Reduction of Synaptojanin 1 Accelerates Aβ Clearance and Attenuates Cognitive Deterioration in an Alzheimer Mouse Model*

Background: Recent studies have linked synaptojanin 1 (synj1) with Alzheimer disease (AD). Results: We report that synj1 reduction decreases amyloid plaque burden and attenuates cognitive deterioration in an AD mouse model. These effects are mediated through accelerating endosomal/lysosomal degradation of Aβ. Conclusion: Our data suggest a novel mechanism by which synj1 reduction promotes Aβ clearance. Significance: These studies implicate a therapeutic strategy for AD. Recent studies link synaptojanin 1 (synj1), the main phosphoinositol (4,5)-biphosphate phosphatase (PI(4,5)P2-degrading enzyme) in the brain and synapses, to Alzheimer disease. Here we report a novel mechanism by which synj1 reversely regulates cellular clearance of amyloid-β (Aβ). Genetic down-regulation of synj1 reduces both extracellular and intracellular Aβ levels in N2a cells stably expressing the Swedish mutant of amyloid precursor protein (APP). Moreover, synj1 haploinsufficiency in an Alzheimer disease transgenic mouse model expressing the Swedish mutant APP and the presenilin-1 mutant ΔE9 reduces amyloid plaque load, as well as Aβ40 and Aβ42 levels in hippocampus of 9-month-old animals. Reduced expression of synj1 attenuates cognitive deficits in these transgenic mice. However, reduction of synj1 does not affect levels of full-length APP and the C-terminal fragment, suggesting that Aβ generation by β- and γ-secretase cleavage is not affected. Instead, synj1 knockdown increases Aβ uptake and cellular degradation through accelerated delivery to lysosomes. These effects are partially dependent upon elevated PI(4,5)P2 with synj1 down-regulation. In summary, our data suggest a novel mechanism by which reduction of a PI(4,5)P2-degrading enzyme, synj1, improves amyloid-induced neuropathology and behavior deficits through accelerating cellular Aβ clearance.

Alzheimer disease (AD) 3 is neuropathologically characterized by senile plaques containing ␤-amyloid peptides (A␤), as well as neurofibrillary tangles consisting of hyperphosphorylated Tau. The amyloidogenic A␤ peptide is proteolytically derived from the amyloid precursor protein (APP) by distinct enzymatic activities known as ␤-secretase (or BACE) and ␥-secretase (1,2). Overproduction and impaired A␤ clearance can both lead to A␤ accumulation. Recent evidence indicates that late onset AD cases are likely caused by an overall impairment in A␤ clearance (3).
Synaptojanin 1 (synj1), a major negative regulator of the levels of PI(4,5)P 2 in the nervous tissue, has been implicated in the regulation of endocytic traffic at synapses (4). Trisomy of synj1 in Down syndrome mouse models causes a deficiency in PI(4,5)P 2 as well as learning deficits (5), whereas synj1 haploinsufficiency is protective against A␤-induced neurotoxicity in mouse models of AD (6,7). In the current study, we link synj1 to regulation of cellular A␤ clearance.
We have found that down-regulation of synj1 leads to reduced A␤ accumulation and amyloid plaque load. synj1 knockdown promotes A␤ uptake and lysosomal trafficking, thereby facilitating cellular A␤ clearance without affecting BACE or ␥-secretase expression or enzymatic activities to generate A␤. These effects are dependent on PI(4,5)P 2 that increases with synj1 down-regulation. As a consequence, reduction of synj1 attenuates amyloid-induced neuropathologic changes and behavior deficits in an AD transgenic mouse model. Our findings uncover a novel regulatory mechanism by a PI(4,5)P 2 -degrading enzyme synj1 that controls cellular A␤ degradation through the endosomal/lysosomal pathway.
Cell Transfection-For siRNA analysis, N2a cells were seeded at 50 -60% confluence and transfected with 200 pmol of synj1 siRNA duplex versus control duplex (per well of a 6-well plate) using Lipofectamine RNAimax (Invitrogen) according to the manufacturer's instructions.
