectroscopic analysis indicates that similar ligands can adopt radically different binding modes, thus complicating the interpretation of structure-activity relationships.     

Molecular determinants for high-affinity block of human EAG potassium channels by antiarrhythmic agents

Undesired block of human ERG1 potassium channels is the basis for cardiac side effects of many different types of drugs. Therefore, it is important to know exactly why some drugs particularly bind to these channels with high affinity. Upon expression in mammalian cells and Xenopus laevis oocytes, we investigated the inhibition of the closely related hEAG1 and hEAG2 channels by agents that have previously been reported to block hERG1 channels. Clofilium inhibited hEAG1 and hERG1 with the same potency, whereas hEAG2 was about 150-fold less sensitive to this antiarrhythmic agent. The molecular determinants for this difference are residues Ser436 and Val437 in the inner cavity of the pore and Ala453, which is located in S6 (i.e., remote from the inner cavity). A modeling approach that allowed for partial conformational relaxation of hEAG model structures upon ligand docking suggests that high-affinity block of ether à go-go channels is mediated by an anchoring of the clofilium alkane tail between S6 and the pore helices. In qualitative agreement with experiments, the mutations of hEAG1 residues Ser436 and Val437 to the corresponding larger hEAG2 residues (Thr432, Ile433) resulted in reduced sterical fit between the ligand and the binding cavity. The model is further supported by functional assays involving (+)-N-[1'-(6-cyano-1,2,3,4-tetrahydro-2(R)-naphthalenyl)-3,4-dihydro-4(R)-hydroxyspiro(2H-1-benzopyran-2,4'-piperidin)-6-yl]methanesulfonamide monohydrochloride (MK-499), terfenadine, quinidine, and tetrabutylammonium that are differentially affected by mutations in the pore pocket.     

Crystal structure of FMN-dependent nitroreductase from Escherichia coli B: a prodrug-activating enzyme

The FMN-dependent flavoprotein nitroreductase from Escherichia coli B (NTR) is used in cancer chemotherapy to activate a range of prodrugs. The crystal structure of this enzyme has been determined, using molecular replacement methods and refined at 2.06 A resolution. The recombinant 24-kDa enzyme was crystallized in the tetragonal space group P4(1)2(1)2, with unit cell dimensions of a = b = 57.74 A and c = 275.51 A and two molecules in the asymmetric unit. The structure has a final R factor of 20.3% (R(free) = 26.7%), for all data between the resolution ranges of 10-2.06 A, and includes 4453 protein atoms, 230 water molecules, and 2 flavin mononucleotide (FMN) molecules. The functional unit is a homodimer, which forms the asymmetric unit in the crystal structure. The tertiary structures of these two monomers and their subunit interactions are nearly identical. The molecular replacement search model, the crystal structure of the major NAD(P)H:FMN oxidoreductase of Vibrio fisheri (FRase 1), was selected on the basis of its high sequence identity to that of NTR. The final superposition of these two enzymes revealed a very similar overall fold, with variation in the structures focused around surface loops and helices near the FMN cofactor. Helix G is implicated in substrate specificity and is better resolved in the present NTR structure than in the previously reported FRase 1 structure. The FMN binding pocket is also well-resolved, showing the presence of two channels leading into the active site. The amino acid side chains and main chain atoms interacting with the FMN are well-ordered. The structure of the substrate binding pocket has been used to examine substrate specificity and enzyme kinetics for prodrugs used in antibody-directed enzyme prodrug therapy (ADEPT) and gene-directed enzyme prodrug therapy (GDEPT).     

New serotonin 5-HT(1A) receptor agonists with neuroprotective effect against ischemic cell damage

We report the synthesis of new compounds 4-35 based on structural modifications of different moieties of previously described lead UCM-2550. The new nonpiperazine derivatives, representing second-generation agonists, were assessed for binding affinity, selectivity, and functional activity at the 5-HT(1A) receptor (5-HT(1A)R). Computational β(2)-based homology models of the ligand-receptor complexes were used to explain the observed structure-affinity relationships. Selected candidates were also evaluated for their potential in vitro and in vivo neuroprotective properties. Interestingly, compound 26 (2-{6-[(3,4-dihydro-2H-chromen-2-ylmethyl)amino]hexyl}tetrahydro-1H-pyrrolo[1,2-c]imidazole-1,3(2H)-dione) has been characterized as a high-affinity and potent 5-HT(1A)R agonist (K(i) = 5.9 nM, EC(50) = 21.8 nM) and exhibits neuroprotective effect in neurotoxicity assays in primary cell cultures from rat hippocampus and in the MCAO model of focal cerebral ischemia in rats.     

