The Low Density Lipoprotein Receptor-related Protein (LRP) Is a Novel -Secretase (BACE1) Substrate*

BACE is a transmembrane protease with -secretase activity that cleaves the amyloid precursor protein (APP). After BACE cleavage, APP becomes a substrate for -secretase, leading to release of amyloidpeptide (A ), which accumulates in senile plaques in Alzheimer disease. APP and BACE are co-internalized from the cell surface to early endosomes. APP is also known to interact at the cell surface and be internalized by the low density lipoprotein receptor-related protein (LRP), a multifunctional endocytic and signaling receptor. Using a new fluorescence resonance energy transfer (FRET)based assay of protein proximity, fluorescence lifetime imaging (FLIM), and co-immunoprecipitation we demonstrate that the light chain of LRP interacts with BACE on the cell surface in association with lipid rafts. Surprisingly, the BACE-LRP interaction leads to an increase in LRP C-terminal fragment, release of secreted LRP in the media and subsequent release of the LRP intracellular domain from the membrane. Taken together, these data suggest that there is a close interaction between BACE and LRP on the cell surface, and that LRP is a novel BACE substrate.

The low density lipoprotein receptor-related protein, LRP, is a type I integral membrane protein with a 515-kDa extracellular ␣-chain non-convalently bound to the 85 kDa membranespanning ␤-chain. It is also found on the cell surface and cycles between the cell membrane and endosomes. Multiple intracellular adaptor and scaffolding proteins bind the LRP 100 amino acid cytoplasmic tail (16,17); its four extracellular binding domains mediate endocytosis of a wide array of ligands, including several of potential importance for Alzheimer disease pathophysiology: APP, apolipoprotein E and ␣ 2 -macroglobulin (16 -18). The LRP ligand binding domains interact with KPIcontaining forms of APP. In addition, an interaction between the C-terminal domain of APP and LRP, mediated by the cytoplasmic adaptor protein Fe65, impacts APP internalization (18 -23). In addition to its role in endocytosis, LRP has an interesting pattern of proteolysis that parallels APP in some ways. Ectodomain shedding of LRP has been described (24) and proteolysis of LRP by matrix metalloproteases was recently reported (25); MT1-MMP also cleaves APP (26) and the postulated ␣-secretases of the ADAM family are also metalloproteinases. Furthermore, and as with APP, ␥-secretase cleavage of LRP leads to release of the LRP intracellular domain (LRP-ICD), which can translocate to the nucleus and interact with Tip60 (27,28).
Given that BACE and APP interact and traffic with one another, and that APP interacts with and traffics with LRP, we now examined whether LRP interacts with BACE. Using both a FRET-based assay of protein proximity and co-immunoprecipitation, we demonstrate that the LRP-ICD interacts with BACE and that this interaction seems to take place in lipid rafts on the cell surface. Surprisingly, the BACE-LRP interaction does not appear to enhance BACE endocytosis from the cell surface. Instead BACE induces LRP extracellular domain cleavage and subsequent release of the LRP intracellular domain from the membrane. Taken together, these data suggest a close interaction between BACE and LRP on the cell surface and suggest that LRP is processed by BACE in a fashion analogous to APP processing.

Generation of Expression Constructs of LRP and BACE and BACE
siRNA-The generation of the LRP light chain with two copies of Myc at its N terminus (amino acids 3844 -4525) (Myc-LC), the minireceptor mLRP1-Myc that encodes the N-terminal cluster of ligand binding repeats fused to the light chain of LRP and tagged with Myc at its C terminus has been described previously (29). These constructs were used instead of full-length LRP because of its functional similarity to and better expression than full-length LRP. To make mLRP1-GFP, mLRP1 was PCRed into pEGFP-N1 (Clontech). To create the LC-LDLR chimera, a unique KpnI restriction site was introduced into the cytoplasmic portion of LC downstream of the transmembrane domain using the QuikChange XL site-directed mutagenesis kit (Stratagene, La Jolla, CA). PCR-generated sequences encoding cytoplasmic domains of the human LDL receptor were then inserted in place of the LRP sequence. To make the LRP165-Myc construct, mLRP1-Myc was digested, and the band containing the vector and the N-terminal 14 amino acids and the C-terminal 165 amino acids of LRP was extracted and self-ligated as described (27). LRP light chain (amino acid 4148 -4544) was fused C-terminally to the Gal4-VP16 synthetic transcription factor, and subcloned into pcDNA3 expression vector (Invitrogen). The leader peptide from LRP was fused to a hemagglutinin epitope, which was then fused to the N terminus of the fusion protein in the construct. Secretory alkaline phosphatase (SEAP) was fused by PCR to the N terminus of LC (amino acids 4018 -4525) in pSecTagB (Invitrogen); the SEAP-cDNA was kindly provided by S. F. Lichtenthaler (A. Butenandt-Institut, Munich, Germany).
