Apolipoprotein E and Low Density Lipoprotein Receptor-related Protein Facilitate Intraneuronal Aβ42 Accumulation in Amyloid Model Mice*

The low density lipoprotein receptor-related protein (LRP) is highly expressed in the brain and has been shown to alter the metabolism of amyloid precursor protein and amyloid-β peptide (Aβ) in vitro. Previously we developed mice that overexpress a functional LRP minireceptor (mLRP2) in their brains and crossed them to the PDAPP mouse model of Alzheimer disease. Overexpression of mLRP2 in 22-month-old PDAPP mice with amyloid plaques increased a pool of carbonate-soluble Aβ in the brain and worsened memory-related behavior. In the current study, we examined the effects of mLRP2 overexpression on 3-month-old PDAPP mice that had not yet developed amyloid plaques. We found significantly higher levels of membrane-associated Aβ42 in the hippocampus of mice that overexpressed mLRP2. Using immunohistochemical methods, we observed significant intraneuronal Aβ42 in the hippocampus and frontal cortex of PDAPP mice, which frequently co-localized with the lysosomal marker LAMP-1. Interestingly, PDAPP mice lacking apolipoprotein E (apoE) had much less intraneuronal Aβ42. We also found that PC12 cells overexpressing mLRP2 cleared Aβ42 and Aβ40 more rapidly from media than PC12 cells transfected with the vector only. Preincubation of apoE3 or apoE4 with Aβ42 increased the rate of Aβ clearance, and this effect was partially blocked by receptor-associated protein. Our results support the hypothesis that LRP binds and endocytoses Aβ42 both directly and via apoE but that endocytosed Aβ42 is not completely degraded and accumulates in intraneuronal lysosomes.

The low density lipoprotein receptor-related protein (LRP) 2 is a large endocytic receptor that is highly expressed in neuronal cell bodies and dendritic processes (1,2). LRP binds and internalizes many distinct ligands, including molecules associated with Alzheimer disease (AD) such as apolipoprotein E (apoE) and ␣2-macroglobulin (3). LRP has also been shown to interact with amyloid precursor protein (APP) (4 -8), and this interaction appears to favor the processing of APP to generate the amyloid ␤-peptide or A␤ (9,10). In turn, A␤ has been shown to directly bind LRP (11) and to form stable complexes with apoE and ␣2-macroglobulin (12,13). These observations suggest that changes in LRP expression levels may affect risk for AD by altering both the processing of APP and the clearance of A␤.
To study the role of LRP in the brain, our laboratory previously cloned a section of the extremely large human LRP receptor (mLRP2) into the MoPrP.Xho transgenic mouse vector and developed transgenic mice that overexpress mLRP2 (14). The mLRP2 section or minireceptor of LRP behaves similarly in vitro to full-length LRP with respect to ligand binding and internalization (14). Mice overexpressing mLRP2 were then bred to the well characterized PDAPP mouse model of AD, which expresses human APP with a mutation that causes familial AD (V717F) and develops amyloid plaques in the brain beginning at ϳ6 months of age (15). We found that 22-month-old PDAPP/ mLRP2 and PDAPP/wild-type mice had similar levels of total amyloid deposition in their brains but that PDAPP/mLRP2 mice had increased levels of A␤ in carbonate-soluble brain extracts (14). Furthermore, this carbonate-soluble A␤ pool was highly correlated with memory deficits in old mice.
In the current study, we examined whether overexpression of mLRP2 also affects A␤ levels in 3-month-old mice that have not yet developed amyloid plaques. We hypothesized that changes in A␤ at this age are more likely to reflect subtle changes in A␤ metabolism resulting from mLRP2 overexpression without the confounding factor of severe pathology as in the older mice. Based on our previous work (14), we expected that overexpression of LRP would increase membrane-associated A␤. Additionally, we investigated whether membrane-associated A␤ could represent intracellular A␤.

