Peroxisome Proliferator-activated Receptor γ Up-regulates the Bcl-2 Anti-apoptotic Protein in Neurons and Induces Mitochondrial Stabilization and Protection against Oxidative Stress and Apoptosis*

Peroxisome proliferator-activated receptor γ (PPARγ) has been proposed as a therapeutic target for neurodegenerative diseases because of its anti-inflammatory action in glial cells. However, PPARγ agonists preventβ-amyloid (Aβ)-induced neurodegeneration in hippocampal neurons, and PPARγ is activated by the nerve growth factor (NGF) survival pathway, suggesting a neuroprotective anti-inflammatory independent action. Here we show that the PPARγ agonist rosiglitazone (RGZ) protects hippocampal and dorsal root ganglion neurons against Aβ-induced mitochondrial damage and NGF deprivation-induced apoptosis, respectively, and promotes PC12 cell survival. In neurons and in PC12 cells RGZ protective effects are associated with increased expression of the Bcl-2 anti-apoptotic protein. NGF-differentiated PC12 neuronal cells constitutively overexpressing PPARγ are resistant to Aβ-induced apoptosis and morphological changes and show functionally intact mitochondria and no increase in reactive oxygen species when challenged with up to 50 μm H2O2. Conversely, cells expressing a dominant negative mutant of PPARγ show increased Aβ-induced apoptosis and disruption of neuronal-like morphology and are highly sensitive to oxidative stress-induced impairment of mitochondrial function. Cells overexpressing PPARγ present a 4- to 5-fold increase in Bcl-2 protein content, whereas in dominant negative PPARγ-expressing cells, Bcl-2 is barely detected. Bcl-2 knockdown by small interfering RNA in cells overexpressing PPARγ results in increased sensitivity to Aβ and oxidative stress, further suggesting that Bcl-2 up-regulation mediates PPARγ protective effects. PPARγ prosurvival action is independent of the signal-regulated MAPK or the Akt prosurvival pathways. Altogether, these data suggest that PPARγ supports survival in neurons in part through a mechanism involving increased expression of Bcl-2.

inflammation (1). In addition, PPAR␥ is the target of the insulin-sensitizing thiazolidinediones (TZDs) drugs, used to treat type II diabetes. Recent studies suggest that treatment of insulin resistance with a PPAR␥ agonist retards the development of Alzheimer disease (2), and TZDs extend survival in a transgenic mouse model of amyotrophic lateral sclerosis (3). TZDs have been proposed as potential therapeutic agents for both Alzheimer disease and multiple sclerosis (4). Most of the neuroprotective effects of TZDs are ascribed to either improved insulin sensitivity or to their anti-inflammatory action through PPAR␥ activation in glial cells (5)(6)(7). However, activation of PPAR␥ by three different TZDs protects rat hippocampal neurons against ␤-amyloid (A␤)-induced damage (8), and the TZD rosiglitazone (RGZ) protects human neuroblastoma SH-SY5Y cells against acetaldehyde-induced cytotoxicity (9). In addition, activation of the nerve growth factor (NGF) survival pathway increases PPAR␥ transcriptional activity and modulates its expression (10), further suggesting that PPAR␥ may be involved in cell survival independently of its anti-inflammatory action.
NGF, a member of the neurotrophin family of growth factors, exerts its effects through activation of its tyrosine kinase TrkA receptor and the neurotrophin receptor p75 (11,12); it is required for sympathetic neurons survival, both in vitro and in vivo (13,14). TrkA is essential to the survival-promoting effect of NGF (15), and PPAR␥ modulation by NGF is dependent on TrkA and independent of p75 (10). Downstream effectors of the NGF signaling pathway such as the extracellular signal-regulated MAPKs 1 and 2 (ERK1/2), and Akt/protein kinase B, induce PC12 cell differentiation to sympathetic-like neurons; inactivate components of the cell death machinery, and up-regulate anti-apoptotic proteins, such as Bcl-2, which increases mitochondrial membrane potential, thereby, protecting cells against apoptosis (11,16,17). Bcl-2-overexpressing neural cells display a reduced redox state, low levels of reactive oxygen species (ROS), and a resistance to mitochondrial injury and cell death induced by oxidative stress (18).
The ERK1/2 pathway is though to be responsible for Bcl-2 up-regulation by activating the CREB (cAMP response element-binding protein) transcription factor (19,20). Sustained activation of ERK1/2 signaling is involved in NGF-induced differentiation and survival, whereas transient activation of this pathway by epidermal growth factor results in an opposite proliferative effect (21). Because NGF but not epidermal growth factor induces PPAR␥ transcriptional activity (10), TZD activation of PPAR␥ can enhance mitochondrial potential (22), and a putative PPAR response element (PPRE) has been reported in the 3Ј-untranslated region of the bcl-2 gene in human colon cancer cells (23), we sought to define whether PPAR␥ is involved in cell survival by modulating Bcl-2 expression. We used hippocampal neurons and dorsal root ganglion (DRG) neurons and PC12 cells, a widely accepted model of NGF signaling (24 -26). We provide evidence that PPAR␥ up-regulates Bcl-2 through a novel ERK1/2-independent mechanism, preventing neuronal degeneration induced by both oxidative and A␤ stress, with a concomitant increase in mitochondrial stability.