Cell Lysate Analysis-After transfection, the cells were harvested in lysis buffer (14). Equal amounts of total protein were loaded onto 10 -20% Tricine SDS-PAGE gels for electrophoresis and transferred to PVDF membranes. The membranes were analyzed by Western blot using 6E10 to detect holoAPP and ␤CTF/C99. A␤ 40 and A␤ 42 levels in media were determined by human ELISA kits (WAKO), according to the manufacturer's instructions. In some experiments, the amount of A␤ in media and lysate with synj1 or control siRNA treatment were determined in the presence of lysosomal inhibitors (pepstatin A, 10 M; leupeptin, 100 M; E-64d, 50 M) to block lysosomal degradation of A␤. Alternatively, a PIP 2 modulator m-3m3FBS or its inactive analog o-3m3FBS was added with or without lysosomal inhibitors to determine whether the degradation of A␤ with synj1 reduction is dependent upon elevated PIP 2 levels.
Immunoprecipitation-Lysates were diluted with immunoprecipitation (IP) buffer (10) and immunoprecipitated using antibody 4G8 followed by immunoblotted with 6E10 for detection of intracellular A␤ and ␤CTF. Media were immunoprecipitated using 4G8 antibodies (Covance) and immunoblotted with 6E10 for detection of media A␤ as described before (15). In some experiments, after siRNA transfection, N2a cells were treated with Me 2 SO lysosomal inhibitors or PIP 2 modulators overnight before analysis of media and lysate A␤ production.
Brain Lysate Preparation and Analysis-Mouse brains were rapidly removed, hemisected, and snap frozen before further analysis. Each frozen hemi-brain was then processed via stepwise solubilization (14,18). Lysates of hemi-brains derived from APP/PS1 ϩ/Ϫ synj1 ϩ/ϩ or APP/PS1 ϩ/Ϫ synj1 ϩ/Ϫ at 9 months of age were analyzed by SDS-PAGE and immunoblotted with 6E10 to determine levels of holoAPP and ␤CTF/ C99. Levels of A␤ 40 and A␤ 42 were determined by human A␤ 40 and high sensitivity human A␤ 42 ELISA kits (Wako), according to the manufacturer's instructions. The results were normalized to wet brain weight.
Immunohistochemistry and LCO Staining of Amyloid Plaque-The hemi-brains of 9-month-old APP/PS1 ϩ/Ϫ synj1 ϩ/ϩ or APP/PS1 ϩ/Ϫ synj1 ϩ/Ϫ mice were processed, embedded, and sectioned at 10 m. For amyloid plaque quantitation, the blocks were serial sectioned across the whole hippocampal regions, and every eight sections were used for staining (ϳ20 sections/ animal). After deparaffination and antigen retrieval process, the brain sections were treated with anti-amyloid antibody AB2454 or 6E10 (1:200 dilution in TBS buffer) overnight at 4°C. Following a thorough rinse in TBS buffer, the sections were incubated with secondary antibodies, i.e., biotinylated goat anti-mouse IgG or biotinylated goat anti-rabbit IgG (diluted 1:200 in TBS buffer), and then incubated with avidin biotinylated enzyme complex and 3,3Ј-diaminobenzidine. The amyloid plaque load density in the hippocampal region, as well as CA1/3 and dentate gyrus subregions, was measured using the Sinq Image Analysis System (Sinq Inc.) (19,20). Alternatively, the brain sections were stained with anti-amyloid antibodies followed by incubated with secondary antibody Texas Red-conjugated antirabbit or anti-mouse IgG as well as LCO reagent (21) for double staining of amyloid before confocal microscopy analysis (LSM510; Zeiss).
Novel Object Recognition Task-Human Swedish APP and FAD-linked PS1 mutant transgenic mice show a deficit in the novel object recognition memory task (22). Littermates at 4 -5 months of age were tested in novel object recognition, with onset of the exploration time defined as the moment the head of the animal approached the object within a 2-cm radius. Briefly, animals were habituated to testing arena for 10 min on day 1. After 24 h, the animals were tested with two trials (identical objects for 10 min followed by one novel per one familiar object for 4 min). The inter-trial interval was 1 h. Each animal was videotaped from overhead cameras and scored for total time spent investigating objects per trial. The objects were randomized left and right, as well as between animals. They were cleaned with alcohol between trials/animals to prevent odor cues.
