Internalized Antibodies to the Aβ Domain of APP Reduce Neuronal Aβ and Protect against Synaptic Alterations*♦

Immunotherapy against β-amyloid peptide (Aβ) is a leading therapeutic direction for Alzheimer disease (AD). Experimental studies in transgenic mouse models of AD have demonstrated that Aβ immunization reduces Aβ plaque pathology and improves cognitive function. However, the biological mechanisms by which Aβ antibodies reduce amyloid accumulation in the brain remain unclear. We provide evidence that treatment of AD mutant neuroblastoma cells or primary neurons with Aβ antibodies decreases levels of intracellular Aβ. Antibody-mediated reduction in cellular Aβ appears to require that the antibody binds to the extracellular Aβ domain of the amyloid precursor protein (APP) and be internalized. In addition, treatment with Aβ antibodies protects against synaptic alterations that occur in APP mutant neurons.

Active immunization for ␤-amyloid peptide (A␤) 2 has been demonstrated to reduce A␤ plaques and improve cognitive function in transgenic mouse models of Alzheimer disease (AD) (1)(2)(3)(4). In human AD patients actively immunized with A␤, subjects with high antibody titers appeared to have slowed cognitive deterioration and reduced plaque burden, although 6% of subjects developed meningoencephalitis (5)(6)(7). Passive immunotherapy in mouse models of AD has provided similar benefits to those seen with active immunization (4). The mechanisms by which A␤ antibodies reduce A␤ plaque pathology in the brain remain unclear (8). Data suggest roles for antibodymediated microglial activation and A␤ efflux from the brain in the reduction of A␤ (9,10). Interestingly, intracerebral injection of A␤ antibody reduced levels of both extracellular and intracellular A␤ in a triple transgenic (3ϫTg) mouse carrying mutations in amyloid precursor protein (APP), presenilin 1 and tau (11), and reduction of intraneuronal A␤ was the better correlate with cognitive improvement (12). How A␤ antibodies reduce intracellular A␤ is not known. Increasing evidence supports that intraneuronal A␤ accumulation is important in the pathogenesis of AD. Intraneuronal A␤ accumulation has been reported in transgenic mouse models of cerebral amyloidosis (13)(14)(15)(16)(17)(18)(19), human AD (20 -22) and Down syndrome (21,(23)(24)(25). Moreover, cultured neurons from APP mutant transgenic mice develop subcellular A␤ accumulation and synaptic alterations that parallel those observed in vivo in the brain with ␤-amyloidosis (26,27).
We now report that A␤ antibodies decrease levels of intracellular A␤ in culture and provide evidence that antibody binding to the A␤ domain of APP and internalization of the antibody/APP complex appear to be required to reduce intracellular A␤. Moreover, A␤ antibody treatment protects against synaptic alterations that occur in APP mutant neurons in culture.
Antibody Treatment-Several well characterized A␤ antibodies were used: monoclonal 6E10 (human A␤ residues 5-10; IgG 1 ) and 4G8 (A␤ residues 17-24; IgG2 B ) (Signet Laboratories), G2-11 (A␤42 C terminus; IgG 1 ; Genetics Co.), anti-A␤42 (A␤42 C terminus; Chemicon), polyclonal anti-A␤40 (A␤40 C terminus; Chemicon), and polyclonal APP-ab2 (human A␤ residues 1-10; Labvision NeoMarkers). Other antibodies used were: monoclonal P2-1 (specific for the N terminus of human APP; IgG 1 ; Affinity BioReagents) and mouse anti-human IgG (Jackson ImmunoResearch). We added fresh medium to cells just prior to each treatment. Final antibody concentration in all treatments was 2 g/ml. Antibodies were added to the culture for different time points, as indicated. Treatment for 24 h with A␤ antibody 6E10 did not induce neuronal death, evaluated using the TUNEL labeling method and lactate dehydrogenase (LDH) assay (see below). For internalization studies, neurons were kept at 4°C for 45 min with the indicated antibody. Cells were then either immediately fixed (0 min) or incubated at 37°C for 10, 30, 60, or 180 min and then fixed in 4% paraformaldehyde in phosphate buffer (PB).
Cell Death Assay-19 DIV Tg2576 neurons grown on poly-D-lysine precoated coverslips were incubated 24 h with or without 6E10 antibody. Neurons were washed twice with PB saline (PBS) and fixed in 4% paraformaldehyde. The TUNEL staining kit (Roche Applied Science) was used to stain apoptotic neurons according to the manufacturer's instructions. Nuclei were stained with the Hoechst stain. Counts of TUNEL positive nuclei and total nuclei were performed with Metamorph (Universal Imaging Co.) on 6 -10 fields per coverslip at 20ϫ magnification. The ratio between TUNEL-positive nuclei and Hoechst-positive nuclei was calculated. LDH kit (Sigma) was used to evaluate whether antibody treatment was inducing cell damage after incubation for 24 h with or without A␤ antibodies. Levels of LDH were measured in the media according to the manufacturer's instructions.
A␤ Immunoprecipitation and Detection-Primary neurons and Sw-N2a cells were washed twice, harvested in ice-cold PBS, and centrifuged. Cell lysates were treated with 6% SDS containing 10 l/ml of ␤-mercaptoethanol, sonicated, and then heated at 95°C for 6 min. After centrifugation, supernatants were either loaded directly (neuron lysates) into 10 -20% Tricine gels (Invitrogen) for A␤ detection or immunoprecipitated (N2a cells) overnight at 4°C with 4G8 antibody (in 190 mM NaCl, 50 mM Tris-HCl pH 8.3, 6 mM EDTA and 2.5% Triton X-100). The latter were then incubated with rabbit anti-mouse secondary antibody (Cappell) together with protein A-Sepharose beads (GE Healthcare) for 2 h at 4°C. Samples were subjected to electrophoresis and transferred to polyvinylidine difluoride membranes (Millipore). Membranes were boiled in PBS for 5 min and immunoblotted as described (26). The immunoreaction was visualized by a chemiluminescence system (Pierce). Band intensities were quantified using Scion Image software. The area under the band peak and above the baseline was quantified.
