The 1.4 Å Crystal Structure of the Human Oxidized Low Density Lipoprotein Receptor Lox-1*

The lectin-like oxidized low density lipoprotein receptor-1 (Lox-1) mediates the recognition and internalization of oxidatively modified low density lipoprotein by vascular endothelial cells. This interaction results in a number of pro-atherogenic cellular responses that probably play a significant role in the pathology of atherosclerosis. The 1.4 Å crystal structure of the extracellular C-type lectin-like domain of human Lox-1 reveals a heart-shaped homodimer with a ridge of six basic amino acids extending diagonally across the apolar top of Lox-1, a central hydrophobic tunnel that extends through the entire molecule, and an electrostatically neutral patch of 12 charged residues that resides next to the tunnel at each opening. Based on the arrangement of critical binding residues on the Lox-1 structure, we propose a binding mode for the recognition of modified low density lipoprotein and other Lox-1 ligands.

The lectin-like oxidized low density lipoprotein receptor-1 (Lox-1) mediates the recognition and internalization of oxidatively modified low density lipoprotein by vascular endothelial cells. This interaction results in a number of pro-atherogenic cellular responses that probably play a significant role in the pathology of atherosclerosis. The 1.4 Å crystal structure of the extracellular C-type lectin-like domain of human Lox-1 reveals a heart-shaped homodimer with a ridge of six basic amino acids extending diagonally across the apolar top of Lox-1, a central hydrophobic tunnel that extends through the entire molecule, and an electrostatically neutral patch of 12 charged residues that resides next to the tunnel at each opening. Based on the arrangement of critical binding residues on the Lox-1 structure, we propose a binding mode for the recognition of modified low density lipoprotein and other Lox-1 ligands.
Atherosclerosis is understood to be a disease of chronic vascular inflammation resulting from the interaction of oxidatively modified low density lipoprotein (oxLDL) 1 with macrophages, lymphocytes, and various cellular components of artery walls, including vascular endothelial cells (1,2). oxLDL causes vascular endothelial cell activation and dysfunction, resulting in pro-inflammatory responses, pro-oxidative conditions, and apoptosis, all of which are pro-atherogenic. The lectin-like oxidized low-density lipoprotein receptor-1 (Lox-1) has been characterized as the primary receptor for oxLDL on the surface of vascular endothelial cells and is up-regulated in atherosclerotic lesions (3,4). It is also expressed to a lower extent in macrophages, smooth muscle cells, and dendritic cells. Upon recognition of oxLDL, Lox-1 is observed to initiate oxLDL internalization and degradation as well as the induction of a variety of pro-atherogenic cellular responses including a reduction of nitric oxide (NO) release (5), secretion of monocyte chemoattractant protein-1 (MCP-1) (6), production of reactive oxygen spe-cies (7), expression of matrix metalloproteinases-1 and -3 (8), monocyte adhesion (6), and apoptosis (9). In addition, Lox-1 expression is up-regulated by various elicitors of vascular stress, including oxLDL (9), reactive oxygen species (10) and fluid shear stress (11), suggesting that Lox-1 may help amplify oxLDL-induced vascular dysfunction.
Moreover, Lox-1 is a multifunctional receptor involved in several other cellular events. In addition to binding oxLDL, Lox-1 is reported to be a dendritic cell receptor for the 70-kDa heat shock protein involved in antigen cross-presentation to naive T cells (12) and a receptor for advanced glycation end products (13), monocytes (14), apoptotic cells (15), and both Gram-negative and Gram-positive bacteria (16).
