The Essential Role of p53-up-regulated Modulator of Apoptosis (Puma) and Its Regulation by FoxO3a Transcription Factor in β-Amyloid-induced Neuron Death*

Background: β-Amyloid-induced neuron death and degeneration is considered to be central to the pathogenesis of Alzheimer disease. Results: p53-up-regulated modulator of apoptosis (Puma), a protein of the B-cell lymphoma-2 family, is induced by transcription factor FoxO3a and participates in neuron death in response to β-amyloid. Conclusion: β-Amyloid-induced neuron death requires induction of Puma. Significance: Puma could be a potential target for disease therapeutics. Neurodegeneration underlies the pathology of Alzheimer disease (AD). The molecules responsible for such neurodegeneration in AD brain are mostly unknown. Recent findings indicate that the BH3-only proteins of the Bcl-2 family play an essential role in various cell death paradigms, including neurodegeneration. Here we report that Puma (p53-up-regulated modulator of apoptosis), an important member of the BH3-only protein family, is up-regulated in neurons upon toxic β-amyloid 1–42 (Aβ(1–42)) exposure both in vitro and in vivo. Down-regulation of Puma by specific siRNA provides significant protection against neuron death induced by Aβ(1–42). We further demonstrate that the activation of p53 and inhibition of PI3K/Akt pathways induce Puma. The transcription factor FoxO3a, which is activated when PI3K/Akt signaling is inhibited, directly binds with the Puma gene and induces its expression upon exposure of neurons to oligomeric Aβ(1–42). Moreover, Puma cooperates with another BH3-only protein, Bim, which is already implicated in AD. Our results thus suggest that Puma is activated by both p53 and PI3K/Akt/FoxO3a pathways and cooperates with Bim to induce neuron death in response to Aβ(1–42).

Neurodegeneration underlies the pathology of Alzheimer disease (AD). The molecules responsible for such neurodegeneration in AD brain are mostly unknown. Recent findings indicate that the BH3-only proteins of the Bcl-2 family play an essential role in various cell death paradigms, including neurodegeneration. Here we report that Puma (p53-up-regulated modulator of apoptosis), an important member of the BH3-only protein family, is up-regulated in neurons upon toxic ␤-amyloid 1-42 (A␤(1-42)) exposure both in vitro and in vivo. Down-regulation of Puma by specific siRNA provides significant protection against neuron death induced by A␤(1-42). We further demonstrate that the activation of p53 and inhibition of PI3K/Akt pathways induce Puma. The transcription factor FoxO3a, which is activated when PI3K/Akt signaling is inhibited, directly binds with the Puma gene and induces its expression upon exposure of neurons to oligomeric A␤(1-42). Moreover, Puma cooperates with another BH3-only protein, Bim, which is already implicated in AD. Our results thus suggest that Puma is activated by both p53 and PI3K/Akt/FoxO3a pathways and cooperates with Bim to induce neuron death in response to A␤(1-42).
Alzheimer disease (AD) 2 is an irreversible cognitive malfunctioning of brain that eventually dysregulates the integrity of the nervous system. Widespread synapse and neuron loss in selective areas of the brain are important characteristic features of AD along with other pathophysiological hallmarks, such as ␤-amyloid (A␤) plaques and neurofibrillary tangles. The amyloid cascade hypothesis claims a central role of A␤ accumulation in the disease pathogenesis (1). It has also been shown that oligomeric A␤ is the main pathologic species that induces neu-rodegeneration in vivo and in vitro (2)(3)(4). However, downstream effectors of A␤ toxicity still remain elusive.
Accumulating evidence implicates a number of death-associated genes in AD-related neuron death (2,4,5). Proteins of the Bcl-2 (B-cell lymphoma 2) family regulate cell death in response to most if not all of the death insults. The Bcl-2 family comprises pro-survival (e.g. Bcl-2 and BclxL), multidomain proapoptotic (e.g. BAX and BAK), and BH3-only proapoptotic (e.g. Bim (Bcl-2-interacting mediator of cell death), Bid, and Puma) proteins (6). Activation of the proapoptotic members BAX and/or BAK is essential for cell death (7). The activity of BAX is controlled by the opposing action of prosurvival members and BH3-only proteins of the family. Recently, essential roles of Bim, Bid, and Puma have been shown in BAX activation and subsequent apoptosis during development (8). More importantly, it has been shown that inhibition of BAX protects neurons from A␤ toxicity in vitro and in vivo (2). Bim has also been shown to be an essential mediator of neuron death, yet knockdown of this molecule provides only transient protection in AD-relevant cell death models (5). This indicates that other death-associated molecules are also necessary in the molecular events of cell death in AD. Interestingly, recently, it has been shown that Puma cooperates with Bim in a critical apoptotic check point in autoreactive thymocytes (9) and in oncogene inactivation-induced apoptosis (10). Puma is induced by various apoptotic stimuli like DNA damage, endoplasmic reticulum stress, aberrant oncogene expression, serum or growth factor deprivation, and oxidative stress (11). However, its role in AD-related neurodegeneration has not yet been explored.
