Mode of Interaction between β2GPI and Lipoprotein Receptors Suggests Mutually Exclusive Binding of β2GPI to the Receptors and Anionic Phospholipids

Structure

Article Mode of Interaction between b2GPI and Lipoprotein Receptors Suggests Mutually Exclusive Binding of b2GPI to the Receptors and Anionic Phospholipids Chang-Jin Lee,1 Alfredo De Biasio,1 and Natalia Beglova1,* 1Department of Medicine, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA 02215, USA *Correspondence: [email protected] DOI 10.1016/j.str.2009.12.013

SUMMARY

Lipoprotein receptors of the LDLR family serve as clearance receptors for b2GPI and as signaling receptors for the b2GPI/antibody complexes in antiphospholipid syndrome. We compared four ligandbinding LA modules from LDLR and ApoER2 for their ability to bind domain V of b2GPI (b2GPI-DV). We found that the LA modules capable of binding b2GPI-DV interact with the same region on b2GPIDV using residues at their calcium-coordination site. The structure of a complex between b2GPI-DV and LA4 of LDLR, solved by molecular docking guided by NMR-derived restraints and extensively validated, represents the general mode of interaction between b2GPI and lipoprotein receptors. We have shown that b2GPI-DV cannot simultaneously bind to lipoprotein receptors and anionic phospholipids, suggesting that the association of b2GPI/anti-b2GPI antibody complexes with anionic phospholipids will interfere with lipoprotein receptors’ signaling in APS.

INTRODUCTION Beta2-glycoprotein I (b2GPI) is an abundant plasma protein with an established function in the pathology of antiphospholipid syndrome (APS). APS is an autoimmune disease characterized by recurrent thrombosis and pregnancy complications, and b2GPI is a major antigen for autoimmune antibodies in patients with APS (Giannakopoulos et al., 2009; Urbanus et al., 2008a). Precise molecular mechanisms leading to thrombotic complications in APS are not well understood and, as a consequence, current options for treatment and prevention of thrombosis are limited (Lim et al., 2006). The presence of anti-b2GPI antibodies is accompanied by cellular activation both in vitro and in vivo (Cugno et al., 2009; Koike et al., 2007; Pierangeli et al., 2004). A number of cellsurface molecules, including annexin A2 and Toll-like receptors (TLR) on endothelial cells and monocytes, and ApoER2 and GPIba on platelets, are suggested to play a role in APS (Cockrell et al., 2008; Lutters et al., 2003; Pennings et al., 2007b; Raschi et al., 2008; Shi et al., 2006; Sorice et al., 2007; Urbanus et al., 2008b; Zhang and McCrae, 2005). The importance of annexin

A2 and TLR-4 for the mediation of thrombogenic effects of APS-related antibodies was demonstrated in vivo in a mouse thrombosis model (Pierangeli et al., 2007; Romay-Penabad et al., 2009). Annexin A2, which is not a transmembrane receptor, binds b2GPI with high affinity even in the absence of antibodies (Ma et al., 2000). Bridging of annexin A2 by either anti-annexin A2 antibodies or by b2GPI/antibody complexes induces expression of adhesion molecules on endothelial cells (Zhang and McCrae, 2005). It was suggested that annexin A2 and TLR-4 may act in concert on endothelial cells (Cockrell et al., 2008; Zhang and McCrae, 2005). On platelets, ApoER2 and GPIba bind a recombinant chimeric dimer of b2GPI that mimics the dimeric b2GPI/ antibody complex, and synergize in platelet activation (Pennings et al., 2007b; Urbanus et al., 2008b). ApoER2 is a member of the low-density lipoprotein receptor (LDLR) family. Closely related members of the family, including LDLR, VLDLR, ApoER2, LRP and megalin, are present in all 79 human tissues analyzed by gene expression array (Su et al., 2004). These receptors not only mediate the cellular uptake of lipids and nutrients from the circulation and function as scavenger receptors, but are also involved in fundamental signaling processes (Herz et al., 2009; May et al., 2005; Stolt and Bock, 2006). Lipoprotein receptors were implicated in the clearance of b2GPI bound to apoptotic cells, and it was demonstrated that ApoER2 contribute to platelet activation by the b2GPI/antibody complexes (Maiti et al., 2008; Pennings et al., 2007a, 2007b; Urbanus et al., 2008b). Biochemical studies have shown that all lipoprotein receptors can bind dimerized b2GPI and these interactions are inhibited by RAP, which is a universal inhibitor of ligand binding by lipoprotein receptors (Pennings et al., 2006). Other lipoprotein receptors besides ApoER2 are expressed on endothelial cells and monocytes and might contribute to the activation of these cells by b2GPI/anti-b2GPI complexes, thus playing a part in the pathology of APS (Korschineck et al., 2001; May et al., 2005; Moestrup et al., 1992; Yang et al., 2009). Lipoprotein receptors interact with their ligands by LA modules of about 40 residues each, which are arranged in sequence to form a ligand-binding domain (Blacklow, 2007). The LA modules have little sequence homology, but adopt a very similar three-dimensional structure. The structure of an LA module is maintained by three disulfide bonds and a bound calcium ion. Lipoprotein receptors interact with domain V of b2GPI (b2GPI-DV) (van Lummel et al., 2005). All five domains of b2GPI have similar fold and few interdomain contacts (Bouma et al., 1999; Hammel et al., 2002; Schwarzenbacher et al., 1999). Compared with the four N-terminal domains, b2GPI-DV has an

366 Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved

Structure Structure of the b2GPI-DV/LA Complex

Table 1. Stoichiometry of the B2GPI-DV/LA4 Complex Derived from NMR Relaxation Data Estimated MW (Da) MW (Da)

NMR Sample

R2/R1

15

1.6 ± 0.2 2.7 ± 0.5 4,400

N-LA4a

15

N-B2GPI-DVb

15

N-B2GPI-DV+LA4

tc(ns)

3.2 ± 0.4 5.6 ± 0.5 9,500 c

5.3 ± 0.7 7.9 ± 0.6

13,402 ± 1,571

N-LA4 +B2GPI-DVd 4.9 ± 0.7 7.5 ± 0.7

12,222 ± 2,405

15

Composition of the NMR sample: a 15 N-LA4 (300 mM). b 15 N-B2GPI-DV (515 mM). c 15 N-B2GPI-DV (306 mM) in the presence of LA4 (917 mM). d 15 N-LA4 (275 mM) in the presence of B2GPI-DV (550 mM).

extra disulfide bond, a flexible loop inserted close to the C terminus of the domain, and an unusually high content of lysine residues (18.1%) in its primary sequence. In these studies, we uncovered the characteristic features of the LA modules capable of binding b2GPI and determined the structural details of the binding interface in the b2GPI/LA complex by comparing different LA modules for their ability to bind b2GPI. We used four different ligand-binding LA modules, LA3, LA4, and LA6 from LDLR, and LA1 from ApoER2, as examples to investigate how LA modules interact with b2GPI. LDLR is a primary internalization receptor for LDL. ApoER2, which has high level of expression in the brain and testis, is critical both for neuronal migration and for selenium uptake through endocytosis of selenoprotein P (Masiulis et al., 2009). The first LA module in ApoER2, ApoER2-A1, was shown to contribute the most to the binding of dimeric b2GPI (Pennings et al., 2007a). Both LA3 and LA4 bind RAP (Fisher et al., 2006), which also inhibits the interaction between dimeric b2GPI and lipoprotein receptors (Pennings et al., 2006). Using solution nuclear magnetic resonance (NMR) spectroscopy, we have shown that the LA modules capable of binding b2GPI-DV interact with the same site on b2GPI-DV by residues at their calcium-coordination region. We solved the structure of a complex between b2GPI-DV and LA4 by molecular docking guided by NMR-derived restraints. In the structure, extensively validated by binding studies and site-directed mutagenesis, the contact site for LA modules on b2GPI-DV is formed by 3 lysine residues: K308, K317, and K282. The structure of the b2GPI-DV/LA4 complex represents how b2GPI-DV interacts with LA modules and establishes the characteristic features that distinguish LA modules capable of binding b2GPI. We have shown that b2GPI-DV cannot simultaneously bind to lipoprotein receptors and anionic phospholipids, suggesting that the association of b2GPI/antibody complexes with anionic phospholipids will interfere with lipoprotein receptors’ signaling in APS. RESULTS Stoichiometry of the b2GPI-DV/LA4 Complex We determined the stoichiometry of binding between b2GPIDV and LA4 by measuring R1 and R2 relaxation rates for the backbone amides of 15N-labeled b2GPI-DV and 15N-labeled LA4 in isolation and in the presence of unlabeled interacting domains at saturating concentrations. The measured relaxation

