Adhesion mechanism of human beta 2-glycoprotein I to phospholipids based on its crystal structure
1999; Springer Nature; Volume: 18; Issue: 19 Linguagem: Inglês
10.1093/emboj/18.19.5166
ISSN1460-2075
Autores Tópico(s)Systemic Lupus Erythematosus Research
ResumoArticle1 October 1999free access Adhesion mechanism of human β2-glycoprotein I to phospholipids based on its crystal structure Barend Bouma Barend Bouma Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Philip G. de Groot Philip G. de Groot Haemostasis and Thrombosis Laboratory, University Medical Center Utrecht, PO Box 85500, 3508 GA, Utrecht, The Netherlands Search for more papers by this author Jean M.H. van den Elsen Jean M.H. van den Elsen Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Raimond B.G. Ravelli Raimond B.G. Ravelli EMBL Grenoble Outstation, 6 Rue Jules Horowitz, B.P. 156, 38042 Grenoble, Cedex 9, France Search for more papers by this author Arie Schouten Arie Schouten Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Marleen J.A. Simmelink Marleen J.A. Simmelink Haemostasis and Thrombosis Laboratory, University Medical Center Utrecht, PO Box 85500, 3508 GA, Utrecht, The Netherlands Search for more papers by this author Ronald H.W.M. Derksen Ronald H.W.M. Derksen Rheumatology and Clinical Immunology, University Medical Center Utrecht, PO Box 85500, 3508 GA, Utrecht, The Netherlands Search for more papers by this author Jan Kroon Jan Kroon Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Piet Gros Corresponding Author Piet Gros Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Barend Bouma Barend Bouma Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Philip G. de Groot Philip G. de Groot Haemostasis and Thrombosis Laboratory, University Medical Center Utrecht, PO Box 85500, 3508 GA, Utrecht, The Netherlands Search for more papers by this author Jean M.H. van den Elsen Jean M.H. van den Elsen Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Raimond B.G. Ravelli Raimond B.G. Ravelli EMBL Grenoble Outstation, 6 Rue Jules Horowitz, B.P. 156, 38042 Grenoble, Cedex 9, France Search for more papers by this author Arie Schouten Arie Schouten Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Marleen J.A. Simmelink Marleen J.A. Simmelink Haemostasis and Thrombosis Laboratory, University Medical Center Utrecht, PO Box 85500, 3508 GA, Utrecht, The Netherlands Search for more papers by this author Ronald H.W.M. Derksen Ronald H.W.M. Derksen Rheumatology and Clinical Immunology, University Medical Center Utrecht, PO Box 85500, 3508 GA, Utrecht, The Netherlands Search for more papers by this author Jan Kroon Jan Kroon Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Piet Gros Corresponding Author Piet Gros Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands Search for more papers by this author Author Information Barend Bouma1, Philip G. de Groot2, Jean M.H. van den Elsen1, Raimond B.G. Ravelli3, Arie Schouten1, Marleen J.A. Simmelink2, Ronald H.W.M. Derksen4, Jan Kroon1 and Piet Gros 1 1Department of Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH, Utrecht, The Netherlands 2Haemostasis and Thrombosis Laboratory, University Medical Center Utrecht, PO Box 85500, 3508 GA, Utrecht, The Netherlands 3EMBL Grenoble Outstation, 6 Rue Jules Horowitz, B.P. 156, 38042 Grenoble, Cedex 9, France 4Rheumatology and Clinical Immunology, University Medical Center Utrecht, PO Box 85500, 3508 GA, Utrecht, The Netherlands *Corresponding author. E-mail: [email protected] The EMBO Journal (1999)18:5166-5174https://doi.org/10.1093/emboj/18.19.5166 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Human β2-glycoprotein I is a heavily glycosylated five-domain plasma membrane-adhesion protein, which has been implicated in blood coagulation and clearance of apoptotic bodies from the circulation. It is also the key antigen in the autoimmune disease anti-phospholipid syndrome. The crystal structure of β2-glycoprotein