Large-scale protein modelling and integration with the SWISS-PROT and SWISS-2DPAGE databases: the example of Escherichia coli

Knowledge-based molecular modelling of proteins has proven useful in many instances, including the rational design of mutagenesis experiments, but it has generally been limited by the availability of expensive computer hardware and software. To overcome these limitations, we developed the SWISS-MODEL server for automated knowledge-based protein modelling. The SWISS-MODEL server uses the Brookhaven Protein Data Bank as a source of structural information and automatically generates protein models for sequences which share significant similarities with at least one protein of known three-dimensional structure. We have now used the software framework of the server to generate large collections of protein models, and established the SWISS-MODEL Repository, a new database for automatically generated and theoretical protein models. This repository is directly integrated with the SWISS-PROT and SWISS-2DPAGE databases through the ExPASy World Wide Web server (URL is http://expasy.hcuge.ch). Here we present an illustration of this process by an application to the Escherichia coli sequences.

Identification of mutations in cystatin B, the gene responsible for the Unverricht-Lundborg type of progressive myoclonus epilepsy (EPM1)

Progressive myoclonus epilepsy (EPM1) is an autosomal recessive disorder, characterized by severe, stimulus-sensitive myoclonus and tonic-clonic seizures. The EPM1 locus was mapped to within 0.3 cM from PFKL in chromosome 21q22.3. The gene for the proteinase inhibitor cystatin B was recently localized in the EPM1 critical region, and mutations were identified in two EPM1 families. We have identified six nucleotide changes in the cystatin B gene of non-Finnish EPM1 families from northern Africa and Europe. The 426G-->C change in exon 1 results in a Gly4Arg substitution and is the first missense mutation described that is associated with EPM1. Molecular modeling predicts that this substitution severely affects the contact of cystatin B with papain. Mutations in the invariant AG dinucleotides of the acceptor sites of introns 1 and 2 probably result in abnormal splicing. A deletion of two nucleotides in exon 3 produces a frameshift and truncates the protein. Therefore, these four mutations are all predicted to impair the production of functional protein. These mutations were found in 7 of the 29 unrelated EPM1 patients analyzed, in homozygosity in 1, and in heterozygosity in the others. The remaining two sequence changes, 431G-->T and 2575A-->G, probably represent polymorphic variants. In addition, a tandem repeat in the 5' UTR (CCCCGCCCCGCG) is present two or three times in normal alleles. It is peculiar that in the majority of patients no mutations exist within the exons and splice sites of the cystatin B gene.

Structure and bioactivity of recombinant human CTAP-III and NAP-2

Connective tissue-activating peptide III (CTAP-III) and neutrophil-activating peptide-2 (NAP-2) are both derived from a common precursor, platelet basic protein (PBP), which is stored in the alpha-granules of platelets and released upon their activation. CTAP-III is an 85-residue peptide which is converted to NAP-2 by enzymic removal of the 15 amino-terminal residues. Both peptides play a role in the early stages of wound healing and inflammation through different activities. We have cloned the cDNA for PBP and expressed constructs coding for the CTAP-III and NAP-2 polypeptides in Escherichia coli. We have purified and renatured these recombinant proteins. The integrity of the recombinant proteins has been ascertained by in vitro bioassays. CTAP-III causes 51% histamine release from the basophilic cell lin KU812 at 10(-7) M, whereas NAP-2 only causes 28% release at the same concentration. In assays on human neutrophils, NAP-2 had an EC50 of 2 x 10(-8) M in chemotaxis, an EC50 of 3 x 10(-8) M for shape change, and could displace IL-8 from neutrophils with a Kd of 7.5 x 10(-9) M. CTAP-III had no activity in these assays. The disulfide bonds have been identified by peptide mapping and sequence analysis, and are in the positions predicted by homology to interleukin-8 and platelet factor 4. Measurement of the molecular mass at physiologic concentrations by gel permeation chromatography has shown that CTAP-III forms predominantly tetramers and dimers, whereas NAP-2 is only dimetric. SDS/PAGE analysis of samples cross-linked with disuccinimidyl suberate support these topologies. We postulate a mechanism for tetramer formation based on the interaction of the amino-terminal extension in CTAP-III involving a helix-helix interaction that could stabilize the association of two CTAP-III dimers.

