Receptor-based design of dihydrofolate reductase inhibitors: comparison of crystallographically determined enzyme binding with enzyme affinity in a series of carboxy-substituted trimethoprim analogues

Chemical labeling of intracellular proteins via affinity conjugation and strain-promoted cycloadditions in live cells

A versatile chemical labeling approach was developed, where intracellular proteins were first incorporated with a bioorthogonal group via affinity conjugation, and subsequently labeled via strain-promoted cycloaddition reactions in live cells.

A bioorthogonal small-molecule-switch system for controlling protein function in live cells

Chemically induced dimerization (CID) has proven to be a powerful tool for modulating protein interactions. However, the traditional dimerizer rapamycin has limitations in certain in vivo applications because of its slow reversibility and its affinity for endogenous proteins. Described herein is a bioorthogonal system for rapidly reversible CID. A novel dimerizer with synthetic ligand of FKBP' (SLF') linked to trimethoprim (TMP). The SLF' moiety binds to the F36V mutant of FK506-binding protein (FKBP) and the TMP moiety binds to E. coli dihydrofolate reductase (eDHFR). SLF'-TMP-induced heterodimerization of FKBP(F36V) and eDHFR with a dissociation constant of 0.12 μM. Addition of TMP alone was sufficient to rapidly disrupt this heterodimerization. Two examples are presented to demonstrate that this system is an invaluable tool, which can be widely used to rapidly and reversibly control protein function in vivo.

Structure-based drug discovery and protein targets in the CNS

Structure-based methods are having an increasing role and impact in drug discovery. The crystal structures of an increasing number of therapeutic targets are becoming available. These structures can transform our understanding of how these proteins perform their biological function and often provide insights into the molecular basis of disease. In addition, the structures can help the discovery process. Methods such as virtual screening and experimental fragment screening can provide starting hit compounds for a discovery project. Crystal structures of compounds bound to the protein can direct or guide the medicinal chemistry optimisation to improve drug-like properties - not only providing ideas on how to improve binding affinity or selectivity, but also showing where the compound can be modified in attempting to modulate physico-chemical properties and biological efficacy. The majority of drug discovery projects against globular protein targets now use these methods at some stage. This review provides a summary of the range of structure-based drug discovery methods that are in use and surveys the suitability of the methods for targets currently identified for CNS drugs. Until recently, structure-based discovery was difficult or unknown for these targets. The recent determination of the structures of a number of GPCR proteins, together with the steady increase in structures for other membrane proteins, is opening up the possibility for these structure-based methods to find increased use in drug discovery for CNS diseases and conditions.

Selective antifolates for chemically labeling proteins in mammalian cells

Antifolate labels: Molecules that bind specifically and with high affinity to proteins can be developed into powerful tools for chemical biology. The interaction between substituted 5‐benzyl pyrimidines and dihydrofolate reductase can be exploited for chemically labeling fusion proteins in mammalian cells.magnified image

Folate-Synthesizing Enzyme System as Target for Development of Inhibitors and Inhibitor Combinations against Candida albicansSynthesis and Biological Activity of New 2,4-Diaminopyrimidines and 4‘-Substituted 4-Aminodiphenyl Sulfones

The paper describes the design, synthesis, and testing of inhibitors of folate-synthesizing enzymes and of whole cell cultures of Candida albicans. The target enzymes used were dihydropteroic acid synthase (SYN) and dihydrofolate reductase (DHFR). Several series of new 2,4-diaminopyrimidines were synthesized and tested as inhibitors of DHFR and compared with their activity against DHFR derived from mycobacteria and Escherichia coli. To test for selectivity, also rat DHFR was used. A series of substituted 4-aminodiphenyl sulfones was tested for inhibitory activity against SYN and the I(50) values compared to those obtained previously against Plasmodium berghei- and E. coli-derived SYN. Surprisingly, QSAR equations show very similar structural dependencies. To find an explanation for the large difference in the I(50) values observed for enzyme inhibition (SYN, DHFR) and for inhibition of cell cultures of Candida, mutant strains with overexpressed efflux pumps and strains in which such pumps are deleted were included in the study and the MICs compared. Efflux pumps were responsible for the low activity of some of the tested derivatives, others showed no increase in activity after pumps were knocked out. In this case it may be speculated that these derivatives are not able to enter the cells.

