Main

The endoplasmic reticulum (ER) serves as the primary processing site for membrane and secreted proteins. The ER recruits translating ribosomes, translocates newly synthesized polypeptides into its lumen, and promotes a variety of post-translational modifications and chaperone-facilitated protein folding.1, 2 Proper ER function is critical for numerous aspects of cell physiology, including vesicle trafficking, lipid and membrane biogenesis, and protein targeting and secretion. Accordingly, cells react rapidly to various forms of ER dysfunction – including the accumulation of unfolded, misfolded or excessive protein, ER lipid or glycolipid imbalances, or changes in the redox or ionic conditions of the ER lumen – through a set of adaptive pathways known collectively as the ER stress response (ESR).2, 3, 4, 5

The ESR promotes cell survival both by increasing the capacity of the ER to fold and process client proteins and by reducing the amount of protein inside the ER. These effects are achieved through three major pathways: (1) the unfolded protein response, a transcription-dependent induction of ER lumenal chaperone proteins and many other components of the secretory apparatus, which augments the polypeptide processing capacity of the ER;5, 6 (2) the activation of proteasome-dependent ER-associated degradation to remove proteins from the ER;7, 8 and (3) the control of protein translation to modulate the polypeptide traffic into the ER.9, 10 Normally, this suite of responses succeeds in restoring ER homeostasis. However, in metazoans, persistent or intense ER stress can also trigger apoptosis.11, 12, 13, 14

ER stress and the apoptotic program coupled to it have been implicated in many important pathologies, including diabetes, obesity, neurodegenerative disorders, viral infection, and a variety of ER storage diseases.5, 15, 16, 17, 18, 19 Nevertheless, the initiation and execution steps of ER stress-induced apoptosis in mammals remain poorly understood. Classical genetics has allowed detailed dissection of the ESR in lower organisms such as Saccharomyces cerevisiae, but greater complexity and intractable genetics have hindered progress in mammalian systems. In particular, there is a need for new research tools to study ER stress-induced apoptosis in mammals and to explore how this pathway might be manipulated for therapeutic benefit. To address these needs, we have undertaken a chemical biology approach to study ER stress-induced apoptosis in mammalian cells.

Previously, we identified salubrinal (sal), a small molecule that protects cells from ER stress-induced apoptosis by inhibiting the dephosphorylation of the translation initiation factor eukaryotic translation initiation factor 2 subunit α (eIF2α) by GADD34/PP1 and related phosphatase complexes, which are involved in the endogenous response to ER stress.20 As we have shown, sal affects eIF2α dephosphorylation but not the dephosphorylation of many other PP1 substrates in the cell.20 Therefore, sal selectively enforces a phosphorylation that is part of the endogenous ESR, but through a novel mechanism. Because sal selectively affects a known ESR pathway, we asked what other previously unknown effects sal might have on cells and whether these additional effects might participate in cell survival signaling during ER stress as well.

The eukaryotic translation elongation factor 2 (eEF-2) is an essential mediator of the ribosomal elongation step during mRNA translation.21 The activity of eEF-2 can be regulated post-translationally: phosphorylation of eEF-2 on Thr56 inhibits its translational function by blocking its ability to promote ribosome translocation.22, 23, 24 The phosphorylation of eEF-2 is catalyzed by eEF-2 kinase (eEF-2K), an unusual calcium/calmodulin-dependent enzyme belonging to the alpha kinase family of atypical protein kinases.25 EEF-2 phosphorylation has been shown to play an important role in coupling protein synthesis to energy metabolism in response to calcium flux,26 cyclic adenosine monophosphate (AMP)/protein kinase A and AMP-activated protein kinase signaling,27, 28, 29, 30 amino acid or glucose limitation,31, 32 and cytoplasmic pH changes,33 but no role for eEF-2 phosphorylation in ER stress has been reported.

Here, we demonstrate that eEF-2 is phosphorylated in response to sal treatment. This result with sal prompted us to investigate the role of eEF-2 phosphorylation in endogenous ESR, leading us to discover that this signaling event occurs during ER stress as well. We also show that eEF-2K is essential for the phosphorylation of eEF-2 in response to both sal and ER stress. In addition, eIF2α phosphorylation is necessary for the phosphorylation of eEF-2 in response to sal, but dispensable in response to ER stress, suggesting that ER stress induces eEF-2 phosphorylation by multiple mechanistically distinct but non-mutually exclusive pathways. Finally, we show that genetic ablation of eEF-2K impairs the downregulation of translation in response to ER stress and confers partial resistance to ER stress-induced cell death, demonstrating that eEF-2K activation by ER stress is physiologically relevant and promotes apoptosis. This work identifies eEF-2K as an important effector of ER stress signaling and underlines the utility of novel small molecules in discovering new biology.