alpha 1 but not alpha 2 or alpha 3 isoforms of Na,K-ATPase are efficiently phosphorylated in a novel protein kinase C motif

Protein kinase C (PKC) phosphorylates the catalytic alpha 1 subunit of Na,K-ATPase in purified enzyme preparations and in intact cells. Little is known, however, whether all three known alpha isoforms are substrates for PKC and whether direct phosphorylation is implicated in the modulation of the transport activity of the different Na,K-ATPase isozymes. In this study, we investigated the structural requirements for PKC phosphorylation of alpha 1, alpha 2, and alpha 3 isoforms of different species after expression in Xenopus oocytes. By using a combination of site-directed mutagenesis and computer-assisted protein modeling, we characterized a novel Ser-X-His motif which in concert with more distantly located basic residues acts as an efficient substrate for PKC-mediated phosphorylation in the N-terminus of most Na,K-ATPase alpha 1 isoforms. As indicated by controlled proteolysis, alpha 2 isoforms are also phosphorylated in the N-terminus but to a much lower extent than alpha 1 isoforms containing the Ser-X-His motif. Phosphorylation and phosphoamino acid analysis of fusion proteins containing the wild-type or mutant N-terminus of alpha 2 reveal that Thr-Thr-Ser-X-Asn or Thr-Thr-Ala-X-Asn motifs represent weak targets for PKC phosphorylation. Finally, our data suggest that, with the exception of rat alpha 3, all alpha 3 isoforms from other species are not substrates for PKC. On the basis of the phosphorylation efficiency, we may speculate that only alpha 1 but not alpha 2 or alpha 3 isoforms of Na,K-ATPase are likely candidates for regulatory PKC phosphorylation.

CD40lbase: a database of CD40L gene mutations causing X-linked hyper-IgM syndrome

X-linked hyper-IgM syndrome (X-HIM) is an immunodeficiency caused by mutations in the gene encoding the CD40 ligand (CD40L). A database (CD40Lbase) of CD40L mutations has now been established, and the resultant information, together with other mutations reported elsewhere in the literature, is presented here.

Asp-49 is not an absolute prerequisite for the enzymic activity of low-M(r) phospholipases A2: purification, characterization and computer modelling of an enzymically active Ser-49 phospholipase A2, ecarpholin S, from the venom of Echis carinatus sochureki (saw-scaled viper)

Several studies have shown that Asp-49 is the residue that controls calcium binding in, and so plays a critical role in the calcium-mediated activation of, low-M(r) group I-III phospholipases A2 (PLA2s). The present paper provides experimental evidence that Asp-49 is not an absolute prerequisite for the enzymic activity of PLA2s, and that proteins with amino acid(s) other than Asp at position 49 can exhibit significant phospholipase activity. The purification, complete amino acid sequence and characterization of ecarpholin S, a PLA2 from Echis carinatus sochureki (saw-scaled viper) venom, is described. This single-chain, 122-amino-acid, basic (pI 7.9) protein is a group II PLA2. Although Asp-49 is replaced by Ser and Tyr-28 by Phe (both of these positions being involved in the Ca(2+)-binding site of PLA2s), the lipolysis of soybean phosphatidylcholine and egg yolk in the presence of 10 mM CaCl2 was 1.5 times and 2.9 times greater respectively with ecarpholin S than with recombinant human group II PLA2. The Ca(2+)-dependencies of the enzymic activities of ecarpholin S and rPLA2 were found to be similar. Ecarpholin S added to washed platelets induced aggregation; the presence of Ca2+ was a prerequisite for this platelet-aggregating effect. Computer modelling of the Ca(2+)-binding site of Ser-49 PLA2 compared with the Asp-49 and Lys-49 forms, for which crystallographic data exist, shows that the Ca(2+)-binding site is sterically blocked by Lys-49 but not by Ser-49; in the latter, the Ser hydroxy group may replace the Asp carboxylate in stabilization of Ca2+ binding. Sequence comparisons of ecarpholin S and other low-M(r) PLA2s predicts the presence of a Ser-49 group in the protein family of low-M(r) PLA2s that is distinct from the Asp-49 and Lys-49 groups.