Spatial Alternation/Y Maze Task-Briefly the Y maze test was performed in a Y-shaped maze with three white opaque plastic arms at a 120°angle from each other (23,24). The test involves animals freely explored the three arms of the Y maze for 8 min (see Fig. 3B). Appropriate external cues were assigned for each arm. All animal activities were recorded by video camera, and the number of arm entries and the number of triads were determined to calculate the percentage of alternation. An entry occurs when all four limbs are within the arm.
A␤ Uptake Assay-N2a Swedish mutant APP cells were transfected with synj1 siRNA or control siRNA duplex as indicated. At 4 day post-transfection, the cells were incubated with biotin-conjugated A␤ 42 in PBS for 1 h at 37°C as described previously (27,28). The cells were subsequently washed by icecold PBS five times followed by incubation with 100 M glutathione for 15 min at 4°C. The cells were washed another two times with PBS and then incubated with 5 mg/ml iodoacet-amide for 15 min followed by washing with ice-cold PBS five times before being harvested in cell lysis buffer (ThermoFisher Scientific) as described (14). Lysates were subsequently immunoprecipitated by streptavidin beads and analyzed by SDS-PAGE and Western blotting with 6E10 for internalized biotin-A␤ 42 . Similar experiments were performed in primary neuron culture derived from embryonic day 17 synj1 ϩ/ϩ versus synj1 Ϫ/Ϫ animals. In some experiments, cells after siRNA treatment were incubated with a PIP 2 modulator (m-3-FBS or an inactive analog o-3-FBS) at 6.25 M overnight before being subjected to an A␤ uptake assay.
Cellular A␤ Turnover Rate Assay-For a time course of A␤ degradation, N2a Swedish mutant APP cells were transfected with synj1 siRNA or control siRNA duplex as indicated. At 4 days post-transfection, the cells were divided into 6-well plates coated with 100 g/ml polyornithine and allowed to recover for 5 h. The cells were subsequently treated with 50 g/ml cycloheximide (CHX) in serum-free and antibiotic-free DMEM (29). At the indicated time points, the cells were washed in ice-cold PBS and then harvested in radioimmune precipitation assay buffer as described (14). Lysates were subsequently analyzed by SDS-PAGE and Western blotting with 6E10 for A␤ and holoAPP.
Neuronal Culture and Confocal Microscopy-Primary hippocampal neurons were obtained from 17-day-old embryos of synj1 ϩ/ϩ , synj1 ϩ/Ϫ , and synj1 Ϫ/Ϫ mice and grown in neurobasal medium (Invitrogen) supplemented with 0.5 mM GlutaMAX (Invitrogen), 2% B27 (Invitrogen), and 1% penicillin-streptomycin (Invitrogen) in dishes containing poly-Dlysine coated cover glasses (Nalge Nunc International, Rochester, NY) for 7 days in vitro. They were then incubated with Alexa-A␤ 555 or Alexa-A␤ 488 for various time periods before fixation and double-stained for endosomal and lysosomal markers Rab5 and LAMP1 as well as a nuclear marker DAPI (blue) before confocal fluorescence microscopy analysis (Zeiss). Similar experiments were performed in N2a cells after control or synj1 siRNA transfection, followed by incubation with Alexa-A␤ 555 or Alexa-A␤ 488 . Fluorescent colocalization analysis was quantified using the Zen program to calculate colocalized pixels (subtracted by background) and colocalization coefficiency as previously described (30,31). The data are presented as percentages of control Ϯ S.E.
Statistical Analysis-Densitometric analysis of Western blot bands (integrated density) was performed using Multigauge v3.1 software (Fujifilm). The levels of holoAPP, ␤CTF/C99, A␤, synj1, BACE1, PS1, biotin-A␤, and C100-/N100-FLAG were normalized to actin levels and expressed as percentages of control. Absolute A␤ 40 and A␤ 42 concentrations were quantitatively determined by sandwich ELISA (Wako) and expressed as percentages of control. For all analysis, independent sample t tests (parametric design) were used to determine significant mean differences (the threshold for significance is set at p Ͻ 0.05). Where two or more variables were compared, a one-way analysis of variance with post hoc tests were used to test group differences for multiple comparisons. All statistical analysis was performed using SPSS v21.0.
