The ER Stress Inducer l-Azetidine-2-Carboxylic Acid Elevates the Levels of Phospho-eIF2α and of LC3-II in a Ca2+-Dependent Manner

Accumulation of misfolded proteins in the endoplasmic reticulum (ER) activates the unfolded protein response (UPR) to reduce protein load and restore homeostasis, including via induction of autophagy. We used the proline analogue l-azetidine-2-carboxylic acid (AZC) to induce ER stress, and assessed its effect on autophagy and Ca2+ homeostasis. Treatment with 5 mM AZC did not induce poly adenosine diphosphate ribose polymerase (PARP) cleavage while levels of binding immunoglobulin protein (BiP) and phosphorylated eukaryotic translation initiation factor 2α (eIF2α) increased and those of activating transcription factor 6 (ATF6) decreased, indicating activation of the protein kinase RNA-like ER kinase (PERK) and the ATF6 arms of the UPR but not of apoptosis. AZC treatment in combination with bafilomycin A1 (Baf A1) led to elevated levels of the lipidated form of the autophagy marker microtubule-associated protein light chain 3 (LC3), pointing to activation of autophagy. Using the specific PERK inhibitor AMG PERK 44, we could deduce that activation of the PERK branch is required for the AZC-induced lipidation of LC3. Moreover, both the levels of phospho-eIF2α and of lipidated LC3 were strongly reduced when cells were co-treated with the intracellular Ca2+ chelator 1,2-bis(O-aminophenoxy)ethane-N,N,N′,N′-tetraaceticacid tetra(acetoxy-methyl) ester (BAPTA-AM) but not when co-treated with the Na+/K+ ATPase inhibitor ouabain, suggesting an essential role of Ca2+ in AZC-induced activation of the PERK arm of the UPR and LC3 lipidation. Finally, AZC did not trigger Ca2+ release from the ER though appeared to decrease the cytosolic Ca2+ rise induced by thapsigargin while also decreasing the time constant for Ca2+ clearance. The ER Ca2+ store content and mitochondrial Ca2+ uptake however remained unaffected.

and phospho-eIF2α levels, but does not affect ERp57 or ERp72 expression. Cells were treated for 6 h with DMSO or 1 µM staurosporine (STS) as control, or for 0-9 h with 5 mM AZC, and during the last 4 h 100 nM Baf A1 was added. A representative blot for six independent experiments each performed in duplicate showing protein levels of ER stress markers ATF6, BiP, ERp72, ERp57, total eIF2α, and phospho-eIF2α. Vinculin was used as loading control. (C) Quantification of ER stress markers in B shown as mean ± SEM relative to 0 h AZC. * p < 0.05, ** p < 0.01, *** p < 0.001. (D) Representative RT-PCR gel of unspliced and spliced XBP1 mRNA levels of three independent experiments each performed in duplicate.

Cell Culture
HeLa cells were cultured at 37 • C and 5% CO 2 in Dulbecco's Modified Eagle Medium supplemented with 10% heat-inactivated fetal bovine serum (FBS), GlutaMAX (Gibco/Invitrogen, Merelbeke, Belgium; # 35050) and penicillin and streptomycin (Gibco/Invitrogen; # 15070-063), as described before [19][20][21]. Cells were washed with phosphate-buffered saline (PBS) and supplied with fresh medium two hours before the start of each experiment. The cell line has been authenticated using autosomal Short Tandem Repeat profiling performed by the University of Arizona Genetics Core and fully matched the DNA fingerprint present in the reference database.

Cell Death Assays
Cell death was assessed by PARP cleavage and by propidium iodide (PI) staining. PARP cleavage was assessed by Western blotting as described in Section 2.3. PI staining was performed essentially as previously described [43]. Briefly, cells were plated in 96-well plates and stained with 2.5 µg/mL PI (Thermo Fisher Scientific; # P3566). Cells were treated with 0-25 mM AZC, and the PI fluorescence was monitored for a total of 72 h in an IncuCyte Zoom (Essen Bioscience, Welwyn Garden City, UK) with a 4 h interval between scans. Individual experiments were always performed in triplicate.