Lox-1 is a member of the scavenger receptor family, a structurally diverse group of cell surface receptors of the innate immune system that recognize modified lipoproteins. It is a disulfide-linked homodimeric type II transmembrane protein with a short 34-residue cytoplasmic region, a single transmembrane region, and an extracellular region consisting of an 80residue domain predicted to be a coiled coil followed by a 130-residue C-terminal C-type lectin-like domain (CTLD) (3,17). Deletion analysis has localized oxLDL recognition to the highly conserved (61-83% sequence identity) CTLD of Lox-1 (18). The human Lox-1 CTLD shares its highest sequence identity (35-42%) with the leukocyte-expressed immunoreceptors dectin-1, NKG2D, DC-SIGN, and DC-SIGNR. Like NKG2D, Lox-1 does not have any of the conserved calcium binding residues observed in classic CTLDs, such as mannose binding protein, and is not known to bind carbohydrate.
The oxidative modification of low density lipoprotein (LDL) results in an increased net negative charge and potential conformational rearrangements of ApoB100 domains (19,20). Although several positively charged Lox-1 residues are known to play a role in the recognition of modified LDL (18,21,22), the lack of a three-dimensional structure for Lox-1 has hindered our understanding of the binding mode(s) employed by Lox-1. A detailed understanding of this interaction could be of significant medical interest, because antagonists could potentially mitigate the progression of atherosclerosis. In an effort to begin elucidating the molecular recognition mechanisms of Lox-1, we have determined the crystal structure of the human Lox-1 CTLD in two crystal forms refined to 1.4 and 3.0 Å respectively.

MATERIALS AND METHODS
Plasmid Construction-cDNA encoding human Lox-1 was amplified by PCR from a pDNR-LIB plasmid (Open Biosystems) to create the constructs Lox1R136 and Lox1A142 (residues 136 -273 and 142-273, respectively). These constructs were respectively subcloned into a pET15b expression vector (Novagen) downstream of a histidine tag and a thrombin cleavage site. The cDNA for leaderless Escherichia coli chaperone/disulfide isomerase DsbC was amplified by PCR from a pET40 plasmid (Novagen) and subcloned downstream of each Lox-1 construct. A separate ribosome binding site was added just upstream of DsbC to facilitate translation. Plasmids pET15b-Lox1R136-DsbC and pET15b-Lox1A142-DsbC were transformed into Origami B (DE3) E. coli (Novagen). Human Lox-1 R136 and A142 were expressed using 0.4 mM isopropyl ␤-D-thiogalactoside in E. coli cells grown in LB medium containing kanamycin, tetracycline, and carbenicillin at 23°C, with 250 rpm agitation, for 20 h.
Purification and Thrombin Digestion-Each of the human Lox-1 fragments were purified by passage through a nickel-nitrilotriacetic acid column followed by a cation exchange column. In brief, cells were harvested by centrifugation and resuspended in sonication buffer (50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, and 15 mM imidazole), sonicated on an ice-water bath, and centrifuged for 25 min at 22,000 ϫ g. Supernatant was applied to a nickel-nitrilotriacetic acid column (Qiagen). The column was then washed with a linear gradient from 20 mM to 250 mM imidazole. The elutions were pooled and concentrated with a 10-kDa cutoff ultrafiltration unit (Amicon), and the buffer was changed to phosphate-buffered saline for thrombin digestion. A ratio of 0.1 unit of thrombin to 1 mg of Lox-1 was applied to the sample and incubated on room temperature for 2 h. Thrombin-digested protein was applied to a nickel-nitrilotriacetic acid column, and the flow-through was loaded on a HiPrep S 16/10 column (Amersham Biosciences) equilibrated with 25 mM HEPES, pH 7.5. Protein was eluted with a linear gradient from 100 mM to 500 mM NaCl. Fractions containing protein were concentrated to 20 mg/ml with an ultrafiltration unit (Amicon). Covalent dimerization of the Arg-136 fragment was confirmed by SDS-PAGE and mass spectrometry under reducing and nonreducing conditions.