I isolated from human plasma reveals an elongated fish-hook-like arrangement of the globular short consensus repeat domains. Half of the C-terminal fifth domain deviates strongly from the standard fold, as observed in domains one to four. This aberrant half forms a specific phospholipid-binding site. A large patch of 14 positively charged residues provides electrostatic interactions with anionic phospholipid headgroups and an exposed membrane-insertion loop yields specificity for lipid layers. The observed spatial arrangement of the five domains suggests a functional partitioning of protein adhesion and membrane adhesion over the N- and C-terminal domains, respectively, separated by glycosylated bridging domains. Coordinates are in the Protein Data Bank (accession No. 1QUB). Introduction Human β2-glycoprotein I (β2GPI), also known as apolipoprotein H, is a membrane-adhesion glycoprotein present in blood plasma at a concentration of ∼150-300 μg/ml (Willems et al., 1996). It consists of 326 amino acid residues (Lozier et al., 1984; Kristensen et al., 1991) with ∼20% w/w carbohydrates attached. β2GPI is the key antigen in the autoimmune disease anti-phospholipid syndrome (APS), defined by thrombo-embolic complications and the presence of anti-phospholipid autoantibodies (aPLs) in the blood. β2GPI has been indicated as a natural anticoagulant (Brighton et al., 1996; Mori et al., 1996) and has a role in the clearance of apoptotic bodies from the circulation (Price et al., 1996; Balasubramanian and Schroit, 1998; Manfredi et al., 1998). β2GPI belongs to a super-family of proteins characterized by repeating stretches of ∼60 amino acid residues, each with a set of 16 conserved residues and two fully conserved disulfide bonds. More than 50 mammalian, mainly complement, proteins belong to this family including CR2, Factor H and CR1, which contain up to 15, 20 and 30 of these consecutive repeating stretches, respectively (Bork et al., 1996). These repeating units have been termed short consensus repeat (SCR), complement control protein or Sushi domains and are, in many cases, involved in protein-protein interactions, with typically two to four consecutive domains forming an interaction site (Iwata et al., 1995; Sharma et al., 1996; Blom et al., 1999; Casasnovas et al., 1999; Van der Poel et al., 1999). NMR structures of two single SCR domains, domains 5 and 16 of human Factor H (HFH) (Norman et al., 1991; Barlow et al., 1992), and two SCR domains in tandem, domains 15-16 of HFH and domains 3-4 of vaccinia virus complement control protein (VCP) (Barlow et al., 1993; Wiles et al., 1997), have been determined. Recently, the crystal structure of the N-terminal two SCR domains of CD46 has been published (Casasnovas et al., 1999). β2GPI consists of five of these SCR domains. The first four domains are regular SCR domains with respect to their amino acid sequences. The fifth C-terminal domain contains a six-residue insertion and a 19-residue C-terminal extension, which is C-terminally cross-linked by a disulfide bond. This aberrant domain is responsible for adhesion to acidic phospholipids (Steinkasserer et al., 1992; Hunt and Krilis, 1994; Sheng et al., 1996). Adhesion to membranes is very likely to be an essential aspect of β2GPI that is common to the observed effects of β2GPI in APS, coagulation and apoptosis. The autoimmune disorder APS is characterized by the presence of a group of heterogeneous autoantibodies in blood plasma and the occurrence of thrombo-embolic complications in both the arterial and venous vasculature of patients (Bick and Baker, 1994). The symptoms appear predominantly in women aged 25-35 years. A particular problem in understanding the pathophysiology of aPLs has been the apparent contradiction between the in vivo observed thrombosis and the in vitro observed prolonged coagulation time (Brighton et al., 1996). An important observation has been that the real antigen for aPLs is