A Lipoprotein Receptor Cluster IV Mutant Preferentially Binds Amyloid-β and Regulates Its Clearance from the Mouse Brain*

Background: Wild type LRP1 cluster IV (WT-LRPIV) binds plasma Aβ and reduces Aβ-related pathology in APPsw+/0 mice. Results: A novel LRPIV-D3674G mutant binds Aβ with a higher affinity than WT-LRPIV and clears brain Aβ better than WT-LRPIV. LRPIV-D3674G reduces effectively Aβ pathology in APPsw+/0 mice. Conclusion: LRPIV-D3674G is an efficient Aβ clearance agent. Significance: Aβ clearance therapy is critical for treatment of Alzheimer disease. Soluble low density lipoprotein receptor-related protein-1 (sLRP1) binds ∼70% of amyloid β-peptide (Aβ) in human plasma. In Alzheimer disease (AD) and individuals with mild cognitive impairment converting to AD, plasma sLRP1 levels are reduced and sLRP1 is oxidized, which results in diminished Aβ peripheral binding and higher levels of free Aβ in plasma. Experimental studies have shown that free circulating Aβ re-enters the brain and that sLRP1 and/or its recombinant wild type cluster IV (WT-LRPIV) prevent Aβ from entering the brain. Treatment of Alzheimer APPsw+/0 mice with WT-LRPIV has been shown to reduce brain Aβ pathology. In addition to Aβ, LRPIV binds multiple ligands. To enhance LRPIV binding for Aβ relative to other LRP1 ligands, we generated a library of LRPIV-derived fragments and full-length LRPIV variants with glycine replacing aspartic acid residues 3394, 3556, and 3674 in the calcium binding sites. Compared with WT-LRPIV, a lead LRPIV-D3674G mutant had 1.6- and 2.7-fold higher binding affinity for Aβ40 and Aβ42 in vitro, respectively, and a lower binding affinity for other LRP1 ligands (e.g. apolipoprotein E2, E3, and E4 (1.3–1.8-fold), tissue plasminogen activator (2.7-fold), matrix metalloproteinase-9 (4.1-fold), and Factor Xa (3.8-fold)). LRPIV-D3674G cleared mouse endogenous brain Aβ40 and Aβ42 25–27% better than WT-LRPIV. A 3-month subcutaneous treatment of APPsw+/0 mice with LRPIV-D3674G (40 μg/kg/day) reduced Aβ40 and Αβ42 levels in the hippocampus, cortex, and cerebrospinal fluid by 60–80% and improved cerebral blood flow responses and hippocampal function at 9 months of age. Thus, LRPIV-D3674G is an efficient new Aβ clearance therapy.

Deposits of amyloid ␤-peptide (A␤) 3 and neurofibrillary tangles are neuropathological features of Alzheimer disease (AD) (1). A␤ is produced in the brain and periphery by proteolytic cleavage from its larger A␤ precursor protein (APP) (2). A body of evidence suggests that soluble A␤ oligomer species contribute to neurodegeneration in AD (3). A␤ concentration in brain interstitial fluid is controlled by its rate of production in brain; influx and/or re-entry of circulating, peripheral A␤ into the brain across the blood-brain barrier (BBB) via receptor for advanced glycation end products (RAGE) (4); and clearance of A␤ from the brain (5,6). A␤ is cleared from the brain by different mechanisms, including transport across the BBB via low density lipoprotein receptor-related protein-1 (LRP1) (5,(7)(8)(9) and enzymatic degradation in the brain (10). A continuous removal of A␤ from the brain and systemic circulation is essential to prevent accumulation of toxic soluble oligomeric A␤ in the brain (3,11).
We and others have reported that cell surface LRP1 at the BBB and soluble LRP1 (sLRP1) in plasma have important functions in A␤ homeostasis (7)(8)(9)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21). sLRP1 is a major transport binding protein for peripheral circulating A␤ and can sequester 70 -90% of plasma A␤ in neurologically intact human subjects (14). sLRP1 levels are reduced in patients with AD and mild cognitive impairment who subsequently progress to AD (MCI-AD) (14,22). In addition, sLRP1 is oxidized in AD and AD-MCI individuals, and oxidized sLRP1 is unable to bind A␤ (14,22). Consequently, an increase in free, unbound A␤ plasma levels relative to sLRP1-bound plasma A␤ levels has been reported in AD and MCI-AD patients (14,22). Experimental studies have shown that free A␤40 and A␤42 re-enter the brain (4,23,24) and contribute to the formation of neurotoxic soluble oligomeric A␤ species. It has been also reported that sLRP1 and its recombinant ligand binding domain cluster IV (LRPIV) prevent free A␤ from entering the brain (14).
