Identification of amino acid residues important for ligand binding to Fas.

The interaction of Fas (CD95), a member of the tumor necrosis factor receptor (TNFR) family, and its ligand (FasL) triggers programmed cell death (apoptosis) and is involved in the regulation of immune responses. Although the Fas-FasL interaction is conserved across species barriers, little is currently known about the molecular details of this interaction. Our aim was to identify residues in Fas that are important for ligand binding. With the aid of a Fas molecular model, candidate amino acid residues were selected in the Fas extracellular domain 2 (D2) and D3 and subjected to serine-scanning mutagenesis to produce mutant Fas molecules in the form of Ig fusion proteins. The effects of these mutations on FasL binding was examined by measuring the ability of these proteins to inhibit FasL-mediated apoptosis of Jurkat cells and bind FasL in ELISA and BIAcore assays. Mutation of two amino acids, R86 and R87 (D2), to serine totally abolished the ability of Fas to interact with its ligand, whereas mutants K84S, L90S, E93S (D2), or H126S (D3) showed reduced binding compared with wild-type Fas. Two mutants (K78S and H95S) bound FasL comparably to wild type. Therefore, the binding of FasL involves residues in two domains that correspond to positions critical for ligand binding in other family members (TNFR and CD40) but are conserved between murine and human Fas.

Complementarity determining region 1 (CDR1)- and CDR3-analogous regions in CTLA-4 and CD28 determine the binding to B7-1.

T cell surface receptors CD28 and CTLA-4 are homologous members of the immunoglobulin superfamily (IgSF), each comprising a single V-like extracellular domain. CD28 and CTLA-4 bind to the B7-1 and B7-2 counter-receptors on antigen presenting cells (APCs), thereby triggering a costimulatory pathway important for optimal T cell activation in vitro and in vivo. Soluble forms of CD28 and CTLA-4 in which the V-like extracellular domains were fused to Ig constant domains (CD28Ig and CTLA4Ig), have been used to study their interactions with B7-1 and B7-2, with CTLA4Ig binding B7-1 more strongly than CD28Ig (approximately 20-fold higher avidity). We have now, by site-specific and homologue mutagenesis, identified regions in CTLA4Ig important for strong binding to B7-1. A hexapeptide motif (MYPPPY) in the complementarity determining region 3 (CDR3)-like region is fully conserved in all CD28 and CTLA-4 family members. Alanine scanning mutagenesis through the motif in CTLA4Ig and at selected residues in CD28Ig reduced or abolished binding to B7-1. Chimeric molecules HS4, HS4-A, and HS4-B were constructed in which CDR3-like regions of CTLA-4, COOH-terminally extended to include nonconserved residues, were grafted onto CD28Ig. These homologue mutants showed stronger binding to B7-1 than did CD28Ig. Grafting of the CDR1-like region of CTLA-4, which is not conserved in CD28 and is predicted to be spatially adjacent to CDR3, into HS4 and HS4-A, resulted in chimeric molecules (HS7 and HS8) which bound B7-1 even better. Inclusion of the CDR2-like domain of CTLA-4 into HS7 and HS8 did not further increase binding. Thus, the MYPPPY motifs of CTLA4Ig and CD28Ig are important for their binding to B7-1, but the increased strength of this binding by CTLA4Ig is mediated by nonconserved residues in the CDR1- and CDR3-analogous regions.

Autolysis and inhibition of proteinase K, a subtilisin-related serine proteinase isolated from the fungus Tritirachium album Limber.

The activity of proteinase K (EC 3.4.21.14), a subtilisin-related serine proteinase, was assayed with azoalbumin that showed non-expected behavior in substrate saturation curve because of interaction between albumin molecules. Succinyl-(Ala)n-p-nitroanilide with n = 2 and 3, yielded specific activities of 3.5, 13 units/mg protein, respectively, reflecting a chain length dependence. The influence of peptide chain length on binding to proteinase K was also observed using mono- and dipeptide chloromethyl ketone inhibitors. They showed a maximum inhibition. They showed a maximum inhibition of proteinase K in solution of only about 50% even at a more than 20-fold molar excess. With the above substrates, the Vmax is not affected in presence of 10, 20 and 30% methanol, but the Km differs remarkably, suggesting competitive inhibition. The activity of proteinase K shows a maximum at 37 degrees C, and a temperature profile with more than 80% maximum activity in the range 20-60 degrees C. Autolysis of the enzyme is observed during sample preparation for SDS-gel electrophoresis and at low concentration (0.01 mg/ml) in aqueous solution. It does not occur at higher proteinase K concentrations at or above 1.0 mg/ml, consistent with crystallographic studies.

Comparison of CH1 domains in different classes of murine antibodies.