synj1 Haploinsufficiency in AD Transgenic Mice Decreases Hippocampal Amyloid Plaque Load-We then investigated whether genetic reduction of synj1 affects A␤ generation in adult animals in vivo. The synj1 haploinsufficient (synj1 ϩ/Ϫ ) mice (1) with AD transgenic background expressing human Swedish APP and PS1⌬E9 mutations (16,17) were generated, and amyloid plaque burden, as well as A␤ 40 and A␤ 42 levels in the brains of 9-month-old mice, were analyzed. As shown in Fig. 2A, amyloid plaque load in hippocampus was determined by both immunoperoxidase (left panels) and immunofluorescence (right panels). There was an obvious reduction in the amount of dense core plaques (doubled stained by LCO and anti-amyloid antibody AB2454). The stereological quantitation of amyloid plaques in 9-month-old APP/PS1 ϩ/Ϫ synj1 ϩ/Ϫ male mouse brains showed that total plaque load in hippocampus, as well as in dentate gyrus subregions, was reduced by 34.3 Ϯ 12.1% (p ϭ 0.039) and 27.1 Ϯ 11.1% (p ϭ 0.014), respectively, compared with their APP/ PS1 ϩ/Ϫ synj1 ϩ/ϩ littermates (Fig. 2B, left panel). However, total amyloid plaque load in the cortices of APP/PS1 ϩ/Ϫ synj1 ϩ/Ϫ mice was not clearly reduced when compared with wild type littermates, suggesting that partial reduction of synj1 may not be sufficient to alter heavy amyloid load in cortical regions at this stage of pathological progression.
synj1 Haploinsufficiency in AD Transgenic Mice Attenuates Learning and Memory Deficits-Next, we investigated the effects of synj1 down-regulation on learning and memory impairments in AD transgenic mice. Previously studies have shown that one strain of human Swedish APP and FAD-linked PS1 double mutant transgenic mice (Swedish APP/PS1 M146V) exhibit deficits in the novel object recognition memory task (22). Similarly, our double transgenic mice (APP/ PS1⌬E9 ϩ/Ϫ /synj1 ϩ/ϩ ) spent a comparable amount of time between novel and familiar objects ( Fig. 4A; 45.6 Ϯ 14.6% of total time with novel object, n ϭ 5), indicating that they were unable to discriminate between these two objects. In contrast, APP/PS1⌬E9 ϩ/Ϫ /synj1 ϩ/Ϫ animals displayed improved exploratory behavior compared with APP/PS1⌬E9 ϩ/Ϫ /synj1 ϩ/ϩ littermates, spending more time with the novel object (66.0 Ϯ 11.8%, n ϭ 5, p ϭ 0.019), similar to the degree of wild type and synj1 ϩ/Ϫ mice without AD transgenic background (74.5 Ϯ 11.1%, n ϭ 5, versus 68.6 Ϯ 11.8%, n ϭ 5). However, the total amount of time spent in exploring objects (seconds) was comparable among the four groups (data not shown).
synj1 Haploinsufficiency Does Not Affect ␥-Secretase Activity in Vitro-Previous reports indicate that A␤ generation and APP processing by PS1 can be modulated by PI(4,5)P 2 (11). Because PI(4,5)P 2 is a major substrate for synj1 and knockdown of synj1 elevates brain PI(4,5)P 2 levels (Refs. 1 and 2 and Fig.  3C), we next investigated whether down-regulation of synj1 affected ␥-secretase processing of APP.
Utilizing a well established in vitro ␥-secretase assay (25, 26), we characterized the activity of solubilized ␥-secretase derived from membrane fractions of N2a wild type cells transfected with synj1 siRNA or control siRNA toward the substrates C100-FLAG (Fig. 5A) and N100-FLAG (Fig. 5B). Synj1 knockdown did not affect ␥-secretase cleavage of C100 to generate A␤ (101 Ϯ 11.1% of controls, n ϭ 6, p ϭ 0.74) or of N100 to generate its cleavage product (recognized by Val 1744 antibody only; 129.9 Ϯ 27.4, n ϭ 5, p ϭ 0.11). A ␥-secretase inhibitor, DAPT (12,13), was used as a control, and reduced cleavage of both C100 and N100 substrates (69.3 Ϯ 2.3% and 38.2 Ϯ 16% of controls, respectively; Fig. 5, A and B, lane 4). It should also be noted that there was no change in levels of ␥-secretase/PS1 total proteins in APP/PS1 ϩ/Ϫ synj1 ϩ/Ϫ mouse brains (Fig. 3A). Moreover, the levels of ␤CTF (precursor of A␤) are not changed with synj1 knockdown in cells and in transgenic mouse brains (Figs. 1 and 3). These data altogether suggest that synj1 down-regulation does not affect ␥-secretase cleavage of APP to generate A␤ in our system.