the plasma-protein β2GPI (Galli et al., 1990) and not phospholipids (Roubey, 1996). The affinity of β2GPI for acidic phospholipids increases strongly in the presence of aPLs, which is explained by the formation of divalent (β2GPI)2-aPL complexes (Willems et al., 1996). Binding of these complexes to phospholipids interferes with the binding of other phospholipid-binding proteins in plasma, such as coagulation proteins, resulting in the in vitro prolongation of coagulation (Takeya et al., 1997). In vivo, the (β2GPI)2-aPL complexes possibly inhibit the anticoagulant activity of protein C at phospholipid surfaces, explaining the thrombo-embolic risk (Esmon et al., 1997). We have determined the crystal structure of the glycosylated five-domain human β2GPI purified from blood plasma to a resolution of 2.7 Å. The structure aids the characterization of the epitopes for the heterogeneous pool of aPLs and yields insights into the spatial arrangement and functional partitioning over the multiple consecutive SCR domains and the mechanism of binding to acidic phospholipids. Results Structure determination Crystals of β2GPI grew in the orthorhombic space group C2221. Significant non-isomorphism was observed between crystals, as indicated by an Riso of 20.9% between data sets Native I and Native II (Table I). Structure determination of β2GPI with the multiple-isomorphous replacement method using anomalous scattering (MIRAS) revealed one β2GPI molecule in the asymmetric unit with a remarkably high solvent content of 86% and a large Vm value of 8.5 Å3/Da (Table II, Figure 1). The initial map using Native I and derivative set I was of low quality due to poor phasing statistics (Table II) and was improved dramatically by solvent flattening using a solvent fraction of 70%. A first model was built at 3.75 Å resolution using the NMR structure of the 15th SCR domain of HFH (Barlow et al., 1993). Rigid-body refinement of this initial model against data set Native II yielded a decrease in R-factor from 52.6 to 48.0% and in Rfree from 52.0 to 47.8%. Phase information to 2.7 Å resolution obtained at a later stage (derivative sets II) was used to validate and correct the model (Figure 2). Refinement used the maximum-likelihood method and a bulk-solvent correction (see Materials and methods). Electron density corresponding to residues Ser311-Lys317 is not visible in the final 2Fo−Fc map and, therefore, these residues have not been included in the final model. Seven carbohydrate units are identified in the electron density maps at the four N-glycosylation sites. The final structure is refined to 2.7 Å resolution with an R-factor of 24.9% and an Rfree of 26.9% and displays good stereochemistry (Table II). Coordinates and structure factors have been deposited with the Protein Data Bank (accession No. 1QUB). Figure 1.Structural representations of human blood plasma β2GPI revealing the extended chain of the five SCR domains. (A) Ribbon drawing of β2GPI with consecutive domains labelled I-V. N-linked glycans, as well as the position of the putative O-linked glycan, Thr130, are indicated by a ball-and-stick model. β-strands are shown in red and helices in green. (B) Topology diagram of β2GPI. The central β-sheets of all five domains are labelled B2(-B2″)-B3-B4(-B5), the N- and C-terminal β-sheets are labelled B1′-B2′ and B4′-B5′, the α-helix and the 3/10 helix are denoted A1 and A2 and numbers of residues delimiting secondary structure elements are given. Disulfide bonds are indicated with dashed lines. The positions of N-glycosylation are given by hexagons; a diamond indicates the putative O-glycan. Horizontal dashed lines mark domain boundaries. (C) Ribbon representation of domain III of β2GPI with labelled secondary structure elements. The two fully conserved disulfide bonds are shown in yellow. (D) Ribbon representation of domain V of β2GPI with labelled secondary structure elements. The three disulfide bonds are