EXPERIMENTAL PROCEDURES
Materials-Chemicals and drugs were from Sigma, and culture media and sera were from Invitrogen. Rosiglitazone and GW9662 were from Cayman Chemical (Ann Arbor, MI). Stock solutions of drugs were prepared in Me 2 SO and added to the culture medium (0.01% final Me 2 SO concentration). Synthetic A␤  peptide, corresponding to the human A␤ wild-type sequence, was obtained from Bachem Inc. (Torrance, CA). The fluorescent probes 2,7-DCF and MitoTracker Orange, and TO-PRO-3 were from Molecular Probes (Leiden, The Netherlands). NGF 2.5 S was from Alomone (Jerusalem, Israel). Antibodies were from Affinity Bio-Reagents (Golden, Co; anti-PPAR␥), Santa Cruz Biotechnology (San Diego, CA; anti-bcl-2, anti-neurofilaments), ABCAM (Cambridge, UK; anti-␤-actin), Cell Signaling Technology (Beverly, MA, anti-phospho ERK1/2 and anti-total ERK1/2, anti-total Akt, and anti-phospho Thr-305 Akt), and Jackson ImmunoResearch (West Grove, PA, Rhodamine Red-X donkey anti-rabbit IgG and Cyanine 5 donkey antigoat IgG).
Reporter Gene Assay and Plasmids-PPAR reporter gene assays were performed as previously described (10). Full-length murine PPAR␥1 and dominant negative murine PPAR␥1 (10) were cloned into a pcDNA3 vector for generating stably PPAR␥-expressing PC12 cells. PC12 cells, stably expressing the pcDNA3 vector alone, were used as control.
Cell Culture and Generation of Cell Lines-PC12 cell cultures (24) were maintained, as previously described (10). Stable transfection was carried out using Lipofectamine 2000 (Invitrogen), according to the manufacturer's instructions. PC12 cells were separately transfected with 0.2 g each of the PPAR␥-pcDNA3 (mP␥ϩ cells) or the dominant negative PPAR␥-pcDNA3 (mP␥Ϫ cells) vector. Stably transfected PC12 cells were selected under 0.5 mg/ml Geneticin for 2 weeks until resistant colonies appeared. Clones of each transfection were selected on the basis of their relative level of murine PPAR␥ mRNA expression, as assessed by RT-PCR, using specific murine PPAR␥ primers that do not recognize endogenous rat PPAR␥ mRNA in PC12 cells (forward, 5Ј-CTG GCC TCC  CTG ATG AAT AA-3Ј; reverse, 5Ј-CAG ATC GTC ACA  GAG CAC GT-3Ј).
Primary Neuronal Cultures-Rat embryonic hippocampal cells were obtained and processed for immunofluorescence as previously described (8). In brief, cells were seeded in polylysine-coated wells. On day 3 of culture they were treated with 2 M 1-␤-D-arabinofuranosylcytosine for 24 h to reduce almost all the glial cells present in the culture. Seven days later, hippocampal cells were used for experiments. The average number of neurons in each experiment corresponded to 98% of total cells present in the cultures. DRG neuronal cultures were obtained using published procedures with modifications (27,28). Briefly, newborn DRGs (lumbar, cervical, and thoracic) were removed rapidly, placed in Ham's F-12 medium (Invitrogen), and digested in 0.125% collagenase (Sigma-Aldrich) in Ham's F-12 medium for 25 min followed by 20 min in 0.25% trypsin (Invitrogen), all at 37°C. Ganglia were washed in Ham's F-12 medium, and then gently triturated through a glass-polished pipette. The cell suspension was centrifuged at 5000 rpm for 5 min. The cell pellet was resuspended in Ham's F-12 medium containing 100 units/ml penicillin, 100 g/ml streptomycin, and 50 ng/ml NGF, and cells were plated immediately on T-75 flasks (Nunc International, Rochester, NY). After 15-20 h at 37°C and 5% CO 2 , non-neuronal cells were firmly attached to the flask, whereas neurons were weakly adherent to the dish or to flattened non neuronal cells (28). Medium was carefully removed, and the attached neurons were selectively dislodged with a gentle stream of Ham's F-12 culture medium. Further neuronal enrichment was achieved by centrifugation of the cell suspension at 1000 rpm for 5 min. Cells were maintained for 48 h in Ham's F-12 containing 10% horse serum (Invitrogen), 100 units/ml penicillin, 100 g/ml streptomycin, 50 ng/ml NGF, 5 M cytosine arabinoside (Sigma-Aldrich), and 20 M fluorodeoxyuridine (Sigma-Aldrich-Fluka) at 37°C, 5% CO 2 , either in 3.5-cm dishes (real-time PCR) or polylysinecoated covers (immunofluorescence studies). After this step over 65% of cells were neurons, as assessed by counting cells with round, phase-bright bodies and intact neurites. Culture medium was replaced with medium without mitotic inhibitors or NGF, maintained for 24 h, and treated as follows. For immunofluorescence, cells were treated for 18 h with RGZ 0.1 or 1 M or vehicle (Me 2 SO, 0.01% final concentration). For real-time PCR, cells were treated for the same time in the absence or presence of 50 ng/ml NGF, and the presence of 0.1 or 1 M RGZ, or vehicle. For all treatments cells were cultured in Ham F-12 medium containing 5% horse serum.
Immunofluorescence Studies and Image Analysis-Hippocampal or DRG neurons were immunostained with anti Bcl-2 or anti neurofilaments antibodies, incubated with appropriate secondary antibodies, and covers were mounted and analyzed using a LSM Pascal Zeiss model 510 confocal microscope (Carl Zeiss Ltd., Germany). Bcl-2 mean fluorescence intensity, under non-saturated conditions, was determined in at least 50 NFH positive control or RGZ-treated cells in each experiment, in digital images from random fields using the Image-Pro express software (Media Cybernetics Inc., Silver Spring, MD).
