Recognition of α2-Macroglobulin by the Low Density Lipoprotein Receptor-related Protein Requires the Cooperation of Two Ligand Binding Cluster Regions*

The low density lipoprotein receptor-related protein (LRP) is a scavenger receptor that binds several ligands including the activated form of the pan-proteinase inhibitor α2-macroglobulin (α2M*) and amyloid precursor protein, two ligands genetically linked to Alzheimer's disease. To delineate the contribution of LRP to this disease, it will be necessary to identify the sites on this receptor which are responsible for recognizing these and other ligands to assist in the development of specific inhibitors. Structurally, LRP contains four clusters of cysteine-rich repeats, yet studies thus far suggest that only two of these clusters (clusters II and IV) bind ligands. Identifying binding sites within LRP for certain ligands, such as α2M*, has proven to be difficult. To accomplish this, we mapped the binding site on LRP for two inhibitors of α2M* uptake, monoclonal antibody 8G1 and an amino-terminal fragment of receptor-associated protein (RAP D1D2). Surprisingly, the inhibitors recognized different clusters of ligand binding repeats: 8G1 bound to repeats within cluster I, whereas the RAP fragment bound to repeats within cluster II. A recombinant LRP mini-receptor containing the repeats from cluster I along with three ligand binding repeats from cluster II was effective in mediating the internalization of125I-labeled α2M*. Together, these studies indicate that ligand binding repeats from both cluster I and II cooperate to generate a high affinity binding site for α2M*, and they suggest a strategy for developing specific inhibitors to block α2M* binding to LRP by identifying molecules capable of binding repeats in cluster I.

structurally related endocytic receptors. LRP, like all members of the LDL receptor gene family, consists of five common structural units: 1) clusters of ligand binding cysteine-rich repeats; 2) epidermal growth factor (EGF) receptor like cysteine-rich repeats; 3) YWTD domains; 4) a single membrane-spanning segment; and 5) a cytoplasmic tail that harbors two NPXY motifs. The ligand binding regions in LRP occur in four clusters (clusters I-IV) containing between 2 and 11 individual ligand binding repeats. Most of the ligands for LRP for which the binding sites have been mapped interact with ligand binding repeats in clusters II and IV (1)(2)(3)(4).
The first member of this receptor family to be identified was the LDL receptor, which plays an important role in cholesterol homeostasis (5). In contrast, other members of this gene family seem to have more diverse functions. In the case of LRP, its biological role includes a function in lipoprotein metabolism (6), in the homeostasis of proteinases and proteinase inhibitors (7,8), in the cellular entry of viruses (9) and toxins (10), in activation of lysosomal enzymes (11), in cellular signal transduction (12), and in the pathology of Alzheimer's disease (13).
LRP contributes to the pathology of Alzheimer's disease by influencing both the production (13) and clearance (14) of the ␤ amyloid peptide (A␤). This molecule is a 40 -42-amino acid peptide that is the prominent component of senile plaques (15,16), which are a major pathological hallmark of Alzheimer's disease (17). A␤ is derived from proteolytic processing of a ubiquitous transmembrane protein termed ␤-amyloid precursor protein (18). The association of LRP with amyloid precursor protein isoforms containing a Kunitz-type proteinase inhibitor domain alters amyloid precursor protein processing, leading to increased A␤ production (13,19). At the same time, the A␤ peptide binds avidly to LRP ligands, especially ␣ 2 M and the activated form of the molecule (termed ␣ 2 M*). LRP-mediated clearance of the ␣ 2 M*⅐A␤ complex contributes to a reduction in A␤ levels (14,20). Not only does ␣ 2 M* appear to mediate clearance of the A␤ peptide, but the association of ␣ 2 M* with LRP on neurons also leads to an influx of calcium via N-methyl-D-aspartate receptors (12).
To delineate further the contribution of LRP to pathological processes such as Alzheimer's disease, it will be important to define the structural basis for the ␣ 2 M* interaction with LRP to assist in the development of specific inhibitors of this process. Defining regions within LRP responsible for ␣ 2 M* binding has proven to be difficult, and conflicting studies have been reported (4,21). The objective of the current investigation was to identify the specific region on LRP responsible for binding ␣ 2 M*. To accomplish this, we identified the binding sites on LRP for two inhibitors of ␣ 2 M* uptake, monoclonal antibody 8G1 and an amino-terminal fragment of RAP (RAP D1D2). The studies indicate that the cooperation of ligand binding repeats * This work was supported in part by National Institutes of Health Grants HL50784 and HL54710 (to D. K. S.) and Scientist Development Award 0030115N from the American Heart Association (to I. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  from both cluster I and II is required for the high affinity binding of ␣ 2 M* to LRP.

