Redox regulation of cAMP-responsive element-binding protein and induction of manganous superoxide dismutase in nerve growth factor-dependent cell survival.

Reactive oxygen species (ROS) act as both signaling molecules and mediators of cell damage in the nervous system and are implicated in the pathogenesis of neurodegenerative diseases. Neurotrophic factors such as the nerve-derived growth factor (NGF) support neuronal survival during development and promote regeneration after neuronal injury through the activation of intracellular signals whose molecular effectors and downstream targets are still largely unknown. Here we present evidence that early oxidative signals initiated by NGF in PC12 cells, an NGF-responsive cell line, play a critical role in preventing apoptosis induced by serum deprivation. This redox-signaling cascade involves phosphatidylinositol 3-kinase, the small GTPase Rac-1, and the transcription factor cAMP-responsive element-binding protein (CREB), a molecule essential to promote NGF-dependent survival. We found that ROS are necessary for NGF-dependent phosphorylation of CREB, an event directly correlated with CREB activity, whereas hydrogen peroxide induces a robust CREB phosphorylation. Cells exposed to NGF show a late decrease in the intracellular content of ROS when compared with untreated cells and increased expression of the mitochondrial antioxidant enzyme manganese superoxide dismutase, a general inhibitor of cell death. Accordingly, serum deprivation-induced apoptosis was selectively inhibited by low concentrations of the mitochondrially targeted antioxidant Mito Q (mitoquinol/mitoquinone). Taken together, these data demonstrate that the oxidant-dependent activation of CREB is a component of NGF survival signaling in PC12 cells and outline an intriguing circuitry by which a cytosolic redox cascade promotes cell survival at least in part by increasing mitochondrial resistance to oxidative stress.

produced by incomplete reduction of oxygen to water in a multiplicity of cellular reactions. Unstable and highly nucleophilic ROS attack nearly all the cellular molecules, thereby exerting important toxic effects when generated in excess with respect to the cellular antioxidant defenses and repairing systems. Accordingly, oxygen species have been implicated in a large number of pathological mechanisms such as senescence, ischemic cell death, and neoplastic transformation (1). Neuronal cells are particularly sensitive to oxidative insults, and ROS are involved in many neurodegenerative processes, including those associated with Alzheimer's, Parkinson's, and Huntington's diseases, acute brain ischemia, and excitotoxicity (Ref. 2 and references therein and Refs 3 and 4). In these pathological conditions, cellular stress triggers mitochondrial oxidative damage, which eventually results in apoptosis and/or necrosis (5). Moreover, the analysis of mice lacking the mitochondrial antioxidant enzyme manganese-dependent superoxide dismutase (Mn-SOD) (SOD2) (6) has demonstrated lesions reminiscent of human neurodegenerative diseases.
In apparent contradiction with the role of ROS in human and experimental pathology, a large body of evidence indicates that oxygen species, when produced in low amounts and in a controlled fashion, are not toxic and can function as physiological components of the signaling generated by cytokine and growth factors (7,8). ROS, and hydrogen peroxide in particular, act as critical signaling molecules in cellular signaling generated by a wide array of stimuli, including tumor necrosis factor-␣, interleukin-1 (9), platelet-derived growth factor, epidermal growth factor, insulin (Ref. 7 and references therein), vascular endothelial growth factor (10), antigens, mitogenic lectins (11), and in growth control by cell-cell contact (12).
Although the molecular mechanisms of oxygen radical-mediated signaling is still largely unclear, much evidence suggests that ROS strongly regulates protein tyrosine phosphorylation by directly inhibiting protein-tyrosine phosphatase activity (13,14). Moreover, the small GTPase Rac-1 (15), whose activity is linked to ROS generation by the combined effect on membrane oxidases and metabolism of arachidonic acid (16), mediates cellular responses to oxidative stimuli. In neurons oxidant species such as NO (17) and ROS regulate signaling events, including PI 3-kinase/Akt pathway, c-Jun NH 2 -terminal kinase/ stress-activated protein kinase pathway, and activation of the nuclear factor B (18,19), suggesting a dual role as endogenous toxins and signaling molecules.
Target-derived neurotrophins, including nerve growth factor (NGF) and brain-derived neurotrophic factor act as major regulators of neuronal cell differentiation and survival in the developing nervous system and exert a protective effect after acute neuronal injury (20,21). Neurotrophins exert their effects through a complex network of intracellular signals involving the activation of the PI 3-kinase/Akt pathway, Ras/mitogen-activated protein kinase pathway, and the small GTPases of the Rho family (Rho, Rac, and CDC42) (20,22). Many NGFdependent survival signals converge on the transcription factor cAMP-responsive element-binding protein (CREB), a key molecule responsible for the expression of antiapoptotic genes in sympathetic neurons and cerebellar granule cells (23,24).