Puma is known to be a key transcriptional target of the tumor suppressor p53, and it carries out the cell death cascade in response to p53 activation (12)(13)(14)(15)(16). However, an increasing body of evidence shows that Puma is also regulated in a p53independent manner (11,17,18). In healthy cells, Puma is kept in check by survival signals (19,20). Inhibiting these signals through growth factor or cytokine withdrawal leads to p53independent transcriptional activation of Puma. It has been shown that Forkhead transcription factor FoxO3a (Forkhead box, class O3a) is activated and up-regulates Puma in response to such survival factor withdrawal (21,22). In surviving cells, FoxO3a is phosphorylated by Akt and remains in cytosol bound to 14-3-3 protein. However, in response to deprivation of growth factors or cytokines, it becomes dephosphorylated, translocates into the nucleus, and induces its target genes like Bim, Puma, etc. (22,23). Recently, we have shown that FoxO3a is activated by multiple post-translational modifications, translocates to the nucleus, and mediates neuron death via Bim in response to A␤ (4). However, whether Puma is a target of activated FoxO3a in A␤-treated neurons is yet to be discovered.
Because A␤ is widely considered to be the toxic species in AD, a number of drugs targeted to A␤ metabolism are in clinical trials to reduce the A␤ load. Although most of them have failed in clinical trials, targeting A␤ is still considered to be a very useful therapeutic strategy (24). In addition to that, it is believed that a complementary therapy may be necessary to block toxicity of the A␤ that cannot be removed completely. Therefore, it is essential to understand the molecular mechanism of A␤-induced neuron death. In this study, we have investigated the role of Puma in A␤-induced neuron death. Our findings suggest that Puma is induced transcriptionally by FoxO3a and cooperates with Bim to induce neuron death in response to A␤ toxicity.

EXPERIMENTAL PROCEDURES
Materials-A␤  was purchased from American Peptide. Insulin, progesterone, putrescine, selenium, transferrin, NGF, and poly-D-lysine were purchased from Sigma. Anti-Puma and anti-FoxO3a antibodies were from Cell Signaling Technology. Protein A-agarose and HRP-conjugated secondary antibodies were from Santa Cruz Biotechnology, Inc. LY 294002 and pifithrin-␣ were from Calbiochem. Lipofectamine 2000, Alexa Fluor 488, Alexa Fluor 568, culture media, and serum were purchased from Invitrogen. Brain tissues of A␤PPswe-PS1de9 mice and control littermates were kindly gifted by Dr. Anant B. Patel (Council of Scientific and Industrial Research-Centre for Cellular and Molecular Biology (CSIR-CCMB), Hyderabad, India).
Cell Culture-Cortical neurons from the neocortex of rat brain were cultured as described previously (25,26). Briefly, neurons were isolated from the neocortex of day 18 embryos. The cells were plated on poly-D-lysine-coated culture plates and maintained in DMEM/F-12 medium supplemented with insulin (25 g/ml), glucose (6 mg/ml), transferrin (100 g/ml), progesterone (20 ng/ml), putrescine (60 g/ml), and selenium (30 ng/ml). Cultured neurons were subjected to treatment after 6 days. Rat pheochromocytoma (PC12) cells were cultured as described previously (27) in RPMI medium supplemented with 10% heat-inactivated horse serum and 5% heat-inactivated fetal bovine serum. Neuronal differentiation was induced by NGF (100 ng/ml) in medium containing 1% horse serum for 6 days before the treatment, as described previously (28).