rates were used to calculate an overall rotational correlation time, tc, for the labeled molecules (Table 1). The values of tc for LA4 and b2GPI-DV in isolation are consistent with previously published data (Beglova et al., 2001; Hoshino et al., 2000). The effective molecular weight of the b2GPI-DV/LA4 complex estimated from the NMR relaxation measurements indicates that the complex has 1:1 stoichiometry (Table 1). Mapping the Binding Site for LA Modules onto b2GPI-DV To determine which residues of b2GPI-DV form the contact site for LA modules, we monitored changes in the spectra of the 15 N-labeled b2GPI-DV in the presence of different LA modules. The residues of the flexible C-terminal loop of b2GPI-DV from H310 to A320 are either weak or missing from the spectra and, therefore, excluded from the analysis. Comparison of the chemical shift perturbation spectra of b2GPI-DV revealed that the same subset of residues is perturbed by LA4, LA3, and ApoER2-A1 (Figures 1A–1E). In addition to these core residues that likely comprise the major contact site for LA modules on b2GPI-DV, another group of residues is perturbed by LA4 and ApoER2-A1, suggesting that both LA4 and ApoER2-A1 form additional contacts with b2GPI-DV that are not present in the b2GPI-DV/LA3 complex. Finally, several additional b2GPI-DV residues are perturbed when ApoER2-A1 is used for titration. The perturbation of the residues I263, M271, and I301, which form a hydrophobic cluster in b2GPI-DV, suggests that the binding of ApoER2-A1 to b2GPI-DV affects the backbone structure of b2GPI-DV distant from the contact site. The extent of perturbations caused by different LA modules suggests that ApoER2-A1 binds b2GPI-DV with a higher affinity than LA4 or LA3, and that LA3 has the weakest affinity for b2GPI-DV. The Binding Affinities of b2GPI-DV to LA Modules Measured by Fluorescence Polarization Changes in the global mobility of the fluorescently labeled LA modules upon binding to b2GPI-DV are reflected in an increase in the measured fluorescence polarization (Figure 2). Fitting of the binding curves resulted in the Kd values of 1.14 ± 0.06 mM and 46 ± 7 mM for the b2GPI-DV/ApoER2-A1 and b2GPI-DV/LA4 complexes, respectively. The binding of LA3 to b2GPI-DV did not reach saturation even when 1 mM b2GPI-DV was added to LA3 (data not shown), suggesting that Kd for this complex is in a high micromolar range. The increase in viscosity of the protein solution at high protein concentration, which would lead to an overestimation of Kd (Lefebvre, 1982), precluded us from measuring Kd of the b2GPI-DV/LA3 complex by fluorescence polarization. Mapping the Binding Site for b2GPI-DV onto LA Modules The 15N-labeled LA modules, LA4 and ApoER2-A1, were titrated with unlabeled b2GPI-DV to identify the residues perturbed by complex formation (Figures 3A–3E). The relatively small changes in the chemical shifts of the backbone amides of the LA modules in the presence of b2GPI-DV suggest that these modules bind b2GPI-DV predominantly by the side chains. For both LA modules, the majority of perturbed residues are either directly involved in coordination of the calcium ion or are adjacent to the calcium-binding residues, suggesting that the contact site for b2GPI-DV on the LA modules overlaps with the calciumbinding site. Two residues from the hydrophobic core of the

Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved 367

Structure Structure of the b2GPI-DV/LA Complex

Figure 1. Interaction of b2GPI-DV with LA Modules (A–C) Superposition of 15N-HSQC spectra of 184 mM b2GPI-DV in isolation (black) and in the presence of 400 mM LA4 (A), LA3 (B), and ApoER2-A1 (C). Perturbed residues of b2GPI-DV are labeled. (D and E) Perturbed residues of b2GPI-DV mapped onto the structure (D) and the primary sequence of b2GPI-DV (E). Residues of b2GPI-DV that are perturbed when b2GPI-DV is titrated with LA3, LA4, and ApoER2-A1 (red); with LA4 and ApoER2-A1 (orange); with ApoER2-A1 (purple). Residues that are missing from the 15N-HSQC spectra of b2GPI-DV are colored gray. The NH groups of tryptophan side chains are marked 3W on (E).

populated first cluster, strongly suggesting that it represents a correct solution. The ensemble of the 20 best-scoring structures from the first cluster has 1.08 ± 0.47 A˚ of an overall backbone root-meansquare deviation (rmsd) from the average structure, indicating that the complex is well defined (Figure 4A and Table 2). The calculated structures have good geometry and occupy the allowed and most favorable regions on the Ramachandran plot.

LA4 module, F132 and I140, and the corresponding I26 of ApoER2-A1 are also perturbed in the presence of b2GPI-DV. Because the structure of the LA modules is dependent on the bound calcium ion, the perturbation of the residues in the hydrophobic core of the LA modules supports our observation that the binding interface for b2GPI-DV on LA modules is centered at the calcium-binding site. Docking Model of the b2GPI-DV/LA4 Complex Because the structures of LA4 and b2GPI-DV are available in the Protein Data Bank (PDB) database, we used these molecules for docking. We built a docking model of the b2GPI-DV/LA4 complex using the program HADDOCK2.0 and the interaction restraints derived from the NMR data. The clustering of the 200 refined structures resulted in only two clusters with more than 10 structures in each. The first cluster of 141 structures had an average HADDOCK score of 97.7 ± 13.5, and the second cluster of 26 structures had an average HADDOCK score of 70.3 ± 12.4. The best-scoring structure belongs to the most

Structural Features of the b2GPI-DV/LA4 Complex The intermolecular interface in the b2GPIDV/LA4 complex (PDB ID 2KRI) is maintained by a number of hydrogen bonds and hydrophobic contacts (Table 3). As clearly seen from Table 3, the majority of intermolecular contacts are created by K308 and K317 of b2GPI-DV. These contacts are further augmented by the hydrogen bonds between K282 of b2GPI-DV and aspartates 147 and 149 of LA4. The intermolecular interface in the complex is relatively small and maintained primarily by electrostatic interactions between positively charged lysine residues of b2GPI-DV and negatively charged aspartates of LA4 grouped around the bound calcium ion (Table 3 and Figures 4B and 4C). A solvent-accessible surface area of b2GPI-DV buried in the interface of the b2GPI-DV/LA4 complex with a minimal HADDOCK score is 455 A˚2. Validation of the Model We used four different approaches to validate the docking model of the b2GPI-DV/LA4 complex. First, we mutated the critical lysine residues of b2GPI-DV, K308, K317, and K282 to aspartates and evaluated the binding of mutants to LA4 (Figures 5A and 5B). The absence of perturbations on the 15N-HSQC spectra of the mutants in the presence of LA4 suggests that the mutants are incapable of binding LA4 as predicted by the model. Second, the b2GPI-DV/LA4 structure shows that the N148/E mutation of LA4 may create an additional hydrogen bond with K282, leading

