|Home | About | Journals | Submit | Contact Us | Français|
The accumulation of the β-amyloid peptide (Aβ) in Alzheimer’s disease (AD) is thought to play a causative role in triggering synaptic dysfunction in neurons leading to their eventual demise through apoptosis. Aβ is produced and secreted upon sequential cleavage of the amyloid precursor protein (APP) by β- and γ-secretases. However, while Aβ levels have been shown to be increased in AD patients’ brains, little is known about how the cleavage of APP and the subsequent generation of Aβ is influenced, or if the cleavage process changes over time. It has been proposed that Aβ can bind APP and promote amyloidogenic processing of APP, further enhancing Aβ production. This idea has remained controversial due to a lack of a clear mechanism and complicated by the promiscuous nature of Aβ binding. To work around this problem, we used an antibody-mediated approach to bind and cross-link cell surface APP in cultured rat primary hippocampal neurons. Here we show that cross-linking of APP is sufficient to raise the levels of Aβ in viable neurons with a concomitant increase in the levels of the β-secretase BACE1. This appears to occur as a result of a sorting defect, due to the caspase-3-mediated inactivation of a key sorting adaptor protein, namely GGA3, which prevents the lysosomal degradation of BACE1. Taken together, our data suggest the occurrence of a positive pathogenic feedback loop involving Aβ and APP in affected neurons possibly allowing Aβ to spread to nearby healthy neurons.
The progressive accumulation and deposition of the β-amyloid peptide (Aβ) leading to senile plaques (SP) is an invariant feature of Alzheimer’s disease. Processing of the amyloid precursor protein (APP; Kang et al., 1987) by β- and γ-secretases leading to the production of Aβ is a well-characterized process (reviewed by Walsh et al., 2007). Although APP is conventionally thought to exist and function as a monomer, biochemical and structural data suggest that APP may exist as functional dimeric or oligomeric complexes (Rossjohn et al., 1999; Scheuermann et al., 2001; Wang, 2004; Chen et al., 2006).
APP contains three GxxxG motifs at the junction between the juxtamembrane and transmembrane (TMS) sequences (Liu et al., 2005; Marchesi, 2005; Sato et al., 2006). These motifs mediate sequence-specific dimerization between transmembrane helices by direct glycine-glycine contacts (Bormann et al., 1989; Lemmon et al., 1994) facilitating oligomerization of many transmembrane proteins (Russ, 2000), including APP (Munter et al., 2007; Kienlen-Campard et al., 2008). APP homo- and heterotypic interactions have traditionally been implicated in cellular adhesion (Soba et al., 2005), but APP multimerization has also been proposed to play a factor in the pathogenesis of Alzheimer’s disease (AD) by driving the overproduction of the Aβ (Scheuermann et al., 2001; Kaden et al., 2008). Interestingly, one of the identified mutations (Flemish) in early-onset AD involves an alanine to glycine substitution, creating a fourth GxxxG motif in APP (Hendriks et al., 1992). Additionally, inducing dimerization of APP by introducing cysteine residues within the Aβ domain of APP (K28C and L17C) have been found to dramatically increase Aβ production (Scheuermann et al. 2001). Conversely, disrupting dimerization either by mutating a glycine in the GxxxG motifs to an alanine (Munter et al., 2010) or by chemical means (Richter et al., 2010) has been shown to attenuate Aβ production. However, It should be noted that others have reported the opposite results, either with the K28C mutation (Ren et al., 2007) or by using a chimeric APP construct (Eggert et al. 2009).
These conflicting studies led us to revisit the issue in a native system, using endogenous APP in cultured rat hippocampal neurons (PHNs). Since a definitive ligand for APP has not been identified, we used a commercially available monoclonal antibody (22C11) raised against the extracellular domain (amino acids 66–81) of APP (Hilbich et al., 1993) to cross-link cell-surface APP and mimic Aβ binding of APP. Here, we show that multimerization of APP triggers the amyloidogenic pathway in viable neurons, resulting in increased Aβ production with a concomitant increase in BACE1 levels. This appears to be due to decreased lysosomal degradation of BACE1 due to a sorting defect, which allows BACE1 to accumulate in recycling endosomes. Our data suggest that aberrant APP signaling in affected neurons may play an important role in not only triggering but propagating the amyloidogenic pathway to other neurons as well.
All chemicals used were of the highest grade available. Poly-D-lysine, chloroquine, ammonium chloride (NH4Cl) and Drebrin A rabbit antibody (rAb) were purchased from Sigma; APP N-terminal mouse antibody (mAb) 22C11, Aβ mAb 4G8, APP C-terminal mAb, sAPPβ mAb, neprilysin rAb from Millipore Corporation; PSD-95 rAb, GGA3 rAb, caspase-3 rAb, cleaved-caspase-3 rAb, anti-myc-tag rAb from Cell Signaling Technology; BACE1 rAb, IDE rAb, EEA1 mAb, GAPDH mAb from Abcam; ERK1 rAb, Aph-1 rAb from Santa Cruz Biotechnology; dendra2 rAb from Evrogen; PS1 mAb from Abgent; βIII-tubulin mAb TUJ1 from Covance.