An siRNA corresponding to the BACE1 gene was designed as described in Kao et al. (31) and synthesized by Dharmacon (Lafayette, CO). The following sense sequence was used: 5Ј-gctttgtggagatggtgga-3Ј.
Cell Culture Conditions and Transient Transfection-H4 cells derived from human neuroglioma cells, mouse neuroblastoma N2a, pac1, LRP-deficient CHO cells (13-5-1), and HEK293 cells are used in this study. Transient transfection of the cells was performed using a liposome-mediated method (FuGENE 6; Roche Applied Science) according to the manufacturer's instruction. Cells were passaged 24 h prior to transfection and harvested or stained 24 h post-transfection. Primary cortical neurons were prepared as described previously (32). Cortical neurons were isolated from embryonic day 16 CD1 mice (Charles River). Staining was performed 8 days after preparation. For experiments requiring delivery of siRNA, cells were transfected by electroporation per the manufacturer's instructions (AMAXA, Gaithersburg, MD). Specific knockdown of Ͼ70% of overexpressing BACE was observed starting 24 h after transfection by Western blot and immunostaining (data not shown).
Immunocytochemistry and Antibodies-Cells were fixed in 4% paraformaldehyde, permeabilized by 0.5% Triton X-100 in TBS, and blocked with 1.5% normal goat serum. For surface staining Triton-X treatment was omitted. The following antibodies were used: the Golgi organelle marker GM130 mAb and the FITC-conjugated endosomal marker EEA1 mAb (BD Transduction Laboratories, San Diego, CA). The tag antibodies used were rabbit anti-Myc Ab (Upstate Biotechnology, Lake Placid, NY), anti-Myc mAb, anti-V5 mAb (both from Invitrogen), and rabbit anti-V5 Ab (Abcam, Cambridge, MA). Antibodies raised in rabbit against the N (46 -56) and C termini (487-501) of BACE were obtained from Calbiochem. A hybridoma secreting an mAb to the LRP intracellular domain (11H4) was obtained from the American Type Culture Collection. Alexa-555-labeled cholera toxin B (CTx-B, Molecular Probes, Eugene, OR) was used to visualize lipid rafts. Secondary antibodies used were labeled with FITC and Cy3 (Jackson Immunoresearch, West Grove, PA) or Alexa 488 (Molecular Probes). Immunostained cells were coverslipped and mounted for confocal or two photon microscopic imaging. The immunostained cells were observed with the appropriate filters by confocal microscopy using a Bio-Rad 1024 confocal 3-channel instrument.
Isolation of Caveolae and Noncaveolae Membrane (CM and NCM, respectively)-Caveolae and noncaveolae membrane fractions were isolated from rat smooth muscle cells (pac1) using the method of Smart et al. (33). Briefly, 70% confluent rat smooth muscle cells were collected and Dounce-homogenized in a hypertonic buffer containing protease and phosphatase inhibitors. First, the plasma membrane were isolated and sonicated. CM and NCM fractions were prepared using an optiprep gradient with the sonicated plasma membrane samples. 300-l membrane fractions (PM, CM, NCM) were solubilized with 100 l of 4ϫ lysis buffer (4% Nonidet P-40, 200 mM Tris, and 600 mM NaCl) to attain final concentration of 1ϫ (1% Nonidet P-40, 50 mM Tris, and 150 mM NaCl). Following the addition of protease and phosphatase inhibitors, CM and NCM lysates were precleared and immunoprecipitated with anti-LRP monoclonal IgG (5A6)-protein G complex overnight. CM and NCM immunoprecipitates along with PM lysate were separated on 4 -12% SDS-PAGE under nonreducing conditions and transferred onto nitrocellulose membrane. The membranes were probed with 125 I-labeled 11H4 and exposed to Biomax MR film (Kodak).
Tissue Staining-The Massachusetts Alzheimer Disease Research Center Brain Bank provided temporal cortex. Tissue staining was performed with 11H4 labeled by Alexa-488 and BACE-CT (Calbiochem) labeled by Cy3.
