PAR-4 Is Involved in Regulation of β-Secretase Cleavage of the Alzheimer Amyloid Precursor Protein*

Mounting evidence indicates that aberrant production and aggregation of amyloid β-peptide (Aβ)-(1–42) play a central role in the pathogenesis of Alzheimer disease (AD). Aβ is produced when amyloid precursor protein (APP) is cleaved by β- and γ-secretases at the N and C termini of the Aβ domain, respectively. The β-secretase is membrane-bound aspartyl protease, most commonly known as BACE1. Because BACE1 cleaves APP at the N terminus of the Aβ domain, it catalyzes the first step in Aβ generation. PAR-4 (prostate apoptosis response-4) is a leucine zipper protein that was initially identified to be associated with neuronal degeneration and aberrant Aβ production in models of AD. We now report that the C-terminal domain of PAR-4 is necessary for forming a complex with the cytosolic tail of BACE1 in co-immunoprecipitation assays and in vitro pull-down experiments. Overexpression of PAR-4 significantly increased, whereas silencing of PAR-4 expression by RNA interference significantly decreased, β-secretase cleavage of APP. These results suggest that PAR-4 may be directly involved in regulating the APP cleavage activity of BACE1. Because the increased BACE1 activity observed in AD patients does not seem to arise from genetic mutations or polymorphisms in BACE1, the identification of PAR-4 as an endogenous regulator of BACE1 activity may have significant implications for developing novel therapeutic strategies for AD.

The complete coding sequence of human BACE1 mRNA is ϳ1.5 kb long and encodes a pro-BACE1 protein, which is further cleaved by furin and other members of the furin family of convertases to remove the N-terminal propeptide domain within the trans-Golgi network (5,48,49). The cleavage occurs at the sequence RLPR2E, a furin recognition motif. Mature BACE1 is formed when cleaved immature BACE1 is further modified by N-linked glycosylation (46,55,57,66). BACE1 mRNA is found in neurons of all brain regions (47,64,72). Cleavage of APP by ␤-secretase (BACE1) generates a soluble N-terminal fragment (soluble APP␤) and a membrane-bound C-terminal fragment (known as CTF99). Cleavage of CTF99 by ␥-secretase produces A␤ . In vitro, BACE1 was shown to be able to generate two APP C-terminal fragments (CTFs) starting at either Asp 1 (CTF99) or Glu 11 (CTF89) of the A␤ sequence (47,49). Alternatively, cleavage of APP at the ␣-secretase site generates a soluble N-terminal fragment (soluble APP␣) and leaves the C-terminal fragment (known as CTF83) containing the APP transmembrane domain and cytoplasmic tail (89 -93). BACE1 is a type I membrane protein with a single transmembrane segment linking a luminal catalytic unit to a C-terminal cytosolic tail. The newly synthesized pro-BACE1 is processed to mature protease by furin during transit through the secretory pathway to the cell surface (1,5,6,16,18,34,. The observation that mice with a targeted deletion of BACE1 are viable and fertile but fail to produce any A␤ indicates that BACE1 is required for A␤ generation (6,40,51,62). Because neither genetic mutations nor polymorphisms in BACE1 seem to be responsible for the increased BACE1 activity observed in AD patients (1,42,47,51), the identification of factors that effectively regulate BACE1 activity is apparently a highly promising approach to treating AD.
PAR-4 (prostate apoptosis response-4) is a leucine zipper protein that was initially identified to be associated with neuronal degeneration in AD (94). The levels of Par-4 mRNA and protein were found to be increased in tissue from vulnerable brain regions of AD patients compared with age-matched control patients. Double labeling analysis using antibodies against phosphorylated tau (antibody PHF-1, a marker of neurofibrillary tangle-bearing neurons) revealed that ϳ30 -50% of tanglebearing neurons are also PAR-4 positive (94). The levels of PAR-4 in vulnerable neurons are effectively induced by insults relevant to the pathogenesis of AD, such as trophic factor withdrawal or aberrant elevations in intracellular calcium levels. Additional data suggest that PAR-4 exerts its cell deathpromoting action in the early stages of cell death prior to caspase activation and mitochondrial alterations (94). PAR-4 also increases the production of neurotoxic A␤ in transfected neural cells (95,96). An essential role for PAR-4 in aberrant A␤ production and neuronal apoptosis was demonstrated in studies of cultured primary rat hippocampal neurons and PC12 cells since inhibition of PAR-4 activity significantly diminishes A␤ production and attenuates apoptosis induced by A␤ or Alzheimer presenilin-1 (PS-1) mutations (95,96). Of importance, the adverse effects of PAR-4 seem to require its interaction with other proteins since a deletion mutant of PAR-4 lacking a protein/protein interaction domain in its C-terminal region does not enhance A␤ production, and overexpression of this domain of PAR-4 blocks PAR-4 activity in a dominant-negative fashion (95,96). Apoptosis-antagonizing transcription factor (AATF), another leucine zipper protein, is an interaction partner and potent inhibitor of PAR-4 activity (95,96). AATF was initially identified as an interaction partner of Dlk (DAP-like kinase), a member of the death-associated protein kinase family of pro-apoptotic serine/threonine kinases (96 -98). AATF binds directly with PAR-4 via the leucine zipper domain and blocks the activity of PAR-4 in enhancing A␤ production (96). These data strongly suggest that PAR-4 may, via protein/protein interactions, regulate not only apoptotic pathways, but also amyloidogenic processing of APP. We now report that the C-terminal domain of PAR-4 formed a complex with the cytosolic tail of BACE1 and increased ␤-secretase cleavage of APP. In hippocampal neurons, RNA interference (RNAi)-mediated silencing of Par-4 resulted in decreased APP cleavage activity of BACE1.

