The Low Density Lipoprotein Receptor-related Protein 1 Mediates Uptake of Amyloid β Peptides in an in Vitro Model of the Blood-Brain Barrier Cells*

The metabolism of amyloid β peptide (Aβ) in the brain is crucial to the pathogenesis of Alzheimer disease. A body of evidence suggests that Aβ is actively transported from brain parenchyma to blood across the blood-brain barrier (BBB), although the precise mechanism remains unclear. To unravel the cellular and molecular mechanism of Aβ transport across the BBB, we established a new in vitro model of the initial internalization step of Aβ transport using TR-BBB cells, a conditionally immortalized endothelial cell line from rat brain. We show that TR-BBB cells rapidly internalize Aβ through a receptor-mediated mechanism. We also provide evidence that Aβ internalization is mediated by LRP1 (low density lipoprotein receptor-related protein 1), since administration of LRP1 antagonist, receptor-associated protein, neutralizing antibody, or small interference RNAs all reduced Aβ uptake. Despite the requirement of LRP1-dependent internalization, Aβ does not directly bind to LRP1 in an in vitro binding assay. Unlike TR-BBB cells, mouse embryonic fibroblasts endogenously expressing functional LRP1 and exhibiting the authentic LRP1-mediated endocytosis (e.g. of tissue plasminogen activator) did not show rapid Aβ uptake. Based on these data, we propose that the rapid LRP1-dependent internalization of Aβ occurs under the BBB-specific cellular context and that TR-BBB is a useful tool for analyzing the molecular mechanism of the rapid transport of Aβ across BBB.

recapitulate the transport of A␤ and other macromolecules across the BBB and with which a precise molecular mechanism of A␤ transport across the BBB could be elucidated.
To verify the molecular and cellular mechanisms of A␤ transport across the BBB, application of cell biological and physiological techniques at a single cell level is mandatory. In this study, we adopted the TR-BBB cells, a conditionally immortalized cell line derived from brain capillary endothelial cells of transgenic rats expressing temperature-sensitive large T antigen (19), whose inactivation upon incubation at 37°C renders the cells into a nonimmortalized state similar to primary BMECs. TR-BBB cells have been shown to express a number of receptors and transporters expressed in endothelial cells comprising the BBB (e.g. GLUT-1, P-glycoprotein (P-gp), and other influx or efflux transporters, allowing for the characterization of their functions in vitro (20).
Using TR-BBB cells, we established an in vitro model of A␤ uptake and found that LRP1 is involved in the rapid and robust A␤ internalization in TR-BBB cells. In contrast, fibroblasts or neuroblastoma cells that express LRP1 did not internalize A␤, suggesting that LRP1 is not sufficient for the rapid internalization of A␤, implicating cell type specificity in the LRP1-dependent A␤ uptake. Our observations provide a new clue to the molecular mechanism of A␤ uptake and transport across the BBB.
Preparation of A␤-125 I-A␤ peptides were solubilized in distilled water at a concentration of 0.1 nM and stored at Ϫ80°C until use. Synthetic A␤ peptides were solubilized in 1,1,1,3,3,3hexafluoro-2-propanol (Kanto Chemical) at a concentration of 1 mg/ml, dried, and resolubilized in phosphate-buffered saline containing 3% (v/v) Me 2 SO (Kanto Chemical) upon use. We confirmed that 125 I-A␤ migrates at 4 kDa on SDS-PAGE and does not show signs of aggregation at concentrations used in this study (see supplemental Fig. 1). In addition, synthetic A␤ utilized in this study showed no sign of fibrillization as examined by a thioflavin T fluorescence assay (data not shown).