Only few target genes for neurotrophin signaling have been identified so far; because neurotrophins increase cell resistance to oxidative stress (25), and application of NGF to sympathetic neurons reduce the generation of oxygen radicals in mitochondria upon serum withdrawal (26), it is conceivable that these trophic factors modulate cellular antioxidant defense. Transcriptional regulation of the mitochondrial superoxide scavenger Mn-SOD represents a general protective event in response to cellular stress and to pro-apoptotic stimuli (27)(28)(29)(30). Interestingly, a cAMP-responsive element is present within the Mn-SOD promoter, and 12-O-tetradecanoylphorbol-13-acetatemediated gene expression in lung carcinoma cell line (28) is dependent upon CREB activation.
Here we show that induction of Mn-SOD in serum-deprived PC12 cells is NGF-and CREB-dependent and may therefore represent a general modality of neurotrophin signaling. Importantly, the molecular pathway linking NGF to Mn-SOD expression reveals an important role for ROS in triggering CREB activation and gene expression, thereby outlining a redox circuitry in which cytosolic oxidants are used as messenger intermediates to increase mitochondrial protection from oxidative stress.

EXPERIMENTAL PROCEDURES
Plasmids, Antibodies, and Chemicals-Myristylated human Akt expression construct was kindly provided by Dr. A. Bellacosa. Expression constructs for CREB-VP16 and CREB-VP16/⌬LZ were obtained from Dr. D. D. Ginty. Rac-1 N17 cDNA was provided by Dr. A. Hall and subcloned in the EcoRI site of the pLPC retroviral expression vector (a gift of Dr. S. Lowe). The retroviral construct encoding the oncogenes V12 H-Ras and E1A on the pLPC backbone was provided by Dr. Lowe. The rat Mn-SOD cDNA was obtained by reverse transcription-PCR from PC12 total RNA and cloned in antisense orientation in the EcoRI site of pLPC. Plasmid encoding the green fluorescent protein (GFP) under the transcriptional control of the cytomegalovirus promoter was purchased from Promega.
Rat nerve growth factor was a kind gift of Drs. D. D. Ginty and D. Mercanti. Dichlorofluorescein diacetate (DCF-DA) and dihydroethidium (DHE) were from Molecular Probes, 2-methoxyestradiol (2-ME), hydrogen peroxide, N-acetylcysteine, and propidium iodide were purchased from Sigma. The synthetic mitochondrial antioxidant Mito Q, described in Kelso et al. (43), was obtained from Drs. M. Murphy and R. Smith. Transfection reagent LipofectAMINE was purchased from Promega.
Cell Lines-PC12 were maintained in RPMI 1640 with 10% fetal bovine serum and 10% horse serum in collagen-coated cell culture dishes. 293T kidney adenocarcinoma cells, and the derivative ecotropic packaging line Phoenix (kindly provided by Dr. S. Lowe) were maintained in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum. The E1A/Ras-transformed, SOD2 ϩ/ϩ and Ϫ/Ϫ MEF lines were derived by in vitro transformation of primary embryonic fibroblasts from either wild-type or SOD2 Ϫ/Ϫ C57Bl 6J mice (kind gifts of Dr. C. J. Epstein, University of California, San Francisco, CA). Transformed clones of each genotype were transfected with the rat Trk cDNA (provided by Dr. Ginty), selected in puromycin, and assessed for Trk expression by flow cytometry. Wild-type and mutant cells expressing comparable levels of the receptor were chosen for further experimental analysis.
Cell Transfection and Infection-PC12 cells and E1A/Ras MEFs were transfected with LipofectAMINE (Promega) according to the manufacturer's instructions. Average transfection efficiency was 15-20%, based on cell positivity for GFP. Cells were left to recover for 24 -48 h after transfection before functional or biochemical assays. 293T and Phoenix cells were transfected by calcium-DNA co-precipitation according to the standard procedure (50% efficiency).
For in vitro transformation, SOD ϩ/ϩ and SOD2 Ϫ/Ϫ MEFs were infected with the pLPC-E1A/H-Ras construct, which harbors no selectable marker, and left in culture for 2 weeks. Transformed foci were then isolated and further expended.
Measurement of ROS-For detection of reactive oxygen species, cells were maintained for 16 h in low (0.5%) serum. After medium replacement with Hanks' balanced salt solution or serum-free RPMI, NGF was added, and cells were incubated for an additional 60 or 120 min. During the last 30 min of incubation, DCF-DA (1 g/ml) or DHE (10 M) was added to cells. After quick detachment from the plate, cells were immediately subjected to flow cytometry (FL-1 for DCF-DA and FL-2 for DHE) using a COULTER-EPICS flow cytometer equipped with an argon laser lamp (emission 488 nm). The mean cell fluorescence values were expressed as raw data or as fluorescence variations (stimulated Ϫ unstimulated).
