Drug-like inhibitors of protein-protein interactions: a structural examination of effective protein mimicry.

Protein-protein interactions represent targets for drug discovery that are highly relevant in a biological sense, but have proven difficult in a practical sense. Nevertheless, there have been recent successes in discovering drug-like small molecule inhibitors of protein-protein systems. To build on this progress, it is worth analyzing successful cases to understand at a molecular level the strategies by which these compounds effectively interfere with protein-protein pairing. A commonly observed situation is one wherein the small molecule acts as a direct mimic of one of the protein partners. This review focuses exclusively on cases where this strategy is employed, and examines the structural characteristics of the binding sites and the conformational attributes of the small molecule ligands. Common traits shared among these successful examples are identified, and formulated into potentially useful guidance for drug discovery efforts within this target class.

Structure of the Ras-binding domain of c-Raf-1 as determined by NMR spectroscopy and identification of the region that interacts with Ras.

The structure of the Ras-binding domain of human c-Raf-1 (residues 55 to 132) as determined in solution by NMR spectroscopy is presented. It consists of a five-stranded beta-sheet, a twelve residue alpha-helix, and an additional one-turn helix. The fold belongs to a known family whose members include ubiquitin and protein G. The surface of Raf55-132 that interacts with Ras has been identified by resonance perturbation mapping. The binding site is a spatially contiguous patch comprised of the two-N-terminal beta-strands, the loop between them, and the C-terminal end of the alpha-helix. A model of the Raf-Ras complex is presented, which was derived by analogy to the complex between protein G and a Fab fragment of IgG. In the model, edge beta-strands of each protein align in an antiparallel orientation, forming a unified beta-sheet, and side chains from both proteins are able to participate in ionic and hydrophobic interactions at the interface.

Solution structure of the Ras-binding domain of c-Raf-1 and identification of its Ras interaction surface.

The structure of the Ras-binding domain of human c-Raf-1 (residues 55-132) has been determined in solution by nuclear magnetic resonance (NMR) spectroscopy. Following complete assignment of the backbone and side-chain 1H, 15N, and 13C resonances, the structure was calculated using the program CHARMM. Over 1300 NOE-derived constraints were applied, resulting in a detailed structure. The fold of Raf55-132 consists of a five-stranded beta-sheet, a 12-residue alpha-helix, and an additional one-turn helix. It is similar to those of ubiquitin and the IgG-binding domain of protein G, although the three proteins share very little sequence identity. The surface of Raf55-132 that interacts with Ras has been identified by monitoring perturbation of line widths and chemical shifts of 15N-labeled Raf55-132 resonances during titration with unlabeled Ras-GMPPNP. The Ras-binding site is contained within a spatially contiguous patch comprised of the N-terminal beta-hairpin and the C-terminal end of the alpha-helix.

Chemical shift assignments and folding topology of the Ras-binding domain of human Raf-1 as determined by heteronuclear three-dimensional NMR spectroscopy.

Raf-1 is a 74-kDa serine-threonine kinase which serves as the immediate downstream target of Ras in the cell growth signal transduction pathway. Recent genetic and biochemical experiments have demonstrated that (1) Ras interacts directly with the amino-terminal domain of Raf and (2) residues 51-131 of the Raf sequence are sufficient to mediate this interaction [Vojtek, A. B., Hollenberg, S. M., & Cooper, J. A. (1993) Cell 74, 205-214]. We have expressed a corresponding segment of the human Raf sequence (Raf55-132) in Escherichia coli as a fusion with maltose binding protein. The fusion protein was purified by affinity chromatography and cleaved at a pre-engineered site with factor Xa protease to liberate the 78-residue fragment of Raf. Raf55-132 bound to Ras with high affinity in a competition assay with GAP. An unlabeled version of Raf55-132 was studied by 2D homonuclear NMR, and uniformly 15N- and 13C/15N-labeled versions of Raf55-132 were studied by 2D and 3D heteronuclear NMR. Nearly complete sequence-specific assignments were made for the backbone HN, H alpha, 15N, and 13C alpha resonances. NOEs were used to determine regions of secondary structure and the overall folding topology. Raf55-132 is an independently folded domain composed of a five-stranded beta-sheet, a three-turn alpha-helix, and possibly an additional one-turn helix. Its structure resembles that of ubiquitin, even though there is no more than 11% sequence homology between the two proteins.

Solution structures of cyclic and dicyclic analogues of growth hormone releasing factor as determined by two-dimensional NMR and CD spectroscopies and constrained molecular dynamics.