Hippocampal neuron cultures from both male and female rats pups were prepared following a slightly modified version of the method of Brewer (Brewer et al., 1993). Hippocampal neurons were kept in culture at 37°C with 5% CO2 in Neurobasal® medium (Invitrogen) with B27 supplement and Glutamax™ (Invitrogen) and plated at a density of 2.5×105 cell/ml on dishes coated with poly-L-lysine (Sigma). For our experiments, neurons were used after approximately 14–21 days in vitro (DIV). Human embryonic kidney 293 (HEK 293) and murine neuroblastoma B103 cells were cultured and maintained in Dulbecco’s Modified Eagle Medium (DMEM; Invitrogen) supplemented with 10% fetal bovine serum (FBS; Invitrogen) and 10% FBS + 5% horse serum (HS; Invitrogen) respectively. For stably transfected B103 cells expressing APP, the media was supplemented with 50 μg/ml Geneticin (Invitrogen).
This protocol was adapted from Gan and Grutzendler (Gan et al., 2000). Briefly, 100 mg of tungsten particles (1.1 μm diameter; Bio-Rad) were thoroughly precipitated with 5.0 mg of lipophilic dye (DiI; Invitrogen) and dissolved in 100 μl of methylene chloride. Dye-coated particles were shot 2 to 3 times into the cells using the Helios gene gun system (BioRad) and left in 0.1 M PBS overnight to allow dye diffusion along neuronal processes. Cells were post fixed with 4% paraformaldehyde (PFA; Pierce) for 1 hour to preserve staining then mounted onto glass slides using Gel Mount™ (Biomeda Corporation) and stored at 4°C in the dark.
Labeled neurons were imaged using the Laser Scanning Microscope (LSM) 510 Meta confocal microscope (Zeiss), equipped with 40X 1.3 NA and 100X 1.4 NA oil immersion objectives. The NIH image software program Image J was used the quantify DiO labeled cultured neurons. An average of 12 compressed images (20 μm thick) consisting of pyramidal neurons in the hippocampus were quantified for each treatment. Statistical analysis was performed using the statistics package GraphPad Prism®.
Cell death was assayed using the LIVE/DEAD® Cell Viability Assay kit (Invitrogen) following the manufacturer’s protocol. After the appropriate treatments, cells were treated with a solution containing 2.0 μM calcein AM and 4.0 μM ethidium homodimer 1 (EthD-1) in PBS for 30 minutes. Live cells are distinguished by the presence of ubiquitous intracellular esterase activity, determined by the enzymatic conversion of the virtually non-fluorescent cell-permeant calcein AM to the intensely fluorescent calcein.
Oligonucleotide sequences were custom ordered (Dharmacon) with a thiol functionality at the 5′-end. Caspase-3 siRNA sequence: 5′ Th-AGCCGAAACUCUUCAUCAUUU. Oligonucleotide stocks were solubilized in water at a 10 mM concentration. The delivery peptide Penetratin-1 (MP Biomedicals) was cross-linked via a Cys-Cys bond to the desired oligonucleotide as previously described (Davidson et al., 2004).
γ-secretase activity in hippocampal neurons was detected as AICD-myc formation using as substrate a C100-myc construct from HEK293 cells transfected with a C100-myc plasmid. CHAPS-solubilized (20 mM HEPES, pH 7.0; 150 mM KCl; 2.0 mM EGTA; 1.0% CHAPS and 2x protease inhibitor cocktail) HEK293 cells were incubated together with solubilized 22C11-treated and untreated hippocampal neuron membranes (1:1 ratio) at −20°C or 37°C, with or without the γ-secretase inhibitor (DAPT; 1.0 μM) for 2 hours. AICD fragments were detected by Western blot analysis using an anti-myc rAb (Hansson et al., 2006).
Caspase-3 activity in hippocampal neurons was assessed with the NucView™ 488 Caspase Detection kit (Biotium Inc.) following the manufacturer’s protocol. Briefly, hippocampal neurons cultured in chamber slides were treated with a 5.0 μM solution of the NucView caspase-3 substrate for 30 minutes. Cells were then fixed with 4.0% paraformaldehyde (Pierce), mounted and observed under a fluorescence microscope using a fluorescein isothiocyanate (FITC) filter.