In some experiments, cell were plated in serum-free medium with or without NGF, and the fluorescent probes were added directly to the medium 24 h later for 30 min. Cells were then processed as described above.
Assays for Cell Viability-Cells were seeded in serum-free medium with or without NGF and/or antioxidants. In experiments with Mito Q, cell were pretreated with the drug for 8 h in complete medium before serum withdrawal. 48 h later, cell viability was determined by flow cytometry.
Live and dead cells were distinguished by flow cytometry according to two criteria, (a) forward/side scatter profile and (b) exclusion of propidium iodide. The majority of cells positive for propidium iodide (dead cells) fell in a distinct region on the forward/side scatter histogram (R2), which could also be easily identified in the forward/side scatter plot in the absence of the fluorescent marker. This latter population was found to be slightly more abundant than the population of propidium iodide-positive cells, likely due to the presence of early apoptotic cells non permeable to propidium iodide. Cell debris was gated out in all the measurements. The percentage of surviving cells was determined according to the forward/scatter profile, with the following formula: [% cells in the "live" region R1/(% cells in "dead" region R2 ϩ % cells in the "live" region R1)] ϫ 100.
In some experiments GFP was expressed in cells together with a gene of interest to allow the identification of transfected cells by flow cytometry. Because GFP can leak out of apoptotic cells, the absolute number of live, GFP-positive cells was then determined using fluorescent microspheres (Flow-Count fluorospheres, Coulter) as an internal standard, according to the manufacturer's recommendations.
Cell Stimulation and Lysis-PC12 cells were stimulated in 0 or 0.5% fetal bovine serum with NGF and in 0% FCS with hydrogen peroxide for the indicated times after 16 h of incubation in low (0.5%) serum. Antioxidants, when necessary, were added either 16 h (N-acetylcysteine (NAC)) or 5 h (Mito Q) before stimulation.
For Western blot analysis cells were lysed in lysis buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotinin, leupeptin, and pepstatin, 1 mM sodium orthovanadate) containing 1% Triton X-100 and 0.1% SDS; lysates were cleared by centrifugation, and equal amounts of protein for each sample were subjected to SDS-PAGE followed by electrotransfer, immunoblotting, and ECL detection. Band quantitation was performed using the Quantity. One software (Bio-Rad). For total RNA extraction, cells were lysed in Trizol (Invitrogen) and further processed according to the manufacturer's recommendations.
For "in gel" determination of SOD activities, cell pellets were resuspended in phosphate buffer, pH 8, containing 1 mM phenylmethylsulfonyl fluoride and sonicated. Equal amounts of protein were loaded on a non-denaturing polyacrylamide gel and processed as described (46).
Northern Blot Analysis of Mn-SOD mRNA-15-20 g of total RNA were subjected to Northern blot analysis. The rat Mn-SOD cDNA and the rat actin cDNA were obtained by reverse transcription-PCR and random-prime-labeled using the Redi-Prime labeling kit (Amersham Biosciences). After hybridization for Mn-SOD (65°C overnight) filters were stripped and reprobed for actin to verify that comparable amounts of RNA had been loaded in all lanes.

NGF Rescues PC12 Cells from Apoptosis Induced by Serum
Withdrawal-PC12 cells are dependent on calf serum for survival in vitro and rapidly undergo apoptosis when placed in serum-free medium (25). The addition of NGF to the culture medium largely prevents death induced by serum starvation, with 70% of the cells remaining viable after 48 h of fetal bovine serum withdrawal (Fig. 1, A and B).
NGF activates a multiplicity of signaling cascades in PC12 cells, mainly through the interaction with the tyrosine kinase receptor TrkA and the low affinity receptor p75 (20). As previ-ously shown, NGF addition to serum-deprived PC12 cells dramatically induces the phosphorylation of Akt/protein kinase B and extracellular signal-regulated kinases 1/2 ( Fig. 1C), an event mirroring the increase in enzymatic catalytic activity. Although phosphorylation of mitogen-activated protein kinases likely reflects the activation of the Ras/Raf pathway by NGF, Akt is a direct target for the lipid products of the PI 3-kinase, a major mediator of survival signaling by NGF (31). Overexpression of a membrane-targeted, constitutively active mutant of human Akt nearly completely abolishes cell death induced by serum deprivation (Fig. 1D). Interestingly, a negligible or at best marginal effect was upon transfection of an active form of Raf (CAAX-Raf, data not shown), further suggesting that NGFmediated cell survival in serum-deprived condition is mainly mediated by the PI 3-kinase/Akt pathway (24).
NGF Modulates the Intracellular Content of ROS in PC12 Cells-Reactive oxygen species generated in mitochondria trigger apoptotic cell death in response to many stimuli (32). Moreover, ROS are involved in regulation of gene expression, proliferation, and cell differentiation induced by cytokine and growth factor receptors (7).