368 Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved

Structure Structure of the b2GPI-DV/LA Complex

Figure 2. Binding of b2GPI-DV to LA Modules Determined by Fluorescence Polarization Atto488-labeled LA4 (triangles) and Atto488-labeled ApoER2-A1 (circles) were titrated with unlabeled b2GPI-DV. Results are expressed as mean and standard deviation (± SD) (n = 3).

to a tighter complex. Kd of the N148/E mutant is 9.0 ± 1.6 mM compared with 46 ± 7 mM of the b2GPI-DV/LA4 complex (Figure 5C). Interestingly, N148 of LA4 corresponds to E34 in the ApoER2-A1 module, which binds b2GPI-DV with 1.14 ± 0.06 mM affinity. Third, we confirmed that the lysine residues from b2GPI-DV are critical for the interaction with the LA modules and that the binding interface of the complex overlaps with the calcium coordinating residues of the LA module. We selected the sixth LA module from the LDLR. The structure of LA6 is very similar to LA4 (Clayton et al., 2000; North and Blacklow, 2000), but LA6 contains an arginine residue in the predicted binding interface. This arginine corresponding to W144 in the primary sequence of LA4 will interfere with the lysines in the binding site of b2GPI-DV, precluding complex formation. The absence of any perturbations in the spectra of 15N-LA6 in the presence of b2GPI-DV confirms that LA6 does not bind b2GPI-DV (Figure 5D). Fourth, we monitored changes in the intrinsic fluorescence of W316, the only tryptophan residue of b2GPI-DV, upon complex formation. For the titration experiments we used LA3, which does not have tryptophan in its primary sequence. The fluorescence of W316 decreases gradually upon titration with LA3, confirming that W316 is close to the binding interface in the complex (Figure 5E and Table 3). The emission maximum of W316 is 357 nm and does not change in the presence of LA3, suggesting that W316 remains in the polar environment even when LA3 is bound. This observation is in agreement with the structure of the b2GPI-DV/LA complex, which shows that W316 is in the periphery of the binding interface. Kd of the b2GPI-DV/LA3 complex roughly estimated from the concentration of LA3 at half-maximum binding is about 75 mM, which is in accordance with our observation that LA3 interacts with b2GPI-DV with a weaker affinity than LA4. LA Modules Inhibit Binding of b2GPI-DV to Cardiolipin A sequence of hydrophobic residues 313–316 (Mehdi et al., 2000) and a lysine-rich region from residue 281 to 288 (Hunt and Krilis, 1994; Sheng et al., 1996) are both important for

b2GPI-DV binding to anionic phospholipids. In the complex, LA4 wedges between these two regions. Therefore, we evaluated, if LA modules interfere with the binding of b2GPI-DV to cardiolipin. b2GPI-DV bound to cardiolipin was detected by a horseradish peroxidase (HRP)-conjugated antibody directed to HA tag added at the N terminus of b2GPI-DV (HA-b2GPI-DV). First, we compared the binding affinities for cardiolipin of a monomeric HA-b2GPI-DV and of HA-b2GPI-DV dimerized by anti-HA antibodies. A concentration of HA-b2GPI-DV at halfmaximum binding was 22 ± 4 nM and 1.6 ± 0.3 nM for monomeric HA-b2GPI-DV and for HA-b2GPI-DV preincubated with antibodies, respectively (Figure 6A). As expected, the binding affinity of domain V for cardiolipin is close to that of a full-length b2GPI (Hagihara et al., 1995) and the dimerization of HA-b2GPIDV increased the apparent affinity of HA-b2GPI-DV for cardiolipin more than 10-fold (van Lummel et al., 2005). When the total concentration of HA-b2GPI-DV is 10 mM and above, HAb2GPI-DV saturates all available antibodies and free HAb2GPI-DV competes with HA-b2GPI-DV/antibody complexes for binding sites on cardiolipin, resulting in diminished binding of the complexes to cardiolipin and lower measured OD (Figure 6A). To investigate whether LA modules interfere with the binding of HA-b2GPI-DV to cardiolipin, we used either LA4 or a pair of the modules, LA3-4. We have already shown above that each of these modules, LA3 and LA4, can bind b2GPI-DV. A gradual decrease of bound HA-b2GPI-DV indicates that both LA4 and LA3-4 interfere with the binding of HA-b2GPI-DV to cardiolipin. The shape of the inhibition curve strongly suggests that LA3-4 competes with cardiolipin for the same binding site on HA-b2GPI-DV. The concentration of LA3-4 that inhibits 50% of binding of HA-b2GPI-DV to cardiolipin was 106 ± 9 mM (Figure 6B). The apparent Ki of LA3-4 for HA-b2GPI-DV in the presence of anti-HA antibodies was calculated using an affinity partitioning approach to estimate amount of HA-b2GPI-DV bound to LA3-4 (Olson et al., 1991). Assuming that the amount of HA-b2GPI-DV bound to cardiolipin is much smaller than the total amount of HA-b2GPI-DV, the calculated apparent Ki of LA3-4 was 6 mM. This experiment provides additional validation for the structure of the b2GPI-DV/LA4 complex, confirms that LA3 and LA4 interacts with the same site on b2GPI-DV, and shows that a pair of LA modules binds b2GPI-DV dimerized by antibody significantly stronger than a single LA module binds b2GPI-DV. It also demonstrates that the interactions of b2GPI-DV with lipoprotein receptors and negative phospholipids are mutually exclusive. DISCUSSION ApoER2, a lipoprotein receptor expressed on platelets, functions as a signaling receptor for dimeric b2GPI (Lutters et al., 2003; Pennings et al., 2007a, 2007b; Urbanus et al., 2008b). Besides platelets, several members of the LDLR family are expressed on endothelial cells and monocytes, and might contribute to thrombosis in APS. We used solution NMR spectroscopy, molecular docking, and binding studies to investigate and compare the structural features of the complexes between b2GPI-DV and the ligand-binding LA modules from lipoprotein receptors.

Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved 369

Structure Structure of the b2GPI-DV/LA Complex

Figure 3. Mapping the Contact Site For b2GPI-DV onto LA Modules (A and B) Superposition of the 15N-HSQC spectra of 100 mM LA4 in isolation (black) and in the presence of 1:0.75 (green) and 1: 2 (red) molar ratio of LA4 to b2GPI-DV (A). Superposition of the 15N-HSQC spectra of 100 mM ApoER2-A1 in isolation (black) and in the presence of 1:0.5 (green) and 1:2 (red) molar ratio of ApoER2-A1 to b2GPI-DV (B). Residues of the LA modules that moved above average in the presence of b2GPI-DV are labeled. The NH groups of tryptophan side chains are labeled 3Trp. (C and D) Combined, 1H and 15N, chemical shift perturbations of LA4 (C) and ApoER2-A1 (D) in the presence of saturating amount of b2GPI-DV. Chemical shift difference is calculated as D = sqrt(DH**2+(0.2*DN)**2). The last bars on both panels correspond to tryptophan side chains and are marked by asterisks. (E) Chemical shift perturbations mapped onto the primary sequence of LA modules. Residues that are shifted above average are colored pink and the residues shifted more than 1.5 standard deviations above average are colored green. The calcium coordinating residues are labeled by filled black circles and the disulfide bonds are indicated by lines. The sequence alignment was done with ClustalW.