In Vitro Uptake Assays of 125 I-Labeled Proteins in TR-BBB Cells-The uptake of 125 I-A␤ or 125 I-tPA into TR-BBB cells was examined as reported previously (33). Briefly, TR-BBB cells cultured onto collagen I coated 24-well dishes were grown to 90 -100% confluence. Prior to experiments, cells were washed three times with 1 ml of ECF buffer (138 mM NaCl, 5.0 mM KCl, 1.3 mM CaCl 2 , 0.8 mM MgCl 2 , 0.3 mM KH 2 PO 4 , 0.3 mM Na 2 HPO 4 , 5.6 mM D-glucose, 10 mM HEPES, pH 7.4) and incubated at 37°C with 125 I-labeled ligand proteins for a predetermined time period. After incubation, 125 I-labeled ligand proteins were removed, and the cells were washed three times with 1 ml of ice-cold ECF buffer and an additional three times with acid wash buffer (28 mM CH 3 COONa, 120 mM NaCl, 20 mM sodium barbital, pH 3.0). In some experiments, cells were further treated with Pronase (1 mg/ml; Sigma) for 1 h at 4°C to completely remove cell surface-attached radioligands. The amount of surface-bound labeled ligands was calculated as the amount of ligands released by this treatment, and the amount of internalized ligands was defined as the amount of ligands that remained associated with the cell pellet following this treatment. Cells were solubilized with 200 l of 5 M NaOH for 12 h, and the protein amount in the cells was measured by the Lowry method. The amount of the cell-associated ligands was expressed as the cell/medium ratio (see below).
The cell/medium ratio (l/mg of protein) is equal to 125 I counts in the cells (cpm/mg protein)/ 125 I counts in the incubation medium (cpm/l). When the effects of various inhibitors or siRNA against LRP1 were studied, TR-BBB cells were incubated with 125 I-A␤ and 125 I-tPA for 5 and 10 min, respectively.
In the efflux assay, TR-BBB cells were first incubated with 125 I-A␤ for 10 min (first incubation). After incubation, the media were replaced with the ECF buffer (incubation medium), and cells were further incubated at 37°C for a various time period to allow efflux of A␤ (second incubation). Degradation of 125 I-A␤ was studied by a trichloroacetic acid precipitation assay (15). Incubation medium was mixed with trichloroacetic acid (final concentration, 10%) and bovine serum albumin (final concentration, 1.25%) and centrifuged for 20 min, and the radioactivities in the precipitate and supernatant were quantitated by a ␥ counter. We confirmed that ϳ85-90% of total 125 I-A␤ in medium prior to the first incubation was trichloroacetic acid-precipitable. Thus, we quantitated the extent of trichloroacetic acid precipitability of A␤ prior to the first incubation upon each experiment and normalized the level of trichloroacetic acid-precipitable A␤ in incubation medium at each time point upon the second incubation by the precipitation extent and indicated the ratio of trichloroacetic acid-precipitable A␤ as the percentage of that at time 0 of the first incubation. For the uptake assay of A␤⅐apoE complex, A␤ was preincubated with apoE4-containing lipoproteins (2 g/ml) for 30 min at room temperature.
To examine the involvement of heparan sulfate proteoglycans or chondroitin sulfate proteoglycans in 125 I-A␤ uptake, TR-BBB cells were preincubated in medium containing heprinase I (Seikagaku), heparinase II (Sigma), heparinase III (Sigma), or condroitinase ABC (Seikagaku) at a concentration of 1 sigma unit/ml for 4 h at 37°C, followed by three washes with ECF buffer.
Transfection of Small Interference RNA (siRNA) against Rat LRP1-Control siRNA (Stealth TM RNAi negative control medium GC duplex 2) and two Stealth RNAis against rat LRP1 (siRNA 1, 5Ј-UUGACAUUCGGAUCACAGACGUGGG-3Ј; siRNA 2, 5Ј-AUGAAUGCCCGUUUGAUGGCUUGGG-3Ј) were purchased from Invitrogen. TR-BBB cells cultured in 24-well dishes were transfected with 42 pmol of siRNA duplex using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions. The Lipofectamine RNAiMAX and siRNA mixture were incubated with TR-BBB cells for 30 h. Transfected TR-BBB cells were collected for immunoblotting or uptake assays as described above.
Binding of A␤ to Immobilized LRP1 in Vitro-An in vitro binding assay was performed as described previously (29). Briefly, microtiter wells were coated with purified full-length human LRP1 or recombinant sLRP2, sLRP4, or bovine serum albumin (negative control) at 4°C. Unoccupied sites were blocked with Block Ace (Snow Brand, Sapporo, Japan) and washed with phosphate-buffered saline containing 0.05% Tween 20. Wells were then incubated with 100 nM A␤ or tPA in the presence or absence of 1 M RAP for 12 h, washed with phosphate-buffered saline containing 0.05% Tween 20, and then reacted with an anti-A␤ antibody (BA27; gift of Takeda Pharmaceutical Co., Ltd.) (34) or anti-tPA (Oxford Biomedical Research) antibodies for 3 h. After incubation with a horseradish peroxidase-tagged secondary antibody (GE Healthcare), the levels of ligands bound to LRP1 were quantitated by development using a TMB microwell system (KPL, Inc.).