indicated with yellow lines. The aberrant face, which contains the membrane-binding site, is located on the right-hand side. Download figure Download PowerPoint Figure 2.Electron density map near Trp 53 calculated with phases from the refined model. The 2Fo−Fc map at 2.7 Å resolution is contoured at 1σ. Download figure Download PowerPoint Table 1. Crystal characteristics showing non-isomorphism Crystal Res. (Å) m (°)a a, b, c (Å) ΔV/Vb Tc Native I 29-3.75 1.0 161.53 163.73 114.99 0.0 120 Native II 40-2.7 1.0 161.17 166.49 114.51 +1.0 100 K2OsO4 (I) 39-3.0 0.6 160.95 161.33 114.65 −2.1 100 Na3IrCl6 (I) 38-3.1 0.5 160.55 163.11 113.98 −1.9 100 K2PtCl6 (I) 39-3.2 0.8 161.38 161.86 113.98 −2.1 100 K2OsO4 (II) 40-2.7 0.4 160.86 166.26 115.35 +1.0 100 K2PtCl6 (II) 40-2.9 0.5 162.43 165.42 114.76 +1.0 100 a Mosaicity. b Native I is taken as a reference, cell volume differences are given as a percentage and are mainly caused by changes in the b-axis. c Temperature in K during data collection. Table 2. Structure determination statistics Diffraction data statisticsa Crystal Resolution (Å) Redundancy No. unique reflections I / σ(I) Completeness (%) Rmerge (%)b Native I 29-3.75 4.0 15 447 5.2 (4.1) 97.8 (97.7) 14.7 (31.1) Native II 40-2.7 3.8 42 494 5.8 (2.0) 99.8 (99.6) 8.9 (36.3) K2OsO4 (I) 39-3.0 3.3 30 431 8.8 (2.6) 96.6 (88.4) 7.4 (28.9) Na3IrCl6 (I) 38-3.1 3.7 23 790 9.0 (4.4) 83.8 (77.3) 6.9 (20.0) K2PtCl6 (I) 39-3.2 3.2 23 516 8.2 (2.6) 93.6 (67.0) 7.8 (25.1) K2OsO4 (II) 40-2.7 8.9 42 498 9.7 (3.3) 100 (99.8) 7.4 (40.3) K2PtCl6 (II) 40-2.9 10.4 34 609 12.0 (5.4) 99.5 (97.4) 8.9 (33.1) Phasing statistics Derivative Risoc Ranod No. sites Phasing powere Rcullisf FOMg Centric Acentric K2OsO4 (I) 0.230 0.050 3 0.65 0.73 0.67 0.43 Na3IrCl6 (I) 0.130 0.031 2 0.26 0.35 0.81 0.43 K2PtCl6 (I) 0.181 0.047 1 0.40 0.40 0.79 0.43 K2OsO4 (II) 0.207 0.063 4 0.68 0.76 0.70 0.34 K2PtCl6 (II) 0.213 0.056 1 0.47 0.53 0.76 0.34 Refinement statistics Resolution 40-2.7 Å R-factor/Rfreeh 0.249/0.269 r.m.s.d. bond distances 0.019 Å r.m.s.d. angles 1.92° Average B-factor 49 Å2 No. non-hydrogen atoms 2608 No. protein residues 319 No. sugar moieties 7 No. solvent molecules 32 a Numbers in parentheses indicate statistics for highest resolution shells. b Rmerge = ΣhΣi|Ihi− |/ΣhΣi|Ihi| c Riso = Σ| Fph − Fp | / | Σ(Fph + Fp) d Rano = Σ| F(+) − F(−) | / Σ(F(+) + F(−)) e Phasing power is the r.m.s. value of the heavy atom structure factor amplitude divided by the r.m.s. residual lack of closure. f Rcullis is the mean residual lack of closure error divided by the isomorphous difference. g FOM, figure of merit. h A 5% test set of reflections was used for calculation of Rfree. Structure description The structure of β2GPI shows an extended chain of five SCR domains with an overall fish-hook-like appearance with dimensions of 130 Å (vertical direction in Figure 1A), 85 Å (horizontal) and 130 Å from the N- to C-terminal extremity. The distance from the N- to C-terminal end along the curve of the molecule is ∼190 Å. The fish-hook shape is mostly flat, but is bent slightly around domain III. The elongated structure contradicts a hypothetical model of Koike et al. (1998), who described the molecule as being folded into a compact particle. Domains I-IV of β2GPI have common SCR folds (Figure 1B). They consist of a central anti-parallel β-sheet, comprising strands B2-B3-B4, with two extended loops typically flanked by short two-stranded anti-parallel β-sheets, B1′-B2′ and B4′-B5′, at the N- and C-terminal side (Figure 1C). Disulfide bridges located at opposite ends of a domain cross-link the short-flanking β-sheets with the central β-sheet (B1′-B4 and B3-B5′). Domain II has slightly deviating structural elements; its central sheet is extended by one more anti-parallel strand denoted B5 and residues at the position of strand B1′-B2′ do not adopt a β-sheet conformation. These four domains of β2GPI have a sequence homology ranging from 24 to 