EXPERIMENTAL PROCEDURES
Animals-The mLRP2 transgene was expressed by the MoPrP.Xho vector as previously described (14). mLRP2 mice were crossed to PDAPP ϩ/Ϫ mice (15) to produce PDAPP/ mLRP2 and PDAPP/wild-type mice. Apoe Ϫ/Ϫ mice obtained from the Jackson Laboratory (Bar Harbor, ME) were crossed with PDAPP mice to produce PDAPP/Apoe ϩ/ϩ and PDAPP/ Apoe Ϫ/Ϫ mice. All the mice used for this study were on a C57Bl/6 genetic background. Mice were screened for the presence of the mLRP2 and PDAPP transgenes by PCR. Tissues were obtained following transcardial perfusion with ice-cold PBS (14). The right hemisphere was immersion fixed for 24 h in 4% paraformaldehyde and then cryoprotected in 30% sucrose in PBS for histological analysis. The left hemisphere was further dissected, and brain regions were frozen for biochemical analysis. Cerebrospinal fluid (CSF) was isolated from the cisterna magna as previously described (16).
Cells-PC12 cell were transfected with the pcDNA vector only or the pcDNA vector encoding mLRP2 cDNA tagged with the hemagglutinin (HA) epitope. Stably expressing clones were selected based on cell morphology and mLRP2 expression levels, which were evaluated by Western blotting and immunofluorescent staining with anti-HA antibody. High expressing lines were maintained in RPMI 1640 containing 10% horse serum, 5% fetal calf serum, 1% glutamine, 1% penicillin, 1% streptomycin, and 400 g/ml G418. Sequential Brain Extraction-Hippocampi were homogenized with 25 strokes in 15 volumes of cold Tris-buffered saline (TBS), pH 7.4, containing the protease inhibitors leupeptin (10 g/ml) and aprotinin (20 g/ml) using a 1-ml dounce homogenizer. Homogenates were transferred to microcentrifuge tubes and spun at 20,000 ϫ g for 20 min at 4°C. Supernatants (TBS extracts) were transferred to new tubes and kept on ice. After a single wash with 50 l of cold TBS, pellets were resuspended in 15 volumes of cold TBS, pH 7.4, containing 1% Triton X-100 and protease inhibitors and incubated for 30 min at 4°C with agitation. The homogenates were spun again at 20,000 ϫ g for 20 min at 4°C. Supernatants (TBS-Triton X-100 extracts) were transferred to new tubes and kept on ice. Pellets were spun once for 5 min, and all leftover supernatant was removed. 400 l of 5 M guanidine solution containing protease inhibitors were added to the final pellet. Samples were vortexed to detach the pellet from the bottom of the tubes and incubated for 4 h at room temperature. The homogenates were spun again at 20,000 ϫ g for 20 min at 4°C. The supernatants (guanidine extracts) were transferred to new tubes and kept on ice.
A␤40 and A␤42 Determinations-Human A␤40 and A␤42 levels were determined in all hippocampal extracts (TBS, TBS-Triton X-100, and guanidine) and CSF by sandwich ELISA under denaturing conditions (0.5 M guanidine). The capturing antibody was 2G3 for A␤40 and 21F12 for A␤42, and the detection antibody was biotinylated 3D6 as previously described (14). Samples were diluted 10-fold in PBS containing 0.25% bovine serum albumin, and human A␤40 and A␤42 standards were also solubilized in PBS containing 0.25% bovine serum albumin as well as 10% of the buffer used to obtain each sample (TBS, TBS-Triton X-100, or 5 M guanidine).
Mouse ApoE Determination-A previously described sandwich ELISA for mouse apoE with a sensitivity of ϳ1 ng/ml was used (17). Brain samples were sonicated in PBS containing 0.05% Tween 20 and protease inhibitors. Samples were spun to pellet cell debris (20,000 ϫ g for 25 min at 4°C), and supernatants were diluted in PBS containing 0.025% Tween 20 and 0.5% bovine serum albumin. Standards were based on pooled plasma from C57/Bl6 mice containing 68 g/ml apoE, and brain samples from apoE knock-out mice were used for background subtraction.