PCR-Total RNA of each sample is isolated from cultured cortical neurons by using TRI reagent (Sigma). The primers used for PCR amplification of rat Puma were 5Ј-GCGGA-GACAAGAAGAGCAAC-3Ј and 5Ј-CAAGGCTGGCAGTC-CAGTAT-3Ј. The primers for ␣-tubulin were 5Ј-ATGAGGC-CATCTATGACATC-3Ј and 5Ј-TCCACAAACTGGATGG-TAC-3Ј, and those for 18S were 5Ј-GCTTAATTTGACTCAAC-ACGGGA-3Ј and 5Ј-AGCTATCAATCTGTCAATCCTGTC-3Ј. Equal amounts of cDNA template were used for each PCR analysis of Puma or ␣-tubulin/18S. Primers were used at 0.2 M concentration. For semiquantitative PCR, products were analyzed on a 1.5% agarose gel and visualized by staining with ethidium bromide. Quantitative PCR was performed using One Step SYBR Ex Taq qRT-Takara by using an Applied Biosystems 7500 Fast Real Time PCR System following the manufacturer's specifications.
Western Blotting-Cortical neurons were lysed, and proteins were analyzed by Western blotting as described previously (30). For each condition, 50 g of protein were resolved in 12% SDS-PAGE and then transferred to PVDF membrane (Hybond, GE Healthcare). HRP-conjugated secondary antibodies against the primary antibodies were used. Detection was done by Amersham Biosciences ECL Western blotting detection reagent, according to the manufacturer's protocol. Bands were detected on an x-ray film (Eastman Kodak Co.) or Geldoc (4000 Pro, Carestream).
Immunocytochemical Staining-Cortical neurons were immunostained as described previously (28,31). Briefly, the cells were fixed with 4% paraformaldehyde for 10 min and then were washed with PBS three times for 5 min each. The cells were then blocked in 3% goat serum in PBS containing 0.1% Triton X-100 for 2 h at room temperature. The cells were incubated with anti-Puma antibody in a blocking solution overnight at 4°C. Alexa Fluor 546 was used as secondary antibody, and the nuclei were stained with Hoechst.
The intensities of staining for control or treated cells were quantified separately by ImageJ software (National Institutes of Health, Bethesda, MD). The corrected total cell fluorescence (CTCF) was determined by considering the integrated density of staining, area of the cell, and the background fluorescence for the different experimental conditions with the equation, CTCF ϭ integrated density Ϫ (area of selected cell ϫ mean fluorescence of background readings).
Chromatin Immunoprecipitation-Chromatin immunoprecipitation (ChIP) assays were done by using a ChIP assay kit from Millipore (Billerica, MA), following the manufacturer's protocol with a few exceptions. 5-8 ϫ 10 6 cortical neurons were used after treatment with or without A␤. Rabbit polyclonal anti-FoxO3a antibody was used to immunoprecipitate the protein-DNA complexes. The primers used for PCR amplification of the rat Puma promoter were 5Ј-AACTT-GCATTCTCGCAGCTT-3Јand 5Ј-GCTGCTCCCCAGTCT-CACT-3Ј. PCR products were analyzed on a 1.5% agarose gel and visualized by staining with ethidium bromide. Association of FoxO3a with the Puma gene was also quantified by quantitative RT-PCR.
Transfection-The plasmid maxikit (Qiagen) was used to isolate DNA. For the survival assay, cortical neurons were transfected with 0.5 g of either pSIREN-Puma-shRNA-zsgreen (shPuma) or pSIREN-Rand-shRNA-zsgreen (shRand). For the reporter assay, cells were cotransfected with 0.3 g of Puma-luc reporter, 0.1 g of Renilla vector, and 0.3 g of either shPuma or shRand. Transfections were done in 500 l of serum-free medium/well of a 24-well plate using Lipofectamine 2000. Six hours later, Lipofectamine containing medium was replaced by a fresh complete medium. Transfection was performed on the third day of culture. For endogenous FoxO3a down-regulation, naive PC12 cells were transfected with 1 g of either shFoxO3a or shRand. Transfections were done in 1 ml of serum-free medium/well of a 12-well plate using Lipofectamine 2000. 24 h post-transfection, cells were differentiated in the presence of NGF. After 5 days of priming, cells were treated with either A␤ (5 M) or left as untreated control.
Luciferase Assay-Cortical neurons were transfected with the constructs as mentioned above. Forty-eight hours after transfection, the cells were treated with or without A␤(1-42) for 24 h. The cells were lysed in passive lysis buffer (Promega). The dual luciferase assay was done according to the manufacturer's protocol using a luminometer (PerkinElmer Life Sciences). Relative luciferase activities were obtained by normalizing the firefly luciferase activity against Renilla luciferase activity.
Site-directed Mutagenesis-The mutated Puma promoter reporter was generated by incorporating mutations into the consensus sequence for the FoxO binding site by PCR-based site directed mutagenesis using Pfu Turbo DNA polymerase (QuikChange site-directed mutagenesis kit, Stratagene) according to the manufacturer's protocol and was verified by sequencing. The primers used for generating mutations were 5Ј-GGCGGGT-TTGTTTACAGGGAATGGGGTTCGGC-3Ј and 5Ј-GCCCG-AACCCCATTCCCTGTAAACAAACCCGCC-3Ј.
