Novel Roles for Protein Kinase Cδ-dependent Signaling Pathways in Acute Hypoxic Stress-induced Autophagy*

Macroautophagy, a tightly orchestrated intracellular process for bulk degradation of cytoplasmic proteins or organelles, is believed to be essential for cell survival or death in response to stress conditions. Recent observations indicate that autophagy is an adaptive response in cells subjected to prolonged hypoxia. However, the signaling mechanisms that activate autophagy under acute hypoxic stress are not clearly understood. In this study, we show that acute hypoxic stress by treatment with 1% O2 or desferroxamine, a hypoxia-mimetic agent, of cells renders a rapid induction of LC3-II level changes and green fluorescent protein-LC3 puncta accumulation, hallmarks of autophagic processing, and that this process involves protein kinase Cδ (PKCδ), and occurs prior to the induction of BNIP3 (Bcl-2/adenovirus E1B 19-kDa interacting protein 3). Interestingly, hypoxic stress leads to a rapid and transient activation of JNK in Pa-4 or mouse embryo fibroblast cells. Acute hypoxic stress-induced changes in LC3-II level and JNK activation are attenuated in Pa-4 cells by dominant negative PKCδKD or in mouse embryo fibroblast/PKCδ-null cells. Intriguingly, the requirement of PKCδ is not apparent for starvation-induced autophagy. The importance of PKCδ in hypoxic stress-induced adaptive responses is further supported by our findings that inhibition of PKCδ-facilitated autophagy by 3-methyladenine or Atg5 knock-out renders a greater prevalence of cell death following prolonged desferroxamine treatment, whereas PKCδ- or JNK1-deficient cells exhibit resistance to extended hypoxic exposure. These results uncover dual roles of PKCδ-dependent signaling in the cell fate determination upon hypoxic exposure.

Although autophagy has been characterized in many contexts, the signaling pathway that activates autophagy in response to hypoxic stress has not yet been studied extensively. Oxygen deprivation by chronic hypoxia is known to induce autophagy. For example, Beclin-1 levels in the cortex and striatum are increased in cerebral ischemia (11). The early changes of Beclin-1 levels in the penumbra occurred at 6 h, peaked at 24 h, and lasted for at least 2 days in the neuronal cells (12). Beclin-1 (also called Atg6) is a phylogenetically conserved protein that is essential for the initiation of autophagy, presumably via its interaction with the class III PI3K/Vps34 (1,13). Originally, human Beclin-1 was identified as an interactor of Bcl-2 (14). Beclin-1 reportedly possesses a so-called Bcl-2 homology region-3 (BH3) domain (amino acids 114 -123) that mediates the interaction with Bcl-2 and other Bcl-2 homologues, such as Bcl-X L and Mcl-1 (15). It has recently become clear that overexpression of Beclin-1 or depletion of Bcl-2 stimulates autophagy (16). Herein, we demonstrate a novel pathway by which acute hypoxic stress utilizes a rapid activation of the PKC␦ signaling pathway to release Beclin-1 from Bcl-2, leading to autophagy induction.
Myc-Bcl-2 Wild Type and Myc-Bcl2-S70A Mutant-Human Bcl-2 cDNA was reverse transcribed from total RNA with the gene-specific primer, 5Ј-GGTACATCACTGACAATGCA-3Ј. Subsequently, the Bcl-2-encoding fragment, tagged with appropriate restriction enzyme sites, was amplified from the cDNA by PCR and inserted between EcoRI and HindIII sites of expression vector pCMV-Tag3A (Stratagene). The S70A mutant was generated using the Transformer site-directed mutagenesis kit (Clontech) and the mutagenic primer, 5Ј-CGCCAGGACCGCGCCGCTGCAG-3Ј. DNA sequences of these constructs were verified by sequencing reactions.
GFP-LC3 Puncta Analyses-Pa-4, MCF-7/caspase-3, or MCF-7/Neo cells were transfected with GFP-LC3 or GFP-N1 expression constructs by using Lipofectamine 2000 (Invitrogen) per the manufacturer's instructions. For GFP-LC3 puncta analyses, cells at 36 h post-transfection or stably transfected MEF/GFP-LC3 cells were seeded on coverslips placed in the 6-well plates and cultured overnight, followed by various treatments for different time periods. Cells were then fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) for 30 min, incubated with 4Ј-6-diamidino-2-phenylindole (Molecular Probes) for 10 min and mounted with the Prolong Gold Antifade reagent (Molecular Probes). Images were acquired with a Zeiss inverted LSM 510 Meta 2 photon microscope and analyzed using a Zeiss LSM image examiner. The GFP-LC3 punctae and GFP-LC3-positive cells were examined and quantified in more than five fields per slide on five slides. The increase in GFP-LC3 punctae represents the relative accumulation of autophagosomes.
Measurement of GFP-LC3 Intensity by Fluorescence-activated Cell Sorter (FACS) Analysis-MEF cells that stably expressed GFP-LC3 were seeded subconfluently on the day before treatment in 6-well plates. Following the indicated treatment periods, cells were harvested with trypsin/EDTA, washed with PBS, and fixed with 2% paraformaldehyde in PBS for 30 min at room temperature. Fixed cells were washed with PBS twice prior to subjecting them to FACS. Analysis of 1 ϫ 10 5 cells/sample was performed by a CyAn TM ADP 9 Color Flow cytometer (DAKO; Analytic Cytometry Core Facility in City of Hope Medical Center), and viable cell counts were plotted as GFP fluorescence intensity by FlowJo Software (Tree Star, Inc., Ashland, OR). The level of GFP fluorescence intensity in each treated sample was normalized to the level of resting, vehicletreated controls set at 100%. The relative level of GFP-LC3 intensity in each treatment was calculated from at least three independent experiments and represents mean Ϯ S.D. in the graphs.