Western Blot Analysis-Hippocampi were dounce homogenized in 15 volumes of TBS containing 1% Triton X-100 with the protease inhibitors leupeptin (10 g/ml) and aprotinin (20 g/ml). Equal amounts of protein from the tissue lysates were separated using 10% Tris-Tricine gels and transferred to polyvinylidene difluoride membranes. Membranes were blotted with anti-APP antibody (Zymed Laboratories, San Francisco, CA), and immunoreactive bands were detected using ECL (Amersham Biosciences).
Immunostaining-Right hemispheres were cut in the coronal plane on a freezing sliding microtome. Floating brain sections (50 m) were blocked with 10% normal goat serum (Vector Laboratories, Burlingame, CA) in TBS containing 0.25% Triton X-100 for 45 min. The sections were then incubated overnight at 4°C with a polyclonal antibody to A␤42 (Chemicon International, Temecula, CA) and a monoclonal antibody to neuronspecific nuclear protein (NeuN) (Chemicon International), both at 1:500 dilution in TBS containing 0.25% Triton X-100 and 2% normal goat serum. Secondary antibodies AlexaFluor 488 goat anti-rabbit IgG and AlexaFluor 568 goat anti-mouse IgG (Molecular Probes, Eugene, OR) were diluted 1:1000 for detection. Fluorescence was visualized by confocal laser scanning microscopy. A mouse-specific rat antibody to LAMP-1 (1D4B) was used as hybridoma tissue culture supernatant (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA) at 1:4 dilution with secondary antibody Alexa-Fluor 568 goat anti-rat IgG at 1:1000 dilution (Molecular Probes) for detection. Quantitation of A␤42 immunoreactivity associated with neurons from PDAPP/Apoe ϩ/ϩ and PDAPP/ Apoe Ϫ/Ϫ mice was performed using the ImageJ software from the National Institutes of Health (Bethesda, MD). Briefly, each of the images obtained under identical settings by confocal microscopy using ϫ40 magnification was split into RGB channels (red, green, and blue) and a total of 25 neurons/section were randomly circled in the red channel (NeuN). The intensity of fluorescence signal associated with each neuron was then calculated by redirecting measurements to the green channel (A␤42).
In Vitro A␤ Clearance-In vitro A␤ clearance experiments were performed using PC12 cells stably transfected with mLRP2 or the pcDNA vector. PC12 cells were seeded in 24-well plates at a density of ϳ300,000 cells/well and allowed to attach overnight. Serum-containing medium was then replaced with neurobasal medium containing N2 supplement and nerve growth factor (100 ng/ml) to stimulate differentiation overnight. Cell-secreted A␤ was obtained from PDAPP primary neuronal cultures that contained ϳ60% A␤42. This high A␤42: A␤40 ratio is typical of A␤ produced by PDAPP mice (18,19). The cell-secreted A␤ was diluted 10-fold in neurobasal medium plus B27 and nerve growth factor and was added to the differentiated PC12 cells (final A␤ concentration ϳ500 pg/ml). In some conditions, diluted A␤-containing medium was preincu-bated for 2 h at 37°C with apoE particles (10 g/ml) derived from human embryonic kidney cells stably transfected with human apoE3 or apoE4 (12,20), and then the apoE-and A␤-containing medium was added to the cells. In other conditions, cells were pretreated for 15 min with 0.5 M receptorassociated protein (RAP), a LRP antagonist (21), prior to the addition of A␤-conditioned medium preincubated with apoE3 particles. After 24 h, A␤ levels in the incubation medium were quantified by ELISA and compared with the values of the original A␤ medium, kept at 37°C. The difference in A␤ levels was then divided by cell protein content in each well (n ϭ 3).