The Molecular Basis of the Chemokine/Chemokine Receptor Interaction-Scope for Design of Chemokine Antagonists

Chemokines are a family of small proteins that are present in a variety of inflammatory conditions and have been shown to activate and recruit a wide variety of cell types. They bind to a family of seven transmembrane G-protein-coupled receptors. Models for the interaction of the chemokines with their receptors suggest a two-step mechanism. Initially, the main body of the chemokine interacts with the outside of the receptor (Site 1), and this interaction directs receptor selectivity. Subsequently, the flexible amino-terminus of the chemokine interacts with the receptor core (Site 2) to initiate the signaling response. Mutagenesis studies of IL-8, the archetypal CXC chemokine, show that altering the protein on the third beta-sheet can change the receptor selectivity from that of a CXC chemokine and introduce CC chemokine activity-confirming the role of this region in Site 1. Mutagenesis studies of the amino-terminal region of IL-8 showed that a tripeptide, ELR, was essential for the interaction with Site 2. We have shown, using synthetic peptides and site-directed mutagenesis, that the amino-terminus of RANTES is important in the signaling response (Site 2). Mutations that alter only the interaction with Site 2 are capable of binding the receptor and not signaling and are therefore potential antagonists. Such antagonists have now been made by several groups, for a number of the chemokine receptors, and are active at nanomolar concentrations. These can now be used to test the hypothesis that antagonism of chemokine receptors will lead to a reduction in inflammation in vivo.

The carboxy-terminal region of human interleukin-5 is essential for maintenance of tertiary structure but not for dimerization

The C-terminal region of interleukin-5 has previously been suggested to be important for biological activity [Mackenzie et al., (1991), Mol. Immunol. 28, 155-158; Kodama et al. (1991), Biochem. Biophys. Res. Commun. 178, 514-519]. We have investigated this region by making a series of truncation mutants. The proteins were expressed in Escherichia coli, purified from inclusion bodies, and were able to refold with the disulfide homodimeric topology typical of interleukin-5. Analysis of the truncated carboxy-terminal proteins in an interleukin-5-dependent proliferation assay on TF-1 cells showed a rapid loss of activity as the C-terminal was shortened by more than two amino acids. This loss of biological activity correlated with a drop in binding affinity to both the alpha chain of the receptor and the high-affinity complex consisting of the alpha and beta subunits. Analysis of the proteins by 1H-NMR showed that the truncated mutants have higher exchange rates with solvent, indicating a less rigid structure. The carboxy-terminal region is therefore necessary to maintain the stability of the four-helix bundle and to orient correctly the important residues of the fourth helix. Inspection of the structure determined by X-ray crystallography shows that Trp-110 acts as the major residue in anchoring the fourth helix.