Co-immunoprecipitation-Immunoprecipitation experiments were carried out with BioMag beads conjugated to goat anti-mouse IgG (PerSeptive Biosystems, Framingham, MA). The magnetic beads were incubated overnight at 4°C with anti-V5 or anti-Myc mAb or TBS alone. Lysates from H4 cells co-transfected with BACE-V5 and mLRP1-Myc or pure lysis buffer were added to the bead-antibody complex for 2 h at 4°C. After the supernatants were collected, the beads were washed in lysis buffer and then boiled with 2ϫ Tris-glycine SDS sample buffer (Invitrogen) for 3 min. The supernatants were loaded onto 10 -20% Tris-glycine polyacrylamide gels (Novex, San Diego, CA) under denaturing and reducing conditions. The proteins were transferred to polyvinylidene difluoride membrane (Millipore, Bedford, MA) and blocked in 5% nonfat dried milk. mLRP1-Myc was detected by rabbit anti-Myc Ab. BACE was detected by rabbit anti-BACE-NT Ab. Secondary antibodies conjugated to horseradish peroxidase were applied and visualized by chemiluminescence. The Massachusetts Alzheimer Disease Research Center Brain Bank provided temporal cortex. Our protein solubilization procedure was adapted from previously reported studies (35) with minor modifications. The tissue was homogenized at 1 ml/100 mg tissue in ice-cold TEVP-sucrose buffer (containing 10 mM Tris, pH 7.4, 5 mM NaF, 1 mM Na 3 VO 4 , 1 mM EDTA, 1 mM EGTA, and 320 mM sucrose). The homogenates were centrifuged at 4°C, and the supernatants were removed. The pellets were resuspended in 800 l of TEVP with 1% SDS, sonicated for 10 s, and then boiled for 5 min. The samples were centrifuged, and the supernatant was collected for immunoprecipitation after the protein concentration was determined by protein assay (Bio-Rad). Co-immunoprecipitation in human brain tissue was performed as described above with rabbit anti-BACE-CT as pull-down Ab and probed with 11H4 mAb.
FRET Measurements using Fluorescence Lifetime Imaging Microscopy (FLIM)-FRET is observed when two fluorophores are in very close proximity, i.e. Ͻ0 nm. FRET measurements using FLIM relies on the observation that fluorescence lifetimes (the time of fluorophore emission after brief excitation, measured in picoseconds) are shorter in the presence of a FRET acceptor. We have utilized a new FLIM technique that can detect protein-protein proximity using multiphoton microscopy (36,37). A mode-locked Ti-sapphire laser (Spectra Physics) sends a ϳ100-fs pulse every ϳ12.5 ns to excite the fluorophore. Images were acquired using a Bio-Rad Radiance 2000 multiphoton microscope. We used a high speed Hamamatsu MCP detector (MCP5900; Hamamatsu, Ichinocho, Japan) and hardware/software from Becker and Hickl (SPC 830, Berlin, Germany) to measure fluorescence lifetimes on a pixel-by-pixel basis. Excitation at 800 nm was empirically determined to excite GFP, Alexa 488 and FITC, but not Cy3. Donor fluorophore (GFP, Alexa 488, or FITC) lifetimes were fit to two exponential decay curves to calculate the fraction of fluorophores within each pixel that interact with an acceptor. As a negative control, GFP, Alexa 488, or FITC lifetimes were measured in the absence of acceptor (Cy3), which showed lifetimes equivalent to GFP, Alexa 488-IgG, or FITC IgG alone, in solution or with co-transfection with an empty vector (pEGFP) measured in the presence of Cy3-labeled BACE-V5 or LC-Myc. No bleedthrough or mis-excitation of Cy3 was observed under these conditions. Statistical testing was performed by Student's t test.
Internalization Assay-To quantitate BACE internalization we modified a previously reported protocol (38). CHO 13-5-1 (LRP-null cells) were grown to 70% confluency in 6-well plates and transiently transfected with Myc-BACE and either empty vector or LC-GFP. Cells were then washed once with ice-cold PBS containing 1 mM CaCl 2 and 1 mM MgCl 2 , 0.2% bovine serum albumin, and 5 mM glucose (PBSϩϩϩϩ) and 0.4 g/ml biotinylated Myc-mAb (Upstate Biotechnologies) in PBSϩϩϩϩ was applied for 30 min on ice. After that the cells were allowed to endocytose at 37°C for the indicated times. Returning the plates to ice stopped endocytosis. Surface biotin was masked with streptavidin (Roche Applied Science) for 1 h on ice. Avidin was quenched with 0.5 mg/ml biocytin (Sigma). Cells were harvested in blocking buffer (1% Triton X-100, 0.1% SDS, 0.2% bovine serum albumin, 50 mM NaCl, 1 mM Tris, pH 7.4) and incubated on IgG-coated 96-well plates at 4°C overnight. After three washes in PBS, the plates were incubated in streptavidin-peroxidase 1:5000 (Roche Applied Science) in blocking buffer for 1 h. After another wash cycle 3ϫ in PBS, the plates were incubated with 200 l of 10 mg of o-phenyldiamine HCl (Sigma), 10 l of 30% H 2 O 2 (Sigma) in 25 ml of 50 mM Na 2 HPO 4 , 27 mM citrate, pH 5.0. The reaction was terminated by the addition of 50 l of H 2 SO 4 and the A 490 was read. BACE internalization was then graphed as the percentage of internalized Myc-BACE of total surface Myc-BACE.