EXPERIMENTAL PROCEDURES
Proteins and Antibodies-LRP was isolated from human placenta as described by Ashcom et al. (22). Human ␣ 2 M was isolated from plasma and activated with trypsin as described (22). Human RAP and RAP fragments D1D2 and D4 were expressed in bacteria as fusion proteins with glutathione S-transferase and purified as described previously (23). Pro-urokinase provided by Jack Henkin (Abbott Laboratories) was activated by incubation with plasmin-Sepharose, and high molecular weight urokinase (uPA) was purified over a benzamadine-Sepharose column. Plasminogen activator inhibitor type I (PAI-1) was generously provided by Dan Lawrence (American Red Cross). R2629, a rabbit polyclonal IgG against LRP, was affinity purified over LRP-Sepharose as described (10). Monoclonal antibodies 5A6 and 8G1 have been raised against human LRP and described previously (7). Cells producing the anti-myc IgG 9E10 were obtained from the American Type Culture Collection (Rockville), and the IgG was purified by chromatography on protein G-Sepharose.
Proteins were labeled with 125 I to a specific activity ranging from 2 to 10 Ci/g protein using IODO-GEN (Pierce Chemical Co.). To generate the radiolabeled uPA⅐PAI-1 complex 125 I-uPA was incubated with PAI-1 (1:1.5 molar ratio) for 30 min at room temperature in Tris-buffered saline, pH 8.0. To generate the iodinated ␣ 2 M⅐trypsin complex (designated ␣ 2 M*) 125 I-␣ 2 M was incubated with trypsin (1:4 molar ratio) for 5 min at room temperature, then the soybean proteinase inhibitor was added, and the 125 I-␣ 2 M* was purified by size exclusion chromatography.
Solid Phase Binding Assay-The binding of ␣ 2 M* to LRP immobilized on plastic was performed essentially as described earlier (23). Microtiter wells were coated with LRP (10 g/ml in Tris-buffered saline, pH 8.0, 100 l/well) overnight and then blocked with 3% bovine serum albumin in Tris-buffered saline. 5 nM 125 I-␣ 2 M* was added to the wells in the absence or presence of the indicated antibody (300 g/ml) and incubated for 4 h at 37°C. After incubation the microtiter wells were washed and counted. Nonspecific binding was measured in presence of 1 M RAP and subtracted.
Construction of cDNAs for LRP Mini-receptors and Soluble Fragments-cDNA of human LRP (13) was used as a template. To generate expression vectors for the soluble fragments of LRP a commercial plasmid pSecTagB (Invitrogen) was utilized. The plasmid contains Ig k-chain leader sequence to ensure the secretion of expressed protein. cDNAs encoding selected portions of LRP (see Fig. 3) were produced by polymerase chain reaction and subcloned into pSecTagB using HindIII and XhoI sites in the case of sLRP1, sLRP1a, and sLRP1b. BamHI and XhoI sites were used for subcloning of sLRP2, sLRP3, and sLRP4 fragments. sLRP1 encodes amino acids 1-172 of mature LRP peptide. The other fragments used in this study contain the following amino acid sequences of LRP: sLRP1a, 1-786; sLRP1b, 1-955; sLRP2, 787-1244; sLRP3, 2462-3004; sLRP4, 3274 -3843. Recombinant proteins sLRP1, sLRP1a, and sLRP1b contain extra 13 amino acids (AAQPARRARRTKL) encoded by a polylinker sequence at their amino terminus. SLRP2, sLRP3, and sLRP4 contain 19 amino acids (AAQPARRARRTKLGTELGS) encoded by a polylinker sequence. All soluble LRP fragments have a myc epitope and polyhistidine tag at the carboxyl terminus for detection. The construction of cDNAs for the mini-receptors is illustrated in Fig. 4. First, a polymerase chain reaction product that encodes the entire region from after the cluster IV of ligand binding domains to the carboxyl terminus of LRP (residues 3844 -4525) was subcloned into the XhoI and XbaI sites of pSecTagB. This plasmid was designated pSecTagLC (light chain). Then, cDNA encoding the selected portion of LRP containing ligand binding repeats was inserted into pSecTagLC using HindIII and XhoI sites for mini-receptors mLRP1 and mLRP1b or BamHI and XhoI sites for mLRP2, mLRP3, and mLRP4.