Predicting relative binding free energies of tacrine-huperzine A hybrids as inhibitors of acetylcholinesterase

The binding of the 9-methyl derivative of tacrine-huperzine A hybrid to Torpedo californica acetylcholinesterase (AChE) has been studied by computational methods. Molecular dynamics simulations have been performed for the AChE-drug complex considering two different ionization states of the protein and two different orientations of the drug in the binding pocket, which were chosen from a previous screening procedure. Analysis of structural fluctuations and of the pattern of interactions between drug and enzyme clearly favor one binding mode for the tacrine-huperzine A hydrid, which mixes effectively some of the binding features of tacrine and huperzine A. The differences in inhibitory activity for a series of related derivatives have been successfully predicted by free energy calculations, which reinforces the confidence in the binding mode and its usefulness for molecular modeling studies. The same techniques have been used to make de novo predictions for a new 3-fluoro-9-ethyl derivative, which can be used to verify a posteriori the goodness of the binding mode. Finally, we have also investigated the effect of replacing Phe300 in the Torpedo californica enzyme by Tyr, which is present in the human AChE. The results indicate that the Phe330-->Tyr mutation is expected to have little effect on the binding affinities. Overall, the whole of results supports the validity of the putative binding model to explain the binding of tacrine-huperzine A hybrids to AChE.     

Recognition of privileged structures by G-protein coupled receptors

Privileged structures are ligand substructures that are widely used to generate high-affinity ligands for more than one type of receptor. To explain this, we surmised that there must be some common feature in the target proteins. For a set of class A GPCRs, we found a good correlation between conservation patterns of residues in the ligand binding pocket and the privileged structure fragments in class A GPCR ligands. A major part of interior surface of the common ligand binding pocket of class A receptors, identified in many GPCRs, is lined with variable residues that are responsible for selectivity in ligand recognition, while other regions, typically located deeper into the binding pocket, are more conserved and retain a predominantly hydrophobic and aromatic character. The latter is reflected in the chemical nature of most GPCR privileged structures and is proposed to be the common feature that is recognized by the privileged structures. Further, we find that this subpocket is conserved even in distant orthologs within the class A family. Three pairs of ligands recognizing widely different receptor types were docked into receptor models of their target receptors utilizing available structure- activity relationships and mutagenesis data. For each pair of ligands, the ligand-receptor complexes reveal that the nature of the privileged structure binding pocket is conserved between the two complexes, in support of our hypothesis. Only part of the privileged structures can be accommodated within the conserved subpocket. Some contacts are established between the privileged structure and the nonconserved parts of the binding pocket. This implies that any one particular privileged structure can target only a subset of receptors, those complementary to the full privileged structure. Our hypothesis leads to a valuable novelty in that ligand libraries can be designed without any foreknowledge of the structure of the endogenous ligand, which in turn means that even orphan receptors can in principle now be addressed as potential drug targets.     

Genomic drug discovery for apoptosis regulation using a new computer screening amino acid complement wave method

A novel caspase-3-specific inhibitory peptide and an agonistic peptide that binds to the Fas molecule were discovered using our computer screening strategy called the amino acid complement wave (ACW) method, which is based on the complementarities of interacting amino acids between comprehensive testing peptides and a target protein surface pocket. The precise binding configurations of the designed peptides on the three-dimensional (3D) structure of the target protein and the prediction of binding affinities (DeltaG) are determined by the molecular docking program. A designed novel tetrapeptide inhibitor of caspase-3, Ac-DNLD-CHO, was revealed to have potent and specific inhibitory activity. When a designed Fas ligand mimic peptide (Fas reactive peptide-4, FRAP-4) was multimerized by carboxyl terminal linkages of polylysine branches (MAP), the octamer (FRAP-4)(8)-MAP effectively induced apoptosis of human ovarian cancer cell line NOS4 cells. Thus the ACW method for structure-based design of optimized small peptides can be used to further develop small peptidomimetic and nonpeptidic organic forms into a new generation of effective pharmaceuticals.     

Discovery of high-affinity ligands of sigma1 receptor, ERG2, and emopamil binding protein by pharmacophore modeling and virtual screening

ERG2, emopamil binding protein (EBP), and sigma-1 receptor (sigma(1)) are enzymes of sterol metabolism and an enzyme-related protein, respectively, that share high affinity for various structurally diverse compounds. To discover novel high-affinity ligands, pharmacophore models were built with Catalyst based upon a series of 23 structurally diverse chemicals exhibiting K(i) values from 10 pM to 100 microM for all three proteins. In virtual screening experiments, we retrieved drugs that were previously reported to bind to one or several of these proteins and also tested 11 new hits experimentally, of which three, among them raloxifene, had affinities for sigma(1) or EBP of <60 nM. When used to search a database of 3525 biochemicals of intermediary metabolism, a slightly modified ERG2 pharmacophore model successfully retrieved 10 substrate candidates among the top 28 hits. Our results indicate that inhibitor-based pharmacophore models for sigma(1), ERG2, and EBP can be used to screen drug and metabolite databases for chemically diverse compounds and putative endogenous ligands.     

