Ligand-specific scoring functions: improved ranking of docking solutions.

Molecular docking is a powerful computational method that has been widely used in many biomolecular studies to predict geometry of a protein-ligand complex. However, while its conformational search algorithms are usually able to generate correct conformation of a ligand in the binding site, the scoring methods often fail to discriminate it among many false variants. We propose to treat this problem by applying more precise ligand-specific scoring filters to re-rank docking solutions. In this way specific features of interactions between protein and different types of compounds can be implicitly taken into account. New scoring functions were constructed including hydrogen bonds, hydrophobic and hydrophilic complementarity terms. These scoring functions also discriminate ligands by the size of the molecule, the total hydrophobicity, and the number of peptide bonds for peptide ligands. Weighting coefficients of the scoring functions were adjusted using a training set of 60 protein-ligand complexes. The proposed method was then tested on the results of docking obtained for an additional 70 complexes. In both cases the success rate was 5-8% better compared to the standard functions implemented in popular docking software.

Structure of the complex of leech-derived tryptase inhibitor (LDTI) with trypsin and modeling of the LDTI-tryptase system.

BACKGROUND: Tryptase is a trypsin-like serine proteinase stored in the cytoplasmic granules of mast cells, which has been implicated in a number of mast cell related disorders such as asthma and rheumatoid arthritis. Unlike almost all other serine proteinases, tryptase is fully active in plasma and in the extracellular space, as there are no known natural inhibitors of tryptase in humans. Leech-derived tryptase inhibitor (LDTI), a protein of 46 amino acids, is the first molecule found to bind tightly to and specifically inhibit human tryptase in the nanomolar range. LDTI also inhibits trypsin and chymotrypsin with similar affinities. The structure of LDTI in complex with an inhibited proteinase could be used as a template for the development of low molecular weight tryptase inhibitors. RESULTS: The crystal structure of the complex between trypsin and LDTI was solved at 2.0 A resolution and a model of the LDTI-tryptase complex was created, based on this X-ray structure. LDTI has a very similar fold to the third domain of the turkey ovomucoid inhibitor. LDTI interacts with trypsin almost exclusively through its binding loop (residues 3-10) and especially through the sidechain of the specificity residue Lys8. Our modeling studies indicate that these interactions are maintained in the LDTI-tryptase complex. CONCLUSIONS: The insertion of nine residues after residue 174 in tryptase, relative to trypsin and chymotrypsin, prevents inhibition by other trypsin inhibitors and is certainly responsible for the higher specificity of tryptase relative to trypsin. In LDTI, the disulfide bond between residues 4 and 25 causes a sharp turn from the binding loop towards the N terminus, holding the N terminus away from the 174 loop of tryptase.

Comparative analysis of the X-ray structures of HIV-1 and HIV-2 proteases in complex with CGP 53820, a novel pseudosymmetric inhibitor.

BACKGROUND: The human immunodeficiency virus (HIV) is the causative agent of acquired immunodeficiency syndrome (AIDS). Two subtypes of the virus, HIV-1 and HIV-2, have been characterized. The protease enzymes from these two subtypes, which are aspartic acid proteases and have been found to be essential for maturation of the infectious particle, share about 50% sequence identity. Differences in substrate and inhibitor binding between these enzymes have been previously reported. RESULTS: We report the X-ray crystal structures of both HIV-1 and HIV-2 proteases each in complex with the pseudosymmetric inhibitor, CGP 53820, to 2.2 A and 2.3 A, respectively. In both structures, the entire enzyme and inhibitor could be located. The structures confirmed earlier modeling studies. Differences between the CGP 53820 inhibitory binding constants for the two enzymes could be correlated with structural differences. CONCLUSIONS: Minor sequence changes in subsites at the active site can explain some of the observed differences in substrate and inhibitor binding between the two enzymes. The information gained from this investigation may help in the design of equipotent HIV-1/HIV-2 protease inhibitors.

A new structural class of serine protease inhibitors revealed by the structure of the hirustasin-kallikrein complex.