To determine secreted APP␤ levels, media were centrifuged 5 min at 1,000 ϫ g to pellet cellular debris. 1 ml of supernatant was collected and incubated with an antibody against secreted APP␤ (Signet) overnight at 4°C (in 190 mM NaCl, 50 mM Tris-HCl, pH 8.3, 6 mM EDTA, and 2.5% Triton X-100). Samples were incubated with rabbit anti-mouse secondary antibody together with protein A-Sepharose beads for 2 h at 4°C. The same media were further used to immunoprecipitate secreted APP␣ fragments with antibody 6E10 in the same conditions described for secreted APP␤. Western blot analyses were performed as described above using 22C11 (Roche Applied Science) as primary antibody.
ELISA Analyses-19 DIV Tg2576 primary neurons or Sw-N2a cells were incubated for 24 h in the presence or absence of antibody 6E10 or 4G8 and collected as described previously. Concentrations of A␤1-40 and A␤1-42 were measured by using the respective ELISA kits (BIOSOURCE) according to manufacturer's instructions.
Biochemical Measurements of Surface APP-19 DIV primary neurons or Sw-N2a cells were incubated for 3 h with antibody 6E10. After two washes with PBS containing 1 mM CaCl 2 and 0.5 mM MgCl 2 (PBS-Ca-Mg), cells were placed on ice to block endocytosis and incubated with PBS-Ca-Mg containing 1 mg/ml Sulfo-NHS-LC-Biotin (Pierce) for 20 min. Cultures were rinsed in ice-cold culture medium to quench the biotin reaction. Cultures were lysed in 200 l of 3% SDS. The homogenates were centrifuged at 14,000 ϫ g for 15 min at 4°C. Fifteen microliters of the supernatant were removed to measure total protein levels; the remaining supernatant was incubated with 100 l of Neutravidin agarose (Pierce) overnight at 4°C. Samples were then washed three times with a buffer containing: 150 mM NaCl, 10 mM Tris-HCl, pH 8.3, 5 mM EDTA, 0.1% Triton X-100, 0.01% bovine serum albumin, and protease inhibitor mixture (Roche Applied Science). Bound proteins were resuspended in 30 l of SDS sample buffer and boiled. Quantitative Western blots were performed on both total and biotinylated (surface) proteins using APP N-terminal antibody 22C11 and tubulin antibody (Sigma). Immunoreactive bands were visualized by enhanced chemiluminescence (ECL, Amersham Biosciences) and captured on autoradiography film (Amersham Biosciences Hyperfilm ECL). Digital images, produced by densitometric scans of autoradiographs on a Scan-Maker 8700 (Microtek) were quantified using NIH Image 1.63 software. The surface/total APP ratio was calculated for each culture.
Degradation Assay for APP and C99-Sw-N2a or C99-N2a cells were treated with biotin as described above. After cells were rinsed in ice-cold culture medium to quench the biotin reaction, fresh ice-cold medium containing 6E10 antibody (2 g/ml) was added. Cells were kept on ice for 5 min to allow the antibody recognition of surface APP or C99. Cells were then placed at 37°C for 45 min. After treatment, cells were collected and lysed as described for the assay of surface APP, above. Western blot analyses of full-length APP were performed using 22C11 antibody; Western blot analyses of C99 peptides were performed using 6E10 antibody.
Immunofluorescence-Cells were grown on poly-D-lysinecoated coverslips (Fisher). After antibody treatment, cells were washed in ice-cold PBS and fixed for immunofluorescence, as described previously (27). The following antibodies were used: A␤42, synapsin I, and PSD-95 (Chemicon), APP 369 (anti-C terminus of APP, (30), EEA1 (BD Transduction), Tsg101 (Genetex), and Lamp2 (Zymed Laboratories Inc.). Fluores-cent secondary antibodies were either Alexa-488 or Alexa-546 (Molecular Probes) or Cy2-or Cy3-conjugated (Jackson ImmunoResearch). To determine antibody uptake, cells were incubated with secondary antibodies with or without prior permeabilization with 0.1% saponin. Cells were viewed with an Olympus Optical IX-70 microscope equipped with an ORCA-ER CCD camera (Hamamatsu Photonics) and a 60ϫ, 1.4 NA plan apochromat objective. Metamorph software was used for quantitative analysis. To quantify A␤ immunofluorescence, 5-10 neurons or Sw-N2a cells were randomly picked from each of three independent experiments, and average intensities were measured in selected areas. To quantify A␤ staining in Sw-N2a cells after transfection with the dynamin-1 cDNA containing GFP, we only considered cells that were transfected (GFP-positive). For quantification of 6E10 antibody internalization and A␤42 staining, intensity threshold was set using Metamorph 6.1 so that fluorescence of neurons was above background fluorescence. Total fluorescence per 30 m of a thresholded neurite was automatically quantified. PSD-95 puncta were quantified as described previously (27). One or two coverslips from each culture were analyzed, 5-10 neurons per coverslip. From each neuron, 3-5 neuritic segments 30 m in length were selected from areas where single puncta could be outlined. Images were thresholded so that only the brightest puncta, with intensity at least twice that of the neuritic shaft, were outlined. Using the integrated morphometric analysis feature in Metamorph, puncta density was automatically measured.
Confocal Microscopy-Immunofluorescence A␤ antibody internalization and intracellular A␤ reduction were examined by confocal microscopy using an Axiovert 100 M inverted microscope equipped with an LSM 510 laser scanning unit and a 63 ϫ 1.4 NA plan apochromat objective (Carl Zeiss, Inc.), Ar488, HeNe1543 nm lasers, and LP560 and BP505-530. Optical sections were acquired at 0.7 m thickness.