Preparation of Radiolabeled tPA-Human tPA was iodinated, using IODO-GEN-precoated tubes, according to the manufacturer's instructions. The unincorporated 125 I was removed using a D-salt polyacrylamide desalting column.

A␤ Is Internalized by TR-BBB Cells in a Specific Manner-
We reasoned that if A␤ is transported across the BBB, A␤ should be internalized by BMECs. To test this, we used TR-BBB cells, an in vitro cellular model in which the internalization in BBB endothelial cells can be analyzed (19). We first evaluated the uptake of 125 I-A␤ in monolayer cultures of TR-BBB cells.
First, we checked whether 125 I-A␤-(1-40) does not form oligomers at the concentration used in this study (0.1 nM) by a light-induced chemical cross-linking experiment, PICUP, an efficient method for detecting SDS-sensitive A␤ assemblies by forming covalently cross-linked A␤ oligomers (30,31). The oligomerization state of each peptide after cross-linking was determined by immunoblot analysis or autoradiography. When 20 M of unlabeled A␤ was subjected to PICUP reaction, several covalently cross-linked oligomer bands (i.e. dimer, trimer, and tetramer) appeared in an irradiation-dependent manner, suggesting that a fraction of A␤ formed SDS-sensitive oligomers in the solution (supplemental Fig. 1). In contrast, 0.1 nM 125 I-A␤-(1-40) did not exhibit any cross-linked oligomers, and only a single band migrating at 4 kDa was detected even after the PICUP reaction (supplemental Fig. 1). This suggested that the concentration of 125 I-A␤-(1-40) used in this experiment was low enough to eliminate formation of oligomers.
Cells grown in collagen-coated dishes were incubated with 125 I-A␤- . After incubation, cells were extensively washed with acid wash buffer to remove 125 I-A␤ that remained on cell surface. The remaining radioactivities in TR-BBB cells that represent 125 I-A␤ were calculated as cell/medium ratio (l/mg) (see "Experimental Procedures"). Incubation with 125 I-A␤ at 37°C elicited a rapid and robust uptake of 125 I-A␤ by TR-BBB cells with total uptake reaching a plateau within 8 min (Fig. 1A). This rapid uptake of 125 I-A␤ was not observed at 4°C, suggesting that the A␤ uptake is a receptor-mediated event. The uptake was competed by the addition of excess unlabeled synthetic A␤-(1-40) in a concentration-dependent manner, with an IC 50 of ϳ600 nM (Fig. 1B). Because the uptake assays
To more rigorously estimate the amount of internalized A␤ into TR-BBB cells, we further treated TR-BBB cells with Pronase after incubation with 125 I-A␤ at 37°C to remove surfacebound A␤. We defined "cell-bound" A␤ as the sum of surfacebound A␤ that includes the binding to specific receptor(s) as well as nonspecific binding and defined "internalized" A␤ as the amount of A␤ taken up into cells and resistant to Pronase treat-ment (Fig. 1C). TR-BBB cells retained certain levels of 125 I-A␤ radioactivity even after Pronase treatment, which was reduced by co-incubation with unlabeled A␤- , confirming that A␤ is actually internalized into cells (Fig. 1D). We also found that unlabeled A␤-(1-42) blocked internalization of 125 I-A␤-  with an efficacy similar to that of A␤-(1-40) (Fig. 1D). Taken together, our data show that TR-BBB cells rapidly internalize A␤ possibly through a receptor-mediated mechanism and that A␤-(1-40) and A␤-(1-42) probably share the same receptor.
TR-BBB Cells Partially Degrade Internalized A␤ and Release Fragments and Intact A␤ into Culture Medium-To examine the fate of internalized A␤, we performed an efflux assay. TR-BBB cells were first incubated with 125 I-A␤ for 10 min, and the media were then replaced with the ECF buffer. Cells were further incubated at 37°C for various periods of time to allow efflux of A␤. We found that the cell/medium ratio of A␤ internalized into TR-BBB cells was rapidly decreased over the incubation time (shown as cell in Fig. 2A). Concurrently, the radioactivity in the incubation medium was increased (shown as medium in Fig. 2A), suggesting that the internalized A␤ is rapidly effluxed into the culture medium ( Fig. 2A). We next precipitated A␤ released into culture medium by trichloroacetic acid and found that ϳ80% of 125 I-A␤ in the culture medium escaped trichloroacetic acid precipitation and that the ratio of 125 I-A␤ precipitated by trichloroacetic acid was constant for 15 min (Fig. 2B). However, SDS-PAGE analysis of culture medium revealed that a portion of effluxed A␤ remained intact and migrated at 4 kDa, which increased in a time-dependent manner (Fig. 2C). These data suggest that a significant proportion of internalized A␤ was degraded within TR-BBB cells before being released into the extracellular space.