BACKGROUND: Hirustasin belongs to a class of serine protease inhibitors characterized by a well conserved pattern of cysteine residues. Unlike the closely related inhibitors, antistasin/ghilanten and guamerin, which are selective for coagulation factor Xa or neutrophil elastase, hirustasin binds specifically to tissue kallikrein. The conservation of the pattern of cysteine residues and the significant sequence homology suggest that these related inhibitors possess a similar three-dimensional structure to hirustasin. RESULTS: The crystal structure of the complex between tissue kallikrein and hirustasin was analyzed at 2.4 resolution. Hirustasin folds into a brick-like structure that is dominated by five disulfide bridges and is sparse in secondary structural elements. The cysteine residues are connected in an abab cdecde pattern that causes the polypeptide chain to fold into two similar motifs. As a hydrophobic core is absent from hirustasin the disulfide bridges maintain the tertiary structure and present the primary binding loop to the active site of the protease. The general structural topography and disulfide connectivity of hirustasin has not previously been described. CONCLUSIONS: The crystal structure of the kallikrein-hirustasin complex reveals that hirustasin differs from other serine protease inhibitors in its conformation and its disulfide bond connectivity, making it the prototype for a new class of inhibitor. The disulfide pattern shows that the structure consists of two domains, but only the C-terminal domain interacts with the protease. The disulfide pattern of the N-terminal domain is related to the pattern found in other proteins. Kallikrein recognizes hirustasin by the formation of an antiparallel beta sheet between the protease and the inhibitor. The P1 arginine binds in a deep negatively charged pocket of the enzyme. An additional pocket at the periphery of the active site accommodates the sidechain of the P4 valine.

The 1.2 A crystal structure of hirustasin reveals the intrinsic flexibility of a family of highly disulphide-bridged inhibitors.

BACKGROUND: Leech-derived inhibitors have a prominent role in the development of new antithrombotic drugs, because some of them are able to block the blood coagulation cascade. Hirustasin, a serine protease inhibitor from the leech Hirudo medicinalis, binds specifically to tissue kallikrein and possesses structural similarity with antistasin, a potent factor Xa inhibitor from Haementeria officinalis. Although the 2.4 A structure of the hirustasin-kallikrein complex is known, classical methods such as molecular replacement were not successful in solving the structure of free hirustasin. RESULTS: Ab initio real/reciprocal space iteration has been used to solve the structure of free hirustasin using either 1.4 A room temperature data or 1.2 A low temperature diffraction data. The structure was also solved independently from a single pseudo-symmetric gold derivative using maximum likelihood methods. A comparison of the free and complexed structures reveals that binding to kallikrein causes a hinge-bending motion between the two hirustasin subdomains. This movement is accompanied by the isomerisation of a cis proline to the trans conformation and a movement of the P3, P4 and P5 residues so that they can interact with the cognate protease. CONCLUSIONS: The inhibitors from this protein family are fairly flexible despite being highly cross-linked by disulphide bridges. This intrinsic flexibility is necessary to adopt a conformation that is recognised by the protease and to achieve an optimal fit, such observations illustrate the pitfalls of designing inhibitors based on static lock-and-key models. This work illustrates the potential of new methods of structure solution that require less or even no prior phase information.

Three-dimensional structure of the bifunctional enzyme phosphoribosylanthranilate isomerase: indoleglycerolphosphate synthase from Escherichia coli refined at 2.0 A resolution.