The Docking Model of the b2GPI-DV/LA4 Complex Is Supported by Independent Experimental Data We solved the structure of the b2GPI-DV/LA complex by molecular docking guided by the NMR-derived restraints. This is the preferred approach in the case of the b2GPI-DV/LA complexes. Our NMR titration studies have shown that the binding interface in the complexes is relatively small and formed predominantly by the side chains, suggesting that it might not yield enough distance restraints for accurate structure determination by NMR. The relatively weak affinity of b2GPI-DV for the LA modules makes crystallization of the complex difficult. For the modeling studies, we selected LA4 because its structure is available in the PDB database. The carbohydrate chains of b2GPI are not expected to interfere with the b2GPI-DV/LA complex and, therefore, are not taken into account for the modeling. b2GPI

has four N-linked glycosylation sites at the residues N143, N164, and N174 in domain III and N234 in domain IV. All four glycosylation sites are far from the binding interface for LA modules. The structure of the complex is consistent with independent experimental observations. First, we confirmed the structure by site-directed mutagenesis. Based on the structure of the b2GPI-DV/LA4 complex we selected two types of mutations. One type of mutations disrupts the complex formation and the other creates additional hydrogen bonds improving the binding affinity of the complex. Mutation of the critical lysine residues of b2GPI-DV, K308, K317, and K282, to aspartates completely abolished the binding of LA4 to the b2GPI-DV mutant. The N148/E mutation of LA4 resulted in a tighter complex with Kd 9.0 ± 1.6 mM compared with 46 ± 7 mM of LA4. Second, the

370 Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved

Structure Structure of the b2GPI-DV/LA Complex

Figure 4. Structure of the b2GPI-DV/LA4 Complex (A) Superposition of 20 best structures. The polypeptide chains of b2GPI-DV and LA4 are green and blue, respectively, and the calcium ions bound to LA4 are gray spheres. (B) Stereoview of the overall structure of the b2GPI-DV/LA4 complex. b2GPI-DV (brown) and LA4 (green) are illustrated as ribbons. The transparent molecular surface of b2GPI-DV is colored light brown. The calcium ion bound by LA4 is shown as a gray sphere. (C) Interface between b2GPI-DV and LA4. The backbone traces of b2GPI-DV (brown) and LA4 (green) are shown as ribbons and the calcium ion is a gray sphere. The interacting side chains are rendered as sticks and labeled.

structure of the b2GPI-DV/LA4 complex is consistent with previously published data for the W316/Ser mutant of b2GPI (Pennings et al., 2006; van Lummel et al., 2005). The mutation had no effect on the binding of dimeric b2GPI to immobilized soluble ApoER2, VLDLR, LRP, and megalin and to ApoER2 expressed on the platelet surface. In the b2GPI-DV/LA4 structure, this mutation does not create clashes between b2GPI-DV and LA4 and, therefore, does not disrupt the complex. Third, the b2GPI-DV/LA4 complex built by two different computational docking approaches HADDOCK and PIPER (Beglov et al., 2009) converged into nearly identical structures with a backbone rmsd of only 0.5 A˚. Structure of the b2GPI-DV/LA4 Complex Represents the General Mode of Interaction between b2GPI and Lipoprotein Receptors Our results show that not all LA modules can bind b2GPI-DV, and that those LA modules capable of binding b2GPI-DV will interact with it in the same fashion as observed in the structure of the b2GPI-DV/LA4 complex. We have demonstrated by solution NMR spectroscopy that LA3, LA4, and ApoER2-A1 bind to the same site on b2GPI-DV,

and that the binding site for b2GPI-DV on these LA modules is centered on the residues involved in coordination of the calcium ion. The dissociation constants measured here for different LA modules are in the range from 1 mM (for ApoER2-A1) to more than 100 mM (for LA3). The difference in the binding affinity is due to the auxiliary contacts that augment the interactions of the residues at the core of the binding interface. These auxiliary contacts depend on the primary sequence of the LA modules. When we mutated N148 of LA4 to glutamate to resemble E34 at the corresponding position in ApoER2-A1, which binds b2GPI-DV with 1 mM affinity, we improved Kd of LA4 by a factor of five. The changes in the intrinsic fluorescence of W316 of b2GPI-DV upon titration with LA3 confirmed that W316 is close to the binding interface and remains solvent exposed in the b2GPI-DV/LA3 complex, similarly to what is observed in the b2GPI-DV/LA4 complex. LA modules capable of binding b2GPI could be distinguished based on the primary sequence in the vicinity of the calciumcoordinating residues of the LA modules. Specifically, the side chains of amino acids at the position in the primary sequence equivalent to D147, D149, and D151 of LA4, the immediate neighbors of these residues as well as the residues equivalent

Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved 371

Structure Structure of the b2GPI-DV/LA Complex

Table 2. Structural Statistics for the Twenty Best-Scoring B2GPIDV/LA4 Structures B2GPI-DV/LA4 Complex Backbone rmsd (A˚) with respect to average structure

Table 3. Statistics of the Intermolecular Contacts Calculated for the Ensemble of the Twenty Best-Scoring Structures Donor Lys282 (B2GPI-DV)

All atoms

1.08 ± 0.47

Atoms in the interface

1.04 ± 0.47

Atoms in the interface of B2GPI-DV

0.84 ± 0.28

Atoms in the interface of LA4

1.49 ± 0.49

Lys308 (B2GPI-DV)

Lys317 (B2GPI-DV)

CNS intermolecular energies after water refinement (kcal/mol)

Hydrogen Bonds S-Sa S-Bb

Nonbonded Contacts

Asp147 (LA4)

11

11

Asp149 (LA4)

13

Acceptor

Asp147 (LA4)

20

Asp149 (LA4)

17

Asp151 (LA4)

19

Asp149 (LA4)

16

Pro150 (LA4

20

Asp151 (LA4

19

Evdw

6 ± 6

Eelec

599 ± 31

Lys284 (B2GPI-DV)

Asp147 (LA4)

15

887 ± 66

Lys305 (B2GPI-DV

Leu143 (LA4)

13

Lys308 (B2GPI-DV)

Trp144 (LA4)

20

0.00362 ± 0.00008

Trp316 (B2GPI-DV)

Pro150 (LA4)

15

Angles (deg)

0.56 ± 0.02

Lys317 (B2GPI-DV)

Asp149 (LA4)

17

Impropers (deg)

0.50 ± 0.02

Intermolecular contacts are reported only if present in at least 10 out of 20 structures. a Side chain/side-chain contacts. b Side chain/backbone contacts.

Buried surface area (A˚2) Rmsd from idealized covalent geometry Bonds (A˚)

Ramachandran analysis of the average structure Residues in the favored region (%)

84.5

Residues in additionally allowed region (%)

15.5

to L143 and W144 should not interfere with lysine side chains of b2GPI-DV. Based on the outlined criteria, each lipoprotein receptor contains several LA modules that can bind b2GPI. LDLR has the least number of LA modules (LA3, LA4, and possibly LA5) available for interaction with b2GPI, which might explain why LDLR has the weakest binding affinity for the b2GPI/antibody complexes in binding studies (Pennings et al., 2006). The spatial orientation of the side chains in the contact interface of the b2GPI-DV/LA4 complex strongly resembles that observed in the crystal structure of the RAP/LA3-4 complex (Fisher et al., 2006). The arrangement of aspartates from LA4 around K308 from b2GPI-DV is very similar to that around K256 of RAP. As in the RAP/LA3-4 structure, two other lysines of b2GPI, K282 and K317, in the vicinity of K308 additionally stabilize the b2GPI-DV/LA4 complex. Our data strongly suggest that the general mode of recognition of basic ligands by the LA modules is directly applicable to the complexes between b2GPI and lipoprotein receptors. Effect of Avidity in Interaction between b2GPI and LA Modules We have shown that a single LA module binds b2GPI-DV with only micromolar affinity. Compared with a single LA module, a pair of modules connected by a flexible linker was able to efficiently bind HA-tagged b2GPI-DV dimerized by anti-HA antibodies and inhibit its attachment to cardiolipin. Although LA3 binds b2GPI-DV with low affinity (Kd > 100 mM), its presence in the LA3-4 pair significantly improved the binding of the LA module-pair to dimerized b2GPI-DV. Typically, lipoprotein receptors use several LA modules connected by flexible linkers to bind diverse ligands which, in turn, have binding sites for several LA modules, suggesting that a single module would be insufficient to confer high-affinity binding (Andersen et al., 2003; Fisher et al., 2004). The modular organization of the