LRP Ectodomain Secretion Assay-HEK cells passaged into 12-well plates were transfected with a ␤-galactosidase reporter, LRP␤-fused N-terminally to secreted alkaline phosphatase and either empty vector, BACE, or a catalytically inactive BACE mutant. Each condition was transfected in triplicate except for siRNA experiments, which were transfected in duplicate. Media was changed 24 h later, and then collected after another 24 h. Measurement of SEAP activity in the conditioned media was carried out in triplicate by chemiluminescent assay (Roche Applied Science) according to the manufacturer's instructions. SEAP activity was normalized to ␤galactosidase activity, which

FIG. 2. Localization of BACE and mLRP1 in transfected cells.
A, BACE-V5 (green) and mLRP1-Myc (red) co-transfected H4 cells were immunostained with anti-V5 mAb (visualized by Cy5) and rabbit anti-Myc Ab (visualized by Cy3) followed by a FITC-conjugated antibody to the endosomal marker EEA1 (shown in blue). B, BACE-GFP (green) and mLRP1-Myc (red) were co-transfected and then immunostained with rabbit anti-Myc (visualized by Cy3) and an antibody to the Golgi marker GM130 (visualized by Cy5, shown in blue). C, to demonstrate cell surface localization, BACE-V5 (blue) and Myc-(N terminus)-LC (green) were co-transfected and then immunostained with anti-Myc mAb (visualized by FITC) and rabbit anti-BACE-NT Ab (visualized by Cy5) without permeabilizing the cell membrane. Alexa-555-labeled cholera toxin B (CTx-B), used to visualize lipid rafts, was added for 20 min after thoroughly washing the primary and before adding the secondary Ab. D, total plasma membrane (PM) fraction along with CM and NCM membrane fractions immunoprecipitated with anti-LRP monoclonal 5A6 were separated on 4 -12% SDS-PAGE and analyzed by immunoblot analysis using 125 I-anti-LRP monoclonal 11H4. The 60-kDa band represents endogenous BACE and the 75-kDa band transfected BACE-V5 (lane 5). C, human brain extracts were immunoprecipitated with rabbit anti-BACE-CT Ab and probed with 11H4. LC is recognized as an 85-kDa band brain homogenate (lane 5). A specific band of the same size was found after co-immunoprecipitation with anti-BACE-Ab (lane 1). Identical results were observed when probing with 5A6, another LC-specific Ab (data not shown). Negative controls as described above are shown in lanes 2 and 3. Supernatants are shown in lane 4. D, BACE and LRP co-localization in human brain tissue is shown by confocal microscopy.
was measured by hydrolysis of o-nitrophenyl-␤-D-galactopyranoside in cells lysed with reporter lysis buffer (Promega). Pharmacologic inhibition of LRP cleavage was assessed after overnight treatment with vehicle (Me 2 SO) or a cell-permeable, peptidomimetic inhibitor of BACE (Calbiochem) (39).
Western Blotting-N2a cells co-transfected with LC-Myc and either empty vector, BACE, or a catalytically inactive BACE mutant and treated with 1 M ␥-secretase inhibitor DAPT (40) for 12 h (a generous gift from M. Wolfe, Brigham and Women's Hospital, Boston, MA) were lysed in 1% Triton X-100 in TBS buffer and proteinase inhibitor tablets (Roche Applied Science) and then loaded onto 4 -20% Tris-glycine polyacrylamide gels (Novex) under denaturing and reducing conditions. The proteins were transferred to polyvinylidene difluoride membrane and LRP light chain was detected by rabbit anti-Myc Ab with Alexa 680 (Molecular Probes) goat anti-rabbit secondary and visualized on a Licor Odyssey near-infrared gel reader (Lincoln, Nebraska).
Luciferase Assay-HEK293 cells were transfected with LRP-Gal4-VP16 (LRP-GV) in the absence or presence of BACE and relative luciferase activity determined (28). Activity relative to ␤-galactosidase is shown and averaged for triplicate transfection. In all cases transfection was confirmed by immunoblotting.