Whole Cell Lysate Preparation and Western Analyses-For protein phosphorylation studies, whole cell lysates were extracted by SDS lysis buffer, as previously described (17), containing both the Complete protease inhibitor mixture (Roche Applied Science) and phosphatase inhibitor Na 3 VO 4 (2 mM; Sigma) and subjected to SDS-PAGE and subsequently immunoblotted with respective antibodies for BNIP3 (Abcam), Bcl-2, or c-Myc (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), phospho-SAPK/JNK (Thr-183/Tyr-185), SAPK/JNK, phospho-c-Jun (Ser-63), c-Jun, phospho-Bcl-2 (Ser-70), phospho-AMPK␣ (Thr-172), or AMPK␣ (Cell Signaling Technology). Anti-PKC␦ and anti-GFP antibodies were purchased from Santa Cruz Biotechnologies. A mouse anti-Beclin-1 antibody was purchased from BD Biosciences. A mouse anti-phosphotyrosine antibody (clone 4G10) was purchased from Upstate Biotechnology. To detect endogenous LC3-I and LC3-II, whole cell lysates were prepared using a Triton X-100-based lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100) plus the Complete protease inhibitor mixture (Roche Applied Sci-ence). Equal amounts of proteins were aliquoted and added with SDS-PAGE loading buffer immediately before boiling, followed by SDS-PAGE and subsequent immunoblotting with an anti-LC3 antibody (MBL International Co.). Anti-tubulin (Santa Cruz Biotechnology) and anti-actin (Chemicon International) antibodies were used to assess equal protein loading. Immunoblots using an enhanced chemiluminescence detection kit (ECL-Plus; Amersham Biosciences) were imaged with the VersaDoc 5000 Imaging System (Bio-Rad). Densitometric data were captured, quantitated with Quantity One Software (Bio-Rad), and normalized with internal control proteins, individually, in each experiment. The relative level of a particular protein altered by various treatments was then calculated by setting the normalized value in the control as 1, assuming equal variances.
Immunoprecipitation-Whole cell lysates were prepared using radioimmune precipitation buffer (25 mM Tris, 125 mM NaCl, 1% Nonidet P-40 (Nonidet P-40), 0.1% SDS, 0.5% sodium deoxycholate, 0.004% sodium azide, pH 8.0) containing both the Complete protease inhibitor mixture and PhosSTOP phosphatase inhibitor mixture (Roche Applied Science). Protein concentrations were determined by a BCA assay (Bio-Rad). Whole cell lysates (1 mg) were incubated with 1-5 g/ml of a specific antibody for Beclin-1 (Abcam), PKC␦, or c-Myc at 4°C for 2 h, followed by incubation with 20 l of Protein A/G PLUSagarose beads (Santa Cruz Biotechnologies) at 4°C overnight to capture immune complexes. Immune complexes were then washed three times with PBS containing 1% Nonidet P-40, 0.5% sodium deoxycholate, and the Complete protease inhibitor mixture, dissolved by adding SDS-PAGE loading buffer, boiled for 6 min, resolved by SDS-PAGE, transferred to membranes, and subsequently blotted with appropriate antibodies. The levels for proteins of interest were determined using an ECL-Plus kit. The relative level of immunoprecipitated complex was normalized with total input proteins in lysates individually and normalized by setting the nontreatment sample value as 1.
Cell Viability Assay-Cell viabilities were determined by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. Pa-4, MEF/WT and various knock-out cell lines were seeded into 24-well plates to reach 35% confluence on the day of treatments. The cells were treated with different regimens or vehicle as indicated for 0, 24, and 48 h, followed by the 3--(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Sigma) assay according to the manufacturer's recommendation. Absorbance of each well was read at 540 nm in a scanning multiwell spectrophotometer. The results were depicted as percentage for cell viability and reported as the mean Ϯ S.D. of three independent experiments performed in triplicate.
Statistical Analyses-Experiments were independently carried out at least three times, and one representative data set of the three independent experiments was presented where appropriate. The results were evaluated for statistical significance by two-way analyses of variance with randomized sample blocks, followed by post hoc comparisons based on Fisher's least squared difference method of protected t-tests. The error bars were marked as the S.D. of the mean. p values less than 0.05 were regarded as significant.

Increased Autophagosome Accumulation during Acute
Hypoxic Stress-We previously reported that treatment of cells with DFO, a hypoxia-mimetic agent, confers a cross-talk between the activation of PKC␦ and apoptotic caspase-3 pathways (17). Since signaling pathways that regulate cell proliferation and apoptosis are frequently associated with the regulation of autophagy, we hypothesized that hypoxia or DFO treatment induces autophagy. To test this hypothesis, induction of autophagy detectable by Western analyses on the time course of the changes in the amount of membrane-bound LC3-II (the phosphatidylethanolamine-conjugated form (5) and by GFP-LC3 puncta accumulation (reviewed in Ref. 18) was investigated. In Pa-4 cells, endogenous LC3 is mainly presented as the lipidated form of LC3-II in the normoxic and nutrient-rich condition (Fig. 1A, top, lane 1), similarly reported in A431, HeLa, and MDA-MB-231 cells (19).
First, we monitored the time course of the changes in LC3-II level as a means to assess autophagy progression, since it was reported that LC3-II level is decreased inside autolysosomes during the autophagic process (18). Due to its high basal LC3-II level in Pa-4 cells under normoxic conditions, we pretreated cells with lysosomal protease inhibitors, E64d and pepstatin A, for 2 h prior to hypoxic exposure to ascertain that autophagic flux under hypoxic stress can be reliably detected (18). Upon acute hypoxia treatment with 1% O 2 , LC3-II markedly accumulated in the presence of E64d and pepstatin A compared with those in the absence of these lysosomal protease inhibitors (Fig.  1A, second panel, lanes 4, 6, and 8 versus lanes 3, 5, and 7). Under hypoxic conditions, the amount of LC3-II in the presence of these inhibitors was about 4-fold higher than that in their absence in Pa-4 cells (Fig. 1A, second panel, lanes 6 and 8 versus lanes 5 and 7), suggesting that LC3-II is degraded by lysosomal hydrolases much faster under hypoxic conditions. Second, we monitored GFP-LC3 puncta formation as an index for autophagosome accumulation in Pa-4 cells that had been transiently transfected with an expressing vector for GFP-LC3 (4), followed by DFO exposure. The formation of GFP-LC3-labeled structures, GFP-LC3 punctae representing autophagosomes, was induced in cells exposed to DFO (Fig. 1B, top panels, and supplemental Fig. S1A, indicated by arrows). Morphometric analyses of the GFP fluorescence images indicated that GFP-LC3 punctae were transiently induced at 2 h posttreatment but tapered off after reaching a maximal induction at 4 h post-treatment (Fig. 1C). Since CQ has been shown to inhibit the fusion of lysosomes with autophagosomes and to amplify GFP-LC3 puncta signals (18), treatment with CQ for 4 h was used to confirm the expected increase in GFP-LC3 puncta accumulation (Figs. 1, B (middle panels) and C). In addition, transfection with the parental vector GFP-N1 followed by DFO exposure was used to validate the specificity of our assays in response to DFO (Fig. 1B, bottom panels).