ligand-binding domains of lipoprotein receptors, in combination with the flexibility of the linkers connecting individual LA modules, is tailored to accommodate ligands that contain multiple binding sites. The correlation between the presence of anti-b2GPI antibodies and the development of APS might underscore the role of lipoprotein receptors in the pathology of APS. b2GPI/antibody complex engages two LA modules, significantly stabilizing the b2GPI/anti-b2GPI/receptor complex on the cellular surface. In neurons, clustering of ApoER2 and VLDLR by Reelin starts the intracellular signaling (Strasser et al., 2004). It is possible that the dimerization of lipoprotein receptors by b2GPI/antibody complexes also contributes to the pathology of APS. The Binding of b2GPI to Lipoprotein Receptors versus Anionic Phospholipids The structure of the b2GPI-DV/LA complex and the observation that the LA modules inhibit binding of b2GPI-DV to cardiolipin indicate that b2GPI cannot simultaneously interact by its domain V with lipoprotein receptors and anionic phospholipids. When bound to b2GPI-DV, LA modules prevent two phospholipidbinding regions on b2GPI (residues 282–287 and 313–316) from simultaneously reaching the cellular membrane. On the other hand, when b2GPI is bound to the cellular surface, the binding site for LA modules is not accessible even though LA does not bind residues that interact with phospholipids. We performed cardiolipin binding and inhibition studies at low ionic strength to match the buffer used for NMR titration studies and to minimize the amount of proteins used in the experiment. We do not expect that the presence of sodium ions at physiological concentration alters the structure of the b2GPI-DV/LA complex. The LA module inserted between the two phospholipid-binding regions of b2GPI-DV in the complex will still interfere with the binding of b2GPI to anionic phospholipids, although the binding of b2GPI-DV to both LA and phospholipids will be attenuated by the increased ionic strength.

372 Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved

Structure Structure of the b2GPI-DV/LA Complex

Figure 5. Validation of the b2GPI-DV/LA4 Structure (A and B) Site-directed mutagenesis in combination with NMR studies. 15N-HSQC spectrum of 200 mM b2GPI-DV K282,308/D mutant (black) is overlaid with the spectrum of 190 mM b2GPI-DV K282,308/D in the presence of 400 mM LA4 (red) (A). 15N-HSQC spectrum of 275 mM b2GPI-DV K282,308,317/D mutant (black) is overlaid with the spectrum of 200 mM b2GPI-DV K282,308,317/D in the presence of 445 mM LA4 (red) (B). (C) Site-directed mutagenesis in combination with fluorescence polarization measurements. Fluorescence polarization of Atto488-labeled LA4 (triangles) and Atto488-labeled LA4 Asn148/Glu mutant (circles) titrated with b2GPI-DV. Results are expressed as mean ± SD (n = 3). (D) Superposition of the 15N-HSQC spectra of 200 mM 15N-LA6 in isolation (black) and in the presence of 200 mM b2GPI-DV (red). The side chains are shown by horizontal lines. (E) Superposition of the fluorescence emission spectra of 100 mM b2GPI-DV. For each spectrum, the corresponding amount of LA3 is indicated. Fluorescence maximum recorded at 357 nm is expressed as mean ± SD (n = 3) and is labeled by a black circle with an error bar.

Normally, anionic phospholipids are present on the inner leaflet of the cellular membrane and become exposed on activated or apoptotic cells (Zwaal and Schroit, 1997). The content of phosphatidylserine on the outer membrane of nonactivated cells is about 7% (Murphy et al., 1992). The measured binding affinity of b2GPI for phospholipid membrane composed of 10%/90% PS/PC in the presence of 120 mM NaCl and 3 mM CaCl2 was 14 mM for monomeric b2GPI and 0.4 mM for b2GPI/ antibody complexes (Willems et al., 1996). The increase of phosphatidylserine to 20% resulted in a much better binding of b2GPI

to phospholipids. The measured affinity of b2GPI/antibody complexes improved to 75 nM and 5 nM in the presence and absence of calcium ions (Willems et al., 1996). Similarly, the measured Kd of the dimeric b2GPI for 20%/80% PS/PC in the absence of calcium was 18 nM (van Lummel et al., 2005). Because lipoprotein receptors bind dimeric b2GPI with a low nanomolar affinity (Pennings et al., 2006), the interactions of b2GPI/antibody complexes with lipoprotein receptors will not be efficient on activated cells exposing high levels of phosphatidylserine.

Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved 373

Structure Structure of the b2GPI-DV/LA Complex

Figure 6. Binding of b2GPI-DV to Cardiolipin (A) Binding of HA-b2GPI-DV (inverted black triangles) and HA-b2GPI-DV preincubated with antiHA antibody (black circles) to cardiolipin. Results are expressed as mean ± SD (n = 3). (B) Inhibition of binding of HA-b2GPI-DV to cardiolipin. Monomeric HA-b2GPI-DV was incubated with increasing amounts of LA4 (inverted black triangles); HA-b2GPI-DV was first preincubated with anti-HA antibodies and then incubated with increasing amounts of the LA3-4 modulepair (black circles). Results are expressed as mean ± SD (n = 3).

On platelets, the signaling by dimeric b2GPI involves two receptors, ApoER2 and GPIba (Urbanus et al., 2008b). Both receptors act independently and neither receptor alone is sufficient to support platelet adhesion to fibronectin. The dimeric W316/Ser mutant of b2GPI that cannot bind to anionic phospholipids does not support platelet adhesion either (Lutters et al., 2003; van Lummel et al., 2005). Because the W316/Ser mutant binds ApoER2 similarly to wild-type b2GPI (van Lummel et al., 2005), it is likely that the mutation perturbs the binding of b2GPI to GPIba. Because the binding site for GPIba on b2GPI is in domain V (Pennings et al., 2007b), it is conceivable that the W316/Ser mutation interferes with the binding of dimeric b2GPI to GPIba, abolishing platelet activation. Current knowledge indicates that APS-related antibodies only predispose to thrombosis. Despite the presence of persistent antibodies, APS patients do not continuously suffer from thrombosis, which is often triggered by infection or trauma (Sherer et al., 2007; Shoenfeld et al., 2008; Urbanus et al., 2008a). In in vivo experiments, thrombosis must be provoked by infection or vessel injury (Fischetti et al., 2005; Jankowski et al., 2003; Pierangeli et al., 1996). Adhesive surfaces are required for detectable platelet activation by dimeric b2GPI in vitro (Lutters et al., 2003; Urbanus et al., 2008b; van Lummel et al., 2005). Lipoprotein receptors are constitutively expressed in different combinations on cellular membranes of endothelial cells, monocytes, and platelets. Acting in parallel with other receptors, b2GPI/antibody complexes bound to lipoprotein receptors might continuously sensitize cells, making them vulnerable to other prothrombotic factors. On endothelial cells annexin A2 provides high-affinity binding sites for b2GPI/antibody complexes (Ma et al., 2000; Zhang and McCrae, 2005). The binding site for annexin A2 on b2GPI and whether lipoprotein receptors act alongside with annexin A2 on stimulation of endothelial cells are not yet known. The role of lipoprotein receptors expressed on endothelial cells and monocytes in the pathology of APS and the signaling pathways initiated by lipoprotein receptors in response to b2GPI/antibody complexes are not well understood. Our structural and binding studies might be helpful in further research addressing the contribution of lipoprotein receptors to the pathology of APS. EXPERIMENTAL PROCEDURES Proteins The fragments of the human LDLR (residues 86–122 [LA3], 126–165 [LA4], and 212–251 [LA6]) and mouse ApoER2 (ApoER2-A1 residues 12–47) were ex-