The effect of relaxation on the epitope mapping by saturation transfer difference NMR

The effect of longitudinal relaxation of ligand protons on saturation transfer difference (STD) was investigated by using a known binding system, dihydrofolate reductase and trimethoprim. The results indicate that T1 relaxation of ligand protons has a severe interference on the epitope map derived from a STD measurement. When the T1s of individual ligand protons are distinctly different, STD experiments may not give an accurate epitope map for the ligand-target interactions. Measuring the relaxation times prior to mapping is strongly advised. A saturation time shorter than T1s is suggested for improving the potential epitope map. Reduction in temperature was seen to enhance the saturation efficiency in small to medium size targets.

Epitope mapping of ligand-receptor interactions by diffusion NMR

A novel method based on diffusion NMR for the epitope mapping of ligand binding is presented. The intermolecular NOE builds up during a long diffusion period and creates a deviation from the linearity. The ligand proton nearest the protein generates the strongest NOE from protein during the diffusion period and has the largest deviation. Therefore, this diffusion artifact can be used to characterize the ligand binding epitope. The concept was investigated using dihydrofolate reductase (DHFR) and its ligand trimethoprim (TMP), and the epitope map of TMP on DHFR generated with this method is in excellent agreement with the structural and dynamic studies by crystallography and NMR, as well as the medicinal chemistry results.

A family competition evolutionary algorithm for automated docking of flexible ligands to proteins

In this paper, we study an evolutionary algorithm for flexible ligand docking. Based on family competition and adaptive rules, the proposed approach consists of global and local strategies by integrating decreasing mutations and self-adaptive mutations. To demonstrate the robustness of the proposed approach, we apply it to the problems of the first international contests on evolutionary optimization. Following the description of function optimization, our approach is applied to a dihydrofolate reductase enzyme with the anti-cancer drug methotrexate and with two analogs of the antibacterial drug trimethoprim. Our numerical results indicate that the proposed approach is robust. The docked lowest energy structures have rms derivations ranging from 0.72 A to 1.98 A with respect to the corresponding crystal structure.

1H/15N HSQC NMR studies of ligand carboxylate group interactions with arginine residues in complexes of brodimoprim analogues and Lactobacillus casei dihydrofolate reductase

