Molecular modelling and experimental studies of mutation and cell-adhesion sites in the fibronectin type III and whey acidic protein domains of human anosmin-1.

Anosmin-1, the gene product of the KAL gene, is implicated in the pathogenesis of X-linked Kallmann's syndrome. Anosmin-1 protein expression is restricted to the basement membrane and interstitial matrix of tissues affected in this syndrome during development. The anosmin-1 sequence indicates an N-terminal cysteine-rich domain, a whey acidic protein (WAP) domain, four fibronectin type III (FnIII) domains and a C-terminal histidine-rich region, and shows similarity with cell-adhesion molecules, such as neural cell-adhesion molecule, TAG-1 and L1. We investigated the structural and functional significance of three loss-of-function missense mutations of anosmin-1 using comparative modelling of the four FnIII and the WAP domains based on known NMR and crystal structures. Three missense mutation-encoded amino acid substitutions, N267K, E514K and F517L, were mapped to structurally defined positions on the GFCC' beta-sheet face of the first and third FnIII domains. Electrostatic maps demonstrated large basic surfaces containing clusters of conserved predicted heparan sulphate-binding residues adjacent to these mutation sites. To examine these modelling results anosmin-1 was expressed in insect cells. The incorporation of the three mutations into recombinant anosmin-1 had no effect on its secretion. The removal of two dibasic motifs that may constitute potential physiological cleavage sites for anosmin-1 had no effect on cleavage. Peptides based on the anosmin-1 sequences R254--K285 and P504--K527 were then synthesized in order to assess the effect of the three mutations on cellular adhesion, using cell lines that represented potential functional targets of anosmin-1. Peptides (10 microg/ml) incorporating the N267K and E514K substitutions promoted enhanced adhesion to 13.S.1.24 rat olfactory epithelial cells and canine MDCK1 kidney epithelial cells (P<0.01) compared with the wild-type peptides. This result was attributed to the introduction of a lysine residue adjacent to the large basic surfaces. We predict that two of the three missense mutants increase the binding of anosmin-1 to an extracellular target, possibly by enhancing heparan sulphate binding, and that this critically affects the function of anosmin-1.

Structural models for carcinoembryonic antigen and its complex with the single-chain Fv antibody molecule MFE23.

MFE23 is a single chain Fv antibody that has a high affinity for carcinoembryonic antigen (CEA). A full homology model for CEA based on V-type, I-type and C2-type immunoglobulin folds, 28 oligosaccharides and the interdomain angle of CD2 was validated using solution scattering data. The superimposition of the intermolecular contacts observed in our recent crystal structure of MFE23 with the N-terminal domain of CEA permitted the MFE23-CEA complex to be modelled. Good surface and electrostatic complementarity and carbohydrate-unhindered access of MFE23 with the indentation between the first two CEA domains was observed. The model is supported by biochemical data and provides insight on the high affinity of MFE23 for CEA.

Crystal structure of the anti-(carcinoembryonic antigen) single-chain Fv antibody MFE-23 and a model for antigen binding based on intermolecular contacts.

MFE-23 is the first single-chain Fv antibody molecule to be used in patients and is used to target colorectal cancer through its high affinity for carcinoembryonic antigen (CEA), a cell-surface member of the immunoglobulin superfamily. MFE-23 contains an N-terminal variable heavy-chain domain joined by a (Gly(4)Ser)(3) linker to a variable light-chain (V(L)) domain (kappa chain) with an 11-residue C-terminal Myc-tag. Its crystal structure was determined at 2.4 A resolution by molecular replacement with an R(cryst) of 19.0%. Five of the six antigen-binding loops, L1, L2, L3, H1 and H2, conformed to known canonical structures. The sixth loop, H3, displayed a unique structure, with a beta-hairpin loop and a bifurcated apex characterized by a buried Thr residue. In the crystal lattice, two MFE-23 molecules were associated back-to-back in a manner not seen before. The antigen-binding site displayed a large acidic region located mainly within the H2 loop and a large hydrophobic region within the H3 loop. Even though this structure is unliganded within the crystal, there is an unusually large region of contact between the H1, H2 and H3 loops and the beta-sheet of the V(L) domain of an adjacent molecule (strands DEBA) as a result of intermolecular packing. These interactions exhibited remarkably high surface and electrostatic complementarity. Of seven MFE-23 residues predicted to make contact with antigen, five participated in these lattice contacts, and this model for antigen binding is consistent with previously reported site-specific mutagenesis of MFE-23 and its effect on CEA binding.