The CH1 domains of antibodies belonging to the following five murine immunoglobulin (Ig) classes IgG1, IgG2a, IgG2b, IgG3 and IgA have been compared. The IgG CH1 domain structures are, as would be expected, similar overall, but show local conformational variations. When compared with IgG CH1 domain structures, the IgA CH1 domain displays several significant structural differences, which are a consequence of insertions/ deletions and specific structural constraints. In regions of structural differences in the IgG CH1 domains, the spatial correspondence of residues is not reflected by conventional (Kabat) sequence number. Thus the sequence alignment and numbering for CH1 domains has been revised to be consistent with the three-dimensional alignments.

X-ray structure of the uncomplexed anti-tumor antibody BR96 and comparison with its antigen-bound form.

The X-ray structure of the uncomplexed human chimeric Fab' of the anti-tumor antibody BR96 has been determined at 2.6 A resolution. The structure has been compared with Lewis Y antigen-complexed structures of BR96 which were determined previously. The comparison reveals segmental motions and/or conformational rearrangements of three CDR loops (L1, L3, and H2), whereas CDR H3 does not undergo changes upon complexation despite its significant main-chain contacts to the carbohydrate antigen. In light of the uncomplexed chimeric Fab' structure reported here, the previously observed high mobility of the CL:CH1 domains of the complexed chimeric BR96 Fab is rationalized as a "swinging" motion approximately about the axis of the elbow bend.

Crystallization and preliminary X-ray analysis of the monoclonal anti-tumor antibody BR96 and its complex with the Lewis Y determinant.

The monoclonal anti-tumor antibody BR96 binds a tetrasaccharide, Lewis y (Le(y)), in vitro and recognizes a Le(y)-bearing or Le(y)-related tumor-associated antigen in vivo. The Fab of the murine monoclonal antibody, mBR96 (IgG3, kappa), and the Fab' of its human chimera, cBR96 (IgG1, kappa), and their complexes with Le(y) have been screened for crystallization conditions. Crystals suitable for X-ray diffraction have been obtained for uncomplexed cBR96 Fab', cBR96 Fab' in complex with Le(y) and mBR96 Fab in complex with Le(y). The symmetry of the cBR96 Fab' crystals is consistent with space group P2(1)2(1)2, a = 61.1 A; b = 174.3 A; c = 45.6 A; the symmetry of the cBR96 Fab'-Le(y) complex crystals with space group P4(3)2(1)2 (or its enantiomorph), a = b = 82.2 A; c = 167.1 A and the symmetry of the mBR96 Fab-Le(y) complex crystals with space group P2(1)2(1)2(1), a = 69.4 A; b = 84.9 A; c = 86.8 A.

Profiles for the analysis of immunoglobulin sequences: comparison of V gene subgroups.

A format for the structure-oriented analysis of immunoglobulin (Ig) variable region sequences is presented and applied to generate sequence profiles for comparison of heavy- and light-chain subgroups. The profile allows simultaneous evaluation of sequences and structural information and can be used for a number of different applications.

Structure-based modeling of the ligand binding domain of the human cell surface receptor CD23 and comparison of two independently derived molecular models.

CD23, a type II membrane receptor protein, recognizes four different ligands via its extracellular C-type lectin domain: immunoglobulin E (IgE), CD21, and the beta 2-integrins CD11b and CD11c. CD23 specifically interacts in a calcium-dependent manner, "lectin-like" with carbohydrate moieties expressed on CD21 and CD11b/c, but also "lectin-unlike" with protein epitopes on IgE. As a first step in analyzing the multiple binding specificities associated with CD23 in more detail, we report a detailed molecular model of the lectin-like domain of human CD23 (hCD23). The model was built based on information provided by X-ray structures of mannose binding protein (MBP) and E-selectin, both of which are members of the calcium-dependent (C-type) lectin superfamily. Sequence-structure comparisons suggest that hCD23 is structurally more similar to MBP than to E-selectin. The hCD23 model is compared to an independently derived model. Although the CD23-carbohydrate and CD23-protein interactions are both calcium dependent, analysis of the model suggests the presence of distinct binding sites for these ligands.

Knowledge-based model building of proteins: concepts and examples.

We describe how to build protein models from structural templates. Methods to identify structural similarities between proteins in cases of significant, moderate to low, or virtually absent sequence similarity are discussed. The detection and evaluation of structural relationships is emphasized as a central aspect of protein modeling, distinct from the more technical aspects of model building. Computational techniques to generate and complement comparative protein models are also reviewed. Two examples, P-selectin and gp39, are presented to illustrate the derivation of protein model structures and their use in experimental studies.

Immunoglobulin fold characteristics of B7-1 (CD80) and B7-2 (CD86).