Hippocampal neurons grown in chamber slides were treated with 22C11 (100 ng/ml) only or 22C11 (100 ng/ml) and Penetratin-1-linked-siCasp3 (80 nM) for 8 hours. After fixing with 4.0% paraformaldehyde (Pierce) cells were permeabilized with 0.5% Triton® X-100 (Sigma) in PBS and incubated the anti-BACE1 rAb and anti-EEA1 mAb overnight. Anti-mouse and anti-rabbit secondary antibodies conjugated with Alexa® Fluor dyes (Invitrogen) were used. Labeled neurons were imaged using the Laser Scanning Microscope (LSM) 510 Meta confocal microscope (Zeiss), equipped with 40X 1.3 NA and a 100X 1.4 NA oil immersion objectives. Scanning used two excitation lines of the argon laser (488 nm for Alexa Fluor-488, 568 nm for Alexa Fluor-568). Z stacks were collected at 1.0 μm intervals and then compressed into a single image for analysis using the Volocity® Imaging software (Volocity® Acquisition and Volocity® Visualization; Perkin-Elmer).
Prior studies using antibodies to cross-link APP have shown a significant amount of apoptotic death in cells exposed to high antibody concentrations (Rohn et al., 2000; Sudo et al., 2000). While neuronal death is invariably an important feature of Alzheimer’s disease, synaptic dysfunction is thought to be the earliest event in the progression of the disease (Delaere et al., 1989; Terry et al., 1991). We thus wanted to determine a concentration of the antibody sufficient to cause synaptic dysfunction, as measured by the loss of dendritic spines, without affecting their viability. To that end, we treated cultured rat primary hippocampal neurons (PHNs) with various concentrations of the monoclonal antibody, 22C11 (100 pg/ml, 100 ng/ml and 1.0 μg/ml), and performed a direct count for cell viability after 24 hours. We found that 22C11 triggered a significant amount of neuronal death (~20% cell death) at 1.0 μg/ml, but not at lower concentrations of 100 ng/ml and 100 pg/ml (Figure 1a). We next investigated the effects of lower concentrations of 22C11 on synaptic function by once again treating PHNs with 100 ng/ml of 22C11 for 24 hours. A direct count of the number of dendritic spines in diOlistically labeled neurons (Figure 1b) revealed a decrease in density, by as much as ~40%, in 22C11-treated neurons compared with control neurons treated with PBS (Figure 1c). The decline in dendritic spine density was accompanied by a corresponding reduction in the protein levels of the dendritic spine markers PSD-95 (Figure 1d,e) and Drebrin A (Figure 1d,f) after the 24-hour period. As a control for cross-linking, PHNs were treated with isolated monovalent Fab fragments of the APP antibody, which had no effect on spine density (Figure 1c), suggesting that cross-linking of APP is required for its effects in neurons.
We next examined whether cross-linking of APP with sub-apoptotic levels of 22C11 increased the levels of Aβ in neurons. PHNs were treated with 100 ng/ml of 22C11 for 24 hours and harvested to measure the levels of neuronal Aβ by immunoprecipitation (IP) with a monoclonal antibody raised against the Aβ domain of APP, 4G8 (Figure 2a). We detected a significant increase in Aβ production (~2-fold) in 22C11-treated neurons compared to PBS-treated control neurons (Figure 2b). There were no significant changes in the levels of full-length APP, either in 22C11 or in PBS-treated control neurons (Figure 2a), suggesting that 22C11 treatment increases Aβ levels by directly promoting its production or by affecting its clearance. In a second approach, the levels of Aβ in lysates were measured using a commercially available enzyme-linked immunosorbent assay (ELISA) kit, which measures the levels of Aβ(x-42). Consistent with our immunoprecipitation results, we found a significant increase in the levels of intraneuronal Aβ(x-42) in the 22C11-treated cells compared to PBS-treated control cells (Figure 2c). Interestingly, we did not detect a significant change in the levels of secreted Aβ in the culture medium of either 22C11-treated neurons nor PBS-treated control cells, as measured by ELISA or immunoprecipitation (data not shown), suggesting that Aβ is accumulating intracellularly as a result of 22C11 treatment.
Our next objective was to determine the mechanism by which Aβ levels were increased in 22C11-treated neurons. Since the levels of APP remained unchanged in treated cells, we hypothesized that the changes in Aβ levels were due to a perturbation in the equilibrium between its clearance and production rates. Several studies have implicated the metalloproteases neprilysin and IDE in the degradation of Aβ (reviewed in Miners et al., 2008). The levels of these two proteins have also shown to be reduced in several AD mouse models and in AD patients (Kurochkin and Goto, 1994; Yasojima et al., 2001; Apelt et al., 2003; Cook et al., 2003). However, we did not did not detect any significant changes in the protein levels of either IDE or neprilysin (Data not shown) in 22C11-treated neurons after 8 hours, indicating that increased β- and/or γ-secretase processing of APP was the likely cause of the elevated Aβ levels.
