Oxidative stress induces neuronal death by recruiting a protease and phosphatase-gated mechanism.

Reactive oxygen species (ROS) cause death of cerebellar granule neurons. Here, a 15-min pulse of H(2)O(2) (100 microm) induced an active process of neuronal death distinct from apoptosis. Oxidative stress activated a caspase-independent but calpain-dependent decline of calcium/calmodulin-dependent protein kinase IV and cAMP- responsive element-binding protein (CREB). Calpain inhibitors restored calcium/calmodulin-dependent protein kinase IV and CREB but did not influence phosphorylated CREB levels or survival, indicating recruitment of an additional dephosphorylation process. Co-treatment with calpain and serine/threonine phosphatase inhibitors restored pCREB levels and rescued neurons. This phosphatase-activated signaling pathway was shown to be dependent on de novo protein synthesis. Further, gene transfer studies revealed that CREB is a common final effector of both apoptosis and ROS-induced death. Our data indicate that dephosphorylation and proteolytic signaling mechanisms underlie ROS-induced programmed cell death.

Reactive oxygen species (ROS) cause death of cerebellar granule neurons. Here, a 15-min pulse of H 2 O 2 (100 M) induced an active process of neuronal death distinct from apoptosis. Oxidative stress activated a caspaseindependent but calpain-dependent decline of calcium/ calmodulin-dependent protein kinase IV and cAMPresponsive element-binding protein (CREB). Calpain inhibitors restored calcium/calmodulin-dependent protein kinase IV and CREB but did not influence phosphorylated CREB levels or survival, indicating recruitment of an additional dephosphorylation process. Co-treatment with calpain and serine/threonine phosphatase inhibitors restored pCREB levels and rescued neurons. This phosphatase-activated signaling pathway was shown to be dependent on de novo protein synthesis. Further, gene transfer studies revealed that CREB is a common final effector of both apoptosis and ROS-induced death. Our data indicate that dephosphorylation and proteolytic signaling mechanisms underlie ROS-induced programmed cell death.
Oxidative stress refers to the cytotoxic consequences of exposure to reactive oxygen species (ROS). 1 The latter are generated as by-products of normal and aberrant metabolic processes that utilize molecular oxygen. There is now evidence for a potential role of ROS in acute neurological events, such as ischemia, anoxia and reperfusion injury, and chronic neurodegenerative disease, such as Alzheimer's disease (1), Parkinson's disease (2), and amyotrophic lateral sclerosis (3). The precise mechanism by which free radical-induced neurodegeneration occurs is not known and remains a subject of controversy; some groups report features of necrosis (4,5), while others describe apoptotic features (6 -9). Apoptotic cells are characterized by condensed and fragmented nuclei, whereas necrotic cells present loss of membrane integrity without apparent nuclear damage. Apoptosis is also characterized by the activation of specific cysteinyl proteases (caspases), which have been shown to be ROS-activated in several models (7,10).
Calpains represent another group of cysteinyl proteases. Their up-regulation by oxidative stress has recently been demonstrated in PC12 pheochromocytoma cells (11) and in the mouse embryonic carcinoma P19 cell line (12). Calpain is one of the most abundant neutral proteinases in the central nervous system and occurs as a proenzyme in association with its endogenous inhibitor, calpastatin. Two forms of calpains, microcalpain (-calpain) and millicalpain (m-calpain), are expressed ubiquitously in animal tissues. Calpain expression is increased in neuropathological processes, including Alzheimer's (13) and Parkinson's (14) diseases, suggesting involvement of this protease in neuronal stress responses.
Oxidative stress induces rapid increases in [Ca 2ϩ ] i by stimulating Ca 2ϩ influx from the extracellular environment and efflux from intracellular stores (15,16), thereby probably leading to calpain activation. These Ca 2ϩ fluxes may also elicit the activity of other enzymes such as calcineurin. The latter, also called protein phosphatase 2B, belongs to the family of calcium/calmodulin-dependent serine/threonine phosphatases. Calcineurin, a heterodimeric enzyme composed of a catalytic and a regulatory subunit (17) has been implicated in neuronal death. In fact, calcineurin is highly expressed in the central nervous system, especially in neurons vulnerable to ischemic and traumatic insult (for a review, see Ref. 18). Glutamate-induced early necrosis and delayed apoptosis in cerebellar granule cells (CGC) can be blocked by calcineurin inhibitors (19). Furthermore, adenovirus-mediated expression of constitutively activated calcineurin mutant induces cytochrome c/caspase-3-dependent apoptosis in neurons (20).