RNA Interference-Custom on-target plus SMARTpool small interfering RNA (siRNA) to target rat Bcl-2 (GenBank TM accession number NM_016993) was designed and synthesized by Dharmacon (Chicago, IL). A non-targeting siRNA duplex, used as control, and a modified fluorescence RNA duplex (siGLO Green transfection indicator), used to assess transfection efficiency, were also from Dharmacon (Chicago, IL). PC12 cells were seeded in polylysine-coated coverslips in 24-well plates (30,000 cells). After 24 h cells were transfected with 50 pmol of either Bcl-2 siRNA, control siRNA, or siGLO green RNA, using DharmaFECT 1 transfection reagent (Dharmacon) according to the protocol provided by the manufacturer, in PC12 basal medium containing 2% horse serum.
Apoptosis and Cell Viability Detection-In hippocampal neurons the MitoCapture TM Apoptosis Detection Kit (Molecular Probes, Eugene, OR) was used to distinguish between healthy and apoptotic neurons by detecting the changes in the mitochondrial membrane potential. In healthy neurons, Mito-Capture TM accumulates and aggregates in the mitochondria, emitting a bright red fluorescence. In apoptotic cells, this reagent cannot aggregate in the mitochondria due to the altered mitochondrial membrane potential, and thus it remains in the cytoplasm in its monomeric form and emits green fluorescence. Fluorescence changes and images were acquired and analyzed using confocal microscopy as described above. The filters used were BP 505-530 and HFT 488 for green fluorescence and LP 560 and HFT488/543 for red fluorescence. Laser excitations were adjusted for 488 nm to 5% and for 543 nm 60%, respectively. In PC12 cells, mitochondrial membrane potential was estimated with the specific mitochondrial probe MitoTracker Orange, detected in the confocal microscope. PC12 cells were grown on poly-L-lysine-coated glass coverslips and cultured for 5 days. The cells were then loaded for 30 min with MitoTracker Orange in Krebs-Ringer-Hepes-glucose, containing 0.02% pluronic acid; they were then washed and allowed to equilibrate for 30 min. Coverslips were then mounted in a chamber on the stage of the confocal laser scanning microscope. The fluorescence changes, determined by MitoTracker Orange fluorescence, indicated the mitochondria potential changes (31,32). Images were acquired using a 543-nm He-Ne laser to excite MitoTracker Orange, and the signals were collected at 570 nm. Signals from control cells and cells treated with the complex were compared using identical settings for laser power, confocal thickness, and detector sensitivity. The images were analyzed, and the mean MitoTracker signal was measured per live cell. Estimation of fluorescence intensity was presented like the pseudoratio (⌬F/F), indicated by the following formula: ⌬F/F ϭ (F Ϫ F base )/(F base Ϫ B), where F is the measured fluorescence intensity of the indicator, F base is the fluorescence intensity before the stimulation, and B is the background signal determined from the average of areas adjacent to the cells (33).
In DRG neurons, changes in nuclear morphology characteristics of apoptosis were visualized using Hoechst 33342 (Sigma) staining in cells also immunostained with the anti-neurofilaments (NFH) antibody to distinguish between DRG neurons and contaminating glial cells. At least 10 fields containing 10 -50 NFH-positive cells were photographed in a fluorescent microscope. Apoptotic nuclei in NFH-positive cells were determined in blinded photograph. In each experimental condition, at least 300 NFH-positive cells were analyzed.
ROS Measurements-PC12 cells were incubated with the fluorescent probe 2,7-DCF (10 M) for 30 min (34). Then, wells were washed three times with phosphate-buffered saline solution; 50 M H 2 O 2 was added to the wells, and cells were incubated for 4 h. Cells were photographed in a Confocal Microscope. Photographs were analyzed by the LSM Image Program (Zeiss Co.). Fluorescence signal of treated cells was obtained using a 488 nm argon laser to excite 2,7-DCF fluorescence, and signals were collected at 505-530 nm.

The PPAR␥ Agonist Rosiglitazone Up-regulates Bcl-2 and Protects DRG Neurons from NGF Withdrawal-induced
Apoptosis-To determine whether the protective effect of RGZ, previously described in neuronal cells, might be attributed to PPAR␥-induced up-regulation of Bcl-2, we first determined the effect of RGZ in Bcl-2 protein content and mRNA expression in DRG neurons. We used double immunofluorescence staining for Bcl-2 and the neuronal marker NFH. Because DRG cultures are contaminated by non-neuronal cells (30 -40% under our conditions), quantitative immunofluorescence of Bcl-2 was assessed in NFH-positive cells. RGZ (1 M) induced increased Bcl-2 expression in DRG neurons (Fig. 1A). Quantification of Bcl-2 immunostaining in NFH-positive cells is shown in Fig. 1B. RGZ induced a concentration-dependent increase in Bcl-2 immunostaining, which was 3-to 4-fold higher than the one observed in control cells, for 1 M RGZ-treated cells. No difference in NFH immunoreactivity was found between control and RGZ-treated DRG cells (not shown). This result shows that RGZ induces increased Bcl-2 expression specifically in DRG neurons, discarding the possible contribution of contaminating non-neuronal cells. To gain some evidence of whether Bcl-2 mRNA expression was increased in RGZ-treated DRG cultures we performed real-time RT-PCR, using as a positive control NGF, which is known to increase Bcl-2 in neurons (20). NGF-induced a 2-fold increase in Bcl-2 mRNA (Fig. 1C). A similar increase was observed in RGZ-treated cells (1 M), suggesting that it might correspond to neuronal Bcl-2 mRNA. To assess whether RGZ-induced up-regulation of Bcl-2 was reflected in decreased sensitivity to apoptosis we used NGF withdrawal, a widely used model to induce apoptosis in embryonic DRG neurons (35)(36)(37). Changes in nuclear morphology characteristics of apoptosis were visualized using Hoechst staining in NFHpositive cells and quantified as described under "Experimental Procedures" (Fig. 1D). In the presence of NGF, RGZ alone was without effect in the number of apoptotic cells, whereas in NGF-deprived cells a 3-fold increase in apoptotic cells, amounting to ϳ60% of total NFH-positive cells was observed, which was completely reversed in the presence of RGZ. These results show that RGZ prevents apoptotic death induced by NGF withdrawal in embryonic DRG neurons and suggest that PPAR␥ is participating by up-regulating Bcl-2.