Transfection of IMR-32 Cells and Trophic Factor
Withdrawal-The methods used were described in our previous studies (94, 96, 99 -101). In brief, human neuroblastoma IMR-32 cells (American Type Culture Collection) were maintained at 37°C in an atmosphere of 95% air and 5% CO 2 in Eagle's minimal essential medium supplemented with nonessential amino acids and 10% heat-inactivated fetal bovine serum. A full-length Par-4 cDNA was subcloned into the expression vector pRc/ CMV, yielding the recombinant construct pCMV-Par-4, which encodes a 1.2-kb RNA species and a full-length 38-kDa PAR-4 protein (94,102). A cDNA fragment containing Par-4 lacking nucleotides 541-1267 of Par-4 cDNA (Par-4⌬CTF, which encodes a C-terminal half-deletion mutant of PAR-4) was similarly subcloned into the pRc/CMV expression vector, yielding the recombinant construct pCMV-Par-4⌬CTF, which encodes an ϳ700-bp RNA species or the N-terminal half (i.e. ϳ18 kDa) of the PAR-4 protein (94,102). A cDNA fragment containing the complete full-length human pro-BACE1 coding sequence (a generous gift from Dr. Jordan Tang, Oklahoma Medical Research Foundation) was subcloned into the pREP4 expression vector (Invitrogen), yielding the recombinant construct pREP4-BACE1, which encodes full-length human BACE1. Human IMR-32 cell lines stably expressing Par-4 and Par-4⌬CTF were established by transfection using Lipofectamine 2000 reagent (Invitrogen) with pCMV-Par-4 or pCMV-Par-4⌬CTF. Transfected cells were selected with G418 (400 g/ml) for 4 weeks, and surviving clones were selected. IMR-32 cells expressing human BACE1 were similarly established, except that the transfected cells were selected with hygromycin (400 g/ml). Additional double-transfected IMR-32 cell lines were generated in which two proteins (PAR-4/BACE1 and PAR-4⌬CTF/BACE1) were coexpressed. For control purposes, parallel cultures of IMR-32 cells were stably transfected with pRc/CMV and pREP4 vectors alone. After the cells became confluent in the culture flasks, the culture medium was replaced with fresh medium and incubated for 48 h at 37°C to condition the medium for A␤ measurement. Trophic factor withdrawal was initiated by washing cultures four times with Locke's buffer (154 mM NaCl, 5.6 mM KCl, 2.3 mM CaCl 2 , 1.0 mM MgCl 2 , 3.6 mM NaHCO 3 , 5 mM glucose, and 5 mM HEPES, pH 7.2) with subsequent incubation in 1 ml of Locke's buffer.