Molecular modeling studies of the Akt PH domain and its interaction with phosphoinositides

The serine-threonine protein kinase Akt is a direct downstream target of phosphatidylinositol 3-kinase (PI3-K). The PI3-K-generated phospholipids regulate Akt activity via directly binding to the Akt PH domain. The binding of PI3-K-generated phospholipids is critical to the relocalization of Akt to the plasma membrane, which plays an important role in the process of Akt activation. Activation of the PI3-K/Akt signaling pathway promotes cell survival. To elucidate the structural basis of the interaction of PI3-K-generated phospholipids with the Akt PH domain with the objective of carrying out structure-based drug design, we modeled the three-dimensional structure of the Akt PH domain. Comparative modeling-based methods were employed, and the modeled Akt structure was used in turn to construct structural models of Akt in complex with selected PI3-K-generated phospholipids using the computational docking approach. The model of the Akt PH domain consists of seven beta-strands forming two antiparallel beta-sheets capped by a C-terminal alpha-helix. The beta1-beta2, beta3-beta4, and beta6-beta7 loops form a positively charged pocket that can accommodate the PI3-K-generated phospholipids in a complementary fashion through specific hydrogen-bonding interactions. The residues Lys14, Arg25, Tyr38, Arg48, and Arg86 form the bottom of the binding pocket and specifically interact with the 3- and 4-phophate groups of the phospholipids, while residues Thr21 and Arg23 are situated at the wall of the binding pocket and bind to the 1-phosphate group. The predicted binding mode is consistent with known site-directed mutagenesis data, which reveal that mutation of these crucial residues leads to the loss of Akt activity. Moreover, our model can be used to predict the binding affinity of PI3-K-generated phospholipids and rationalize the specificity of the Akt PH domain for PI(3,4)P2, as opposed to other phospholipids such as PI(3)P and PI(3,4,5)P3. Taken together, our modeling studies provide an improved understanding of the molecular interactions present between the Akt PH domain and the PI3-K-generated phospholipids, thereby providing a solid structural basis for the design of novel, high-affinity ligands useful in modulating the activity of Akt.     

Requirements for specific binding of low affinity inhibitor fragments to the SH2 domain of (pp60)Src are identical to those for high affinity binding of full length inhibitors

Results from a novel approach which uses protein crystallography for the screening of a low affinity inhibitor fragment library are analyzed by comparing the X-ray structures with bound fragments to the structures with the corresponding full length inhibitors. The screen for new phospho-tyrosine mimics binding to the SH2 domain of (pp60)src was initiated because of the limited cell penetration of phosphates. Fragments in our library typically had between 6 and 30 atoms and included compounds which had either millimolar activity in a Biacore assay or were suggested by the ab initio design program LUDI but had no measurable affinity. All identified fragments were located in the phospho-tyrosine pocket. The most promising fragments were successfully used to replace the phospho-tyrosine and resulted in novel nonpeptidic high affinity inhibitors. The significant diversity of successful fragments is reflected in the high flexibility of the phospho-tyrosine pocket. Comparison of the X-ray structures shows that the presence of the H-bond acceptors and not their relative position within the pharmacophore are essential for fragment binding and/or high affinity binding of full length inhibitors. The X-ray data show that the fragments are recognized by forming a complex H-bond network within the phospho-tyrosine pocket of SH2. No fragment structure was found in which this H-bond network was incomplete, and any uncompensated H-bond within the H-bond network leads to a significant decrease in the affinity of full length inhibitors. No correlation between affinity and fragment binding was found for these polar fragments and hence affinity-based screening would have overlooked some interesting starting points for inhibitor design. In contrast, we were unable to identify electron density for hydrophobic fragments, confirming that hydrophobic interactions are important for inhibitor affinity but of minor importance for ligand recognition. Our results suggest that a screening approach using protein crystallography is particularly useful to identify universal fragments for the conserved hydrophilic recognition sites found in target families such as SH2 domains, phosphatases, kinases, proteases, and esterases.     

Exploration of a 14-3-3 PPI Pocket by Covalent Fragments as Stabilizers

The systematic discovery of functional fragments binding to the composite interface of protein complexes is a first critical step for the development of orthosteric stabilizers of protein–protein interactions (PPIs). We have previously shown that disulfide trapping successfully yielded covalent stabilizers for the PPI of 14-3-3 with the estrogen receptor ERα. Here we provide an assessment of the composite PPI target pocket and the molecular characteristics of various fragments binding to a specific subpocket. Evaluating structure–activity relationships highlights the basic principles for PPI stabilization by these covalent fragments that engage a relatively large and exposed binding pocket at the protein/peptide interface with a “molecular glue” mode of action.