pressed in E. coli and purified from inclusion bodies as previously described (North and Blacklow, 2000). b2GPI-DV (residues 244–326) was subcloned into a pET15b vector (Novagen). The encoded protein has an N-terminal histidine tag followed by the sequence recognized by the tobacco etch virus (TEV) protease so that the His tag can be removed. b2GPI-DV expressed in E. coli was recovered from inclusion bodies, cleaved with TEV, and refolded under conditions permitting disulfide exchange before final purification by reversephase high-performance liquid chromatography (HPLC) on a C18 column. NMR Spectroscopy For NMR spectroscopy, the labeled protein domains were produced in M9 minimal media supplemented either with 13C6-D-glucose and/or 15N ammonium chloride. All NMR experiments were conducted on a Varian Inova 500 MHz spectrometer equipped with four RF channels and a triple-resonance probe with pulsed-field gradients. Spectra were acquired at 298 K using BioPack (Varian Inc.) pulse sequences and processed with GIFA v4.3 software (Pons et al., 1996). Chemical shifts were referenced to 2,2-dimethyl-2-silapentane-5-sulfonate sodium salt (DSS) used as an internal reference. The buffer used for the NMR experiments contained 90%H2O/10%D2O, 10 mM bisTris with 10 mM CaCl2 (pH 7.1). Backbone 1H, 13C, and 15N resonance assignments of b2GPI-DV, LA3, LA4, and ApoER2-A1 were carried out by analyzing HNCA, NHCOCA, HNCACB, and HNCOCACB spectra. 15 N-R1 and 15N-R2 relaxation rates were measured in a series of 7–11 experiments with increased relaxation time delays. To determine the relaxation rates, the experimentally measured resonance intensities were fitted to a monoexponential decay model using the xcrvfit program (http://www. bionmr.ualberta.ca/bds/software/xcrvfit). The overall rotational correlation time tc was calculated from the R2/R1 ratio using the r2r1_tm program. Fluorescence Polarization Measurements Experiments were performed on an EnVision plate reader (Perkin-Elmer) using 480 nm excitation, 535 nm emission filters with 30 nm and 40 nm bandwidths, respectively, and a FP FITC dual optics module. The LA modules were labeled with Atto488 (Sigma) according to the manufacturer’s protocol and the labeled modules were purified by reverse-phase HPLC. Fluorescence polarization measurements were carried out in 10 mM bis-Tris buffer with 10 mM CaCl2 (pH 7.0). Experiments were performed in triplicates on a 384 well plate using 40 ml of a reaction mixture per well containing either 80 nM LA4, 150 nM ApoER2-A1, or 130 nM of the LA4 N148/E mutant. GNUPLOT was used to fit the Kd values to the experimental data via a one-site binding model. Measurements of the Intrinsic Tryptophan Fluorescence of b2GPI-DV Fluorescence measurements were performed on a Spectramax GeminiXS (Molecular Devices) using a 96 well plate. Each sample contained 100 mM of b2GPI-DV in 10 mM bis-Tris buffer with 10 mM CaCl2 (pH 7.2) and varied amounts of LA3. Samples were allowed to equilibrate for 25 min before spectra were acquired. The excitation wavelength was set to 280 nm and the emission scan was read from 330 nm to 400 nm with 1 nm step. To account for the fluorescence of the C-terminal tyrosine of LA3, the measured fluorescence at each titration point was corrected for the fluorescence of the blank sample that contained the same amount of LA3 in the absence of b2GPI-DV. Measurements for

374 Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved

Structure Structure of the b2GPI-DV/LA Complex

each sample and blank were done in triplicates using 250 ml sample volume per well.

Beglova, N., North, C.L., and Blacklow, S.C. (2001). Backbone dynamics of a module pair from the ligand-binding domain of the LDL receptor. Biochemistry 40, 2808–2815.

Molecular Docking with HADDOCK The coordinates of b2GPI-DV were extracted from the crystal structure of the full-length protein (PDB code 1c1z), and the structure of LA4 including a calcium ion was taken from the X-ray structure of the RAP/LA3-4 complex (PDB code 2fcw). The docking was performed with HADDOCK 2.0 using standard protocols (Dominguez et al., 2003). The active and passive residues were selected based on the NMR chemical shift perturbation and the solvent accessibility data. The relative solvent-accessible surface area of residues was calculated with NACCESS v2.1.1 (Hubbard and Thornton, 1993). Only residues with more than 50% relative solvent accessibility of the main chain or the side chain were considered. The following solvent-accessible residues perturbed by complex formation in the NMR titration experiments were designated as active: K251, K286, K287, K305, K308, E309, W316, D322, K324, and C326 (for b2GPI-DV), and L143, W144, and D149 (for LA4). The passive residues, defined as solvent-exposed neighbors of the active residues, were P248, K250, K284, D293, E302, S311, S312, L313, F315, and P325 (for b2GPI-DV), and Q142, N148, P150, and D151 (for LA4). A 6.0 A˚ cutoff of the pairwise backbone rmsd in the interface was used for clustering of 200 refined structures.

Beglov, D., Lee, C.J., De Biasio, A., Kozakov, D., Brenke, R., Vajda, S., and Beglova, N. (2009). Structural insights into recognition of beta2-glycoprotein I by the lipoprotein receptors. Proteins 77, 940–949.

Cardiolipin Binding Assay We used 96 well plates precoated with cardiolipin available as part of the ImmunoWELL cardiolipin IgG test kit (GenBio). We constructed a recombinant b2GPI-DV with HA epitope tag added to the N terminus of b2GPI-DV (HA-b2GPI-DV). HA-b2GPI-DV bound to cardiolipin was detected by HRPconjugated anti-HA-tag antibodies (Abcam ab1265) using 2-20 -azino-di[3-ethylbenzthiazoline] sulfonate (ABTS) chromogenic reagent by measuring OD at 405 nm. All measurements were done in triplicate and corrected to blank before data fitting. The blank contained all components except for HA-b2GPIDV. The assay buffer was 10 mM bis-Tris with 10 mM CaCl2 (pH 7.6). Cardiolipin-coated wells were blocked for 1 hr with 0.5% skim milk and 2% bovine serum albumin in the assay buffer. The binding and inhibition data were fitted to one-site models using GNUPLOT. For the binding studies, 50 ml HA-b2GPI-DV was applied to wells and incubated for 30 min. After washing, 50 ml anti-HA antibody was added to wells and incubated for 30 min before detection. In the second set of experiments, 50 ml HA-b2GPI-DV was first incubated for 30 min with the anti-HA antibodies. Then, the mixture was applied to cardiolipin, incubated for 30 min, washed, and bound HA-b2GPI-DV was detected. For the inhibition studies, 44 nM HA-b2GPI-DV was incubated with LA4 for 30 min. Then, 50 ml of the mixtures was applied to wells and incubated for an additional 30 min. After washing, 50 ml anti-HA antibody was added to wells and incubated for 30 min before detection. In the second set of experiments, 50 ml 3.5 nM of HA-b2GPI-DV was first incubated for 20 min with the anti-HA antibodies. Then, various concentrations of the LA3-4 module-pair were added to the mixture and incubated for an additional 40 min. The mixture was applied to cardiolipin, incubated for 30 min, washed, and bound HA-b2GPI-DV was detected.

