Differential Effects of Endoplasmic Reticulum Stress-induced Autophagy on Cell Survival*

Autophagy is a cellular response to adverse environment and stress, but its significance in cell survival is not always clear. Here we show that autophagy could be induced in the mammalian cells by chemicals, such as A23187, tunicamycin, thapsigargin, and brefeldin A, that cause endoplasmic reticulum stress. Endoplasmic reticulum stress-induced autophagy is important for clearing polyubiquitinated protein aggregates and for reducing cellular vacuolization in HCT116 colon cancer cells and DU145 prostate cancer cells, thus mitigating endoplasmic reticulum stress and protecting against cell death. In contrast, autophagy induced by the same chemicals does not confer protection in a normal human colon cell line and in the non-transformed murine embryonic fibroblasts but rather contributes to cell death. Thus the impact of autophagy on cell survival during endoplasmic reticulum stress is likely contingent on the status of cells, which could be explored for tumor-specific therapy.

teins induce ER stress. ER stress is frequently observed in pathological conditions where protein misfolding is caused by genetic mutations either in the molecule to be processed or in the machinery processing the folding (3,4).
The major protective and compensatory mechanism during ER stress is the unfolded protein response (UPR) (1,5), which leads to translational attenuation and selective up-regulation of a number of bZip transcription factors (1,5). UPR serves multiple functions, including the assistance of protein folding via the up-regulated ER protein chaperones and the enhanced degradation of misfolded proteins via the up-regulation of molecules involved in the ER-associated degradation pathway (1,5). However, if the stress is excessive, the compensatory mechanisms may not be able to fully sustain ER function, and ER decompensation could lead to cell death (2,6). It is not clear whether there are other mechanisms that can regulate ER stress.
Macroautophagy (referred as autophagy hereafter) is mainly responsible for the degradation of long-lived proteins and subcellular organelles (7)(8)(9). Autophagy is frequently activated in response to adverse environment or stress (10 -13) and has been shown to be involved in many physiological and pathological processes (8,14). However, whether autophagy serves a protective or detrimental role is controversial (15)(16)(17). Although some studies indicate that autophagy is responsible for the non-apoptotic cell death (13,15,(17)(18)(19), others indicate that autophagy is protective against cell death (12, 20 -24). The condition under which autophagy may be pro-survival or prodeath is not clear.
In the current study, we found that autophagy could be activated by the classical ER stress inducers in mammalian cells. However, autophagy alleviates ER stress and reduces cell death in cancer cells but not in non-transformed cells. This unique feature may be explored for certain types of cancer therapy in which ER stress constitutes a major cause of cell death.
Cell Culture and Microscopy-HCT116 Bax-positive and Bax-negative cell lines (27) were maintained in McCoy's 5A with the routine supplements. DU145 cell lines were maintained in Dulbecco's modified Eagle's medium with routine supplements. Wild type and Atg5-deficient MEFs were immortalized through SV40 large T overexpression and cultured in Dulbecco's modified Eagle's medium with standard supplements (11). The non-immortalized human colon cell line CCD-18Co was purchased from ATCC (CRL-1459 TM ) and cultured in Dulbecco's modified Eagle's medium with standard supplements. All cell lines were maintained in a 37°C incubator with 5% CO 2 . Cells (2 ϫ 10 5 /well) were seeded into 12-well plates. After 24 h, cells were treated as indicated in the figure legends.
For electron microscopy, cells were fixed with 2% paraformaldehyde and 2% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4) followed by 1% OsO 4 . After dehydration, thin sections were stained with uranyl acetate and lead citrate for observation under a JEM 1016CX electron microscope. To examine distribution of GFP-LC3B, cells were observed under a fluorescence microscope and digital images were acquired for analysis (SPOT, Diagnostic Instruments, Inc.). To examine and quantify cellular vacuolization, digital phasecontrast images were recorded.
Analysis of Cell Death-Cell death was determined using propidium iodide staining (1 g/ml) and quantified by digital microscopy. Apoptotic cells with condensed or fragmented nuclei were quantified after Hoechst 33342 (5 g/ml) staining. Caspase activities were measured using 30 g of proteins and 20 M fluorescent substrates (Ac-DEVD-AFC, and Ac-LEVD-AFC for caspase-3 and -4, respectively). The fluorescence signals were detected by a fluorometer (Tecan GENios) at 400/510 nm (excitation/emission), as described previously (28).