To avoid nonspecific accumulation (but not dependent on autophagy) of GFP-LC3 in cells transiently transfected (18), MEF/GFP-LC3 cells (MEF cells stably transfected with GFP-LC3) were used to examine the ability of DFO for the indicated time periods to induce GFP-LC3 punctae. As shown in Fig. 1D (a), GFP-LC3 signals were equally distributed throughout the cells prior to DFO treatment. Upon exposure to DFO, green signals redistributed from nuclei and aggregated in perinuclear areas, where the intensity of signals reached a maximum at 4 h posttreatment (supplemental Fig. S1B), followed by a gradual decrease in signals for GFP-LC3 punctae by 6 h post-treatment (Fig. 1D, a-e). The MEF/GFP-LC3 cells were also subjected to amino acid deprivations by EBSS treatment for 4 h to induce GFP-LC3 punctae (Fig. 1D, f), serving as the control.
Last, to further characterize the DFO-induced autophagic process, the relative amount of lipidated LC3-II in MEF/Atg5 Ϫ/Ϫ cells during DFO exposure was assessed by Western analyses. As illustrated in Fig. 1E (left top two panels), the level of lipidated LC3-II transiently increased in wild type MEF cells, confirming that the hypoxic effectmediated change in LC3-II level is not restricted to Pa-4 cells. Consistent with the thesis that Atg12-Atg5 conjugates have an E3-like activity to enhance LC3 lipidation (6), no LC3 processing was observed in Atg5 Ϫ/Ϫ cells during DFO stress ( Fig. 1E, right top panel). We also observed a reproducible increase in LC3-I level ( Fig. 1E) in response to DFO treatment. Although it is possible that DFO induces LC3 expression, the exact nature of this phenomenon is still unclear. Taken together, we conclude that autophagic process is stimulated in Pa-4 and MEF cells upon acute exposure to 1% O 2 or DFO.
PKC␦ Is Required for DFO (but Not EBSS)-induced Autophagy-To examine whether or not PKC␦ is involved in DFO-induced autophagic process, we then demonstrated that the time-dependent changes in LC3-II level are almost completely abrogated in DFO-treated Pa-4/ PKC␦KD cells ( Fig. 2A, left). The lack of LC3 processing in DFOtreated Pa-4/PKC␦KD cells was in agreement with our previous observations that the K376R mutation of A, hypoxia (1% O 2 )-induced degradation of LC3-II is inhibited by a combined treatment of lysosomal protease inhibitors, E64d and pepstatin A. Pa-4 cells were pretreated with a mixture of E64d and pepstatin A (10 g/ml each) or vehicle for 2 h prior to exposure of 1% O 2 for the indicated time periods. Whole cell lysates were harvested and analyzed by Western analyses with anti-LC3 and anti-actin antibodies, respectively. Images were visualized with long exposures (L, top panel) to show both LC3-I and LC3-II bands, whereas the intensity of each LC3-II band was also acquired with shorter exposures (S, second panel) for quantitation purposes. Relative levels of LC3-II are shown as -fold induction (numbers in italic type), whereas nontreatment (control) is designated as 1. B, DFO treatment induces GFP-LC3 puncta accumulation in Pa-4 cells. Pa-4 cells were transfected with either GFP-LC3 or GFP-N1 and treated with either DFO (50 M) or chloroquine (CQ; 10 M) for the indicated time periods. These cells were then examined by confocal microscopy. Autophagosome formation was defined by the accumulation of GFP-LC3 punctae (indicated by arrows). C, quantitation of GFP-LC3 punctae in Pa-4 cells. The GFP-LC3 punctae were counted individually by three investigators blind to experimental conditions. The ratio of counted GFP-LC3 punctae over the number of GFP-positive cells was shown as a percentage. D, DFO treatment induces GFP-LC3 puncta accumulation in MEF/GFP-LC3 cells. MEF/GFP-LC3 cells were treated with DFO (100 M) for the indicated time periods and subjected to confocal microscopic analyses. Green fluorescence represents GFP-LC3 signals, whereas the blue color represents nuclear staining (4Ј,6-diamidino-2-phenylindole (DAPI) staining). Images shown are representatives from one of three independent experiments. Autophagosome formation was indicated by the accumulation of GFP-LC3 punctae (indicated by arrows). E, DFO induces the conversion of LC3-I to LC3-II in MEF, but not in MEF/Atg5 Ϫ/Ϫ , cells. Cells were treated with DFO (100 M) for the indicated time periods before harvesting. Equal amounts of whole cell lysates were subjected to Western analyses with anti-LC3 and anti-actin antibodies, respectively. Images were taken with long exposures (L, top panel) to show both LC3-I and LC3-II bands, whereas the intensity of each LC3-II band was also acquired with shorter exposures (S, middle panel) for quantitation purposes. Relative levels of LC3-II were quantitated and shown as described in Fig. 1A. DECEMBER 5, 2008 • VOLUME 283 • NUMBER 49 PKC␦, PKC␦KD, is a dominant negative form of PKC␦ (17). In contrast, EBSS rendered a comparable LC3 processing in both Pa-4 and Pa-4/PKC␦KD cells ( Fig. 2A, right). These results suggest that the autophagy pathway is functional in Pa-4/PKC␦KD cells, whereas the DFO-activated signaling pathway(s) to stimulate autophagy is disrupted by the dominant negative PKC␦KD. We next sought to address the stimulatory role of PKC␦ in DFO-induced autophagy directly, rather than an offtarget of PKC␦KD, using both MEF and MEF/PKC␦ Ϫ/Ϫ cells to assess the LC3 processing in response to DFO stress. The expression level of PKC␦ in MEF and MEF/PKC␦ Ϫ/Ϫ cells was confirmed by Western analyses (data not shown). As shown in Fig. 2C, LC3-II levels transiently peaked in DFO-exposed MEF cells (top, lanes 1-4), whereas the LC3-II level remained unchanged during the course of DFO treatment in MEF/PKC␦ Ϫ/Ϫ cells (top, lanes [5][6][7][8]. Pretreatment of cells with 3-MA, a nucleotide derivative that inhibits the early stage of autophagosome formation (1), largely inhibited the DFO-induced changes of LC3-II level in both Pa-4 (Fig. 2, B versus A) and MEF (Fig. 2C, third panel, lanes  1-4) cells, suggesting the critical role of PI3K in the induction of autophagy by hypoxic stress. Taken together, we conclude that hypoxic stress utilizes a PKC␦-dependent and 3-MA-sensitive pathway to induce autophagy.