The Fab and Fc fragments of IgA1 exhibit a different arrangement from that in IgG: a study by X-ray and neutron solution scattering and homology modelling.

Human immunoglobulin A (IgA) is an abundant antibody that mediates immune protection at mucosal surfaces as well as in plasma. The IgA1 isotype contains two four-domain Fab fragments and a four-domain Fc fragment analogous to that in immunoglobulin G (IgG), linked by a glycosylated hinge region made up of 23 amino acid residues from each of the heavy chains. IgA1 also has two 18 residue tailpieces at the C terminus of each heavy chain in the Fc fragment. X-ray scattering using H2O buffers and neutron scattering using 100 % 2H2O buffers were performed on monomeric IgA1 and a recombinant IgA1 that lacks the tailpiece (PTerm455). The radii of gyration RG from Guinier analyses were similar at 6.11-6.20 nm for IgA1 and 5.84-6.16 nm for PTerm455, and their cross-sectional radii of gyration RXS were also similar. The similarity of the RG and RXS values suggests that the tailpiece of IgA1 is not extended outwards in solution. The IgA1 RG values are higher than those for IgG, and the distance distribution function P(r) showed two distinct peaks, whereas a single peak was observed for IgG. Both results show that the hinge of IgA1 results in an extended Fab and Fc arrangement that is different from that in IgG. Automated curve-fit searches constrained by homology models for the Fab and Fc fragments were used to model the experimental IgA1 scattering curves. A translational search to optimise the relative arrangement of the Fab and Fc fragments held in a fixed orientation resembling that in IgG was not successful in fitting the scattering data. A new molecular dynamics curve-fit search method generated IgA1 hinge structures to which the Fab and Fc fragments could be connected in any orientation. A search based on these identified a limited family of IgA1 structures that gave good curve fits to the experimental data. These contained extended hinges of length about 7 nm that positioned the Fab-to-Fab centre-to-centre separation 17 nm apart while keeping the corresponding Fab-to-Fc separation at 9 nm. The resulting extended T-shaped IgA1 structures are distinct from IgG structures previously determined by scattering and crystallography which have Fab-to-Fab and Fab-to-Fc centre-to-centre separations of 7-9 nm and 6-8 nm, respectively. It was concluded that the IgA1 hinge is structurally distinct from that in IgG, and this results in a markedly different antibody structure that may account for a unique immune role of monomeric IgA1 in plasma and mucosa.

Analogy and solution scattering modelling: new structural strategies for the multidomain proteins of complement, cartilage and the immunoglobulin superfamily.

Many immunologically relevant proteins possess multidomain structures. Molecular structures both at the level of the individual domain and that of the intact protein are required for a full appreciation of function and control. Two recently developed structural approaches are reviewed here. Analogy modelling methods are based on the current understanding of many protein structures, and make possible the identification of folds for superfamilies of unknown structures. An integrated multidisciplinary predictive approach has been successfully applied to the von Willebrand factor type A, proteoglycan tandem repeat and factor I/membrane attack complex domains. The available experimental and predictive evidence is assembled in order to identify a known three-dimensional structure related to the unknown one of interest. Neutron and X-ray scattering curve modelling provides information on the full multidomain structure in solution. As scattering curves can be calculated from known atomic structures, the present availability of structures for many domains in conjunction with tight constraints based on these structures and the covalent connections between them results in a small family of allowed best-fit structures for a given scattering curve. The curve-fit procedure can be automated, and whole multidomain structures can be determined to a positional precision of the order of 0.2-1 nm. Such models are informative on the steric accessibility of each domain and their functional activity, and this is illustrated for antibody, cell-surface and complement proteins.

The solution structure of human coagulation factor VIIa in its complex with tissue factor is similar to free factor VIIa: a study of a heterodimeric receptor-ligand complex by X-ray and neutron scattering and computational modeling.