Ca 2+ Measurements at the Population Level
Cells were plated in 96-well plates and treated as indicated. During the last hour of treatment the cells were loaded for 30 min with 1.8 µM Fura-2 AM followed by 30 min de-esterification in modified Krebs solution (150 mM NaCl, 5.9 mM KCl, 1.2 mM MgCl 2 , 11.6 mM Hepes (pH 7.3), 11.5 mM glucose, 1.5 mM CaCl 2 ). Fluorescence was measured on a FlexStation 3 microplate reader (Molecular Devices, Sunnyvale, CA, USA) by alternate excitation at 340 nm and 380 nm and recording emission at 510 nm. Compounds were added as indicated. Individual experiments were always performed in triplicate.

Single-Cell Ca 2+ Measurements
ER and mitochondrial Ca 2+ levels were measured with the genetically encoded Ca 2+ indicators G-CEPIA1er and R-GECO1mt [44] kindly provided by Dr. M. Iino (The University of Tokyo, Tokyo, Japan). Cells were transfected with 300 ng G-CEPIA1er and 600 ng R-GECO1mt using X-treme Gene HP DNA (Roche; # 06366546001) according to the manufacturer's protocol. After 48 h single-cell measurements were performed on a Zeiss Axio Observer Z1 Inverted Microscope equipped with a 20× air objective and a high-speed digital camera (Axiocam Hsm, Zeiss, Jena, Germany). Extracellular Ca 2+ was chelated with 3 mM ethylene glycol tetraacetic acid (EGTA) and one minute later the indicated compound was added. Changes in G-CEPIA1er fluorescence were followed after excitation at 480 nm and measurement of emission at 520 nm. Changes in R-GECO1mt fluorescence were followed after excitation at 377 nm and measurement of emission at 466 nm. The traces were normalized to baseline fluorescence (F/F 0 ) where the baseline was calculated as the average of the first six time points. At least 99 cells per condition were measured in the course of six individual experiments.

Statistics
Results are presented as mean ± standard error of the mean (SEM). Significance was tested using a one-way analysis of variance with Tukey post-hoc test. Results were considered significant when p < 0.05.

AZC Upregulates the Levels of BiP and Phospho-eIF2α while Decreasing the Level of Full-Length ATF6
To analyze the induction of ER stress, HeLa cells were treated with 5 mM AZC for up to 9 h. The cell lysates were analyzed by Western blotting for the ER stress markers ATF6, binding immunoglobulin protein (BiP), ER protein (ERp) 72, ERp57, and total and phospho-eIF2α ( Figure 1B,C). AZC treatment led to a 2.5-fold decrease in full-length ATF6 (already significantly decreased after 3 h), a 1.5-fold increase of BiP protein levels (already significantly increased after 6 h), and a 2-fold increase in the levels of phospho-eIF2α (already significantly increased after 3 h). The total level of eIF2α only decreased after treatment with AZC for 9 h, while ERp57 or ERp72 protein levels did not change in response to AZC treatment. To evaluate the activation of the IRE1α arm of the UPR, we performed an X-box protein (XBP) 1 splicing assay. AZC treatment did not significantly induce XBP1 splicing ( Figure 1D). Staurosporine (STS, 1 µM) was used as a positive control for activation of ER stress [45] and cell death [46][47][48]. STS treatment for 6 h resulted in a 10-fold reduction in the level of full-length ATF6, a two-fold downregulation of ERp57 protein levels, an approximately nine-fold increase in the levels of phospho-eIF2α and a substantial production of spliced XBP1, but did not modify BiP levels.
Taken together, these results indicate that a 6 h treatment with AZC activates both the PERK and the ATF6 arm of the UPR, resulting in upregulated BiP protein expression and increased phosphorylation of eIF2α.