RESULTS
Localization of BACE and LRP Constructs-We first tested the localization of BACE and LRP in co-expressing H4 cells. When expressed individually, both mLRP1 and BACE were localized mainly in punctate structures in the cells. mLRP1-positive structures largely overlapped with BACE-positive structures when they were co-expressed. To determine the subcellular distribution we immunostained co-expressing H4 cells with organelle markers or, in cell surface stained without Triton X-100 treatment, Alexa555-labeled CTx-B as a raft marker. mLRP1 and BACE co-localized in the endosomal compartments stained by EEA1 (Fig. 2A). To a lesser extent, the Golgi marker GM130 also overlapped with mLRP1 and BACE (Fig. 2B). On the cell surface Myc-LC and BACE are partly co-localized with one another in lipid rafts. The results of the immunocytochemistry suggest that LC and BACE are co-localized in distinct compartments of the cell including lipid rafts (Fig. 2C), Golgi and prominently in the endosomal compartment.
To confirm that LRP localizes to lipid rafts we prepared total membrane and separated CM and NCM fractions using an optiprep gradient. LRP was present in caveolae as well as in noncaveolae fractions (Fig. 2D), which is in accordance with our confocal data showing partial overlap with the lipid raft marker CTx-B. We then looked for co-localization under physiological conditions. By staining human brain sections, includ-ing the hippocampal formation, we were able to observe similar results in neurons expressing endogenous levels of LRP and BACE (Fig. 3D).
Co-immunoprecipitation of BACE and LRP in Human Brain Tissue-From the immunohistochemical experiments that showed robust co-localization in both transfected cells and human brain tissue (Fig. 3D), we hypothesized that there may be a close interaction between LRP and BACE. To test whether LRP interacts with BACE, we immunoprecipitated BACE from co-transfected H4 cells and probed for mLRP1 (lane 1), controlling for nonspecific interactions by assessing lysates incubated without the pull-down antibody (lane 2) or pure lysis buffer (lane 4). Immunoreactive bands of ϳ100 kDa (resembling mature furin-cleaved mLRP1-Myc) and ϳ140 kDa (resembling unprocessed Golgi and ER forms of mLRP1-Myc) (29) were detected in the immunoprecipitated sample and the whole cell lysate (Fig. 3A, lane 4). To confirm this interaction, the complementary pull-down experiment was performed. mLRP1 immunoprecipitates were probed by an anti-BACE Ab with the same controls. The doublet bands of ϳ60 and 75 kDa were detected only in lane 1 and in the control lysate lane (Fig. 3B), suggesting that BACE is present in the mLRP1 immunoprecipitates. The 60-kDa band represents endogenous BACE, whereas the 75-kDa band represents transfected BACE containing a V5-His tag.
To demonstrate a direct interaction of LRP and BACE under physiological conditions in brain, where BACE function is presumed to be important in the pathogenesis of Alzheimer disease, we immunoprecipitated BACE from human brain tissue and probed with an antibody to the LRP light chain. A strong immunoreactive band of ϳ85 kDa (the expected size of the mature endogenous light chain of LRP, i.e. furin-cleaved form) (29) was detected in the sample lane (Fig. 3C, lane 1). Control precipitate from samples lacking anti-BACE Ab (lane 2) showed only a weak immunoreactive band, and samples lacking cell extract (lane 3) did not contain this band. Thus, endogenous LRP and BACE co-immunoprecipitate.