Expression of Recombinant Forms of LRP-Secreted fragments of LRP containing ligand binding domains were transiently expressed in COS-1 cells using FuGENE 6 transfection reagent (Roche Molecular Biochemicals, Indianapolis) according to the manufacturer's protocol. Cells growing in 100-mm dishes (ϳ50% confluence) were transfected in serum-containing medium with 30 g of pSecTagB carrying cDNA for various LRP fragments. Cells were washed 24 h after transfection, and the medium was changed for plain Dulbecco's modified Eagle's medium supplemented with 1% Nutridoma ® -NS medium supplement (Roche Molecular Biochemicals). This medium was harvested after 48 h of incubation, subjected to immunoblot analysis using anti-myc antibody to detect recombinant proteins, and used in further experiments. LRP mini-receptors were expressed in CHO 13-5-1 cells using FuGENE 6 transfection reagent (Roche Molecular Biochemicals) as follows. Cells were plated in six-well plates (5 ϫ 10 4 cells/well) 24 h prior to the transfection. Transfections were performed using 2 g of DNA/well in 1.5 ml of serum-containing medium. 36 -40 h after the beginning of transfection cells were washed and used in the ligand internalization experiments.
Cellular-mediated Ligand Internalization and Degradation Assays-Cellular internalization and degradation assays were generally conducted as described previously (24). Human foreskin fibroblasts were seeded into 12-well culture dishes (5 ϫ 10 4 cells/well) and grown in Dulbecco's modified Eagle's medium supplemented with 10% bovine calf serum and penicillin/streptomycin for 2 days. Cells were washed and incubated in serum-free medium for 1 h before the assay. 0.4 ml of Dulbecco's modified Eagle's medium containing 1% bovine serum albumin and various antibodies at selected concentrations was added to the corresponding wells and incubated for 15 min at 37°C. Then 125 Ilabeled ␣ 2 M* was added to each well (2 nM final concentration). After incubation for 2 h at 37°C, cells were washed with phosphate-buffered saline and detached from plastic using 0.5 mg/ml trypsin, 0.5 mg/ml proteinase K, and 5 mM EDTA-containing buffer. Internalized 125 I-␣ 2 M* was defined as radioactivity associated with the cell pellet. Nonspecific uptake of 125 I-␣ 2 M* was determined in the presence of 1 M RAP and was subtracted from the total internalization. The cell numbers for each experimental condition were measured in parallel wells that did not contain radioactivity to ensure that antibodies did not cause cell detachment during treatment.
CHO 13-5-1 cells (25) transiently transfected with mini-LRP constructs were incubated for 3 h at 37°C with 125 I-labeled ligands at the concentrations indicated in each experiment, and cellular internalization was measured as described above. Nonspecific internalization of 125 I-labeled IgG was determined in the presence of excess unlabeled antibody. Nonspecific internalization of 125 I-␣ 2 M* and 125 I-labeled fragments of RAP was determined in the presence of 1 M RAP. To estimate the relative amount of different mini-receptors on the cell surface, transfected CHO 13-5-1 cells were first chilled on ice for 1 h and then incubated with ice-cold medium containing 20 nM 125 I-labeled monoclonal antibody 5A6. After incubation for 2 h, cells were washed with phosphate-buffered saline and detached from the plastic wells using 0.5 mg/ml trypsin, 0.05 mg/ml proteinase K, and 5 mM EDTA-containing buffer. Bound 125 I-5A6 was defined as radioactivity released from the cell surface by trypsin and proteinase K. Nonspecific binding of 125 Ilabeled IgG was determined in the presence of excess unlabeled antibody and subtracted.
To measure the cellular degradation of 125 I-␣ 2 M* and 125 I-uPA⅐PAI-1 complexes, mouse embryonic fibroblasts were incubated with 125 I-labeled ligand in the absence or presence of RAP or the fragments of RAP for 6 h at 37°C. After incubation the degraded ligands were detected as radioactivity in the medium that is soluble in 10% trichloroacetic acid. The amount of degradation products generated in the absence of cells was also measured and subtracted from the total.