The three-dimensional structure of the monomeric bifunctional enzyme N-(5'-phosphoribosyl)anthranilate isomerase:indole-3-glycerol-phosphate synthase from Escherichia coli has been refined at 2.0 A resolution, using oscillation film data obtained from synchrotron radiation. The model includes the complete protein (452 residues), two phosphate ions and 628 water molecules. The final R-factor is 17.3% for all observed data between 15 and 2 A resolution. The root-mean-square deviations from ideal bond lengths and bond angles are 0.010 A and 3.2 degrees, respectively. The structure of N-(5'-phosphoribosyl)anthranilate isomerase: indole-3-glycerol-phosphate synthase from E. coli comprises two beta/alpha-barrel domains that superimpose with a root-mean-square deviation of 2.03 A for 138 C alpha-pairs. The C-terminal domain (residues 256 to 452) catalyses the PRAI reaction and the N-terminal domain (residues 1 to 255) catalyses the IGPS reaction, two sequential steps in tryptophan biosynthesis. The enzyme has the overall shape of a dumb-bell, resulting in a surface area that is considerably larger than normally observed for monomeric proteins of this size. The active sites of the PRAI and the IGPS domains, both located at the C-terminal side of the central beta-barrel, contain equivalent binding sites for the phosphate moieties of the substrates N-(5'-phosphoribosyl) anthranilate and 1-(o-carboxyphenylamino)-1-deoxyribulose-5-phosphate. These two phosphate binding sites are identical with respect to their positions within the tertiary structure of the beta/alpha-barrel, the conformation of the residues involved in phosphate binding and the hydrogen-bonding network between the phosphate ions and the protein. The active site cavities of both domains contain similar hydrophobic pockets that presumably bind the anthranilic acid moieties of the substrates. These similarities of the tertiary structures and the active sites of the two domains provide evidence that N-(5'-phosphoribosyl)anthranilate isomerase:indole-3-glycerol-phosphate synthase from E. coli results from a gene duplication event of a monomeric beta/alpha-barrel ancestor.

Refined crystal structures of subtilisin novo in complex with wild-type and two mutant eglins. Comparison with other serine proteinase inhibitor complexes.

The crystal structures of the complexes formed between subtilisin Novo and three inhibitors, eglin c, Arg45-eglin c and Lys53-eglin c have been determined using molecular replacement and difference Fourier techniques and refined at 2.4 A, 2.1 A, and 2.4 A resolution, respectively. The mutants Arg45-eglin c and Lys53-eglin c were constructed by site-directed mutagenesis in order to investigate the inhibitory specificity and stability of eglin c. Arg45-eglin became a potent trypsin inhibitor, in contrast to native eglin, which is an elastase inhibitor. This specificity change was rationalized by comparing the structures of Arg45-eglin and basic pancreatic trypsin inhibitor and their interactions with trypsin. The residue Arg53, which participates in a complex network of hydrogen bonds formed between the core and the binding loop of eglin c, was replaced with the shorter basic amino acid lysine in the mutant Lys53-eglin. Two hydrogen bonds with Thr44, located in the binding loop, can no longer be formed but are partially restored by a water molecule bound in the vicinity of Lys53. Eglin c in complexes with both subtilisin Novo and subtilisin Carlsberg was crystallized in two different space groups. Comparison of the complexes showed a rigid body rotation for the eglin c core of 11.5 degrees with respect to the enzyme, probably caused by different intermolecular contacts in both crystal forms.

Changes in interactions in complexes of hirudin derivatives and human alpha-thrombin due to different crystal forms.

The three-dimensional structures of D-Phe-Pro-Arg-chloromethyl ketone-inhibited thrombin in complex with Tyr-63-sulfated hirudin (ternary complex) and of thrombin in complex with the bifunctional inhibitor D-Phe-Pro-Arg-Pro-(Gly)4-hirudin (CGP 50,856, binary complex) have been determined by X-ray crystallography in crystal forms different from those described by Skrzypczak-Jankun et al. (Skrzypczak-Jankun, E., Carperos, V.E., Ravichandran, K.G., & Tulinsky, A., 1991, J. Mol. Biol. 221, 1379-1393). In both complexes, the interactions of the C-terminal hirudin segments of the inhibitors binding to the fibrinogen-binding exosite of thrombin are clearly established, including residues 60-64, which are disordered in the earlier crystal form. The interactions of the sulfate group of Tyr-63 in the ternary complex structure explain why natural sulfated hirudin binds with a 10-fold lower K(i) than the desulfated recombinant material. In this new crystal form, the autolysis loop of thrombin (residues 146-150), which is disordered in the earlier crystal form, is ordered due to crystal contacts. Interactions between the C-terminal fragment of hirudin and thrombin are not influenced by crystal contacts in this new crystal form, in contrast to the earlier form. In the bifunctional inhibitor-thrombin complex, the peptide bond between Arg-Pro (P1-P1') seems to be cleaved.