Several A␤ antibody therapies are directed at facilitating A␤ clearance from the brain (25,26). A␤ antibodies that act principally by sequestering the peripheral A␤ pool do not cross the BBB and are generally thought to clear A␤ from the brain by the so-called "peripheral sink" mechanism through binding of peripheral A␤, which according to some studies lessens the risk of potential central side effects of the antibody clearance therapy, such as neuroinflammation, vasogenic edema, and cerebral microhemorrhages (27)(28)(29)(30). Recent studies in patients with mild AD have shown beneficial effects of intravenous immunoglobulin preparation (Gammagard), which has been suggested to act as a peripheral sink agent containing sLRP1 and antibodies against diverse A␤ conformations (31,32).
In a previous study, we have demonstrated that that wild type recombinant LRPIV (WT-LRPIV) effectively binds A␤40 and A␤42 in vitro (8,14) and sequesters free A␤ in plasma of AD patients and AD transgenic mice in vivo, which reduces A␤ pathology in mice (14). LRPIV contains 11 complement-type repeats (CRs) (CR21-CR31), which can participate in binding of LRP1 ligands (33,34). For example, CR24 -CR28 efficiently binds LRP1 ligands Factor IXa, receptor-associated protein (RAP), and activated ␣ 2 -macroglobulin (␣ 2 M*) (35,36). Each CR has ϳ40 amino acids and a single calcium ion (35). Calcium ion binding domains are required for proper folding and structural integrity of LRP1 (33,34). The alterations in calcium binding sites alter folding of the CRs (33). In an attempt to enhance A␤ binding to LRPIV and improve its A␤-clearing properties, we generated a library of recombinant LRPIV fragments and full-length LRPIV variants with single glycine replacement with an aspartic acid residue in the calcium binding site in either CR 21,22,26,27,28,or 29. We expressed these LRPIV analogs in Chinese hamster ovary (CHO) cells. Next, LRP-IV-derived analogs that have been well secreted by CHO cells into the medium were purified, and their binding affinity for A␤40 and A␤42 and other LRP1 ligands, including apolipoprotein E2-4 (apoE2-4), tissue plasminogen activator (tPA), matrix metalloproteinase-9 (MMP-9), and Factor IXa, was determined. A lead LRPIV-D3674G mutant had the highest in vitro binding affinity for A␤ peptides relative to other ligands and cleared mouse endogenous A␤ more efficiently than WT-LRPIV. Moreover, subcutaneous LRPIV-D3674G treatment significantly reduced A␤ levels in brain and cerebrospinal fluid (CSF) in Alzheimer APPsw ϩ/0 mice and improved cerebral blood flow responses and hippocampal function. Our findings suggest that LRPIV-D3674G is an efficient new A␤ clearance therapy for AD.

Materials
Human A␤40 and A␤42 were synthesized at the W. M. Keck Facility (Yale University), using solid-phase N-t-butyloxycarbonyl chemistry, and purified by HPLC. Primers were synthesized by Integrated DNA Technologies (Coralville, IA), and dNTPs were obtained from Invitrogen. All other reagents were from Sigma-Aldrich unless otherwise indicated. Purified recombinant full-length wild type ligand binding cluster IV of LRP1 (WT-LRPIV) was used for generating rabbit polyclonal LRPIV antibody (GeneScript, Piscataway, NJ). Another rabbit polyclonal anti-LRPIV antibody (37) was kindly provided by Dr. Elizabeth Komives (University of California San Diego, La Jolla, CA).

Production of LRPIV Variants
Synthesis of cDNA and Cloning of LRPIV-LRP1 cDNA was synthesized from human spleen total RNA (Clontech, Mountain View, CA) using SuperScript II RT (Invitrogen). Primers were designed based on LRP1 sequence (NM_002332) ( Table  1). LRPIV (Fig. 1A) was amplified from cDNA by polymerase chain reaction (PCR) using Pfx-DNA polymerase (Invitrogen) TABLE 1 Primers used in the PCR to generate constructs of the truncated fragments or mutants of LRPIV CRs of LRPIV were cloned into pSecTag2 B vector. Full-length LRPIV (WT-LRPIV) was cloned into pcDNA3.3 TOPO vector. The forward primer has Kozak sequence (acc) and the tPA signal peptide (in italic type) followed by the LRPIV sequence.