Increased p34cdc2-dependent kinase activity during apoptosis: a possible activation mechanism of DNase I leading to DNA breakdown

Cells undergoing apoptosis typically exhibit distinctive morphological characteristics. Early events include the rounding up of the cell, chromatin condensation, nuclear membrane breakdown and blebbing of the cellular membrane. Strikingly similar changes take place in the cell cycle progression, at the entry into mitosis, suggesting a link between mitosis and apoptosis. Here we show that expression of active p34cdc2 at inappropriate phases during the cell cycle leads to morphological changes reminiscent of apoptosis, including DNA degradation. Cells cotransfected with the active mutant of p34cdc2 and DNase I displayed degraded DNA, which was absent in p34cdc2 wild-type and DNase I-transfected cells, in spite of similar DNase activities. Upon induction of apoptosis in thymocytes, transient p34cdc2 activation was detected prior to lamina breakdown and nuclease activation. P34cdc2 activation was also observed during APO-1 (Fas/CD95)-induced apoptosis in a B lymphoblastoma cell line. Our results suggest that unscheduled activation of p34cdc2 may participate in the initiation of the typical apoptotic phenotype.

Selectivity and antagonism of chemokine receptors

The chemokine superfamily can be subdivided into two groups based on their amino terminal cysteine spacing. The CXC chemokines are primarily involved in neutrophil-mediated inflammation and, so far, two human receptors have been cloned. The CC chemokines tend to be involved in chronic inflammation, and recently we have cloned a fourth leukocyte receptor for this group of ligands. Understanding what makes one receptor bind its range of agonists is important if we are to develop potent selective antagonist. We have started to investigate the molecular basis of this receptor selectivity by looking at why CC chemokines do not bind to the CXC receptors in several ways. First, we looked at the role of the three-dimensional structure of the ligand, and have solved the three dimensional structure of RANTES using nuclear magnetic resonance spectroscopy. The structure is similar to that already determined for the CC chemokine macrophage inflammatory protein-1 beta, and it has a completely different dimer interface to that of the CXC chemokine interleukin-8 (IL-8). However, the monomer structures of all the chemokines are very similar, and at physiological concentrations the proteins are likely to be monomeric. Second, by examining all the known CC and CXC chemokines, we have found a region that differs between the two subfamilies. Mutations of one of the residues in this region, Leu-25 in IL-8, to tyrosine (which is conserved at this position in CC chemokines) enables the mutant IL-8 to bind CC chemokine receptor-1 (CC-CKR-1) and introduces monocyte chemoattractant activity. Using other mutations in this region, we can show a direct interaction with the N-terminus of CC-CKR-1. Third, we have found that modification of the amino terminus of RANTES by addition of one amino acid makes it into an antagonist with nanomolar potency. Taken together, this data suggests a two-site model for receptor activation and for selectivity between CC and CXC chemokines, with an initial receptor contact provided by the main body of the chemokine, and activation provided by the amino terminal region.

Automated modelling of the transmembrane region of G-protein coupled receptor by Swiss-model

Molecular modelling of the transmembrane helices of G-protein coupled receptors is an increasingly used method to identify the possible three-dimensional environment of key residues. Thereby site-directed mutagenesis experiments, aimed at the understanding of the receptor-ligand interactions, can be designed in a rational way. The modelling methods are however not generally available to experimentalists, and often require expensive software and hardware. To overcome these limitations, we have constructed a World Wide Web server for the automated protein modelling of user-defined transmembrane helices. The service is freely available at this address: http:/(/)expasy.hcuge.ch/swissmod/SWISS-MODEL.++ +html.

Characterization of the non-functional Fas ligand of gld mice

Mice homozygous for either the gld or lpr mutation develop autoimmune diseases and progressive lymphadenopathy. The lpr mutation is characterized by the absence of functional Fas, whereas gld mice exhibit an inactive FasL due to a point mutation proximal to the extracellular C-terminus. The structural repercussions of this amino acid substitution remain unknown. Here we report that FasL is expressed at similar levels on the surface of activated T lymphocytes from gld and wild-type mice. Using a polyclonal anti-FasL antibody, indistinguishable amounts of a 40 kDa protein are detected in both gld and wild-type splenocytes. The molecular model of FasL, based on the known structure of TNF-alpha, predicts that the Phe --> Leu gld mutation is located at the protomer interface which is close to the FasR interaction site. We conclude that the gld mutation allows normal FasL biosynthesis, surface expression and oligomerization, but induces structural alterations to the Fas binding region leading to the phenotypic changes observed.