AZC Does Not Induce Cell Death within 6 h of Treatment
Because it was important to study ER stress and subsequent autophagy in conditions where cell death was not induced, we assessed the levels of cleaved poly adenosine diphosphate ribose polymerase (PARP) upon AZC treatment during the same time period. While STS led to clear PARP cleavage, we observed virtually no cleavage upon AZC treatment (Figure 2A). The quantification of these results is shown in Figure 2B. Additionally, we evaluated the cellular toxicity of various AZC concentrations as a function of time by monitoring the increase of PI-positive cells. These results indicated a dose-dependent increase of PI-positive cells with time starting only after at least 12 h of treatment with AZC, even at high AZC concentrations (10-25 mM). Moreover, even after 72 h, 5 mM AZC induced less than 10% cell death above the vehicle control condition.
From both the PARP-cleavage analysis and the PI staining, we can conclude that treatment with 5 mM AZC for 6 h did not lead to any significant cell death.

AZC Increases the Levels of Lipidated Autophagy Marker LC3
To avoid interference of ER stress-induced cell death, we wanted to use conditions that allowed induction of the UPR and of autophagy in a time frame as short as possible. Based on the results presented in Figure 1 (activation of the UPR) and in Figure 2 (absence of cell death), we decided to select a 6 h incubation time with 5 mM AZC to assess the levels of the autophagy marker lipidated microtubule-associated protein light chain 3 (LC3-II). In order to evaluate autophagy induction upon

AZC Increases the Levels of Lipidated Autophagy Marker LC3
To avoid interference of ER stress-induced cell death, we wanted to use conditions that allowed induction of the UPR and of autophagy in a time frame as short as possible. Based on the results presented in Figure 1 (activation of the UPR) and in Figure 2 (absence of cell death), we decided to select a 6 h incubation time with 5 mM AZC to assess the levels of the autophagy marker lipidated microtubule-associated protein light chain 3 (LC3-II). In order to evaluate autophagy induction upon AZC treatment, autolysosomal degradation was inhibited by addition of 100 nM bafilomycin A1 (Baf A1) during the last 4 h of AZC treatment. In the absence of Baf A1, AZC increased the levels of LC3-II 3-fold as assessed by Western blotting ( Figure 3A). Co-treatment with Baf A1 increased the levels of LC3-II about 8-and 13-fold in control and AZC-treated cells, respectively. Figure 3B shows the quantification from twelve independent experiments.
The larger increase of LC3-II levels in AZC-treated cells in the presence of Baf A1 as compared to the control conditions suggests that the upregulation of LC3-II levels does not depend on a block of autophagosome-lysosome fusion and/or of lysosomal degradative activity but is related to the induction of autophagy.  shown as mean ± SEM relative to untreated control. ** p < 0.01, *** p < 0.001.

AZC Upregulates the Levels of LC3-II Subsequently to Activation of the PERK Pathway
To investigate the relation between the UPR and the increased LC3-II levels after application of AZC, we used the PERK-selective inhibitor AMG PERK 44 [49]. AMG PERK 44 (1-5 µ M) did neither affect ATF6 cleavage nor BiP induction ( Figure 4A,B). Not only the AZC-induced increase in phospho-eIF2α was abolished, as was anticipated, but also the AZC-induced increase in LC3-II in spite of the fact that the ATF6 pathway remained active.
These results indicate that activation of the PERK pathway is a prerequisite for the occurrence of the AZC-induced increase in LC3-II levels, and that the ATF6 branch of the UPR is not sufficient for obtaining this effect.