Crystallization-Crystals of Lox-1 were grown by vapor diffusion using 2 l of 20 mg/ml protein and an equal volume of precipitant and were fully grown within 24 h. Lox-1R136 crystals grew from two different crystallization conditions yielding the same monoclinic spacegroup (C2) and identical cell dimensions (a ϭ 71.0 Å, b ϭ 49.1 Å, c ϭ 76.3 Å, ␤ ϭ 98.5°). In the first condition, crystals were grown at 4°C using a precipitant solution containing 0.1 M Bicine, pH 9.0, 5% dioxane, and 10% polyethylene glycol 10,000 (w/v). For the second condition (referred to as "dioxane free" in the text), the same protein was crystallized at room temperature in a precipitant solution containing 0.2 M ammonium acetate, 0.1 M sodium acetate, pH 4.6, and 30% polyethylene glycol 3000. Crystals of Lox1A142 grew in a trigonal space group (P3 1 ) by the addition of 0.1 M ammonium acetate, 0.1 M Bis-Tris, pH 5.5, and 17% polyethylene glycol 10,000 at room temperature.
Data Collection and Structure Determination-Crystals were briefly transferred to reservoir solutions containing an additional 30% glycerol before cryofreezing. A data set from the first crystal condition of human Lox-1 R136 with 1.4 Å resolution was collected on a MAR225 CCD detector at the Southeast Regional Collaborative Access Team (SER-CAT) 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory. Data sets from the dioxane-free monoclinic crystals of Lox-1 R136 and trigonal crystals of Lox-1A142 were collected inhouse with a Rigaku RU-H3R generator and an RAXIS IV detector. Data sets were processed with HKL2000 (23). The structure of Lox-1 Arg-136 was solved by the molecular replacement method using MOL-REP (24). A composite search model was generated by combining the CTLD structures of hNKG2D (PDB entry 1MPU), hDC-SIGN (PDB entry 1K9I), and hCD69 (PDB entry 1Q03) into a single polyalanine model using SwissModel (25). Refinement was performed in CNS using a maximum likelihood target, bulk solvent correction, an overall anisotropic B-factor, and individual B-factors combined with iterations of manual rebuilding (26). No noncrystallographic symmetry restraints were applied. Water molecules were assigned using CNS. The structure of human Lox-1A142 (trigonal crystal) was also determined by molecular replacement with MOLREP. A polyalanine version of the refined Lox-1 Arg-136 monomer A with loop L8 deleted was used as the search model. The eight Lox-1A142 monomers were refined with CNS using grouped B-factors and no noncrystallographic symmetry restraints. Manual rebuilding and adjustment of the Lox-1 structures were performed using graphics program O (27). Data processing and refinement statistics for all three crystals structures are given in Table I. Molecular figures were created using PyMOL (28). Superpositions and buried surface area was calculated using LSQMAN (29) and SURFACE (24), respectively. The cavity volume was determined using VOIDOO with a 1.2 Å probe radius (30).