RESULTS
We have previously shown that overexpression of mLRP2 significantly increases a carbonate-soluble A␤ pool in the hippocampus and cortex of 22-month-old PDAPP mice with abundant plaque deposition (14). In the current study, we examined whether overexpression of mLRP2 affects A␤ levels in 3-month-old PDAPP mice that have not yet developed amyloid plaques in the brain.
To examine the different pools of A␤, we performed sequential extraction of the brain tissue. Previously, we examined carbonate (soluble) and guanidine (insoluble) extracts from brain. For these experiments, we added an extra step into our protocol so that we could separate the soluble extract into a water-soluble extract with few membrane-associated proteins and a detergent-soluble extract enriched for membraneassociated proteins. Instead of carbonate, we used TBS buffer containing 1% Triton X-100 because this buffer is more widely used to extract membrane proteins and yields results similar to those obtained with carbonate. APP, as well as other membrane proteins such as transferrin, PS1, and calnexin, were extracted by both carbonate and TBS with 1% Triton X-100 ( Fig. 1A and data not shown). The hippocampus was first homogenized in TBS and centrifuged, which provided a supernatant with water-soluble proteins that contained almost none of the intramembrane protein APP (Fig.  1A). Next, the pellet from the TBS extraction was rehomogenized in TBS with 1% Triton X-100 and centrifuged, resulting in a supernatant containing detergent-soluble, mainly membrane-associated proteins such as APP (Fig. 1A). Finally, the pellet was homogenized again in 5 M guanidine and centrifuged, allowing extraction of relatively insoluble proteins. The TBS, TBS-Triton X-100, and guanidine extracts from the brain were then assayed for A␤40 and A␤42 by ELISA.
We found that overexpressing mLRP2 in PDAPP mice does not significantly affect levels of A␤40 in any of the three different extracts from the hippocampus (Fig. 1, B-D). TBS-soluble A␤42 was also not different in PDAPP/mLRP2 and PDAPP/wild-type mice (Fig. 1B). However, detergent-and guanidine-soluble A␤42 was significantly higher in PDAPP mice overexpressing mLRP2 compared with non-transgenic littermates. Specifically, detergent-soluble A␤42 levels were increased by ϳ20% in PDAPP/ mLRP2 mice, and guanidine-soluble A␤42 levels were increased by ϳ12% in PDAPP/mLRP2 mice compared with PDAPP/wildtype control mice (Fig. 1, C and D). The selective increase of detergent-and guanidine-soluble A␤42 suggests that mLRP2 affects A␤42 that is associated with membranes and not freely soluble A␤42. Although these effects are small, they demonstrate the involvement of LRP in regulating brain A␤ levels. A, levels of APP in brain extracts made with carbonate and with TBS containing 1% Triton X were similar and were not affected by overexpression of mLRP2. Equal amounts of total protein from hippocampal extracts made with carbonate and TBS with 1% Triton X-100 from young 3-month-old PDAPP/wild-type (n ϭ 2) and PDAPP/mLRP2 mice (n ϭ 2) were separated by SDS-PAGE electrophoresis followed by Western blotting with an antibody to APP. B-D, the hippocampus underwent a three-step serial extraction with TBS, TBS with 1% Triton X-100, and 5 M guanidine. A␤40 and A␤42 levels were determined by ELISA. For both the PDAPP/mLRP2 and PDAPP/wild-type groups, n ϭ 12 mice. E, CSF was obtained from the cisterna magna and assayed for A␤40 and A␤42 by ELISA. n ϭ 5 PDAPP/mLRP2 and PDAPP/wild-type mice. Results are shown as mean Ϯ S.E.; statistical analysis was by Student's two-tailed t test.