The crystal structure of TGF-beta 3 and comparison to TGF-beta 2: implications for receptor binding.

Transforming growth factors beta belong to a group of cytokines that control cellular proliferation and differentiation. Five isoforms are known that share approximately 75% sequence identity, but exert different biological activities. The structure of TGF-beta 3 was solved by X-ray crystallography and refined to a final R-factor of 17.5% at 2.0 A resolution. Comparison with the structure of TGF-beta 2 (Schlunegger MP, Grütter MG, 1992, Nature 358:430-434; Daopin S, Piez KA, Ogawa Y, Davies DR, 1992, Science 257:369-373) reveals a virtually identical central core. Differences exist in the conformations of the N-terminal alpha-helix and in the beta-sheet loops. In TGF-beta 3, the N-terminal alpha-helix has moved approximately 1 A away from the central core. This movement can be correlated with the mutation of Leu 17 to Val and Ala 47 to Pro in TGF-beta 3. The beta-sheet loops rotate as a rigid body 9 degrees around an axis that runs approximately parallel to the dimer axis. If these differences are recognized by the TGF-beta receptors, they might account for the individual cellular responses. A molecule of the precipitating agent dioxane is bound in a crystal contact, forming a hydrogen bond with Trp 32. This dioxane may occupy a carbohydrate-binding site, because dioxane possesses some structural similarity with a carbohydrate. The dioxane is in contact with two tryptophans, which are often involved in carbohydrate recognition.

Recombinant hirustasin: production in yeast, crystallization, and interaction with serine proteases.

A synthetic gene coding for the 55-amino acid protein hirustasin, a novel tissue kallikrein inhibitor from the leech Hirudo medicinalis, was generated by polymerase chain reaction using overlapping oligonucleotides, fused to the yeast alpha-factor leader sequence and expressed in Saccharomyces cerevisiae. Recombinant hirustasin was secreted mainly as incompletely processed fusion protein, but could be processed in vitro using a soluble variant of the yeast yscF protease. The processed hirustasin was purified to better than 97% purity. N-terminal sequence analysis and electrospray ionization mass spectrometry confirmed a correctly processed N-terminus and the expected amino acid sequence and molecular mass. The biological activity of recombinant hirustasin was identical to that of the authentic leech protein. Crystallized hirustasin alone and in complex with tissue kallikrein diffracted beyond 1.4 A and 2.4 A, respectively. In order to define the reactive site of the inhibitor, the interaction of hirustasin with kallikrein, chymotrypsin, and trypsin was investigated by monitoring complex formation in solution as well as proteolytic cleavage of the inhibitor. During incubation with high, nearly equimolar concentration of tissue kallikrein, hirustasin was cleaved mainly at the peptide bond between Arg 30 and Ile 31, the putative reactive site, to yield a modified inhibitor. In the corresponding complex with chymotrypsin, mainly uncleaved hirustasin was found and cleaved hirustasin species accumulated only slowly. Incubation with trypsin led to several proteolytic cleavages in hirustasin with the primary scissile peptide bond located between Arg 30 and Ile 31. Hirustasin appears to fall into the class of protease inhibitors displaying temporary inhibition.

A Lys27-to-Glu27 mutation in the human insulin-like growth factor-1 prevents disulfide linked dimerization and allows secretion of BiP when expressed in yeast.

Recombinant human insulin-like growth factor-1 (IGF1) secreted from yeast contains only 10-15% of the active monomer. A majority of the IGF1-like molecules are disulfide bonded dimers. These dimers are not formed in an IGF1 mutant where Lys27 has been replaced by glutamic acid. However, increased levels of secreted BiP (the yeast KAR2 gene product) are seen in cells expressing the mutant. These results imply that by preventing ionic interactions between two IGF1 molecules, intermolecular disulfide bonds do not form in yeast, and that in the mutant there is a structural change which induces BiP, allowing its secretion.