Clone
Forward primer Reverse primer and their respective primer sets and cloned into pcDNA3.3 TOPO vector. Using this construct, full-length LRPIV consisting of 11 CRs (CR21-CR31), two four-repeat fragments (CR24 -CR27 and CR25-CR28), one three-repeat fragment (CR25-CR27), two two-repeat fragments (CR25-CR26 and CR26-CR27), and two single-repeat fragments (CR25 and CR26) (Fig. 1A) were amplified using PCR and cloned in mammalian expression vector pSecTag2 B (Invitrogen) between HindIII and BamHI restriction sites to express soluble proteins. pSecTag2 B vector has the IgK leader peptide on the N terminus and a Myc tag and His 6 tag on the C terminus. In addition, a full-length LRPIV was amplified using 129 bp of forward primer (which has Kozak sequence, start codon, tPA signal peptide sequence, and LRPIV sequence) and reverse primer with the HindIII restriction site and cloned into pcDNA3.3 TOPO vector (WT-LRPIV). Full-length LRPIV variants with glycine replacing aspartic acid residues 3354, 3394, 3556, 3595, 3633, and 3674 ( Fig. 1B) in the calcium binding sites were made by site-directed mutagenesis using the QuikChange Lightning site-directed mutagenesis kit (Stratagene, LA Jolla, CA). WT-LRPIV was used as a template along with their respective primer sets. Protein Expression-CHO cells were grown in CDOpti CHO medium supplemented with 1 mM CaCl 2 and 2 mM Glutamax at 37°C in a shaker flask. The cells were then stably transfected with each construct using FreeStyle MAX reagent (Invitrogen). Five days after transfection, cells were grown into medium supplemented with 700 g/ml Geneticin (for selection of cells with pcDNA 3.3 TOPO vector) or 200 g/ml hygromycin (for selection of cells with pSecTag2 vector). After 12-15 days, around 5000 antibiotic-resistant cells were plated on a 100 ϫ 10-mm Petri plate containing CloneMatrix (catalog no. K8510, Genetix Molecular Devices, Inc., Sunnyvale, CA) mixture (40% Clone-Matrix, 50% 2ϫ CDOpti CHO, and antibiotics). Three weeks after plating, 50 -60 single clones were transferred into CDOpti CHO medium in 48-well plates. Three days later, media were collected and tested for expression of LRPIV by Western blot analysis. Selected clones were subsequently transferred into 12-well plates. A single selected clone was transferred into a Fernbach flask and grown as a suspension culture. Culture was started with 1 ϫ 10 6 /ml cell density in CDOpti CHO medium containing 2 mM Glutamax, 1 mM CaCl 2 , and 10% CHO CD Efficient Feed A (Invitrogen). The cells were counted daily using a hemocytometer (Hausser Scientific Partnership, Horsham, PA), and glucose levels were determined using a Gluc-Cell TM test strip (CESCO Bioengineering Co., Taichung, Taiwan). When the glucose level fell below 2 g/liter in the conditioned medium, cells were supplemented with 10% Feed A containing 2 mM Glutamax and 1 mM CaCl 2 . After 10 days of culture, the conditioned medium was collected, centrifuged, filtered through 0.2-m membrane, and stored frozen at Ϫ20°C until analysis. The LRPIV-derived peptides CR25-CR26, CR26-CR27, CR25, CR26, and mutant proteins D3354G, D3595G, and D3633G were not included in the present study due to their poor secretion in the medium.

Screening of LRPIV Variants for Ligand Binding by ELISA-
Microtiter plates were coated with 5 g/ml LRPIV analog overnight in 55 mM sodium bicarbonate buffer at 4°C. All wells were blocked with protein-free buffer (catalog no. 37570, Pierce) for 1 h at room temperature. Varying concentrations of A␤40, A␤42, or ApoE2, -3, or -4 (catalog nos. P2002, P2003, and P2004, respectively, Invitrogen), tPA (catalog no. 10-633-45291, Genway Biotech, Inc., San Diego, CA), MMP-9 (catalog no. 911-MP-010, R&D Systems, Minneapolis, MN), and Factor IXa (catalog no. RP-43110, Pierce) were incubated in Hanks' balanced solution culture medium (HBSC), pH 7.4, containing 0.05% Tween 20 (HBSCT) at room temperature for 2 h. Anti-A␤ (1 g/ml; catalog no. 2454, Cell Signaling Technology Inc.), anti-ApoE (0.1 g/ml; catalog no. K74180, Biodesign Meridian LifeScience, Memphis, TN), anti-tPA (catalog no. Ab62763, Abcam, Cambridge, MA), anti-MMP-9 (catalog no. Ab5707, Abcam), or anti-Factor IXa (catalog no. LS-C23381, LifeSpan Biosciences, Inc., Seattle, WA) overnight at 4°C in HBSCT containing 0.25% BSA. After washing plates four times with HBSCT, wells were incubated with goat anti rabbit-HRP or donkey anti-goat (1:3000 dilution) in HBSCT and 0.25% BSA for 30 min at room temperature. After washing plates four times with HBSCT, 100 l of 3,3Ј,5,5Ј-tetramethylbenzidine substrate (catalog no. 53-00-01, KPL, Gaithersburg, MD) was added. The reaction was developed for 10 min and stopped with 100 l of 1 M HCl. The absorbance was read at 450 nm. To calculate the K d (ligand concentration that binds to half of the microtiter plate immobilized LRPIV receptor sites at equilibrium), we utilized GraphPad Prism software (version 3) based on the Marquardt method of nonlinear one-site binding (hyperbola) regression (curve fit) analysis of the ELISA-based absorbance values against specific concentrations of the various ligands studied. GraphPad Prism software uses the follow- where B max is the maximum specific binding in the same units as Y (Y is the specific binding extrapolated to very high concentrations of ligand, so its value is almost always higher than any specific binding measured in the experiment). K d is the equilibrium binding constant, in the same units as X (X is the ligand concentration needed to achieve a half-maximum binding at equilibrium). We only measured LRP-bound ligand for the affinity calculation and did not measure free and unbound LRP constructs.