We used two different approaches to evaluate γ-secretase processing of APP. Since active endogenous presenilin (PS) exists mainly as a heterodimeric complex formed from the endoproteolytically processed N- and C-terminal fragments of PS (Thinakaran et al., 1996), we first measured the levels of N-terminal fragments (~28 kDa) by Western blot using a commercially available N-terminal antibody for presenilin-1. There was no statistical difference in the levels of cleaved PS1 N-terminal fragments between neurons treated with 22C11 and those treated with PBS (Figure 3a). To assess whether 22C11 could enhance the activity of γ-secretase, we used a previously described cell-free assay (Sastre et al., 2001; Hansson et al., 2006) designed to directly measure cleavage of CTFβ (C100) by γ-secretase (Figure 3b). Briefly, HEK293 cells were transiently transfected with a C100-myc construct and membrane fractions were prepared from these cells 48 hours post-transfection. These membrane fractions, enriched with APP C100-myc fragments (data not shown), were combined with membrane extracts (γ-secretase complex source) from 22C11-treated and PBS-treated PHNs and incubated for 2 hours at 37°C to allow γ-secretase cleavage of C100-myc. The AICD-myc fragments generated were then detected by Western blot using a commercially available polyclonal anti-myc antibody (Figure 3c). We did not observe a significant difference across the samples (Figure 3d). γ-Secretase-mediated cleavage of C100-myc was confirmed by the addition of the γ-secretase inhibitor DAPT (Dovey et al., 2001), which completely abolished generation of AICD-myc fragments (Figure 3c).
To assess β-secretase cleavage of APP, we measured the total levels of cellular BACE1 protein by Western blot in PHNs after an 8-hour treatment with 100 ng/ml of 22C11 (Figure 3e). BACE1 protein levels were significantly increased in 22C11-treated PHNs compared with PBS-treated control PHNs (Figure 3f). To correlate increased BACE1 levels with increased activity, we also measured the levels of sAPPβ in the culture medium by Western blot (Figure 3e). Concomitant with the increase in BACE1, sAPPβ levels were significantly higher in the culture medium from 22C11-treated PHNs compared to that of control PHNs (Figure 3g), consistent with increased activity of BACE1 and increased processing of APP.
Taken together, these results suggest that the observed increases in the levels of intracellular Aβ in 22C11-treated PHNs is unlikely to be caused by a deficiency in the clearance mechanism of Aβ. This is supported by the unchanged levels of IDE and neprilysin in 22C11-treated neurons. Instead, our results point towards increased processing of APP by BACE1 at the β-secretase cleavage site of APP as the main element driving the elevated Aβ levels.
Having demonstrated elevated levels of BACE1 in 22C11-treated PHNs, we turned our focus on determining the cause. We first investigated whether 22C11 activated transcription of BACE1 in hippocampal neurons. qRT-PCR analysis of BACE1 mRNA levels in 22C11 treated-neurons did not reveal any significant changes over control neurons after 8 hours (data not shown), ruling out the possibility that increased transcription was the likely cause of increased BACE1 levels. This is in line with several findings that failed to show a significant increase in BACE1 mRNA levels in AD brains and in AD mouse models, in spite of elevated BACE1 protein levels (Holsinger et al., 2002; Zhao et al., 2007; Hébert et al., 2008).
We therefore reasoned that the increase in BACE1 protein levels was due to increased stability of the protein. To test this hypothesis, we performed a cycloheximide degradation assay to directly examine changes in the turnover rate of BACE1. PHNs were treated either with cycloheximide (CHX, 150 μM) alone, or with both CHX and 22C11 (100 ng/ml) over the course of 18 hours and BACE1 protein levels were evaluated by Western blot (Figure 4a). When we compared the rate of degradation of BACE1, we found a significant delay in its clearance in 22C11-treated PHNs compared to PHN treated with CHX alone (Figure 4b), consistent with the idea of increased protein stability. The apparent increase in BACE1 in CHX/22C11-treated PHNs was likely due to the fact that all the samples were normalized for equal amounts of protein. Since BACE1 is stabilized following 22C11 treatment while most other proteins continue to be degraded at their normal rates, the levels of BACE1 in relation to the total amount of proteins are seemingly elevated. We also compared the turnover rate of APP between CHX alone and CHX/22C11-treated PHNs (Figure 4c) and found a significant decrease in the levels of APP, suggesting that APP was processed at a faster rate in 22C11-treated PHNs (Figure 4d). APP levels were decreased by about 50% and 31% after 6 and 18 hours respectively in PHNs treated with 22C11 (Figure 4e).
These results are consistent with our initial hypothesis that 22C11 treatment affects the turnover rate of BACE1, increasing the half-life of endogenous BACE1 protein and argues against the possibility that increased synthesis is the cause of elevated BACE1 levels.