Live Cell Microscopy of Antibody Uptake-Cells were imaged using an Olympus Optical IX-70 microscope equipped with an ORCA-ER CCD camera, a 60ϫ, 1.4 NA plan apochromat objective and a 37°C heated chamber. Images were obtained using a Hamamatsu Orca ER digital camera. N2a cells were transfected with human APP-YFP for 3-4 h. Cells were incubated for 10 min on ice with Alexa-555-conjugated 6E10 antibody (6E10 was labeled using an Alexa Fluor 555 Monoclonal Antibody Labeling Kit, Invitrogen) in live imaging solution (120 mM NaCl, 3 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, 10 mM Hepes). Cells were washed twice in ice cold live imaging solution before imaging. Frames were automatically and sequentially acquired every 20 s with the FITC and Rhodamine filters using Metamorph.
Metabolic Labeling-Primary neurons were plated in 10-cm dishes. 7-Day-old cultures after 30-min starvation in cysteine/ methionine-free medium (Invitrogen) were pulsed in fresh cysteine/methionine free medium with 1 mCi [ 35 S]methionine/ cysteine (PerkinElmer Life Sciences) in the presence or absence of 2 g/ml A␤ antibody 6E10 for 20 -30 min and then chased in Neurobasal medium (Invitrogen) for 15 min, 45 min, or for media studies, 90 min. Media were centrifuged for 10 min transferred to clean tubes and immunoprecipitated with A␤ antibody 4G8. Cells were collected in ice-cold PBS and lysed. Samples were immunoprecipitated with A␤ antibody 4G8. 35 S signal was visualized using a phosphoimager system (Hewlett Packard Cyclone).
Statistical Analysis-Statistical comparisons were made using unpaired t tests with significance placed at p Ͻ 0.05. A set of cultures prepared from one mouse embryo was considered as one independent experiment (n ϭ 1). One section prepared from one Tg2576 mouse was considered as one independent experiment (n ϭ 1). Data were expressed as mean Ϯ S.E. Statistical analysis was performed using GraphPad Prism 3.0 software (GraphPad Software).

RESULTS
Treatment with monoclonal A␤ antibodies (6E10, 4G8; Fig.  1A) reduced levels of A␤ in primary neurons at 19 DIV derived from Tg2576 mice harboring the Swedish mutant human APP and in N2a neuroblastoma cells stably transfected with AD Swedish mutant human APP (Sw-N2a) ( Fig. 1, B-G). Specifically, treatment of cultured Tg2576 neurons with 2 g/ml of A␤ antibodies for 24 h resulted in a 36 Ϯ 12% (6E10) and 30 Ϯ 11% (4G8) decrease in levels of A␤ as quantified by Western blot (Fig. 1B, E). In contrast to these N-terminal to mid A␤ domain/ APP antibodies, treatment of neurons with G2-11, a C-terminal-specific A␤42 antibody (Fig. 1A), did not induce changes in intracellular levels of A␤ (Fig. 1E), suggesting that binding to the exposed extracellular domain of A␤ was required for A␤ antibody-mediated reduction in intracellular A␤. Similar reductions in levels of A␤ after treatment with A␤ antibodies were evident by ELISA analyses, where reduction of both A␤40 and A␤42 was observed with either 6E10 or 4G8 treatment (Fig.  1D). To further confirm these biochemical results, we also evaluated intracellular A␤ immunofluorescence (32) following A␤ antibody treatment, which revealed a reduction in intracellular A␤42 immunofluorescence in cultured Tg2576 neurons as shown by confocal imaging (Fig. 1C). Treatment for 24 h with A␤ antibody 6E10 did not alter neuronal morphology or induce neuronal death as evaluated using the TUNEL labeling method or the LDH assay (supplemental Fig. S1). Western blot analyses demonstrated that incubation of Sw-N2a cells with A␤ antibodies had similar reductions in levels of A␤. Specifically, 24 h of treatment with A␤ antibodies 6E10 or 4G8 reduced levels of intracellular A␤ by 51 Ϯ 10% and 54 Ϯ 11%, respectively, compared with untreated controls (Fig. 1F). Since these A␤ antibodies bind A␤ peptides and the A␤ domain within APP, we also treated Sw-N2a cells with a monoclonal antibody directed against the N-terminal ectodomain of human APP (antibody P2-1, Fig. 1A); this treatment did not reduce levels of intracellular A␤ (Fig. 1F). ELISA analyses confirmed the reductions of both A␤40 and A␤42 after 24 h treatment with A␤ antibodies 6E10 or 4G8 in Sw-N2a cells (Fig. 1G).
To investigate the mechanism whereby A␤ antibodies reduced intracellular A␤, we examined whether the binding of cell surface APP by A␤ antibodies might be involved. Sw-N2a or untransfected N2a cells were first incubated with 6E10 or 4G8 for 24 h and then fixed, rinsed and stained with a fluorescent secondary antibody without cell permeabilization. The humanspecific antibody 6E10 stained the cell surface of Sw-N2a cells (expressing human APP) but not the surface of untransfected N2a cells (expressing only mouse APP) ( Fig.  2A). Remarkably, when the secondary antibody was applied after cell permeabilization, both A␤ antibodies 6E10 and 4G8 were pronounced within Sw-N2a cells (Fig. 2B), consistent with the internalization of the antibodies. In contrast, incubation for 24 h with human IgG revealed absence of labeling inside cells (Fig. 2B), indicating that nonspecific antibody uptake was not occurring. Moreover, treatment of N2a cells (lacking the human APP transgene) with human specific 6E10 antibody did not reveal significant intracellular labeling, indicating that 6E10 antibody uptake only occurred in the presence of human APP (Fig. 2B). To evaluate whether this result might be due to less intracellular A␤ in untransfected N2a cells, we treated N2a cells with A␤ antibody 4G8, which is not human specific and also recognizes the murine A␤ domain. Confocal Z stack images indicated a cytoplasmic staining of 4G8 in permeabilized N2a cells (supplemental Fig.  S2), supporting that lower intracellular A␤ levels do not preclude A␤ antibody uptake.