Characterization of LRPIV-D3674G Binding to Monomer, Oligomers, and Fibrils by a Dot Blot Assay-Oligomers and fibrils of A␤40 were prepared as described earlier (41,42) and confirmed by a dot blot assay using oligomer-and fibril-specific antibodies A11 and OC, respectively (41,42). In the dot blot assay, 1 g of A␤40 monomers, oligomers, or fibrils was applied to a nitrocellulose membrane, and the membrane was blocked with protein-free blocking buffer (Pierce). The membrane was incubated with 10 g/ml LRPIV-D3674G in HBSC, pH 7.4, for 2 h at room temperature, and the membrane was probed with a rabbit polyclonal anti-LRPIV antibody.

In Vivo Studies
Animals-C57Bl6 and APPsw ϩ/0 mice were purchased from Jackson Laboratories and Taconic Farms, respectively. Mice were housed under standard conditions (12-h light/dark cycle starting at 7:00 a.m.; 21 Ϯ 2°C; 55 Ϯ 10% humidity) in solid bottomed cages on wood chip bedding. All studies were performed according to National Institutes of Health guidance using protocols approved by the University of Rochester Committee on Animal Resources. Mice were anesthetized by intraperitoneal injections with a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg).
LRP-IV Pharmacokinetic Studies-Mice (C57BL6, 2-3 months old) were anesthetized as above, and a single bolus of 125 I-WT-LRP-IV or 125 I-LRPIV-D3674G in PBS was administered via femoral vein, and blood samples were collected from the retro-orbital sinus at different time points after injection within 24 h, as described earlier (14). Plasma samples were counted, and trichloroacetic acid (TCA)-precipitable radioactivity was determined. Pharmacokinetic parameters were determined using a non-compartmental intravenous bolus module of Kinetica version 5.1 software (Thermo Fisher Scientific).
Permeability Surface Area (PS) Product for LRPIV-D3674G and WT-LRPIV-125 I-WT-LRPIV (20 Ci/kg) or 125 I-LRPIV-D3674G (20 Ci/kg) was administered intravenously and compared with [ 14 C]sucrose (11 Ci/kg) uptake by the brain and CSF 1 h after injection of tracers. 125 I TCA-precipitable counts/ min were determined in the plasma, CSF, and brain. In these studies, the brain was perfused briefly with cold PBS at the end of the experiment to remove residual vascular radioactivity, as confirmed by undetectable radioactivity of [ 14 C]sucrose (11 Ci/kg). The BBB PS product (l/min/g) of [ 14 C]sucrose, 125 I-WT-LRPIV, and 125 I-LRPIV-D3674G was calculated by using the equation, where V D is the volume of distribution of [ 14 C]sucrose, 125 I-WT-LRPIV, and 125 I-LRPIV-D3674G after 1 h of intravenous injection in whole brain homogenate; C PL (T) is the terminal plasma concentration of TCA-precipitable 125 I-WT-LRPIV, 125 I-LRPIV-D3674G, or [ 14 C]sucrose; and AUC is area under the curve on a plot of concentration versus time calculated as we described (43).
LRPIV Treatment of Wild Type Mice-Ten-week-old C57BL6 mice (n ϭ 5 mice/group) were treated daily with WT-LRPIV or LRPIV-D3674G (20 g/mouse, intravenously via the tail vein) or vehicle for 5 days. At the end of treatment, mice were sacrificed under anesthesia. Blood and CSF samples were collected, and plasma was immediately separated from blood at 4°C. All samples were stored immediately at Ϫ80°C. Mice were perfused intracardially with ice-cold heparinized saline and hemibrains homogenized in 2% SDS containing protease inhibitor mixture (Roche Applied Science). Mouse endogenous A␤40 and A␤42 levels in the plasma and brain were determined by mouse A␤-specific ELISA as described below. WT-LRPIV and LRPIV-D3674G levels in CSF were determined by Western blot analysis as described earlier (14).

Safety Studies
Plasma Cholesterol-Total cholesterol levels in plasma were determined using a kit (catalog no. 439-17501, Wako Chemicals USA Inc., Richmond, VA).
Mouse Plasma tPA-Plasma levels of active mouse tPA were determined using a kit (catalog no. MTPAKT, Molecular Innovations, Novi, MI).
Mouse Plasma Pro-MMP-9-Plasma levels of mouse pro-MMP-9 were determined using a kit (Quantinine MMP900, R&D Systems) and following the manufacturer's instructions.
Activated Partial Thromboplastin Time (aPTT)-Mouse blood samples were collected in citrate buffer (250 mM, pH 7.2) and centrifuged at 10,000 ϫ g for 10 min to separate plasma. Plasma (25 l) was incubated with 25 l of TriniCLOT aPTT HS reagent (Trinity Biotech plc., Bray, Ireland) for 3 min at 37°C in a Start4 coagulometer (Diagnostica Stago, Inc.). The aPTT reaction was started by adding 50 l of 50 mM calcium chloride, and the time of clotting was recorded when the movement of the iron ball ceased under the magnetic field.