RESULTS
Monoclonal Antibody 8G1 Blocks Binding of ␣ 2 M* to LRP-As a first step in localizing the ␣ 2 M* binding site in LRP, we investigated the abilities of various antibodies prepared against LRP to block the cellular-mediated uptake of ␣ 2 M* (Fig. 1A). As a positive control, rabbit polyclonal antibody R2629, which recognizes multiple epitopes on LRP, effectively inhibited ␣ 2 M* uptake, whereas the non-immune IgG had no effect on ␣ 2 M* internalization by these cells. Monoclonal antibody 8G1 was also able to inhibit LRP-mediated ␣ 2 M* internalization in a dose-dependent manner. In contrast, monoclonal antibody 5A6 did not inhibit ␣ 2 M* internalization but rather seemed to enhance its uptake somewhat. Western blot analysis confirmed our previous report (7) that monoclonal antibody 8G1 recognizes the 515-kDa ␣-chain of LRP to which all of the ligands have been found to bind, whereas 5A6 recognizes the 85-kDa ␤-chain (Fig. 1B).
To determine if monoclonal antibody 8G1 blocks the direct interaction between ␣ 2 M* and LRP, an in vitro binding assay was employed in which the binding of 125 I-labeled ␣ 2 M* to LRP immobilized on microtiter wells was measured. The results of this experiment (Fig. 2) demonstrated that monoclonal antibody 8G1 blocked the binding of ␣ 2 M* to LRP. In contrast, the ␤-subunit-specific antibody 5A6 had no effect on the binding of ␣ 2 M* to LRP in this solid phase assay. Curiously, when comparing the data from Fig. 1 with those of Fig. 2, it appears that 5A6 seemed to stimulate LRP-mediated ␣ 2 M* uptake in cells but had little effect on ␣ 2 M* binding to LRP immobilized on plastic. Although the reason for this effect is not known, we speculate that 5A6 may dimerize LRP on the cell surface, which is likely to increase the affinity of this receptor for the multivalent ligand, ␣ 2 M*.
Monoclonal Antibody 8G1 Binds to the First Cluster of Ligand Binding Repeats-We next set out to map the region on LRP recognized by monoclonal antibody 8G1. For these experiments, a series of secreted receptor fragments was prepared (Fig. 3A). Accumulating data demonstrate that a number of LRP ligands bind to the clusters of ligand binding cysteine-rich repeats, so we focused on LRP fragments which contain the four clusters of ligand binding repeats (sLRP1, sLRP2, sLRP3, sLRP4). We also designed a fragment that included cluster I along with EGF repeats and the YWTD region (sLRP1a) and a fragment with a portion of cluster II (sLRP1b) because a similar fragment of chicken LRP has been reported to display some ␣ 2 M* binding activity (26). The constructs contained a myc epitope tag at their carboxyl-terminal region. After transfection of cells, the conditioned medium was collected, and secreted LRP fragments were detected by immunoblotting with antimyc IgG (Fig. 3B, left). When the same samples were probed with 8G1 antibody (Fig. 3B, right), sLRP1, sLRP1a, and sLRP1b were recognized by 8G1. In contrast, sLRP2, sLRP3 and sLRP4 were not recognized by 8G1. These results indicate that the 8G1 binding site is located within a stretch of amino acids encompassing residues 1-172 in the LRP sequence, which contains the first cluster of ligand binding repeats along with two EGF-like repeats. When sLRP-containing medium was subjected to SDS-polyacrylamide gel electrophoresis under reducing conditions prior to immunoblotting, monoclonal antibody 8G1 no longer recognized any of the LRP fragments (data not shown), indicating that the 8G1-binding epitope is not comprised of a linear stretch of amino acids. Experiments were also performed to determine if ␣ 2 M* was capable of binding to full-length LRP or any secreted receptor fragment using a ligand blotting protocol. The ligand blot experiments failed to detect ␣ 2 M* binding to full-length LRP or any secreted receptor fragment, indicating that ␣ 2 M* binding to LRP is sensitive to the conformation of the LRP molecule.