NGF treatment of PC12 cells maintained in the absence of serum is initially (15 min to 2 h) accompanied by increased intracellular content of ROS (Fig. 2, A and B), as evaluated by the oxidation of two different redox-sensitive fluorescent probes, DCF-DA (Fig. 2, A and B, a) and DHE (Fig. 2B, b). Because DCF-DA is mainly sensitive to hydrogen peroxide, whereas DHE detects superoxide (33), it is likely that both reactive species are rapidly produced in response to NGF. During the first 2 h of NGF treatment, the increase of intracellular ROS is concentration-dependent, significant at 10 ng/ml and more pronounced at 100 ng/ml ( Fig. 2A). Pretreatment of PC12 cells with the glutathione precursor and radical scavenger NAC drastically reduced early NGF-dependent oxidative events ( Fig. 2A, b), and a similar decrease was also observed in cells transfected with a dominant negative mutant of GTPase Rac-1 (N17Rac1) (Fig. 2A, c).
Interestingly, the early increase of ROS after NGF treatment is followed by a decrease of intracellular oxidants (both peroxides and superoxide) with respect to untreated cells at 24 h after NGF application (Fig. 2B, a and b). The late antioxidant effect of NGF coincides with the onset of morphological signs of death in untreated serum-starved cells. This observation suggests that elevated levels of oxidants detected 24 h after serum deprivation in PC12 cells are causally linked to cell death and are counteracted by application of NGF.
NGF Signaling to Akt and CREB Requires ROS-Upon treatment with NGF, PC12 cells respond with a robust activa-tion of the PI 3-kinase pathway, which couples receptor engagement to cell survival through the activation of Akt/protein kinase B (34,36). A critical downstream event in the survival signaling initiated by NGF is represented by the phosphorylation of CREB, a transcription factor necessary and sufficient for neurotrophin-dependent survival (23,35,36). Accordingly, CREB may be a direct target of Akt (24,37).
To determine whether the early burst of ROS plays a role in NGF-dependent signaling to Akt and CREB, PC12 cells were treated with glutathione precursor and the ROS scavenger NAC to prevent intracellular oxidations, and cell signaling ability was assessed. Removal of ROS resulted in a dramatic decrease of Akt phosphorylation and activity, as evaluated by Ser-256 phosphorylation of the Akt substrate FKHR (38) (Fig.  3A). Concomitantly, pretreatment of PC12 cells with antioxidant NAC resulted in a decrease of CREB phosphorylation on its critical residue, Ser-133, as revealed by immunoblotting using a phosphorylation site-specific antibody (Fig. 3A). Moreover, inhibition of Rac activity, an event necessary for NGF-dependent redox signaling ( Fig. 2A, c), resulted in inhibition of CREB phosphorylation (Fig. 3B). Because CREB phosphoryla- tion is critical for transcriptional activation (39), our data suggest that CREB activity in response to NGF could also be impaired.
To further test the possibility that redox signals regulates CREB in PC12 cells, we assessed whether exogenous hydrogen peroxide is sufficient to induce CREB activation. Cells exposed to 1 mM H 2 O 2 show increased activation of CREB and Akt, to an extent comparable with NGF (Fig. 3C). Hydrogen peroxidedependent Akt and CREB phosphorylation is inhibited by the PI 3-kinase inhibitor wortmannin (Fig. 3D), suggesting that oxidants activate multiple survival pathways upstream of PI 3-kinase.
NGF Induces Mn-SOD Expression in PC12 Cells-CREB-dependent transcriptional activation of antiapoptotic genes contributes to the survival effect induced by NGF. Late reduction of intracellular ROS in NGF-treated PC12 cells (Fig. 2B, a and  b) suggests that neurotrophins might regulate antioxidant enzyme expression. Interestingly, the mitochondrial protein Bcl2, which may have some antioxidant functions, is among the few NGF-dependent antiapoptotic proteins hitherto identified and is modulated in a CREB-dependent fashion.
The mitochondrial antioxidant enzyme and putative survival factor Mn-SOD is induced in many cell types by cytokines (tumor necrosis factor-␣, interleukin-1) (27, 29) and growth factors (platelet-derived growth factor, vascular endothelial growth factor) (40, 41) through signaling cascades that often involve ROS and redox-sensitive transcriptional regulators (36,52). Moreover, analysis of the human Mn-SOD gene promoter region has revealed the presence of a perfect CRE, a binding site for CREB, which accounts for gene response to 12-O-tetradecanoylphorbol-13-acetate in a human lung carcinoma cell line (28).