Interaction of LRP and BACE by FLIM Analysis-We next used an alternative technique to probe protein-protein proximity to test the idea that the LRP-BACE interaction detected by co-immunoprecipitation occurs in specific cell compartments, and to further evaluate the biochemical parameters of this interaction. We utilized FLIM, a morphology-based FRET tech- nique that can reveal close protein-protein proximity in intact cells. Fluorescence lifetime is influenced by the surrounding microenvironment and is shortened in the immediate vicinity of a FRET acceptor molecule. The degree of lifetime shortening can be displayed with very high spatial resolution in a pseudocolor-coded image. As shown in Fig. 2, double immunostaining showed subcellular compartment co-localization of BACE and LRP predominantly in endosomal compartments, but this does not necessarily imply a close interaction. We measured changes in the lifetime of the donor fluorophore (either FITC, Alexa 488, or GFP) under different experimental conditions. In the absence of an acceptor fluorophore, the lifetime of FITC conjugated to IgG (hereafter referred to simply as FITC) is ϳ2300 ps, GFP ϳ2200 ps, and Alexa 488 is ϳ1900 ps. If an acceptor fluorophore is present but remains too distant from the donor (i.e. there is no interaction), donor lifetimes remain in this range. The lifetime of FITC attached to the C terminus of BACE-V5 alone (2334 Ϯ 72 ps) was significantly shortened when co-expressed mLRP1-Myc was C-terminally labeled by Cy3 (2153 Ϯ 55 ps, p Ͻ 0.0001), indicating FRET between the two fluorophores (Table I). Equivalent results were obtained when BACE was tagged with GFP and when the acceptor and donor fluorophores were exchanged (Table I). In order to test the idea that the decrease in lifetime observed in the mLRP1-GFP/BACE-V5 FLIM assay was because of FRET, we performed an additional control. mLRP1-GFP and Myc-BACE cotransfected cells were stained with anti-Myc Ab, then labeled with Cy3. In this experiment, the Myc tag was at the N terminus of BACE, across the membrane and, therefore, too distant from the C-terminal GFP on mLRP1-GFP to be detected by FRET. Although there was striking co-localization, no lifetime change was observed. This experiment demonstrates the specificity of the proximity assay in this FLIM-based method of measuring FRET. FLIM allows analysis of FRET localization by recording the distribution of donor lifetimes on a pixel-by-pixel basis. FITC bound to BACE-V5 in the absence of acceptor has a uniform lifetime and a single lifetime peak (Fig. 4A); in the presence of mLRP1-Cy3 acceptor, FITC has two distinct lifetimes and is faster, representing "FRETting" molecules (Fig. 4B). Examination of the FLIM images of transfected cells suggests that LRP and BACE interact (red pseudocolor) at the cell surface and in endosomal compartments. Because LRP and BACE co-localize in lipid rafts at the cell surface, we suggest that the interaction detected by FLIM also is lipid raft-associated.
Indeed, additional experiments in which only cell surface Myc-LC (N-terminal Myc) and BACE-NT are immunostained demonstrate FRET (Table II) between BACE and LRP specifically in punctuate structures at the cell surface (Fig. 5B). To test the hypothesis that this interaction occurs in rafts we cholesterol depleted the cells and repeated the cell surface FLIM experiment. Cholesterol depletion weakened the interaction of BACE and LRP at the cell surface (Fig. 5C). This result strongly suggests that BACE and LRP interact on the cell surface distinctively in lipid rafts.
To further confirm the physiologic interaction of LRP and BACE we performed FLIM in untransfected N2a cells as well as primary cortical neurons, which have relatively abundant BACE and LRP; Alexa 488 was chosen as donor fluorophore because it is somewhat brighter than FITC under the conditions utilized.  (Table III). In the absence of acceptor, Alexa 488 has a uniform lifetime; in the presence of acceptor a second peak appears, reflecting an interaction (Fig. 4, C-F). The interaction, pseudocolored red, also appears to be stronger in the distal compartments at or near the cell surface. This result demonstrates close protein-protein interaction between endogenous LRP and endogenous BACE at the cell surface in a neuronal cell type, paralleling the co-immunoprecipitation results. To identify the domain of LRP interacting with BACE, we utilized LRP deletion constructs. LC, which contains the ␤-chain of LRP, and LRP165, a construct that contains only the 100 amino acid intracellular domain, the transmembrane domain and a very small extension beyond the membrane, both interacted strongly with BACE (Table I). This result implicates the intracellular or the transmembrane domain of LRP as the site of interaction. To further test this hypothesis, we utilized a chimeric protein in which the extracellular and transmembrane domains of LC are fused to the intracellular domain of the low density lipoprotein receptor (LDLR) (Fig. 1) and performed FLIM on the cell surface as described above. This construct did not FRET with BACE (Table II), further supporting the importance of the intracellular domain of LRP for this interaction.