Immunoblot Assay-This is essentially performed as described previously (28). Cells were washed in phosphate-buffered saline and lysed in radioimmune precipitation buffer. Forty micrograms of protein was separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes. The membranes were stained with the indicated primary and secondary antibodies and developed with SuperSignal West Pico chemiluminescent substrate (Pierce).

Induction of Macroautophagy by ER Stress Inducers in Mam-
malian Cells-A characteristic feature of autophagy activation is the conversion of the Atg8/LC3B from the unconjugated form (LC3B-I) to the phosphatidylethanolamine-conjugated form (LC3B-II) (25,29). The mammalian Atg8/LC3B is first cleaved by Atg4 to expose the conserved Gly 120 and then conjugated to phosphatidylethanolamine via a ubiquitination-like reaction mediated by Atg7, a ubiquitin-activating enzyme (E1)like protein, and Atg3, a ubiquitin carrier protein (E2)-like protein. This process is also affected by another conjugation system, Atg5-Atg12-Atg16, and Atg6/Beclin 1 (25,29). The non-cleaved unconjugated form of Atg8/LC3B (LC3B-I) is in the cytosol, whereas the cleaved, conjugated form (LC3B-II) targets to the autophagosomal membrane (25).
The ER stress inducer A23187 could elicit a strong formation of the lipidated LC3B-II in a dose-dependent manner in a colon cancer cell line, HCT116 (Fig. 1A), independent of the presence of the pro-apoptotic molecule, Bax, which is critical to apopto- sis induction in this cell line (see below). Other ER stress inducers, including tunicamycin (TM), thapsigargin (TG), and brefeldin A (BA), could also induce such an accumulation ( Fig.  1B and data not shown). This conversion could be suppressed by the autophagy-inhibiting agent, 3-methyadenine (3-MA) (12,13,30) (Fig. 1B), or by the specific siRNA against Atg6/ Beclin 1 or Atg8/LC3B (data not shown), suggesting the speci-ficity of the phenomenon caused by the ER stress inducers. Consistently, cells expressing GFP-LC3B showed a transition of the GFP-LC3B signals from the diffusive cytoplasm pattern to the punctated membrane pattern following the application of the ER stress inducers (Fig. 1, C and D), suggesting the formation of autophagic vacuoles. Similarly, GFP-LC3B puncta could be suppressed by 3-MA or a specific siRNA against Atg6/Beclin 1 (Fig. 1, D and E). Electron microscopic study indicated that a significant amount of autophagic vacuoles was present in HCT116 as well as in DU145, a prostate cancer cell line, following the treatment of A23187, TM, TG, or BA (Fig. 2). Both double-membrane and multimembrane structures with different intracellular contents could be observed.
Disturbing ER homeostasis or function causes the accumulation of misfolded proteins, which in turn induce ER stress (1)(2)(3)(4)(5). Limiting the protein influx into the ER network could reduce the stress level and thus autophagy induction. Indeed, treatment of cells with the general transcriptional inhibitor, actinomycin D, or the translational inhibitor, cycloheximide, suppressed the formation of LC3-II (Fig. 3A) and the translocation of GFP-LC3B to autophagic vacuoles (Fig. 3B). Together, these data strongly indicate that a diverse array of chemicals could activate macroautophagy via the induction of ER stress.
Autophagy Protects Cancer Cells from ER Stress and Cell Death-We then examined the significance of autophagy in the regulation of ER stress. In HCT116 and DU145 cells, ER stress inducers could cause cellular vacuolization to different degrees (Fig. 4, A-C). These vacuoles represented dilated ER lumens under stress based on electron microscopic examination (data not shown). Notably, suppression of autophagy with either 3-MA or a specific siRNA against Atg6/Beclin 1 (Fig. 4, B and C) or Atg8/LC3B (data not shown) increased the percentage of cells with cytoplasmic vacuolization, suggesting an enhanced level of ER stress.