The PPAR␥ Agonist Rosiglitazone Up-regulates Bcl-2 and Prevents A␤-induced Mitochondrial Damage in Hippocampal
Neurons-Next, we determined whether our previous findings may be extended to other neuronal cells. We used primary cultures of embryonic hippocampal cells, because we have previously shown that several TZDs, including RGZ, protects this neurons from A␤-induced loss of viability (8). Immunofluorescence against Bcl-2 revealed a significant RGZ-induced increase in Bcl-2 immunostaining intensity (1.8 Ϯ 0.3-fold, p Ͻ 0.01; quantified as described under "Experimental Procedures"), whereas NFH immunostaining intensity remained unchanged (Fig. 1E). Similarly, RGZ also up-regulated Bcl-2 mRNA, as determined using real-time PCR (Fig. 1F) as well as Bcl-2 protein, as assessed by Western blots (Fig. 1G). Increased Bcl-2 mRNA and protein expression can be ascribed to hippocampal cells, because in contrast to DRG neurons, roughly 98% of the population in this primary culture are hippocampal cells (8) (see "Experimental Procedures"). In this relatively pure population, it was also possible to directly assess whether RGZinduced Bcl-2 up-regulation was reflected in changes in mitochondrial membrane potential, an important parameter of Mitochondrial viability changes were measured by using the MitoCapture TM assay kit. In apoptotic cells MitoCapture TM cannot aggregate in the mitochondria due to the altered mitochondrial membrane potential, and thus it remains in the cytoplasm in its monomeric form and emits green fluorescence, whereas in healthy neurons, this reagent accumulates and aggregates in the mitochondria, emitting a bright red fluorescence (see "Experimental Procedures"). RGZ induced no significant differences in the mitochondrial membrane potential (compare control with RGZ-treated cells). Nevertheless, a severe reduction in the mitochondrial viability was observed in the cultures treated with A␤, as assessed by a sustained loss of red fluorescence pattern and an increase in green fluorescence levels, an effect that was reverted by co-treatment with RGZ. Confocal images are representative from three independent experiments. mitochondrial function, which is known to be stabilized by Bcl-2, thus preventing apoptosis (18). Using the Mito-Capture TM assay kit (see "Experimental Procedures"), we found no significant changes in mitochondrial membrane potential in RGZ-treated hippocampal neurons, when compared with control cells, whereas a severe loss was induced by A␤, which was reversed in neurons co-treated with RGZ (Fig. 1H). These results are consistent with previous data and with the known Bcl-2-induced stabilizing effect on mitochondria, which results in protection of neurons from both physiological and toxic insult (38,39). They suggest a possible involvement of PPAR␥ in neuroprotection trough up-regulation of Bcl-2, which would result in mitochondrial stabilization and prevention of apoptosis.
PPAR␥ Regulates Bcl-2 Expression and Prevents A␤-induced Damage in PC12 Cells-Because TZDs presents both PPAR␥dependent and independent effects (40 -44), we used a gainand loss-of-PPAR␥ function approach, using PC12 cells, to support our previous conclusions. First, we determined whether RGZ has similar effects in PC12 cells and primary neuron cultures. RGZ (1 M) was found to up-regulate both Bcl-2 mRNA expression ( Fig. 2A) and protein content (Fig. 2B), in a time-dependent manner. Also consistent with previous results, PC12 cells cultured for 24 h in the presence of 1 M RGZ were protected from A␤-induced degeneration, as assessed by the MTT assay (Fig. 2C). Further, the PPAR␥ antagonist GW9662 suppressed this effect (Fig. 2C).
Second, PC12 cell clones stably expressing either mouse PPAR␥ (mP␥ϩ cells) or its dominant negative form (mP␥Ϫ cells) (10), were produced by Lipofectamine 2000 transfection and Geneticin selection (see the "Experimental Procedures"). Positive clones were chosen on the basis of mouse PPAR␥ mRNA expression, as assessed by RT-PCR (Fig. 3A, upper  panel). To establish the functionality of the PPAR␥ clones, we determined PPAR␥ transcriptional activity at increasing concentration of the PPAR␥ agonist RGZ, using a PPAR reportergene assay (Fig. 3A, lower panel). Overexpression of PPAR␥ (mP␥ϩ cells) induced an RGZ concentration-dependent increase of PPAR␥ transcriptional activity, which was 6-to 7fold higher than the one observed in controls cells expressing the vector alone (control (Ct) cells). Cells expressing the dominant negative form of PPAR␥ (mP␥Ϫ cells) showed almost no activity, indicating that the PPAR␥ mutant effectively blocks PPAR␥ transcriptional activity.