Exposure of serum-deprived PC12 cells to NGF results in a significant increase of expression of Mn-SOD mRNA transcripts of 4.1 and 1.1 kilobases (Fig. 4A). SOD2 mRNA induction is present 1 h after NGF treatment and peaks at 12 and 24 h after induction. An increase in RNA transcripts is mirrored by the increase of both Mn-SOD protein (Fig. 4B) and enzyme activity, as assessed by an in-gel nitro blue tetrazolium-based SOD assay (Fig. 4C) (42). Accordingly, the intracellular concentration of superoxide is significantly decreased 24 h after NGF stimulation (Fig. 4D). Taken together, these data suggest that Mn-SOD is a downstream target of NGF signaling in PC12 cells.
Modulation of Mn-SOD Expression by PI 3-Kinase and CREB-PI 3-kinase and the transcription factor CREB play a pivotal role in mediating antiapoptotic signaling generated by NGF. PC12 cell pretreatment with the PI 3-kinase inhibitor

FIG. 3. Requirement of ROS for NGF signaling to AkT and CREB.
A, effect of NAC on NGF signaling to AkT and CREB. Cells were preincubated as described in Fig. 2A, b, with 10 mM NAC or left untreated. After 16 h, NGF was added for 15 min at the indicated concentrations. Cells were quickly lysed, and protein extracts were subjected to SDS-PAGE, electrotransfer, and immunoblotting with the indicated phosphospecific antibodies. Equal protein loading throughout the gel was verified by reversible Ponceau S staining or anti-actin immunostaining. The bands of interest (Ser-473 ϩ Thr-308 AkT, Ser-256 FKHR, Ser-133 CREB) are indicated by arrows. p-, phosphorylated. B, inhibition of Rac-1 attenuates CREB phosphorylation by NGF. Cell were processed as in Fig. 2A, c, until stimulation with 100 ng/ml for the indicated times. After cell lysis, protein extracts were subjected to SDS-PAGE and immunoblotting with anti-phospho-Ser-133 CREB. Equal protein loading was verified by reversible Ponceau S staining and by anti actin immunoblotting of the same nitrocellulose filter. Densitometric band intensities for Ser(P)-133 CREB are indicated. C, phosphorylation of AkT and CREB in PC12 cells exposed to hydrogen peroxide. Cells were incubated for 16 h in RPMI, 0.5% FCS, switched to serum-free Hanks' balanced salt solution, and stimulated with either 100 ng/ml NGF or 1 mM hydrogen peroxide for the indicated times. Analysis of protein phosphorylation was performed as described in B. Equal protein loading was verified by anti-actin immunoblotting. D, effects of H 2 O 2 on AkT and CREB are inhibited by wortmannin (WO). Serum-starved PC12 cells were incubated for 10 min with 10 M wortmannin or the relative vehicle (Me 2 SO); hydrogen Peroxide was then added for 15 min, and protein phosphorylation was analyzed as described above. Wortmannin was occasionally found to increase the base-line phosphorylation and activity (as assessed by FKHR phosphorylation) of AkT, whereas peroxide-induced phosphorylation was consistently and significantly reduced. wortmannin significantly inhibited NGF-induced Mn-SOD expression (Fig. 5A). Similarly, pretreatment of PC12 cells with NAC, which diminished both ROS production and CREB phosphorylation in response to NGF (see Figs. 2 and 3), resulted in a significant decrease of Mn-SOD transcription (Fig. 5A). These data suggest that NGF-dependent induction of Mn-SOD requires both PI 3-kinase activity and oxidants in a fashion that mirrors CREB phosphorylation.
To determine whether CREB is sufficient to induce Mn-SOD mRNAs, we overexpressed a constitutively active form of CREB (VP-16 CREB) (23) in both PC12 (Fig. 5B, a) and 293T cells (Fig. 5B, b). Activation of CREB resulted in a significant increase of Mn-SOD mRNAs when compared with cells overexpressing a mutant form of CREB lacking the DNA binding domain (⌬LZ-VP16, Fig. 5B, b). Taken together these data suggests that CREB is required to regulate Mn-SOD expression. Importantly, induction of Mn-SOD by active CREB is insensitive to NAC (5B, b), suggesting that NAC does not indiscriminately repress Mn-SOD gene and reinforcing the idea that signaling roles for ROS in the NGF cascade are upstream of CREB activation.