AZC Upregulates the Levels of LC3-II Subsequently to Activation of the PERK Pathway
To investigate the relation between the UPR and the increased LC3-II levels after application of AZC, we used the PERK-selective inhibitor AMG PERK 44 [49]. AMG PERK 44 (1-5 µM) did neither affect ATF6 cleavage nor BiP induction ( Figure 4A,B). Not only the AZC-induced increase in phospho-eIF2α was abolished, as was anticipated, but also the AZC-induced increase in LC3-II in spite of the fact that the ATF6 pathway remained active.
These results indicate that activation of the PERK pathway is a prerequisite for the occurrence of the AZC-induced increase in LC3-II levels, and that the ATF6 branch of the UPR is not sufficient for obtaining this effect. showing protein levels of ER stress markers ATF6, BiP, total eIF2α, and phospho-eIF2α and of autophagy marker LC3-II. Vinculin was used as loading control. (B) Quantification of ER stress and autophagy markers in (A) as well as of independent experiments performed in absence or presence of AMG PERK 44 (1-5 μM), shown as mean ± SEM relative to untreated control. * p < 0.05, ** p < 0.01, *** p < 0.001.

AZC Upregulates the Levels of LC3-II in a Ca 2+ -Dependent Manner
To investigate the role of Ca 2+ in the elevation of LC3-II levels triggered by AZC, we co-treated the cells with increasing concentrations of the intracellular Ca 2+ chelator BAPTA-AM. This led to a shown as mean ± SEM relative to untreated control. * p < 0.05, ** p < 0.01, *** p < 0.001.

AZC Upregulates the Levels of LC3-II in a Ca 2+ -Dependent Manner
To investigate the role of Ca 2+ in the elevation of LC3-II levels triggered by AZC, we co-treated the cells with increasing concentrations of the intracellular Ca 2+ chelator BAPTA-AM. This led to Cells 2018, 7, 239 9 of 19 a dose-dependent reduction of the AZC-induced increase in LC3-II ( Figure 5A). Figure 5B shows the quantification from six individual experiments.
These results strongly suggest that the AZC-induced increase in LC3-II requires intracellular Ca 2+ .
Cells 2018, 7, x 10 of 20 dose-dependent reduction of the AZC-induced increase in LC3-II ( Figure 5A). Figure 5B shows the quantification from six individual experiments. These results strongly suggest that the AZC-induced increase in LC3-II requires intracellular Ca 2+ .

AZC-Induced Elevation of Phospho-eIF2α Levels Is Ca 2+ Dependent
The link between ER stress induction and autophagy involves several steps. In order to identify which steps are sensitive to Ca 2+ , we assessed the effect of BAPTA-AM on the AZC-induced upregulation of BiP expression levels, and on the levels of phospho-eIF2α. While there was no effect of BAPTA-AM treatment on BiP levels, we observed a significant reduction of AZC-induced elevation of phospho-eIF2α upon co-treatment with BAPTA-AM ( Figure 6A,B). Additionally, we verified that BAPTA-AM treatment did not result in PARP cleavage ( Figure 6A).

AZC-Induced Elevation of Phospho-eIF2α Levels Is Ca 2+ Dependent
The link between ER stress induction and autophagy involves several steps. In order to identify which steps are sensitive to Ca 2+ , we assessed the effect of BAPTA-AM on the AZC-induced upregulation of BiP expression levels, and on the levels of phospho-eIF2α. While there was no effect of BAPTA-AM treatment on BiP levels, we observed a significant reduction of AZC-induced elevation of phospho-eIF2α upon co-treatment with BAPTA-AM ( Figure 6A,B). Additionally, we verified that BAPTA-AM treatment did not result in PARP cleavage ( Figure 6A). These results indicate that levels of phospho-eIF2α induced by AZC is dependent on Ca 2+ , and may thus constitute an upstream event to the sensitivity of the subsequent autophagic process to Ca 2+ .

Effects of BAPTA-AM Treatment Are Not Related to Na + /K + ATPase Inhibition
Recently, it was shown that BAPTA-AM can also affect Na + /K + ATPase activity [50]. To discriminate the Ca 2+ -chelating role of BAPTA-AM from its inhibitory effect on the Na + /K + ATPase, we combined AZC treatment with various concentrations of the specific Na + /K + ATPase inhibitor ouabain and assessed the levels of total and phospho-eIF2α and the levels of LC3-II. In contrast to the effects observed with BAPTA-AM, increasing ouabain concentrations resulted in strongly increased levels of phospho-eIF2α, while LC3-II levels generally increased up to 500 nM ouabain followed by a decrease at the highest used concentration of ouabain ( Figure 7A,B). Moreover, these effects occurred independently of the absence or presence of AZC. These results indicate that levels of phospho-eIF2α induced by AZC is dependent on Ca 2+ , and may thus constitute an upstream event to the sensitivity of the subsequent autophagic process to Ca 2+ .