Immunoprecipitation and Western Blot Analysis-The levels of expression of PAR-4 and BACE1 were determined by Western blot analysis as described (94,96,101). The antibodies used that specifically recognize PAR-4 were polyclonal rabbit and mouse monoclonal antibodies raised against full-length rat PAR-4 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). These antibodies react with PAR-4 of mouse, rat, and human origin and recognize both full-length PAR-4 as well as the deletion mutant of PAR-4. The anti-APP antibody used was a rabbit polyclonal antibody raised against a 22-amino acid synthetic peptide derived from the C terminus of human APP (Zymed Laboratories Inc., South San Francisco, CA). The anti-BACE1 antibodies used were rabbit antiserum raised against mature human BACE1 (a generous gift from Dr. Jordan Tang) and rabbit polyclonal antibody raised against a synthetic peptide corresponding to amino acids 458 -501 of human BACE1 (Chemicon International, Inc., Temecula, CA). For immunoprecipitation, aliquots of cell lysates containing 200 g of protein were incubated for 1 h at 4°C with appropriate dilutions of anti-PAR-4 or anti-BACE1 antibody in immunoprecipitation buffer (150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, 5 g/ml leupeptin, 5 g/ml aprotinin, 2 g/ml pepstatin A, 0.25 mM phenylmethylsulfonyl fluoride, and 50 mM Tris, pH 7.6). Protein A-Sepharose resin (30 l/sample; Amersham Biosciences) was then added to the extracts for collection of the immunocomplexes, which were then washed three times with immunoprecipitation buffer and solubilized by heating in Laemmli buffer containing 2-mercaptoethanol at 100°C for 4 min. The solubilized proteins were separated by electrophoresis on a 4 -12% gradient SDS-polyacrylamide gel and transferred to a nitrocellulose sheet. For Western blot analysis, the nitrocellulose sheet was blocked with 5% milk, followed by a 1-h incubation in the presence of anti-APP, anti-BACE1, or anti-PAR-4 primary antibody. The membrane was further processed using horseradish peroxidaseconjugated secondary antibody, and immunoblotted proteins were detected by chemiluminescence using the ECL system (Amersham Biosciences). To examine whether PAR-4 interacts with PS-1, an affinitypurified rabbit anti-PS-1 polyclonal antibody (103) was used in the immunoprecipitation/Western blot analysis. Equal loading was verified by probing the blots with anti-tubulin antibody (Sigma). Western blot images were acquired and quantified using Image Station 2000R and Digital Science 1D Version 3.6 software (Eastman Kodak Co).
Hippocampal Neuronal Cultures and Quantification of A␤-(1-40) and A␤-  by Sandwich Enzyme-linked Immunosorbent Assays (ELISAs)-Dissociated hippocampal cell cultures were prepared from postnatal day 1 mouse pups using methods similar to those described previously (100). Briefly, hippocampi were removed from mouse brain and incubated for 15 min in Ca 2ϩ -and Mg 2ϩ -free Hanks' balanced saline solution (Invitrogen) containing 0.2% papain. Cells were dissociated by trituration and plated into polyethyleneimine-coated plastic or glass-bottom culture dishes containing minimal essential medium with Earle's salts supplemented with 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, 1 mM pyruvate, 20 mM KCl, 10 mM sodium bicarbonate, and 1 mM HEPES, pH 7.2. Following cell attachment (3-6 h post-plating), the culture medium was replaced with Neurobasal Medium with B27 supplements (Invitrogen). Experiments were performed in 7-day-old cultures. A␤-(1-40) and A␤-(1-42) levels in the conditioned culture medium were measured using a fluorescence-based sandwich ELISA described in detail in our previous studies (95,96). The C terminus-specific sandwich ELISAs use a monoclonal antibody directed against the N-terminal region of human A␤ and two other antibodies specific for A␤-(1-40) and A␤-(1-42). The ratio (percent) of A␤-(1-42) to total A␤ (A␤-(1-40) plus A␤-(1-42)) was used to measure the changes in the relative amount of A␤-(1-42) secreted from transfected IMR-32 cells (95,96).
Pull-down Assays-The methods used were essentially the same as those described previously (69). In brief, the peptide from the BACE1 C-terminal cytosolic tail (referred to as BACE1 CT, -CLRQQHDDFAD-DISLLK) was covalently linked by thiol groups to Sulfolink coupling gel (Pierce) according to the manufacturer's instructions. Full-length PAR-4 was expressed in IMR-32 cells; the cell lysates were prepared; and 200 g of protein samples were incubated with 100 l of gel bearing immobilized peptide in 1.5 ml of phosphate-buffered saline (PBS) at room temperature for 2 h. The gel beads were pelleted by centrifugation at 750 ϫ g for 1 min and washed three times with PBS. The proteins on the gel beads were eluted with SDS sample buffer and subjected to SDS-PAGE. PAR-4 immunoreactivity was identified by Western blotting using anti-PAR-4 monoclonal antibody. A cysteine-blocked gel was used as a negative control.