Reduction of synj1 Accelerates A␤ Uptake into the Cells-We then investigated the rate of cellular A␤ uptake with reduced synj1 expression. As shown in Fig. 6A, the amount of biotin-A␤ 42 after 1 h of incubation was increased by 25.2% in N2a cells with synj1 siRNA knockdown (125.2 Ϯ 13.4% of controls, n ϭ 3, p ϭ 0.03). A similar increase in biotin-A␤ 42 uptake was seen in primary neurons derived from embryonic day 17 synj1 Ϫ/Ϫ animals in comparison to wild type cells (119.6 Ϯ 9.0%, n ϭ 4, p ϭ 0.039).
It has been reported that the majority of internalized A␤ traffics through Rab5-and Rab7-positive early and late endosomes, respectively. Most internalized A␤ is delivered to the lysosomal pathway for degradation (32). We then studied whether synj1 reduction would affect the amount of Alexa 488conjugated A␤ 42 (green fluorescent signals) delivered to early endosomes (recognized by an early endosomal marker Rab5, red fluorescent signals) after 1 h of incubation. As shown in Fig.  6B, the amount of fluorescent-conjugated A␤ 42 delivered to Rab5 ϩ early endosomes was increased by 76.2 Ϯ 43.4% (p ϭ 0.03) with synj1 siRNA treatment when compared with controls. A magnified image of individual cells showed increased colocalization of Alexa 488 -A␤ and Rab5 ϩ early endosomes (punctate structures shown in Fig. 6B, right panels). In addition, immunofluorescence staining of a second endosomal marker, EEA1, showed a similar pattern of increase in the amount of A␤ colocalized with EEA1 ϩ endosomes with synj1 reduction. However, there were comparable amounts of Alexa 488 -conjugated A␤ 42 colocalized with Rab11 ϩ recycling endosomes after 3 h of incubation (data not shown), suggesting that the rate of A␤ recycling between endosomes and plasma membrane is not changed by syn1j reduction.
We next studied whether increased A␤ uptake with synj1 reduction is dependent upon elevated PIP 2 levels, using a pharmacological reagent m-3m3FBS that activates phospholipase C to deplete PIP 2 inside the cells (11). As shown in Fig. 6C, the amount of A␤ taken up by control cells was modestly decreased with reduction of PIP 2 (86.4 Ϯ 7.8%, p ϭ 0.24). An inactive analog of m-3m3FBS was used as a control (o-3m3FBS). However, in synj1 knockdown cells, the amount of A␤ taken up by cells remained increased even with treatment to reduce PIP 2 (149.6 Ϯ 8.0% in m-3-FBS treated versus 144.3 Ϯ 24.3% in o-3-FBS-treated conditions).
Together, our data suggest that reduction of synj1 accelerates A␤ uptake into cells and delivery to early endosomes. However, the increased A␤ uptake is independent from PIP 2 increase induced by synj1 reduction. The results represent three independent experiments, and the data are presented as means Ϯ S.D., *, p Ͻ 0.05. synj1 knockdown did not affect ␥-secretase cleavage of C100 to generate A␤ (101 Ϯ 11.1% of controls, n ϭ 6, p ϭ 0.74) or N100 to generate its cleavage product (recognized by Val 1744 antibody only; 129.9 Ϯ 27.4, n ϭ 5, p ϭ 0.11). ctrl, control.