Next, we determined the expression of Bcl-2 mRNA in control and PC12 clones using real-time PCR. As shown in Fig. 3B, Bcl-2 mRNA expression was increased by ϳ2-fold in mP␥ϩ cells, whereas in mP␥Ϫ cells, it was decreased to less than half of the amount present in Ct cells. Bcl-2 protein content was also  In A, upper panel, procedure for PC12 clones selection. Expression of mouse PPAR␥ (mP␥ϩ), or its dominant negative mutant (mP␥Ϫ) was assessed by RT-PCR, using specific primers for mouse PPAR␥ that do not recognize rat PPAR␥. Control (Ct) cells are also shown. Total mouse adipose tissue mRNA was used as-positive control (AdT). Lower panel, PPAR␥ transcriptional activity in mP␥ϩ, mP␥Ϫ, or Ct cells, in response to increasing RGZ concentration. PC12 clones were transfected with a PPAR reporter gene and a cytomegalovirus-␤-Gal vector for normalization and assayed for luciferase activity. Transcriptional activity is expressed relative to Ct cells, in the absence of RGZ. No transcriptional activity is observed in mP␥Ϫ cells, whereas mP␥ϩ cells present ϳa 7-fold increase in transcriptional activity. B, overexpression of PPAR␥ increases Bcl-2 mRNA, whereas expression of its dominant negative has the opposite effect. Total mRNA from PC12 cell clones, overexpressing PPAR␥ (mP␥ϩ), its dominant negative (mP␥Ϫ), or Ct cells were submitted to realtime PCR. A significant decrease of Bcl-2 expression is observed in mP␥Ϫ cells, whereas an almost 2-fold increase is observed in mP␥ϩ cells. Results are presented as mean Ϯ S.D. of three independent experiments, performed in triplicate. C, overexpression of PPAR␥ increases, while its dominant negative decreases Bcl-2 protein content. A representative Western blot of proteins extracted from two separate flasks of Ct, mP␥Ϫ, and mP␥ϩ cells is presented. ␤-Actin was used as the loading control. D, quantification of three separate Western blot experiments in duplicate flask is presented as mean Ϯ S.D., relative to Ct cells. Similar up-regulation or down-regulation was observed in NGF-differentiated mP␥ϩ and mP␥Ϫ cells, respectively, when compared with NGF-differentiated control cells (not shown).
up-regulated by ϳ5-fold in mP␥ϩ cells, whereas it was barely detected in mP␥Ϫ cells, as assessed by Western blot (Fig. 3C, quantified in Fig. 3D). A similar up-regulation and repression of Bcl-2 was found, respectively, in NGF-differentiated mP␥ϩ and mP␥Ϫ cells (data not shown).
PPAR␥ Overexpression Protects PC12 Cells from A␤-induced Damage-A␤-peptide has been shown to induce neuronal stress, reflected in lost of neurites and decreased survival. To determine whether PPAR␥-induced protection against damage is reflected in A␤-induced neuronal stress, we used NGF-differentiated Ct, mP␥ϩ, and mP␥Ϫ PC12 cells. NGF induced differentiation in mP␥ϩ and mP␥Ϫ PC12 clones, as determined by morphological examination (Fig. 4A). NGF differentiated Ct, and the NGF-differentiated mP␥ϩ and mP␥Ϫ clones were treated with 5 M A␤. After 4-h treatment, NGF-differentiated Ct and mP␥Ϫ cells have lost their morphology, revealing extensive neurite disruption, whereas almost no morphological changes were observed in NGF-differentiated mP␥ϩ cells (Fig.  4A). Quantification of mean neurite length after A␤ treatment confirmed this observation, showing that mP␥Ϫ cells are significantly more sensitive than Ct cells to A␤-induced morphological changes (Fig. 4B). Similarly, NGF-differentiated mP␥ϩ cells also presented increased resistance to A␤-induced apoptosis, as determined by the TUNEL assay (Fig. 4C). More than 95% of mP␥ϩ cells were resistant to A␤-induced apoptosis, whereas NGF-differentiated mP␥Ϫ and Ct cells were almost completely apoptotic under the same conditions, although we observed a significant increased sensitivity in NGF-differentiated mP␥Ϫ cells, when compared with control cells (Fig. 4D).

PPAR␥ Overexpression Prevents Oxidative Stress-induced Neurodegeneration and Increases Mitochondrial Stability-Be-
cause Bcl-2 anti-apoptotic protein is known to regulate survival in a number of cell types, we investigate whether overexpression or repression of PPAR␥ differentially affects PC12 cell survival. Several reports imply oxidative insult and impairment in mitochondrial function in A␤-induced degeneration in AD and other neurodegenerative diseases (45) (Fig. 5B). Next, we investigated whether mP␥ϩ cells resistance to oxidative stress is reflected in mitochondrial stability. Control (Ct), mP␥Ϫ and mP␥ϩ cells were loaded with a MitoTracker (see "Experimental Procedures") and exposed to 50 M H 2 O 2 . A clear time-dependent decrease in fluorescence is detected in Ct and mP␥Ϫ cells, which is more severe for mP␥Ϫ cells after 30 min of H 2 O 2 exposure (Fig. 5C). In contrast, mP␥ϩ cells presented no detectable loss of fluorescence. To quantify this observation, Ct and PC12 cell clones were loaded with 250 nM MitoTracker and set for real-time confocal microscopy. After a 5-min control period, 50 M H 2 O 2 was added, and changes in fluorescence were registered for 30 min (Fig. 5D). A strong decrease in mP␥Ϫ cell fluorescence was observed almost immediately after H 2 O 2 addition, whereas the Ct PC12 cells presented a less pronounced decrease. In agreement with confocal images, mP␥ϩ cell fluorescence remained unchanged during the whole incubation period, whereas Ct cells presented an intermediate behavior. These results show that PPAR␥ induces PC12 cells mitochondrial stabilization and decreases ROS generation, most probably by up-regulating Bcl-2. . A␤ has no effect on mP␥ϩ cell morphology, whereas a severe loss of neurites occurs in mP␥Ϫ and Ct cells. B, quantification of neurites. Results are presented as mean Ϯ S.D. (bars) from data obtained from 10 different micrographs, as shown in A. *, indicate statistical significance (p Ͻ 0.001, Student's t test). **, indicates statistical difference from Ct A␤-treated cells (p Ͻ 0.01). C, NGF-differentiated mP␥ϩ cells show increased resistance to A␤-induced apoptosis. NGF-differentiated Ct, mP␥Ϫ, and mP␥ϩ cells were exposed to A␤ 5 M for 4 h and then were stained for nuclei and for apoptotic cells using a TUNEL assay kit. Less than 5% of mP␥ϩ cells were apoptotic, whereas 90 -95% of Ct and mP␥Ϫ cells presented the apoptotic staining, as quantified in D, as mean Ϯ S.D. (bars). Similar results were obtained in undifferentiated PC12 cells and PPAR␥ clones.