RESULTS AND DISCUSSION
Overview of the Structure-Fully soluble human Lox-1 CTLD fragments were co-overexpressed with the E. coli chaperone/ disulfide isomerase DsbC in the cytoplasm of Origam cells, a  mutant E. coli strain with a non-reducing cytoplasm (see "Materials and Methods"). This expression method provided up to 9 mg of Lox-1 per liter of culture, negating the need for refolding from inclusion bodies despite the presence of seven disulfide bonds per dimer. A CTLD fragment (residues 136 -273) was crystallized in the monoclinic spacegroup C2, with one disulfide-bonded dimer in the asymmetric unit. Data were collected with (1.4 Å) and without (2.1 Å) 5% dioxane. A second Lox-1 fragment truncated two residues beyond the interchain disulfide bond (residues 142-273) crystallized in a trigonal spacegroup with eight independent molecules (chains A-H) in the asymmetric unit and diffracted to 3.0 Å. Both crystal forms were phased using molecular replacement (see Table I and "Materials and Methods"). Unless otherwise stated, the descriptions in this article refer to the high resolution monoclinic form with dioxane. Human Lox-1 forms a heart-shaped homodimer with a tunnel running through the center of the molecule (Fig. 1, a and b). As expected from the sequence, the Lox-1 monomer has a CTLD fold consisting of two antiparallel ␤-sheets, ␤0-␤1-␤5-␤1Ј and ␤2Ј-␤2-␤3-␤4, respectively, flanked by two ␣-helices, ␣1 and ␣2 (Fig. 1, a and b). The fold is further stabilized by three conserved intrachain disulfide bonds (Cys-144 -Cys-155, Cys-172-Cys-264 and Cys-243-Cys-256). A cysteine at position 140, present only in human Lox-1, forms an interchain disulfide between the monomers at the N terminus of the CTLD. In terms of structure, Lox-1 is highly homologous to a broad range of carbohydrate-and protein-binding long-form CTLDs. Superpositions generally resulted in root-mean-square deviations of 1.6 Å or less for 100 C␣ atoms, despite low sequence identities, with Lox-1 of around 20% for several CTLDs. The most notable conformational differences in Lox-1 are observed in the ␤2Ј-␤3 loop and the ␣2 helix. The second half of the Lox-1 loop (L8, following ␤2Љ) has two completely different conformations in the A and B monomers in the asymmetric unit, neither of which resembles any of the loops seen in any other known CTLD structures. Monomer B makes crystal lattice contacts at loop L8, suggesting an influence from crystal packing. In addition, both the C-terminal end of helix ␣2 and the following loop (L5) are oriented further from the main fold than in most other CTLDs. These latter characteristics contribute to an unusual dimer interface for Lox-1 described below.
Lox-1 Homodimer-The general orientation of the monomers within the Lox-1 homodimer is similar to that observed in the dimeric NK cell receptors NKG2D, CD69, CD94, and Ly49A in that the N termini, the ␤0 ␤-strands, and the C-terminal ends of the ␣2-helices approach each other with nearly 2-fold symmetry (169.8°). However, two major differences are observed in the human Lox-1 dimer structure. First, relative to the NKG2D, CD69, and CD94 homodimers, the Lox-1 monomers are translated away from each other by 4 -6 Å in the plane of the two ␤0 strands, in the direction perpendicular to the long axis of the dimer. This translation is facilitated in part by an elongation of the L5 loop C-terminal to the ␣2-helix by two to four residues relative to other dimer-forming CTLDs. As a result, the ␤0 ␤-strands do not directly hydrogen-bond to each other to form a continuous ␤-sheet across the dimer interface as seen in other CTLD dimer structures (Fig. 1b). A similar translation is observed in both the Ly49C and Ly49I dimer interfaces, but that translation is in the opposite direction relative to Lox-1, and a continuous ␤-sheet across the interface is still formed. Second, as described in detail below, this particular dimeric association forms a 20 Å mostly nonpolar tunnel through the center of the dimer interface (Fig. 2, a and b). This is a unique feature, not yet observed in any other CTLD dimers. The tunnel is created largely by the monomer translation noted above and the elongated L5 loop that helps keep the ␣2 helices separated from each other.
The dimer interface is composed primarily of nonpolar loop residues and results in a buried surface area of 1945 Å 2 . The distinctive Lox-1 dimer arrangement is stabilized by multiple factors, including the following: 1) a disulfide bond between Cys-140 in monomers A and B, 2) a completely conserved Trp-150 side chain unique to Lox-1 that forms an integral part of the hydrophobic core just below the tunnel, 3) two completely conserved salt bridges (Asp-147A-His-151B and Asp-147B-His-151A), 4) seven protein-protein hydrogen bonds, and 5) five buried interface waters involved in an extensive hydrogen-bonding network with a total of nine protein atoms. These five buried waters reside in a completely enclosed internal cavity of 137 Å 3 just below the tunnel floor. The above interactions at the dimer interface are expected to be conserved across species except for the disulfide bond at position 140. Other Lox-1 sequences have phenylalanine, tyrosine, or serine at this position.