1H and 15N NMR studies have been undertaken on complexes of Lactobacillus casei dihydrofolate reductase (DHFR) formed with analogues of the antibacterial drug brodimoprim (2,4-diamino-5-(3', 5'-dimethoxy-4'-bromobenzyl)pyrimidine) in order to monitor interactions between carboxylate groups on the ligands and basic residues in the protein. These analogues had been designed by computer modeling with carboxylated alkyl chains introduced at the 3'-O position in order to improve their binding properties by making additional interactions with basic groups in the protein. Specific interactions between ligand carboxylate groups and the conserved Arg57 residue have been detected in studies of 1H/15N HSQC spectra of complexes of DHFR with both the 4-carboxylate and the 4, 6-dicarboxylate brodimoprim analogues. The spectra from both complexes showed four resolved signals for the four NHeta protons of the guanidino group of Arg57, and this is consistent with hindered rotation in the guanidino group resulting from interactions with the 4-carboxylate group in each analogue. In the spectra of each complex, one of the protons from each of the two NH2 groups and both nitrogens are considerably deshielded compared to the shielding values normally observed for such nuclei. This pattern of deshielding is that expected for a symmetrical end-on interaction of the carboxylate oxygens with the NHeta12 and NHeta22 guanidino protons. The differences in the degree of deshielding between the complexes of the two structurally similar brodimoprim analogues and the methotrexate indicates that the shielding is very sensitive to geometry, most probably to hydrogen bond lengths. The 1H/15N HSQC spectrum of the DHFR complex with the brodimoprim-6-carboxylate analogue does not feature any deshielded Arg NHeta protons and this argues against a similar interaction with the Arg57 in this case. It has not proved possible to determine whether the 6-carboxylate in this analogue is interacting directly with any residue in the protein. 1H/15N HSQC spectra have been fully assigned for the complexes with the three brodimoprim analogues and chemical shift mapping used to explore interactions in the binding site. The 1H signals of the bound ligands for all three brodimoprim analogues have been assigned. Their 1H chemical shifts were found to be fairly similar in the different complexes indicating that the 2, 4-diaminopyrimidine and the benzyl ring are binding in essentially the same binding sites and with the same overall conformation in the different complexes. The rotation rate about the NepsilonCzeta bond in the brodimoprim-4,6-dicarboxylate complex with DHFR has been determined from a zz-HSQC exchange experiment, and its value is quite similar to that observed in the DHFR.methotrexate complex (24 +/- 10 s-1 at 8 degrees C and 50 +/- 10 s-1 at 15 degrees C, respectively). The 1H and 15N chemical shift differences of selected amide and guanidino NH groups, measured between the DHFR complexes, provided further evidence about the interactions involving Arg57 with the 4-carboxylate and 4,6-dicarboxylate brodimoprim analogues.

The solution structure of the complex of Lactobacillus casei dihydrofolate reductase with methotrexate

We have determined the three-dimensional solution structure of the complex of Lactobacillus casei dihydrofolate reductase (18.3 kDa, 162 amino acid residues) formed with the anticancer drug methotrexate using 2531 distance, 361 dihedral angle and 48 hydrogen bond restraints obtained from analysis of multidimensional NMR spectra. Simulated annealing calculations produced a family of 21 structures fully consistent with the constraints. The structure has four alpha-helices and eight beta-strands with two other regions, comprising residues 11 to 14 and 126 to 127, also interacting with each other in a beta-sheet manner. The methotrexate binding site is very well defined and the structure around its glutamate moiety was improved by including restraints reflecting the previously determined specific interactions between the glutamate alpha-carboxylate group with Arg57 and the gamma-carboxylate group with His28. The overall fold of the binary complex in solution is very similar to that observed in the X-ray studies of the ternary complex of L. casei dihydrofolate reductase formed with methotrexate and NADPH (the structures of the binary and ternary complexes have a root-mean-square difference over the backbone atoms of 0.97 A). Thus no major conformational change takes place when NADPH binds to the binary complex. In the binary complex, the loop comprising residues 9 to 23 which forms part of the active site has been shown to be in the "closed" conformation as defined by M. R. Sawaya & J. Kraut, who considered the corresponding loops in crystal structures of complexes of dihydrofolate reductases from several organisms. Thus the absence of the NADPH does not result in the "occluded" form of the loop as seen in crystal studies of some other dihydrofolate reductases in the absence of coenzyme. Some regions of the structure in the binary complex which form interaction sites for NADPH are less well defined than other regions. However, in general terms, the NADPH binding site appears to be essentially pre-formed in the binary complex. This may contribute to the tighter binding of coenzyme in the presence of methotrexate.