Redox Regulation of NGF-dependent Cell Survival-Our data suggest that NGF regulates ROS in a complex fashion such that oxygen species generated early after neurotrophin stimulation signal to CREB and act as pro-survival mediators. Later on, when toxic oxidants are generated, NGF contributes to their elimination through up-regulation of compartmentspecific, mitochondrial antioxidant, such as Mn-SOD. To assess the physiological relevance of these redox changes in PC12 cells, we used two different antioxidants, the generic, mem-brane-permeant scavenger NAC and a mitochondria-targeted derivative of ubiquinol (Mito Q) (43). PC12 cells treated with 10 mM NAC, a condition that severely impairs NGF signaling to Akt and CREB (Fig. 3A), showed decreased ability to survive in both the presence and the absence of NGF (Fig. 6, A and B). It is therefore conceivable that NAC promotes cell death by blocking protective redox signals initiated by NGF. Conversely, cell exposure to Mito Q mimicked NGF treatment and resulted in an increased resistance to cell death (Fig. 6A, gray columns). Because Mito Q is present at a low concentration (1 M) in the culture medium and preferentially accumulates in mitochondria (43), it is unlikely to interfere with oxidative reactions that take place in the cytosol, where early ROS are thought to be produced. Still, Mito Q selectively removes harmful oxidants leaking from the respiratory chain during apoptosis, thereby promoting cell survival. In line with this view, Mito Q, unlike NAC, had no effect on NGF-dependent CREB phosphorylation (Fig. 6C) nor did it block the early rise of ROS in cells treated with NGF (data not shown). Along similar lines of evidence, we found that cell exposure to the recently described SOD inhibitor 2-ME (33) drastically reduced NGF-dependent survival without significantly affecting cells grown in standard medium (10% fetal bovine serum) (Fig. 6D). Taken together, these data suggest that oxidative events generated during cell survival and cell death are distinct and physically compartmentalized between cytosol and mitochondria and that NGF survival signaling is critically dependent on the elimination of intracellular superoxide.
Mn-SOD Is Required for Trk-mediated Antiapoptotic Signaling-To confirm that antioxidant defense against mitochon-

FIG. 4. Transcriptional up-regulation of Mn-SOD by NGF in serum-depleted PC12 cells.
A, time course for Mn-SOD mRNA expression in PC12 cells exposed to NGF. Cells were preincubated in low (0.5%) serum for 16 h before exposure to NGF in serum-free medium for the indicated times. Total mRNA was extracted and subjected to Northern blot analysis. Two waves of induction at 1 and 12 h, respectively, were observed. The two main transcripts of 4.1 and 2.1 kb are indicated by arrows. Equal RNA loading was verified by filter re-hybridization for a housekeeping gene (glyceraldehyde 3-phosphate dehydrogenase). Fold induction (F.I.) with respect to the unstimulated sample was determined by band densitometry after normalization for the housekeeping gene. B and C, increases in Mn-SOD immunoreactive protein (B) and in-gel activity (C) in response to NGF. In B band intensities are indicated. In C, a slight increase of Mn-SOD activity in serum-starved cells in comparison to cells grown in standard medium was observed, in agreement with previous reports (30,32). Note that CuZn-SOD, unlike the mitochondrial, manganese-dependent enzyme, is marginally or not affected by serum deprivation and NGF. Ctrl, control. D, a reduction in intracellular superoxide accompanies up-regulation of Mn-SOD by NGF. Intracellular superoxide was measured as in Fig. 2 24 h after serum withdrawal and incubation with (NGF 24 h) or without (ctrl) NGF. The figure is representative of several independent experiments. drial oxidants is important for antiapoptotic signaling by NGF, we first inhibited expression of Mn-SOD by retrovirus-mediated expression of an antisense cDNA (pBabe/Mn-SODAS), as assessed by Western blot analysis (Fig. 7A, b). PC12 cells expressing low levels of Mn-SOD (Mn-SODAS) were more sensitive to cell death in the presence of NGF when compared with mock-transfected cells (Fig. 7A, a). The important role of Mn-SOD in NGF trophic signaling was further demonstrated by making SOD2-deficient transformed fibroblasts or their wildtype counterparts responsive to the neurotrophin through transfection of the NGF receptor Trk. Transformed fibroblasts are highly sensitive to apoptosis by serum deprivation (44). As indicated in Fig. 7B, NGF signaling through Trk protects SOD2-proficient cells from serum withdrawal, whereas no significant protection was observed SOD2 null cells. A block of trophic signaling in SOD2 Ϫ/Ϫ cells occurred downstream of NGF-induced CREB phosphorylation, which occurred at comparable levels in both cell lines (Fig. 7B, b). Although in part artificial, the MEF/Trk model strongly confirms the involvement of Mn-SOD in NGF signaling, extending the above observations to a non-neuronal cell background. These data are therefore consistent with the idea that elimination of harmful oxygen species from mitochondria through the transcriptional up-regulation of Mn-SOD plays a critical role in NGF-dependent cytoprotection (Fig. 8). DISCUSSION Experimental observations described in the present paper lead to two major conclusions as follows. 1) Reactive oxygen species have a role in transducing survival signals released by NGF in PC12 cells. Because PC12 cells respond to NGF at least in part through the Trk-A receptor, the above finding is in line with a growing body of evidence concerning the involvement of ROS in intracellular signaling downstream of PTK-receptor molecules.