The aim of the present study was to investigate the interactions between oxidative stress signaling and neurotrophic pathways in CGC. We previously demonstrated the antiapoptotic role of calcium/calmodulin-dependent protein kinase IV (CaMKIV) in these neurons (21). CaMKIV is a serine/threonine kinase that phosphorylates a large variety of substrates and is implicated in the control of gene transcription because of its ability to phosphorylate several transcription factors (22)(23)(24). CaMKIV is expressed in cerebellar granule neurons (for a review, see Refs. 25 and 26) and is activated under depolarizing conditions (high extracellular potassium (HK): [K ϩ ] ϭ 30 mM). CaMKIV mediates HK-induced CGC survival by phosphorylating CREB. Several reports have indicated that both CaMKIV activity (27) and CREB phosphorylation at serine 133 (pCREB) (28,29) are involved in cell survival. We show here that oxidative stress induces a caspase-independent but calpain-dependent cleavage of CaMKIV and one of its nuclear targets, CREB. We also show that the levels of CaMKIV and CREB, but not those of pCREB, can be restored by treatment with calpain inhibitors, indicating the recruitment of an active dephosphorylating mechanism. Further, we demonstrate that co-treat-* This work was supported by ARC Grants 9821 and 4306. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported by the Ministère de la Recherche et de l'Enseignement Supérieur.

Colorimetric MTT Assay
Cells were cultured in 96-well culture dishes (Costar) and treated as indicated. Following treatment, mitochondrial activity (MTT) was assayed using a slight modification of the method described by Mossmann (31). Briefly, cultures were incubated for 1 h (37°C) with freshly prepared culture medium containing 0.5 mg/ml MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide; Sigma). After aspiration of the medium, the dark blue crystals formed were dissolved by adding 100 l/well of 0.04 N HCl in isopropyl alcohol. Plates were shaken at room temperature to ensure that all crystals were dissolved before taking absorbance readings on a Metertech S960 MicroELISA plate reader (test wavelength, 490 nm; nonspecific wavelength, 650 nm). Results are presented as a percentage of survival, taking control as 100%. Statistical analysis of the data was carried out using Graphpad Instat 2 software. Significant differences were assessed by means of analysis of variance, followed by the Student-Newman-Keuls test for multiple comparisons.

DNA Laddering
Cerebellar neurons cultured in 10-cm diameter plates were either treated with a pulse of 100 M H 2 O 2 (15 min) or switched to LK medium. After the indicated times of culture, cells were harvested, washed twice with phosphate-buffered saline (PBS), and lysed with 20 mM Tris-HCl (pH 7.5) containing 20 mM EDTA and 0.4% Triton X-100 for 15 min on ice. Centrifugation at 15,000 rpm eliminated cell debris and genomic DNA. The supernatant was incubated (1 h at 45°C) with an equal volume of a buffer containing 150 mM NaCl, 200 g/ml Proteinase K, 10 mM Tris, pH 8, 40 mM EDTA, and 1% SDS. After a phenol/chloroform (v/v) treatment, the upper phase was precipitated with ethanol and analyzed on a 1% agarose gel in the presence of RNase A (1 g/ml).

Hoechst Staining
Condensed and fragmented nuclei were evaluated in situ by intercalation of the fluorescent probe bisbenzimide, Hoechst 33342 (Sigma) into nuclear DNA (32). Briefly, after fixation with 4% paraformaldehyde in phosphate-buffered saline (30 min), cells were incubated with Hoechst 33342 (1 g/ml) for 45 min at room temperature. Hoechst 33342, which is cell-permeant and labels both intact and apoptotic nuclei, was visualized with an AMCA filter (excitation, 350 nm; emission, 450 nm). The nuclear surface sizes were assessed using the NIH Image Software. Nuclei were considered to be condensed if their surface area was at least 25% surface lower than that of the mean found for control cells. Nuclei showing distinct fragmentation were considered to be "fragmented nuclei." The percentage of condensed and fragmented nuclei in each treatment condition was computed from cell counts in at least five randomly chosen fields (40ϫ objective).