Sholl Analysis-Neural networks of cortical neurons were analyzed by Sholl analysis using ImageJ software as reported previously by Cuesto et al. (32), excepting few modifications as mentioned below. In brief, transfected cortical neurons either with shPuma or shRand were imaged at low magnification under a fluorescence microscope. Sholl analysis was performed on single neurons imaged at 0 and 72 h of A␤(1-42) treatment using the ImageJ software (Sholl analysis plugin). Several concentric circles were drawn from a point of the cell body with gradually increasing radius of 40 m in length. The number of branches that intersect the successive concentric circle was counted.
Survival Assay-Primary cortical neurons (5 days in vitro) were transfected with shRand or shPuma as mentioned above. After 48 h of transfection, the neurons were exposed to A␤ , and the number of transfected neurons (green) was counted (0 h). The number of surviving transfected neurons was also counted after 24, 48, and 72 h of treatment. Control and A␤-treated transfected neurons were imaged under a fluorescence microscope (Leica, Wetzlar, Germany).
The cell viability was also checked by the intact nuclear counting method. This assay was performed as described previously (33). In brief, a detergent containing the buffer was added to the cells that dissolve only the cell membrane, leaving the nuclear membrane intact. The intact nuclei were then counted on a hemocytometer. The number of live cells was expressed as a percentage of the total cell population.
Oligomeric A␤ Infusion in Animals-Male Sprague-Dawley rats (300 -380 g) were infused with oligomeric A␤ as described previously (4). Briefly, rats were anaesthetized by injecting a mixture of xylazine-ketamine and placed on a stereotaxic frame, and then a volume of 5 l of 100 M A␤ in PBS was infused in the right cerebral cortex at stereotaxic coordinates from bregma (AP, Ϫ4.1 mm; L, 2.5 mm; DV, 1.3 mm) according to the rat brain atlas. Control animals were injected with an equal volume of PBS. After 21 days of injection, animals were sacrificed. Following cardiac perfusion, the brains were dissected out and fixed in 4% paraformaldehyde for 24 h. The brains were further incubated in a 30% sucrose solution for 24 h, and then cryosectioning was done by using a cryotome (Thermo, West Palm Beach, FL).
Immunohistochemistry of Brain Slices-Cryosections of the brains from A␤-infused or PBS-infused rats and wild-type or transgenic mice were immunostained as described previously (4). In brief, sections were blocked with 5% goat serum in PBS containing 0.3% Triton X-100 for 1 h at room temperature, incubated in primary antibody in a blocking solution overnight at 4°C, washed with PBS, and then incubated with a fluorescence-tagged secondary antibody for 2 h at room temperature. After PBS wash and Hoechst staining for the nucleus, the sections were mounted and observed under fluorescence microscope.
Statistics-All experimental results are reported as mean Ϯ S.E. Student's t test was performed as unpaired, two-tailed sets of arrays to evaluate the significance of difference between the means and presented as p values.

Puma Is Induced in Neurons in Response to A␤ in Cultures and in
Vivo-Recent findings suggest that A␤ oligomers play an important role in the development of AD (34 -36). Cortical neurons are severely affected in AD brains and undergo death after exposure to oligomeric A␤ in vitro (5,26,28,29,(37)(38)(39). We have recently shown that the oligomeric A␤(1-42) at a concentration of 1.5 M induced significant neuron death after 24 h. Neuron death first becomes apparent by 12-16 h of A␤ exposure, and more than 40% of neurons die within 24 h (4). A time course reveals that Puma is up-regulated by oligomeric A␤  in cultured neurons. Puma transcript is increased by 2-fold within about 2-4 h of A␤ treatment, as detected by semiquantitative PCR (Fig. 1A) as well as by real-time PCR (Fig. 1B). Protein level is increased by more than 2-fold within 8 h of A␤ exposure (Fig. 1, C and D). Thus, Puma induction by A␤ precedes overt signs of neuron death. Next, we checked whether Puma level is also elevated in vivo in response to A␤. We and others have shown that adult rats infused with oligomeric A␤(1-42) in brain resulted in A␤ deposition, caspase-3 activation, and loss of neuronal cells in the vicinity of the infusion site of A␤(1-42) (4,40). In this study, the right hemispheres of the brains of adult rats were infused with either A␤  or PBS, and after 21 days, the animals were sacrificed, and brains were fixed, cryosectioned. The brain sections were co-immunostained with Puma and neuronal marker NeuN antibodies. Hoechst dye was used to stain nuclei. Results showed that Puma levels were greatly enhanced in neurons in A␤(1-42)-infused but not in PBS-infused rat brains (Fig. 2). The presence of A␤ deposition in adjacent sections of A␤(1-42) infused rat brains was verified by staining with Congo red (data not shown).