X-ray crystal structure of the serine proteinase inhibitor eglin c at 1.95 A resolution.

The crystal structure of eglin c, naturally occurring in the leech Hirudo medicinalis, is known from its complexes with various serine proteinases, but the crystallization of free eglin c has not yet been reported. A method is described for growing well-diffracting crystals of free eglin c from highly concentrated protein solutions (approximately 200 mg/ml). The space group of the orthorhombic crystals was determined to be P2(1)2(1)2(1) with unit cell parameters a = 32.6, b = 42.0, c = 44.1 A. The structure of free eglin c was resolved at 1.95 A resolution by Patterson search methods. The final model contains all 70 amino acids of eglin c and 125 water molecules. In comparison to the eglin structure known from its complexes with proteinases, only small differences have been observed in free eglin c. However, the reactive site-binding loop and a few residues on the surface of eglin have been found in different conformations due to crystal contacts. In contrast to the complex structures, the first seven amino acids of the highly flexible amino terminus can be located. Crystallographic refinement comprised molecular dynamics refinement, classical restrained least-squares refinement and individual isotropic atomic temperature refinement. The final R-factor is 15.8%.

Active site binding loop stabilization in the subtilisin inhibitor eglin c: structural and functional studies on specifically designed mutants in complex with subtilisin and the uncomplexed inhibitor.

As known from the x-ray crystal structure in complex with a proteinase and from NMR studies, the serine proteinase inhibitor eglin c has a wedge-like shape with a hydrophobic core and a solvent exposed active site binding loop which is stabilized by a network of non-covalent core-binding loop interactions. Previous studies implied a crucial role of the P1'-residue Asp-46 for binding loop stabilization and high inhibitory potency of eglin c towards serine proteinases such as subtilisin. In the present study, the formation of specific eglin core-binding loop interactions was modulated by replacing the wildtype Asp-46 by asparagine, glutarnate and glutamine. The x-ray crystal structures of these mutants were solved in complex with subtilisin, and the inhibitory potency towards this enzyme was determined. Our results imply a reduction of inhibitory potency with declining core-binding loop interactions. We succeeded in crystallizing free wildtype eglin c. The 1.95 angstroms x-ray crystal structure indicates that the transition from the free to the bound form of eglin is accompanied by a concerted conformational change in the binding loop, implying an induced fit to the accessible enzyme surface. Except for the binding loop domain and a few residues on the surface of eglin, the differences observed between the uncomplexed and bound form of the inhibitor are only small.

The three-dimensional structure of interleukin-1 beta.

The three-dimensional structure of human recombinant interleukin-1 beta has been determined at 0.24 nm resolution by X-ray crystallographic techniques. The partially refined model has a crystallographic R-factor of just under 19%. The structure is composed of 12 beta-strands forming a complex network of hydrogen bonds. The core of the structure can best be described as a tetrahedron whose edges are each formed by two antiparallel beta-strands. The interior of this structure is filled with hydrophobic side-chains. There is a 3-fold repeat in the folding of the polypeptide chain. Although this folding pattern suggests gene triplication, no significant internal sequence homology between topologically corresponding residues exists. The folding topology of interleukin-1 beta is very similar to that described by A. D. McLachlan [(1979) J. Mol. Biol. 133, 557-563] for soybean trypsin inhibitor.

Crystal structure of the cytokine interleukin-1 beta.

The crystal structure of human recombinant interleukin-1 beta has been determined at 3.0 A resolution by the isomorphous replacement method in conjunction with solvent flattening techniques. The model prior to refinement has a crystallographic R-factor of 42.3%. The structure is composed of 12 beta-strands forming a complex network of hydrogen bonds. The core of the structure can best be described as a tetrahedron whose edges are each formed by two antiparallel beta-strands. The interior of this structure is filled with hydrophobic side chains. There is a 3-fold repeat in the folding of the polypeptide chain. Although this folding pattern suggests gene triplication, no strong internal sequence homology between topologically corresponding residues exists. The folding topology of interleukin-1 beta is very similar to that described by McLachlan (1979) J. Mol. Biol., 133, 557-563, for soybean trypsin inhibitor.