Western Blot Analysis-For detection of APP, frozen brain cortical tissues were homogenized in lysis buffer (Roche Applied Science) containing a complete protease inhibitor mix-ture (Roche Applied Science). The homogenate was centrifuged for 20 min at 20,000 ϫ g. The supernatant was collected in a separate tube, and total protein concentration was determined by a BCA protein assay (Pierce). Brain lysate proteins were subjected to 4 -12% NuPAGE BisTris SDS-PAGE (Invitrogen) and transferred to nitrocellulose membrane (Bio-Rad). Membrane was blocked with 5% nonfat milk in TBST for 1 h and incubated overnight with the primary monoclonal antibody against APP (MAB348, Chemicon, Temecula, CA). The membranes were washed and incubated with horseradish peroxidase-conjugated secondary antibody for 1 h. Immunoreactivity was detected using SuperSignal West Pico chemiluminescent substrate (Thermo Scientific).
CSF Levels of WT-LRPIV and LRPIV-D3674G-CSF samples (15 l) were denatured in the SDS buffer by heating at 70°C for 10 min. Samples were subjected to electrophoresis on an SDSpolyacrylamide gel and transferred to nitrocellulose. LRPIV was detected with a rabbit polyclonal anti-LRPIV antibody. In a separate experiment, 125 I TCA-precipitable counts in 20 l of plasma and CSF were analyzed 1 h after a single bolus of 125 Ilabeled WT-LRPIV or LRPIV-D3674G (20 Ci/kg) in PBS.
Immunogenicity Testing-Immunogenicity testing was outsourced (QED Bioscience Inc., San Diego, CA). Approximately 8 -10-week-old BALB/c female mice (average weight 0.02 kg) were dosed (20, 40, and 60 g/kg, intraperitoneally) four times biweekly. Retro-orbital blood samples of 250 l each were collected before dosing and 1 week after following each injection. Each mouse was bled a total of five times. Blood samples were collected in heparinized tubes and centrifuged immediately after collection. Serum samples were frozen at Ϫ20°C. After the last blood sample collection, all of the serum samples were tested for anti-LRPIV-D3674G antibody by ELISA. The plates were coated with three different LRPIV-D3674G concentrations (1, 5, or 20 g/ml), and plasma samples were tested from 1:100 to 1:6400 dilutions for anti-LRPIV-D3674G antibodies. A rabbit LRPIV antiserum was used as a positive control.
LRPIV Treatment of APPsw ϩ/0 Mice-APPsw ϩ/0 mice (n ϭ 9 mice/group) were treated subcutaneously with LRPIV-D3674G (40 g/kg per day) or saline for 3 months beginning at the age of 6 months. After 3 months of treatment, we measured cerebral blood flow responses to brain activation by whisker stimulation and performed a burrowing behavioral test. After these functional tests, mice were sacrificed under anesthesia and brain, CSF, and plasma were collected and analyzed for human A␤40 and A␤42 levels as described earlier (14).
Burrowing Test-APPsw ϩ/0 mice after 3 months of treatment with LRPIV-D3674G were subjected to a burrowing behavior test. The burrowing test was performed as described (46). Briefly, mice were individually placed in the cages equipped with a burrow made from a 200-mm-long, 68-mm-diameter tube of PVC plastic. One end of the tube was closed by a PVC cap. The open end of the tube was raised ϳ30 mm by a PVC ring. The burrow was filled with 200 g of mouse food pellets, and the mice were allowed to burrow for 3 h. The weight of the remaining food pellets inside the burrow was determined to obtain a measurement of the amount burrowed.
Prussian Blue Staining-Briefly, acetone-fixed saggital brain sections from vehicle-treated and LRPIV-D3674-treated mice were incubated in a 5% potassium ferrocyanide and 5% hydrochloric acid solution (1:1 working solution) for 30 min, as reported (48). Hemosiderin shows a blue color. To analyze the abundance of Prussian blue-positive deposits per section, the total numbers of Prussian blue-positive spots were divided by the number of sections analyzed.
Statistical Analysis-The results were compared by one-way multifactorial analysis of variance followed by Tukey's post hoc analysis for more than two groups of data or Student's t test for unpaired data of two groups. The differences were considered to be significant at p Ͻ 0.05. All values are mean Ϯ S.E.