Solution structures were determined for a linear analogue of growth hormone releasing factor (GRF), and cyclic and dicyclic analogues in which the side chains of aspartyl and lysyl residues spaced at positions i-(i + 4) were joined to form a lactam. The four analogues were [Ala15]-GRF-(1-29)-NH2 and its cyclo8-12, cyclo21-25, and dicyclo8-12;21-25 derivatives. The peptides were studied in two solvent systems: 75% methanol/25% water at pH 6.0; and 100% water at pH 3.0. CD spectroscopy was used to assess the overall alpha-helical content. Nuclear magnetic resonance spectroscopy was used to determine the structures in more detail. Nearly complete proton resonance assignments were made for each of the peptides, in both solvents. Nuclear Overhauser effects were converted into distance constraints and applied in the molecular dynamics program CHARMM to evaluate the range of low-energy structures that satisfied the nmr data. In 75% methanol, all of the peptides are comprised of a single alpha-helical segment with fraying of one to three residues at each end. The linear analogue has a tendency to kink. In water, the analogues have two helical segments with flexible regions between them and at the termini of the peptides. The linear analogue is helical at residues 7-14 and 21-28. In the cyclo8-12 analogue, the N-terminal helical region extends to include residues 7-19, while the other helical region is slightly shortened. In the cyclo21-25 analogue, the C-terminal helical region is extended to include residues 19-28, while the N-terminal helical region is destabilized. The dicyclic analogue has the largest N-terminal helix, spanning residues 7-20, but its helical segment at residues 21-28 is not well ordered. All of the analogues exhibit substantial biological activity. The cyclic and dicyclic analogues show dramatically increased resistance to degradation during incubation with human plasma. The i-(i + 4) lactam, therefore, appears to be a synthetic means of stabilizing a local alpha-helical conformation, which may be of general use in the design of active, stable peptides.

Solution structure of an analogue of vasoactive intestinal peptide as determined by two-dimensional NMR and circular dichroism spectroscopies and constrained molecular dynamics.

Structures have been determined for a potent analogue of vasoactive intestinal peptide (VIP), Ac-[Lys12, Lys14, Nle17, Val26, Thr28]VIP (VIP'), in methanol/water solutions. In CD studies, both VIP and VIP' were helical in methanol/water, with the percentage of alpha-helix increasing with percentage methanol. The pH had little effect on the structure. Complete 1H NMR assignments were made for VIP' in 25% methanol at pH 4 and 6 and in 50% methanol at pH 6, using two-dimensional COSY, NOESY, and relay-COSY experiments. There were no widespread changes in chemical shifts between the samples at pH 4 and 6; however, widespread changes were observed between the samples in 25% and 50% methanol. Complete sets of NOEs were obtained for VIP' in 25% methanol, pH 4, and in 50% methanol, pH 6. These NOEs were converted into distance constraints and applied in molecular dynamics and energy minimization calculations using the program CHARMM. A set of low-energy structures was obtained for VIP' in each solvent system. In 25% methanol, VIP' has two helical segments at residues 9-17 and 23-28. The remainder of the structure is not well determined. In 50% methanol, residues 8-26 form a regular, well-defined alpha-helix and residues 5-8 form a type III beta-turn. The remaining residues are not ordered. These structural assessments agree with the CD data. In the lowest energy structure in 50% methanol, the side chains of Asp3, Phe6, Thr7, and Tyr10 are clustered together--these residues are conserved throughout the family of peptide hormones homologous to VIP.

Solution structure of the 45-residue MgATP-binding peptide of adenylate kinase as examined by 2-D NMR, FTIR, and CD spectroscopy.