Crystallographic refinement of interleukin 1 beta at 2.0 A resolution.

The structure of human recombinant interleukin 1 beta (IL-1 beta) has been refined by a restrained least-squares method to a crystallographic R factor of 17.2% to 2.0 A resolution. One-hundred sixty-eight solvent molecules have been located, and isotropic temperature factors for each atom have been refined. The overall structure is composed of 12 beta-strands that can best be described as forming the four triangular faces of a tetrahedron with hydrogen bonding resembling normal antiparallel beta-sheets only at the vertices. The interior of this tetrahedron is filled by hydrophobic side chains. Analysis of sequence alignments with IL-1 beta from other mammalian species shows the interior to be very well conserved with the exterior residues markedly less so. There does not appear to be a clustering of invariant amino acid side chains on the surface of the molecule, suggesting an area of interaction with the IL-1 receptor. Comparison of the IL-1 beta structure with IL-1 alpha sequences indicates that IL-1 alpha probably has a similar overall folding as IL-1 beta but binds to the receptor in a different fashion. The three-dimensional structure of the IL-1 beta is analyzed in light of what has been suggested by previously published work on mutants and fragments of the molecule.

Crystallization and preliminary x-ray diffraction studies of recombinant human interleukin-1 beta.

Recombinant human interleukin-1 beta has been crystallized into a tetragonal cell. The unit cell constants are a = b = 54.9 A, c = 76.8 A, and alpha = beta = gamma = 90 degrees. The crystals diffract to better than 1.9 A and are suitable for high resolution data collection. The crystallization conditions and general crystal data are presented.

The crystal structures of recombinant glycosylated human renin alone and in complex with a transition state analog inhibitor.

Recombinant human glycosylated renin has been crystallized in complex with CGP 38'560, a transition state analog inhibitor (IC50 = 2 x 10(-9) M), in a tetragonal crystal form. The structure has been determined to a resolution of 2.4 A and refined to a crystallographic Rfactor of 17.6%. It reveals the conformation of the inhibitor as well as its interactions with the enzyme active site. The active site is a deep cleft between the N- and the C-terminal domains to which the inhibitor binds in an extended conformation filling the S4 to S2' pockets. The structure of the complex is compared with that of the related uninhibited enzyme pepsin. Significant changes in the relative orientation of the N- and C-terminal domains are observed. In the inhibited renin structure the C-terminal loop segments forming the active site are closer to those from the N-terminal domain than in the related "open" pepsin structure. In addition, the structure of uninhibited glycosylated renin has been determined at 2.8 A resolution from a cubic crystal form with two renin molecules in the asymmetric unit. The two independent renin molecules show different conformations with respect to the relative orientation of their N- and C-terminal domains; one molecule is found in the "closed inhibited" conformation, the other in the "open uninhibited" conformation.

Mutational analysis of the structure-function relationship in interferon-alpha.

A number of mutants of the recombinant human hybrid interferon-alpha 8[60]alpha 1[92]alpha 8 have been expressed in Escherichia coli, purified to homogeneity and characterized. The introduction of one or two negative charges results in an anomalously lower electrophoretic mobility than that expected from the molecular mass. The mutations Lys84-->Glu, Cys86-->Tyr, Cys86-->Ser, Thr87-->Ile, Tyr90-->Asp and Lys84-->Glu/Tyr90-->Asp (double mutant) result in a marked decrease of the biological activity in murine cells compared to the unmutated protein. Comparisons with published structural and homology models of interferon-beta and -alpha, imply that these residues are located primarily on the external surface of the carboxylterminus of helix C. We propose a model of interferon-receptor interaction in which these residues define a potential binding site.