NMR detection of arginine-ligand interactions in complexes of Lactobacillus casei dihydrofolate reductase

1H-NMR and 15N-NMR signal assignments have been made for the eight arginine residues in Lactobacillus casei dihydrofolate reductase in its binary complex with methotrexate and in its ternary complex with methotrexate and NADPH. 1H-NMR chemical shifts for the guanidino groups of two of the arginines (Arg57 and Arg43) were sensitive to different modes of binding of the guanidino groups with charged oxygen atoms of the ligands. In the complexes formed with methotrexate, Arg57 showed four non-equivalent NH eta proton signals indicating hindered rotation about the N epsilon-C zeta and C zeta-N eta bonds. The NH eta 12 and NH eta 22 protons showed large downfield shifts, which would be expected for a symmetric end-on interaction of these protons with the charged oxygen atoms of a carboxylate group in methotrexate. These effects were not observed for the complex formed with trimethoprim, which does not contain any carboxylate groups. In the complex formed with NADPH present, Arg43 showed a large downfield chemical shift for its NH epsilon proton and a retardation of its rate of exchange with water. This pattern of deshielding contrasts with that detected for Arg57 and is that expected for a side-on interaction of the guanidino group protons with charged oxygen atoms of the ribose 2'-phosphate group of NADPH.

New antifolate 4,4'-diaminodiphenyl sulfone substituted 2,4-diamino-5-benzylpyrimidines. Proof of their dual mode of action and autosynergism

New 4,4'-diaminodiphenylsulfone substituted 2,4-diamino-5-benzylpyrimidines were synthesized. These compounds are highly active inhibitors of both bacterial dihydrofolate reductase (DHFR) and dihydropteroic acid synthase (SYN). The simultaneous inhibition of both enzymes leads to autosynergism in whole cells in the same way as known for combinations of sulfonamides with trimethoprim. The inhibitory activity is demonstrated in cell-free systems of DHFR and SYN derived from various species (M. lufu, E. coli, C. albicans) and in whole cell systems of the mycobacterial strain M. lufu. The compounds are rare examples for the combination of two mechanisms of action in one molecule.

Dihydrofolate reductase: sequential resonance assignments using 2D and 3D NMR and secondary structure determination in solution

Three-dimensional (3D) heteronuclear NMR techniques have been used to make sequential 1H and 15N resonance assignments for most of the residues of Lactobacillus casei dihydrofolate reductase (DHFR), a monomeric protein of molecular mass 18,300 Da. A uniformly 15N-labeled sample of the protein was prepared and its complex with methotrexate (MTX) studied by 3D 15N/1H nuclear Overhauser-heteronuclear multiple quantum coherence (NOESY-HMQC), Hartmann-Hahn-heteronuclear multiple quantum coherence (HOHAHA-HMQC), and HMQC-NOESY-HMQC experiments. These experiments overcame most of the spectral overlap problems caused by chemical shift degeneracies in 2D spectra and allowed the 1H-1H through-space and through-bond connectivities to be identified unambiguously, leading to the resonance assignments. The novel HMQC-NOESY-HMQC experiment allows NOE cross peaks to be detected between NH protons even when their 1H chemical shifts are degenerate as long as the amide 15N chemical shifts are nondegenerate. The 3D experiments, in combination with conventional 2D NOESY, COSY, and HOHAHA experiments on unlabelled and selectively deuterated DHFR, provide backbone assignments for 146 of the 162 residues and side-chain assignments for 104 residues of the protein. Data from the NOE-based experiments and identification of the slowly exchanging amide protons provide detailed information about the secondary structure of the binary complex of the protein with methotrexate. Sequential NHi-NHi+1 NOEs define four regions with helical structure. Two of these regions, residues 44-49 and 79-89, correspond to within one amino acid to helices C and E in the crystal structure of the DHFR.methotrexate.NADPH complex [Bolin et al. (1982) J. Biol. Chem. 257, 13650-13662], while the NMR-determined helix formed by residues 26-35 is about one turn shorter at the N-terminus than helix B in the crystal structure, which spans residues 23-34. Similarly, the NMR-determined helical region comprising residues 102-110 is somewhat offset from the crystal structure's helix F, which encompasses residues 97-107. Regions of beta-sheet structure were characterized in the binary complex by strong alpha CHi-NHi+1 NOEs and by slowly exchanging amide protons. In addition, several long-range NOEs were identified linking together these stretches to form a beta-sheet. These elements align perfectly with corresponding elements in the crystal structure of the DHFR.methotrexate.NADPH complex, which contains an eight-stranded beta-sheet, indicating that the main body of the beta-sheet is preserved in the binary complex in solution.