Immunocytochemistry and Fluorescence Microscopy-For immunocytochemical analysis of PAR-4 expression, the cultured cells were fixed for 30 min in 4% paraformaldehyde and PBS, and membranes were permeabilized by incubation in 0.2% Triton X-100 in PBS. Cells were incubated for 1 h in blocking serum (5% normal goat serum in PBS). Cells were then exposed overnight at 4°C to a 1:100 dilution of mouse anti-PAR-4 monoclonal antibody, followed by incubation for 1 h with fluorescein-labeled anti-mouse secondary antibody (Vector Laboratories, Burlingame, CA). Images of PAR-4 immunofluorescence were acquired using a Nikon TS 100 fluorescence microscope. The average pixel intensity of PAR-4 immunoreactivity per cell was determined using LSM 510 software (Carl Zeiss MicroImaging, Inc., Thornwood, NY). Throughout the experiments, the specificity of the immunoreactivity was confirmed by subjecting additional samples to the immunostaining procedures without primary antibody.
PAR-4 Knockdown by RNAi-The methods used have been described in our previous study (104). In brief, small interfering RNAs (siRNAs) were generated by in vitro transcription using a Silencer TM siRNA mixture kit (Ambion Inc., Austin, TX) following the manufacturer's instructions. T7 promoter sequences were added to a DNA template (a cDNA fragment containing nucleotides 511-990 of the mouse Par-4 coding sequence) by PCR using the following primers: forward primer, 5Ј-taatacgactcactatagggtactgaagatgatgaagca-3Ј; and reverse primer, 5Јtaatacgactcactatagggtactctgcccaacaacctt-3Ј. The transcription reaction was assembled, and the resulting cRNA was annealed according to the manufacturer's instructions for maximal duplex yield. The RNA samples were treated with DNase/RNase A to remove DNA, and single-and double-stranded RNAs were then purified using a transcription reaction filter cartridge (Ambion Inc.). An siRNA mixture was obtained by RNase III digestion and further purified using an siRNA purification unit (Ambion Inc.). Primary neurons were transfected with siRNA mixtures at a concentration of 100 nM using TransMessenger transfection reagent (Qiagen Inc., Valencia, CA). To determine the transfection efficiency, siRNAs were fluorescently labeled with Cy3 using a Silencer TM siRNA labeling kit (Ambion Inc.) following the manufacturer's instructions. Labeled siRNAs were then transfected into cells and analyzed by fluorescence microscopy. The effectiveness of the RNAi experiments under our experimental conditions was confirmed using a glyceraldehyde-3-phosphate dehydrogenase-positive control DNA template (Ambion Inc.) according to the manufacturer's instructions. A FIG. 1. Increased BACE1 cleavage of APP induced by PAR-4. a, PAR-4 increases ␤-secretase cleavage of APP. Representative Western blot analysis was carried out with rabbit polyclonal antibody raised against a 22-amino acid synthetic peptide derived from the C terminus of human APP showing increased levels of the ␤-secretase cleavage products CTF99 and CTF89 in IMR-32 cells transfected with PAR-4. The lower bands represent the C-terminal ␣-secretase cleavage product CTF83, which was not significantly altered by PAR-4. C1 and C2, clones C1 and C2, respectively. b, PAR-4 significantly increases secretion of A␤-(1-42) from transfected IMR-32 cells following trophic factor withdrawal. Cultures of the indicated clones of transfected IMR-32 cells were deprived of trophic support for the indicated time periods, and the A␤-(1-42)/total A␤ ratios in the conditioned culture media of transfected IMR-32 cells were measured by sandwich ELISAs. Note that overexpression of PAR-4 drastically increased the A␤-(1-42)/total A␤ ratios in the conditioned media following trophic factor withdrawal (TFW). ***, p Ͻ 0.001 compared with the corresponding A␤-(1-42)/total A␤ ratios in untransfected and vector-transfected control cell groups. Similar data were obtained from at least three separate transfected cell lines. c, cotransfection of BACE1 and PAR-4 increases the relative amount of A␤-(1-42) secreted by IMR-32 cells under basal culture conditions. Whereas transfection of PAR-4 alone significantly increased the levels of A␤ production following apoptotic insults, coexpression of BACE1 with PAR-4 altered A␤ production under basal non-apoptotic culture conditions. The percentage of A␤- non-silencing siRNA and a validated siRNA against green fluorescent protein (GFP; Qiagen Inc.) were used as negative controls. Using these methods, an average of 78.6 Ϯ 6.2% of siRNA transfection efficiency has been obtained in primary neurons (104). Primary mouse hippocampal neurons carrying human wild-type APP695 (APPwt) or human APP with Swedish (K670N/M671L) mutations (APPmut) were prepared using an adeno-associated virus vector system as described (105) and were employed to examine the effect of Par-4 siRNAs on the levels of CTF99 and A␤ production using Western blotting and fluorescence-based sandwich ELISAs as described above.