High Affinity ␣ 2 M* Binding to LRP Requires Ligand Binding Repeats in Both Cluster I and Cluster II-To identify the region on LRP responsible for binding ␣ 2 M*, LRP mini-receptors were constructed in which various clusters of ligand binding repeats were fused with the entire ␤-chain containing a myc epitope at its carboxyl terminus (Fig. 4). In addition, a mini-receptor (termed mLRP1b) was also prepared which contained the first cluster of repeats along with the first three repeats from cluster II. Plasmids containing the cDNA encoding these receptors were transfected into LRP-deficient CHO 13-5-1 cells, and the expression and processing of the receptors were analyzed by immunoblotting with anti-myc IgG. This revealed that all minireceptors were expressed at similar levels and processed by furin (Fig. 4B, left). Immunoblotting of the same cell lysates with 8G1 (Fig. 4B, right) indicated that constructs containing the first cluster of ligand binding repeats were recognized by 8G1, data that are consistent with the results obtained with the soluble receptors. Further, the 8G1 blot confirmed furin processing of mLRP1 and mLRP1b because ␣-chains of mature mini-receptors were readily detectable (Fig. 4B, right).
Next, CHO 13-5-1 cells transfected with various mini-receptors were employed to measure the internalization of 125 Ilabeled 8G1 (Fig. 5A) and 125 I-labeled 5A6. Cells expressing mLRP1 and mLRP1b mediated the internalization of 125 I-labeled 8G1 (Fig. 5A), whereas cells expressing mLRP2 or LC were unable to internalize antibody 8G1. Studies using 125 Ilabeled ␤-chain monoclonal antibody 5A6 confirmed that both mLRP2 and LC are effectively delivered to the cell surface, and capable of mediating the endocytosis of this antibody (Fig. 5B), confirming that these mini-receptors are functional. We next measured the ability of various mini-receptors to mediate the internalization of 125 I-labeled ␣ 2 M* and found that cells transfected with mLRP1b were much more efficient in internalizing ␣ 2 M* than cells expressing mLRP1, mLRP2 or LRP light chain (LC) (Fig. 6A). To correct for different expression levels of mini-receptors, transfected cells were chilled to 4°C, and the amount of 125 I-labeled 5A6 bound to the cells was measured (Fig. 6B). Fig. 6C shows the internalization of ␣ 2 M* adjusted to  the difference in the expression levels of mini-receptors (i.e. normalized to the amount of 5A6 IgG bound at 4°C). The result suggests that optimal binding of ␣ 2 M* to LRP requires ligand binding repeats from cluster I as well as from cluster II. In separate experiments, we also examined cells transfected with mLRP3 and mLRP4 and found that these mini-receptors were unable to bind and internalize ␣ 2 M* (data not shown).
The Amino-terminal Region of RAP Inhibits the LRP-mediated Internalization of ␣ 2 M* upon Binding to Repeats in Cluster II-RAP is composed of four domains (27) and contains two LRP binding sites, one within the first two domains (D1D2) and one in the fourth domain (D4) (27). We tested the ability of each of these domains to inhibit the uptake of 125 I-labeled uPA⅐PAI-1 complexes and 125 I-labeled ␣ 2 M* (Fig. 7). Both D1D2 and D4 inhibited the LRP-mediated internalization of 125 I-labeled uPA⅐PAI-1 complexes (Fig. 7A), a ligand that binds to repeats within cluster II as well as cluster IV. 2 In contrast to uPA⅐PAI-1, the internalization of ␣ 2 M* was only blocked by the D1D2 fragment of RAP, whereas D4 had no inhibitory effect on LRP-mediated uptake of ␣ 2 M* (Fig. 7B).
To identify the portions of LRP which bind the RAP fragments D1D2 and D4, we transfected cells with various LRP mini-receptors and measured internalization of radiolabeled D1D2 and D4. In the same experiment uptake of 125 I-5A6 was determined to adjust for difference in the levels of receptor expression. The results of this experiment are shown in Fig. 8. 125 I-Labeled D1D2 was effectively internalized by cells expressing mLRP1b and mLRP2 but not by cells expressing mLRP1 or LC (Fig. 8A), indicating that the D1D2 binding site is located within three amino-terminal ligand binding domains of the cluster II. Thus, D1D2, which binds to cluster II, can effectively compete for ␣ 2 M* binding. In the same experiment, D4 was also found to bind to mLRP1b (Fig. 8B), but remarkably it failed to inhibit ␣ 2 M* binding.