Project GROPEHaptic displays for scientific visualization

We began in 1967 a project to develop a haptic+display for 6-D force fields of interacting protein molecules. We approached it in four stages: a 2-D system, a 3-D system tested with a simple task, a 6-D system tested with a simple task, and a full 6-D molecular docking system, our initial goal. This paper summarizes the entire project---the four systems, the evaluation experiments, the results, and our observations. The molecular docking system results are new.Our principal conclusions are:• Haptic display as an augmentation to visual display can improve perception and understanding both of force fields and of world models populated with impenetrable objects.• Whereas man-machine systems can outperform computer-only systems by orders of magnitude on some problems, haptic-augmented interactive systems seem to give about a two-fold performance improvement over purely graphical interactive systems. Better technology may give somewhat more, but a ten-fold improvement does not seem to be in the cards.• Chemists using GROPE-III can readily reproduce the true docking positions for drugs whose docking is known (but not to them) and can find very good docks for drugs whose true docks are unknown. The present tool promises to yield new chemistry research results; it is being actively used by research chemists.• The most valuable result from using GROPE-III for drug docking is probably the radically improved situation awareness that serious users report. Chemists say they have a new understanding of the details of the receptor site and its force fields, and of why a particular drug docks well or poorly.• We see various scientific/education applications for haptic displays but believe entertainment, not scientific visualization, will drive and pace the technology.• The hardware-software system technology we have used is barely adequate, and our experience sets priorities for future development.• Some unexpected perceptual phenomena were observed. All of these worked for us, not against us.

NMR studies of interactions of ligands with dihydrofolate reductase

NMR spectroscopy is a useful technique for studying interactions, conformations and dynamic processes within ligand-protein complexes. Several examples of the application of the method to studies of complexes of anti-folate drugs with their target enzyme, dihydrofolate reductase, are discussed.

Interactions between inhibitors of dihydrofolate reductase

The binding of substrates and inhibitors to dihydrofolate reductase was studied by steady-state kinetics and high-field 1H-n.m.r. spectroscopy. A series of 5-substituted 2,4-diaminopyrimidines were examined and were found to be 'tightly binding' inhibitors of the enzyme (Ki less than 10(-9) M). Studies on the binding of 4-substituted benzenesulphonamides and benzenesulphonic acids also established the existence of a 'sulphonamide-binding site' on the enzyme. Subsequent n.m.r. experiments showed that there are two binding sites for the sulphonamides on the enzyme, one of which overlaps the coenzyme (NADPH) adenine-ring-binding site. An examination of the pH-dependence of the binding of sulphonamides to the enzyme indicated the influence of an ionizable group on the enzyme that was not directly involved in the sulphonamide binding. The change in pKa value from 6.7 to 7.2 observed on sulphonamide binding suggests the involvement of a histidine residue, which could be histidine-28.

The geometries of interacting arginine-carboxyls in proteins

The geometries are reported for interacting arginine-carboxyl pairs obtained from 37 high resolution protein structures solved to a resolution of 2.0 A or better. The closest interatomic distance between the guanidinium and carboxyl is less than 4.2 A for 74 arginine and carboxyl groups, with the majority of these lying within hydrogen-bonding distance (2.6-3.0 A). Interacting pairs have been transformed into a common orientation, and arginine-carboxyl, and carboxyl-arginine geometries have been calculated. This has been defined in terms of the spherical polar angles T theta, T phi, and the angle P, between the guanidinium and carboxyl planes. Results show a clear preference for the guanidinium and carboxyl groups to be approximately coplanar, and for the carboxyl oxygens to hydrogen bond with the guanidinium nitrogens. Single nitrogen-single oxygen is the most common type of interaction, however twin nitrogen-twin oxygen interactions also occur frequently. The majority of these occur between the carboxyl oxygens and the NH1 and NE atoms of the arginine, and are only rarely observed for NH1 and NH2. The information presented may be of use in the modelling of arginine-carboxyl interactions within proteins.