Bcl-2 Knockdown Using siRNA Decreases Survival of PPAR␥overexpressing PC12 Cells in Response to A␤ and Oxidative
Stress-To obtain further evidence that Bcl-2 up-regulation mediates enhanced survival in response to PPAR␥ we targeted Bcl-2 using siRNA in mP␥ϩ cells, which present increased levels of Bcl-2 and show increased resistance to oxidative insult. Transfection efficiency, determined using a control fluorescent RNA duplex (see the "Experimental Procedures"), was over 80%, as measured either at 24 or 48 h after transfection for both mP␥ϩ cells or NGF-differentiated mP␥ϩ cells (not shown). Bcl-2 protein expression slightly decreased 24 h after transfection, was almost suppressed 48 h after transfection, and recovered 96 h after transfection, as assessed by Western blot (Fig.  6A). No significant changes were observed in mP␥ϩ cells transfected with a control siRNA, when compared with non-transfected cells (Fig. 6A). As shown in Fig. 6B (quantified in Fig. 6C) 5 M A␤ or H 2 O 2 (100 and 200 M) strongly induced apoptosis in mP␥ϩ cells, 48 h after transfection with Bcl-2 siRNA, as assessed using the TUNEL assay, whereas in cells transfected with the control siRNA, A␤ was without effect and 100 and 200 M H 2 O 2 treatment induced a 30% and 60% apoptosis, respectively, when compared with untreated cells. It is worth noting that in untreated cells Bcl-2 knockdown alone induced a significant increase in apoptotic cells, which amounted to ϳ20% of total cells, whereas in control cells it was less than 10% of total cells. The increased sensitivity of Bcl-2 siRNA transfected mP␥ϩ cells to A␤, and H 2 O 2 -induced damage is comparable to that already observed in mP␥Ϫ cells and is consistent with their low Bcl-2 content. As for mP␥ϩ cells, transfection of NGFdifferentiated mP␥ϩ cells with Bcl-2 siRNA resulted in almost total suppression of Bcl-2 expression 48 h after transfection (not shown). However, a severe detachment of cells was observed, whereas transfection with the control siRNA was without comparative effect. ϳ50% of the remaining Bcl-2 siRNA-transfected cells were apoptotic, as assessed using the TUNEL assay. Exposure to 50 M H 2 O 2 resulted in 100% of apoptosis, treatment that was without effect in control siRNAtransfected cells, which presented Ͻ10% spontaneous apoptosis 48 h after transfection (not shown). Altogether, data in this section support the view that Bcl-2 up-regulation mediates enhanced survival in response to PPAR␥, because ablation of Bcl-2 effectively suppressed the protecting effect induced by PPAR␥ overexpression.
RGZ or PPAR␥ Overexpression Does Not Activate the ERK1/2 or the Akt Survival Pathways-Phosphorylation and activation of ERK1/2 by downstream effectors of the NGF/TrkA signaling pathway is also known to increase Bcl-2 expression (19,20). Because NGF/TrkA signaling also induces PPAR␥ transcriptional activity (10), and MAPKs interact with PPAR␥ (46,47), it is conceivable that activated PPAR␥ might induce activation of this MAPK and Bcl-2 up-regulation. On the other hand, another of the main anti-apoptotic pathways in neurons is the Akt/protein kinase B-dependent pathway. Akt activation is governed by phosphorylation at Thr-308, which is mediated by multiple signaling routes (recently reviewed in Ref. 48, and results in the inactivation of pro-apoptotic pathways and in up-regulation of Bcl-2 (49,50).
On this basis we determined the effect of RGZ and of the overexpression or repression of PPAR␥ in the phosphorylation status of ERK1/2 and Akt. As shown in Fig. 7A (left panel), exposure of PC12 Ct cells to RGZ for 5 min or 24 h had no effect on ERK1/2 phosphorylation. NGF treatment for 5 min was used as a positive phosphorylation control. Down-regulation (Fig.  7A, mP␥Ϫ cells, middle panel) or up-regulation of PPAR␥ (Fig.  7A, mP␥ϩ cells, left panel) had no effect on ERK1/2 phosphorylation and was not affected by RGZ (1 M) either after 5-min or 24-h exposure. As for control cells, NGF treatment strongly increased ERK1/2 phosphorylation in both PC12 clones, suggesting that PPAR␥ overexpression or repression does not alter TrkA/NGF activation of ERK1/2. Similar results were obtained for Akt phosphorylation in Thr-308 (Fig. 7B). These results strongly suggest that PPAR␥-induced Bcl-2 up-regulation does not result from ERK1/2 activation and that the Akt-dependent survival pathway is not likely involved in PPAR␥-induced increased survival of neurons.