To compare ␣ 2 M* binding properties of full-length LRP and mLRP1b, we expressed these receptors in CHO 13-5-1 cells and measured 125 I-␣ 2 M* uptake in presence of excess D1D2 and D4 fragments of RAP and 8G1 IgG (Fig. 9A). Consistent with results shown earlier ( Figs. 1 and 7), monoclonal antibody 8G1 and D1D2 inhibited the interaction of ␣ 2 M* with both fulllength LRP and mini-receptor mLRP1b, but D4 did not. To compare the relative affinity of ␣ 2 M* for LRP and mLRP1b, CHO 13-5-1 cells were transfected with full-length LRP and mLRP1b, and the transfected cells were incubated with increasing concentrations of 125 I-labeled ␣ 2 M*. As expected, cells transfected with either LRP or mLRP1b showed a dose-dependent increase in ␣ 2 M* uptake, which approached saturation at higher ␣ 2 M* concentrations (Fig. 9B). The concentration of ␣ 2 M* required for half-saturation in the LRP-transfected cells was 17 nM, whereas the concentration required for the mLRP1b transfected cells was 34 nM. These results indicate similar affinity of the mLRP1b and LRP for ␣ 2 M*. DISCUSSION A major question that remains unanswered is how LRP can bind multiple structurally distinct ligands with such high affinity. An important component in answering this question lies in identifying regions on LRP which recognize specific ligands. Identification of the regions on LRP involved in recognizing ␣ 2 M* has proven to be difficult because unlike most other LRP ligands, ␣ 2 M* binding to LRP appears sensitive to LRP conformation. Previous work indicates that ␣ 2 M* binds a unique region on LRP because its binding to this receptor is not competed with other ligands, such as apoE (28), tissue-type plasminogen activator (29), or with a Fab fragment that binds to a region within cluster II (30).
Earlier studies by Moestrup and Gliemann (31) demonstrated that a 75-kDa proteolytic fragment of LRP containing cluster II (residues 776 -1399) recognized the 125 I-labeled rat ␣ 1 -macroglobulin light chain. Later studies, using surface plasmon resonance reported measurable binding of LRP fragments containing either cluster II or cluster IV to ␣ 2 M* immobilized on sensor chips (4). Together, these in vitro binding studies suggest that the ␣ 2 M* binding site in LRP may be contained within cluster II or cluster IV. However, in contrast to these experiments, cells transfected with LRP mini-receptors containing clusters II or clusters IV failed to bind ␣ 2 M* or mediate its cellular internalization (21,32). This failure was not caused by incorrect folding or cellular processing of these mini-recep- tors, as they were fully functional in mediating the internalization of other ligands.
The objective of the current investigation was to map the binding site on LRP for ␣ 2 M*. Our strategy relied on the generation of fully functional LRP deletion mutants. Each mini-receptor was appropriately processed and delivered to the cell surface and functioned to endocytose a monoclonal antibody recognizing an epitope on the ␤-subunit. We used these mini-receptors to map the binding sites on LRP for two inhibitors that block the LRP-mediated uptake of 125 I-labeled ␣ 2 M* in cells. The first inhibitor, monoclonal antibody 8G1, was found to bind to a fragment of LRP containing the first two ligand binding repeats, along with two EGF-type repeats. A second inhibitor, an amino-terminal fragment of RAP, was found to bind to repeats within cluster II. These data suggest that regions of both cluster I and cluster II form a high affinity ␣ 2 M* binding site. This was confirmed by constructing a minireceptor containing ligand binding repeats within cluster I as well as the first three ligand binding repeats of cluster II (residues 1-954 of the LRP ␣-chain) and demonstrating that this receptor is effective in internalizing ␣ 2 M*. Thus, ␣ 2 M* appears to be the first known ligand to recognize repeats within cluster I.