Results

Because sal selectively promotes eIF2α phosphorylation, a cell signaling pathway that participates in the endogenous ESR,20 we reasoned that other, as yet unknown, molecular effects of sal might also be relevant to endogenous ER stress signaling. Therefore, we took a proteomics approach based on difference gel electrophoresis (DIGE) to search for additional effects of sal treatment. In a DIGE experiment, two protein samples are separately covalently labeled with one of two dyes derived from the Cy3 and Cy5 fluorophores.34 The dyes are identical in isoelectric point and nearly identical in molecular mass, and so they do not differentially affect the migration of polypeptides on a two-dimensional (2D) gel, consisting of isoelectric focusing (IEF) followed by standard SDS-PAGE.34 The two separately labeled protein samples are mixed and run on a single 2D gel. Then, the differences in protein spots can be visualized by comparing images of the same gel on both the Cy3 and Cy5 wavelengths to detect changes between samples.34 In principle, DIGE experiments can reveal changes in individual proteins' abundance, as well as any post-translational modification that affects their migration on a 2D gel. DIGE is therefore well suited for identifying changes to the proteome induced by a small molecule because it provides a sensitive, simultaneous survey of thousands of proteins while avoiding the reproducibility problems that plague the comparison of spots from different 2D gels.34

We used DIGE to search for sal-induced changes to the proteome in the rat pheochromocytoma cell line PC12. Replicate plates of cells were treated either with 75 μM sal or DMSO vehicle control, collected at regular intervals over a 36-h time course (starting at 0 h), lysed and labeled with Cy3- and Cy5-based dyes essentially as described.35 At each time point, the sample was split in half, and one half was labeled with Cy3 and the other with Cy5. This allowed us to run two complementary 2D gels per time point with reciprocal dye labeling for the DMSO- and sal-treated samples, in order to rule out any trivial dye-dependent effects on protein mobility in the gel.

Equal protein amounts of the complementary samples from each time point were mixed, run on a single 2D gel and analyzed using a custom-built imaging robot to visualize the Cy3- and Cy5-labeled protein spots, as described.35 A representative pair of images from the 17.5 h time point, with Cy3-labeled lysate from DMSO-treated cells and Cy5-labeled lysate from sal-treated cells is shown (Figure 1a). By comparing the gel images from each time point (see Materials and Methods and Supplementary Figure 1), we found that only three to five of the hundreds of detectable protein spots reproducibly shifted in response to sal treatment over a period of 36 h, the time point at which we observe optimal protection by sal from ER stress in this cell type.20 The modest extent of these changes indicates that sal treatment does not grossly perturb cell physiology or protein modification in this context, consistent with our previous observations using other assays.20 However, the reproducible shifts we observed were dye-independent and, in a few cases, increased over time up to 36 h. We concluded that sal affects a handful of proteins besides eIF2α within the time frame relevant for ER stress protection.

Figure 1
figure 1

Sal induces phosphorylation of eEF-2 but few other changes to the proteome in PC12 cells. (a) Single DIGE gel comparing DMSO- (top, Cy3 image) and sal-treated (bottom, Cy5 image) samples. PC12 cells were treated as indicated for 17.5 h, lysed, stained with DIGE dyes, mixed and run on a 2D gel. The spots corresponding to eEF-2 are marked (white arrowheads). (b) PC12 cells were treated with DMSO only (–) or 75 μM sal. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein

One set of protein spots shifted particularly prominently in our DIGE experiments, moving toward the anode of the IEF dimension in response to sal (i.e., to the left in Figure 1a, white arrowheads; see also Supplementary Figure 1). We excised these spots, analyzed them by mass spectrometry and identified them as eEF-2 (Supplementary Figure 2a). A shift toward the anode in the IEF dimension is consistent with a phosphorylation event, which increases the negative charge on a protein. Therefore, we re-examined our mass spectrometry data to search for signals consistent with phosphorylated peptides from the eEF-2 sequence. We found one peak consistent with a phosphorylation on the tryptic peptide FTDTR, corresponding to residues 55–59 of rat eEF-2 (Supplementary Figure 2b). This site matches exactly with a previously described phosphorylation site on eEF-222, 36, 37 and suggests that our DIGE experiment had detected a sal-induced phosphorylation of eEF-2. We confirmed that sal induces eEF-2 phosphorylation, using an immunoblot for phosphorylated Thr56, the major site of eEF-2 phosphorylation in vivo.38, 39 Indeed, sal treatment induced robust phosphorylation of Thr56 without altering the level of total eEF-2 protein (Figure 1b).