Western blotting of APP showed no differences between PDAPP/mLRP2 and PDAPP/wild-type mice (Fig. 1A). We also examined the levels of CTF-␤, a fragment of APP produced by ␤-secretase cleavage of APP that is cleaved by ␥-secretase to form A␤. CTF-␤ levels were not different in PDAPP/mLRP2 and PDAPP/wild-type mice by Western blotting (data not shown). These data suggest that APP expression and APP processing is not drastically altered by mLRP2 overexpression, although it does not rule out the possibility that subtle changes in APP expression or APP processing are occurring that affect A␤ levels.
Interestingly, although A␤42 levels in the brain significantly increased in the PDAPP/mLRP2 overexpressing mice, total A␤ was significantly decreased in the CSF of mLRP2 versus wildtype mice, with a trend toward decreased A␤40 and A␤42 (Fig.  1E). The increased amount of A␤42 in the brain tissue of the PDAPP/mLRP2 mice, coupled with the decreased amount of A␤ in the CSF of the PDAPP/mLRP2 mice, suggests that mLRP2 overexpression increases the internalization of extracellular A␤. These findings are consistent with the hypothesis that mLRP2 binds and internalizes A␤, either directly or via endocytosis of apoE-A␤ complexes, and that mLRP2 overexpression increases the internalization of A␤42.
To further examine the mechanism by which A␤42 levels are altered in young PDAPP mice, we visualized and quantified intracellular A␤ using immunohistochemical methods. We observed significant A␤42 associated with neurons in the CA1 region of the hippocampus and frontal cortex (Fig.  2, A and B). Double labeling of A␤42 and LAMP1 revealed that much of the A␤42 is localized to the lysosomes within neurons (Fig. 2C). Double labeling of A␤42 and the transferrin receptor (an endosomal marker) or BiP (an endoplasmic reticulum marker) did not show any significant co-localization (data not shown). Importantly, the anti-A␤42 antibody showed no staining when preincubated with an excess of A␤42 peptide (Fig. 2D), supporting the conclusion that we were staining A␤42. Additionally, the anti-A␤42 antibody did not stain brain sections from wild-type mice that do not express human A␤ (data not shown). Furthermore, previous studies have shown that neither APP nor CTF-␤ are significantly changed in Apoe Ϫ/Ϫ mice when compared with wild-type mice, further supporting our conclusion that we were staining A␤42 and not APP or CTF-␤ (19).
Because overexpression of mLRP2 is likely to affect levels of ligands such as apoE, and apoE levels are known to affect A␤ metabolism, we determined the apoE levels in mLRP2 and wildtype mice. We found that apoE levels were 25% lower in mice overexpressing mLRP2 (Fig. 3A). This is consistent with the known role of LRP as an apoE receptor that facilitates internalization and degradation of apoE. Furthermore, we examined whether the presence of the intraneuronal A␤42 we previously observed (Fig. 2) depended on apoE. We found that PDAPP/ Apoe Ϫ/Ϫ mice had ϳ50% less intraneuronal A␤42 (Fig. 3, B and FIGURE 2. Significant A␤42 immunoreactivity was associated with neurons in 3-month-old PDAPP mice prior to plaque deposition. Brain sections (50 m) were stained with a polyclonal anti-A␤42 antibody and monoclonal anti-neuron-specific nuclear protein (NeuN) antibody, followed by secondary detection with goat anti-rabbit AlexaFluor 488 and goat anti-mouse AlexaFluor 568, respectively. Immunofluorescence was detected by confocal microscopy. A, A␤42 immunoreactivity (green) was associated with neurons (red) from the CA1 hippocampal region. B, A␤42immunoreactive neurons were also evident within the frontal cortex region. C, neuronal-associated immunostaining for A␤42 (green) was largely co-localized (yellow) with the lysosomal marker LAMP-1 (red). D, preincubation of anti-A␤42 antibody with 200-fold excess molar concentration of A␤42 peptide completely blocked neuronal-associated A␤42 immunoreactivity. FIGURE 3. Neuron-associated A␤42 immunoreactivity was significantly decreased in PDAPP mice lacking apoE. A, mice overexpressing mLRP2 had decreased levels of apoE in the cortex as measured by ELISA. n ϭ 5 mLRP2 and wild-type mice. C, brain sections (50 m) from 6-monthold PDAPP/Apoe Ϫ/Ϫ and PDAPP/Apoe ϩ/ϩ mice were stained with a polyclonal anti-A␤42 antibody and a monoclonal anti-neuron-specific nuclear protein (NeuN) antibody followed by secondary detection with goat antirabbit AlexaFluor 488 and goat anti-mouse AlexaFluor 568, respectively. Immunofluorescence was detected by confocal microscopy at ϫ40 magnification. A␤42 immunoreactivity (green) associated with neurons (red) in the frontal cortex region was markedly reduced in PDAPP mice lacking apoE. B, quantitation of A␤42-associated fluorescent intensity in the frontal cortical region showed a 54% significant reduction in neuronal-associated A␤42 immunoreactivity in PDAPP/Apoe Ϫ/Ϫ (n ϭ 4) compared with PDAPP/Apoe ϩ/ϩ mice (n ϭ 5). Results are shown as mean Ϯ S.E.; statistical analysis was by Student's two-tailed t test. NOVEMBER 24, 2006 • VOLUME 281 • NUMBER 47 C). This demonstrates that apoE plays a major role in the uptake of A␤42 by neurons, possibly by forming apoE-A␤ complexes that bind to LRP and other receptors and are endocytosed.

Effects of LRP and ApoE on Intraneuronal A␤
To directly address the question of whether mLRP2 overexpression affects intraneuronal A␤42 levels, we examined whether PDAPP/mLRP2 and PDAPP/wild-type mice had different levels of intraneuronal A␤42 staining. We did not see any significant differences in staining (data not shown). However, the sensitivity of this technique is relatively low and we were expecting only a small difference, so we performed another experiment to answer this question. We stably transfected PC12 cells with mLRP2 or the vector only and induced differentiation of the PC12 cells with nerve growth factor. We then added A␤ secreted by primary neurons derived from PDAPP mice. PC12 cells overexpressing mLRP2 cleared significantly more A␤42 than cells expressing the vector only (Fig. 4A). Similar results were obtained when preparations of oligomeric A␤42 were added to the cells (data not shown). Clearance of A␤42 by mLRP2-overexpressing cells was increased by preincubation of A␤ with apoE3 or apoE4 particles (Fig. 4A). It is not surprising that A␤42 preincubated with apoE3 or apoE4 had a similar rate of clearance because apoE3 and apoE4 have similar binding properties to LRP (22,23). The addition of RAP partially blocked clearance of A␤42 preincubated with apoE3 (Fig.  4A). We also examined the clearance of A␤40 and found that it was much slower than A␤42 clearance (Fig. 4B). A␤40 clearance was increased by the overexpression of mLRP2, but to a lesser extent than A␤42 (54% increase in clearance for A␤40, 80% increase for A␤42). Additionally, clearance of A␤40 was not affected by preincubation with apoE3 particles (Fig. 4B). This suggests that both A␤42 and A␤40 are cleared by an LRPmediated process but that apoE plays a larger role in clearance of A␤42. Overall, these data support two conclusions. 1) mLRP2 overexpression increases the internalization of A␤42, even in the absence of apoE. This suggests that A␤42 binds directly to mLRP2 and undergoes endocytosis. 2) ApoE enhances the clearance of A␤42 via mLRP2. We hypothesize that A␤ binds to apoE and forms an apoE-A␤ complex, which then binds to mLRP2 and undergoes endocytosis. However, while the endocytosed apoE is rapidly degraded, a portion of the A␤42 accumulates and possibly aggregates inside lysosomes.