45% (Corpet, 1988) and superpose within 1.2-1.9 Å root-mean-square (r.m.s.) coordinate difference (see Materials and methods). A similar range of homology, 21-41%, and a similar range of r.m.s. coordinate difference, 1.4-2.3 Å, are observed when comparing β2GPI I-IV with SCR domains available in the Protein Data Bank, HFH-15, 16 (Barlow et al., 1993) and VCP-3, 4 (Wiles et al., 1997). This indicates that the SCR fold is very well conserved. The largest differences with respect to amino acid length and structure are observed in the B2-B2′ loops, referred to as ‘hyper-variable’ by Barlow et al. (1993). The fifth C-terminal domain deviates significantly from the common SCR fold (Figure 1D). Similar to the SCR fold, it has the central anti-parallel β-sheet encompassing β-strands B2, B3 and B4 and the two disulfide bonds common to all SCR domains. In domain V, strands B3 and B4 of the central sheet are extended and are part of a larger and strongly twisted anti-parallel β-sheet. Domain V has an insertion of six residues in the region of the hyper-variable loop. This region now forms an additional β-strand (B2″), which participates in the central β-sheet, followed by a short α-helix (A1). The C-terminal extension contains a short 3/10 helix (A2) and is cross-linked C-terminally by a third disulfide bond. Residues 311-317 of this extension are not visible in the electron density map. This exposed loop is possibly mobile or disordered due to potential cleavage at Ala314-Phe315 or Lys317-Thr318 (Hunt et al., 1993; Hunt and Krilis, 1994). In conclusion, domain V may be considered to consist of a core that is reminiscent of the consensus SCR fold with unique structural elements A1, A2 and B2″ and the exposed loop 311-317 forming a completely new face of this domain with respect to other known SCR domains. Oligosaccharide antennae As indicated by Kristensen et al. (1991), human β2GPI contains four N-glycosylation sites. The electron density shows seven carbohydrate units at these four sites, namely GlcNac-α(1-N)-Asn143, GlcNac-α(1-N)-Asn164, Man-β(1-4)-GlcNac-β(1-4)-GlcNac-β(1-N)-Asn174 and GlcNac-β(1-4)-GlcNac-β(1-N)-Asn234. Weak density at Thr130 indicates the position of an O-linked sugar (Gambino et al., 1997), which is in agreement with the sequence context of glycosylated threonines (Hansen et al., 1995). Four glycans (at positions 130, 143, 164 and 174) are located on domain III and one (at 234) on domain IV. Three glycans are positioned in the inner curve of the fish-hook, filling the niche formed by the molecule (Figure 1A). The remaining two glycans are located on domain III at the outer curve of the fish-hook. The five glycans point into large solvent channels present in the crystal. Crystal packing is dominated by interactions involving domains I, II and V. Very few crystal contacts involve domains III and IV (Figure 3). Moreover, domains III and IV, which carry the glycans, appear to bridge between the contacts made by the N- and C-terminal domains I, II and V. The observed shielding of domains III and IV may reflect an indirect functional role of the glycans. Figure 3.Crystal packing of β2GPI. Cα traces of one molecule with its four neighbouring molecules are shown. Crystal contacts involve predominantly domains I, II and V. Contacts with glycosylated domains III and IV are restricted to the N-terminal top and C-terminal bottom of the domains III and IV, respectively. In the crystal the glycosylated domains bridge the crystal contacts made by the N- and C-terminal domains. Download figure Download PowerPoint Interdomain flexibility The domains in β2GPI are connected by short linker regions of three (between domains IV and V) and four residues (all others), i.e. counting the number of residues between the C-terminal cysteine of the first domain and the N-terminal cysteine of the second domain. Between domains II-III and III-IV these linking residues form β-strands that connect sheets