The structure of a synthetic peptide corresponding to residues 1-45 of rabbit muscle adenylate kinase has been studied in aqueous solution by two-dimensional NMR, FTIR, and CD spectroscopy. This peptide, which binds MgATP and is believed to represent most of the MgATP-binding site of the enzyme [Fry, D.C., Kuby, S.A., & Mildvan, A.S. (1985) Biochemistry 24, 4680-4694], appears to maintain a conformation similar to that of residues 1-45 in the X-ray structure of intact porcine adenylate kinase [Sachsenheimer, W., & Schulz, G.E. (1977) J. Mol. Biol. 114, 23-26], with 42% of the residues of the peptide showing NOEs indicative of phi and psi angles corresponding to those found in the protein. The NMR studies suggest that the peptide is composed of two helical regions of residues 4-7 and 23-29, and three stretches of beta-strand at residues 8-15, 30-32, and 35-40, yielding an overall secondary structure consisting of 24% alpha-helix, 38% beta-structure, and 38% aperiodic. Although the resolution-enhanced amide I band of the peptide FTIR spectrum is broad and rather featureless, possibly due to disorder, it can be fit by using methods developed on well-characterized globular proteins. On this basis, the peptide consists of 35 +/- 10% beta-structure, 60 +/- 12% turns and aperiodic structure, and not more than 10% alpha-helix. The CD spectrum is best fit by assuming the presence of at most 13% alpha-helix in the peptide, 24 +/- 2% beta-structure, and 66 +/- 4% aperiodic. The inability of the high-frequency FTIR and CD methods to detect helices in the amount found by NMR may result from the short helical lengths as well as from static and dynamic disorder in the peptide. Upon binding of MgATP, numerous conformational changes in the backbone of the peptide are detected by NMR, with smaller alterations in the overall secondary structure as assessed by CD. Detailed assignments of resonances in the peptide spectrum and intermolecular NOEs between protons of bound MgATP and those of the peptide, as well as chemical shifts of peptide resonances induced by the binding of MgATP, are consistent with the previously proposed binding site for MgATP on adenylate kinase.

Distinguishing among protein kinases by substrate specificities.

In the previous paper, N-methylated peptides were shown to be sensitive probes of substrate conformation within the adenosine cyclic 3',5'-phosphate dependent protein kinase (A-kinase) active site. While it has been shown that other protein kinases will catalyze the phosphorylation of the same peptide sequences as A-kinase, there is as yet little information as to whether the protein kinases differentiate between substrates on the basis of conformation. For this reason, the conformationally restricted N-methylated peptides were used to probe the active site of guanosine cyclic 3',5'-phosphate dependent protein kinase (G-kinase), which is homologous in sequence to [Takio, K., Wade, R. D., Smith, S. B., Krebs, E. G., Walsh, K. A., & Titani, K. (1984) Biochemistry 23, 4207-4218] and which has substrate specificities similar to [Lincoln, T. M., & Corbin, J. D. (1977) Proc. Natl. Acad. Sci. U.S.A. 74, 3239-3243] those of A-kinase. Although this enzyme appears to bind the peptides in a conformation resembling that of conformation A, it is more able to accommodate backbone methylation than is A-kinase. A peptide substrate at least 700-fold selective for G-kinase over A-kinase was found. Backbone methylation may, therefore, represent a way of making peptide substrates and inhibitors selective for a particular kinase.

Conformation of Leu-Arg-Arg-Ala-Ser-Leu-Gly bound in the active site of adenosine cyclic 3',5'-phosphate dependent protein kinase.

Studies utilizing NMR spectroscopy have shown that adenosine cyclic 3',5'-phosphate dependent protein kinase (A-kinase) probably binds Leu-Arg-Arg-Ala-Ser-Leu-Gly (peptide 1) in one of two extended coil conformations (A or B). The relative reactivities of a series of N-methylated peptides based on the structure of peptide 1 might, therefore, be related to how well each can assume the A or B conformation. From estimates of the magnitude of steric interactions that would be induced by N-methylation of an amide in peptide 1 that is locked in either conformation, the ability of each peptide to form that conformation was predicted. The ability of A-kinase to catalyze phosphorylation of the N-methylated peptides correlated well with the ability of each peptide to form conformation A, but not conformation B. In accord with these findings, the reactivity of an unreactive N-methylated peptide was partially restored by a second change, which allowed the peptide to assume conformation A. These results suggest that, when bound in the enzymatic active site, peptide 1 has a conformation that resembles structure A much more closely than structure B.

NMR studies of the AMP-binding site and mechanism of adenylate kinase.