Measurement of Caspase-3 Activity
Caspase-3 activity was determined by measuring the release of 7-amino-4-methylcoumarin (AMC) from the caspase tetrapeptide CPP32 substrate DEVD-AMC (Calbiochem), as previously described (33). Briefly, neurons grown in 30-mm culture dishes were washed with phosphate-buffered saline and lysed with 100 l of lysis buffer (0.5% Igepal NP-4 (Sigma), 0.5 mM EDTA, 150 mM NaCl, 50 mM Tris, pH 7.5) on ice. One portion of the lysate was used for protein determination (Bradford method; Bio-Rad), and 50 l of lysate were added to a reaction mixture containing 100 l of 2ϫ reaction buffer (20 mM Hepes, pH 7.5, 50 mM NaCl, 2.5 mM dithiothreitol), 40 l of H 2 O, and 10 l of DEVD-AMC (final concentration 50 M) and incubated for 1 h at 37°C. The reaction was stopped by a 10ϫ dilution with ice-cold lysis buffer. Fluorescence of free AMC was measured (excitation and emission wavelengths: 360 and 465 nm, respectively), using a PerkinElmer Life Sciences HTS 7000 Bioassay Reader.

Transfections
Expression Vectors-pSV2-CREB-VP16 (a generous gift from C. Bancroft, Mount Sinai School of Medicine, New York) encodes a chimera protein where CREB-(1-341) is fused to the transcription activator region of VP16 and thus generates constitutively active CREB protein (34). CREB M1 was kindly provided by J. R. Lundblad (Vollum Institute, Portland, OR) This plasmid encodes a phosphorylation-deficient CREB protein mutated on serine 133 (changed to alanine) and has been reported as a dominant inhibitory mutant of CREB (35).
Gene Transfer-Neuronal gene transfer was performed as reported previously, using polyethyleneimine (25 kDa) as DNA carrier (36). After 3 days in vitro, cell cultures on glass coverslips, placed in 12-well plates, were transfected for 30 min with 0.75 g/ml of expression vector and 0.75 g/ml of EGFP-expression vector (CLONTECH). Plates were spun for 5 min at 1500 rpm. Twenty-four h after transfection and recovery in HK, cells were transferred to LK for 10 h or treated with a pulse of H 2 O 2 for 8 h. Apoptotic nuclei were visualized by Hoechst 33342 staining, and transfected cells were visualized by EGFP fluorescence.

Oxidative Stress Induces Necrotic-like Caspase-independent
Neuronal Cell Death-The characteristics of cerebellar granule neuron death induced by oxidative stress were analyzed and compared with those observed for cell death induced by exposure to low extracellular potassium levels (LK; [K ϩ ] ϭ 5 mM) (37,38). Oxidative stress was induced with a 15-min pulse of H 2 O 2 , and cell death was monitored by mitochondrial activity in the MTT assay. Fig. 1A   of H 2 O 2 to study the cell death pathways. Fig. 1B shows the temporal evolution of cell death after a pulse of 100 M H 2 O 2 ; neuronal death (22%) is clearly detectable 8 h after the initial H 2 O 2 pulse.
To further characterize the mechanisms underlying H 2 O 2induced neuronal death, We sought to identify typical morphological features of apoptosis. Using Hoechst 33342 nuclear staining, we observed that 78% of the cells showed condensed nuclei (without obvious nuclear fragmentation) within 10 h of a pulse of H 2 O 2 (15 min, 100 M) (Fig. 1C). In contrast, LK conditions induced a lower degree of chromatin condensation (24%), although nuclear fragmentation was detectable in 20% of the cells, consistent with our earlier findings (21,40). In the DNA laddering assay, we observed that H 2 O 2 failed to induce any laddering typical of apoptosis at 10 and 15 h after treatment (Fig. 1D), although obvious cell death was observable at these time points (Fig. 1B); however, an accumulation of large DNA fragments was detectable. In contrast to the results obtained with H 2 O 2 , exposure of cells to LK for 10 h resulted in a clear DNA ladder (of ϳ180-base pair fragments), characteristic of apoptosis.