DISCUSSION
Neuronal cell death plays a central role in several neurodegenerative disorders (51), and PPAR␥ agonists have been proposed as an alternative for the treatment of these pathologies because of its anti-inflammatory effect (4 -7). We show, for the first time, a connection between PPAR␥ signaling and a prosurvival protein expression in neurons, suggesting that PPAR␥ might be a key factor in inducing neuronal cells survival under neuropathological conditions, independently of PPAR␥ anti-inflammatory effects. This proposal, which relates PPAR␥ to mitochondrial function, is further supported by a recent report showing that RGZ induces mitochondrial biogenesis in mouse brain (52), and is consistent with the resistance to mitochondrial injury and cell death already reported in Bcl-2-overexpressing neural cells (18). A large body of literature supports the protective effects of TZDs directly in neurons. TZDs can protect hippocampal neurons from A␤-induced neurodegeneration (8); cerebellar granule neurons and cortical neurons from post-glutamate neurotoxicity and low potassiuminduced apoptosis (53,54), and HT-22 cells from oxidative insult (55). Furthermore, RGZ protects human neuroblastoma SH-SY5Y cells against acetaldehyde-induced cytotoxicity and increases Bcl-2 mRNA expression as assessed by semi-quantitative RT-PCR (9).
In contrast, in non-neuronal normal or tumor cell lines and tissues, there is conflicting data showing TZD-induced induction of apoptosis in some cases, or TZD-induced protection in others (reviewed in Ref. 56). In fact, different reports using the same cell model show a damaging effect (T cells (57,58) and endothelial cells (59) or protection (T cells (22) and endothelial cells (60)). These discrepancies most probably result from either the high concentration of TZDs used in most studies, which is known to substantially down-regulate PPAR␥ (61), and that, according to our model, would lead to down-regulation of Bcl-2 and apoptosis, or because the observed TZD effect is PPAR␥-independent, as shown in a number of cases (40 -44), including the expression of apoptosis-related genes (62).
Our results in dominant negative PPAR␥and PPAR␥-overexpressing PC12 cells, and in NGF-differentiated PC12 cells and clones, strongly suggest that RGZ-induced Bcl-2 up-regulation and the resulting mitochondrial stabilization are PPAR␥dependent. PPAR␥ loss of function results in Bcl-2 down-reg-FIGURE 6. Bcl-2 knockdown using siRNA decreases survival of PPAR␥-overexpressing PC12 cells in response to A␤ and oxidative stress. mP␥ϩ cells were transfected with a Bcl-2 or a control siRNA, as described under "Experimental Procedures." After 24 h cells were changed to PC12 basal medium containing 2% horse serum, and cultured for various times. A, Bcl-2 siRNA transfection strongly decreases Bcl-2 protein content. Proteins were extracted 24, 48, and 96 h after transfection, and Western blots were performed against Bcl-2. NT corresponds to un-transfected cells, and Ctr and Bcl-2 represent cells transfected with control and Bcl-2 siRNA, respectively. ␤-Actin was used as a loading control. Bcl-2 expression was strongly repressed 48 after transfection (80 -85%; in two independent experiments) and recovered its expression 96 h after transfection. No comparative effect was observed in control siRNA-transfected cells. B, Bcl-2 knockdown increases susceptibility of mP␥ϩ cells to A␤-and H 2 O 2 -induced stress. Cells were transfected with control or Bcl-2 siRNA, and 48 h after transfection they were exposed for 4 h to 5 M A␤, 100 M, or 200 M H 2 O 2 . Cell were fixed and processed for nuclei and TUNEL staining. Representative pictures are presented for control siRNA-transfected cells (left panel) and Bcl-2 siRNA-transfected cells (right panel) for nuclei and TUNEL staining for untreated cells (control) and cells exposed to various treatments, as indicated. C, quantification of non-apoptotic cells after various treatments in siRNA mP␥ϩ-transfected cells. The percent of apoptotic cells was determined by counting total cells (nuclei staining) and apoptotic cells (TUNEL staining) in random fields of two independent experiments as described in A. At least 500 total cells were counted for each treatment. Results are presented as mean Ϯ S.D. (bars). *, statistical significance (p Ͻ 0.005, or less). A clear increase in susceptibility to stress is observed in Bcl-2 siRNA-transfected cells when compared with cells transfected with a control siRNA. Furthermore, in untreated cells Bcl-2 siRNA transfection is by itself sufficient to induce a significant decrease in cell viability. ulation, concomitantly increases H 2 O 2 -induced ROS levels and susceptibility to A␤-induced toxicity, and impairs mitochondrial membrane potential, whereas PPAR␥ overexpression up-regulates Bcl-2, prevents H 2 O 2 -induced ROS, stabilizes mitochondria, and decreases A␤-induced toxicity. These observations make it very likely that the RGZ-induced increased expression of Bcl-2 in primary neurons cultures, leading to protection against NGF withdrawal in DRG neurons and to mitochondrial stabilization in hippocampal cells, is also PPAR␥-dependent. Moreover, ablation of the increased resistance of PPAR␥-overexpressing cells to insult by siRNA knockdown of Bcl-2 suggests that PPAR␥-induced up-regulation of this prosurvival factor mediates enhanced survival in response to PPAR␥.