Effects of BAPTA-AM Treatment Are Not Related to Na + /K + ATPase Inhibition
Recently, it was shown that BAPTA-AM can also affect Na + /K + ATPase activity [50]. To discriminate the Ca 2+ -chelating role of BAPTA-AM from its inhibitory effect on the Na + /K + ATPase, we combined AZC treatment with various concentrations of the specific Na + /K + ATPase inhibitor ouabain and assessed the levels of total and phospho-eIF2α and the levels of LC3-II. In contrast to the effects observed with BAPTA-AM, increasing ouabain concentrations resulted in strongly increased levels of phospho-eIF2α, while LC3-II levels generally increased up to 500 nM ouabain followed by a decrease at the highest used concentration of ouabain ( Figure 7A,B). Moreover, these effects occurred independently of the absence or presence of AZC. These results are therefore distinct from those obtained with BAPTA-AM ( Figures 5 and 6) and thus support a role for intracellular Ca 2+ in AZC-induced ER stress and autophagy.

Pretreatment with AZC Reduces the Cytosolic Amount of Ca 2+ after ER Store Release
To determine the relation between AZC treatment and intracellular Ca 2+ handling, we investigated whether AZC treatment directly impacts the Ca 2+ stores. We therefore loaded the cells with the cytosolic Ca 2+ indicator Fura-2 AM and followed its fluorescence on a FlexStation 3 microplate reader. To prevent influx of Ca 2+ in the cell, the extracellular Ca 2+ was first chelated by addition of 3 mM EGTA, and the SERCA inhibitor TG was subsequently used to uncover Ca 2+ release from the ER. While TG led to an increase in cytosolic [Ca 2+ ], acute application of either 5 mM or 10 mM AZC did not provoke a rise in cytosolic [Ca 2+ ] ( Figure 8A). However, when cells were pre-treated These results are therefore distinct from those obtained with BAPTA-AM ( Figures 5 and 6) and thus support a role for intracellular Ca 2+ in AZC-induced ER stress and autophagy.

Pretreatment with AZC Reduces the Cytosolic Amount of Ca 2+ after ER Store Release
To determine the relation between AZC treatment and intracellular Ca 2+ handling, we investigated whether AZC treatment directly impacts the Ca 2+ stores. We therefore loaded the cells with the cytosolic Ca 2+ indicator Fura-2 AM and followed its fluorescence on a FlexStation 3 microplate reader. To prevent influx of Ca 2+ in the cell, the extracellular Ca 2+ was first chelated by addition of 3 mM EGTA, and the SERCA inhibitor TG was subsequently used to uncover Ca 2+ release from the ER. While TG led to an increase in cytosolic [Ca 2+ ], acute application of either 5 mM or 10 mM AZC did not provoke a rise in cytosolic [Ca 2+ ] ( Figure 8A). However, when cells were pre-treated for 6 h with 5 mM or 10 mM AZC and TG-induced [Ca 2+ ] elevations were monitored, we noticed a lower rise of cytosolic Ca 2+ in the pre-treated samples as compared to the control samples ( Figure 8B). This was reflected by a~35% lower area under the curve ( Figure 8C). Additionally, the clearance of Ca 2+ from the cytosol appeared faster after AZC pretreatment, as indicated by a~42% lower τ value for the decline phase of the Fura-2 AM signal ( Figure 8D). for 6 h with 5 mM or 10 mM AZC and TG-induced [Ca 2+ ] elevations were monitored, we noticed a lower rise of cytosolic Ca 2+ in the pre-treated samples as compared to the control samples ( Figure 8B). This was reflected by a ~35% lower area under the curve ( Figure 8C). Additionally, the clearance of Ca 2+ from the cytosol appeared faster after AZC pretreatment, as indicated by a ~42% lower τ value for the decline phase of the Fura-2 AM signal ( Figure 8D). In short, AZC does not have an acute effect on intracellular Ca 2+ , though a prolonged incubation with the compound appears to modulate intracellular Ca 2+ signaling in a complex way.