Dominguez, C., Boelens, R., and Bonvin, A.M. (2003). HADDOCK: a proteinprotein docking approach based on biochemical or biophysical information. J. Am. Chem. Soc. 125, 1731–1737.

Blacklow, S.C. (2007). Versatility in ligand recognition by LDL receptor family proteins: advances and frontiers. Curr. Opin. Struct. Biol. 17, 419–426. Bouma, B., de Groot, P.G., van den Elsen, J.M., Ravelli, R.B., Schouten, A., Simmelink, M.J., Derksen, R.H., Kroon, J., and Gros, P. (1999). Adhesion mechanism of human beta(2)-glycoprotein I to phospholipids based on its crystal structure. EMBO J. 18, 5166–5174. Clayton, D., Brereton, I.M., Kroon, P.A., and Smith, R. (2000). Three-dimensional NMR structure of the sixth ligand-binding module of the human LDL receptor: comparison of two adjacent modules with different ligand binding specificities. FEBS Lett. 479, 118–122. Cockrell, E., Espinola, R.G., and McCrae, K.R. (2008). Annexin A2: biology and relevance to the antiphospholipid syndrome. Lupus 17, 943–951. Cugno, M., Borghi, M.O., Lonati, L.M., Ghiadoni, L., Gerosa, M., Grossi, C., De Angelis, V., Magnaghi, G., Tincani, A., Mari, D., et al. (2009). Patients with antiphospholipid syndrome display endothelial perturbation. J. Autoimmun. Published online August 3, 2009. 10.1016/j.jaut.2009.07.004.

Fischetti, F., Durigutto, P., Pellis, V., Debeus, A., Macor, P., Bulla, R., Bossi, F., Ziller, F., Sblattero, D., Meroni, P., and Tedesco, F. (2005). Thrombus formation induced by antibodies to beta2-glycoprotein I is complement dependent and requires a priming factor. Blood 106, 2340–2346. Fisher, C., Abdul-Aziz, D., and Blacklow, S.C. (2004). A two-module region of the low-density lipoprotein receptor sufficient for formation of complexes with apolipoprotein E ligands. Biochemistry 43, 1037–1044. Fisher, C., Beglova, N., and Blacklow, S.C. (2006). Structure of an LDLR-RAP complex reveals a general mode for ligand recognition by lipoprotein receptors. Mol. Cell 22, 277–283. Giannakopoulos, B., Passam, F., Ioannou, Y., and Krilis, S.A. (2009). How we diagnose the antiphospholipid syndrome. Blood 113, 985–994. Hagihara, Y., Goto, Y., Kato, H., and Yoshimura, T. (1995). Role of the N- and C-terminal domains of bovine beta 2-glycoprotein I in its interaction with cardiolipin. J. Biochem. 118, 129–136. Hammel, M., Kriechbaum, M., Gries, A., Kostner, G.M., Laggner, P., and Prassl, R. (2002). Solution structure of human and bovine beta(2)-glycoprotein I revealed by small-angle X-ray scattering. J. Mol. Biol. 321, 85–97. Herz, J., Chen, Y., Masiulis, I., and Zhou, L. (2009). Expanding functions of lipoprotein receptors. J. Lipid Res. 50 (Suppl ), S287–S292. Hoshino, M., Hagihara, Y., Nishii, I., Yamazaki, T., Kato, H., and Goto, Y. (2000). Identification of the phospholipid-binding site of human beta(2)-glycoprotein I domain V by heteronuclear magnetic resonance. J. Mol. Biol. 304, 927–939.

ACKNOWLEDGMENTS

Hubbard, S.J., and Thornton, J.M. (1993). NACCES (computer program) (Department of Biochemistry and Molecular Biology, University College London).

This work was supported in part by grants from the American Heart Association and the American Society of Hematology.

Hunt, J., and Krilis, S. (1994). The fifth domain of beta 2-glycoprotein I contains a phospholipid binding site (Cys281-Cys288) and a region recognized by anticardiolipin antibodies. J. Immunol. 152, 653–659.

Received: August 14, 2009 Revised: November 30, 2009 Accepted: December 29, 2009 Published: March 9, 2010

Jankowski, M., Vreys, I., Wittevrongel, C., Boon, D., Vermylen, J., Hoylaerts, M.F., and Arnout, J. (2003). Thrombogenicity of beta 2-glycoprotein I-dependent antiphospholipid antibodies in a photochemically induced thrombosis model in the hamster. Blood 101, 157–162.

REFERENCES

Koike, T., Bohgaki, M., Amengual, O., and Atsumi, T. (2007). Antiphospholipid antibodies: lessons from the bench. J. Autoimmun. 28, 129–133.

Andersen, O.M., Benhayon, D., Curran, T., and Willnow, T.E. (2003). Differential binding of ligands to the apolipoprotein E receptor 2. Biochemistry 42, 9355–9364.

Korschineck, I., Ziegler, S., Breuss, J., Lang, I., Lorenz, M., Kaun, C., Ambros, P.F., and Binder, B.R. (2001). Identification of a novel exon in apolipoprotein E receptor 2 leading to alternatively spliced mRNAs found in cells of the vascular wall but not in neuronal tissue. J. Biol. Chem. 276, 13192–13197.

Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved 375

Structure Structure of the b2GPI-DV/LA Complex

Lefebvre, J. (1982). Viscosity of concentrated protein solutions. Rheologica Acta 21, 620–625. Lim, W., Crowther, M.A., and Eikelboom, J.W. (2006). Management of antiphospholipid antibody syndrome: a systematic review. JAMA 295, 1050–1057. Lutters, B.C., Derksen, R.H., Tekelenburg, W.L., Lenting, P.J., Arnout, J., and de Groot, P.G. (2003). Dimers of beta 2-glycoprotein I increase platelet deposition to collagen via interaction with phospholipids and the apolipoprotein E receptor 20 . J. Biol. Chem. 278, 33831–33838. Ma, K., Simantov, R., Zhang, J.C., Silverstein, R., Hajjar, K.A., and McCrae, K.R. (2000). High affinity binding of beta 2-glycoprotein I to human endothelial cells is mediated by annexin II. J. Biol. Chem. 275, 15541–15548. Maiti, S.N., Balasubramanian, K., Ramoth, J.A., and Schroit, A.J. (2008). Beta-2-glycoprotein 1-dependent macrophage uptake of apoptotic cells. Binding to lipoprotein receptor-related protein receptor family members. J. Biol. Chem. 283, 3761–3766. Masiulis, I., Quill, T.A., Burk, R.F., and Herz, J. (2009). Differential functions of the Apoer2 intracellular domain in selenium uptake and cell signaling. Biol. Chem. 390, 67–73. May, P., Herz, J., and Bock, H.H. (2005). Molecular mechanisms of lipoprotein receptor signalling. Cell. Mol. Life Sci. 62, 2325–2338. Mehdi, H., Naqvi, A., and Kamboh, M.I. (2000). A hydrophobic sequence at position 313-316 (Leu-Ala-Phe-Trp) in the fifth domain of apolipoprotein H (beta2-glycoprotein I) is crucial for cardiolipin binding. Eur. J. Biochem. 267, 1770–1776. Moestrup, S.K., Gliemann, J., and Pallesen, G. (1992). Distribution of the alpha 2-macroglobulin receptor/low density lipoprotein receptor-related protein in human tissues. Cell Tissue Res. 269, 375–382. Murphy, E.J., Joseph, L., Stephens, R., and Horrocks, L.A. (1992). Phospholipid composition of cultured human endothelial cells. Lipids 27, 150–153. North, C.L., and Blacklow, S.C. (2000). Solution structure of the sixth LDL-A module of the LDL receptor. Biochemistry 39, 2564–2571.