Since endocytosis of surface proteins does not occur at 4°C, we also tested for A␤ antibody uptake at 4°C. After 45-min incubation of Sw-N2a cells with 6E10 at 4°C, 6E10 was now only evident at the plasma membrane and not in intracellular compartments with permeabilization, compared with 37°C where internalization was evident (supplemental Fig. S3). Treatment of Sw-N2a cells for 24 h with antibodies to the C terminus of A␤40 (anti-A␤40) or A␤42 (G2-11), but not to APP, did not reveal intracellular staining after permeabilization (Fig.  2B), supporting that antibody binding to the extracellular A␤ domain of APP was important for internalization. Indeed, Sw-N2a cells incubated with antibody P2-1, directed at the ectodomain of human APP, revealed both a cell surface pattern of staining when secondary anti-FIGURE 1. A␤ antibody treatment reduces cellular levels of A␤. A, sequences in APP/A␤ recognized by the antibodies used in this study (not drawn to scale). G2-11 and anti-A␤40 are specific for the C terminus of A␤42 and A␤40 peptides, respectively. 6E10 (human-specific), APP-ab2 (human-specific), and 4G8 are specific for the extracellular A␤ domain of APP. P2-1 (human specific) and 369 are specific to the N-and C-terminal regions of APP, respectively. B, treatment with A␤ antibodies reduces levels of intracellular A␤. Incubation of Tg2576 neurons with the indicated antibodies 6E10 or 4G8 (2 g/ml for 24 h), directed at the extracellular portion of the A␤ domain within APP, reduces intraneuonal A␤. Levels of full-length APP were unchanged. C, representative images of A␤42 immunofluorescence. Tg2576 neurons treated for 24 h with A␤ antibody 6E10 revealed reduced A␤42 immunofluorescence compared with untreated control Tg2576 neurons. Scale bar: 10 m. D, ELISA analysis revealed reduction of A␤40 and A␤42 in Tg2576 neuron cell lysates after 24-h incubation with either antibody 6E10 or 4G8 (n ϭ 3; *, p Ͻ 0.05; **, p Ͻ 0.01). E, Tg2576 neurons were treated with the indicated antibodies (2 g/ml) for 24 h (n ϭ 9 for 6E10 and 4G8; n ϭ 3 for G2-11) and neuron cell lysates were then analyzed by Western blot. A␤ levels were reduced only by antibodies 6E10 and 4G8 but not G2-11, directed at the C terminus of A␤42. Densitometric quantitation of Western blots performed on Tg2675 neuronal lysates is expressed as relative amount of A␤ in treated compared with untreated cells (*, p Ͻ 0.05; **, p Ͻ 0.01). F, immunoprecipitation followed by Western blot analysis on Sw-N2a cells also revealed reductions in levels of A␤ after 24-h incubation with antibodies against A␤ (6E10, 4G8) but not by an antibody against the APP ectodomain (P2-1). Densitometric quantitation is expressed as relative amount of A␤ in treated compared with untreated cells (n ϭ 4; **, p Ͻ 0.01). G, ELISA analysis confirmed reduction of intracellular A␤ (A␤40 and A␤42) in Sw-N2a cells after 24-h incubation with either antibody 6E10 or 4G8 (n ϭ 4; *, p Ͻ 0.05; **, p Ͻ 0.01).
body was applied without permeabilization ( Fig. 2A) and antibody internalization when secondary antibody was applied after permeabilization (Fig. 2B). To better delineate A␤ antibody uptake in Sw-N2a cells and cultured Tg2576 neurons, we analyzed Z stack confocal images that more clearly reveal the intracellular localization of antibody (supplemental Fig. S4, A  and B). These results supported that recognition of antibodies to the extracellular exposed A␤ domain of cell surface APP was required for antibody internalization.
To investigate whether A␤ antibodies remain associated with APP after internalization, live cell imaging was done on N2a cells transiently transfected with APP-YFP and treated with fluorescently conjugated A␤ antibody 6E10 (Fig. 2C and supplemental movie S1). As indicated by the arrows, internalized fluorescent 6E10 co-localized with APP-YFP within the same vesicle transported in the retrograde direction within a neurite. Internalized A␤ antibody 6E10 also co-localized with APP (369) in neuronal processes in fixed neurons (supplemental Fig. S5). To determine whether there was a correlation between antibody internalization and reduction in A␤, we performed a time course of incubation with 6E10 (10 min, 1 h, or 7 h) in Tg2576 neurons, which revealed a progressive accumulation of A␤ antibody in neurons, especially evident within neuronal processes (Fig. 2, D and E). Quantification of A␤42 immunofluorescence in neuronal processes after antibody treatment revealed a reduction of intraneuronal A␤42 over time, which inversely correlated with antibody uptake in these neurites (Fig.  2, E and F). Internalized A␤ antibody 6E10 did not co-localize with intracellular A␤42 in neurites after 10 min, at which time 6E10 would be expected to localize in early endosomes. However, at 1 h some co-localization of 6E10 with A␤42 was evident (Fig. 2E), consistent with a late endosomal localization at this time (32).
To investigate whether treatment with A␤ antibody promoted internalization of full-length APP from the plasma membrane, we used biotin to specifically label APP at the cell surface of Tg2576 neurons. After 3-h incubation, there was a 40 Ϯ 14% reduction in surface levels of full-length APP in antibody 6E10-treated compared with untreated neurons (n ϭ 5; p Ͻ 0.05), consistent with increased APP internalization upon A␤ antibody binding (Fig. 3A).