The use of synthetic A␤ may result in different effects compared with naturally secreted A␤. Therefore, we also examined the brain sections of A␤PPswe-PS1de9 (Swedish mutation in APP and PS1 mutation) transgenic mice for Puma expression. Brains of transgenic or control littermate mice were cryosectioned and stained with Congo red to confirm deposition of A␤ plaques (Fig. 3A). Then the brain sections were co-immunostained with Puma and NeuN antibodies. Nuclei were stained with Hoechst dye. Our results showed that the levels of Puma expression were much higher in transgenic mice compared with control littermates (Fig. 3B). Collectively, these findings clearly establish that Puma is induced in neurons in response to A␤  in vitro and in vivo.
Puma Knockdown Prevents Cortical Neuron Death upon A␤ Treatment-Next, we employed an shRNA-mediated knockdown strategy to assess whether Puma plays any role in A␤(1-42)-mediated neuron death. Cultured cortical neurons were transfected with a previously described shRNA construct against the Puma gene (shPuma) (12) along with a control shRNA construct (shRand), maintained for 48 h, and then subjected to A␤ treatment (1.5 M) for 72 h. The number of surviving cells (green) was counted under a fluorescence microscope. Down-regulation of Puma by shRNA provided significant protection of cortical neurons from death evoked by A␤ . A significant number of shPuma-expressing neurons survived with intact neurites even after 72 h of A␤ exposure compared with shRand-expressing neurons (Fig. 4, A  and B).
We have also quantitatively measured the preservation of neural networks by shPuma in A␤ treated cells by Sholl analysis (Fig. 4C). A single cortical neuron transfected with shPuma or shRand was analyzed by ImageJ as described under "Experimental Procedures." The analyses showed that shPuma-transfected cells retain most of the neurites even after 72 h of A␤ treatment compared with shRand (control)-transfected cells, where diminution of neural networks is evident. Taken together, these results suggest that Puma is required for neuron death and degeneration evoked by A␤ toxicity.
FoxO3a Regulates Puma Expression upon A␤ Toxicity-It is known that Puma is regulated in both a p53-dependent and -independent manner (11,17,18). We have also found that basal expression of Puma was down-regulated when p53 was inhibited by pifithrin-␣ (Fig. 5, A and B). However, inhibition of p53 did not completely block the A␤-mediated up-regulation of Puma. Fig. 5, A and B, shows that the increase in the level FIGURE 1. Puma is induced by A␤ in cortical neurons. A, rat cortical neurons (7 days in vitro) were subjected to oligomeric A␤ (1.5 M) for the indicated times, and total RNA was isolated, reverse-transcribed, and analyzed by semiquantitative PCR for Puma transcripts. ␣-Tubulin was used as a loading control. B, graphical representation of changes in Puma transcript level upon A␤ (1.5 M) treatment on primary cultured rat cortical neurons at the indicated times by real-time PCR. 18 S was used as a loading control. Data are presented as -fold increase relative to untreated control (C) and represent mean Ϯ S.E. of three independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01. C, cortical neurons were subjected to A␤ (1.5 M) for the indicated times, and total tissue lysates were analyzed by Western blot for Puma level. A representative immunoblot shows Puma protein level at the indicated time points. ERK2 was used as a loading control. D, graphical representation of -fold increase of Puma protein level after A␤ treatment at different time points expressed relative to untreated control. Data represent mean Ϯ S.E. (error bars) of three experiments. *, statistically significant differences from 0 h control; p Ͻ 0.05.

FIGURE 2. Puma is induced following A␤(1-42) infusion in vivo.