Factor VIIa (FVIIa) is a soluble four-domain plasma serine protease coagulation factor that forms a tight complex with the two extracellular domains of the transmembrane protein tissue factor in the initiating step of blood coagulation. To date, there is no crystal structure for free FVIIa. X-ray and neutron scattering data in solution for free FVIIa and the complex between FVIIa and soluble tissue factor (sTF) had been obtained for comparison with crystal structures of the FVIIa-sTF complex and of free factor IXa (FIXa). The solution structure of free FVIIa as derived from scattering data is consistent with the extended domain arrangement of FVIIa seen in the crystal structure of its complex with sTF, but is incompatible with the bent, less extended domain conformation seen in the FIXa crystal structure. The FVIIa scattering curve is also compatible with a subset of 317 possible extended structures derived from a constrained automated conformational search of 15 625 FVIIa domain models. Thus, the scattering data support extended domain models for FVIIa free in solution. Similar analyses showed that the solution scattering derived and crystal structures of the FVIIa-sTF complex were in good agreement. An automated constrained search for allowed structures for the complex in solution based on scattering curves showed that only a small family of compact models gave good agreement, namely those in which FVIIa and sTF interact closely over a large surface area. The general utility of this approach for structural analysis of heterodimeric complexes in solution is discussed. Analytical ultracentrifugation data and the modeling of these data were consistent with the scattering results. It is concluded that in solution FVIIa has an extended or elongated domain structure, which allows rapid interaction with sTF over a large surface area to form a high-affinity complex.

Molecular structures from low angle X-ray and neutron scattering studies.

Molecular structures can be extracted from solution scattering analyses of multidomain or oligomeric proteins by a new method of constrained automated scattering curve fits. Scattering curves are calculated using a procedure tested by comparisons of crystal structures with experimental X-ray and neutron data. The domains or subunits in the protein of interest are all represented by atomic coordinates in order to provide initial constraints. From this starting model, hundreds or thousands of different possible structures are computed, from each of which a scattering curve is computed. Each model is assessed for steric overlap, radii of gyration and R-factors in order to leave a small family of good fit models that corresponds to the molecular structure of interest. This method avoids the tedium of curve fitting by hand and error limits on the ensuing models can be described. For single multidomain proteins, the key constraint is the correct stereochemical connections between the domains in all the models. Successful applications to determine structures are summarised for the Fab and Fc fragments in immunoglobulin G, the three domain pairs in the Fc subunit of immunoglobulin E and the seven, domains in carcinoembryonic antigen. For oligomeric proteins, the key constraint is provided by symmetry and successful analyses were performed for the association of the monomers of the bacterial amide sensor protein AmiC to form trimers and pentameric serum amyloid P component to form decameric structures. The successful analysis of the heterodimeric complex of tissue factor and factor VIIa required the use of constraints provided from biochemical data. The outcome of these analyses is critically appraised, in particular the biological significance of structures determined by these solution scattering curve fits.

Pentameric and decameric structures in solution of serum amyloid P component by X-ray and neutron scattering and molecular modelling analyses.

Human serum amyloid P component (SAP) is a normal plasma glycoprotein and the precursor of amyloid P component which is a universal constituent of the abnormal tissue deposits in amyloidosis. X-ray and neutron scattering data showed that pentameric or decameric ring structures for SAP in solution are readily distinguished. Further neutron data collection showed that SAP pentamers were reproducibly obtained in the presence of Ca2+ at pH 5.5 or in the presence of methyl 4,6-O-(1-carboxyethylidene)-beta-d-galactopyranoside (MObetaDG) and Ca2+ at pH 6.0 to 8.0, while SAP decamers were obtained in the presence of EDTA between pH 5.5 and 8.0. SAP pentamers have a mean X-ray RG of 3.99(+/-0.11) nm and a mean neutron RG of 3.69(+/-0.12) nm in 100% 2H2O. SAP decamers have a mean X-ray RG of 4.23(+/-0.12) nm and a mean neutron RG of 4.09(+/-0.14) nm in 100% 2H2O. The absorption coefficients of SAP pentamers and decamers differ by 10%. If we infer that the two alpha-helical A-faces are in contact with each other in the SAP decamer, the lack of structural change of the decamer with pH may be explained by the absence of His residues from the A-face of the SAP pentamer, and the change in absorption coefficients is compatible with the presence of Trp residues at this A-face. The rigid ring structure of pentameric SAP provided a test of scattering curves calculated from crystal structures. The only structural unknown is the orientation of the five chemically homogeneous oligosaccharide chains relative to the protein, but extended oligosaccharide structures were found to account for its scattering curve. X-ray scattering curves were best calculated using a hydrated structure, while neutron scattering curves were best calculated using an unhydrated structure. The outcome of these analyses was used to model the structure of decameric SAP. The evaluation of 640 structures for two SAP pentamers brought face-to-face to form SAP decamers gave better curve fits for structures in which the two A-faces were in contact with each other, in which it is likely that the two pentamers were out of alignment by a rotation of 36 degrees and the oligosaccharide chains were extended.