Previously, we showed that sal induces the phosphorylation of eIF2α on Ser51,20 an event that is also a part of the endogenous ESR. Because sal mimics some aspects of the ESR and also because our DIGE results showed that sal induces eEF-2 phosphorylation, we asked whether eEF-2 phosphorylation might be a component of the endogenous ESR as well. Indeed, we found that ER stress mediated by the protein N-linked glycosylation inhibitor tunicamycin (Tm) resulted in eEF-2 phosphorylation in PC12 cells in a dose- (Figure 2a) and time-dependent (Figure 2b) manner, consistent with a specific signaling event triggered by ER stress. Furthermore, pharmacological inducers of ER stress such as Tm and thapsigargin (Tg), an inhibitor of the ER-resident Ca2+-dependent ATPase, induced eEF-2 phosphorylation with kinetics similar to those of sal (Supplementary Figure 3). EEF-2 phosphorylation has been observed in response to several stimuli,26, 27, 28, 29, 30, 31, 32, 33 but never in the context of ER stress. Our observations suggest that eEF-2 might be regulated by phosphorylation during ESR.

Figure 2
figure 2

ER stress induces dose- and time-dependent phosphorylation of eEF-2. (a) PC12 cells were treated with Tm for 24 h as indicated. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein. (b) ER stress induces time-dependent eEF-2 phosphorylation in wild-type MEFs. Cells were treated with 100 ng/ml Tm as indicated. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein. Reduction in total eEF-2 protein at 24 h is due to cell death (see also Supplementary Figure 4)

The ER-resident kinase PERK is a direct sensor of perturbations in ER homeostasis10 and is indispensable for eIF2α phosphorylation and translational control during ER stress.40 Therefore, we asked whether the phosphorylation of eEF-2 in response to ER stress also requires PERK. As with PC12 cells, wild-type mouse embryonic fibroblasts (MEFs) induced eEF-2 phosphorylation in response to Tm in a dose-dependent manner (Figure 3a). However, unlike the case of eIF2α phosphorylation,40 PERK−/− MEFs induced eEF-2 phosphorylation in response to Tm to a similar extent as did wild-type MEFs (Figure 3a). These results demonstrate that eEF-2 phosphorylation during ER stress does not depend on PERK. We concluded that eEF-2 phosphorylation may represent another form of translational control induced by ER stress, separate from the previously described PERK-dependent phosphorylation of eIF2α.

Figure 3
figure 3

EEF-2 phosphorylation during ER stress is independent of PERK but dependent on eEF-2K. (a) Wild-type and PERK−/− MEFs were treated with Tm as indicated for 24 h. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein. Reduction in total eEF-2 protein at high Tm doses is due to cell death (see also Supplementary Figure 4). (b) Wild-type and eEF-2K−/− MEFs were treated with sal as indicated for 24 h. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein. (c) Wild-type and eEF-2K−/− MEFs were treated with Tm as indicated for 24 h. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein. Reduction in total eEF-2 protein at high Tm doses is due to cell death (see also Supplementary Figure 4)

EEF-2K is the only kinase known thus far to phosphorylate eEF-2 on Thr56. Mammals have a single eEF-2K gene, and homozygous deletion of the murine eEF-2K gene completely eliminates detectable eEF-2 phosphorylation in a variety of tissues.25 We therefore asked whether eEF-2K is required for the phosphorylation of eEF-2 in response to sal or ER stress. Sal treatment induced eEF-2 phosphorylation in wild-type MEFs in a dose-dependent manner (Figure 3b). In contrast, eEF-2K−/− MEFs41 showed no eEF-2 phosphorylation in response to sal at any dose (Figure 3b). Similarly, Tm also induced eEF-2 phosphorylation in wild-type MEFs in a dose-dependent manner, while eEF-2K−/− MEFs showed no eEF-2 phosphorylation in response to any Tm dose over a wide range (Figure 3c). We concluded that eEF-2K is essential for eEF-2 phosphorylation in response to sal and Tm.

Because sal induces eIF2α phosphorylation as an early event,20 we asked whether eIF2α phosphorylation is required for subsequent eEF-2 phosphorylation in response to sal or Tm. We treated wild-type MEFs or MEFs harboring a homozygous Ser51 → Ala mutation in the eIF2α gene (eIF2αA/A MEFs), which ablates the phosphorylation site,42 with sal or Tm and assessed the induction of eEF-2 phosphorylation by immunoblot (Figure 4a). As before, wild-type MEFs induced eEF-2 phosphorylation in response to both sal and Tm (Figure 4a). Interestingly, however, eIF2αA/A MEFs induced eEF-2 phosphorylation in response to Tm but not to sal (Figure 4a). These results demonstrate that the induction of eEF-2 phosphorylation by sal requires eIF2α phosphorylation, suggesting that a specific signal downstream of eIF2α phosphorylation can activate eEF-2K. In contrast to sal, Tm can induce eEF-2 phosphorylation independently of eIF2α phosphorylation (Figure 4a). Therefore, while sal triggers eEF-2 phosphorylation exclusively through the eIF2α phosphorylation pathway, additional pathways are involved in the induction of eEF-2 phosphorylation downstream of bona fide ER stress inducers such as Tm.