Hydrophobic Tunnel-The Lox-1 tunnel is 7-8 Å in diameter except for a constriction caused by the side chains of Ile-149A, Ile-149B, and Tyr-197A that narrows the middle of the tunnel to a diameter of 4 Å (Fig. 2a). In addition, dimer asymmetry narrows the opening at one end of the tunnel to 3 Å. Adjacent to the constriction, a small nonpolar side pocket opens into monomer A. Monomer B has no side pocket on its half of the tunnel because Phe-158B plugs the pocket entrance. The amino acid side chains that line the inside of the tunnel are essentially nonpolar and highly conserved (Figs. 2a and 3). Tyr-197A and Tyr-197B are the only residues that provide a potential hydrogen bonding partner through their side chains. It is intriguing that residue 197 is a histidine in the other five known Lox-1 orthologs. Several other residues are situated to provide main chain hydrogen bonding partners primarily through backbone carbonyls to potential tunnel occupants. The tunnel mouths are lined with conserved, mostly hydrophilic side chains (Figs. 2b and 3).
A comparison of the 3 Å trigonal crystal form of Lox-1 with the high resolution monoclinic crystal form reveals apparent tunnel plasticity. The tunnel side pocket found in the monoclinic form is not observed in any of the four dimers in the trigonal form. Furthermore, a new similarly sized nonpolar pocket located between the Tyr-197A and Tyr-197B side chains is observed in the AB dimer of the trigonal form. These changes are facilitated by different rotamers for the Phe-158A and Tyr-197B side chains. In addition, the dimers in the trigonal crystal have less asymmetry (176.8 -178.7°compared with 169.8°), resulting in more symmetric tunnel openings.
Bound Dioxane-Our Lox-1 structure was serendipitously observed to have a dioxane molecule bound within the larger tunnel chamber (Fig. 2c). One dioxane oxygen is hydrogen bonded to the main chain nitrogen of Phe-158. The other oxygen is hydrogen-bonded through two water molecules to the hydroxyl of Tyr-197A and the main chain carbonyls of Asp-147A and Ala-194B. The dioxane ring packs against the side chains of Phe-158A, Leu-157A, Tyr-197B, and Ile-149A. A dioxane-free crystal structure of Lox-1 in the same space group reveals no conformational changes to Lox-1 relative to the dioxane-bound structure. Hence, dioxane does not influence the symmetry of the dimer nor does it bind with an induced fit.
Potential Tunnel Ligands-The apolar Lox-1 tunnel appears large enough to accommodate a cholesterol molecule, a fatty acid chain, or possibly a six-to seven-residue nonpolar peptide. The Lox-1 tunnel may recognize the lipophilic portion of Stereoview of the tunnel outer molecular surface viewed from the top (a; orientation similar to Fig. 1b) and from the tunnel end (b; as in Fig. 1a). The semitransparent tunnel molecular surface is colorcoded with green for carbon, red for oxygen, and blue for nitrogen. Selected residues and bound dioxane are shown as stick models. c, stereoview of the interactions bound dioxane makes in the Lox-1 tunnel including the 1.4 Å 2F o Ϫ F c electron density map contoured to 1.0 . Hydrogen bonds are represented by dotted lines and the electron density for dioxane is colored dark blue for clarity.