Extended glycoprotein structure of the seven domains in human carcinoembryonic antigen by X-ray and neutron solution scattering and an automated curve fitting procedure: implications for cellular adhesion.

Carcinoembryonic antigen (CEA) is one of the most widely used cell-surface tumour markers for tumour monitoring and for targeting by antibodies. It is heavily glycosylated (50% carbohydrate) and a monomer is constructed from one V-type and six C2-type fold domains of the immunoglobulin superfamily. The solution arrangement at low resolution of the seven domains in CEA cleaved from its membrane anchor was determined by X-ray and neutron scattering. Guinier analyses showed that the X-ray radius of gyration RG of CEA was 8.0 nm. The length of CEA was 27 to 33 nm, and is consistent with an extended arrangement of seven domains. The X-ray cross-sectional radius of gyration RXS was 2.1 nm, and is consistent with extended carbohydrate structures in CEA. The neutron data gave CEA a relative molecular mass of 150,000, in agreement with a value of 152,500 from composition data, and validated the X-ray analyses. The CEA scattering curves were analysed using an automated computer modelling procedure based on the crystal structure of CD2. The V-type and C2-type domains in CD2 were separated, and the C2-type domain was duplicated five times to create a linear seven-domain starting model for CEA. A total of 28 complex-type oligosaccharide chains in extended conformations were added to this model. By fixing the six interdomain orientations to be the same, three-parameter searches of the rotational orientations between the seven domains gave 4056 possible CEA models. The best curve fits from these corresponded to a family of zig-zag models. The long axis of each domain was set at 160(+/-25) degrees relative to its neighbour, and the two perpendicular axes were orientated at 10(+/-30) degrees and -5(+/-35) degrees. Interestingly, the curve fit from this model is within error of that calculated from a CEA model generated directly from the CD2 crystal structure by the superposition of adjacent domains. Zig-zag models of this type imply that the protein face of the GFCC' beta-sheet in neighbouring CEA domains lie on alternate sides of the CEA structure. Such a model has implications for the adhesion interactions between CEA molecules on adjacent cells or for the antibody targeting of CEA.

Properties of the reversible K(+)-competitive inhibitor of the gastric (H+/K+)-ATPase, SK&F 97574. I. In vitro activity.

SK&F 97574 (3-butyryl-4-(2-methylamino)-8-(2-hydroxyethoxy)quinoline), is a potent inhibitor of the (H+/K+)-ATPase in membrane vesicles isolated from porcine gastric mucosa. It inhibits (H+/K+)-ATPase activity in lyophilised vesicles in a kinetically competitive manner with respect to the activating cation, K+, with an inhibition constant (Ki) of 0.46 +/- 0.003 microM. Inhibition of (H+/K+)-ATPase activity is freely reversible. Binding of SK&F 97574 was shown to be mutually exclusive and the previously reported reversible (H+/K+)-ATPase inhibitors, SCH 28080 and MDPQ. Therefore, despite its structural dissimilarity, SK&F 97574 appears to bind to the same lumenal region of the (H+/K+)-ATPase identified as the binding site for these compounds. SK&F 97574 is a weak base (pKa = 6.86), and would therefore be expected to accumulate in the acidic compartment at the lumenal face of the parietal cell. In intact gastric vesicles (which have the lumenal face of the ATPase on the interior), SK&F 97574 inhibited ATP-dependent H(+)-transport with a similar potency to ATPase activity. SK&F 97574 is therefore relatively membrane permeable, and would be predicted to gain access readily to its site of action in vivo. The effect of pH on inhibition of H+/K(+)-ATPase activity by SK&F 97574 is consistent with its being active only in its protonated form. The selectivity of SK&F 97574 for the gastric (H+/K+)-ATPase was tested by examining its ability to inhibit a closely related p-class pump, the (Na+/K+)-ATPase from dog kidney. SK&F 97574 was found to have a 60-fold greater sensitivity for the former enzyme. The (Na+/K+)-ATPase was not inhibited in a K(+)-competitive manner by SK&F 97574, indicating an entirely different, probably nonspecific, mechanism.