We next assessed caspase-3 activation, another hallmark of the apoptotic pathway, by Western blotting and enzymatic detection. Fig. 1E (inset), shows a Western blot using an antibody against the active fragment of caspase-3 (p20). No difference in caspase-3 processing was detectable 6 h after the initial 15-min H 2 O 2 pulse (concentrations ranging from 25 to 200 M). This result was confirmed by in vitro measurements of caspase-3 activity. In contrast, LK induced a time-dependent increase of caspase-3 activity, as shown in Fig. 1E (41), and according to this author it is indicative of a calpain-dependent proteolysis.
The loss of CaMKIV during potassium deprivation (LK) induces a marked reduction in the levels of pCREB, although CREB stability remains unaffected (21). Since pCREB is the transcriptionally active form of CREB, and since it promotes cell survival (28,29), the correspondence between H 2 O 2 -induced loss of CaMKIV and pCREB levels were assessed. We found that H 2 O 2 treatment decreased pCREB levels in a doseand time-dependent manner (Fig. 2, C and D), with the maximal decrease occurring 8 h after the H 2 O 2 pulse (Fig. 2D). We also measured total levels of CREB (N-CREB, i.e. independent of phosphorylation state). In contrast to LK, H 2 O 2 induced a clear time-and dose-dependent loss of N-CREB (Fig. 2, C and  D). No N-CREB cleavage products were detectable (even after longer exposures). A time course analysis showed that the loss of pCREB precedes the loss of total N-CREB protein levels. Indeed, a loss of pCREB is observed 1 h after treatment, whereas the loss of N-CREB protein is only observed 2 h after H 2 O 2 treatment (Fig. 2D), suggesting that dephosphorylation and regulation of protein levels are not closely linked. These reductions in the levels of these two essential components of cell survival (CaMKIV and N-CREB) may well account for the death-promoting effects of oxidative stress.
Analysis of the mechanism involved in the regulation of CaMKIV and CREB protein levels was undertaken subsequently. CaMKIV has already been described to be cleaved by caspase-3 and calpain proteases (42). Since H 2 O 2 does not induce caspase-3 in the model used in these studies (Fig. 1E), we explored the possible contribution of calpains to the loss of CaMKIV and CREB. H 2 O 2 -induced CaMKIV degradation was completely reversed by both calpain inhibitors I and II (2 M) (Fig. 3A). Treatment with these inhibitors also abolished H 2 O 2induced loss of N-CREB, suggesting that both CaMKIV and N-CREB are sensitive to calpains. Interestingly, levels of pCREB were not reinstated after treatment with either of the calpain inhibitors (Fig. 3A), indicating that, even when N-CREB is present and CaMKIV is functional, an impairment of CREB phosphorylation persists. These results strongly suggest that the decreases in pCREB and N-CREB protein levels occur independently of each other. Interestingly, the inability of the calpain inhibitors to restore pCREB levels correlates with their inability to maintain cell viability (Fig. 3B); this observation suggests that another oxidative stress-induced pathway underlies the cell death observed, and it clearly points to a role for kinase/phosphatase-dependent regulatory mechanisms.
H 2 O 2 Induces CREB Dephosphorylation through Calcineurin-Since CREB phosphorylation is dependent on both kinases and phosphatases, we next monitored the consequences of phosphatase inactivation on CREB phosphorylation and cell survival. Fig. 4A shows that pretreatment of cells with microcystin-LR, a broad spectrum and cell-permeable serine/threonine phosphatase inhibitor (protein phosphatase 1, protein phosphatase 2A, and to a lesser extent protein phosphatase 2B) has little or no effect on N-CREB levels after either H 2 O 2 or calpain inhibitor treatments. Further, microcystin-LR treatment does not enhance pCREB levels during oxidative stress. However, in the presence of calpain inhibitor and under oxidative conditions, microcystin-LR significantly and strongly increases pCREB levels. The lack of effect of the phosphatase inhibitor (in the absence of calpain inhibitor) is thus best explained by the absence of N-CREB under these conditions. Together, these data indicate that H 2 O 2 recruits both calpain and serine/threonine phosphatasemediated signaling mechanisms.