To determine whether endocytosis of the A␤ antibody/APP complex upon A␤ antibody treatment was required to reduce intracellular A␤, Sw-N2a cells were transfected with either wild-type GFP-dynamin (wtDyn) or the dominant negative K44E mutant GFP-dynamin (DynK44E) cDNA, which blocks endocytosis (28). Sw-N2a cells transfected with the wtDyn construct and incubated for 3 h with A␤ antibody 6E10 displayed a similar pattern of antibody internalization to the untransfected Sw-N2a cells, while cells transfected with the dominant negative DynK44E construct revealed only surface staining and no internalization of the antibody (Fig. 3B). In Sw-N2a cells transfected with wtDyn, treatment with 6E10 reduced levels of intracellular A␤ by 61 Ϯ 1% compared with untreated cells (n ϭ 3, p Ͻ 0.01). In contrast, in Sw-N2a cells transfected with DynK44E, 6E10 treatment did not reduce levels of intracellular A␤ (Fig. 3C). These results were confirmed by immunofluorescence experiments, which revealed a 41 Ϯ 8% decrease in cel-lular A␤42 in wtDyn transfected cells after 6E10 treatment, whereas antibody treatment did not reduce levels of A␤42 in DynK44E transfected Sw-N2a cells (Fig. 3D). The overall reduction in A␤ levels upon transfection with DynK44E (Fig. 3, C and D) confirms recent work (33) and is consistent with previous studies indicating that APP internalization is important for A␤ generation (34,35). At the same time, the reduction in A␤ generation by the dynamin mutant also limits the interpretation of the data obtained with regards to effects of A␤ antibody treatment.
We used antibody co-localization studies to examine whether internalization of the A␤ antibody/APP complex followed the classical endocytic pathway for cell surface receptors. Since antibodies against subcellular markers are mostly mouse monoclonal antibodies, APP mutant neurons were incubated with a 6E10-like rabbit A␤ antibody, APP-ab2, for different time points (0, 10, 30, 60, 180 min) and then co-stained with markers for subcellular compartments. After 10-min incubation, A␤ antibody was internalized within neurites and showed a punctate pattern that co-localized with the early endosomal marker EEA1 (supplemental Fig. S6). In contrast, after 30-min incubation, A␤ antibody co-localization was more pronounced with the late endosomal/multivesicular body marker Tsg101. At later time points (60 and 180 min), internalized APP-ab2 co-stained with the late endosomal/lysosomal marker Lamp2 within processes and cell bodies (supplemental Fig. S6).
It has been reported that cell surface binding by antibodies can promote internalization and degradation of receptors in lysosomes (36). Antibody binding to the N terminus of cell surface APP was previously used to follow APP internalization to endosomal-lysosomal compartments (37,38). To investigate whether A␤ antibody binding targeted internalized APP to the degradative pathway, we carried out biotin surface labeling on Sw-N2a cells at 4°C followed by incubation of cells at 37°C in the presence or absence of A␤ antibody 6E10 or the APP N-terminal antibody P2-1. After 15-min incubation there was no significant difference in levels of biotinylated APP within cells with 6E10 or P2-1 treatment (supplemental Fig. S7). In contrast after 45-min incubation with antibody 6E10, levels of biotinylated APP decreased by 45 Ϯ 10% compared with untreated control cells (Fig. 4A, n ϭ 3; p Ͻ 0.01), suggesting either increased degradation and/or increased secretase cleavage of APP. Although P2-1 did not reduce levels of A␤ (Fig. 1E), it did decrease levels of biotinylated APP. However, the magnitude of the effect (28 Ϯ 9% compared with control) was less than that for 6E10 (Fig. 4A, n ϭ 3, p Ͻ 0.05).
To further investigate the mechanism whereby A␤ antibodies reduce intracellular A␤, we assessed A␤ generation at earlier time points of treatment using metabolic labeling. Primary neurons were pulsed for 20 min in [ 35 S]methionine/cysteine containing media and then chased for 15 min and 45 min in the presence or absence of 6E10. Although there was no significant change in levels of newly generated intraneuronal A␤ at 15 min (not shown), there was a significant 30 Ϯ 9% reduction in levels of [ 35 S]methionine/cysteine-labeled intraneuronal A␤ at 45 min (Fig. 4B). Another explanation for the reduction of intracellular A␤ levels after A␤ antibody treatment could have been an increase of A␤ secretion. To investigate this hypothesis, we pulsed primary neurons for 30 min in [ 35 S]methionine/cysteine containing media and then chased for 90 min in the presence or absence of 6E10. In the media of 6E10-treated neurons, there was a trend for decreased levels of secreted A␤ compared with untreated cells which did not reach significance (supplemental Fig. S8A) suggesting that antibody treatment either reduced generation and/or increased degradation of newly generated A␤. Therefore, we examined the effect of A␤ antibody treatment on the generation of A␤ by ␤and ␥-secretases.