Right hemispheres of brains of adult rats were infused with either A␤  or PBS, and after 21 days, the animals were sacrificed, and brains were taken out following cardiac perfusion. The brains were cryosectioned and co-immunostained with Puma and NeuN antibodies; nuclei were stained with Hoechst dye. Representative images of five sections from three animals of each group with similar results are shown here. Scale bar, 100 m. Images were taken for each case by using an inverted fluorescence microscope and camera set to the same exposure. . Puma is elevated in AD transgenic mice brain. The brain sections of A␤PPswe-PS1de9 transgenic mice and wild-type mice were analyzed for level of Puma expression. A, transgenic and wild-type brain slices were stained with Congo red to see the deposition of A␤ plaques. These plaques (arrows) were seen in transgenic brains, whereas they were absent in wild-type brain tissue. B, the brain sections of transgenic mice were co-immunostained with Puma and NeuN antibodies. Hoechst was used to stain nuclei. Representative images from six sections from three animals of each group with similar results are shown here. Scale bar, 100 m. Inset, magnified version of the same section in each case. Images were taken for each case using an inverted fluorescence microscope and a camera set to the same exposure. of Puma by A␤ was only partially suppressed by pifithrin ␣, suggesting that other regulatory pathways independent of p53 are operating in the induction of Puma by A␤ (see below). Previous reports showed that the detrimental effect of A␤ can be abolished by pifithrin ␣ to a considerable extent (41). Under stress stimuli, p53 gets activated upon phosphorylation at serine 15 (42,43). The rapid increase of p53 phosphorylation at serine 15 evoked by A␤ is evident in our study and can be inhibited by pifithrin ␣ in the presence of A␤ (Fig. 5, C and D). We also checked whether pifithrin-␣ has any effect on cell viability in this model. We found that this compound significantly protected neurons from death induced by A␤ toxicity (Fig. 5E). These results suggest that p53 is induced and is required for neuron death in response to A␤. It is known that both A␤ and p53 induce BAX (2,44,45), which acts as a downstream effector of PUMA in the cell death cascade (46,47). Therefore, we checked the level of Bax and found that it was elevated in response to A␤ (Fig. 5, F and G).
Besides p53, it has been shown that dysregulated PI3K/Akt signaling in response to growth factor withdrawal activates transcription factor FoxO3a, which in turn up-regulates Puma (10,21,22). It has also been reported that the PI3K/Akt signaling pathway is inhibited by A␤ toxicity (48). Therefore, we investigated whether Puma is regulated by the PI3K/Akt/ FoxO3a pathway in A␤-treated neurons. First we checked whether inhibition of PI3K signaling is capable of inducing Puma. We found that PI3K inhibition by a specific PI3K inhibitor, LY294002, led to the elevation of Puma about 2-fold compared with control (Fig. 5, A and B). The extent of the increase was comparable with the Puma level induced by A␤ treatment. Interestingly, treatment with A␤ in presence of LY294002 did not enhance the Puma level further (Fig. 5, A and B). This result implies that Puma induction in A␤-treated neurons also occurs through inhibition of PI3K pathway. Because the inhibition of PI3K signaling is known to activate FoxO3a, we investigated the role of FoxO3a in the induction of Puma by A␤. Cultured cortical neurons were transfected with a previously described shRNA construct against FoxO3a (4) and then immunostained with Puma antibody. Results revealed that shFoxO3a-expressing cells had less intense staining of Puma compared with shRand-expressing cells upon A␤ treatment (Fig. 6A). A quantitative analysis showed that about 80% of shFoxO3a-transfected neurons and 20% of shRand-expressing neurons had low staining of Puma after 8 h of A␤ exposure (Fig. 6B). We further quantified the total cell fluorescence intensity of Puma in shFoxO3a-and shRand-expressing cells by ImageJ and found significantly less intense Puma staining in FoxO3a-expressing cells compared with shRand-expressing cells upon A␤ treatment (Fig. 6C). We also validated the down-regulation of Puma by shFoxO3a by Western blot. PC12 cells were transfected with shFoxO3a or shRand and then primed with NGF, and we checked the level of FoxO3a and Puma. In this condition, more than 60% cells were transfected, and FoxO3a was markedly down-regulated by shFoxO3a (Fig. 6, D and E). We found that the Puma level was significantly reduced by shFoxO3a-compared with shRand-transfected cells (Fig. 6, F and G). Collectively, these results suggest that Puma is a target of FoxO3a in A␤-treated neurons.
FoxO3a Directly Activates Puma Gene in Response to A␤-A FoxO response element (FHRE) in intron 1 of the Puma gene, which is conserved between human and mouse, has been reported (22). We have also found a similar FHRE in intron 1 of the rat gene. Therefore, we were interested to see whether FoxO3a directly binds with the Puma gene and is required for its activation during A␤-induced neuron death. First, we prepared a construct of luciferase reporter driven by a segment of the rat Puma gene that contains the FHRE (Fig. 7A). Cultured cortical neurons were co-transfected with the luciferase reporter and shRand or shFoxO3a and exposed to A␤. Puma promoter-driven luciferase activity was increased severalfold upon A␤(1-42) treatment for 24 h when neurons were cotransfected with shRand (Fig. 7B). However, this increase in luciferase activity in response to A␤(1-42) is diminished when neurons were co-transfected with shFoxO3a (Fig. 7B).