Temporal and Dual Roles of PKC␦ in Adaptive Responses
Tyrosine phosphorylation of PKC␦ at a specific Tyr residue(s) is one of the activation mechanisms in response to different stimuli, such as UV, H 2 O 2 , and etoposide (20). As shown in Fig. 2D, Tyr phosphorylation of PKC␦ was detected between 0.5 and 1 h post-DFO treatment. Moreover, when WT or EGFP-PKC␦ mutated at Y64F, Y155F, or Y64F/Y155F was transfected into MEF/PKC␦ Ϫ/Ϫ cells and subjected to DFO treatment, both Y-64 and Y-155 appeared to be involved in DFO-induced PKC␦ activation, as seen by immunoprecipitation, followed by Western analyses (Fig. 2E). The specific mechanisms underlying the observed interdependence between DFO-induced Y-64 and Y-155 phosphorylation remain to be investigated.
Hypoxia Stress Induces PKC␦-dependent Activation of JNK-We next investigated which of the PKC␦-activated downstream signaling pathways is required for the formation of autophagosomes in response to hypoxic stress. Previously, Liu et al. (21) demonstrated that PKC␦ augments UV-induced JNK activation. We thus tested whether or not DFO activates JNK in Pa-4 and Pa-4/PKC␦KD cells. Since DFO treatment conveyed a preferential activation on short forms of JNK1/2 (JNK1/2 S ) in both Pa-4 and MEF cells, only p-JNK1/2 S images are shown in Fig. 3. Notably, the exposure of these cells to DFO results in a transient c-Jun phosphorylation at Ser-63, in relation with JNK1/2 S activation (phosphorylation of Thr-183/Tyr-185) in Pa-4 (Fig. 3A, lanes 1-4), but not in Pa-4/ PKC␦KD (Fig. 3B, lanes 1-4), cells. To our surprise, treatment with 3-MA almost abolished both JNK activation and c-Jun phosphorylation (Fig. 3A, lanes 5-8 versus lanes 1-4). An identical observation was made in the pair of MEF and MEF/ Equal amounts of whole cell lysates were subjected to immunoprecipitation (IP) with an anti-PKC␦ antibody, followed by immunoblotting analyses (IB) with an anti-phosphotyrosine 4G10 antibody. Western analyses of 5% input lysates using an anti-PKC␦ or anti-actin antibody served as protein expression controls. The relative level of immunoprecipitated complex was normalized with total input lysates individually and calculated by setting the basal phosphorylation level prior to DFO treatment as 1. E, both Tyr-64 and Tyr-155 of PKC␦ are involved in DFO-induced PKC␦ phosphorylation. MEF/PKC␦ Ϫ/Ϫ cells were transiently transfected with WT or Y64F-, Y155F-, or Y64F/Y155F-EGFP-PKC␦. Transfected cells were treated with DFO (200 M) for the indicated time periods. Equal amounts of whole cell lysates were subjected to immunoprecipitation with an anti-GFP antibody, followed by immunoblotting with an anti-phosphotyrosine 4G10 antibody. Western analyses of 5% input lysates using an anti-GFP or anti-actin antibody served as protein expression controls. The relative level of DFO-induced Tyr phosphorylation was calculated after normalizing with total input proteins and by setting the intensity obtained in nontreated EGFP-PKC␦ WT transfectant as 1.
PKC␦ Ϫ/Ϫ cells upon DFO exposure (Fig. 3, C and D). Importantly, exposure of Pa-4 and MEF cells to 1% O 2 also rendered a rapid and transient induction of JNK1/2 S , which was almost completely abrogated in MEF/PKC␦ Ϫ/Ϫ cells (Fig. 3E). The enhancement afforded by PKC␦ on UV-induced activation of both long (JNK1/ 2 L ) and short forms of JNK1/2 was demonstrated (supplemental Fig. S2A) in support of validating our observations. BNIP3 (Bcl-2/adenovirus E1B 19-kDa interacting protein 3), a prodeath member of the Bcl-2 family, is reportedly regulated by the hypoxia-induced transcription factor HIF-1␣ and plays a critical role in hypoxia-induced autophagy (22). A timedependent induction of BNIP3 expression was observed in DFO-or 1% O 2 -treated MEF and Pa-4, but not MEF/PKC␦ Ϫ/Ϫ , cells (Fig. 3F), suggesting its PKC␦ dependence. Intriguingly, the maximal induction of BNIP3 by hypoxic stress fell behind the observed rapid stimulation of LC3 changes (Figs. 1A and 2A) and GFP-LC3 puncta accumulation (Fig. 1, B-D). The discordance between delayed BNIP3 induction and rapid autophagy stimulation under hypoxic context suggests that BNIP3 may play a lesser role in acute hypoxic stressinduced autophagy.