Since calpain is a calcium-dependent enzyme, it is reasonable to expect that massive calcium influxes may be implicated in H 2 O 2 -induced cell death. Calcineurin, a calcium-dependent phosphatase, has been shown to participate in CREB dephosphorylation upstream of protein phosphatase 1 (43). We therefore investigated the effects of a calcineurin inhibitor, cyclosporin A, on H 2 O 2 -induced loss of pCREB and cell death. Like microcystin, cyclosporin A had no effect on N-CREB levels, whereas in the presence of calpain inhibitors, pCREB levels were fully restored (Fig. 4B). These results suggest that a 15-min pulse of H 2 O 2 , induces both calpain and calcineurin activation. Neither of the phosphatase inhibitors (microcystin-LR and cyclosporin A) affected H 2 O 2 -induced CaMKIV cleavage (data not shown).
Dephosphorylation and Proteolysis Are Cooperative Mechanisms That Mediate ROS-induced Neuronal Death-Given that the phosphorylation of CREB on serine 133 promotes cell survival (29), the loss of pCREB observed in oxidative conditions could account for the H 2 O 2 -induced cell death. We therefore tested whether blockade of pCREB loss by calpain and phos-phatases inhibitors would induce an attenuation of H 2 O 2 -induced cell death. As shown in Fig. 4C, phosphatase or calpain inhibitors alone had no effect on H 2 O 2 -induced cell death. However, when phosphatase and calpain inhibitors were added simultaneously, a significant reduction of H 2 O 2 -induced cell death was observed. Further, microcystin-LR, in the presence of calpain II inhibitor, reduced cell death by 40%, whereas, under the same conditions, cyclosporin A reduced cell death by only 22%. This result shows that microcystin-LR is more potent than cyclosporin A in maintaining cell viability, suggesting that calcineurin may not be the only phosphatase involved in H 2 O 2 -mediated CREB dephosphorylation.

ROS-controlled Dephosphorylation and Cell Death Are Dependent on de Novo
Protein Synthesis-To test whether oxidative stress operates through an active process of programmed cell death, we investigated whether H 2 O 2 -dependent cell death relies on protein synthesis. We thus examined the effect of the protein synthesis inhibitor cycloheximide on pCREB levels that are down-regulated 1 h after the H 2 O 2 pulse (Fig. 2D). Cycloheximide reverses the oxidative stress-induced loss of pCREB at this short period of H 2 O 2 -treatment (1 h) (Fig. 5A). Simultaneous treatment with cycloheximide and cyclosporin A did not enhance the recovery of pCREB levels further suggesting that cycloheximide interferes with the phosphatase pathway. In contrast, cycloheximide alone did not reverse either the H 2 O 2 -induced loss of pCREB or the loss of N-CREB (measured at 5 h after the H 2 O 2 pulse; Fig. 5B). However, when cycloheximide was used in the presence of a calpain inhibitor that prevents N-CREB loss, the protein synthesis inhibitor allowed pCREB to recover to control levels (Fig. 5B). Similarly, cotreatment with cyclosporin A did not influence the effects of cycloheximide (not shown). Together, these results suggest that phosphatase activation depends on de novo protein synthesis, whereas calpain activation does not.
Cycloheximide treatment was next tested on neuronal survival. As shown in Fig. 5C, cycloheximide alone was not able to inhibit H 2 O 2 -induced cell death; this result agrees well with the fact that the drug was found not to permit the recovery of pCREB levels at 5 h after H 2 O 2 treatment. When co-applied with the calpain inhibitor cpII, cycloheximide was found to reverse cell death by 25%. In contrast, cyclosporin A, when applied together with cycloheximide, was found to be inefficient at promoting survival, a result consistent with the notion of sustained, de novo protein synthesis-dependent phosphatase activity. Taken together, these results demonstrate that two independent pathways are activated during stress-induced neuronal death; one of these pathways clearly relies on the execution of a genetic program.