Collectively, the above data suggested that 22C11 drove Aβ production by altering the steady-state levels of BACE1, the rate-limiting enzyme in the biogenesis of Aβ. It also indicated that the increase in BACE1 levels might be, at least in part, the result of increased stabilization of BACE1. At least three different mechanisms have been reported to control the degradation of BACE1: (1) endoproteolysis within its catalytic domain (Huse et al., 2003), (2) the lysosomal pathway (Koh et al., 2005) and (3) the ubiquitin-proteasomal pathway (Qing et al., 2004). The latter study by Qing and colleagues was largely based on over-expression of BACE1 in several cell lines and it remains unclear whether endogenous BACE1 would normally be degraded through the proteasomal pathway. In addition, since the effects of lysosomal inhibitors were not studied, any potential contribution from lysosomal degradation of BACE1 in unclear. Moreover, the interpretation of experiments involving proteasome inhibitors can be problematic since many proteasome inhibitors can also inhibit lysosomal cathepsins (Kisselev, 2001; Kozlowski et al., 2001).
To overcome these shortcomings, we evaluated the respective contributions of proteasomal and lysosomal degradative pathways in regulating the levels of endogenous BACE1 in our model system, namely cultured rat primary hippocampal neurons (PHNs). To assess lysosomal degradation of endogenous BACE1, PHNs were treated for 8 hours with ammonium chloride (NH4Cl; 500 μM) and chloroquine (100 μM), weak bases known to inhibit lysosomal hydrolases by reducing the acidification of the endosomal/lysosomal compartments (Ohkuma, 1978). After treatments, neurons were harvested and the levels of BACE1 protein were evaluated by Western blot analysis (Figure 5a). Both chloroquine and NH4Cl treatments induced a marked increase in endogenous BACE1 protein levels after 8 hours (Figure 5b) compared to control PHNs. Increasing the concentrations of each inhibitor did not appear to significantly affect the build-up of BACE1. However, longer treatments did lead to higher BACE1 levels (data not shown). These results were in line with previous studies that demonstrated a similar effect in several cell lines, including CHO and SY5Y cells and in cortical neurons (Koh et al., 2005). To assess proteasomal degradation, a similar protocol was followed and PHNs were treated with the proteasomal inhibitor MG132 (1.0 μM) for 8 hours. Western blot analysis (Figure 5c) showed a slight increase in BACE1 levels after 8 hours (Figure 5d) that did not reach statistical significance. Increasing the concentration of MG132, or extending the treatment over a longer time frame proved problematic due to toxicity. Comparable results were also obtained with another proteasomal inhibitor, lactacystin (data not shown).
The results above indicated that endogenous BACE1 degradation in PHNs was primarily regulated by the activity of lysosomal hydrolases. We hypothesized that 22C11 treatment was interfering with lysosomal degradation of BACE1. The dileucine (DXXLL) motif of BACE1 is believed to play a critical role in its degradation, since mutating the leucine residues to alanine residues increases the levels of BACE1 protein (Pastorino et al., 2002). Mutagenesis of the dileucine motif has been shown to interfere with the lysosomal degradation of BACE1 by blocking its transport from endosomes to lysosomes (Koh et al., 2005). This trafficking pathway is mediated by the Vps-27, Hrs and STAM (VHS) domain of the Golgi-localized gamma-ear-containing ARF-binding (GGA) proteins (GGA1, 2, and 3), which binds the dileucine motif of BACE1 (reviewed in Bonifacino, 2004). Depletion of GGA proteins by RNAi increases accumulation of BACE1 in early endosomes (He et al., 2005; Wahle et al., 2005; Tesco et al., 2007) and increases Aβ secretion in neurons (Wahle et al., 2006). Alternatively, over-expression of GGA proteins reduces CTFβ generation and Aβ production (Wahle et al., 2006). Interestingly, of the three GGA family member proteins, GGA3 was highlighted in a recent report by Tesco et al., for its potential role in AD pathology. GGA3 protein levels were found to be significantly decreased in AD brains and inversely correlated with increased levels of BACE1 (Tesco et al., 2007).
Since our data suggested that BACE1 was predominantly degraded through the lysosomal pathway in PHNs, we investigated the fate of GGA3 in response to 22C11 treatment. Tesco et al. had demonstrated that GGA3 was a substrate for caspase-3 (Tesco et al., 2007), which led us to determine whether treatment with 22C11 could lead to the activation of caspase-3 in PHNs. To that end, we analyzed the levels of cleaved caspase-3 in PHNs treated with sub-lethal levels of 22C11 (100 ng/ml) for 1 and 8 hours by Western blot, using an antibody, which recognizes endogenous levels of the large fragment (17/19 kDa) of cleaved caspase-3 (Figure 6a). We found a significant increase in cleaved caspase-3 levels after 8 hours in PHNs treated 22C11 compared to PBS-treated PHNs (Figure 6b).