Reduction of synj1 Increases Cellular A␤ Degradation-
The rate of A␤ turnover was next determined in N2a cells with synj1 siRNA transfection by CHX time course experiments (Fig. 7A). synj1 knockdown resulted in decreased cellular A␤ levels at time point 0 (59.7 Ϯ 13.9%; p ϭ 0.015), consistent with prior observations (Fig. 1A). At 60 min of incubation, the amount of A␤ was diminished by 37.5% with synj1 knockdown (reduced to 22.2 Ϯ 12.1% of control levels at time point 0), whereas A␤ levels in controls were only reduced by 11.5% (88.5 Ϯ 15.2% of control levels at time point 0). However, synj1 reduction does not affect APP turnover rate. As shown in Fig. 7B, the amount of holoAPP in lysates was comparable between control and synj1 siRNA-treated conditions at all time points (0 -180 min) after exposure of cells to CHX. These results in combination with previous data (Figs. 3 and  5) suggest that the reduced A␤ levels with synj1 down-regu-  NOVEMBER 1, 2013 • VOLUME 288 • NUMBER 44 lation are likely due to accelerated degradation of A␤ instead of reduced generation.

Synaptojanin 1 Reduction Accelerates A␤ Clearance
Delivery of Internalized A␤ 42 to LAMP1 ϩ Lysosomes Increases in synj1 Ϫ/Ϫ Primary Neurons-We next investigated whether A␤ trafficking through the endosomal/lysosomal pathways was increased with synj1 reduction. Immunostaining with antibodies against LAMP1 was performed in mouse cultured cortical neurons after incubation with Alexa 555 -A␤ 42 (500 nM) for 24 h. We found that A␤ 42 was colocalized with a lysosomal marker LAMP1 (Fig. 8, A and B, top panels). However, the amount of A␤ 42 colocalized with LAMP1 ϩ lysosomes was significantly increased in both cell bodies and neurites of synj1 ϩ/Ϫ and synj1 Ϫ/Ϫ neurons (Fig. 8, A and B, middle and bottom panels) compared with synj1 ϩ/ϩ cells. Upon quantification, we confirmed that the amounts of fluorescent-conjugated A␤ 42 colocalized with LAMP1 ϩ lysosomes was increased by 130.2 Ϯ 42.4% in synj1 Ϫ/Ϫ neurons if compared with synj1 ϩ/ϩ cells (p ϭ 0.016). These results suggest that internalized A␤ 42 is transported more rapidly to the endosome/lysosome degradation pathway in neurons with synj1 reduction.
Reduction of synj1 Increases A␤ Degradation in Lysosomes Partially through Elevated PIP 2 Levels-Furthermore, to assess lysosomal degradation of A␤, we tested the effects of lysosomal inhibitors (leupeptin, pepstatin A, and E-64d), which block lysosomal enzyme activities (32,33). N2a cells transfected with synj1 or control siRNA for 4 days were then incubated in the presence or absence of lysosomal inhibitors for 24 h. The amount of media and cell-associated A␤ in the presence of lysosomal inhibitors, determined by IP/Western blotting, were significantly increased (Fig. 9A, middle and right panels), indicating that lysosomal degradation of A␤ is a significant mechanism for elimination of A␤ following cellular A␤ uptake. More importantly, inhibition of A␤ degradation through lysosomal pathways abolishes the A␤-lowering effects of synj1 reduction (with lysosomal inhibitors, media A␤ 192.5 Ϯ 8.8% in controls versus 185.3 Ϯ 2.6% in synj1 siRNA; lysate A␤ 156.1 Ϯ 10.5% in controls versus 185.0 Ϯ 14.0% in synj1 siRNA).
The amount of A␤ degraded by lysosomes was determined by subtracting the mean values of A␤ in the absence of lysosomal inhibitors from the values in the presence of inhibitors. Synj1 reduction increases lysosomal degradation of cellular A␤ by 112.6 Ϯ 57.6% (p ϭ 0.021) and media A␤ by 30.6 Ϯ 17.5% (p ϭ 0.043) if compared with controls (Fig. 9B). These data suggest that the A␤-lowering effects of synj1 reduction are mainly mediated through promoting lysosomal degradation.