"core aldehydes" that have derivatized lysine side chains of ApoB100. Core aldehydes are formed in oxLDL when a polyunsaturated fatty acid chain of a phospholipid or a cholesterol ester undergoes peroxidation, resulting in a short chain aldehyde still esterified to its parent lipid (32,33). Further oxidation of core aldehydes may also form potential ligands. The scavenger receptor CD36 was recently shown to specifically recognize phospholipids that incorporate a terminal ␥-hydroxy-(or oxo-) ␣,␤-unsaturated carbonyl at the sn-2 position (34). In addition, short chain peroxidation products such as malondialdehyde, 4-hydroxy-2-nonenal, and acrolein can form ApoB100 adducts with lipophilic, ringed moieties that may specifically bind the Lox-1 tunnel as dioxane does (35). Lox-1 recognition of ApoB100 adducts is consistent with a report that bovine Lox-1 recognizes delipidated ApoB100 from oxLDL (36). Moreover, similar epitopes are formed on the surfaces of apoptotic cells (37) and advanced glycation end products-modified proteins (35), raising the possibility that the tunnel could participate in the recognition of these Lox-1 ligands. Potential peptide ligands for the tunnel include bacterial peptidoglycan or ApoB100 oxidation fragments still attached to oxLDL.
Acid-Base Patch-The sides of the dimer are decorated with 14 acid-base residue pairs, each consisting of a surface exposed acidic and basic residue respectively that are close enough to be able to form a salt bridge. On each monomer, six pairs cluster to a region referred to as the acid-base patch, located adjacent to each tunnel opening (Fig. 4a). A seventh pair, Asp-147-His-151, straddles the dimer interface just below each tunnel opening and is the only invariant acid-base pair. Mapping of the other five known Lox-1 sequences onto the human Lox-1 structure suggests that the total number of acid-base pairs remains between five and seven, including five new putative pairs that would cluster to the same region. The electrostatically neutral character of this highly charged patch suggests the potential for interacting with the zwitterionic phospholipid head groups commonly observed on the surface of LDL, such as phosphatidylcholine, sphingomyelin, or phosphatidylethanolamine.
Basic Residues Critical for Ligand Recognition-The top of Lox-1 (opposite the N and C termini) is saddle-shaped and has a distinctly basic character. A ridge of six arginines (229, 231, and 248 from each monomer) and two histidines (226) runs diagonally across the entire saddle (Fig. 4b). Two more arginines, 208 and 209, reside at each end of the saddle. Of these six residues, only Arg-209 and Arg-229 are completely conserved as basic residues in other species of Lox-1 (Fig. 3). However, mapping of Lox-1 ortholog sequences to the crystal structure reveals that each Lox-1 species has two to four basic residues at each saddle end and two to three basic residues (four to six per dimer) along the basic ridge in addition to one or more histidines. Thus, the overall basic character of the saddle is preserved along the top and ends. Mutation studies of both human and bovine Lox-1 have implicated three clusters of conserved basic residues as playing a significant role in recognition of acetylated LDL or oxLDL, respectively (18,21,22). The Lox-1 crystal structure reveals that the first cluster, residues 208 -209, maps to the saddle end and the second cluster, including residues 229, 231, and 248, maps to the basic ridge (Figs. 3 and 4c). Additional basic residues mapping to the surface region as cluster 1 include Lys-210, present in bovine, rabbit, porcine, and murine Lox-1, and Arg-236 and Arg-237, present in bovine and rabbit Lox-1, respectively. The third cluster involves Lys-266 and Lys-267, which reside at the bottom edge of the acidbase patch near the N and C termini (Figs. 3 and 4c). Simultaneous point mutations of two to four residues within any one of these three clusters are observed to cause a drastic reduction of Lox-1 binding to modified LDL (18,21,22). This indicates a cooperative binding mechanism that requires two or more residues from each of the three basic clusters. Oxidatively modified LDL has a greater negative charge because of the loss of lysine, arginine, and histidine side chains from ApoB100 and the addition of acidic adducts such as carboxymethyllysine or nitrotyrosine (19,35,38). Thus, it is likely that these critical basic residues on Lox-1 (especially clusters 1 and 2) interact directly with modified ApoB100 on oxLDL. It is noteworthy that advanced glycation end product modification of proteins creates a similar negative charge, as does the exposure of phosphatidylserine on apoptotic cells and the presence of phosphate groups within the repeating units of various types of teichoic and lipoteichoic acids found on the surfaces of Grampositive bacteria (35,39). Therefore, it would not be surprising to discover that the conserved basic regions of Lox-1 also play a role in the recognition of these other Lox-1 ligands. FIG. 3. Sequence alignment of the CTLDs of Lox-1 orthologs. Cysteines are highlighted in yellow, basic residues in blue, histidines in cyan, acidic residues in red, and completely conserved residues in brown. Secondary structures are labeled as in Fig. 1, a and b. ␣-Helices are blue cylinders, and ␤-strands are light green arrows. Key residues are labeled as follows: tunnel resides (including data from both the monoclinic and trigonal crystal forms) are marked in red with T for tunnel, M for tunnel mouth, and P for tunnel pocket. Basic saddle residues are labeled with a blue B. Residues in the hydrophobic saddle patch and the acid-base patch are designated by a brown f and a magenta •, respectively.