RESULTS
Generation of Recombinant LRPIV-derived Analogs-We generated eight recombinant CR fragments of LRPIV with Myc and His 6 tags (i.e. CR24 -CR27, CR25-CR28, CR25-CR27, CR25-CR26, CR26-CR27, CR25, CR26, and CR21-CR31, corresponding to full-length LRPIV (Fig. 1, A and B). We also generated WT-LRPIV without Myc and His 6 tags or the 16-amino acid peptide tag derived from factor VIII (tagged WT-LRPIV) that has been described previously (38,49,50). Two single CRs (CR25 and CR26), two double CRs (CR25-CR26 and CR26-CR27), and CR21-CR31 were eliminated from further screening because of their poor secretion by CHO cells into the medium. The remaining three LRPIV fragments were purified from CHO-secreted medium, including two four-CR fragments (CR24 -CR27 and CR25-CR28) and one three-CR fragment (CR25-CR27), and, along with the full-length WT-LRPIV (Fig. 1C), were used for the binding studies. We set the criterion that a lead analog has to have a higher binding affinity for both A␤40 and A␤42 (i.e. lower K d value) compared with previously described tagged WT-LRPIV, which has been shown to bind A␤40 and A␤42 in vitro with K d values of 2 and 5 nM, respectively (14). The K d for A␤40 binding to CHO-secreted WT-LRPIV was 1.90 Ϯ 0.40 nM (Table 2), which was comparable The aspartic acid residues in italic type (*) are replaced with glycine by site-directed mutagenesis to generate LRPIV variants. C, representative silver staining of the purified recombinant LRPIV-derived peptides/proteins used for A␤ binding studies. D and E, binding curves for human A␤40 and A␤42 to immobilized LRPIV mutants obtained by ELISAs, n ϭ 3 assays/group. Binding constants (K d ) were calculated from corresponding binding curves (see Tables 2  and 3 and "Experimental Procedures" for details). F, A␤ binding to immobilized LRPIV-D3674G in the presence of an A␤ N terminus-specific antibody, C terminus-specific antibody, or non-immune IgG (NI-IgG). G, binding of LRPIV-D3674G to immobilized A␤ monomers, oligomers, and fibrils by a dot blot assay. Error bars, S.E.
To determine which region of A␤ is recognized by LRPIV-D3674G, we have studied in vitro binding of A␤40 to immobilized LRPIV-D3674G with and without N terminus anti-A␤40, C terminus anti-A␤40, and control NI-IgG antibodies (Fig. 1F). This experiment has shown that LRPIV-D3674G binds to the C-terminal sequence of A␤, which is consistent with a previous report showing that the wild type form of recombinant LRP-IV containing an extraneous 16-amino acid "tag," which contains the antigenic determinant of the mouse monoclonal antibody CLB-CAg 69 against Factor VIII (35,51), also binds the A␤ C-terminal domain (40). The precise amino acid sequence that LRPIV-D3674G detects will be determined by future epitope mapping studies. We also show that LRPIV-D3674G can bind, in addition to A␤ monomers, also A␤ oligomers and A␤ fibrils (Fig. 1G).
Plasma Elimination of WT-LRPIV and LRPIV-D3674G-There was no significant difference in the systemic clearance rate (t1 ⁄ 2 ϳ3.4 h) or mean residence time (ϳ4.8 h) of 125 I-WT-LRP-IV or 125 I-LRPIV-D3674G after an intravenous injection of tracers ( Fig. 2 and Table 4). However, the t1 ⁄ 2 was shorter compared with previously reported t1 ⁄ 2 for the wild type form of recombinant LRP-IV containing an extraneous 16-amino acid tag, which contains the antigenic determinant of the mouse monoclonal antibody CLB-CAg 69 against Factor VIII used to Values are mean Ϯ S. E., n ϭ 3 mice/group.

TABLE 2 K d values for A␤ binding to WT-LRPIV, LRPIV CRs, and LRPIV full-length variants with aspartic acid replaced with glycine in the calcium-binding sites
Values are means Ϯ S.E., n ϭ 3 assays/group.

CRs and LRPIV full-length variants
a p Ͻ 0.05, A␤ binding affinity to LRPIV-D3674 was significantly higher compared with WT-LRPIV or any other studied LRPIV analog.
purify the peptide from the culture medium (14). It remains unclear how the Factor VIII tag prolongs the rate of systemic clearance of WT-LRP-IV (t1 ⁄ 2 ϭ 11.8 h) (14), but it is of note that none of the presently studied forms of LRP-IV contained this extraneous 16-amino acid tag of Factor VIII. Clearance of Mouse Endogenous Brain A␤ by LRPIV-D3674G-Five-day intravenous treatment of C57Bl6 mice (20 g/mouse) with WT-LRPIV or LRPIV-D3674G at a dose comparable with that used previously to treat C57Bl6 mice with tagged WT-LRPIV (14) significantly reduced mouse endogenous brain A␤ levels. For example, compared with vehicle, LRPIV-D3674G mutant lowered brain A␤40 and A␤42 levels by 42 and 45%, respectively (Fig. 3, A and B), which correlated with the corresponding increases in total plasma levels of A␤40 and A␤42 (Fig. 3, C and D), most of which was bound to the LRPIV-D3674G mutant (not shown), similar to what was reported previously for tagged WT-LRPIV in C57Bl6 mice treated with tagged WT-LRPIV (14). Compared with WT-LRPIV, LRPIV-D3674G was more efficacious in removing mouse endogenous brain A␤, as indicated by ϳ25 and 27% greater reductions in brain A␤40 and A␤42 levels compared with the respective A␤ peptide reductions with WT-LRPIV (Fig. 2, A and B).