These data suggest that other already described TZD-induced protective effects in neurons, discussed above, might result from Bcl-2 up-regulation and are consistent with previous reports showing (a) that Bcl-2 increases mitochondrial stability and prevents apoptosis induced by oxidative stress (18) and (b) that Bcl-2 overexpression induces resistance to A␤ neurotoxicity in cortical cells (63) and prevents A␤-induced stress in PC12 cells (64). Further, it might have clinical implications, because A␤ induces the generation of free radicals and leads to oxidative damage in AD (45).
At present we cannot discard that other pathways or proteins acting in the mitochondrial-dependent intrinsic apoptosis pathway are also affected by PPAR␥ activation, and might contribute to the observed effects. Expression or the phosphorylation status of the pro-or the anti-apoptotic members of the bcl-2 gene family might also be affected by PPAR␥ overexpression or repression or by RGZ. Nevertheless, several considerations point to PPAR␥-induced Bcl-2 up-regulation as a key factor. First, Bcl-2 supports membrane mitochondrial poten-tial, and its disruption initiates the intrinsic mitochondrial-dependent apoptosis pathway preceding other events such as caspase activation. Second, in the Bcl-2 knockout mice the loss of a substantial number of neurons indicates that other antiapoptotic members of the bcl-2 gene family do not seem to compensate fully for the absence of Bcl-2 in the nervous system (38), a notion that is further supported by studies in different cells models using siRNA against Bcl-2 and showing in all cases increased apoptosis (65)(66)(67)(68). The virtual absence of Bcl-2 protein expression in PPAR␥-dominant negative expressing PC12 cells correlates with their sensitivity to toxic insult and is consistent with the suppression of PPAR␥-induced protection against stress by Bcl-2 knockdown. On the other hand, we show that pro-survival Akt signaling, the other main anti-apoptotic pathway in neurons (39,48), which also stabilizes mitochondria, is not activated by PPAR␥ overexpression or by RGZ. Activation of several survival-signaling pathways results in Akt Thr-308 phosphorylation and activation. Activated Akt in turn phosphorylates and inactivates pro-apoptotic proteins such as the bcl-2 gene family member BAD or glycogen synthase kinase-3␤ (39,48).
The mechanism by which PPAR␥ induces Bcl-2 up-regulation is an important question raised by the results presented in this study. Bcl-2 is a target of CREB (20), and NGF-induced ERK1/2 signaling is thought to be responsible for CREB activation and Bcl-2 up-regulation (19). PPAR␥-induced Bcl-2 upregulation is not likely a consequence of ERK1/2 activation, because phosphorylation of this MAPK is not induced by RGZ or PPAR␥ overexpression. However, interaction of PPAR␥ signaling with other pathways, resulting in CREB activation, cannot be excluded, because other NGF/TrkA downstream kinases cascades can also phosphorylate CREB (16,69). Further, PPAR␥ can also change cell fate through physical association with signaling molecules such as p65 and ERK5 (70 -72).
Whether up-regulation of Bcl-2 results from PPAR␥ crosstalk with other signaling pathways or to binding to a PPRE in the bcl-2 gene, as reported in human colon cancer cells (23), remains to be established. Our previous results, showing that PPAR␥ is a downstream target of NGF/TrkA signaling (10), and data presented here, suggest that, in addition to the canonical up-regulation of Bcl-2 through the TrkA-ERK1/2-CREB pathway, NGF/TrkA up-regulation of PPAR␥, which in turn would also increases Bcl-2 expression, might contribute to mitochondrial stability and cell survival. The relative contribution of these NGF/TrkA-dependent pathways to survival in different embryonic and adult neuronal cells remains a question to be addressed in future studies. In adult neuronal cells, which do not require NGF for survival (27), PPAR␥-Bcl2 signaling might be more relevant, because PPAR␥ or TrkA can be activated independently of NGF. For instance, angiotensin II induces PPAR␥ transcriptional activity in PC12 cells via its type 2 receptor, resulting in neurite outgrowth and suggesting a neuroprotective action (73), whereas TrkA can be transactivated, in the absence of neurotrophins, by ligands of the G-protein-coupled receptor, resulting also in increased neuronal survival (74). Results presented here suggest that this novel connection between PPAR␥ signaling and Bcl-2 expression might consti- FIGURE 7. RGZ or PPAR␥ overexpression does not activate the ERK 1/2 MAPK or the Akt signaling pathways. Control (Ctr), mP␥Ϫ, and mP␥ϩ cells were incubated under various conditions. After protein extraction, Western blots were performed against phospho-ERK 1/2 (P-ERK 1/2), total ERK1/2 (T-ERK1/2), or against total Akt (T-AkT) and phospho-Akt (P-Akt (Thr-308). A, control PC12 cells (left panel), mP␥Ϫ (middle panel), or mP␥ϩ cells (right panel) were incubated, in the absence (Ctr) or presence of RGZ (1 M) for 5 min (5Ј) or 24 h, or with 100 ng/ml NGF for 5min. NGF, used as a positive control, strongly induces Erk1/2 phosphorylation in Ctr, mP␥Ϫ, or mP␥ϩ PC12 cells, whereas RGZ was without effect, either after 5-min or 24-h treatment, in all cases. No significant changes were found in total ERK1/2 in control or the PC12 clones. B, as for ERK1/2, neither RGZ nor overexpression or repression of PPAR␥ had any effect on the Akt phosphorylation status. In control (Ctr, left panel), mP␥Ϫ (middle panel), or mP␥ϩ (right panel) PC12 cells, only NGF induced Akt phosphorylation, as expected, without changes in total Akt. tute a new target for the pharmacological treatment of neurodegenerative disorders.