To assess whether A␤ antibody decreased ␤-site amyloid cleaving enzyme (BACE) processing of APP (39), we measured the amount of the N-terminal APP fragment secreted after ␤-site cleavage of fulllength APP (sAPP␤) using an antibody specific for sAPP␤. Remarkably, there was a 41 Ϯ 3.5% increase in levels of secreted sAPP␤ (n ϭ 4; p Ͻ 0.05; Fig. 4C and supplemental Fig. S8B, upper panel) in A␤ antibody 6E10 treated (3 h) compared with untreated Sw-N2a cells. We also measured levels of N-terminal APP secreted after ␣-site cleavage of full-length APP (sAPP␣). There was a trend for reduced levels of sAPP␣ in the media of 6E10 treated compared with untreated Sw-N2a cells, although this did not reach significance (supplemental Fig. S8B, lower  panel). To further investigate the effect of A␤ antibody treatment on APP processing, we measured levels of APP C-terminal fragments (CTFs) after BACE cleavage (␤CTFs or C99) or ␣-secretase cleavage (␣CTFs or C83) in Tg2576 neurons treated for 24 h with A␤ antibody 6E10. In 6E10 treated neurons, there was a 92 Ϯ 34% increase of ␤CTFs compared with untreated neurons (supplemental Fig. S8C, middle and lower panels). In contrast, there was a trend for decreased levels of ␣CTFs which did not reach significance (supplemental Fig. S8C, lower panel). The lack of increase in ␣CTFs in the presence of increased ␤CTFs argues against a generalized inhibition of ␥-secretase, FIGURE 2. A␤ antibodies that bind at the cell surface are internalized into cells. A, untransfected N2a and Sw-N2a cells were treated with the indicated antibodies for 24 h and then stained with fluorescent anti-mouse IgG without permeabilization. A␤ antibodies 6E10 and 4G8 revealed a cell-surface pattern of staining in Sw-N2a cells. The human-specific antibody 6E10 did not show such surface labeling in untransfected mouse N2a compared with the Sw-N2a cells. APP ectodomain antibody P2-1 revealed surface staining in Sw-N2a cells. B, untransfected N2a and Sw-N2a cells were treated with the indicated primary antibodies for 24 h, fixed, and then permeabilized and stained with fluorescent secondary antibodies. Only antibodies directed against the extracellular domain of APP (P2-1), including those to the A␤ domain (6E10, 4G8), were internalized. In contrast, mouse IgG, antibody G2-11 (A␤42), and antibody A␤40 were not internalized, and untransfected N2a cells lacking human APP did not demonstrate uptake of the human A␤-specific antibody 6E10. C, live cell imaging of N2a cells transfected with APP-YFP and incubated with Alexa-555-conjugated 6E10 for 30 min with images acquired at 20-s intervals. Internalized fluorescently tagged A␤ antibody 6E10 (red) co-localized with APP-YFP (green) in the same vesicle moving retrograde in a process toward the cell body. The images were offset 5 pixels horizontally to better differentiate the fluorophores. The artificially large and confluent yellow area of co-staining just to the right and under the darkly staining nucleus in the cell body is secondary to the increased gain required to visualize fluorescence in the thinner process below (magnified in the lower set of frames). D, internalization of A␤ antibody 6E10 with time correlated with reduced A␤42 within Tg2576 neurons. Scale bar: 10 m. E, detail of representative processes more clearly revealed the correlation between increasing internalization of A␤ antibody 6E10 (green) and decreasing levels of A␤42 (red). Although internalized A␤ antibody 6E10 co-localized with APP/␤CTF using the APP C-terminal antibody 369 (see supplemental Fig. S5), internalized 6E10 did not co-localize with intracellular A␤42 at 10 min, while at 1 h some co-localization was evident (F, arrowhead, merged panel). F, quantification of A␤42 and antibody 6E10 fluorescence in neurites of Tg2576 neurons over time (n ϭ 3; *, p Ͻ 0.05; **, p Ͻ 0.01; compared with 10-min time point; scale bar: 10 m).

FIGURE 3. APP endocytosis is promoted by A␤ antibody binding and is required for reduction of cellular A␤.
A, Tg2576 neurons (19 DIV) were treated with A␤ antibody 6E10 for 3 h and then incubated on ice with biotin to label surface APP. Western blot analysis revealed decreased levels of surface full-length APP in A␤ antibody 6E10-treated compared with untreated control neurons. Total levels of APP were unchanged in cell lysates. B, Sw-N2a cells transiently transfected with wild-type GFP-dynamin (wtDyn) or dominant negative GFP-dynamin (DynK44E) revealed internalization of A␤ antibody 6E10 after 3 h treatment in both untransfected control cells (untransf) and wtDyn-transfected cells but not in DynK44E-transfected cells, which revealed a surface pattern of staining. C and D, in Sw-N2a cells transfected with wtDyn, treatment with 6E10 reduced levels of intracellular A␤ by 61 Ϯ 1% compared with untreated cells. In contrast, in Sw-N2a cells transfected with DynK44E, 6E10 treatment did not reduce levels of intracellular A␤. A␤ was measured by Western blotting (C) and A␤42 immunofluorescence (D). Corresponding quantitation was expressed as a relative amount of A␤ in treated compared with untreated cells (n ϭ 3; *, p Ͻ 0.05; **, p Ͻ 0.01; scale bar: 10 m).
because if that had occurred both ␣and ␤CTFs should have been increased. Thus, A␤ antibody 6E10 did not inhibit BACE or ␥-secretase processing of APP but rather appeared to augment BACE cleavage of APP. To further confirm that A␤ antibody treatment did not decrease ␥-secretase processing of APP, we examined N2a cells stably transfected with human APP C99, which is cleaved only by ␥-secretase (31). Similar to Sw-N2a cells, when C99-N2a cells were treated with A␤ antibody 6E10, there was a cell surface pattern of staining when secondary antibody was applied without permeabilization and evidence of internalization when secondary antibody was applied after permeabilization (data not shown). Treatment of APP C99-N2a cells with antibody 6E10 for 3 h did not increase levels of C99 compared with untreated cells but rather showed a trend for decrease (Fig.  4D). Had 6E10 decreased ␥-secretase activity, C99 levels should have increased, as was seen when ␥-secretase was inhibited with DAPT (Fig. 4D). To examine the effect of A␤ antibody treatment with inhibition of ␥-secretase activity, we treated Sw-N2a cells for 3 h with ␥-secretase inhibitor DAPT in the presence or absence of 6E10. Levels of ␤CTFs were increased by 19 Ϯ 6% in DAPT and 6E10 treated cells compared with those treated with DAPT alone (supplemental Fig. S9A). This result is consistent with the data demonstrating that A␤ antibody treatment increased levels of sAPP␤ and ␤CTF (Fig. 4C and supplemental Fig. S8, B and C). Performing the same experiment on C99-N2a cells, which precludes ␤-cleavage, there was a trend for a decrease, which did not reach significance, in levels of C99 after 3 h of treatment with DAPT and 6E10 compared with DAPT alone (supplemental Fig. S9B). Since the A␤ antibody mediated reduction of A␤ did not appear to result from decreased secretase cleavage of APP, and since internalized A␤ antibodies trafficked to late endo- . A␤ antibody induced reduction of A␤ does not act via ␤or ␥-secretase inhibition and requires the late endosomal/lysosomal system. A, levels of internalized biotinylated full-length APP in Sw-N2a cells treated on ice with 6E10 or P2-1 antibody, followed by 45-min incubation at 37°C, were reduced compared with untreated Sw-N2a cells (n ϭ 3). Densitometric quantitation is expressed as a ratio of biotinylated APP to total APP in treated compared with untreated cells. B, metabolic labeling of primary neurons pulsed for 20 min with [ 35 S]methionine and chased for 45 min in the presence or absence of antibody 6E10. In the presence of 6E10 antibody, levels of newly generated A␤ were reduced compared with untreated controls (n ϭ 4). Densitometric quantitation is expressed as relative amount of A␤ in treated compared with untreated cells. C, levels of secreted APP␤ (sAPP␤) were increased in conditioned media of Sw-N2a cells after 3-h incubation with A␤ antibody 6E10 (n ϭ 4). Densitometric quantitation of sAPP␤ in treated compared with untreated cells is shown. D, A␤ antibody treatment of C99-N2a cells did not inhibit ␥-secretase cleavage. C99-N2a cells were treated for 3 h with either A␤ antibody 6E10 or ␥-secretase inhibitor DAPT. There was a tendency for reduction in levels of C99 in 6E10-treated cells compared with untreated C99-N2a cells. As expected, DAPT treatment induced an increase in levels of C99 (n ϭ 4). Densitometric quantitation of C99 in treated compared with untreated cells is shown. E, C99-N2a cells were incubated on ice with biotin in the presence or absence of A␤ antibody 6E10. After 45 min of incubation with 6E10 at 37°C, levels of internalized biotinylated C99 were reduced in C99-N2a cells compared with untreated control cells (n ϭ 3). Densitometric quantitation is expressed as a ratio of biotinylated C99 to total C99 in treated compared with untreated cells. F, lysosomal inhibition with chloroquine (100 M) prevented the A␤ antibody-mediated (6E10) reduction of intracellular A␤ (n ϭ 4). Densitometric quantitation is expressed as relative amount of A␤ in treated compared with untreated cells (*, p Ͻ 0.05; **, p Ͻ 0.01).