LRPIV-D3674G Does Not Affect Activity of Other LRP1 Ligands in Mice-We analyzed plasma and tissue samples from C57Bl6 mice treated intravenously with 20 g of LRPIV-D3674G/mouse daily for potential side effects. LRPIV-D3674G did not alter plasma levels of cholesterol, glucose, or LRP1 ligands, such as apoE, tPA, and pro-MMP-9 (Fig. 4, A-E). Plasma clotting time determined by the aPTT remained unchanged within 2 h of intravenous LRPIV-D3674G administration (Fig. 4F). In liver and brain microvessels, there were no changes in the expression levels of low density lipoprotein receptor or LRP1 (Fig. 5, A and B). In addition, in the brain, there were no changes in phosphorylated LRP1 levels after LRPIV-D3674G treatment (Fig. 5C). Expression of RAGE, an A␤ influx transporter (4), was not altered in brain microvessels by LRPIV-D3674G treatment (Fig. 5D). Furthermore, APP levels in the brain were not altered by LRPIV-D3674G treatment (Fig. 5E). LRPIV-D3674G did not enter CSF (Fig. 5F), similar to tagged WT-LRPIV (14), suggesting that its action is on sequestering the peripheral A␤ pool. Using radiolabeled 125 I-WT-LRPIV and 125 I-LRPIV-D3674G, we have independently confirmed that neither peptide is transported into the CSF and brain, as determined by non-detectable radioactivity in the CSF and non-detectable PS products for either form of LRP-IV (Fig.  5, G and H). For immunogenicity testing, blood samples were collected at the time of antigen administration and 1, 3, 5, and 7 weeks afterward. Antibodies to LRPIV-D3674G were not detected in postimmunization plasma samples in mice after treatment with increasing doses of LRPIV-D3674G, as described under "Experimental Procedures" (Fig. 5I).
LRPIV-D3674G Treatment of APPsw ϩ/0 Mice-Compared with vehicle-treated APPsw ϩ/0 mice, APPsw ϩ/0 mice treated daily with a subcutaneous low dose of LRPIV-D3674G (40 g/kg) for 3 months beginning at 6 months of age showed significant reductions of A␤40 and A␤42 levels in hippocampus, cortex, and CSF by 60 -80% (Fig. 6, A-D), as determined at 9 months of age. As expected, LRPIV-D3674G treatment significantly increased plasma A␤40 and A␤42 levels (Fig. 6, E and F). Treatment with LRPIV-D3674G also increased cerebral blood flow (CBF) responses to whisker stimulation by 75% (Fig. 6G) and significantly improved hippocampal function compared with vehicle-treated mice, as indicated by the burrowing test showing an approximately 60% improvement (Fig. 6H). The LRPIV-D3674 treatment in APPsw ϩ/0 mice did not influence inflammatory response and did not increase microhemorrhages, as shown by non-significant differences in the number of Iba1-positive microglia (Fig. 7, A and B) and Prussian bluepositive hemosiderin deposits (Fig. 7, C and D) in LRPIV-D3674-treated compared with vehicle-treated animals.

DISCUSSION
Plasma sLRP1 levels are reduced in AD and MCI-AD patients, and sLRP1 is oxidized, which results in diminished A␤ FIGURE 3. LRPIV-D3674G clears mouse endogenous brain A␤ more efficiently than WT-LRPIV in C57BL/6 mice. Shown are brain levels (A and B) and total plasma levels (C and D) of A␤40 and A␤42 in vehicle-treated (white), WT-LRPIV-treated (gray), and LRPIV-D3674G-treated (green) mice (intravenously for 5 days, 20 g/day/mouse). Values are mean Ϯ S.E. (error bars), n ϭ 3-5 mice/group. The differences were considered to be significant at p Ͻ 0.05. peripheral binding and higher levels of free A␤ in plasma (14,22). In addition, WT-LRPIV binds free A␤ in plasma of AD patients (14). Experimental studies have shown that free A␤ re-enters the brain (4, 23, 24, 52) and that sLRP1 and/or its recombinant LRPIV ligand binding domain prevent plasma A␤ from entering the brain (14). These studies underscored the need for development of sLRP1 replacement therapy for AD to maintain low levels of free, unbound A␤ in plasma and restore  . LRPIV-D3674G clears human brain A␤ and improves function in APPsw ؉/0 mice. LRPIV-D3674 was administered subcutaneously, 40 g/kg body weight/day, for 3 months beginning at 6 months of age. Shown are brain (hippocampus and cortex) levels (A and B), cerebrospinal fluid levels (C and D), and plasma levels (E and F) of A␤40 and A␤42 in vehicle-treated (white) and LRPIV-D3674G-treated (green) APPsw ϩ/0 mice. G, percentage increase in CBF in response to whisker stimulation in vehicle-treated (white) and LRPIV-D3674G-treated (green) APPsw ϩ/0 mice. H, weight of pellets burrowed in vehicle-treated (white) and LRPIV-D3674G-treated (green) APPsw ϩ/0 mice. Values are mean Ϯ S.E. (error bars), n ϭ 6 -9 mice/group. The differences were considered to be significant at p Ͻ 0.05. an important homeostatic mechanism regulating brain A␤ (11). Circulating endogenous A␤ antibodies that bind a fraction of A␤ in plasma (14,22,(53)(54)(55)(56) are also decreased in early stages of AD (22,57,58), further emphasizing the need for peripheral A␤ binding treatments. Peripheral A␤-binding agents, such as A␤ antibodies (25, 48, 59 -61), gelsolin (62,63), soluble RAGE (4), and LRPIV (14), have been shown to promote A␤ clearance from the brain, reducing brain A␤ and amyloid load in different APP-overexpressing mice.