Cotransfection of PAR-4 Increases BACE1 Cleavage of APP
and Production of A␤-We previously found that overexpression of PAR-4 increases the production of A␤-(1-42) following apoptotic insults (95,96). This effect of PAR-4 is effectively inhibited by AATF, an interaction partner of PAR-4 (95,96). These results suggest that PAR-4 is involved in regulation of APP processing, and this effect of PAR-4 is likely mediated through protein/protein interactions. Because BACE1 cleaves APP at the N terminus of the A␤ domain, it catalyzes the first step in A␤ generation. To test whether PAR-4 alters BACE1 cleavage of APP, we examined the relative levels of CTF99/ CTF89, two major APP C-terminal products of BACE1 cleavage, in IMR-32 cells transfected with PAR-4. As shown in Fig.  1a, although the levels of both CTF99 and CTF89 were significantly increased by overexpression of PAR-4, the increase in the levels of CTF99 was much more pronounced compared with CTF89. The levels of CTF83, the C-terminal ␣-secretase product of APP, were not significantly altered. These results indicate that overexpression of PAR-4 increases ␤-secretase cleavage of APP. To further examine whether a PAR-4-induced increase in ␤-secretase cleavage of APP leads to increased production of A␤-(1-40) and/or A␤-  in these cells, we measured the A␤-(1-42)/total A␤ ratio in the conditioned culture medium of transfected IMR-32 cells using sensitive sandwich ELISAs as described previously (95,96). We found that although expression of PAR-4 alone did not significantly alter the basal levels of A␤- (1-42), the secretion of A␤-(1-42) was significantly increased following trophic factor withdrawal (Fig. 1b). Transfection of BACE1 alone in IMR-32 cells increased the production of both A␤-(1-40) and A␤-(1-42) under basal culture conditions (Table I). The levels of A␤-  seemed to be altered more dramatically than those of A␤-(1-40) by overexpression of BACE1, which led to a slight but significant increase in the A␤-(1-42)/total A␤ ratio ( Fig. 1c and Table I). Of importance, cotransfection of PAR-4 with BACE1 further enhanced the production of A␤, resulting in a significantly exacerbated secretion of A␤-(1-42) under basal and non-apoptotic culture conditions (Fig. 1c and Table I). Taken together, these results demonstrate that PAR-4 alters BACE1 cleavage of APP and increases A␤ production in transfected neural cells.
PAR-4 Is Associated with BACE1 in Transfected Neural Cells and Primary Neurons-To examine whether PAR-4 alters ␤-secretase cleavage of APP by interacting with BACE1, we performed co-immunoprecipitation/Western blot experiments using specific anti-PAR-4 and anti-BACE1 antibodies. Because the basal levels of BACE1 and PAR-4 in untransfected IMR-32 cells are relative low, co-immunoprecipitation studies were performed first on homogenates of transfected IMR-32 cell clones overexpressing PAR-4 and BACE1. When immunoprecipitation was performed using anti-BACE1 antibody, a 38-kDa PAR-4 band was clearly detected on the immunoblot (Fig. 2a). The reverse order of immunoprecipitation/Western blot analysis of the same transfected cell lines showed similar PAR-4⅐BACE1 complex formation (Fig. 2b). To exclude the possibility that the PAR-4⅐BACE1 complex formation was due to an artifact of overexpression of both proteins, we performed further immunoprecipitation/Western blot analysis in primary hippocampal neurons expressing physiological concentrations of these proteins. As shown in Fig. 2 (c and d), similar PAR-4⅐BACE1 complex formation was clearly observed. Note that anti-BACE1 antibody recognized predominantly the mature form of BACE1 in both transfected IMR-32 cells as well as primary neurons. These results demonstrate the interaction between endogenous PAR-4 and BACE1 and indicate that PAR-4⅐BACE1 complex formation is physiologically relevant. A control antibody against AATF failed to precipitate BACE1 (Fig.  2d, lane 3), indicating the specificity of the PAR-4/BACE1 interaction. To further confirm the specificity of the PAR-4/ BACE1 interaction, we examined whether PAR-4 would interact with presenilin-1 using a specific anti-PS-1 antibody. As shown in Fig. 2e, immunoprecipitation/Western blot experiments in transfected IMR-32 cells expressing PAR-4 showed that both full-length PS-1 and its cleavage products were detected in proteins from total lysate, but not in those immunoprecipitated with anti-PAR-4 antibody, indicating that PAR-4 does not interact with PS-1. Similar data were obtained in the reverse order of immunoprecipitation/Western blot analyses of the same transfected cell lines (data not shown). These negative control experiments confirmed the selectivity of PAR-4⅐BACE1 complex formation.