It is often assumed that increased caspase-3 cleavage equates to increased caspase-3 activity. However, IAPs such as xIAP, c-IAP1/2 and Survivin can bind processed caspase-3 and block its activity (reviewed in Liston et al., 2003). To address this issue, we directly measured the activity of caspase-3 in live PHNs using a commercially available fluorescent caspase-3 substrate, DEVD-NucView. The DEVD peptide is attached to a DNA-binding dye, which is unable to produce fluorescence in the absence of DNA. Upon entering the cell cytoplasm, it is cleaved by active caspase-3 to release the high-affinity DNA dye. The released dye migrates to the nucleus and brightly stains it. Consistent with the observed increased in cleaved caspase-3 levels, we found that PHNs treated with 22C11 had notably higher fluorescence compared to control PHNs (Figure 6c), indicative of increased DEVD cleavage by active caspase-3. Fluorescence was significantly attenuated by pre-treating PHNs for 1 hour with the competing non-fluorescent caspase-3 inhibitor Z-DEVD-FMK (25 μM; Figure 6c).
Since caspase-3 activity was increased in 22C11-treated PHNs, we decided to look at the cleavage of endogenous GGA3 protein in these cells. As described earlier, PHNs were treated with of 22C11 (100 ng/ml) for 8 hours and cleavage of endogenous GGA3 was assessed by Western blot. Cleavage of recombinant human GGA3 by recombinant active caspase-3 occurs at three major sites within its hinge domain, generating three C-terminal fragments of ~50 kDa, ~48 kDa and ~37 kDa, as well as an N-terminal fragment, which also function as a dominant negative form of GGA3 (Tesco et al., 2007). We used a commercially available antibody directed against the C-terminus of GGA3, which recognizes the full-length protein as well as C-terminal fragments (Figure 6f). PHNs treated with 22C11 had increased levels of cleaved GGA3 (~37 kDa fragment) compared to control PHNs (Figure 6g). We did not detect the larger ~50 kDa fragment of GGA3 possibly because the levels were below the limit of detection. This was consistent with the findings of Tesco and colleagues, who were unsuccessful at detecting endogenous levels of the larger fragment (~50 kDa) in lysates from the H4 human glioblastoma cell line (Tesco et al., 2007). Interestingly, no significant change in the levels of the ~48 kDa fragment was observed in 22C11-treated PHNs, perhaps due to a preference for the third cleavage site to produce the smaller fragment.
To confirm the role of caspase-3 in the cleavage of GGA3 PHNs, we used a siRNA to deplete the levels of endogenous caspase-3 in PHNs prior to treating with 22C11. The siRNA oligonucleotide was delivered to PHNs by linking it to the cell-penetrating peptide Penetratin-1, previously used as an effective method for delivering siRNA sequences to primary neurons (Davidson et al., 2004). We achieved ~40% knockdown of endogenous caspase-3 protein in hippocampal neurons after 8 hours (Figure 6d,e), which was sufficient to significantly abrogate the cleavage of GGA3 in 22C11-treated PHNs, as demonstrated by the reduced levels of the ~37 kDa fragment (Figure 6g).
The results above suggested a mechanism whereby 22C11 increased caspase-3 activity in PHNs, resulting in the cleavage and blocking of GGA3 function. We predicted that depletion of caspase-3 by siRNA as shown should reverse the effects of 22C11 on both BACE1 and Aβ levels. To test this, we treated PHNs with 22C11 (100 ng/ml), in the presence or absence of the caspase-3 siRNA and analyzed the levels of BACE1 by Western blot and Aβ by IP (Figure 7a). The increases in BACE1 and Aβ levels in 22C11-treated PHNs were completely blocked in PHNs treated with siCasp3 (Figure 7b,c).
Our model also predicted that flooding the neurons with GGA3 by over-expression should block the increase in BACE1 and Aβ production triggered by 22C11. Because of the low transfection efficiency in PHNs, we used the B103 murine neuroblastoma cell line (Schubert et al., 1974), stably expressing APP. These cells were treated with 500 ng/ml of 22C11 for 8 hours and the levels of neuronal Aβ and BACE1 were measured (Figure 7d). Similar to PHNs, B103 cells treated with 22C11 showed a significant increase in both BACE1 levels and Aβ production (Figure 7e,f). To test our prediction, B103 cells were transfected with a plasmid expressing full-length human GGA3 and treated after 48 hours with 22C11 as above. Consistent with our proposed model, B103 cells over-expressing GGA3 did not exhibit an increase in BACE1 levels or Aβ production in response to 22C11 treatment (Figure 7e,f).
Taken collectively, these results provide consistent support for a mechanism in which 22C11 promotes BACE1 accumulation in hippocampal neurons by interfering with its normal degradation process. Activation of caspase-3 by 22C11 leads to increased cleavage and inactivation of GGA3, delaying the degradation of BACE1 and allowing it to accumulate.