The accumulated misfolded proteins in the ER lumen are normally degraded through ER-associated degradation     33342 staining (B). C, overall cell death was determined by propidium iodide staining. D and E, caspase-3 (D) and caspase-4 (E) activities were determined using Ac-DEVD-AFC or Ac-LEVD-AFC as the substrate, respectively, and expressed as the -fold over the non-treated control. All values were shown as mean Ϯ S.D. pathway via the proteasomes following ubiquitination (31). Thus an accumulation of polyubiquitinated proteins could indicate the level of misfolded proteins and therefore the level of ER stress. Indeed, both A23187 and TM could elevate the level of polyubiquitinated proteins in HCT116 cells and DU145 cells (Fig. 4, D and E). Ubiquitinated protein aggregates could be also readily detected in treated cells (Fig. 4, F and G), which suggested that the accumulation of these proteins exceeded the capacity of proteasomes for degradation (32). Importantly, suppressing autophagy by 3-MA or a specific siRNA against Atg6/Beclin 1 (Fig. 4, D, F, and G) led to further increases in polyubiquitinated protein aggregates. Taken together, it seems that autophagy can function to reduce ER stress based on its effects on cellular vacuolization and polyubiquitinated protein accumulation.
As the level of ER stress could correlate with the extent of cell death, we examined whether autophagy induced in these cancer cells could have any effects on ER stress-induced cell death. Apoptosis in HCT116 cells is heavily dependent on the presence of Bax (27). The syngeneic Bax-deficient cells derived from the parental Bax-positive cells require more potent stimuli to go into apoptosis through the alternate Bak-mediated mechanism (33). In the regular dose ranges where ER stress and autophagy could be activated, Bax-positive HCT116 cells were much more sensitive to A23187-, TG-, TM-, or BA-induced apoptosis than the Bax-defi-cient HCT116 cells (Fig. 5A). Suppression of autophagy with either 3-MA (data not shown) or a specific siRNA against Atg6/Beclin 1 or Atg8/LC3B (Fig. 5, B and C) led to increased apoptosis in the Baxpositive HCT116 cells. The activities of both effector caspases and the ER stress-specific caspase-4 were enhanced (Fig. 5, D and E). Interestingly, under this condition, a significant amount of Baxdeficient HCT116 cells also became apoptotic with enhanced caspase activation (Fig. 5), perhaps reflecting an elevated level of death stimulation when autophagy was suppressed. Similar enhancement in cell death could be also observed in the Bax-deficient DU145 cells (data not shown). These observations suggest that autophagy can protect against cell death in cancer cells, likely by curtailing the level of ER stress and therefore the potency of death stimulation.
Autophagy Promotes Cell Death in Non-transformed Cells Treated with ER Stress Inducers-To determine the significance of autophagy in ER stress-induced cell death in different types of cells, we took the advantage of the availability of immortalized MEFs that are Atg5-deficient (11). All four ER stress inducers could induce GFP-LC3B punctation and the lipidated LC3B-II formation in the MEFs in an Atg5-dependent manner (Fig. 6, A and B). Electron microscopic examination of these cells indicated that there was an Atg5-dependent accumulation of intracellular autophagic vacuoles (Fig.  6C). However, treatment of ER stress inducers did not induce cellular vacuolization in the MEFs (Fig. 7A), no matter whether Atg5 is expressed or not. Furthermore, in contrast to what was observed in HCT116 and DU145 cells, deficiency of autophagy in MEFs led to reduced apoptosis. Thus Atg5-deficient MEFs were much less susceptible to apoptosis and caspase activation induced by A23187, TG, TM, or BA than the wild type MEFs (Fig. 7, B-D). Consistently, reintroduction of Atg5 into the Atg5-deficient MEFs restored their sensitivity to the cytotoxic effects of ER stress (Fig. 7E).
To support the hypothesis that this pro-death character of autophagy may be related to the non-cancerous status of the MEFs, we examined a non-immortalized normal human colon cell line (CCD-18Co), which would be more comparable with HCT116 colon cancer cells. Treatment of this cell line with A23187 or TM induced autophagy that could be suppressed by 3-MA (Fig. 8A). However, 3-MA co-treatment did not lead to increased cell death but rather resulted in reduced cell death (Fig. 8, B-D), consistent with the obser-  FEBRUARY 16, 2007 • VOLUME 282 • NUMBER 7 vations in MEFs. Interestingly, these treatments did not induce noticeable cellular vacuolization in this cell line either. Taken together, these observations in the non-transformed cells are in stark contrast to those in HCT116 and DU145 cells, suggesting that the role of ER stress-induced autophagy in cell survival is contingent on the status of the cells and can be different in cancer cells and in non-transformed cells under a given set of stimuli.