To further validate the dependence of Puma induction on FoxO3a, we generated mutations in the consensus FoxO3a binding site in the Puma promoter reporter construct. The mutations abolished the Puma-luc reporter activity induced by A␤ (Fig. 7C). Thus, our result suggests that FoxO3a is necessary for activation of the Puma promoter in response to A␤.
Next, we performed a ChIP assay to see whether FoxO3a directly binds with the Puma gene in response to A␤. Results showed that a significant amount of FoxO3a was bound to the endogenous Puma gene after 8 h of A␤(1-42) exposure, whereas there was negligible binding in control cells (Fig. 7D). As a negative control, an irrelevant antibody was used, which did not precipitate Puma DNA, and PCR of FoxO3a immunoprecipitate with primers specific for ␣-tubulin did not produce any product (data not shown). Quantitative PCR analyses of immunoprecipitated DNA also showed that there was a severalfold increase in binding of FoxO3a with the Puma promoter when cortical neurons were exposed to A␤ for 8 h compared with control conditions, where cells were not treated with A␤ (Fig. 7E). Taken together, our results indi- cate that Puma is induced in response to A␤, at least in part by FoxO3a.
Puma Cooperates with Bim in Neuron Death Induced by A␤-It has been reported that Bim and Puma cooperate in developmental apoptosis of lymphocytes (49) and autoreactive thymocytes (9) and tumor regression in response to tyrosine kinase inhibitors in breast and lung cancers (10). We therefore determined the effects of Bim and Puma knockdown, either alone or together, on A␤-induced neuron death. We used a previously described shRNA construct that effectively silences rat Bim (31) and protects neurons from A␤-induced death (5). Silencing both Bim and Puma led to a modest but significant increase in protection at 24 and 48 h of A␤ treatment (Fig. 8). These data indicate that Bim and Puma cooperate in neuron death evoked by A␤ and that Puma and Bim may activate separate apoptotic pathways.

DISCUSSION
Although recent research in multidisciplinary areas points out that interactions of multiple factors, including A␤, Tau, apoE4, and aging result in development of AD, it does not rule out the original hypothesis that accumulation of A␤ due to altered metabolism and clearance is central to the disease pathogenesis (1,24). Cultured neurons or animals exposed to aggregated A␤ and transgenic mice that produce more A␤ all manifest neurodegeneration and AD-like pathology. Recent radiological tools clearly demonstrate the accumulation of A␤ in the affected areas of patients' brain (50). However, reducing A␤ load by various means does not improve the conditions of patients in various clinical trials. These observations clearly indicate the need of alternative measures to counter the toxicity of A␤ and prevent the progressive loss of neurons by the residual A␤ that cannot be removed completely and safely by any A␤-reducing drug.
The BH3-only proteins of the Bcl-2 family that act as sentinels of apoptotic stimuli and are required for mitochondrial pathway of apoptosis could be appropriate targets to reduce the toxicity of A␤. In this report, we showed that the BH3-only protein Puma is induced by A␤ in vitro and in vivo and plays an essential role in AD-associated neurodegeneration. We have observed significant increase of Puma transcripts as well as its protein levels in cortical neurons upon A␤ treatment in a timedependent manner. The expression of Puma is also enhanced in AD transgenic mice and A␤-infused rat brains. We further observed that knockdown of Puma by the shRNA construct provides significant protection against neuron death and helps in retention of the neural network for longer time points. The cultures were maintained for 48 h and then subjected to overnight A␤ treatment, after which luciferase activity was assayed and represented as -fold change of luciferase activity. Data represent mean Ϯ S.E. of four experiments. *, p Ͻ 0.05. D, primary cultures of rat cortical neurons were treated with or without A␤ for 8 h. An equal number of cells were processed for ChIP assay using anti-FoxO3a antibody for immunoprecipitation. The immunoprecipitated materials were subjected to PCR using primers against the portion of the Puma promoter that flanks the FoxO3a-binding site. PCR products were verified by agarose gel electrophoresis. Templates were DNA from cells before ChIP (Input) or DNA from immunoprecipitated (IP) materials. PCR assays were conducted after ChIP, using samples from cells that were either left untreated (Control) or treated with A␤. E, graphical representation of FoxO3a association with the Puma gene. Quantitative PCR was performed using material derived from cultured cortical neurons treated as in D. Association of FoxO3a with Puma Forkhead response element (Puma FHRE level) in the presence or absence of A␤ was determined by quantitative PCR after ChIP, using samples from cells that were either left untreated (Control) or treated with A␤. Numbers on the y axis represent the levels of FoxO3a association with the Puma promoter region after normalizing to Ct values from input samples. Data shown are means Ϯ S.E. *, p Ͻ 0.05. Accumulating evidence also implicates Puma in several other neurodegenerative disorders. Puma is essential for neuron death evoked by 6-OHDA, a model of Parkinson disease (12). Deletion of Puma rescues motor neurons from apoptosis caused by endoplasmic reticulum stress in amyotrophic lateral sclerosis (51). It has also been reported that Puma is up-regulated in hippocampal CA1 neurons after transient global cerebral ischemia and thus could be a potential molecular target for therapy in stroke (52). Our study now implicates its involvement in AD-related neurodegeneration.