Role of a hydrophobic pocket of the human Y1 neuropeptide Y receptor in ligand binding

We are investigating the nature of the chemical interactions between the neuropeptide Y (NPY) and its cell surface receptor (Y1). A previous study involving site-directed mutagenesis and computer-aided modelling (Walker et al., 1994) suggested that the C-terminal Tyr36 of NPY, known to be a key residue for receptor binding, might dock at a pocket formed by hydrophobic amino acids of transmembrane domains (TM) 1, 2, 6 and 7 of the Y1 receptor. To investigate which residues were required for ligand binding, we mutated the sequences encoding F41, L43, F96, Y100, F286 and H298 of the human Y1 receptor. The mutant cDNAs were transiently expressed in Hela cells and the ability of the encoded proteins to bind NPY was evaluated. Replacing F41, L43 or F96 with alanines had no effect on NPY binding. On the contrary, Y100, F286 and H298 appeared to be residues critical for ligand binding. In particular, the removal of the hydroxyl group of Y100 (Tyr100-->Phe100 mutation) yielded a protein devoid of affinity for the ligand. The level of expression and the presence on the cell surface of mutants lacking NPY binding activity was assessed by immunological techniques. In addition, we tested the ability of synthetic analogues of neuropeptide Y with substitutions at position 36 to bind to the Y1 receptor. To get spatial insight into the relative positions of the above mentioned residues we constructed a molecular model of the interaction between NPY:Y36 and the elements of the hydrophobic pocket surrounding this residue.

Comparative molecular modelling of the Fas-ligand and other members of the TNF family

A number of proteins with significant similarity to the tumour necrosis factor (TNF) have been identified over the last years. Upon interaction with their cognate receptor (members of the TNF-receptor family), all members of this protein family induce either cell death or proliferation/differentiation of the receptor-bearing cells. One of the last identified members of the TNF family is the apoptosis-inducing ligand of the Fas-receptor, termed Fas-ligand (FasL). Here we report the cloning and sequencing of the mouse cDNA for the FasL. Using knowledge-based protein modelling, we demonstrate that all members of the TNF family form trimeric complexes, and define the residues located at the subunit interfaces. The resulting structurally corrected multiple sequence alignment allows the identification of residues potentially involved in receptor recognition, and should help design mutagenesis experiments for structure-function relationship studies.

Structural analysis of TCR-ligand interactions studied on H-2Kd-restricted cloned CTL specific for a photoreactive peptide derivative

To study the interaction of the TCR with its ligand, the complex of a MHC molecule and an antigenic peptide, we modified a TCR contact residue of a H-2Kd-restricted antigenic peptide with photoreactive 4-azidobenzoic acid. The photoreactive group was a critical component of the epitope recognized by CTL clones derived from mice immunized with such a peptide derivative. The majority of these clones expressed V beta 1-encoded beta chains that were paired with J alpha TA28-encoded alpha chains. For one of these TCR, the photoaffinity labeled sites were mapped on the alpha chain as a J alpha TA28-encoded tryptophan and on the beta chain as a residue of the C' strand of V beta 1. Molecular modeling of this TCR suggested the presence of a hydrophobic pocket that harbors this tryptophan as well as a tyrosine on the C' strand of V beta 1 between which the photoreactive side chain inserts. It is concluded that this avid binding principle may account for the preferential selection of V beta 1 and J alpha TA28-encoded TCR.