We next studied whether the increased A␤ degradation with synj1 reduction is dependent upon elevated PIP 2 levels, using Synj1 knockdown N2a cells treated with a PIP 2 -depleting reagent (m-3-FBS). As shown in Fig. 9C, the addition of m-3-FBS completely abolished the A␤-lowering effects induced by synj1 reduction (109.4 Ϯ 13.6% in synj1 siRNA ϩm-3-FBS versus 94.6 Ϯ 6.5% in control siRNAϩm-3-FBS). Moreover, the amount of A␤ accumulated in the presence of lysosomal inhibitors was not increased further by the addition of a PIP 2 -depleting reagent (m-3-FBS/lyso inhibitor columns, 146.9 Ϯ 11.8% in lyso inhibitors alone versus 138.5 Ϯ 12.8% in m-3-FBSϩlyso inhibitors), suggesting that the effects of PIP 2 on A␤ levels act through the same endosomal/lysosomal degradation pathway. In summary, our data suggest that reduction of synj1 increases cellular A␤ uptake, endosomal/lysosomal trafficking and degradation, partially by increasing the levels of PIP 2 .

DISCUSSION
Recent studies indicate that late onset AD is correlated with an overall impairment in A␤ clearance (3). Furthermore, several studies report pathological changes of the lysosomal network, which develop in neurons as Alzheimer disease progresses and include dysregulation of endocytosis and progressive failure of lysosomal clearance mechanisms (34 -36). A close connection between lysosomal protein clearance failure and mechanisms of neurodegeneration is also well documented (36 -39).
In this study, we defined a novel mechanism by which a PI(4,5)P 2 -degrading enzyme, synj1, regulates the cellular itinerary of A␤. We show that down-regulation of synj1 in the brain promotes A␤ uptake by neurons, accelerates A␤ delivery to the endosomal/lysosomal pathway, increases A␤ degradation, and thereby reduces amyloid plaque load and attenuates behavioral deficits in AD transgenic animals.
As proposed in our model (Fig. 10), in WT cells, extracellular A␤ is internalized into the early endosomes, and either recycled back to the plasma membrane through recycling endosomes or delivered to late endosomes/lysosomes for degradation. Under synj1 knockdown conditions, extracellular A␤ is rapidly internalized to the early endosomes and delivered to the late endosomal/lysosomal pathway for degradation. However, the amount of A␤ generated by APP processing/cleavage, as well as the rate of cellular A␤ recycling from the early endosomes to the plasma membrane, remains unchanged. As a consequence, A␤ is more rapidly degraded, whereas the rate of A␤ production is not affected. The amount of extracellular and intracellular A␤ at steady state is subsequently decreased by synj1 knockdown.
Synj1 has been linked to AD and Down's syndrome (5)(6)(7). It is a phosphatase that catalyzes the dephosphorylation of the signaling phospholipid PI(4,5)P 2 , which is shown to control clathrin-mediated endocytosis (40,41). Studies in the mouse and other model organisms have found that synj1 plays a critical role in synaptic vesicle recycling, actin regulation, and glutamate receptor trafficking (4,42). Intriguingly, our results suggest that reduction of synj1 can promote A␤ uptake and internalization through the endosomal pathway (Fig. 6), and these effects are independent upon PI(4,5)P 2 elevation (Fig. 6C). It is unclear to us whether this synj1-  . Model for synj1-regulated intracellular A␤ trafficking and lysosomal degradation. In WT cells, secreted A␤ is internalized into early endosomes, followed by either recycling back to plasma membrane through recycling endosomes or delivered to late endosomes/lysosomes for degradation. With synj1 knockdown (KD), A␤ is more rapidly internalized to the early endosomes and delivered to the late endosomal/lysosomal pathway for degradation. As a consequence, the amount of secreted and cellular A␤ is decreased with synj1 reduction. The increased A␤ uptake and internalization to early endosomes with synj1 reduction are independent upon elevated PIP 2 , whereas the effects of accelerated A␤ degradation in lysosomes by synj1 reduction are PIP 2 -dependent. enhanced internalization is specific to A␤. Further characterization in these aspects is currently ongoing.
Interestingly, others reported that haploinsufficiency of synj1 protects against A␤-induced defects in long term potentiation (7) and ameliorates learning and memory deficits in a transgenic mouse model of AD (6). It was also suggested that increased PI(4,5)P 2 caused by down-regulation of synj1 suppresses A␤ oligomer-induced neurotoxic effects (5-7). Our data suggest that the promotion of cellular A␤ degradation with synj1 reduction is partially dependent upon elevated PIP 2 because pharmacological reduction of PIP 2 abolishes the A␤-lowering effects in synj1 knockdown conditions (Fig. 9C). Together, these findings support the notion that elevation of PI(4,5)P 2 by synj1 reduction is neuroprotective against AD.