NMR has previously been used to determine the conformation of enzyme-bound MgATP and to locate the MgATP-binding site on adenylate kinase [Fry, D. C., Kuby, S. A., & Mildvan, A. S. (1985) Biochemistry 24, 4680-4694]. To determine the conformation and location of the other substrate, AMP, distances have been measured from Cr3+AMPPCP, a linear competitive inhibitor with respect to MgATP, to six protons and to the phosphorus atom of AMP on adenylate kinase, with the paramagnetic probe-T1 method. Time-dependent nuclear Overhauser effects (NOEs) have been used to measure five interproton distances on enzyme-bound AMP. These distances were used to determine the conformation of bound AMP in addition to its position with respect to metal-ATP. Enzyme-bound AMP exhibits a high anti-glycosyl torsional angle (chi = 110 +/- 10 degrees), a 3'-endo,2'-exo ribose pucker (delta = 105 +/- 10 degrees), and gauche-trans orientations about the C4'-C5' bond (gamma = 180 +/- 10 degrees) and the C5'-O5' bond (beta = 170 +/- 20 degrees). The distance from Cr3+ to the phosphorus of AMP is 5.9 +/- 0.3 A, indicating a reaction coordinate distance of approximately 3 A, which is consistent with an associative SN2 mechanism for the phosphoryl transfer. Ten intermolecular NOEs, from protons of the enzyme to those of AMP, were detected, indicating the proximity of at least three hydrophobic amino acids to bound AMP. These constraints, together with the conformation of AMP and the intersubstrate distances, were used to position AMP into the X-ray structure of adenylate kinase. The AMP binding site is found to be near (less than or equal to 4 A from) Leu-116, Arg-171, Val-173, Val-182, and Leu-190; all of these residues have been found to be invariant in muscle-type rabbit, calf, human, porcine [Kuby, S. A., Palmieri, R. H., Frischat, A., Fischer, A. H., Wu, L. H., Maland, L., & Manship, M. (1984) Biochemistry 23, 2393-2399], and chicken adenylate kinase [Kishi, F., Maruyama, M., Tanizawa, Y., & Nakazawa, A. (1986) J. Biol. Chem. 261, 2942-2945].

ATP-binding site of adenylate kinase: mechanistic implications of its homology with ras-encoded p21, F1-ATPase, and other nucleotide-binding proteins.

The MgATP binding site of adenylate kinase, located by a combination of NMR and x-ray diffraction, is near three protein segments, five to seven amino acids in length, that are homologous in sequence to segments found in other nucleotide-binding phosphotransferases, such as myosin and F1-ATPase, ras p21 and transducin GTPases, and cAMP-dependent and src protein kinases, suggesting equivalent mechanistic roles of these segments in all of these proteins. Segment 1 is a glycine-rich flexible loop that, on adenylate kinase, may control access to the ATP-binding site by changing its conformation. Segment 2 is an alpha-helix containing two hydrophobic residues that interact with the adenine-ribose moiety of ATP, and a lysine that may bind to the beta- and gamma-phosphates of ATP. Segment 3 is a hydrophobic strand of parallel beta-pleated sheet, terminated by a carboxylate, that flanks the triphosphate binding site. The various reported mutations of ras p21 that convert it to a transforming agent all appear to involve segment 1, and such substitutions may alter the properties of p21 by hindering a conformational change at this segment. In F1-ATPase, the flexible loop may, by its position, control both the accessibility and the ATP/ADP equilibrium constant on the enzyme.

NMR studies of the MgATP binding site of adenylate kinase and of a 45-residue peptide fragment of the enzyme.

Proton NMR was used to study the interaction of beta,gamma-bidentate Cr3+ATP and MgATP with rabbit muscle adenylate kinase, which has 194 amino acids, and with a synthetic peptide consisting of residues 1-45 of the enzyme, which has previously been shown to bind MgepsilonATP [Hamada, M., Palmieri, R. H., Russell, G. A., & Kuby, S. A. (1979) Arch. Biochem. Biophys. 195, 155-177]. The peptide is globular and binds Cr3+ATP competitively with MgATP with a dissociation constant, KD(Cr3+ATP) = 35 microM, comparable to that of the complete enzyme [KI(Cr3+ATP) = 12 microM]. Time-dependent nuclear Overhauser effects (NOE's) were used to measure interproton distances on enzyme- and peptide-bound MgATP. The correlation time was measured directly for peptide-bound MgATP by studying the frequency dependence of the NOE's at 250 and 500 MHz. The H2' to H1' distance so obtained (3.07 A) was within the range established by X-ray and model-building studies of nucleotides (2.9 +/- 0.2 A). Interproton distances yielded conformations of enzyme- and peptide-bound MgATP with indistinguishable anti-glycosyl torsional angles (chi = 63 +/- 12 degrees) and 3'-endo/O1'-endo ribose puckers (sigma = 96 +/- 12 degrees). Enzyme- and peptide-bound MgATP molecules exhibited different C4'-C5' torsional angles (gamma) of 170 degrees and 50 degrees, respectively. Ten intermolecular NOE's from protons of the enzyme and four such NOE's from protons of the peptide to protons of bound MgATP were detected, which indicated proximity of the adenine ribose moiety to the same residues on both the enzyme and the peptide. Paramagnetic effects of beta,gamma-bidentate Cr3+ATP on the longitudinal relaxation rates of protons of the peptide provided a set of distances to the side chains of five residues, which allowed the location of the bound Cr3+ atom to be uniquely defined. Distances from enzyme-bound Cr3+ATP to the side chains of three residues of the protein agreed with those measured for the peptide. The mutual consistency of interproton and Cr3+ to proton distances obtained in metal-ATP complexes of both the enzyme and the peptide suggests that the conformation of the peptide is very similar to that of residues 1-45 of the enzyme. When this was assumed to be the case and when molecular models and a computer graphics system were used, MgATP could be fit into the X-ray structure of adenylate kinase in a unique manner such that all of the distances determined by NMR were accommodated.(ABSTRACT TRUNCATED AT 400 WORDS)