There are several scaffold/adaptor proteins known to interact with LRP including Fe65 (22,23,41) and mammalian disabled 1 (mDab1) (23). If the intracellular domain is responsible for the interaction between LRP and BACE, these adaptor proteins may play a role in the interaction. We have demonstrated previously that LRP and Fe65 interact using a FRET based cell assay (22), and that Fe65 is responsible for mediating an LRP-Fe65-APP heterotrimeric complex. We therefore examined the possibility that these molecules may interact with BACE by the FRET assay. However, under the conditions utilized we did not detect any FRET between BACE and Fe65 or between BACE and mDab1 (data not shown). Recent data also suggest that phosphorylation of BACE at Ser 498 changes its trafficking possibly by altering its interactions with GGA by its C-terminal dileucine motif (30,42). We generated the S498D, S498A, and L499A/L500A mutants of BACE to evaluate if these mutants, which mimic or block phosphorylation of Ser 498 (15), alter interaction with LRP. No changes in FRET measures were observed with these manipulations (data not shown). BACE Internalization Assay-Because we observed LRP-BACE interactions dependent on a domain of LRP that mediates APP endocytosis, we hypothesized that LRP might also influence BACE endocytosis and thereby regulate APP cleavage. In order to assess the effect of LRP on BACE endocytosis we assessed internalization of BACE after biotinylation of its N-terminal Myc tag at the cell surface and assayed internalized versus cell surface BACE over time (Fig. 6). Co-transfection with LC did not enhance BACE internalization from the cell surface in LRP-null CHO cells (13-5-1) in contrast to known enhancement of APP endocytosis with LC (20). The same results were obtained using PEA13 (LRPϪ/Ϫ) fibroblasts (data not shown). Thus it appears that BACE internalization is not mediated by LRP under these conditions. LRP Shedding-Because LRP is a known ␥-substrate, we hypothesized that it might also be cleaved by ␤-secretase. In order to assess the effect of BACE and LRP interaction on LRP processing, we measured shedding of the extracellular domain of LRP with the cDNA of SEAP fused to the N terminus of the LRP ␤-chain. An analogous construct has been used to study BACE cleavage of APP (43). After co-transfection of the SEAP-LRP construct with a ␤-galactosidase reporter construct and either empty vector, WT-BACE or BACE D93/289A, SEAP activity was measured in the medium and normalized to ␤-galactosidase activity. WT-BACE led to a significant increase in LRP ectodomain shedding compared with baseline, whereas, as expected, catalytically inactive BACE D93/289A exhibited no effect (Fig. 7A). To confirm that LRP is cleaved by BACE, we used two additional approaches. Overnight treatment with a cell-permeable BACE inhibitor (39) reduced LRP cleavage   when treating cells that expressed BACE at endogenous levels, indicating that the observed effect has physiologic relevance and is not restricted to cells overexpressing BACE (Fig. 7B).
The second approach to inactivating BACE utilized siRNAmediated silencing. Knocking down overexpressed BACE by co-transfection of BACE1-specific siRNA reduced processing of LRP by BACE (Fig. 7C). LRP C-terminal Fragment (CTF) Production after BACE Cotransfection-After treatment with DAPT for 12 h we observed an increase of the LRP-CTF 25-kDa band consistent with the observations of (28) suggesting ␥-cleavage of LRP. We hypothesized that, like APP, the direct substrate of ␥-secretase activ-ity would be an N-terminally cleaved form of LRP. To determine if BACE activity would produce an LRP-derived ␥-secretase substrate, we co-transfected LC with BACE. This led to an increase of LRP-CTF detected by Western blot as a 25-kDa band. To confirm that this band is the substrate of ␥-secretase activity, we repeated this experiment in the presence of the ␥-secretase inhibitor DAPT, and detected an increased amount of the 25-kDa product. The presence of BACE further increased and co-transfection with catalytically inactive BACE mutant did not increase this band intensity (Fig. 8). These results are consistent with a BACE-mediated cleavage of LRP, generating a CTF for ␥-cleavage.

Interaction of ␤-Secretase and LRP Protein
Luciferase Assay-LRP interaction with ␤-secretase leads to LRP cleavage and generation of a truncated form. By analogy to APP, it may then undergo further proteolysis and release the LRP-ICD fragment. Evidence that LRP is itself a ␥-substrate has been presented using an LRP-Gal4/VP16 construct utilizing a luciferase reporter assay for LRP C-terminal cleavage (28). We used this same assay to determine if co-transfection with BACE would alter generation of this putative signaling domain. Co-transfection with BACE led to a substantial increase in luciferase activity suggesting that LRP undergoes BACE cleavage, leading ultimately to release of the cytoplasmic domain and translocation to the nucleus (Fig. 9). DISCUSSION In the present study, we have demonstrated that BACE interacts with the intracellular domain of LRP, a multifunctional endocytic receptor. The interaction was demonstrated by co-immunoprecipitation of BACE and LRP from overexpressing cells and from endogenous BACE and LRP in human brain samples and primary neurons. Co-localization and close proximity in both H4 and N2a cells was shown by confocal microscopy and FRET-based proximity assays, suggesting co-localization primarily in endosomes and at the cell surface. Cholesterol depletion disrupted the cell surface LRP interaction, suggesting that the interaction occurs in lipid rafts.