B4′-B5′ of the N-terminal domain with B1′-B2′ of the C-terminal domain (Figure 1A and B). The interactions observed at the interfaces are hydrophobic contacts with one hydrogen bond between domains I-II, II-III and III-IV and two hydrogen bonds between domains IV-V. Only a small amount (10-15%) of surface is buried at the interdomain interfaces: 422, 242, 445 and 492 Å2 for domains I-II up to IV-V, respectively. The four interdomain orientations observed in β2GPI display tilt angles (φ) varying from 128 to 160° and twist angles (ψ) varying from 41 to 137° (Table III, Figure 4). Slightly different angles (up to 6° difference) are observed for the low-resolution data Native I, for which the b-axis is 2.8 Å shorter. The observed tilt angles (φ) in all known SCR domain-domain structures are obtuse and range from 120 to 162° (Table III). The range observed for the twist angle (ψ) is much larger, 22-180°, which indicates a large variability in precise domain-domain interactions. Electron microscopy for CR2 (Moore et al., 1989) and HFH (Di Scipio, 1992), which contain 15 and 20 SCR repeats, also show elongated and winding structures. In these proteins up to eight residues link the separate domains yielding further flexibility. These electron microscopy, NMR and X-ray data suggest that multiple SCR repeats form elongated and rather flexible chains. Figure 4.Domain-domain orientations in β2GPI. (A) Stereo view of domain-domain orientations of the four sets of consecutive SCR domain pairs. The N-terminal domains of tandems II-III, III-IV and IV-V are superposed on domain I of tandem I-II. C-termini are indicated with the domain labels II-V. (B) The variation in domain orientations is expressed in a tilt angle φ and a twist angle ψ determined by the principal inertia axes a and b (Materials and methods, Table III). Download figure Download PowerPoint Table 3. Interdomain orientations Tilt angle φ (°) Twist angle ψ (°) β2GPI I-IIa 160, 162 137, 136 β2GPI II-III 160, 159 83, 78 β2GPI III-IV 128, 125 41, 46 β2GPI IV-V 131, 137 80, 76 HFH 15-16b 130 130 CD46 1-2c 120 180 VCP 3-4d 121 22 a Two values are given: the first value is obtained from β2GPI in crystal form II (cf. Native II), the second value refers to β2GPI in form I (Native I). b Reported standard deviations are 13° in φ and 17° in ψ (Wiles et al., 1997). c Between the six copies of the molecule in the asymmetric unit a difference of 15° in φ is reported (Casasnovas et al., 1999). d Reported standard deviations are 4° in φ and 6° in ψ (Wiles et al., 1997). Membrane binding The fifth domain of β2GPI has been implicated in membrane binding (Steinkasserer et al., 1992). In β2GPI we observe a large, positively charged area of ∼2000 Å2 on domain V (Figure 5). This patch is formed by side chains of 12 lysines, one arginine and one histidine located at the outer curve of the fish hook. It includes four lysines from the loop Cys281-Cys288 and lysines 308 and 324, which are important for phospholipid binding (Steinkasserer et al., 1992; Hunt and Krilis, 1994; Sheng et al., 1996). Other residues of this patch are Lys246, Lys250, Lys251, Arg260, Lys262, Lys266, Lys268 and His310. The flexible loop Ser311-Lys317, containing Trp316, which is essential for phospholipid binding (Sanghera et al., 1997) is located within this charged region (Figure 5B). Figure 5.Binding of β2GPI to an anionic phospholipid surface. (A) Two views, related by 180° rotation, of the electrostatic potential surface of β2GPI. Domains are labelled I-V. The electrostatic potential is scaled from red for negative to blue for positive. (B) Positively charged patch on the aberrant half of domain V. The 14 residues contributing to this patch and the position of the disordered loop Ser311-Lys317 are indicated. (C) Diagram of the proposed model for binding of β2GPI to acidic phospholipids. The positively charged patch on the surface of domain V is indicated by ‘+’, acidic phospholipids