JNK1 Is Activated by Acute Hypoxia-Two closely related JNK isoforms, JNK1 and JNK2, exhibit shared functions. However, they also have distinctly different biological activities (23). To ascertain whether or not JNK1 or JNK2 is required for DFO-induced autophagy, we examined LC3 processing and the JNK activation profile in response to DFO treatment in MEF that are deficient in either JNK1 or JNK2 allele. Immunoblotting analyses using an anti-JNK antibody, which can detect both long forms (p54) and short forms (p46) of JNK1 or JNK2 confirmed the expression of JNK1 and JNK2 in respective JNK2 Ϫ/Ϫ and JNK1 Ϫ/Ϫ MEF cells (Fig. 4A). As shown in Fig. 4B, we only detected phospho-JNK1/2 S signals in WT and JNK2 Ϫ/Ϫ , and not JNK1 Ϫ/Ϫ , MEF cells (top panels). Immunoblotting analyses using an anti-phospho-c-Jun (Ser-63) antibody revealed that the c-Jun phosphorylation induced by DFO was much weaker in JNK1 Ϫ/Ϫ cells compared with those detected in wild type and JNK2 Ϫ/Ϫ cells (Fig.  4B, second panels). Endogenous LC3 processing was induced by DFO in JNK2 Ϫ/Ϫ , but not JNK1 Ϫ/Ϫ , cells in a manner similar to that in MEF wild type cells (Fig. 4C, top panels). These results shown in Figs. 3 and 4 together suggest that, while PKC␦ is involved in hypoxic stressinduced autophagy, JNK1 activation is necessary for PKC␦-dependent LC3 processing. Our observations that JNK activation was attenuated in both Pa-4/PKC␦KD and PKC␦ Ϫ/Ϫ cells (Fig.  3, B, D, and E)   After normalizing with actin levels, the relative levels of each protein were shown as numbers in italic type, where nontreatment control level is designated as 1. E, 1% O 2 exposure induces JNK and c-Jun phosphorylation in MEF and Pa-4 but not MEF/PKC␦ Ϫ/Ϫ cells. MEF, MEF/PKC␦ Ϫ/Ϫ , and Pa-4 cells were exposed to 1% O 2 for the indicated time periods before harvesting cells with SDS-PAGE lysis buffer containing the phosphatase inhibitors. Equal amounts of whole cell lysates were subjected to Western analyses and blotted with specific antibodies as indicated. Relative levels of individual proteins were shown as numbers in italic type. Nontreatment control level is set as 1. F, prolonged DFO treatment induces BNIP3 expression in MEF and Pa-4 cells. Cells were exposed to either 200 M DFO or 1% O 2 for the indicated time periods before harvesting. Equal amounts of whole cell lysates were subjected to Western analyses and blotted with anti-BNIP3 and anti-actin antibodies, respectively. Individual protein level was normalized with its respective actin level and calculated as -fold increase by setting nontreatment control as 1 and is shown as numbers in italic type. DECEMBER 5, 2008 • VOLUME 283 • NUMBER 49 dependent and -independent autophagic pathways are activated by extracellular stressor and that the noncanonical Beclin-1-independent autophagy is inhibited in the presence of caspase-3. As MCF-7 cells are devoid of caspase-3 activity, we utilized the pair of MCF-7/caspase-3 and MCF-7/Neo cells to examine whether or not noncanonical autophagy pathway is involved in acute hypoxic stress-induced activation of autophagy. To address this question, we evaluated the DFO-induced autophagosome formation in both MCF-7/caspase-3 and MCF-7/Neo cells. As shown in Fig. 5A, treatment with DFO for 4 h induced the accumulation of autophagosomes, evidenced by the GFP-LC3 punctae in MCF-7/caspase-3 cells (left third panels). In contrast, we rarely observed GFP-LC3 punctae following the DFO-treatment in MCF-7/Neo cells (Fig. 5A, right  six top panels). The quantitation analyses are summarized in Fig. 5B. To assess whether or not MCF-7/Neo cells are capable of forming autophagosomes, we examined the autophagosome formation by treating MCF-7/caspase-3 and MCF-7/Neo cells with CQ (10 M) for 4 h. CQ-treated MCF-7/Neo cells exhib-ited a GFP-LC3 puncta profile comparable with that in MCF-7/caspase-3 cells at 4 h post-treatment when autophagosomes accumulated (Fig. 5A, bottom panels). Thus, a caspase-3-dependent autophagic pathway is likely to be utilized during acute hypoxic stress, albeit not quite conclusive.

Temporal and Dual Roles of PKC␦ in Adaptive Responses
It has been postulated that JNK1-mediated Bcl-2 phosphorylation is essential to relieve the inhibitory effect of Bcl-2 on Beclin-1-dependent autophagy (25). Since the commercial antibody against phospho-Ser-70-Bcl-2 only recognizes human phosphorylated Bcl-2, we examined the Bcl-2 Ser-70 phosphorylation profiles during the course of DFO treatment in human breast cancer MCF-7/caspase-3 and MCF-7/Neo cells. As shown in Fig. 5C, it appears that DFO treatment resulted in a relatively rapid induction of JNK1 and Bcl-2 phosphorylation in MCF-7/caspase-3 cells compared with that in MCF-7/Neo cells at 1 h post DFO treatment. Taken together with Fig. 5, A and C, it is very likely that DFO induces JNK1-mediated phosphorylation of Bcl-2, inhibits interaction with Beclin-1 as a result, and stimulates autophagy.
The cross-talk between autophagy and apoptosis is further exemplified by a recent finding that the prototypic apoptosis inhibitor, Bcl-2, inhibits autophagy by binding to Beclin-1 (16). To address whether or not DFO induces autophagy by dissociating Beclin-1 from the inhibitory Bcl-2, immunoprecipitation by an anti-Beclin-1 antibody was performed in MEF, MEF/ PKC␦ Ϫ/Ϫ , and MEF/JNK1 Ϫ/Ϫ cells treated with DFO for different time periods. It has been established that minimal levels of Bcl-2 are co-immunoprecipitated with Beclin-1 under autophagy-inducing conditions, whereas high levels of Bcl-2 are coimmunoprecipitated with Beclin-1 under autophagy-inhibitory conditions (15,16). As shown in Fig. 5D, DFO treatment resulted in a dissociation of Beclin-1 from Bcl-2 exclusively in MEF (but not in MEF/PKC␦ Ϫ/Ϫ and MEF/JNK1 Ϫ/Ϫ cells) cells, as evaluated by the extent of Bcl-2 remaining as a form complexed with Beclin-1. Taken together with results shown in Fig.  2C, these observations suggest that decreased association of Beclin-1 with Bcl-2 in response to DFO-stress is a PKC␦-dependent event, supporting our contention that acute hypoxic stress induces autophagy via PKC␦ activation.