DISCUSSION
Induction of Macroautophagy by ER Stress-As the major organelle processing post-translational modifications and supporting correct protein folding, ER also possesses the mechanisms to retro-transport proteins that fail to be modified and/or folded properly (31). These proteins are then ubiquitinated and degraded by proteasomes. This ERassociated degradation pathway is important for maintaining ER homeostasis. The accumulation of the unfolded/misfolded proteins, such as that due to the use of A23817, TM, TG, or BA, can lead to ER stress. The UPR is a classical mechanism that cells mount to relieve ER stress, which aims to reduce overall cellular protein synthesis, assists protein folding, and promotes ER-associated degradation (1,5). Here we show that autophagy is activated in response to ER stress in the mammalian cells and could be another mechanism regulating ER stress and the outcome. Notably, ER stress could also induce autophagy in the yeast (34), indicating that this response is evolutionarily conserved.
How ER stress leads to the activation of autophagy is not quite understood. Transcriptional up-regulation of certain Atg genes, such as Atg12 (24) and Atg8/LC3B (data not shown), have been observed and could be important. Although the majority of cellular protein synthesis is shut down during ER stress, a selected group of proteins, such as the bZIP transcription factors ATF4 and CHOP, are activated downstream of eIF2␣ kinase, which is an important component of the UPR (1,5). These transcription factors could be involved in the autophagy activation. Indeed, eIF2␣ has been found to be important for autophagy induced by the pathogenic polyglutamine repeats (24), viral infection, and starvation (35). Other UPR response pathways, such as those mediated by Ire1 and ATF6, could also be involved. Although other activation mechanisms may also be required for a full activation of autophagy, our data strongly indicate that signals from ER stress are critical for the triggering of autophagy, which would require further characterization in the future. staining, respectively. Ctrl, control. D, caspase-3 activities were measured using Ac-DEVD-AFC as the substrate. E, atg5-deficient MEFs were transfected with murine Atg5 or the vector control for 24 h and then treated with A23187 or TG for another 24 h before being analyzed for apoptosis as described above. The inset shows an immunoblot assay on Atg5 and ␤-actin expression in these cells. Atg5 was detected as the complex with Atg12. All values were shown as mean Ϯ S.D.
The Significance of Autophagy in ER Stress and ER Stressinduced Cell Death-In the cancer cells, autophagy helps to alleviate ER stress, and subsequently, the cell death. How autophagy mitigates ER stress is not completely clear at this moment. Our observations indicate that autophagy is important for the clearance of ubiquitinated unfolded/misfolded proteins and therefore reduces ER stress induced by these molecules. This notion is also supported by the finding made in mice with conditional deletion of Atg5 or Atg7 in the central nervous system, which leads to the accumulation of polyubiquitinated proteins in the neurons (22,23). In addition, it has been found that misfolded proteins, such as the mutant ␣1-antitrypsin Z protein and the pathogenic polyglutamine repeats, could all induce autophagy in addition to ER stress, independent of their cellular locations (24, 36 -38). Although promotion of autophagy enhances the clearance of the mutant molecules and reduces the toxicity, inhibition of autophagy results in the opposite (21,24,39,40). Thus autophagy can be protective against ER stress in a number of circumstances including cancer cells.
Paradoxically, although disturbing ER homeostasis and/or functions by the chemicals could also elicit autophagy in the primary colon cells, suppression of autophagy does not enhance cell death but instead reduces cell death. In addition, suppression of autophagy induced by the same chemicals in the immortalized but non-transformed MEFs by deletion of Atg5 also reduces cell death. These data suggest that autophagy can contribute to ER stress-induced cell death in different scenarios, which may be dependent on cellular status under a given stimulation. However, the exact mechanisms affecting such a different outcome are yet to be determined. In the context of the current experiment, we speculate that the differential role of autophagy in cell survival in the cancer cells and non-transformed cells may be related to how well the ER stress is compensated. Both MEFs and CCD-18Co might be better compensated in response to the stimulation. One hint for this assumption is that treatment of ER stress inducers did not cause cellular vacuolization, and inhibiting autophagy did not seem to promote vacuolization in these cells (Figs. 7A and 8D). Although autophagy could still be involved in clearing misfolded proteins, its degradation capability might turn out to be "too extensive" for the better compensated cells, in which autophagy may lead to excessive consumption of bystander normal cellular constituents and therefore contribute negatively to cell survival. Future work will be directed to seek experimental evidence to test this hypothesis. However, our current work has indicated the importance of understanding the underlying mechanism and pointed out a potential benefit of such an understanding that the differential effect of autophagy in cancer cells and non-transformed cells may be explored for tumor-specific therapy.