Our novel studies demonstrate that reduction of synj1 accelerates cellular A␤ uptake and delivery to lysosomes and thereby promotes A␤ degradation and reduces amyloid plaque load in an AD transgenic mouse model. It should be noted that the reduction in amyloid plaque load is most prominent in hippocampus, which could contribute to functional rescue of learning and memory deficits in our AD transgenic mouse model (Fig. 4) and is consistent with others' observations (42). The specific changes in hippocampal A␤ caused by synj1 reduction likely indicate potential sensitivity of hippocampal neurons to down-regulation of synj1. However, we did not see any significant difference in plaque load in neocortical regions, possibly because of heavy plaque burden at this stage. In contrast, we did observe a robust reduction of intracellular A␤ within both cortical and hippocampal neurons at the preplaque stage of young Swedish APP/PS1⌬E9 ϩ/Ϫ synj1 ϩ/Ϫ mice (data not shown), suggesting a potential therapeutic strategy targeting synj1 at early stages of AD.
It should be also noted that in studies by McIntire et al., no alterations in A␤ levels were evident in 6-month-old Tg2576 mice with synj1 haploinsufficiency (6), despite the fact that they previously reported that A␤ biogenesis is modulated by PI(4,5)P 2 (11). They measured the whole brain lysates at 6 months of age, whereas we specifically quantified A␤ and amyloid plaque load from hippocampus in older mice (9 -10 months of age). Moreover, our data suggest that the A␤-lowering effects induced by synj1 down-regulation are mainly dependent on the promotion of cellular A␤ degradation without affecting A␤ generation through cleavage of APP by ␤and ␥-secretase. There was no change in levels of ␥-secretase/PS1 or BACE1 total proteins in APP/PS1 ϩ/Ϫ synj1 ϩ/Ϫ mouse brains (Fig. 3, A and B). Furthermore, utilizing a well established in vitro ␥-secretase assay (25,26), we have found that synj1 knockdown did not affect ␥-secretase cleavage of C100 to generate A␤ or N100 to generate its cleavage product (Fig. 5).
On the other hand, we observed that synj1 reduction promotes ␣-secretase cleavage of APP and up-regulates expression levels of ␣-secretase ADAM10. 4 Promoting APP processing through the nonamyloidogenic pathway by synj1 reduction can further strengthen rescue of AD related changes (43,44). For example, the ␣-secretase-derived soluble APP N-terminal fragment, sAPP␣ has been suggested to have neurotrophic and neuroprotective functions, further supporting the therapeutic value of reducing synj1 levels in the brain (45). Although ␣-secretase competes with BACE for APP cleavage and has the capacity to pre-empt A␤ generation, it is unlikely that this is the main mechanism underlying synj1 knockdown-induced A␤-lowering effects. Our data (Figs. 1 and 3) show that the levels of ␤CTF, the immediate precursor of A␤, are not affected by synj1 reduction, suggesting that the BACE cleavage of APP is not affected under our conditions. More importantly, increased synj1 expression has been functionally linked to the enlargement of early endosomes (46). It has been reported that endosome anomalies are the earliest specific pathology reported in AD brain tissue (47,48). We here show that synj1 regulates endosomal trafficking and lysosomal degradation of A␤. Reduction of synj1 can promote intracellular A␤ degradation through the endosomal/lysosomal pathway. Therefore, it would be interesting to investigate whether synj1 reduction can reverse early AD pathological changes such as enlargement of early endosomes.
In summary, our studies demonstrate a novel mechanism by which a PI(4,5)P 2 -degrading enzyme, synj1, regulates the cellular itinerary of A␤. We show for the first time that endosomal/ lysosomal degradation of A␤ can be modulated by phosphoinositol homeostasis. Importantly, endosomal anomalies are considered as one of the earliest AD pathologies, and increased function of synj1 is linked to enlargement of early endosomes (46 -48). Thus, our findings uncover a new therapeutic direction for AD aiming at modulation of cellular A␤ clearance by a phosphoinositol regulator, in the hopes of reversing amyloidinduced pathological changes and cognitive deficits.