A method for fast energy estimation and visualization of protein-ligand interaction

A new computational and graphical method for facilitating ligand-protein docking studies is developed on a three-dimensional computer graphics display. Various physical and chemical properties inside the ligand binding pocket of a receptor protein, whose structure is elucidated by X-ray crystal analysis, are calculated on three-dimensional grid points and are stored in advance. By utilizing those tabulated data, it is possible to estimate the non-bonded and electrostatic interaction energy and the number of possible hydrogen bonds between protein and ligand molecules in real time during an interactive docking operation. The method also provides a comprehensive visualization of the local environment inside the binding pocket. With this method, it becomes easier to find a roughly stable geometry of ligand molecules, and one can therefore make a rapid survey of the binding capability of many drug candidates. The method will be useful for drug design as well as for the examination of protein-ligand interactions.

Molecular cloning and sequencing of a cDNA encoding the acyl carrier protein and its flanking domains in the mammalian fatty acid synthetase

Cloned cDNAs containing coding sequences for domains proximal to the carboxy terminus of the rat fatty acid synthetase have been isolated using an expression vector and domain-specific antibodies. The coding regions were assigned to specific domains of the multifunctional complex by identification of sequences coding for characterized peptide fragments and by recognition of sequences homologous to other monofunctional enzymes. Two clones contain the entire coding region for the acyl carrier protein domain. The sequence is flanked at the 3'-end by a region coding for the thioesterase domain and at the 5'-end by a sequence coding for a reductase, most likely the ketoreductase domain. Thus the ordering of these domain-coding regions in the fatty acid synthetase mRNA is established. The acyl carrier protein domain exhibits about 25% homology with that of the discrete monofunctional acyl carrier proteins of Escherichia coli, spinach and barley, the ketoreductase domain exhibits about 25% homology with bacterial dihydrofolate reductases and the active site of the thioesterase domain exhibits both primary and secondary structural features common to the serine proteases. These findings lend support to the hypothesis that the polyfunctional fatty acid synthetase probably arose by a complex evolutionary process involving fusion of genes coding for seven individual enzymes.

Potential targets for antiviral chemotherapy

Serendipity and random screening have been successful in producing effective antiviral agents. The increase in our knowledge of the basic biochemistry of viral replication and of virus-host interrelationships has revealed not only an understanding of the targets upon which existing antiviral agents exert their inhibitory effect, but also has uncovered new potential targets. The hope is that such molecular understanding will afford the synthesis of compounds with selective antiviral activity. A review of various viral targets which are potentially susceptible to attack, and a few approaches for development of antiviral agents are presented.

Dihydrofolate reductase. 1H resonance assignments and coenzyme-induced conformational changes

Lactobacillus casei dihydrofolate reductase has been studied in solution by one and two-dimensional 1H nuclear magnetic resonance (n.m.r.) spectroscopy at 500 MHz. By using a combination of n.m.r. methods in conjunction with the crystal structure of the enzyme-methotrexate-NADPH complex, resonances have been assigned for 32 of the 162 residues of the enzyme. These are widely distributed throughout the structure of the protein, and include all the histidine and tyrosine residues, as well as several valine, leucine, isoleucine and phenylalanine residues. The assignments have been made for the enzyme-methotrexate and enzyme-methotrexate-NADP+ complexes as well as the enzyme-methotrexate-NADPH complex. Comparison of assigned resonances in the spectra of the three complexes has permitted a preliminary assessment of structural differences between them. The beta-sheet "core" of the protein is unaffected by coenzyme binding, but two regions of the structure that undergo coenzyme-induced conformation changes have been identified. These are the loop comprising residues 13 to 23, and alpha-helix C (residues 42 to 49).