Mapping of PAR-4/BACE1-interacting Domains-Next, we performed additional pull-down and co-immunoprecipitation/ Western blot experiments to determine the interacting domains in PAR-4 and BACE1. Because BACE1 is a type I membrane protein and because PAR-4 is primarily cytosolic, we expected PAR-4 to be associated with the C-terminal cytosolic tail of BACE1. To examine this possibility, we employed the BACE1 CT peptide (-CLRQQHDDFADDISLLK) and examined whether PAR-4 could be pulled down by the BACE1 CT peptide immobilized on gel beads (see "Materials and Methods"). The BACE1 CT peptide was covalently linked to Sulfolink coupling gel. The cell lysates from PAR-4-transfected IMR-32 cells were    incubated with the gel-linked BACE1 CT peptide. The gel samples were washed and then subjected to SDS-PAGE and Western blotting for PAR-4. As shown in Fig. 3a, PAR-4 was clearly and specifically pulled down by the BACE1 CT peptide. To further determine which domain(s) of PAR-4 are involved in the interaction with the cytosolic tail of BACE1, we performed co-immunoprecipitation/Western blot analysis using fulllength PAR-4 as well as a C-terminal half-deletion mutant of PAR-4 lacking nucleotides 541-1267 of Par-4 cDNA (Par-4⌬CTF). IMR-32 cells were cotransfected with BACE1 and Par-4 or Par-4⌬CTF. Cell lysates from cotransfected cells were precipitated with anti-BACE1 antibody, followed by Western blotting with anti-PAR-4 antibody (Fig. 3b). Note that BACE1 interacted only with full-length PAR-4 (Fig. 3b, lane 3), but not with the C-terminal half-deletion mutant of PAR-4, indicating that the C-terminal half of PAR-4 is necessary in the interaction with BACE1. The reverse order of immunoprecipitation/ Western blot analysis of the same transfected cells confirmed that deletion of the C-terminal half of PAR-4 abolished the interaction between BACE1 and PAR-4 (data not shown).
Taken together, our results indicate that the C-terminal half of PAR-4 is necessary for association with the cytosolic tail of BACE1 and that the PAR-4/BACE1 interaction exacerbates BACE1-mediated cleavage of APP.
RNAi-mediated Par-4 Gene Silencing Diminishes BACE1 Cleavage of APP-RNAi has been proved to be a very effective oligonucleotide-based gene silencing technology that allows sequence-specific gene suppression in a variety of organisms and cultured cells. The application of RNAi techniques to primary neurons, a setting in which genetic manipulations have traditionally proven difficult, has become a versatile tool to study gene function and protein expression in neuronal cells. To further determine the crucial role of PAR-4 in regulating ␤-secretase cleavage of APP, we examined whether knockdown of PAR-4 expression by RNAi would result in a significant decrease in ␤-cleavage of APP and in A␤ production. One drawback of the RNAi technology is the need to design, synthesize, and test several siRNAs before an effective sequence can be identified. However, this problem can be avoided altogether with the Silencer TM siRNA mixture kit employed in this study.

FIG. 2. Specific interaction between PAR-4 and BACE1 in transfected IMR-32 cells and in primary hippocampal neurons. a and b,
PAR-4⅐BACE1 complex formation in transfected IMR-32 cells. a, transfected IMR-32 cells coexpressing human BACE1 and PAR-4 were lysed and precipitated with rabbit polyclonal antibody raised against a synthetic peptide corresponding to amino acids 458 -501 of human BACE1, followed by Western blotting with anti-PAR-4 antibody (lanes 3 and 4). Input lanes (lanes 1 and 2) show 10% of the total protein used in immunoprecipitation (IP) experiments. Rabbit preimmune serum was used as a control (lanes 5 and 6). The PAR-4⅐BACE1 complex was clearly observed in two separate clones of transfected cells (lanes 3 and 4). b, the reverse order of immunoprecipitation/Western blot analysis of the same transfected cells showed similar PAR-4⅐BACE1 complex formation (lanes 3 and 4). Input lanes (lanes 1 and 2) show 10% of the total protein used in immunoprecipitation experiments. Note that anti-BACE1 antibody recognized predominantly mature BACE1 (mBACE1). imBACE1, immature BACE1. c and d, interaction between endogenous PAR-4 and BACE1 in primary hippocampal neurons. c, cultures of primary hippocampal neurons were lysed, and proteins from total lysates were immunoprecipitated with rabbit anti-BACE1 antibody (lane  3 and 4), followed by Western blotting with anti-PS-1 antibody. The input lanes (lanes 1 and 2) show 10% of the total protein used in immunoprecipitation experiments. Similar data were obtained in the reverse order of immunoprecipitation/Western blot analysis of the same transfected cells (data not shown). Note that both full-length PS-1 and its cleavage products were detected in proteins from total lysate, but not in those immunoprecipitated with anti-PAR-4 antibody.