DISCUSSION
Our current study demonstrates that overexpression of mLRP2 increases detergent-and guanidine-soluble A␤42 in the hippocampus of 3-month-old PDAPP mice that have not yet developed amyloid plaques. We also found significantly decreased total A␤ in the CSF of PDAPP/mLRP2 versus PDAPP/wild-type mice, with a trend toward decreases in both A␤40 and A␤42. The increase in cellular A␤42 and decrease in extracellular A␤ suggest that A␤42 may be endocytosed more rapidly by mLRP2 overexpressing cells and may accumulate intracellularly. To test this hypothesis, we developed a protocol for staining A␤42 associated with neurons. Using this protocol, we found significant amounts of intraneuronal A␤42 in 3-month-old PDAPP mice, most of which was localized inside lysosomes. Interestingly, we found that the levels of intraneuronal A␤42 were dramatically lower in PDAPP mice lacking apoE. Because mLRP2 mice have slightly higher levels of membrane-associated A␤42, we expected PDAPP/mLRP2 mice to have higher levels of intraneuronal A␤42 than PDAPP/wild-type mice. However, we were not able to detect a significant difference in intraneuronal A␤42 between PDAPP/mLRP2 and PDAPP/wild-type animals, probably because the protocol was not sensitive enough to detect a small effect. To study whether mLRP2 affects A␤42 clearance in another system, we made PC12 cell lines that were stably transfected with mLRP2 or vector only. We found that the mLRP2-overexpressing cells cleared significantly more A␤42 from the medium than cells expressing the vector only. In addition, preincubation of the A␤-containing medium with apoE3 or apoE4 particles increased the clearance of A␤42 by the cells, and this effect was partially blocked by addition of RAP.
Our initial hypothesis based on previous studies was that A␤ in detergent-soluble extracts from the hippocampus of 3-month-old PDAPP/mLRP2 mice would be higher than in PDAPP/wild-type mice. This was essentially correct, but the increase in A␤ was only in the more amyloidogenic A␤42 species and only in the detergent-and guanidine-soluble extracts. We speculate that endocytosed A␤42 may not be degraded as effectively as A␤40 in lysosomes. The finding that only deter-

. Overexpression of mLRP2 in PC12 cells increased the clearance of A␤-apoE complexes in vitro.
A, clearance of naturally secreted A␤42 (ϳ300 pg/ml) by PC12 cells stably transfected with mLRP2 was significantly greater than by PC12 cells transfected with the pcDNA vector only. A␤42 clearance was enhanced by preincubation of the A␤-containing medium with apoE3 or apoE4 particles (10 g/ml). Preincubation of cells with LRP antagonist RAP (0.5 M) significantly reduced A␤42 clearance mediated by apoE3 particles. B, clearance of naturally secreted A␤40 (ϳ200 pg/ml) by PC12 cells was slower than clearance of A␤42 and was moderately increased by overexpression of mLRP2. Preincubation of A␤-containing medium with apoE3 did not affect clearance of A␤40. For each group, n ϭ 3. Results are shown as mean Ϯ S.E.; statistical analysis was by one-way analysis of variance with Newman-Keuls multiple comparison post-hoc testing.
gent-and guanidine-soluble A␤42 was increased in the PDAPP/mLRP2 mice suggests that mLRP2 is specifically affecting a pool of A␤42 that is membrane bound or insoluble. We did not specifically test whether the A␤42 that required guanidine for extraction was in an insoluble, oligomeric state or bound to membrane domains that are poorly soluble in 1% Triton X-100. However, other investigators have found that intraneuronal A␤42 deposits are reactive to antibodies specific for A␤42 oligomers (24).