2) The up-regulation of mitochondrial superoxide dismutase (Mn-SOD) is part of the antiapoptotic genetic program triggered in PC12 cells by nerve growth factor. This is a novel insight that expands our limited knowledge regarding the transcriptional targets of NGF. Moreover, the above observation indicates that mitochondrial protection from oxidative stress is a potentially general mechanism of action for neurotrophins in physiological and pathological settings.
We have postulated that ROS produced shortly after cell exposure to NGF serve signaling functions downstream of activated NGF receptor(s). This idea is in agreement with recent reports on redox signaling by NGF receptors in neuronal cells (45,46). Although not biochemically characterized in detail, these oxidant species include probably both hydrogen peroxide and superoxide, based on the data obtained with fluorescent redox probes with different radical specificity (DCF-DA and DHE). It should be noted that probe oxidation in response to NGF was not prevented by the NO synthase inhibitors L-Nmonomethylarginine and diphenyleneiodonium (not shown), ruling out the possibility that DCF-DA and DHE are oxidized by NO. Interestingly, NGF-dependent generation of ROS involves the small GTPase Rac-1, a general transducer for growth factor signaling repeatedly implicated in liganddependent generation of oxygen species and in the promotion of cell survival (47), already reported to mediate neurite outgrowth (48) and cell differentiation in cells of neuronal origin exposed to NGF. Although the present data suggest a role for Rac-induced ROS in survival signaling by NGF, it is conceivable that other NGF-dependent signals involving Rac utilize, at least to some extent, oxygen species as messenger intermediates. To this end, it is noteworthy that oxygen radicals can also induce cell cytoskeleton reorganization (49) and that differentiation of at least some types of neuronal cells involves the generation of oxygen species (45).
An important point emerging from these studies is that NGF

FIG. 5. Effect of PI 3-kinase and CREB on Mn-SOD gene induction.
A, induction of Mn-SOD by NGF is blocked or attenuated by the PI 3-kinase inhibitor wortmannin (Wo) and by NAC. Cells were preincubated with wortmannin (10 M) for 10 min and with NAC for 16 h before exposure to NGF for the indicated times. The two Mn-SOD transcripts are indicated by arrows. The filter was stripped and re-hybridized for actin, to ensure equal RNA loading. Ctrl, control. B, active CREB is sufficient to induce Mn-SOD in PC12 and 293 T cells. Cells were transfected by LipofectAMINE and calcium/DNA co-precipitation, respectively, and incubated for 48 h in complete medium. 293T cells were also exposed to NAC 10 mM for 24 h before RNA extraction (NACϩ). 20 g total RNA were subjected to Northern blot analysis as in A. In B, b, ribosomal RNA was used as the loading control. Transfection efficiencies for the two cell lines are indicated.
signaling to AkT and CREB requires ROS and is severely impaired upon antioxidant treatment. Although experiments with oxidized NAC (data not shown) confirm that the inhibitory effect of N-acetylcysteine on the NGF cascade is directly linked to the capacity of the compound to lower the intracellular concentration of ROS, the link between oxygen species and antiapoptotic signaling is further strengthened by the fact that exogenous hydrogen peroxide, as already reported (50), is sufficient per se to induce site-specific phosphorylation of Akt and CREB.
The redox modulation of CREB phosphorylation/activation represents an intriguing finding whose molecular mechanism deserves further investigation. As a first hypothesis, it is conceivable that Akt, once activated by oxidants, directly phosphorylates CREB, a possibility supported by the observation that both Akt and CREB phosphorylation in response to H 2 O 2 are PI 3-kinase-dependent ( Fig. 3C and Ref. 47). Another important question is whether redox modulation of CREB operates as well in the process of NGF-dependent neuronal differentiation, because it has been reported that ROS induced by NGF are required for differentiation of PC12 cells (45), a process clearly dependent on CREB activity (51).
The finding that a redox-sensitive antiapoptotic cascade is triggered in PC12 cells by NGF may also have broader pathophysiological implications. In fact, because the nervous system is highly prone to oxidative damage (4), activation of survival pathways by oxidants may represent a sensitive and efficient safeguard mechanism designed to increase cell resistance to environmental stress in response to early changes in the intracellular redox balance. This hypothesis is supported by the finding here presented that NGF induces expression of Mn-SOD, an enzyme able to remove harmful oxidants, that such induction is dependent on PI 3-kinase activity, and that interference with Mn-SOD expression results in impaired NGF survival signaling in both PC12 cells and Trk-expressing transformed fibroblasts (Fig. 7). Moreover, Mn-SOD, which is a well recognized antiapoptotic enzyme promoting mitochondrial integrity, has already been shown to protect other cell types from apoptosis by serum deprivation (30,33), and a decrease in mitochondrial superoxide has been proposed as a mechanism for neuroprotection by NGF (26). Furthermore, signs of neurodegeneration have been reported in the brain of SOD2-deficient mice (6). Interestingly, another well established target for NGF survival signaling, Bcl-2 (23), is also located in mitochondria, which underlines the importance of these organelles as targets for neurotrophin signaling and suggests a possible physical and/or functional interaction between the two survival proteins.