A␤ Internalization Is Inhibited by RAP or Anti-LRP1 Antibody-We next examined whether LRP1 plays a major role in the internalization of 125 I-A␤ into TR-BBB cells. We conducted the experiments in the presence of RAP, an antagonist of low density lipoprotein receptor family proteins, including LRP1 (39). When TR-BBB cells were incubated with 125 I-A␤ in the presence of RAP, the uptake of 125 I-A␤ was significantly inhibited (Fig. 3A). The strong inhibitory effect of A␤ uptake by RAP was observed even after Pronase treatment. The internalization of 125 I-A␤ by TR-BBB cells was significantly inhibited by RAP, at an extent of ϳ90% of that inhibited by unlabeled A␤-(1-40) (Fig. 3B). A␤ binding to the cell surface was also blocked by RAP, at an extent of ϳ95% of that by unlabeled A␤- , supporting the notion that the major proportion of A␤ internalization is a RAP-sensitive process. It has been suggested that apoE is involved in A␤ clearance (36,40). To examine the effect of apoE on 125 I-A␤ internalization, we prepared an 125 I-A␤⅐apoE complex by prein- I-A␤ uptake was quantitated as cell/medium ratio. The mean Ϯ S.E. in three independent assays is shown. **, p Ͻ 0.001. B, the uptake assay of 125 I-A␤ was performed in the presence of 0, 30, 100, 300, 1000, and 3000 nM unlabeled A␤ as a competitor. The mean Ϯ S.E. in three independent assays is shown. C, a schematic diagram depicting the uptake assays using Pronase treatment in TR-BBB cells. "Cell-bound A␤" was defined as radioactivities released by this treatment and "internalized A␤" as radioactivities associated with cell pellet after the treatment. D, the levels of cell-bound and internalized A␤ were measured in the presence or absence of 3 M unlabeled A␤-(1-40) or 3 M unlabeled A␤-(1-42), respectively. The mean Ϯ S.E. in three independent assays is shown. **, p Ͻ 0.01.  DECEMBER 12, 2008 • VOLUME 283 • NUMBER 50 cubating 125 I-A␤ with astrocyte-derived human apoE4 for 30 min and added it to TR-BBB cells. 125 I-A␤ uptake was increased following incubation with apoE, compared with that by 125 I-A␤ alone, suggesting that 125 I-A␤⅐apoE complex formation enhanced 125 I-A␤ uptake (Fig. 3A). However, the uptake of 125 I-A␤⅐apoE complex was not inhibited by RAP (Fig. 3A), suggesting that the uptake of free A␤ and that of A␤⅐apoE complex are mediated by different pathways in TR-BBB cells. We also confirmed that the effect of RAP on A␤ uptake was concentrationdependent, with an IC 50 of ϳ1 nM (Fig. 3C), the latter being almost identical to those previously reported for LRP1 (39,41).

LRP1-mediated A␤ Clearance in TR-BBB Cells
We further examined the possibility of involvement of other A␤ receptor candidates. RAGE and P-gp were reported to mediate the A␤ transport across the BBB from blood to brain (42) and brain to blood (43), respectively. Neither anti-RAGE antibody, which is known to inhibit RAGE function, nor verapamil, a major substrate of P-gp that works as a competitor, affected the internalization of A␤, excluding an involvement of RAGE and P-gp in the A␤ uptake by TR-BBB cells (Fig. 3D).
We further performed A␤ uptake experiments in the presence of a neutralizing antibody that specifically inhibits LRP1 function. Western blot analysis revealed that an antibody RRR, generated against LRP1 holoprotein purified from human placenta, specifically recognized the heavy chain of endogenous LRP1 in TR-BBB cells (data not shown). Coincubation with RRR robustly inhibited A␤ uptake by TR-BBB cells at an extent of ϳ69% of that inhibited by RAP, in a concentration-dependent manner (Fig. 4, A and B). A␤ binding to the cell surface was also inhibited by RRR at ϳ32% compared with that in the presence of control immunoglobulin (data not shown). These data strongly support the notion that A␤ internalization into TR-BBB cells is mediated by LRP1.