somes/lysosomes, we considered that A␤ antibodies promote the late endosomal/lysosomal degradation of APP and APPderived products, such as C99 and A␤. To investigate whether A␤ antibody treatment induced the degradation of C99, we carried out biotin labeling on C99-N2a cells at 4°C followed by incubation of cells at 37°C in the presence or absence of A␤ antibody 6E10. After 45-min incubation with antibody 6E10, levels of biotinylated C99 decreased by 57 Ϯ 7% (Fig. 4E), consistent with A␤ antibody induced degradation of C99. To further investigate whether the late endosomal/lysosomal system is involved in the A␤ antibody mediated reduction of A␤, we tested whether inhibition of late endosomal/lysosomal function would interfere with the ability of A␤ antibodies to reduce intracellular A␤. Indeed, incubation of Sw-N2a cells with 6E10 (3 h) in the presence of the lysosomal inhibitor chloroquine (40) prevented and/or counterbalanced the A␤ antibody-induced reduction of intracellular A␤ ( Fig. 4F; similar results were obtained using ammonium chloride; data not shown).
Synaptic dysfunction is considered to be the earliest neurobiological alteration in AD (41,42), and reduction of intraneuronal A␤ was the best A␤-correlate of cognitive improvement in an AD mouse model (12). Therefore, we examined whether reduction of intraneuronal A␤ by A␤ antibodies could protect against the synaptic alterations that we previously described in APP mutant neurons in culture (27). We confirmed that the number of PSD-95 puncta (an important scaffold protein of the post-synaptic density) was reduced in processes of APP mutant compared with wild-type neurons at 19 DIV (Fig. 5A). Immunofluorescence for synapsin-1, a presynaptic protein that remains unchanged in Tg2576 neurons (27), highlights the similarity of the neurites in the representative images (Fig. 5A). Remarkably, treatment of Tg2576 neurons with A␤ antibody 6E10 (or A␤ antibody 4G8; supplemental Fig. S10) for 24 h restored the number of PSD-95 puncta to 96 Ϯ 6% of wild-type levels ( Fig. 5B; for an additional representative figure, see supplemental Fig. S11). In contrast, treatment with a C terminus-specific A␤42 antibody (Chem42) was unable to restore PSD-95 puncta in Tg2576 neurons (supplemental Fig. S12). Treatment of wildtype neurons with A␤ antibody 6E10 had no effect on levels of PSD-95 puncta (data not shown).

DISCUSSION
A␤ immunotherapy remains an exciting therapeutic direction for AD, although the biological mechanism(s) whereby A␤ antibodies reduce A␤ in brain and improve cognitive function in AD mouse models are incompletely understood. Several hypotheses have been proposed. The "sink hypothesis" suggests that peripherally administrated A␤ antibodies can reduce levels of A␤ in the plasma and drive efflux of A␤ from the brain, where it is more concentrated, to the periphery (9). Another hypothesis is based on evidence that peripherally administrated A␤ antibodies can cross the blood brain barrier and enter the central nervous system (4), where A␤ antibodies can mediate degradation of A␤ aggregates by inflammatory cell activation (4,43). However, evidence also indicates that Fc-mediated antibody-directed microglial activation is not necessary to reduce A␤ plaques, since A␤ immunotherapy was effective in FcR-␥ chain knock-out/APP mutant transgenic mice (44) and F(abЈ)2 fragments reduced plaque pathology in Tg2576 mice (45). This suggests that other mechanism(s) may be occurring as well. Recent evidence and our data suggest an additional scenario where A␤ antibodies reduce intracellular A␤.
Intraneuronal A␤ accumulation is increasingly being linked with early, preplaque electrophysiological, synaptic, and pathological abnormalities (46). For example, A␤42 accumulation and oligomerization were associated with ultrastructural pathology within distal processes and synapses prior to, and in areas devoid of, plaques in AD transgenic mice and human AD brain (17,26). A␤ antibodies were shown to reduce intracellular A␤ in triple transgenic mice (11), and this reduction in intracellular A␤ was the best correlate of cognitive improvement (12). The biological mechanism by which A␤ antibodies reduced intracellular A␤ was unclear. Interestingly, antibodies against the cleavage site of BACE on APP (47) and intrabodies against A␤ (48) were suggested as new cellular therapeutic strategies for AD.