AZC Does Not Affect the ER Ca 2+ Store Content or the ER-Mitochondrial Ca 2+ Transfer
Because we observed a lower rise in cytosolic [Ca 2+ ] upon application of TG after pretreatment with AZC, we hypothesized that AZC may lower the ER Ca 2+ store content, or that Ca 2+ transfer to the mitochondria may have been increased. In order to investigate this, we co-transfected cells with both G-CEPIA1er and R-GECO1mt to simultaneously detect ER Ca 2+ and mitochondrial Ca 2+ , and treated the cells for 6 h with 5 mM or 10 mM AZC. EGTA was then applied and Ca 2+ released from the ER with TG. Surprisingly, we did not detect differences in the ER Ca 2+ -leak rates between cells pretreated with either 5 mM or 10 mM AZC and untreated control cells ( Figure 9A). Furthermore, the amount of Ca 2+ transferred into the mitochondria after TG application was also very similar between AZC-treated cells and untreated cells ( Figure 9B). Finally, the difference in the TG-triggered In short, AZC does not have an acute effect on intracellular Ca 2+ , though a prolonged incubation with the compound appears to modulate intracellular Ca 2+ signaling in a complex way.

AZC Does Not Affect the ER Ca 2+ Store Content or the ER-Mitochondrial Ca 2+ Transfer
Because we observed a lower rise in cytosolic [Ca 2+ ] upon application of TG after pretreatment with AZC, we hypothesized that AZC may lower the ER Ca 2+ store content, or that Ca 2+ transfer to the mitochondria may have been increased. In order to investigate this, we co-transfected cells with both G-CEPIA1er and R-GECO1mt to simultaneously detect ER Ca 2+ and mitochondrial Ca 2+ , and treated the cells for 6 h with 5 mM or 10 mM AZC. EGTA was then applied and Ca 2+ released from the ER with TG. Surprisingly, we did not detect differences in the ER Ca 2+ -leak rates between cells pretreated with either 5 mM or 10 mM AZC and untreated control cells ( Figure 9A). Furthermore, the amount of Ca 2+ transferred into the mitochondria after TG application was also very similar between AZC-treated cells and untreated cells ( Figure 9B). Finally, the difference in the TG-triggered cytosolic [Ca 2+ ] rise between AZC-treated and untreated cells could also not be linked to a change in expression levels of any of the major Ca 2+ -transport proteins, such as IP 3 Rs, PMCAs, SERCA2B, or the mitochondrial Ca 2+ uniporter (MCU) (Figure 9C,D).
Taken together, long-term treatment of cells with AZC (6 h) seems to modify cellular Ca 2+ handling, though without directly affecting the ER Ca 2+ leak, ER-mitochondrial Ca 2+ transfer or the expression of IP 3 Rs, PMCAs, SERCA2B, and the MCU. cytosolic [Ca 2+ ] rise between AZC-treated and untreated cells could also not be linked to a change in expression levels of any of the major Ca 2+ -transport proteins, such as IP3Rs, PMCAs, SERCA2B, or the mitochondrial Ca 2+ uniporter (MCU) (Figure 9C,D). Taken together, long-term treatment of cells with AZC (6 h) seems to modify cellular Ca 2+ handling, though without directly affecting the ER Ca 2+ leak, ER-mitochondrial Ca 2+ transfer or the expression of IP3Rs, PMCAs, SERCA2B, and the MCU.