Identification of key charged residues of human interleukin-5 in receptor binding and cellular activation

Interleukin-5 (IL-5) is a cytokine that plays a major role in the differentiation and activation of eosinophils. In order to identify which charged residues of human IL-5 are important in binding to its receptor and subsequent cellular activation, we have systematically replaced all of the clusters of charged amino acids with alanine residues. The mutants have been expressed in Escherichia coli, renatured, and purified. They were assayed for ability to cause proliferation of the erythroleukaemic cell line TF-1 and the up-regulation of eosinophil adhesion to ICAM-1. In addition, we studied receptor binding using either immobilized recombinant IL-5 receptor alpha-chain or the alpha/beta-receptor complex expressed on TF-1 cells. The key charged residue involved in binding to the beta-chain of the receptor is Glu-12. This residue is in an identical position to those previously identified in IL-3 and granulocyte-macrophage colony-stimulating factor (GM-CSF) involved in binding to the receptor beta-chain. The alpha-chain binding site is shown to involve the side chains Arg-90 and Glu-109, located in the second beta sheet and after the end of the fourth helix, respectively. It is unique to IL-5 and does not occur in IL-3 or GM-CSF. Understanding the topology of the interaction of IL-5 with its receptor chains will help in the search for rationally designed antagonists of IL-5 function.

Genomic organisation and expression of mouse deoxyribonuclease I

Deoxyribonuclease I (DNase I) has recently been implicated in cell death by apoptosis, a process which is frequently accompanied by chromatin DNA degradation. Despite extensive studies on DNase I, its genomic organisation remained unknown. Here we report for the first time on the intron-exon structure of the DNase I gene. The coding region of mouse DNase I is composed of eight introns and eight exons, spanning 2315 base pairs. The deduced protein sequence is 91.5% identical to its rat counterpart, but does not carry the two mutations (Glu13 to D and V67 to I) responsible for the decrease in actin-binding of rat DNase I. The enzymatic activity of mouse DNase I is found in striated muscle, kidney, intestine, liver, lymphnodes, but not in the heart, spleen or pancreas.

Mutational analysis of polymeric immunoglobulin receptor/ligand interactions. Evidence for the involvement of multiple complementarity determining region (CDR)-like loops in receptor domain I

The polymeric Ig receptor (pIgR) mediates the transport of IgA and IgM across a variety of mucosal epithelia. The ectodomain of this receptor consists of five immunoglobulin-like domains (I-V), the first four being structurally similar to immunoglobulin variable regions, and the fifth to Ig constant regions. This study examines the structural features of the pIgR that participate in binding of the ligand, dimeric IgA (dIgA). Recent evidence suggests that a highly conserved region of the first Ig-like domain (domain I) may be important in this process (Bakos, M.A., Kurosky, A., and Goldblum, R. M. (1991) J. Immunol. 147, 3419-3426). In support of this hypothesis, molecular modeling of domain I places this conserved region in an exposed loop analogous to the CDR1 loop of Ig, suggesting that interactions between dIgA and the pIgR may be similar to those between antibodies and their cognate antigens. To test this hypothesis directly, we performed a mutagenic analysis of all three CDR-like loops in domain I of the pIgR. We found that point mutations in multiple residues of CDR1 produced effects on IgA binding ranging from minimal (90% of control) to profound (7%). In addition, we replaced regions corresponding to the CDR2 and CDR3 loops of domain I with their counterparts from domain II (which does not bind IgA), which in both cases resulted in complete abrogation of IgA binding. Taken together, these data suggest that each of the three CDR-like loops of domain I of the rabbit pIgR participates in the binding of dimeric IgA.

About the involvement of deoxyribonuclease I in apoptosis

Cell death by apoptosis is involved in a large variety of developmental events and physiological processes requiring a reduction in cell count. Nuclear collapse, one of the first visible changes denoting irreversible commitment to cell death by apoptosis, is frequently accompanied by chromatin degradation into nucleosome-sized fragments of multiples thereof. The identity of the endonuclease responsible for this DNA digestion has attracted some interest in recent years and several candidate endonucleases have been proposed. The scope of this article is to summarise the present knowledge about deoxyribonuclease I, one of the candidate enzymes.