LRP1 at the BBB, sLRP1 in plasma, and LRP1 in the liver have important roles in homeostasis of brain A␤ (5). Down-regulation of LRP1 at the BBB due to vascular risk factors and/or oxidative stress (17,64), oxidation, or reduced levels of sLRP1 in plasma (14) and reduced expression of LRP1 in the liver (65) lead to accumulation of A␤ in brain in different experimental models. A novel recombinant LRPIV-D3674G variant described here preferentially binds A␤ compared with other ligands and effectively reduces A␤ brain levels in wild type mice and AD mice without noticeable side effects. Moreover, treatment with LRPIV-D3674G increased CBF responses to whisker stimulation in AD mice. Whether the improved CBF responses contribute to reductions in A␤ levels in the brain by promoting A␤ clearance into cerebral circulation remains to be determined by future studies designed to measure A␤ clearance and production under experimentally different CBF conditions. LRPIV-D3674G did not enter CSF or brain, which is not unexpected because WT-LRPIV also does not cross the BBB (14), and peptides and proteins in general need specific transport systems expressed in brain endothelium for their transport into the brain (66).
The data presented here represent the first screening of LRPIV-derived analogs for A␤ binding. LRPIV has been shown to bind a range of functionally distinct ligands (35,37,38,49). The surface plasmon resonance analysis has shown that the CR24 -CR28 LRPIV region most effectively binds RAP, ␣ 2 M*, Factor VIII light chain, and Factor IXa (35). The three triple repeats, CR24 -CR26, CR25-CR27, and CR26 -CR28, interact avidly with RAP compared with the other CRs of LRPIV (33). Although the A␤ binding site on LRPIV has not been characterized, it seems plausible that this site might overlap partially or fully with the RAP binding site(s) because we have previously shown that RAP abolishes A␤ binding to LRPIV (8). Although A␤40 and A␤42 interacted with all studied CR fractions (CR24 -CR27, CR25-CR28, and CR25-CR27), their binding affinities varied. In contrast to A␤42, full-length WT-LRPIV was required for high affinity binding of A␤40. Other studies have reported that CR24 -CR26 is involved in high affinity binding of ␣ 2 M* and Factor VIII light chain (35). Site-directed mutation at a calcium binding site within CR29 (D3674G) increased the affinity for A␤ (A␤42 Ͼ A␤40) but decreased the affinity for other studied LRP1 ligands, whereas mutations within CR22 (D3394G) and CR26 (D3556G) reduced their respective binding affinities for both A␤40 and A␤42.
Calcium is required for proper folding and structural stability of the CR regions (33). The calcium ion binding requires the side chain of four acidic residues and two carbonyl groups, usu- ally an aspartate residue and an aromatic residue. These two residues are involved in ligand interaction with LRP1 (67,68). CRs (CR23, CR30, and CR31) with imperfect calcium coordination have reduced ligand-binding affinities (35). Also, calcium is required for binding of some LRP1 ligands, such as RAP and ␣ 2 M* (37), but not for other LRP1 ligands, such as apoE and lactoferrin (37). Importantly, mutation in CR29 resulted in conformational change in LRPIV, which enhanced binding of A␤, especially A␤42, but reduced binding of other ligands, including apoE2, apoE3, apoE4, tPA, MMP-9, and Factor IXa. The nature of the conformational change in the LRPIV-D3674G variant and mechanism of the enhanced A␤ binding will require future structural and biophysical studies.
In summary, we have demonstrated that LRPIV-D3674G exhibits significantly improved A␤ binding compared with WT-LRPIV. Moreover, LRPIV-D3674G efficiently cleared A␤ from the brain in wild type and AD mice. In addition, our data show that LRPIV-D3674G binds with reduced affinity several studied LRP1 ligands in vitro and does not affect noticeably their activity in vivo in mice. Therefore, LRPIV-D3674G could be developed as a potential therapeutic agent administered as a monotherapy and/or in combination with other A␤-lowering agents, such as Notch-sparing ␥-secretase inhibitors or ␤-secretase (BACE1) inhibitors (57,69,70) to restore the natural peripheral sink mechanism for A␤ in MCI and AD patients (22,71).