CREB Phosphorylation on Serine 133 Is Necessary to Promote Neuronal Survival-To test whether phosphorylation of CREB on serine 133 directly regulates neuronal survival, we performed transient co-transfection experiments with CREB mutants and an EGFP-expressing vector that allows the monitoring of transfected cells. Under these experimental conditions, transfection efficiency evaluated by taking into account only highly fluorescent GFP-positive cells was less than 1%. This is in good agreement with similar previously published studies on primary neuronal cultures (44,45). The effects of mutant CREB overexpression were monitored by scoring the alteration of nuclear morphology by Hoechst 33342 staining in the transfected cell population. Neurons co-transfected with the EGFP vector and an empty expression vector served as controls in the evaluation of cells displaying condensed nuclei. Overexpression of a dominant negative CREB mutant, which cannot be transcriptionally activated upon phosphorylation (S133A; CREB M1), was found to induce nuclear condensation (73% versus 30% in control cells; Fig. 6A). This indicates that genetic inactivation of CREB can promote apoptosis. In contrast, a dominant active mutant of CREB (in which CREB is fused to the transactivation domain of VP16; CREB-VP16) prevented both LK-and H 2 O 2 -induced chromatin condensation and fragmentation (Fig. 6B). It should be noted that the transfection procedure itself appears to have some deleterious side effects, since under these conditions the number of condensed nuclei in control appears slightly elevated (control in Fig. 6A, compared with control of nontransfected neurons (Fig. 1C)). These data conform with the view that phosphorylation of CREB on serine 133 is a key event in promoting cell survival and that the loss of CREB activity by dephosphorylation contributes to cerebellar granule neuron death. Further, these observations show that CREB is a final effector of both neuronal apoptosis (caspase-dependent death in LK) and ROS-induced necrotic-like death.

H 2 O 2 Induces a Programmed Neuronal Death Distinct from
Apoptosis-Potassium-deprivation (LK) induces a very well characterized apoptotic process in cerebellar granule cells (38). Using LK-induced apoptosis, we show here that oxidative stress triggers an active cell death that is morphologically and mechanistically different from apoptosis. Whereas LK-induced death occurs with typical chromatin condensation and fragmentation associated with caspase-3 activation and DNA laddering, a short pulse of H 2 O 2 that generates ROS is not accompanied by identical apoptotic features. Although chromatin condensation is observed after H 2 O 2 treatment, nuclear fragmentation is distinctly absent (Fig. 1C). Moreover, caspase-3 activity was not detected at any dose of H 2 O 2 (Fig. 1E); neither was any other form of caspase activity detectable when a wide range set of caspase substrates was used (data not shown). While participation of another as yet unidentified caspase(s) cannot be excluded with respect to H 2 O 2 -induced cell death, other data obtained in this study strongly indicate that oxidative stress-induced cell death is caspase-independent, at least in the CGC model. Further, our findings suggest that H 2 O 2induced CGC cell death is distinct from apoptosis. Thus, in DNA-laddering assays, we detected a band in the upper region of the gel that corresponded to a high molecular weight fragment of DNA that was not cleaved further. This finding is consistent with previously published data showing that H 2 O 2 treatment is associated with the appearance of 50-kilobase pair DNA fragments, with no further DNA laddering of the type characteristic of apoptosis (46). It seems probable that these high molecular weight DNA fragments result from H 2 O 2 activation of topoisomerase II and subsequent excision of the 30 -100-kilobase pair chromosomal loops (47).
There are conflicting views in the literature as to the type of cell death that occurs after exposure to H 2 O 2 . While some authors report that H 2 O 2 induces necrotic cell death (5), others conclude that this stimulus results in apoptotic cell death (6 -8). These differing views may reflect cell type dependence of the response (4) or differences in H 2 O 2 doses employed. For example, cells treated with low doses of H 2 O 2 (ϳ200 M) reportedly display apoptotic features, whereas cells exposed to higher doses of H 2 O 2 (ϳ1 mM) show necrotic features (48). Under the experimental conditions used here, no cell death was observed at H 2 O 2 concentrations lower than 50 M, and when higher doses were employed, the morphological and biochemical features observed were quite distinct from those of apoptosis.