Discussion
One of the main functions of the ER is the proper folding of nascent proteins. To ensure this, chaperones assist and correct the process when necessary [51]. Stress conditions including oxidative stress or high temperatures can disturb the folding and cause an accumulation of misfolded proteins in the ER and subsequently ER stress. Activation of the UPR can resolve this situation but failure to do so may result in accumulation of protein aggregates and eventually cell death [52]. Prolonged ER stress may also be involved in neurodegenerative diseases including Alzheimer's disease, type 2 diabetes, and cancer [53].

Discussion
One of the main functions of the ER is the proper folding of nascent proteins. To ensure this, chaperones assist and correct the process when necessary [51]. Stress conditions including oxidative stress or high temperatures can disturb the folding and cause an accumulation of misfolded proteins in the ER and subsequently ER stress. Activation of the UPR can resolve this situation but failure to do so may result in accumulation of protein aggregates and eventually cell death [52]. Prolonged ER stress may also be involved in neurodegenerative diseases including Alzheimer's disease, type 2 diabetes, and cancer [53].
Classic chemical ER stress inducers include dithiothreitol, tunicamycin, and TG. In our preliminary experiments however, these compounds also showed induction of apoptosis or they displayed an intrinsic effect on Ca 2+ homeostasis. Therefore, we turned to AZC. This proline analogue contains a four-membered carbon ring in contrast to the five-membered ring in proline ( Figure 1A). Incorporation of AZC in proteins leads to an altered tertiary protein structure and accumulation of protein aggregates [37,38,54]. Our results indicate that this activates in HeLa cells the PERK and the ATF6 arms of the UPR but not the IRE1 arm (Figure 1). This is in line with the results of Qian et al. who in HEK293 cells found a dose-and time-dependent increase of phospho-eIF2α upon AZC treatment [39], and with Shang et al. who showed that significant XBP1 splicing requires very high concentrations of AZC [41]. The former correlated with a decrease in phosphorylation of ribosomal protein S6, indicating inhibition of the autophagy suppressor mTOR. This is in line with our results with Baf A1 (Figure 3), which suggest a role of AZC in the induction of autophagy. Moreover, using the specific PERK inhibitor AMG PERK 44, we can conclude that activation of the PERK arm of the UPR is absolutely needed for the subsequent increase in LC3-II levels by AZC. Finally, Nivon et al. showed that AZC treatment causes the formation of protein aggregates in HeLa cells and that this correlates with an NFκB-dependent increase in LC3-II levels [38].
Ca 2+ plays an important role in both ER stress and autophagy [7,9,14]. In autophagy, depending on the exact conditions both inhibitory and stimulatory effects of Ca 2+ were described [7,55]. To assess the role of Ca 2+ in intracellular processes like autophagy, treatment with the intracellular Ca 2+ chelator BAPTA-AM is the favored technique [7]. We therefore co-treated our cells with AZC and BAPTA-AM, upon which we observed a decrease of AZC-induced elevation of phospho-eIF2α as well as of LC3-II (Figures 5 and 6). However, recently it was shown that various related Ca 2+ -binding molecules, including BAPTA-AM, can inhibit the Na + /K + ATPase, independently of their Ca 2+ -binding affinity [50]. Although in the latter study, BAPTA-AM was a less potent Na + /K + ATPase inhibitor than the other compounds tested, it readily accumulates in the cell and we therefore wanted to verify whether Na + /K + ATPase inhibition by BAPTA-AM could affect the interpretation of our results. One way to discriminate between the Ca 2+ chelating and the Na + /K + ATPase inhibiting effect of BAPTA-AM, is to assess whether the effects provoked by BAPTA-AM are replicated by the well-known specific Na + /K + ATPase inhibitor ouabain [56]. All human Na + /K + ATPase isozymes are characterized by a K d for ouabain between 13 and 37 nM [57]. In order to encompass various levels of Na + /K + ATPase inhibition, we applied a range of concentrations of ouabain between 10 nM and 1 µM. However, at none of these concentrations did ouabain (Figure 7) mimic the effect of BAPTA-AM on AZC-induced elevation of phosphorylated eIF2α and LC3-II ( Figures 5 and 6). Therefore, we conclude that these effects represent a bona fide contribution of Ca 2+ chelation rather than Na + /K + ATPase inhibition and that intracellular Ca 2+ is critical for both AZC-induced ER stress and elevation of LC3-II.