In this system, a population of several siRNAs (instead of a single siRNA sequence) was generated by digesting long double-stranded RNA with RNase III, which cleaves doublestranded RNA into 12-30-bp double-stranded RNA fragments with 2-3-nucleotide 3Ј-overhangs and 5Ј-phosphate and 3Ј-hydroxyl termini. The termini and overhangs of RNase III cleavage products are thus the same as those produced by Dicer in the in vivo RNAi pathway. This effectively eliminates the need to design and screen individual siRNAs. The successful application of this RNAi protocol to knockdown PAR-4 expression in primary neurons has been described (104). The transfection efficiency of siRNAs in neuronal cultures was determined by fluorescence microscopy of neurons transfected with siRNAs fluorescently labeled with Cy3 (104). Under our experimental conditions, an average of 78.6 Ϯ 6.2% siRNA transfection efficiency has been achieved (104). To confirm that the siRNA mixture targeted against Par-4 would effectively inhibit aberrant induction of PAR-4 expression, primary hippocampal neurons from wild-type mice were either mock-transfected (control) or transfected with siRNAs against Par-4 at a concentration of 100 nM. 48 h after siRNA transfection, cells were treated with either vehicle (control) or 50 M glutamate for 8 h. The cells were then processed for PAR-4 immunoreactivity using fluorescence microscopy. As shown in Fig. 4a, the induction of PAR-4 expression induced by glutamate, as assessed by immunofluorescence microscopy, was completely knocked down by siRNAs targeted against Par-4, but not by the non-silencing control siRNA. Further Western blot analysis confirmed that PAR-4 expression was largely knocked down by siRNAs targeted against Par-4, but not by a validated negative control siRNA against GFP (Fig. 4, b and c). To determine the specificity of the siRNA mixture targeted against Par-4, the same protein samples were probed with antibody against another leucine zipper protein, AATF. As shown in Fig. 4 (b and  c), neither the siRNA against Par-4 nor the siRNA against GFP altered the expression of AATF. These results indicate that the Par-4 siRNA mixture specifically and effectively targets Par-4 mRNAs for degradation by RNAi.
To examine whether PAR-4 is involved in regulating ␤-secretase cleavage of APP, we investigated whether RNAi-mediated silencing of Par-4 would lead to a significant decrease in the levels of the ␤-secretase cleavage product CTF99 in hippocampal neurons expressing APPwt or APPmut. Densitometric analysis of Western blots of CTF99 showed that the levels of CTF99 were significantly reduced in hippocampal neurons ex-pressing APPwt or APPmut by siRNAs targeted against Par-4 ( Fig. 4, d and e), but not by a negative control siRNA against GFP (data not shown). Interestingly, whereas a small amount of CTF89 was detected in APPwt neurons, the levels of CTF89 seemed to be significantly diminished in APPmut neurons (Fig.  4d). The significance of this observation is not immediately clear. The levels of CTF83 in APPwt-and APPmut-expressing neurons were not significantly altered by siRNAs targeted against Par-4 (Fig. 4d).
We next examined whether RNAi-mediated depletion of PAR-4 would reduce A␤ production in hippocampal neurons expressing APPmut. As shown in Table II, the levels of A␤ (particularly of A␤-(1-42)) in the conditioned culture medium of primary hippocampal neurons expressing APPmut were significantly reduced by Par-4 siRNAs. DISCUSSION Mounting evidence indicates that aberrant production and aggregation of A␤-(1-42) play a central role in the pathogenesis of AD . A␤ is produced when APP is cleaved by ␤and ␥-secretases at the N and C termini of the A␤ domain, respectively. The observation that mice with a targeted deletion of BACE1 are viable and fertile but fail to produce any A␤ indicates that BACE1 is strictly required for A␤ generation and that inhibition of BACE1 activity might be a safe and effective treatment option for AD (6,40,51,62). Because the increased BACE1 activity observed in AD patients does not seem to arise from genetic mutations or polymorphisms in BACE1 (1,42,47,51), the identification of factors that effectively regulate BACE1 activity is apparently a highly promising approach to treating AD. PAR-4 is a leucine zipper protein that was initially identified to be associated with neuronal degeneration in AD (94). Subsequent studies found that PAR-4, via protein/protein interactions, also increases production of neurotoxic A␤ in transfected neural cells, indicating that PAR-4 may regulate both apoptosis and amyloidogenic processing of APP (95,96). Because BACE1 cleaves APP at the N terminus of the A␤ domain, it catalyzes the first step in A␤ generation. These results suggest that PAR-4 might be involved in regulating BACE1 cleavage of APP.