The increase of A␤42 in brain tissue and decrease in A␤ in CSF suggested that the ratio of intracellular to extracellular A␤ was increasing, perhaps as a result of intracellular A␤42 accumulation. We investigated this possibility using immunohistochemistry to visualize intracellular A␤42. The accumulation of A␤ inside neurons has been observed in both AD patients and mice with brain amyloid deposition (25)(26)(27)(28)(29). Intraneuronal A␤ accumulation appears to occur prior to extracellular amyloid deposition and is a prominent neuropathological feature in brain regions that are vulnerable in AD, such as the frontal cortex and the hippocampus. In a triple transgenic AD mouse model with the PS1 (M146V), APP (Swe), and tau (P301L) transgenes, deficits in synaptic transmission and long term potentiation were observed before plaque and tau pathology, but in the presence of early intraneuronal A␤ accumulation (29). It has been suggested that intraneuronal A␤ pathology causes the onset of early cognitive deficits in this triple transgenic AD mouse model (30). In other experiments, it has been shown that intracellular A␤ accumulation directly affects cell viability, ultimately resulting in neuronal death (31)(32)(33)(34). These observations raise the hypothesis that intraneuronal A␤ accumulation may be one of the initial steps in a cascade of events leading to AD (35).
We found that intraneuronal A␤42 co-localized with the lysosomal marker LAMP-1. Previous studies have shown that intraneuronal A␤ is present within endosomes, multivesicular bodies, and lysosomes (24, 28, 36 -38). If A␤42 is concentrated within small organelles such as lysosomes, this could favor the aggregation of A␤42. Additionally, the acidic pH characteristic of endocytic compartments may provide a favorable environment for A␤ oligomerization (39,40). Interestingly, early abnormalities in the endosomal-lysosomal system that precede amyloid deposition have been reported in sporadic AD and Down syndrome (36,41). Whether the A␤42 that accumulates within the endocytic compartments of neurons results from extracellular uptake of A␤42 or intracellular generation of A␤42 is not yet clear. However, chronic infusion of synthetic A␤42 oligomers with transforming growth factor ␤-2 into the ventricles of wild-type mice results in intraneuronal A␤42 accumulation, suggesting that extracellular uptake of A␤42 may result in intraneuronal A␤42 accumulation. RAP markedly reduces this accumulation, which indicates that LRP plays a role in this process (42).
The finding that PDAPP mice lacking apoE have markedly reduced levels of intraneuronal A␤42 demonstrates that apoE plays an important role in the neuronal uptake and/or accumulation of A␤42. Although it has previously been shown that PDAPP/Apoe Ϫ/Ϫ mice have reductions in both total A␤ and fibrillar A␤, these studies were performed with older mice with prominent A␤ deposition (43)(44)(45). In contrast, our studies were performed in 3-month-old mice prior to plaque deposition and therefore suggest that A␤42 begins to accumulate inside neurons even before plaques are deposited in tissues. Additionally, our in vitro results demonstrate that apoE can increase the uptake of A␤42 via binding to LRP and perhaps other members of the low density lipoprotein receptor family. This supports the hypothesis that extracellular uptake of apoE-A␤ complexes can lead to intraneuronal A␤42 accumulation. However, although increased endocytosis of apoE-A␤ complexes likely leads to lower brain apoE levels, such as we found in mLRP2 mice, A␤42 is not degraded completely following endocytosis and accumulates.
Although our current results support a role of LRP in intraneuronal A␤ accumulation, our studies were largely performed with a LRP minireceptor, mLRP2. The extremely large size of LRP (ϳ600 kDa) makes it technically challenging to overexpress the full-length receptor either in transgenic animals or stable cell lines. Because our previous studies have confirmed the functionality of mLRP2 (46), we believe that mLRP2 is an appropriate tool for studying ligand binding to full-length LRP. Future studies should be directed to studying the function of endogenous LRP in intraneuronal A␤ accumulation by conditional knock out or virally mediated gene knockdown approaches.
In summary, the hypothesis that A␤ clearance by neurons is a protective mechanism against AD needs to be reconsidered. Provided that all internalized A␤ is efficiently degraded within lysosomes, this pathway may indeed be protective. However, our present and previous published studies suggest that this may not be the case and that intraneuronal A␤ accumulation may be pathogenic. Furthermore, reducing the endocytosis of A␤ by neurons may have a therapeutic effect on AD.