Next, we sought to test whether Bcl-2 S70 phosphorylation (Fig. 5C) is essential for hypoxic stress-induced dissociation of Beclin-1 from the Bcl-2⅐Beclin-1 complex. To achieve this goal, we engineered and expressed the WT and S70A-mutated forms of Myc-tagged Bcl-2 in MEF, MEF/PKC␦ Ϫ/Ϫ , and MEF/ JNK1 Ϫ/Ϫ cells and assayed their respective phosphorylation status and association with Beclin-1 under control and hypoxia-stressed conditions. Similar to endogenous Bcl-2, a strong phosphorylation of Myc-Bcl-2, but not Myc-Bcl-2-S70A, was observed in only MEF cells in response to DFO treatment (Fig.  5E). Also, the Myc-Bcl-2-S70A was coimmunoprecipitated with endogenous Beclin-1 under hypoxic conditions (Fig. 5E). In addition, Bcl-2 Ser-70 phosphorylation and dissociation of the Bcl-2⅐Beclin-1 complex were not observed in JNK1 Ϫ/Ϫ and PKC␦ Ϫ/Ϫ MEF cells under hypoxic conditions (Fig. 5F). Accordingly, these results support our contention that hypoxia-induced, PKC␦/JNK1-dependent Ser-70 phosphorylation of Bcl-2 is involved in the induction of autophagy by acute hypoxic stress. Last, in order to provide a "proof of concept," we confirmed the role of JNK1 activation in the acute hypoxia-induced autophagic process. To this end, we adapted a new quantitative assay to assess autophagic activity in living cells. The assay also uses GFP-LC3, a well established autophagosomal marker, and follows its turnover by FACS analysis (26). The reduction in GFP-LC3 fluorescence reflects its delivery into the lysosomes (26). Consistent with results of GFP-LC3 puncta formation (Fig. 1) and Western analyses of the steady-state levels of endogenous LC3-I/II (Figs. 1 and 2), it was clearly demonstrated that the fluorescent signal is reduced during both starvation-and hypoxia-inducedautophagicprocessesinatimedependent manner (Fig. 6, A and B). As depicted in Fig. 6C, starvation and hypoxia resulted in an ϳ70 and 50% decrease in GFP-LC3 level at 6 h post-treatment, respectively. Pretreatment with either an autophagic inhibitor 3-MA (an inhibitor for class III PI3K/Vps34 (9)) or lysosomal inhibitor Bafilomycin A1 (Baf A; an inhibitor of vacuolar H ϩ -ATPase (27)) blocked this reduction (Fig. 6C). We then used two different JNK inhibitors, SP 600125 and a cell membrane-permeable JNKi peptide (28), to validate the role of JNK in hypoxia-induced autophagic responses. Fig.  6D demonstrated the ability of both SP 600125 and JNKi to partially reverse the reduction in GFP-LC3 level at both 3 and 6 h post-hypoxic stress, supporting the involvement of JNK activation in acute hypoxiainduced autophagic response.
PKC␦ Promotes Cell Death from Prolonged Hypoxic Stress by DFO-Paradoxically, autophagy not only serves to protect cells but also may contribute to cell damage (1). To examine whether or not the autophagy induced by DFO stress plays a role in cell survival or cell death, Pa-4 cells in which autophagy was blocked using 3-MA were exposed to DFO. The 3-MA-pretreated cells showed more prevalence in cell death at 24 and 48 h after DFO treat- Equal amounts of cell lysates were subjected to immunoprecipitation with an anti-Beclin-1 antibody, followed by Western analyses with anti-Bcl-2 and anti-Beclin-1 antibodies, respectively. Five percent of total cell lysate inputs analyzed by Western analyses served as protein expression controls. DFO-induced dissociation rates (numbers in italic type) of Beclin-1 from Bcl-2 are assessed by the quantitative analyses of immunoprecipitated Bcl-2, as described above. E, phosphorylation of Bcl-2 is required for DFO-induced dissociation of Beclin-1 from Bcl-2. MEF cells were transfected with Myc-tagged Bcl-2 wild type or S70A mutant for 36 h, followed by treatment of 200 M DFO for the indicated time periods. Equal amounts of cell lysates were subjected to immunoprecipitation (IP) with an anti-Myc or anti-Beclin-1 antibody, followed by Western analyses with anti-pS70-Bcl-2, anti-Myc, and anti-Beclin-1 antibodies, respectively. Three percent of total cell lysate inputs analyzed by Western analyses served as protein expression controls. DFO-induced phosphorylation of Bcl-2 and dissociation rates of Beclin-1 from Bcl-2 (numbers in italic type) were assessed by the quantitative analyses of immunoprecipitated Myc-Bcl-2, as described above. F, PKC␦ and JNK1 are essential for DFO-induced phosphorylation of Bcl-2 and dissociation of Beclin-1 from Bcl-2. MEF/PKC␦ Ϫ/Ϫ and MEF/JNK1 Ϫ/Ϫ cells were transfected with Myc-tagged Bcl-2 wild type or S70A mutant for 36 h followed by treatment of 200 M DFO for the indicated time periods. Equal amounts of cell lysates were subjected to immunoprecipitation with specific antibodies and followed by Western analyses as described in E. Three percent of total cell lysate inputs analyzed by Western analyses served as protein expression controls. The relative level of DFO-induced phosphorylated Bcl-2 and dissociation rates of Beclin-1 from Bcl-2 (numbers in italic type) were assessed by the quantitative analyses of immunoprecipitated Myc-Bcl-2, as described above. DECEMBER 5, 2008 • VOLUME 283 • NUMBER 49 ment compared with vehicle-pretreated cells (Fig. 7A). In addition, cells pretreated with rapamycin, which was previously shown to effectively inhibit mTOR and induce autophagy (9), were slightly resistant to DFO at 24 h following treatment. We further analyzed the sensitivity of MEF WT, JNK1 Ϫ/Ϫ , JNK2 Ϫ/Ϫ , Atg5 Ϫ/Ϫ , and PKC␦ Ϫ/Ϫ cells to DFO at 24 and 48 h post-treatment. Atg5-deficient cells showed significantly increased vulnerability to DFO, compared with wild type MEF cells (Fig. 7B, lanes 2 and 3 versus lanes 11 and 12). However, either JNK1-deficient or PKC␦-deficient cells exhibited a partial resistance to DFO-induced cell death at 48 h post-treatment (Fig. 7B, lane 3 versus lanes 6 and 15). Last, we evaluated the JNK activation profile at 24 and 48 h post-treatment with DFO in Pa-4 and Pa-4/PKC␦KD cells. In contrast to those shown in Fig. 4B, the delayed JNK activation for both short and long forms of JNK1/2 was observed at 48 h post-DFO treatment in both Pa-4 and Pa-4/PKC␦KD cells, albeit at different extents (Fig. 7C). In summary, these results indicate that PKC␦ and JNK1 (short form) activation in the early phase of DFO stress is required to activate autophagy, which plays a transient role in protecting against cell death triggered by hypoxic stress (Fig.  7D, solid lines; see "Discussion" for details). In contrast, a sustained activation of both short and long forms of JNK1/2 via PKC␦-dependent and -independent pathways by DFO could lead to cell death (Fig. 7D, dotted lines).