Although the mechanism of Mn-SOD induction by NGF deserves further investigation, our data imply CREB in the NGFdependent regulation of Mn-SOD. In fact, an ideal CREBresponsive element is present at position Ϫ1258 of the human Mn-SOD promoter (28), and overexpression of CREB leads to Mn-SOD overexpression in PC12 and 293T cells (Fig. 7). Moreover, cell treatment with NAC drastically decreases Mn-SOD expression in response to NGF concomitantly with a significant reduction in the phosphorylation/activation of CREB. Interestingly, Mn-SOD is induced by a constitutively active mutant of CREB in a ROS-independent fashion, suggesting that this factor is a downstream target for redox signaling in the cascade linking NGF to the up-regulation of Mn-SOD. Further studies utilizing CREB inhibitory molecules will be required to assess was consistently observed, maybe due to its properties of radical scavenger. NAC (10 mM) was added to cells upon serum deprivation together with NGF, whereas Mito Q was added 8 h earlier in complete medium to allow accumulation in active mitochondria and removed for the rest of the experiment. C, NAC, but not Mito Q, interferes with NGF signaling to CREB. Cells were treated as in A, except that Mito Q was added for 8 h in serum-free medium. Phosphorylation of CREB (P-CREB) was assessed as described above. Equal protein loading was verified by anti-actin immunoblotting of the same nitrocellulose filter. D, inhibition of NGF rescue by the SOD inhibitor 2-ME. Cells were seeded in either standard medium (10% FCS) or serum-free medium plus 100 ng/ml NGF (SF ϩ NGF), in presence of titrated concentrations of 2-ME. Cell survival was determined 48 h later by flow cytometry. the relative role of this factor in the induction of Mn-SOD by NGF with respect to other potential candidates such as nuclear factor B. However, even though the mechanisms involved are unclear, the evidence of Mn-SOD induction through a ROS-dependent cascade is biologically sound and in line with other findings that oxidants are inducers of antioxidant enzymes in both prokaryotic (52) and eukaryotic cells (53).
The proposed role for ROS as both mediators and targets for survival signaling by NGF may appear contradictory. We suggest that oxygen species involved in early NGF signaling are different from those responsible for cell death in both their timing of production and their subcellular sites of production. Although NGF-induced ROS are produced rapidly after receptor stimulation through Rac-1-dependent mechanisms, late generation of ROS likely takes place in mitochondria and reflects the onset of mitochondrial dysfunction. These latter oxi- dants may be eliminated by Mn-SOD, which is in turn induced by NGF (Fig. 8). Importantly, although Mn-SOD activity is expected to lead to peroxide accumulation, decreased DCF-DA fluorescence in concomitance with elevated intracellular SOD is not surprising and is consistent with important recent reports (54,55).
Although in part speculative, the idea that NGF-triggered redox signaling is intracellularly compartmentalized is in line with similar findings concerning survival signaling by tumor necrosis factor-␣ (56) and is supported by data displayed in Fig.  6, which shows the opposite effect of a generic (NAC) versus a mitochondrially targeted (Mito Q) antioxidant in cell response to serum deprivation and to NGF. Although Mito Q well mimics the compartment-specific effect of Mn-SOD, NAC freely diffuses throughout the cell, inhibiting survival signaling. On the other hand, because NAC can also remove harmful oxidants from mitochondria, this effect could partially mitigate the upstream inactivation of ROS-dependent antiapoptotic signals, thereby explaining the discrepancy between the profound inhibition of Akt, CREB, and Mn-SOD and the limited decrease in cell survival observed in cells treated with NAC (compare Figs. 3 and 6). One final consideration deals with the significance of the PC12 model with respect to the general mechanisms of neuroprotection by NGF. Although NGF plays a key role in target-determined cell survival within the developing nervous system (4), the effect of the neurotrophin on undifferentiated, serum-starved PC12 cells depicts a more general situation of growth factor-dependent cell rescue from an acute insult. However, although the above observations need to be confirmed in vivo or in different cellular models more representative of the physiological role of NGF, it should be noted that a role for neurotrophins in adult cell recovery from damage ("cell rescue") has also been proposed (4). Notwithstanding these limitations, the presented observations on redox-dependent regulation of NGF signaling and on NGF induction of Mn-SOD may contribute to understanding the molecular mechanisms of neuronal protection by neurotrophins and clarification of the intricate connections between oxidative stress, antioxidant enzymes, and neuropathology.