While there is a large interest in apoptosis, much less attention has been given to understanding the process of necrosis, which is considered to be a degenerative response of the cells to insurmountable damage. The general opinion is that necrosis is a passive, nonspecific event, in opposition to apoptosis, which is considered to be an active process. According to these definitions, it is implicit that once a cell receives an insult of a certain quality or intensity, its viability will be irreversibly compromised. Contrasting with this view, our data now suggest that necrotic cell death is in fact genetically programmed and regulated, suggesting that the process may be interrupted at specific regulatory steps. Our results show that ROS-mediated cell death, induced here by treating CGC with H 2 O 2 , recruits two independent mechanisms, proteolysis and phosphatase activation, which eventually lead to cell death; further, our studies demonstrate that (partial) escape from H 2 O 2 -induced cell death can only occur if both pathways (proteolysis and phosphatase activation) are simultaneously inhibited (Fig. 4). Our results are supported by a recent report demonstrating that H 2 O 2induced necrotic cell death in U937 cells can also be modulated by the addition of the PARP inhibitor 3AB (49). Here, we also show that one of the recruited signaling pathways is dependent on de novo protein synthesis. Thus, the requirement of specific genetically programmed signaling pathways strongly suggests that the necrosis-like cell death observed in CGC is not a passive event. Indeed, two previous studies (50,51) have shown that programmed cell death may be seen in cells displaying necrotic morphology. This form of cell death, termed "active necrotic-like cell death," appears to be physiologically relevant, since it has been observed during development (type II cell death, autophagic degeneration) (52) as well as in Alzheimer's and Parkinson's diseases (53,54). We therefore propose that H 2 O 2 can induce active necrotic-like cell death in CGC.
Recruitment of a Preexisting Signaling Pathway and a Genetic Program in Oxidative Stress-induced Cell Death-In exploring the molecular mechanisms that might be implicated in oxidative stress-induced neuronal death, we found that CaMKIV and CREB are down-regulated under oxidative conditions. Both of these proteins are involved in cell survival. CaMKIV has been demonstrated previously to promote the survival of a variety of cell types including CGC (21), T-cells (55) or Purkinje cells (56). CREB, which was initially described as a regulator of synaptic plasticity, also promotes neuronal survival (28,57). The reduced levels of these two neuroprotective proteins may thus account for ROS-induced cell death. We show here that calpain inhibitors can prevent the downregulation of these proteins (Fig. 3A). Further, we demonstrate that H 2 O 2 induces rapid dephosphorylation of CREB by in-creasing the activity of a serine/threonine-dependent phosphatase (Fig. 4C), more precisely, that of calcineurin-like activity. Clearly, the two signaling mechanisms initiated by ROS (proteolysis and dephosphorylation) can independently lead to neuronal death.
The ROS-induced phosphatase activity was shown to be a genetically programmed event, since cycloheximide, a protein synthesis inhibitor, inhibited the rapid loss of pCREB (1 h). The loss of N-CREB observed at later time points seems to be independent of any genetic program, since it is not affected by cycloheximide (Fig. 5). Together, these data show that ROS can induce a novel form of active cell death that is clearly distinct were cultured in the absence or presence of cycloheximide (cx, 10 g/ml, 1-h pretreatment) and, where indicated, in the presence of cyclosporin A (cyclo A; 1 g/ml; 1-h pretreatment) and calpain inhibitor II (cpII, 2 M). A and B, phosphorylated CREB levels (pCREB) and total CREB levels (N-CREB) were monitored by Western blot 1 (A) or 5 (B) h later. C, 12 h after the H 2 O 2 pulse, cell survival was assessed by MTT assay, represented relative to control values (100%). * (p Ͻ 0.05) indicates statistically significant differences compared with data obtained for the H 2 O 2 pulse (no other treatment) (analysis of variance, followed by Student-Newman-Keuls multiple comparison test). from apoptosis. The underlying mechanism(s) appears to be complex, involving the recruitment of at least two signaling pathways (proteolysis and phosphatase activity). Phosphatase activity comes into play first, and surprisingly, this signaling pathway depends on de novo protein synthesis. Interestingly, cell survival is only maintained when both proteolysis and phosphatase activity are blocked. Our data are thus in line with the observations of Lee et al. (58), who showed that protein synthesis inhibitors alone do not attenuate H 2 O 2 -induced cell death, and with those of Ciani et al. (59), who found that glutamate toxicity, which is known to be mediated by ROS, is only partly dependent on protein synthesis.