RNAi Knockdown of Endogenous LRP1 Reduced A␤ Internalization in TR-BBB Cells-To further confirm the involvement of LRP1 in the uptake of A␤ into TR-BBB cells, we knocked down LRP1 in TR-BBB cells by siRNA treatment using two stealth RNAis, a chemically modified RNA molecule that eliminates the induction of interferon pathway, against LRP1 and a stealth negative control RNA. TR-BBB cells were incubated with siRNAs against LRP1 for 30 h. The expression levels of LRP1 in TR-BBB cells were significantly reduced by the two LRP1-specific siRNAs to ϳ34 and ϳ29% of the levels by a control siRNA (Fig. 5A). We incubated the siRNA-treated and control cells with 125 I-A␤ and found that the internalization of A␤ by TR-BBB cells was reduced in siRNA-treated cells by ϳ38% (siRNA 1) and ϳ52% (siRNA 2) of that inhibited by RAP (Fig.  5B). Cell-bound A␤ was also reduced by siRNA treatment for LRP1 (Fig. 5C). These data strongly suggest that LRP1 is involved in the uptake of A␤ by TR-BBB cells. We also confirmed that siRNAs against LRP1 inhibited internalization of 125 I-tPA, one of the well known ligands of LRP1, in TR-BBB cells (Fig. 5D).
Lack of Rapid Uptake of A␤ by MEF Cells That Express Endogenous LRP1-To further examine the role of LRP1 in the cellular internalization of A␤, we performed A␤ uptake assays in MEFs derived from LRP1-deficient and wild-type mice (32). An anti-LRP1 antibody, R488, revealed the expression of light chain of LRP1 in MEF-1 cells derived from wild-type mice, whereas PEA10 cells derived from LRP1 heterodeficient mice exhibited ϳ50% levels of LRP1 compared with that of wild-type mice, and PEA13 cells derived from LRP1 homodeficient mice completely lacked the expression of LRP1 (Fig. 6A). Unexpectedly, however, neither the MEF-1 (wild-type), PEA10, nor PEA13 cells exhibited any significant uptake of A␤ that was observed in TR-BBB cells (Fig. 6B). Furthermore, we confirmed that the levels of cell surface binding of A␤ were at similar levels in either of the MEF cell lines (data not shown), suggesting that MEF cells fail to bind and internalize A␤ regardless of the expression levels of LRP1.
To see whether this inability of A␤ internalization was due to the low expression levels of functional LRP1 in these MEF cells, we performed a tPA internalization assay. As shown in Fig. 6C, three MEF cell lines took up 125 I-tPA at levels proportional to

. Anti-LRP1 antibody reduced A␤ internalization in TR-BBB cells.
A, the internalized 125 I-A␤ in TR-BBB cells was measured in the presence of anti-LRP1 antibody (RRR) or control IgG (each 160 g/ml). The mean Ϯ S.E. in 5-6 independent assays is shown. **, p Ͻ 0.01; ANOVA. B, the uptake assay of 125 I-A␤ was performed in the presence of 0, 5, 10, 20, 40, 80, 160, and 320 g/ml RRR in the uptake medium. The mean Ϯ S.E. in three independent assays is shown.

LRP1-mediated A␤ Clearance in TR-BBB Cells
their LRP1 expression, suggesting that LRP1 is present and functional at the surface of these MEF cells (Fig. 6C). We also conducted A␤ internalization assays using a range of cell lines (i.e. Neuro2a, Chinese hamster ovary, SH-SY5Y, McARH7777, and HUVEC) but failed to detect A␤ internalization in any of the cell lines, regardless of the levels of expression of endogenous LRP1 (supplemental Fig. 2). These data suggested that LRP1-mediated A␤ uptake depends on some specific cellular context in TR-BBB cells.