We have also investigated the regulatory pathways of Puma induction in response to A␤ treatments. Accumulating evidence suggests that Puma is regulated in both a p53-dependent and p53-independent manner in various systems (11,18). Consistent with these reports, we found that inhibition of p53 is not sufficient to block the induction of Puma by A␤, indicating the involvement of additional regulatory pathways. The survival pathway directed by PI3K-Akt is seen to be compromised by A␤ toxicity. This can be correlated with our recent findings that the activation and subsequent translocation of one of its downstream transcription factors, FoxO3a, to the nucleus induce proapoptotic protein Bim (4). The present findings indicate that FoxO3a also up-regulates another proapoptotic protein of the same family, Puma, in response to A␤ and causes consequent neuron death. Therefore, our results suggest that both p53 and FoxO3a regulate Puma expression.
In this context, we thought it would be interesting to know whether Bim and Puma act cooperatively in A␤-induced neuron death. Cooperative function of Bim and Puma has been reported in various apoptotic paradigms. Bean et al. (10) have reported that Bim and Puma mediate oncogene inactivationinduced apoptsosis in vitro and in vivo. They found that full apoptotic response initiated by oncogene deletion required both Bim and Puma. Similarly, it has been found that the combined loss of Bim and Puma is required for accumulation of immature thymocytes (9). They found that Bim Ϫ/Ϫ /Puma Ϫ/Ϫ mice had more extensive splenomegaly and lymphadenopathy than Bim Ϫ/Ϫ mice. In agreement with these findings, we found that combined knockdown of Bim and Puma provided significantly increased protection against A␤-induced death up to 48 h of treatment compared with that of single knockdown of either of the proteins. Thus, our results indicate that both Bim and Puma are required for A␤-induced neuron death. However, it remains to be examined whether they activate separate apoptotic pathways.
In comparison with the prominent role of Puma in p53-dependent apoptosis, the function of it in p53-independent cell death remains to be fully addressed. It has already been shown that under stress conditions like serum deprivation or cytokine withdrawal, Puma gets activated by FoxO3a in a p53-independent fashion (19,22,53). Mast cells deficient in FoxO3a were markedly resistant to cytokine withdrawal compared with wildtype cells (54). Puma knock-out mice were resistant to apoptosis induced by the withdrawal of interleukin IL-3 and IL-6 (18,55). In p53-deficient cancer cells, Puma is induced in response to serum starvation, and cells are resistant to apoptosis after knockdown of Puma (56). Puma is also known to be induced by transcription factors other than FoxO3a like CHOP, E2F1, or even Trb3 (20,57,58). We are currently in search of other probable signaling pathways that may regulate Puma in cooperation with p53 and FoxO3a in response to A␤.
Considering our recent report (4) and the findings of this study, we propose a model for A␤-induced neuron death (Fig.  9). Under A␤-treated conditions, p53 is activated, which partly induces Puma. FoxO3a is also activated as the PI3K/Akt pathway is inhibited in A␤-treated neurons. Activated FoxO3a translocates to the nucleus and binds directly to Bim and Puma genes, leading to their enhanced expression. Puma in cooperation with Bim activates the intrinsic apoptotic pathway that results in neuron death. In view of the fact that both Bim and Puma can be regulated by various transcription factors in other systems (17,59), there is a possibility that multiple transcription factors may be required for activation of both Bim and Puma in response to A␤. Further studies are required to validate this assumption and evaluate if those factors work independently or synergistically to activate Bim and Puma in A␤-induced neuron death.