B7-1 and B7-2 are expressed on antigen-presenting cells and bind to the CD28 and CTLA-4 receptors on T cells. These interactions trigger a costimulatory pathway that is essential for T-cell activation. B7-1 and B7-2 are members of the immunoglobulin superfamily (IgSF) and, despite sharing common function, have only limited sequence similarity. The B7-1 extracellular region was previously subdivided into 2 IgSF domains, an N-terminal V(ariable)-like domain, followed by a C(onstant)-like domain. We recently reported that the V-like domains of B7-1 and B7-2 share some significant sequence similarities with 3 major histocompatibility complex (MHC)-encoded members of the IgSF. We have now applied inverse folding methodology to assess the compatibility of the B7-1 and B7-2 extracellular region sequences with currently available 3-dimensional structures. In these calculations, the sequences of the N-terminal (V-like) domains in B7-1 and B7-2 were not compatible with known structures, including the IgSF V-set. In contrast, the sequences of the C-like domains were compatible with IgSF C-set structures and were best recognized by the beta 2-microglobulin (beta 2m) domain of MHC Class I. A sequence comparison of the C-like domains in the B7 molecules showed that 11 of 17 rigorously conserved residues in B7-1 and B7-2 are not IgSF C-1 set consensus residues. When mapped onto the corresponding positions of the beta 2m structure, the conserved residues in B7 cluster on the surface, where they may interact with the B7 V-like domain or other molecules.

Molecular model of the N-terminal receptor-binding domain of the human CD6 ligand ALCAM.

CD6-ligand interactions have been implicated in the regulation of T-cell adhesion and activation. CD6 is a member of the scavenger receptor family, whereas its human ligand (ALCAM) belongs to the immunoglobulin superfamily. The extracellular region of ALCAM includes five immunoglobulin-like domains. As a fusion protein, the N-terminal extracellular domain of ALCAM (ALCAMD1) binds specifically to CD6. We report the construction, assessment, and analysis of a molecular model of ALCAMD1. The model defines the CDR-analogous loops, the location of N-linked glycosylation sites, and residues that form the beta-sheet faces of the immunoglobulin-like domain. Predicted structural characteristics of the A'GFCC'C" face of the model are consistent with the presence of monomeric and dimeric forms of ALCAMD1, which has implications for the receptor-ligand interactions.

CD6 recognizes the neural adhesion molecule BEN.

CD6 and its ligand activated leukocyte cell adhesion molecule (ALCAM, CD166) have been detected on various immune cells and in the brain. CD6-ligand interactions have been implicated in the regulation of T cell function. ALCAM shares the same extracellular domain organization and significant sequence homology with the chicken neural adhesion molecule BEN. Although ALCAM's CD6 binding site is only partially conserved in BEN, CD6 specifically binds BEN, albeit with approximately 10-fold lower avidity than ALCAM. Differences in binding avidity are not detected when ALCAM and BEN fusion proteins containing the full-length extracellular regions are tested. Homotypic interactions between full-length forms are likely to account for these observations. The identified cross-species interaction between CD6 and BEN suggests that CD6-ligand interactions are highly conserved.

Classification of mutations in the human CD40 ligand, gp39, that are associated with X-linked hyper IgM syndrome.

The interaction between the T cell activation antigen gp39 and CD40, its receptor CD40 on B cells, plays a critical role in the regulation of humoral immune responses. Using a detailed three-dimensional model of the gp39 extracellular region, we have analyzed 20 mutations in gp39 that were, with one exception, isolated from patients with X-linked hyper IgM (XHIM) syndrome. On the basis of this analysis, the mutations were classified according to their predicted locations and effects. Twelve mutations are thought to compromise the gp39 structure by affecting interactions in hydrophobic core regions or at monomer interfaces, whereas seven others map closely to gp39 residues important for interaction with CD40. The latter mutations may thus, directly or indirectly, interfere with CD40 binding. One naturally occurring mutant whose carrier displays normal immune responses maps to a solvent-exposed position in a loop region of the molecule.

Extending the B7 (CD80) gene family.

B7-1 and B7-2 are members of the immunoglobulin superfamily (IgSF) and important regulators of T cell-mediated immune responses. Despite sharing only limited sequence identity, B7-1 and B7-2 bind common receptors, CD28 and CTLA-4, on T cells and have similar functional properties. We have found that the extracellular V (ariable)-like domains of B7-1 and B7-2 share significant sequence similarities with 3 major histocompatibility complex (MHC)-encoded members of the IgSF: butyrophilin, myelin/oligodendrocyte glycoprotein, and the chicken MHC molecule, B-G. This raises the question whether there is an evolutionary link between the MHC, which encodes molecules regulating the antigen specificity of T lymphocyte responses, and B7 molecules, which co-stimulate these responses in antigen-nonspecific fashion.