Molecular Mechanism Underlying ROS-induced Proteolysis and Dephosphorylation-From the data presented in this paper, it becomes clear that calpain activation is one of the mechanisms by which ROS induce neuronal death. The increase of calpain activity under oxidative conditions has recently been demonstrated in PC12 cells (11) and in a mouse embryonic carcinoma P19 cell line (12). It thus appears that calpain activation is a general mechanism that operates not only in proliferating and transformed cell lines but also in postmitotic differentiated neurons.
Oxidative stress not only stimulates calpain activation but also that of phosphatases. Thus, we show that following H 2 O 2 treatment, CREB is dephosphorylated downstream of the calcium/calmodulin-dependent phosphatase calcineurin. This phosphatase has been previously described to regulate dephosphorylation of pCREB in hippocampal neurons (43), and calcineurin has been implicated in neuronal apoptosis (19,20) and in neurodegenerative diseases (18,60). However, activation of this phosphatase during oxidative stress is not in line with other reports that superoxide-induced oxidative stress results in an inhibition of calcineurin activity (61)(62)(63). It should be noted, however, that H 2 O 2 itself does not inhibit basal calcineurin activity in neutrophils (61). Thus, it is likely that different oxidative species (e.g. hydroxyls versus superoxides) can impinge on different signaling pathways (51); in addition, cell and tissue type may be important in determining these differences.
Several findings suggest that an additional phosphatase(s), distinct from calcineurin, is involved in oxidative stress-induced cell death. First, in the paradigm used here, the microcystin-LR inhibitor was found to be more potent than cyclosporin A in preventing cell death (Fig. 4C), suggesting that H 2 O 2 may also activate protein phosphatase 1 or protein phosphatase 2A. This interpretation is in line with other work showing that protein phosphatase 1 mediates pCREB dephosphorylation downstream of calcineurin (43). In addition, it should be noted that protein phosphatase 2A is involved in CaMKIV dephosphorylation and inhibition in T-lymphocytes (64). Thus, microcystin-LR blockade of this pathway may indirectly lead to an increase in the levels of phosphorylated CREB.
The initial mechanism by which ROS activate both proteolysis and dephosphorylation remains an open question. Calpain activation is calcium-dependent and increased levels of intracellular free calcium during oxidative conditions are well documented (15,16). Similarly, activation of calcineurin depends on elevated intracellular Ca 2ϩ levels. It is therefore tempting to speculate that increased free Ca 2ϩ plays a major role in initiating calpain and calcineurin signaling during oxidative stress.
Nuclear Consequences of ROS-induced Proteolysis and Dephosphorylation-At the nuclear level, proteolysis and dephosphorylation ultimately participate in the reduction of transcriptionally active pCREB. The essential role of this transcription factor in governing neuronal survival is shown here by two means (Fig. 6). The CREB-M1 mutant, which cannot be phosphorylated on serine 133, can promote nuclear condensation under culture conditions (HK) that would otherwise promote neuronal survival. Conversely, activation of CREB-dependent gene transcription by the CREB-VP16 mutant reduces cell death in response to oxidative stress. Interestingly, this dominant active CREB not only reduces ROS but also LK-induced neuronal death; in the latter case, all of the signs of classical apoptosis are manifest (38). This strongly suggests that CREB serves as a common final effector for several types of neuronal death, irrespective of the intracellular pathway activated (caspase-dependent or -independent cell death). This view is in line with growing evidence that CREB is implicated in neuronal survival (e.g. PC12 and Neuro 2A cells lines that overexpress CREB are more resistant to okadaic acid-induced cell death (29)). How CREB exerts its neurotrophic effect is still unclear, but it has been reported that CREB is involved in transcription of the antiapoptotic gene bcl-2 (28). Thus, oxidative stress-induced dephosphorylation of CREB could result in transcriptionally inactive CREB, thus silencing bcl-2 transcription and initiating neuronal death. Another study has shown that CREB positively regulates brain-derived neurotrophic factor (65) which, via an autocrine loop, exerts neurotrophic actions on CGC. It can thus be envisaged that inactivation of CREB by ROS interrupts this survival loop by inhibiting brainderived neurotrophic factor production. Further investigations will be required to determine which precise mechanism, secondary to the loss of pCREB in CGC, ultimately mediates ROSinduced neuronal death.