A␤ Does Not Directly Bind LRP1
in Vitro-Based on these data, we reasoned that the expression of functional LRP1 per se is not sufficient for the binding and internalization of A␤. We thus reexamined the binding of A␤ to LRP1 in an in vitro assay. LRP1 holoprotein purified from human placenta was immobilized to microtiter plates, and the binding of A␤ to LRP1 was evaluated by an overlay assay. tPA and apoE, both being authentic ligands of LRP1, exhibited robust binding to immobilized LRP1 holoprotein, which was significantly blocked by RAP. Unexpectedly, however, A␤ did not show any significant binding to LRP1 either in the presence or in the absence of RAP (Fig. 7A). We also performed similar sets of binding experiments using recombinant fragments of ligand-binding clusters of LRP1 (i.e. sLRP2 and sLRP4) (Fig. 7B). tPA bound to immobilized sLRP2 and sLRP4 in a dose-dependent manner (Fig. 7C), whereas no binding of A␤ to sLRP2 and sLRP4 was detected (Fig.  7D). Altogether, these data argue against the notion that A␤ directly interacts with LRP1.

DISCUSSION
To gain insights into the molecular and cellular mechanism of A␤ transport across the BBB, we established a novel cellular model using an immortalized brain endothelial cell line, TR-BBB cells, that recapitulates the robust and rapid uptake of A␤ and found the following: 1) TR-BBB cells rapidly internalize chaperone-free A␤ by a receptormediated mechanism; 2) the internalized A␤ is rapidly degraded and released into the medium; 3) the internalization of A␤ into TR-BBB cells is mediated by LRP1, although A␤ may not directly bind LRP1 on the cell-surface. Based on these findings, we hypothesize that BMECs that delineate BBB rapidly internalize chaperone-free A␤ peptides in the brain parenchyma and efflux them into blood by LRP1-dependent transcytosis.
Internalization of soluble or fibrillar forms of A␤ has been documented in several types of cells (13, 35-38, 40, 45, 46, 48). However, the most striking difference between the previous reports and the present observation consists in the time course of A␤ uptake; A␤ internalization took place over a long incuba-   DECEMBER 12, 2008 • VOLUME 283 • NUMBER 50 tion time of hours to days in previous experiments, whereas TR-BBB cells take up A␤ quite rapidly with the time course of Ͻ10 min (Fig. 1A). Recently, Nazer et al. (46) reported an A␤ transport model using polarized Madin-Darby canine kidney cells expressing LRP1, although the A␤ transport took place over several hours. The rapid time course in our TR-BBB cells is in good agreement with the in vivo observation that A␤ is effluxed from the brain with a half-life of ϳ30 min (15,16), underscoring the relevance of our cellular model in the study of A␤ clearance from brain. We showed that RAP or an anti-LRP1 antibody efficiently inhibited A␤ internalization by TR-BBB cells (Figs. 3A and 4A). RNAi knockdown of LRP1 also decreased A␤ uptake (Fig. 5B). In contrast, the involvement of other known A␤ receptors (i.e. RAGE and P-gp) was not detected (Fig. 3D). Taken together, it is likely that the A␤ internalization in TR-BBB cells is mediated chiefly by LRP1. However, because the inhibitory effect of siRNA against LRP1 was incomplete, we cannot exclude the possibility that other RAP-inhibitable receptors are involved in this process.

LRP1-mediated A␤ Clearance in TR-BBB Cells
We also found that A␤ complexed with lipidated form of apoE was efficiently internalized (Fig. 3A), although this uptake was not inhibited by RAP, suggesting that the uptake of A␤ complexed with lipidated apoE was mediated by an LRP1-inde-pendent mechanism. Although the precise mechanism of the uptake of A␤⅐apoE complex observed here is uncertain, it is possible that other apoE receptors (e.g. low density lipoprotein receptor, which is known as the major receptor for apoE) (49) were involved; RAP might have failed to efficiently block the uptake of A␤⅐lipidated apoE complex because of the relatively low affinity of other apoE receptors for RAP compared with that of LRP1 (50).
Despite the requirement of LRP1 in A␤ uptake by TR-BBB cells, A␤ uptake was not observed in the MEF cells, although the expression of functional LRP1 that can mediate tPA internalization was demonstrated (Fig. 6B). We also tested A␤ uptake in a range of cell lines but failed to observe the rapid LRP1-dependent A␤ internalization in any cells other than TR-BBB (supplemental Fig. 2). These data suggest that LRP1 is required, but not sufficient, to cause a rapid internalization of A␤ and that cell type specificity may significantly affect the LRP1-dependent A␤ internalization. Because we could not detect the direct binding of A␤ to LRP1 in a series of carefully controlled in vitro binding assays using full-length LRP1 or ligandbinding cluster fragments thereof (Fig. 7), we speculate that this specificity of LRP1-dependent internalization of A␤ may best be explained by the intercalation of an A␤-binding molecule (including cell surface LRP1 ligands) that is specifically expressed on the surface of TR-BBB cells and cooperates with LRP1 to internalize A␤ into cells.