Romay-Penabad, Z., Montiel-Manzano, M.G., Shilagard, T., Papalardo, E., Vargas, G., Deora, A.B., Wang, M., Jacovina, A.T., Garcia-Latorre, E., Reyes-Maldonado, E., et al. (2009). Annexin A2 is involved in antiphospholipid antibody-mediated pathogenic effects in vitro and in vivo. Blood 114, 3074–3083. Schwarzenbacher, R., Zeth, K., Diederichs, K., Gries, A., Kostner, G.M., Laggner, P., and Prassl, R. (1999). Crystal structure of human beta2-glycoprotein I: implications for phospholipid binding and the antiphospholipid syndrome. EMBO J. 18, 6228–6239. Sheng, Y., Sali, A., Herzog, H., Lahnstein, J., and Krilis, S.A. (1996). Sitedirected mutagenesis of recombinant human beta 2-glycoprotein I identifies a cluster of lysine residues that are critical for phospholipid binding and anti-cardiolipin antibody activity. J. Immunol. 157, 3744–3751. Sherer, Y., Blank, M., and Shoenfeld, Y. (2007). Antiphospholipid syndrome (APS): where does it come from? Best Pract. Res. Clin. Rheumatol. 21, 1071–1078. Shi, T., Giannakopoulos, B., Yan, X., Yu, P., Berndt, M.C., Andrews, R.K., Rivera, J., Iverson, G.M., Cockerill, K.A., Linnik, M.D., and Krilis, S.A. (2006). Anti-beta2-glycoprotein I antibodies in complex with beta2-glycoprotein I can activate platelets in a dysregulated manner via glycoprotein Ib-IX-V. Arthritis Rheum. 54, 2558–2567. Shoenfeld, Y., Meroni, P.L., and Cervera, R. (2008). Antiphospholipid syndrome dilemmas still to be solved: 2008 status. Ann. Rheum. Dis. 67, 438–442. Sorice, M., Longo, A., Capozzi, A., Garofalo, T., Misasi, R., Alessandri, C., Conti, F., Buttari, B., Rigano, R., Ortona, E., and Valesini, G. (2007). Anti-beta2-glycoprotein I antibodies induce monocyte release of tumor necrosis factor alpha and tissue factor by signal transduction pathways involving lipid rafts. Arthritis Rheum. 56, 2687–2697. Stolt, P.C., and Bock, H.H. (2006). Modulation of lipoprotein receptor functions by intracellular adaptor proteins. Cell. Signal 18, 1560–1571.

Olson, S.T., Bock, P.E., and Sheffer, R. (1991). Quantitative evaluation of solution equilibrium binding interactions by affinity partitioning: application to specific and nonspecific protein-heparin interactions. Arch. Biochem. Biophys. 286, 533–545.

Strasser, V., Fasching, D., Hauser, C., Mayer, H., Bock, H.H., Hiesberger, T., Herz, J., Weeber, E.J., Sweatt, J.D., Pramatarova, A., et al. (2004). Receptor clustering is involved in Reelin signaling. Mol. Cell. Biol. 24, 1378–1386.

Pennings, M.T., van Lummel, M., Derksen, R.H., Urbanus, R.T., Romijn, R.A., Lenting, P.J., and de Groot, P.G. (2006). Interaction of beta2-glycoprotein I with members of the low density lipoprotein receptor family. J. Thromb. Haemost. 4, 1680–1690.

Su, A.I., Wiltshire, T., Batalov, S., Lapp, H., Ching, K.A., Block, D., Zhang, J., Soden, R., Hayakawa, M., Kreiman, G., et al. (2004). A gene atlas of the mouse and human protein-encoding transcriptomes. Proc. Natl. Acad. Sci. USA 101, 6062–6067.

Pennings, M.T., Derksen, R.H., Urbanus, R.T., Tekelenburg, W.L., Hemrika, W., and de Groot, P.G. (2007a). Platelets express three different splice variants of ApoER2 that are all involved in signaling. J. Thromb. Haemost. 5, 1538–1544.

Urbanus, R.T., Derksen, R.H., and de Groot, P.G. (2008a). Current insight into diagnostics and pathophysiology of the antiphospolipid syndrome. Blood Rev. 22, 93–105.

Pennings, M.T., Derksen, R.H., van Lummel, M., Adelmeijer, J., VanHoorelbeke, K., Urbanus, R.T., Lisman, T., and de Groot, P.G. (2007b). Platelet adhesion to dimeric beta-glycoprotein I under conditions of flow is mediated by at least two receptors: glycoprotein Ibalpha and apolipoprotein E receptor 20 . J. Thromb. Haemost. 5, 369–377. Pierangeli, S.S., Liu, S.W., Anderson, G., Barker, J.H., and Harris, E.N. (1996). Thrombogenic properties of murine anti-cardiolipin antibodies induced by beta 2 glycoprotein 1 and human immunoglobulin G antiphospholipid antibodies. Circulation 94, 1746–1751. Pierangeli, S.S., Vega-Ostertag, M., and Harris, E.N. (2004). Intracellular signaling triggered by antiphospholipid antibodies in platelets and endothelial cells: a pathway to targeted therapies. Thromb. Res. 114, 467–476.

Urbanus, R.T., Pennings, M.T., Derksen, R.H., and de Groot, P.G. (2008b). Platelet activation by dimeric beta(2)-glycoprotein I requires signaling via both glycoprotein Ibalpha and Apolipoprotein E Receptor 20 . J. Thromb. Haemost. 6, 1405–1412. van Lummel, M., Pennings, M.T., Derksen, R.H., Urbanus, R.T., Lutters, B.C., Kaldenhoven, N., and de Groot, P.G. (2005). The binding site in {beta}2-glycoprotein I for ApoER20 on platelets is located in domain V. J. Biol. Chem. 280, 36729–36736. Willems, G.M., Janssen, M.P., Pelsers, M.M., Comfurius, P., Galli, M., Zwaal, R.F., and Bevers, E.M. (1996). Role of divalency in the high-affinity binding of anticardiolipin antibody-beta 2-glycoprotein I complexes to lipid membranes. Biochemistry 35, 13833–13842.

Pierangeli, S.S., Vega-Ostertag, M.E., Raschi, E., Liu, X., Romay-Penabad, Z., De Micheli, V., Galli, M., Moia, M., Tincani, A., Borghi, M.O., et al. (2007). Toll-like receptor and antiphospholipid mediated thrombosis: in vivo studies. Ann. Rheum. Dis. 66, 1327–1333.

Yang, X.V., Banerjee, Y., Fernandez, J.A., Deguchi, H., Xu, X., Mosnier, L.O., Urbanus, R.T., de Groot, P.G., White-Adams, T.C., McCarty, O.J., and Griffin, J.H. (2009). Activated protein C ligation of ApoER2 (LRP8) causes Dab1dependent signaling in U937 cells. Proc. Natl. Acad. Sci. USA 106, 274–279.

Pons, J.-L., Malliavin, T.E., and Delsuc, M.A. (1996). Gifa V. 4: A complete package for NMR data set processing. J. Biomol. NMR 8, 445–452.

Zhang, J., and McCrae, K.R. (2005). Annexin A2 mediates endothelial cell activation by antiphospholipid/anti-beta2 glycoprotein I antibodies. Blood 105, 1964–1969.

Raschi, E., Borghi, M.O., Grossi, C., Broggini, V., Pierangeli, S., and Meroni, P.L. (2008). Toll-like receptors: another player in the pathogenesis of the anti-phospholipid syndrome. Lupus 17, 937–942.

Zwaal, R.F., and Schroit, A.J. (1997). Pathophysiologic implications of membrane phospholipid asymmetry in blood cells. Blood 89, 1121–1132.

376 Structure 18, 366–376, March 10, 2010 ª2010 Elsevier Ltd All rights reserved