As mentioned above, depletion of GGA3 by siRNA impairs its normal function of targeting BACE1 to lysosomes for degradation, leading to the accumulation of BACE1 in early endosomes (He et al., 2005; Wahle et al., 2005; Tesco et al., 2007). Since 22C11 treatment promoted GGA3 cleavage, we further reasoned that 22C11 should have a similar effect, namely to increase BACE1 in the endosomal compartments of treated PHNs. To test this final prediction, PHNs were again treated with 22C11 for 8 hours after which they were fixed with paraformaldehyde and co-stained with antibodies against the early endosome marker EEA1 (Figure 8b,f,j) and endogenous BACE1 (Figure 8a,e,i). Confocal analysis of the obtained images (Figure 8c,g,k) revealed a marked increase in the colocalization of BACE1 and EEA1 in 22C11-treated PHNs compared to PBS-treated PHNs (Figure 8d,h,m), indicative of BACE1 accumulation in early endosomal compartments, in agreement with our proposed model.
We also tested whether caspase-3 downregulation could reverse the 22C11-induced accumulation of BACE1 in endosomes. Indeed, consistent with our model, downregulation of caspase-3 in 22C11-treated PHNs resulted in a notable reduction in the colocalization of BACE1 with EEA1 (Figure 8h,l,m), suggesting that BACE1 is prevented from accumulating in early endosomes.
Much progress has been made towards deciphering some of the biological functions of APP, particularly its role in facilitating cell-cell adhesion through homo- and hetero-dimerization with the other APP family members, namely APLP1 and APLP2. However, the focus has remained mostly on its pathogenic role, as the source of the Aβ peptide in the context of Alzheimer’s disease. The identification and characterization of APP-cleaving enzymes, such as secretases (BACE1 and the γ-secretase complex) and caspases, have provided a valuable insight into the complex steps involved in the proteolytic processing of APP and the production of Aβ. Whether the Aβ peptide itself is the cause of AD remains a subject for debate. The identification of several mutations in the APP and presenilin (PS) genes in early-onset familial AD seem to support this view, as these mutations have been shown promote the production of Aβ. In late-onset AD however, the picture becomes a little more muddled. While many risk factor genes have been identified, age remains the greatest risk factor for AD. This raises several important questions: Does the cleavage process of APP change with aging or in AD, and if so what triggers these changes? Do elevated BACE1 levels in the brain play a role in AD pathogenesis, and if so what causes BACE1 to become elevated? Could dimerization of APP provide some answers?
The exact consequences of homo-dimerization of APP on the processing of APP are not fully understood. Introduction of a cysteine mutation in the juxtamembrane (JM) region of APP has been reported to enhance Aβ production through the formation of stable disulfide-linked APP dimers (Scheuermann et al., 2001), consistent with the observation that stable Aβ dimers can be found intracellularly in vitro and in vivo in brains (Walsh et al., 2000). However, other laboratories have reported the opposite effect, where enhanced dimerization of APP leads to decreased APP processing and decreased Aβ levels (Struhl, 2000; Eggert et al., 2009). Reconciling this dichotomy remains difficult, but it could simply be explained by differences in the manner through which APP dimerization is promoted in each model systems.
Here we show that cross-linking of endogenous APP through the use of a divalent antibody triggers the amyloidogenic pathway in cultured hippocampal neurons, resulting in a rise in the levels of intracellular Aβ. Interestingly, this increase in Aβ is observed under non-apoptotic conditions, at least within 48 hours of treatment, but is accompanied by a significant loss of dendritic spine protrusions as well as a decrease in synaptic markers, PSD-95 and Drebrin A. This is an important consideration since neuronal and non-neuronal cells undergoing apoptosis have been shown to overproduce and secrete Aβ, whether triggered by staurosporine or by trophic factor withdrawal (LeBlanc, 1995; Barnes et al., 1998; Galli et al., 1998; Gervais et al., 1999; Guo et al., 2001; Tesco et al., 2003; Sodhi et al., 2004; Matrone et al., 2008a, Matrone et al., 2008b). This wide array of conditions and cell types raises questions as to whether the observed Aβ overproduction is a specific process or simply a general response to apoptotic stimuli. Our result supports the former view, but it is possible that both APP signaling and apoptotic stimuli share common pathways.
Whether driven by apoptosis or by APP signaling, enhanced Aβ production seems to correlate with high levels of BACE1. We show that the observed Aβ overproduction in treated neurons is due primarily to increased processing of APP by BACE1, and not by γ-secretase. This is reflected by a significant increase in the production of sAPPβ fragments correlating with elevated BACE1 protein levels. BACE1 levels rise in response to physiological stress or injury, such as oxidative stress (Tamagno et al., 2002), traumatic brain injury (Blasko et al., 2004), ischemia (Wen et al., 2004), hypoxia (Zhang et al., 2007), and energy impairment (Velliquette et al., 2005). BACE1 is also increased in brains from late-onset and early-onset AD patients compared to cognitively normal individuals (Fukumoto et al., 2002; Holsinger et al., 2002; Tyler et al., 2002; Yang et al., 2003; Li et al., 2004). Our results imply that in addition to age-related stress, aberrant signaling triggered by APP oligomerization may also enhance levels of BACE1 and Aβ in the brain, and drive AD pathogenesis.