DISCUSSION
We show in this report that PKC␦/JNK1 activation in the early phase ensuing hypoxic stress is required for the induction of autophagic process. Several observations support this conclusion. First, acute 1% O 2 or DFO treatment stimulates LC3-II formation/degradation and the accumulation of GFP-LC3 punctae. Second, the induced autophagic process is attenuated in cells stably transfected with PKC␦KD, in cells of PKC␦ Ϫ/Ϫ , Atg5 Ϫ/Ϫ , and JNK1 Ϫ/Ϫ contexts or in cells treated with a class III PI3K inhibitor, 3-MA. Third, acute DFO treatment induces PKC␦-dependent dissociation of Beclin-1 from Bcl-2, correlating with JNK activation. Taken together, by using cells derived from distinct genetic approaches, dominant-negative PKC␦KD and PKC␦ knock-out as well as Atg5-null and distinct JNK-null cell types, we obtain consistent results leading to the conclusion that the hypoxic stress mediates the early induction of autophagy, which is PKC␦/JNK1-dependent and acts as an early cellular adaptive response, suggesting the possibility that autophagy could play a transient prosurvival role against the hypoxic stress-induced cell death.
Based on the data shown herein, a model is depicted in Fig.  7D. In brief, acute hypoxic stress leads to a rapid PKC␦ and JNK1 (short form) activation to induce autophagy. The role of JNK in autophagy has been implicated in other studies (25,29,30). How PKC␦/JNK1 may contribute to autophagy is not completely known at the present time, although it is tempting to propose that activated JNK1 results in Bcl-2 phosphorylation and subsequent dissociation of Beclin-1 from protein complexes of Bcl-2 (Fig. 5). Alternatively, it is possible that PKC␦/ JNK1 may suppress the mTOR pathway via their kinase activity. A recent report, that a nonphosphorylatable Bcl-2 mutant (T69A/S70A/S87A) abolishes the starvation-induced dissociation of Bcl-2 and Beclin-1 and inhibits autophagy (30,31) favors the former possibility and supports our model. Autophagy could promote survival by removing hypoxia-/DFO-damaged organelles, although the effect is transient. In agreement with our proposed model, we demonstrated that 3-MA down-regulates DFO-induced changes in LC3-II level (Figs. 2C and 4C). Intriguingly, 3-MA also attenuates JNK1 activation in DFOtreated MEF and Pa-4 cells, reminiscent of those observed in FIGURE 6. JNK is involved in hypoxia-induced decay of GFP-LC3 fluorescence intensity via autophagy. A, serum starvation induces the reduction of GFP-LC3 fluorescence intensity. MEF/GFP-LC3 cells were incubated in EBSS medium for indicated time periods and subsequently subjected to FACS analysis. The relative levels of GFP-LC3 fluorescence intensity versus cell counts were shown in a histogram from a representative experiment. B, hypoxia (1% O 2 ) induces the decrease in GFP-LC3 fluorescence intensity. MEF/GFP-LC3 cells were exposed to 1% O 2 for indicated time periods and analyzed as described in A. C, autophagy inhibitors reverse the EBSS-or hypoxia-induced decay of GFP-LC3 fluorescence intensity. MEF/GFP-LC3 cells were pretreated with vehicle, Bafilomycin A1 (Baf A; 100 nM) or 3-MA (10 mM) for 1 h following by treated with EBSS medium or exposed to 1% O 2 in the presence of indicated inhibitor for an additional 3 or 6 h. The relative levels of GFP-LC3 intensity were measured by FACS analyses and normalized to that of control cells. Results from at least three independent experiments are calculated as mean Ϯ S.D. and shown as a percentage of fluorescence intensity, where control level was designated as 100%. D, cell membrane-permeable JNK inhibitory peptide (JNKi) or JNK inhibitor, SP 600125, reduces hypoxia-induced decay of GFP-LC3 fluorescence intensity. MEF/GFP-LC3 cells were pretreated with JNKi or SP 600125 for 1 h prior to the combined treatment of inhibitor and 1% O 2 . Relative levels of fluorescence intensity were detected, normalized, and shown as described in C.
Nevertheless, both PKC␦-and JNK1-null cells exhibited resistance to DFO-elicited cell death (Fig. 7B). The apparent contradictory function for PKC␦ could be explained by "dual roles" of PKC␦ in cell survival as well as death; early induction of PKC␦ may contribute to a protective response by stimulating autophagy, whereas the delayed and sustained activation of PKC␦ by cleavage could lead to enhanced cell death. Indeed, we have detected autophagy, evident by the accumulation of GFP-LC3 punctae and changes in LC3-II level, in response to acute hypoxic stress, only in PKC␦-expressing cells. There is supporting evidence for such a prosurvival role of autophagy in hypoxia; when autophagy is blocked by knock-out of Atg5 (this report) or HIF-1␣ (39) or knockdown of Beclin-1 (40), these cells showed increased sensitivity to the stress by prolonged hypoxia. These findings are consistent with various reports showing that PKC␦-deficient mice exhibit resistance to irradiation and are defective in mitochondria-dependent apoptosis (41,42); PKC␦ phosphorylates and activates caspase-3 (43), phosphorylates and targets the antiapoptotic protein Mcl-1 for degradation (44), and suppresses Akt phosphorylation (17,45). Using various genetically modified knock-out cells and pharmacological inhibitors, as reported herein, our findings advance current understanding of the role of PKC␦ in autophagy, suggesting that PKC␦ conveys dual roles, a transient prosurvival via autophagy induction (this report) and an eventual prodeath signal (reviewed in Ref. 34), to regulate the cell fate decision.