LRP1 is known to function as a multifunctional receptor that recognizes at least 30 different ligands, although the mechanism regarding how LRP1 can recognize structurally unrelated ligands remains unknown (51,52). It is noteworthy that LRP1 directly binds to a few of these molecules and that many of the ligands interact with their "co-receptor" and are subsequently internalized via LRP1. For example, urokinasetype plasminogen activator and PAI-1 initially bind urokinase-type plasminogen activator receptor, and then the heterotrimeric complex is rapidly internalized by LRP1 and catabolized (53). Also, factor VIII initially adheres to heparan sulfate proteoglycans on the surface of the hepatocytes, and the complex is then scavenged by LRP1 (54). Based on these previous observations, it is plausible that an as yet unidentified A␤-binding molecule that acts as a co-receptor of LRP1 may best explain the unique feature of TR-BBB cells, which robustly internalize chaperone-free A␤ (supplemental Fig. 3). Among the known LRP1 co-receptors, heparan sulfate proteoglycans have been shown to bind A␤ (55), although treatment of TR-BBB cells with heparin or heparinase or condroitinase did not affect the A␤ uptake, ruling out the possibility that heparan sulfate proteoglycans or chondroitin sulfate proteoglycans represent the cell surface coreceptor for A␤ in TR-BBB cells (supplemental Fig. 4). Notably, HUVEC cells derived from umbilical veins failed to show A␤ uptake, and the expression level of LRP1 in HUVEC cells was extremely low (supplemental Fig. 2). This may support the crucial role of LRP1 in the rapid A␤ internalization in brain endothelial cells. However, it is still possible that some cellular environments, other than an A␤ binding co-receptor, are the major determinants for A␤ internalization in TR-BBB cells. The nature of the cell type specificity for A␤ uptake should further be addressed.
In contrast to our results, previous studies showed that recombinant LRP1 polypeptides, especially the cluster II and IV domains, directly bind A␤ in vitro by the surface plasmon resonance analysis (18,47). Although the reason for this discrepancy is not clear, we believe that our present data based on the plate binding assays that have been extensively used for the characterization of ligand-LRP1 (27) as well as ligand-A␤ (44) interactions, using purified LRP1 and appropriate controls, are quite reliable so far as in vitro interactions are concerned.
Because of the lack of polarity as well as formation of tight junctions in TR-BBB cells, we cannot fully replicate the "transport" across the BBB; in this regard, what we have analyzed here represents the "uptake" of A␤ by endocytosis in brain-derived endothelial cells. However, previous studies showed that the K m constant obtained from the uptake assay in TR-BBB cells is similar to that of an in vivo "transport" experiment (e.g. using 3-O-methyl-D-glucose as a substrate for GLUT-1) (20). Thus, the A␤ uptake in TR-BBB cells may well recapitulate the initial internalization step of the A␤ transport in vivo.
Another problem regarding the role of LRP1 in A␤ clearance is whether LRP1 mediates transcytosis or degradation of A␤, as Nazer et al. (46) studied in polarized Madin-Darby canine kidney cells. Indeed, our trichloroacetic acid precipitation assay suggested that ϳ80% of A␤ released from TR-BBB cells underwent degradation as in the Madin-Darby canine kidney cells (46). However, further study is required for the characterization of transcytosis versus degradation of A␤ in relation to the time course and cell type specificity.
In summary, we established a new in vitro model of rapid A␤ internalization using TR-BBB cells. We believe that A␤ internalization by TR-BBB may recapitulate some aspects of the initial internalization step of A␤ transport across the BBB. Stimulation of the A␤ efflux from brain to blood may be a useful strategy to inhibit A␤ accumulation and prevent the development of pathological changes in AD. TR-BBB cells may serve as a useful tool for the screening of modulation of proteins or genes participating in A␤ transport at the BBB as well as small molecule compounds that facilitate the A␤ clearance. The identification of the mechanism of the cell type specificity in A␤ internalization may be crucial to the understanding of the precise mechanism of A␤ transport across the BBB.