The data presented in this study indicate that PAR-4 is indeed associated with BACE1 and specifically increases ␤-secretase cleavage of APP. This notion is supported by several lines of experimental evidence obtained in this study. 1) Transfection of PAR-4 in IMR-32 cells increased the levels of the ␤-secretase cleavage products CTF99 and CTF89 without  3 and 4). Input lanes (lanes 1 and 2) show 10% of the total protein used in immunoprecipitation (IP) experiments. Rabbit preimmune serum was used as a control (lanes 5 and 6). Note that BACE1 interacted only with full-length PAR-4 (lane necessary in the interaction with the C-terminal cytosolic tail of BACE1. 5) RNAi-mediated silencing of Par-4 expression in primary neurons led to a decrease in ␤-secretase cleavage of APP, resulting in a decreased production of A␤. These results suggest that PAR-4 is a potential endogenous regulator of BACE1 activity. The observation that PAR-4 was specifically associated with BACE1 in both transfected cells as well as primary neurons indicates that the interaction between PAR-4 and BACE1 is physiologically relevant. The negative control experiments showing that BACE1 was not associated with AATF and that PAR-4 did not interact with PS-1 indicate that PAR-4/BACE1 binding is specific. The observations that overexpression of PAR-4 significantly increased, whereas silencing of PAR-4 expression significantly decreased, ␤-secretase cleavage of APP indicate that the association of PAR-4 with BACE1 may accelerate the processing of APP by ␤-secretase.
Overexpression of PAR-4 was shown to increase the vulnerability of neural cells to apoptotic insults relevant to the pathogenesis of AD (94). Although overexpression of PAR-4 alone does not affect the production of A␤-(1-42) under basal conditions, it significantly increases the production of A␤-(1-42) following trophic factor withdrawal (95,96). These findings reconcile the apoptotic and amyloid hypotheses of AD and indicate that factors regulating apoptosis may also be involved in regulating aberrant APP processing and A␤ production. Time course analysis of the increase in A␤-(1-42) induced by PAR-4 showed that overexpression of PAR-4 alone increases the production of A␤ after apoptotic cascades are initiated, indicating that apoptosis and aberrant APP processing may be intimately linked processes (95,96). This notion is further supported by several recent reports showing that ␤-secretase activity and expression in rats are significantly increased following transient cerebral ischemia (106) and that overexpression of the ␤-secretase-derived APP C-terminal fragment CTF100, with and without a signal peptide, increases apoptotic cell death in the neurons (107). In addition, the C-terminal ␤-secretase cleavage product of APP may also be subject to further processing by caspases, leading to the production of apoptotic peptide C31 (108,109). On the other hand, the data presented in this study also show that coexpression of PAR-4 significantly exacerbated the BACE1-induced increase in the production of A␤-(1-42) even under basal non-apoptotic conditions, indicating that PAR-4 may also regulate A␤-(1-42) through non-apoptotic mechanisms.
The precise mechanisms by which the association of PAR-4 with BACE1 alters ␤-secretase cleavage of APP need to be further investigated. One possibility is that binding of PAR-4 to BACE1 might directly or indirectly alter the binding affinity of BACE1 for its substrate APP. In addition, because BACE1 is active in acidic medium up to pH 5, both APP and BACE1 at the cell surface are internalized to endosomes, the main site for APP cleavage by BACE1 and the production of A␤. Mature BACE1 is formed when cleaved immature BACE1 is further modified by N-linked glycosylation at Asn 153 , Asn 172 , Asn 223 , and Asn 354 , and the ␤-secretase activity seems to be dependent on the extent of N-linked glycosylation (1,5,6,16,18,34,. Therefore, it is also possible that binding of PAR-4 to BACE1 may regulate ␤-secretase cleavage of APP by altering intracellular trafficking and/or maturation of BACE1. Because BACE1 cleaves APP at the N terminus of the A␤ domain, it catalyzes the first step in A␤ generation (1,5,6,16,18,34,. The observation that mice with a targeted deletion of BACE1 are viable and fertile but fail to produce any A␤ indicates that inhibition of BACE1 activity might be a safe and effective treatment option for AD (6,40,51,62). Because the increased BACE1 activity observed in AD patients does not seem to arise from genetic mutations or polymorphisms in BACE1 (1,42,47,51), the identification of PAR-4 as an endogenous regulator of BACE1 activity may therefore have significant implications for developing novel therapeutic strategies for AD.