The importance of time-dependent PKC␦ activation in differential cell fate determination is further supported by the report that PKC␦ plays a role in mediating endoplasmic reticulum (ER) stress in ischemic heart (46). Intriguingly, there is an intricate relationship between ER stress and autophagy. For example, several ER stressors, such as hypoxia, thapsigargin, A23187, and tunicamycin, can also activate autophagic process (47). It is widely believed that JNK1 is also activated during ER stress-induced autophagy through the ER membrane protein IRE1 (29). Together, these reports suggest the possibility that the hypoxic stress-induced early autophagic responses could be mediated through ER stress-dependent signaling pathways. However, we failed to detect the transcriptional induction of Grp78/BiP, a known marker for ER stress (48), during the early phase of DFO treatment (supplemental Fig. S3). Thus, it is likely that DFO utilizes PKC␦/JNK1 activation, which is distinct from the IRE1/JNK1 pathway involved in ER stress, to stimulate autophagy. In addition, a very recent report suggests that ER stress-activated autophagy involves PKC, but not PKC␦, phosphorylation (49). Together, our studies reported herein confer significant functional and mechanistic insights into the PKC␦/ JNK1-dependent autophagic event upon acute hypoxic stress.
Hypoxic stress is thought to be encountered during various pathological situations, including cancer, myocardial infarction, and stroke (50). Cells deprived of oxygen will initially employ adaptive and survival strategies, but if hypoxia is sustained, cell death will eventually ensue. Hypoxia-induced autophagy has been reported under a sustained hypoxic condition through HIF-1␣, BNIP3, Atg5, and Beclin-1 (see Refs. 1, 11, and 12 and references therein). Our findings support the notion that PKC␦/JNK1 activation and autophagy induction occur prior to the maximal induction of BNIP3 in response to hypoxic stress. This observation conflicts with recent findings by Zhang et al. (39) who concluded that autophagy in response to prolonged hypoxia occurs through the induction of BNIP3. It is possible that the disparity may stem from differences in the experimental conditions and/or cell types analyzed. Whereas Zhang et al. (39) studied the autophagic responses in WT and HIF-1␣ Ϫ/Ϫ MEF cells under 1% O 2 for Ն24 h, we analyzed responses in Pa-4 and MEF cells after a short time period (Յ6 h) using DFO or 1% O 2 . It appears that BNIP3 expression occurs relatively late in hypoxia (51) or following a prolonged DFO treatment (52). In addition, Azad et al. (51) demonstrated that knockdown of BNIP3 did not affect the formation of GFP-LC3 punctae, an early autophagy event, supporting our contention that BNIP3 is not involved in the early responses to hypoxic stress. Since we have shown that PKC␦ is required for DFOinduced changes of LC3-II level in Pa-4 and MEF cells, we believe that PKC␦ is involved in the early response to hypoxic stress. Moreover, AMPK is a major regulator of energy homeostasis (53), and AMPK is known to be induced by sustained hypoxic stress (54) and is implicated in hypoxia-induced autophagy (55). The fact that 1% O 2 and DFO were able to induce AMPK activation in JNK1 Ϫ/Ϫ and PKC␦ Ϫ/Ϫ cells at the later stage of hypoxic stress (supplemental Fig. S4) raised the possibility that PKC␦/JNK1 pathway activates autophagy machinery (in the early phase), independently of AMPK activation. Thus, it is possible that a PKC␦/JNK1-dependent, but BNIP3-and AMPK-independent, pathway is required for autophagy in the early response to hypoxic stress, whereas a BNIP3-dependent pathway is required for autophagy in response to prolonged hypoxia, possibly related to the accumulation of reactive oxygen species during chronic exposure to either hypoxia or DFO (39).
Chronic hypoxia is a typical microenvironment occurring during tumor development, since rapid proliferation causes the tumor to outgrow its available oxygen supply and triggers a variety of adverse effects arising from metabolic stress. The precise mechanism of hypoxia-induced cell death remains unclear, since apoptosis, necrosis, and autophagy have all been reported to occur in response to hypoxic stress (56). Hypoxia-induced autophagy has been proposed to mitigate genome damage caused by metabolic stress (57). Obviously, hypoxia/metabolic stress-induced autophagy might also play dual roles of cell survival versus cell death; early induction of autophagy, as we demonstrated in this study, may contribute to a protective response, whereas prolonged autophagy could lead to cell death. In line with this scenario, we have shown that autophagy occurs following acute DFO treatment, which is likely to be an initial survival strategy (Fig. 6, A and B). For example, cells are shown to lead to decreased survival when autophagy is blocked by 3-MA, whereas cells temporally exhibit increased survival when autophagy is stimulated by rapamycin. A complete picture of the signaling pathways and associated functions of PKC␦ and/or JNK1 in hypoxia-stressed cells has yet to emerge. The data presented herein expand our understanding of PKC␦/JNK1 biology as well as the hitherto unknown role of PKC␦/JNK1 in regulating hypoxia-induced autophagy.
In summary, our data reported herein collectively demonstrated a significant association between PKC␦/JNK1 activation and autophagic response, which is independent of ER stress and BNIP3, during the early phase of hypoxia and suggest that PKC␦-JNK1 axis contributes to the hypoxia-induced autophagy by releasing Beclin-1 from its inhibitor Bcl-2, presumably via phosphorylation. We previously reported that PKC␦ mediates late, proapoptotic cellular response to DFO/hypoxia treatment (17,58). Our current finding that inhibition of PKC␦-dependent autophagy during the early phase of hypoxia favors cell death may suggest, at least in part, the temporal and dual roles played by PKC␦ to adapt to cell stress. The proposed signaling pathway will help us to understand the regulation of autophagy, especially those activated independently of nutrient starvations and/or ER stress.