Sulfated galactocerebrosides as potential antiinflammatory agents.

Native sulfatides, as well as many sulfated glycolipids, have been shown to avidly bind to the selectin receptors. In vivo, native sulfatides significantly block activity in selectin-dependent inflammatory responses. The fact that nonsulfated galactocerebrosides did not inhibit selectin-mediated adhesion identified a critical role for the anionic sulfate residue. We therefore initiated a program to evaluate the activity of position isomers. This study showed a binding selectivity for the positions 2 and 3 of the sulfate group on the carbohydrate ring as well as enhanced activity for the disulfated analogs. Furthermore, it was discovered that the attachment of lipophilic substituents on the carbohydrate ring was tolerated, consistent with the presence of a lipophilic pocket in the binding activity. This resulted in compounds with a 6-fold increased potency.

Computational techniques for diversity analysis and compound classification.

Molecular similarity and diversity analysis has played a significant role in computer-aided drug discovery for more than a decade. Compound classification methods have also become increasingly important for the design and organization of compound databases and in silico screening. Here we review these related methodologies and discuss selected applications.

Molecular descriptors in chemoinformatics, computational combinatorial chemistry, and virtual screening.

Many contemporary applications in computer-aided drug discovery and chemoinformatics depend on representations of molecules by descriptors that capture their structural characteristics and properties. Such applications include, among others, diversity analysis, library design, and virtual screening. Hundreds of molecular descriptors have been reported in the literature, ranging from simple bulk properties to elaborate three-dimensional formulations and complex molecular fingerprints, which sometimes consist of thousands of bit positions. Knowledge-based selection of descriptors that are suitable for specific applications is an important task in chemoinformatics research. If descriptors are to be selected on rational grounds, rather than guesses or chemical intuition, detailed evaluation of their performance is required. A number of studies have been reported that investigate the performance of molecular descriptors in specific applications and/or introduce novel types of descriptors. Progress made in this area is reviewed here in the context of other computational developments in combinatorial chemistry and compound screening.

Molecular modeling of CD28 and three-dimensional analysis of residue conservation in the CD28/CD152 family.

CD28/CD152-CD80/CD86 receptor-ligand interactions result in costimulatory signals critical for optimal T cell activation. CD28/CD152 and CD80/CD86 are members of the immunoglobulin superfamily (IgSF). Despite common receptor-ligand interactions, both receptor and ligand pairs share only limited sequence identity. A detailed molecular model of the extracellular Ig-like domain of human CD28 was constructed using a combination of different modeling methods. The model was based on the solution structure of CD152 and sequence comparison of the CD28/CD152 family. Assessment of the model revealed good stereochemical quality and sequence-structure compatibility. The CD28 model was used to map surface residues, N-linked glycosylation sites, and to compare residue conservation in CD28 and CD152. The location of N-linked glycosylation sites in CD28/CD152 restricts the surface area available for binding. Rigorous sequence conservation in CD28 and CD152 is limited to core IgSF consensus positions and surface residues implicated in ligand binding. Other surface residues vary greatly in CD28/CD152. Residues critical for ligand binding are surrounded by surface patches conserved only in either CD28 or CD152.

Evaluation of docking strategies for virtual screening of compound databases: cAMP-dependent serine/threonine kinase as an example.

In an effort to establish efficient docking routines for computational screening of compound databases on protein structures, cAMP-dependent protein kinase has been selected as a test case and a variety of docking options and scoring functions were compared. These included rigid-body and flexible docking and scoring based on surface complementarity and/or force field energy. Inhibitors were removed from complex crystal structures and added to compound libraries in their binding conformations and, in addition, deliberately modified conformations. Rigid-body docking and contact scoring well reproduced two of three experimental enzyme-inhibitor complexes. Ligand docking with flexible torsional angles failed to do so but anchored search of some inhibitors converged near to experimental structures, however, only when energy scoring was applied.

Molecular scaffold-based design and comparison of combinatorial libraries focused on the ATP-binding site of protein kinases.

Compound libraries were designed to target specifically the ATP cofactor-binding site in protein kinases by combining knowledge- and diversity-based design elements. A key aspect of the approach is the identification of molecular building blocks or scaffolds that are compatible with the binding site and therefore capture some aspects of target specificity. Scaffolds were selected on the basis of docking calculations and analysis of known inhibitors. We have generated 75 molecular scaffolds and applied different strategies to compute diverse compounds from scaffolds or, alternatively, to screen compound databases for molecules containing these scaffolds. The resulting libraries had a similar degree of molecular diversity, with at most 12% of the compounds being identical. However, their scaffold distributions differed significantly and a small number of scaffolds dominated the majority of compounds in each library.