Interestingly, our data indicate an essential role of Ca 2+ in AZC-induced elevation of phospho-eIF2α. Given that the PERK-arm of the UPR has been reported to be required for ER stress-mediated induction of LC3-II [58,59], this finding may form a link with the requirement for Ca 2+ observed for the subsequent elevation of LC3-II levels. Whereas Ca 2+ was recently shown to be required for non-UPR-related functions of PERK [60], our study indicates that Ca 2+ is also required for efficient eIF2α phosphorylation, and thus for transduction of PERK-arm mediated UPR signaling. On the other hand, Ca 2+ did not seem to be required for AZC-mediated upregulation of BiP. The Ca 2+ sensitivities of the different parts of the UPR signal transduction pathways remain to be determined, but it is already noteworthy that the BiP upregulation is insensitive to both the PERK inhibitor and to BAPTA-AM.
Finally, we found that 6 h of pre-treatment with 5-10 mM AZC reduced the amount of Ca 2+ released from the ER Ca 2+ store with TG, but did not induce Ca 2+ release itself (Figures 8 and 9). These results are in agreement with those of Caspersen et al. who previously demonstrated in PC12 cells that AZC did not induce Ca 2+ release from the ER by itself although they also claim that the ER is not significantly depleted after pre-treatment for 4 h with 5 mM AZC [61]. However, in the latter study, no quantification of the Ca 2+ signal was performed, and on the traces shown the agonist-induced Ca 2+ release is anyway smaller than in control conditions. We initially hypothesized that the reduction in TG-induced cytosolic Ca 2+ rise after pre-incubation with AZC may be due to a reduced Ca 2+ content in the ER and therefore a lower release of Ca 2+ into the cytosol upon TG application, or due to an increased Ca 2+ transfer from ER to mitochondria, thereby preventing detection of the Ca 2+ in the cytosol. However, when measuring Ca 2+ simultaneously in the ER and the mitochondria, we did neither observe a reduction in ER store content nor an increase in mitochondrial Ca 2+ transfer in cells pretreated for 6 h with 5 or 10 mM AZC as compared to control cells ( Figure 9A,B). We therefore speculate that pre-treatment with AZC affects Ca 2+ -transport systems outside the ER and mitochondrial compartments. For instance, it is possible that PMCA, responsible for exporting Ca 2+ from the cytosol to the extracellular environment, is affected. We could exclude a modification in PMCA expression levels ( Figure 9C) though a change in activity remains possible. A more efficient export of Ca 2+ from the cytosol by PMCA may lead to an apparent reduction of the amount of cytosolic Ca 2+ without an actual decrease of the ER Ca 2+ store content, and thus explain the reduced τ value for the clearance of Ca 2+ from the cytosol.
Taken together, our results indicate a critical role for Ca 2+ in the process of eIF2α phosphorylation and LC3-II production upon AZC treatment. AZC however does not directly affect ER Ca 2+ store content or ER-to-mitochondria Ca 2+ transfer and the observed effects on intracellular Ca 2+ handling are likely not related to its effect on UPR and autophagy.
Author Contributions: G.R. performed the experiments with the assistance of E.H., K.W., and T.L., N.E. contributed to the cell toxicity assay. G.R., G.B., and J.B.P. designed the study. The first version of the manuscript was written by G.R. and J.B.P. and subsequently revised and approved by all authors.

Acknowledgments:
We thank E. Kania for helpful input and assistance in the course of this study and M. Iino (Univ. Tokyo, Japan) for the generous gift of plasmids encoding ER-and mitochondria-targeted Ca 2+ probes.

Conflicts of Interest:
The authors declare no conflict of interest. The funding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.