NMR studies of the backbone protons and secondary structure of pentapeptide and heptapeptide substrates bound to bovine heart protein kinase.

The conformations of enzyme-bound pentapeptide (Arg-Arg-Ala-Ser-Leu) and heptapeptide (Leu-Arg-Arg-Ala-Ser-Leu-Gly) substrates of protein kinase have been studied by NMR in quaternary complexes of the type (Formula: see text). Paramagnetic effects of Mn2+ bound at the inhibitory site of the catalytic subunit on the longitudinal relaxation rates of backbone Ca protons, as well as on side-chain protons of the bound pentapeptide and heptapeptide substrates, have been used to determine Mn2+ to proton distances which range from 8.2 to 12.4 A. A combination of the paramagnetic probe-T1 method with the Redfield 2-1-4-1-2 pulse sequence for suppression of the water signal has been used to measure distances from Mn2+ to all of the backbone amide (NH) protons of the bound pentapeptide and heptapeptide substrates, which range from 6.8 to 11.1 A. Paramagnetic effects on the transverse relaxation rates yield rate constants for peptide exchange, indicating that the complexes studied by NMR dissociate rapidly enough to participate in catalysis. Model-building studies based on the Mn2+-proton distances, as well as on previously determined distances from Cr3+-AMPPCP to side-chain protons [Granot, J., Mildvan, A.S., Bramson, H. N., & Kaiser, E. T. (1981) Biochemistry 20, 602], rule out alpha-helical, beta-sheet, beta-bulge, and all possible beta-turn conformations within the bound pentapeptide and heptapeptide substrates. The distances are fit only by extended coil conformations for the bound peptide substrates with a minor difference between the pentapeptides and heptapeptides in the phi torsional angle at Arg3C alpha and in psi at Arg2C alpha. An extended coil conformation, which minimizes the number of interactions within the substrate, would facilitate enzyme-substrate interaction and could thereby contribute to the specificity of protein kinase.

Photoinactivation of agaricus bisporus tyrosinase: modification of the binuclear copper site.

Irradiation of Agaricus bisporus tyrosinase in the presence of citrate at pH 5.6 with 300-420-nm light results in a loss of both catecholase activity and cresolase activity. The light-sensitive species appears to be an enzyme-citrate complex, most likely involving coordination of citrate to the active site copper. One copper ion from each binuclear active site can be removed from the inactivated enzyme, resulting in the formation of a met apo derivative. The electron spin resonance spectrum of met apo tyrosinase resembles that of met apo hemocyanin and half-met Neurospora tyrosinase. It is consistent with a distorted square-planar geometry around the copper and with either nitrogen or nitrogen and oxygen ligands. Amino acid analysis indicates that four histidines on the heavy subunit are destroyed during the inactivation process. Some or all of these histidines may serve as ligands to the copper ion which becomes labile after inactivation. Photoinactivation results in decarboxylation of citrate and does not require the presence of oxygen. The reaction may involve generation of a free radical from the citrate which then attacks nearby histidine residues.

Behaviour variables in superior swimmers.

This study investigated relationships between behavioural responses on the Psychological Inventories for Competitive Swimmers and selected performance classifications and characteristics of male (N = 91) and female (N = 85) swimmers at the 1977 Canadian Winter National Swimming Championships. Responses were divided into four groups representing social, attitudinal, and activity behaviour and motivational events. These responses were compared on the basis of sex, age, years of participation in serious swimming, past performance achievement categories, current performance achievement categories, and performance improvement. Swimmers appeared to differ in their behavioural responses across the dimensions of sex, years of participation, and performance time improvement or non-improvement.