FIG. 4. Surface features of human
Lox-1. a, molecular surface view of Lox-1 highlighting the acid-base patch on monomer A. A dotted line encloses the patch. Basic residues are blue, acid residues are red, nonpolar residues are brown, histidine is slate blue, and the rest are green. Top view (b) and side view (c) of the dimer (same orientations as in Fig. 1, b and a,  respectively). The three conserved basic patches are outlined by dotted lines in c. Residues 210, 236, and 237 that have arginine or lysine side chains in ortholog Lox-1 structures are colored cyan in b and c.
Each half of the saddle also contains a relatively flat, starshaped hydrophobic patch of 15 highly conserved nonpolar residues surrounding an invariant Glu-254 (Figs. 3 and 4b). It is interesting that a mutation of the conserved Tyr-238 to serine within the hydrophobic saddle patch next to Glu-254 and the first basic cluster has been observed to dramatically reduce recognition of acetylated LDL by human Lox-1 (40). In contrast, mutation of Glu-254 was found to have no effect on acetylated LDL binding (22).
Model for Binding oxLDL-Based on the placement of the three conserved basic clusters on the three-dimensional structure of human Lox-1 (Fig. 4c), we propose that one entire face of the dimer interacts with modified LDL to involve all three basic regions simultaneously. Although each basic cluster is slightly recessed from the Lox-1 dimer face, they would probably interact with modified portions of ApoB100 that rise slightly above the LDL phospholipid surface. An alternative recognition mode would involve an entire monomer of the Lox-1 dimer binding in a "canyon" between two or more negatively charged ApoB100 domains on the surface of modified LDL. However, the former binding mode is more attractive for several reasons. This mode potentially enables up to five basic clusters to interact with oxLDL at one time, not just three (Fig. 4c). Furthermore, the tunnel and part of the acid-base patch are situated such that they could potentially aid in the recognition of oxLDL (Fig. 4c). As suggested above, the tunnel may specifically bind ApoB100 adducts. In addition, the acidbase patch, particularly Lys-171, Glu-170, Lys-167, Glu-166, or even Lys-266, Asp-176, Lys-267, and Glu-153 may facilitate direct interaction with the zwitterionic phospholipid surface of oxLDL. However, the role of the tunnel and the acid-base patch in ligand recognition remain to be investigated by mutagenesis studies. Finally, the relatively flat (and slightly concave) faces of the dimer would enable Lox-1 to recognize a variety of negatively charged ligands much larger than itself, such as oxLDL, with limited restrictions on orientation. This type of interaction would suggest a multitude of binding sites on Lox-1 ligands that could serve to increase avidity and enable several receptors to bind one ligand concurrently. Interaction with large ligands such as apoptotic cells could concentrate Lox-1 in a localized region of the cell surface, whereas smaller ligands such as oxLDL may bind both faces of Lox-1 simultaneously, thereby forming a two-dimensional array of receptor/ligand complexes on the cell surface. Either clustering method could potentially serve in a cellular signaling mechanism.