We provide evidence that antibodies against A␤, previously shown by multiple groups to reduce plaque pathology in vivo (8), reduce intracellular levels of A␤ in cultured neurons by binding to the extracellular A␤ domain and protect against synaptic alterations in APP mutant neurons. We demonstrate that A␤ antibodies directed to the N-terminal to mid-domain of A␤ can be specifically endocytosed after binding at the cell surface. Our results could be important in why passive immunization is especially effective when using antibodies directed at the N-terminal region of A␤ (4,11,12,49,50).
A potential mechanism by which A␤ antibodies reduced intracellular A␤ could have been inhibition of ␤or ␥-secretase activities. We did not observe such inhibition. Specifically, the increase in sAPP␤ and, in C99-N2a cells, the lack of an increase in C99 with A␤ antibody treatment argued against ␤or ␥-secretase inhibition, respectively. In fact, the increase of C99 in cells expressing full-length APP in conjunction with increased sAPP␤ secretion suggests that the internalization of full-length APP induced by A␤ antibody treatment actually promotes BACE cleavage. Another possible mechanism for the antibody mediated reduction in A␤ could have been from increased clearance of intracellular to extracellular A␤, although the lack of an increase in levels of extracellular A␤ argues against this possibility. In fact, there was a trend for decreased A␤ secretion with A␤ antibody treatment that did not reach significance. Another potential mechanism for A␤ antibody-mediated clearance of intracellular A␤ is by enhancing cellular degradation after antibody binding to the A␤ domain of cell surface APP. Since A␤ antibodies directed to the C terminus of A␤ did not reduce intracellular A␤, we hypothesized that A␤ antibodies act by binding to full-length APP and/or APP CTFs and not on potentially surface-associated A␤ in our experiments. The binding and internalization of antibodies to the ectodomain of cell surface APP was previously used to study APP internalization and trafficking and was employed to follow APP to endosomal-lysosomal compartments (37,38,51). Recent studies are increasingly suggesting an important role of endosomes in APP processing; for example, the majority of APP transported down axons by fast axonal transport was reported to be full-length, supporting endosomal secretase cleavage in neurons (52). Interestingly, alterations in the endosomal-lysosomal system are among the earliest changes described in AD and Down syndrome brains (53). Our data provide evidence for involvement of the endosomal/lysosomal system in intraneuronal A␤ clearance after A␤ antibody treatment. Co-localization of endocytosed A␤ antibody with A␤42 was evident at 1-h incubation, when the internalized antibody co-stained with late endosomal/lysosomal markers and not at 10 min when internalized antibody co-localized with early endosomes. Our results are most consistent with A␤ antibody-induced increased internalization of APP from the cell surface to early endosomes, followed by increased ␤-cleavage (elevated C99) and then enhancement of C99 trafficking to the late endosomal lysosomal pathway for degradation. We hypothesize that enhanced trafficking of A␤ antibody bound C99 through late endosomes where ␥-secretase components have been localized (54,55) limits ␥-cleavage and thereby also increased A␤ secretion. That inhibition of late endosomes/lysosomes with chloroquine or ammonium chloride prevented A␤ clearance after A␤ antibody treatment supports the involvement of the endosomal-lysosomal system in the reduction of A␤. Since both APP ectodomain and A␤ antibodies promoted surface APP reduction, although the latter were more effective, only A␤ antibodies reduced levels of cellular A␤, and this suggests that the dissociation of the APP ectodomain antibody from ␤CTFs after BACE cleavage might preclude the enhanced degradation that occurs when A␤ antibody remains bound to the A␤ domain of C99. Our results support the scenario of A␤ antibodies inducing the internalization of APP from the plasma membrane to early endosomes where increased BACE cleavage appears to occur, followed by induction of the late endosomal-lysosomaldependent clearance of ␤CTFs, and potentially A␤. Since A␤ antibody alone tends to decrease levels of C99 in C99-N2a cells, our results suggest increased degradation of C99 and A␤ rather than ␥-secretase inhibition. The lack of a statistically significant decrease of ␤CTFs in C99-N2a cells treated with DAPT and A␤ antibody compared with DAPT alone might be due to altered trafficking of C99 upon ␥-secretase inhibition that thereby inhibits C99 degradation. In fact, altered ␤CTF trafficking was reported in neurons of PS1 conditional knock-out mice, where ␤CTFs accumulated abnormally at synapses (56). Our data cannot fully exclude that A␤ antibody treatment promotes ␤CTF processing by ␥-secretase followed by increased degradation of the resultant A␤-A␤ antibody complex rather than, or in addition to, primarily promoting degradation of the ␤CTF-A␤ antibody complex.
A␤ antibodies have been reported to block alterations of synapses induced by extracellular A␤ oligomers (50,57). That C-terminal specific A␤ antibodies can also be protective in A␤ immunotherapy supports that antibody effects on extracellular A␤ are also involved (58,59). Increasing evidence supports an as yet poorly understood dynamic relationship between extracellular and intracellular A␤, modulation of which might be especially important in A␤ antibody induced therapeutic effects (60). High levels of extracellular A␤ were shown to induce up-regulation of newly generated intracellular A␤42 (61). The mechanism whereby extracellular A␤ causes cell death in cultured neurons appears to be related to a dynamic relationship also between extracellular A␤ and cell surface APP, since toxicity did not occur in APP knock-out neurons (62) or cells harboring mutations in the YENPTY motif within the C terminus of APP (63). Thus, neurotoxicity might additionally require effects on intracellular A␤.
In summary, in addition to effects on inflammatory mechanisms of A␤ clearance and on extracellular A␤ oligomers, among others, our data underscores that another mechanism whereby A␤ antibodies may play a critical role in A␤ immunotherapy is via reduction in intracellular A␤. A better understanding of the molecular mechanism(s) whereby A␤ immunotherapy leads to reduced A␤ accumulation and improved cognitive function may lead to novel therapeutic approaches for AD.