are depicted by ‘−’ and the putative membrane-insertion loop Ser311-Ser-Leu-Ala-Phe-Trp-Lys317 is shown to insert into the phospholipid layer. The positions of N-glycans are indicated by hexagons and the putative site for O-linked glycosylation is indicated by a diamond. Download figure Download PowerPoint The structural and biochemical data indicate a relatively simple membrane-binding mechanism. The positive charges on domain V interact with the anionic phospholipid headgroups and the flexible loop Ser311-Ser-Leu-Ala-Phe-Trp-Lys317 putatively inserts into the lipid layer and positions Trp316 at the interface region between the acyl chains and phosphate headgroups of the lipids, thereby anchoring the protein molecule to the membrane (Figure 5C). Furthermore, the combination of Trp, Phe or Tyr, followed by a Lys, is of particular importance for the interaction with the interfacial region between the lipid phosphate group and the acyl chains of lipids (Stopar et al., 1996; Mall et al., 1998; Mangavel et al., 1998; de Planque et al., 1999). Comparison of β2GPI sequences of bovine, canine, mouse, rat and human shows that the putative membrane-insertion loop Ser311-Lys317 is identical among these species and that substitutions with respect to a positively charged patch on domain V are conserved. Interestingly, all residues responsible for the unique function in membrane binding of domain V are located on the aberrant non-SCR-like half of this domain. Reduced affinity for acidic phospholipids, as observed for the two naturally occurring mutants Cys306 to Gly and Trp316 to Ser (Sanghera et al., 1997; Horbach et al., 1998) and for three cleaved isoforms of β2GPI, with scissile bonds between residues 314-315 and 317-318 (Hunt et al., 1993; Hunt and Krilis, 1994), can be readily explained by the proposed membrane-binding model. Both mutations and the two scissile bonds disrupt the integrity of the putative membrane-insertion loop 311-317. We think that the in vitro observed binding properties to heparin of β2GPI with the single mutation Trp316 to Ser (Horbach et al., 1998), and of two isoforms that are proteolytically cleaved between Lys317 and Thr318 and native β2GPI (Horbach et al., 1999), show a non-specific behaviour of the protein, which is not affected by alterations or disruptions of the membrane-insertion loop and is brought about solely by charge interactions. The loop Ser311-Ser-Leu-Ala-Phe-Trp-Lys317, therefore, gives β2GPI its specificity for phospholipid interfaces by introducing specific hydrophobic interactions between amino acid residues and acyl chains of phospholipids, in addition to the large number of charge interactions. Binding of anti-phospholipid autoantibodies The group of autoantibodies, aPLs, detected in blood plasma of patients with APS is both inter- and intra-individually heterogeneous. Indeed, the extended shape of β2GPI offers many potential sites for antibody binding, particularly at the non-glycosylated domains I, II and V. So far, aPL binding to domains I, III, IV and V, and the interdomain region between I and II, has been reported (Hunt and Krilis, 1994; Wang et al., 1995; George et al., 1998; Iverson et al., 1998; Blank et al., 1999). Wang et al. (1995) have identified two potential epitope sequences, Gly274-Phe280 and Ala314-Pro325. Based on the structure, both sequences are unlikely to be epitopes of aPLs. Gly274-Phe280 forms the central β-strand B3 of domain V, which is largely inaccessible to the solvent (17% solvent accessibility). It is, thus, unlikely to be either an epitope or a cryptic epitope, without fully disrupting the fifth domain. The second peptide, Ala314-Pro325 contains part of the putative membrane-insertion loop Ser311-Lys317. Binding of aPL to these residues will directly interfere with, if not fully abolish, membrane binding of β2GPI. This analysis shows that the crystal structure of β2GPI is an important tool for the eval
Referência(s)