Human ␣ 2 M is a tetrameric protease inhibitor composed of four identical subunits. As a consequence of the structural transformation of ␣ 2 M upon formation of a complex with proteinases, the carboxyl-terminal receptor binding domain of each subunit becomes exposed, generating four identical binding sites for LRP. In vitro binding of ␣ 2 M* to LRP (23, 31) fits a model in which high affinity (K D ϭ 100 -500 pM) binding occurs when two domains of ␣ 2 M* recognize adjacent receptors, whereas lower affinity binding (K D ϭ 1-10 nM) occurs when only one of the four domains binds to LRP (31). Together with the current results, these studies suggest two possible models for the binding of ␣ 2 M* to LRP (Fig. 10). In the first model, repeats from cluster I and II combine to form an ␣ 2 M* binding site (Fig. 10A), whereas in the second model (Fig. 10B), one subunit of ␣ 2 M* recognizes repeats in cluster I, whereas another subunit recognizes repeats in cluster II. Both models accommodate the interaction of a single ␣ 2 M* molecule with two LRP molecules. Currently, it is not possible to distinguish between these models, but the sensitivity of ␣ 2 M* binding to the conformation of LRP suggests that the first model ( Fig.  10A) more accurately reflects the binding mechanism. The models are supported by NMR measurements that reported a weak interaction (K D approximately equal to 140 M) between the first repeat of cluster II and the receptor binding domain of ␣ 2 M* (33) and by plasmon resonance measurements reporting that tandems of first and second or second and third repeats of cluster II bind ␣ 2 M* weakly (K D ϭ 20 M) when ␣ 2 M* was coupled to a sensor chip (34).
The current investigation begins to give insight into the complexities of ligand recognition by LRP. Crystallographic and NMR studies of individual ligand binding repeats of LDL receptor and LRP revealed that the sequence variability in short loop regions of each repeat results in a unique contour surface and charge density for each repeat (33). The current study, along with previous work, suggests that the ability of LRP to bind numerous structurally distinct ligands with high affinity results from the presence of 31 ligand binding repeats in the molecule, forming a unique contour surface and charge distribution, and from the multiple interactions between both the ligand and receptor. It is now apparent that certain ligands recognize different combinations of the ligand binding repeats in a sequential fashion, whereas others such as ␣ 2 M* recognize repeats from separate clusters.
Identification of regions on LRP that are important for recognizing ␣ 2 M* will now allow development of specific inhibitors capable of preventing this binding. These inhibitors should be extremely useful for dissecting out the contributions of ␣ 2 M* in normal and pathological processes. For example, ␣ 2 M* associates with LRP in neurons and induces a calcium influx via N-methyl-D-aspartate receptors (12). The influx of calcium caused by LRP-mediated activation of N-methyl-D-aspartate receptor channels is likely to impact a variety of downstream signaling cascades and may provide a mechanism of altering local synaptic plasticity. Assessing the contribution of ␣ 2 M* to neuronal function in vivo using an LRP antagonist such as RAP has not been possible because RAP blocks the binding of all ligands to LRP and thus is not selective for a particular ligand. Other LRP ligands, such as tissue-type plasminogen activator, are also implicated in neuronal function. Tissue-type PA associates with LRP, and this interaction is important for hippocampal late phase long term potentiation. Thus, defining the physiological role of ␣ 2 M* and tissue-type PA in neuronal func-tion will require inhibitors capable of specifically blocking their interaction with LRP.
␣ 2 M* is also thought to promote the catabolism of the A␤ peptide by binding this molecule and facilitating its LRP-mediated uptake. LRP is suspected to play a dual role in Alzheimer's disease (35) by promoting both the synthesis (13) and catabolism (14) of this toxic peptide. The development of inhibitors that specifically block ␣ 2 M* binding to LRP but do not impair the interaction of LRP with other ligands such as amyloid precursor protein will be useful for distinguishing between the opposing roles of LRP in Alzheimer's disease and will more precisely define the physiological role of this receptor in this disease. FIG. 10. Proposed models for the LRP-␣ 2 M* interaction. Ligand binding repeats (shaded cylinders) in LRP are arranged in four clusters (I-IV) containing 2, 8, 10, and 11 repeats, respectively. Each cluster is surrounded by EGF-like repeats (open circles) and YWTD ␤-propeller domains (wavy lines). A, in this model, repeats from cluster I are proposed to be in close proximity with the amino-terminal portion of cluster II to form a high affinity binding site for one subunit of ␣ 2 M*. B, an alternative model in which one subunit of ␣ 2 M* recognizes repeats within cluster I, while another subunit recognizes repeats within the amino-terminal region of cluster II. Although not shown, both models allow for ␣ 2 M* interaction with adjacent receptors.