The FRET technique used here, FLIM, is advantageous because it provides quantitative data on protein-protein proximity with exquisite subcellular localization. The two fluoro-phores must be quite close (Ͻ10 nm) to support FRET; tagging LRP and BACE molecules "across the membrane" from one another or after cholesterol depletion abolishes FRET despite continued co-localization at the light level. Analogous results were obtained using three different fluorophores, multiple different antibody pairs, endogenous or transfected LRP and BACE, and two different cell types. Taken together with the co-immunoprecipitation data, our results strongly support the conclusion that LRP and BACE interact at the cell surface in raft compartments.
In contrast to ␥-secretase, where at least 15 substrates have been described (44), BACE has so far appeared to be relatively specific. Sialyltransferase and PSGL-1, in addition to APP and its homologues APLP1 and 2 are the only other reported substrates of BACE (6,43). A family of GGA adaptor proteins and the phospholipid scramblase 1 (PLSCR1) have been shown to directly interact with the BACE tail (42,45). Interestingly, both BACE and PLSCR1 were localized in a low buoyant lipid microdomain, a potential site of interaction with APP and LRP.
In summary, our data demonstrate a close interaction of LRP and BACE in specific subcellular compartments. Although BACE and LRP are mostly co-localized in the endosomal compartments and to a lesser extent in the Golgi and on the cell surface as shown by conventional immunostaining, our FLIM data suggest that they come into closest proximity at the cell surface in lipid rafts, where it has been shown that amyloidogenic processing seems to occur in raft associated compart- FIG. 8. LRP CTF Western blot. N2a cells co-transfected with LC-Myc and either empty vector, BACE, or a catalytically inactive BACE mutant (deadBACE) were treated with vehicle (DMSO) or DAPT (12 h). Increases of CTF intensity were detected after DAPT treatment and/or BACE co-transfection. Shown is the intensity of a representative blot out of four experiments without DAPT treatment and two with DAPT treatment.
FIG. 9. BACE increases release of LRP-ICD. HEK293 cells were transfected with LRP-Gal4/VP16 (28) in the absence or presence of BACE, and luciferase activity determined. Activity relative to ␤-galactosidase is shown as the average of triplicate analyses. The addition of BACE led to a statistically significant increase in luciferase activity, suggesting that LRP undergoes BACE cleavage, and leading to subsequent release of the cytoplasmic domain and translocation to the nucleus to activate the Gal4 assay. ments (13). This is in good accordance with a recent article showing that ApoE, a LRP ligand, is co-localized with APP and BACE in lipid rafts (46). Moreover, our SEAP-LC assay demonstrates that BACE-mediated cleavage of the LRP extracellular domain leads to secretion of the shed domain into the extracellular milieu. Because even the LRP165 construct (which lacks ligand binding domains and most of the extracellular part of LRP) interacts with the BACE intracellular domain and the LC-LDLR-chimera lacking the LRP intracellular domain did not interact with BACE, we postulate that the major interaction site is the intracellular domain of LRP. Whereas two candidate adaptor proteins, Fe65 and mDab, did not appear to mediate the interaction, multiple other potential adaptor proteins might either mediate or impact this interaction. Surprisingly we did not find enhancement of BACE endocytosis in LRP-null cells co-transfected with LRP. Thus, although LRP co-traffics with BACE to the cell surface, it does not appear to be critical for BACE recycling from the cell surface to endosomes. Our current data, showing an interaction between LRP and BACE, when viewed in the context of previous studies of APP-LRP interactions, suggest that LRP has a complex role in modulating APP processing. The LRP C-terminal domain also mediates an interaction with APP in similar cell compartments (22). Thus, LRP, which is highly enriched in rafts, potentially acts as a scaffolding complex in rafts for APP and for BACE; such an interaction may help explain the observations that directing APP to rafts enhances ␤-cleavage and A␤ generation (13).
We found that BACE activity at endogenous levels leads to an increase of secreted LRP in the medium as well as LRP-CTF, analogous to APP processing. Thus we now show that LRP is a substrate for both BACE and ␥-secretase, identified as APP-processing enzymes. Whether this processing leads to a stable LRP equivalent of A␤ is unknown. Moreover BACE overexpression leads to an increase of ␥-secretase-like cleavage to release LRP-ICD. Although LRP-ICD has been shown to translocate to the nucleus and interact with Fe65 and Tip60 (28,47), whether or not it has a transcriptional role under physiological conditions remains unknown. However, both LRP (48,49) and other members of the LDL receptor-related family (50) have been implicated as having signaling roles in neurons, and it seems likely that the cleavage of LRP we observe could modulate such signaling. Further studies will be needed to elucidate if APP and LRP act co-operatively or competitively for access to these secretases and how interactions of LRP with other ligands impact these processes.