The exact mechanism of this up-regulation is not fully understood and hypotheses vary from transcriptional, post-transcriptional, translational and post-translational modifications of BACE1 (Holsinger et al., 2002; Zhao et al., 2007; Faghihi et al., 2008; Hébert et al., 2008; Wen et al., 2008). Our results indicate that the BACE1 increase may be due, at least in part, to enhanced protein stabilization and accumulation. This is reflected by the fact that no significant changes are observed in BACE1 mRNA in treated hippocampal neurons. Alternatively, BACE1 half-life is significantly prolonged in treated neurons, persisting to near control levels after 18 hours of cycloheximide treatment. These results suggest that oligomerization of APP may trigger a signaling cascade that directly interferes with the normal degradation of BACE1 protein, allowing BACE1 to accumulate in treated neurons. Our data shows that activation of this pathway may lead to the loss of function of the GGA3 protein. GGA family proteins are known to be involved in the trafficking of proteins, such as BACE1, which contain the DXXLL signal between different compartments, e.g. Golgi complex, endosomes and lysosomes (reviewed in Bonifacino, 2004). We demonstrated that activation of caspase-3 in treated neurons promotes cleavage of GGA3, perhaps generating increased amounts of the dominant negative N-terminal fragment (Tesco et al., 2007). The accumulation of BACE1 in early endosomal compartments in treated neurons was consistent with a loss of function of the GGA3 protein. Furthermore, downregulation of caspase-3 by siRNA prevented the accumulation of BACE1 in endosomes as well as an increase in intracellular Aβ. A similar effect was observed in cells over-expressing human GGA3, which did not exhibit an accumulation of BACE1 or overproduction of Aβ.
The mechanism by which GGA3 targets its cargo to lysosomes has been shown to be ubiquitin-dependent (Puertollano, 2004). While there is some evidence that BACE1 is ubiquitinated (Qing et al., 2004), future studies will be required to determine whether GGA3-dependent degradation of BACE1 requires ubiquitination or whether it occurs via an alternate mechanism (e.g. binding the VHS domain of GGA3). Additionally, RNAi silencing of GGA1 and GGA2 has also been shown to lead to the accumulation of BACE1 in endosomes. However, unlike GGA3, which shuttles BACE1 from endosomes to lysosomes, GGA1 and GGA2 appear to regulate retrograde transport of BACE1 from endosomes to the TGN (Wahle et al., 2005). Further studies will be required to determine whether loss of function or depletion of GGA1 and GGA2 contribute to BACE1 accumulation in our model. Finally, phosphorylation of BACE1 at Serine498 facilitates its binding to GGA proteins (He et al., 2002; He et al., 2003; Shiba et al., 2004; von Arnim et al., 2004). However, we did not observe any changes in the phosphorylation status of BACE1 in treated neurons (data not shown).
In summary, our results argue for a well-defined mechanism through which aberrant APP signaling can trigger amyloidogenic processing of APP, without affecting neuronal survival, as depicted in Figure 9. As we discussed earlier, alterations in synaptic density occur early in AD and strongly correlate with the cognitive decline observed in the disease (reviewed in Scheff, 2006). Our results suggest that aberrant signaling through APP oligomerization is sufficient to drive synaptic dysfunction as well as promote Aβ production in hippocampal neurons. Whether, these effects are dependent on each other remains unclear. It is interesting to note however, that in our model system, while Aβ production is dependent on caspase-3, downregulation of caspase-3 did not protect neurons against the synaptotoxic effects of 22C11 suggesting on the surface that the effects we observed are independent of Aβ production. While somewhat puzzling, a simple explanation could be that signaling through APP results in the activation of a parallel pathway, which itself drives synaptic dysfunction and perhaps even neuronal death. Indeed, another consequence of APP cross-linking is the production of another toxic 31 amino acid cytosolic fragment, termed C31 (Bertrand et al., 2001; Lu et al. 2003; Shaked et al., 2006), which exerts its effects independently of Aβ (Park et al., 2009). This raises the interesting possibility that abnormal oligomerization of APP initiates a positive feedback loop in an affected neuronal population, resulting in local synaptic dysfunction while simultaneously raising the levels of Aβ, which can spread and “infect” healthy neurons in a manner similar to Prion’s disease. This may also suggest that the development of drugs aimed at disrupting APP dimerization may be a viable therapeutic approach worth exploring.
This work was supported in part by National Institutes of Health (NIH) Grant P50AG008702 (to M. L. S.). We thank Dr. Edward Koo (University of California San Diego) for providing B103 cells. We are thankful to Dr. Carol Troy (Columbia University) for helpful discussions and advice. We are grateful to Dr. Soline Aubry and AD for critically reading and carefully editing this manuscript.