Figure 4
figure 4

EEF-2 phosphorylation in response to sal, but not ER stress, is dependent on eIF2α phosphorylation and translation. (a) Wild-type or eIF2αA/A MEFs were treated with sal or Tm as indicated for 24 h. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein. (b) Wild-type and ATF4−/− MEFs were treated with Tm as indicated for 24 h. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein. Reduction in total eEF-2 protein at high Tm doses is due to cell death (see also Supplementary Figure 4). (c) Wild-type MEFs were treated as indicated for 24 h. Lysates were analyzed by immunoblot for eEF-2 phosphorylation on Thr56 and total eEF-2 protein

To further dissect the mechanism(s) by which sal and Tm induce eEF-2 phosphorylation, we examined signaling events downstream of eIF2α phosphorylation. The primary downstream effect of eIF2α phosphorylation during ER stress is the selective translational induction of certain ER stress-responsive mRNAs, including that of the C/EBP family transcription factor ATF410, 42, 43 and its target CHOP.43, 44 Once translated, ATF4 and CHOP mediate a broader transcription-dependent program of gene expression that is an important component of the ESR.42, 43, 45, 46 Because eEF-2 phosphorylation in response to sal requires eIF2α phosphorylation (Figure 4a), we asked whether ATF4 or CHOP, which are induced by eIF2α phosphorylation, might be responsible for mediating the signal between eIF2α and eEF-2. We treated wild-type and ATF4−/− MEFs47, 48 with either sal or Tm and measured eEF-2 phosphorylation by immunoblot (Figure 4b). Sal and Tm both induced eEF-2 phosphorylation regardless of ATF4 genotype (Figure 4b), indicating that ATF4 is not required for the eIF2α phosphorylation-dependent signal that results in eEF-2 phosphorylation. Similar results were obtained with CHOP−/− MEFs (Supplementary Figure 5). Interestingly, however, the putative signal between eIF2α and eEF-2 requires new protein synthesis, as the translation inhibitor cycloheximide (CHX) blocked eEF-2 phosphorylation in response to sal in wild-type MEFs (Figure 4c). Taken together, these results suggest that eIF2α phosphorylation induces the translation of a protein other than ATF4 or CHOP that directly or indirectly leads to the activation of eEF-2K, and subsequent eEF-2 phosphorylation.

EEF-2 phosphorylation has not previously been observed during ER stress, but its effect on protein translation in other contexts is well established.22, 23, 24, 49 We therefore asked whether eEF-2K is necessary for the full downregulation of translation observed during ER stress. We treated wild-type and eEF-2K−/− MEFs with Tg to induce rapid, synchronous ER stress and then metabolically labeled newly synthesized proteins with radioactive amino acids. Consistent with previous studies,42 Tg downregulated translation in wild-type MEFs (Figure 5a). However, the Tg-mediated downregulation of translation in eEF-2K−/− MEFs was reduced (Figure 5a). These results suggest that eEF-2K may participate in downregulating translation during ER stress by inhibiting peptide chain elongation in response to Tg.

Figure 5
figure 5

Loss of eEF-2K diminishes translational shutdown and promotes cell survival during ER stress. (a) Equal numbers of wild-type or eEF-2K−/− MEFs were incubated with 1 μM Tg for 2 h, washed and incubated with 35S-labeled methionine/cysteine mix for 15 min. Lysate samples of equal protein amount were analyzed by autoradiography to measure new protein synthesis. (b) Transformed wild-type (black diamonds) or eEF-2K−/− (gray squares) MEFs were treated as indicated for 24 h. Cell viability was assessed via ATP content and was normalized to matched vehicle-only controls. Error bars represent standard deviations here and throughout. (c) Transformed wild-type (black diamonds) or eEF-2K−/− (gray squares) MEFs were treated as indicated for 24 (top) or 48 (bottom) h and the activity of caspase-3/7-type enzymes was measured. Values are given as fold enzyme activity over the vehicle-treated control. (d) Transformed wild-type or eEF-2K−/− MEFs were treated with DMSO or 200 ng/ml Tm for 25.5, 50, or 74 h. Cell viability was assessed by bright field microscopy. (e) The living (i.e., attached) and dead (i.e., floating) cells in the experiment described in (d) were counted at the 25.5- and 50-h time points

Translational regulation by PERK is critical for mammalian cell survival during ER stress.40, 45 We therefore asked whether eEF-2K might also regulate cell survival during ER stress. We examined the sensitivity of primary wild-type and eEF-2K−/− MEFs41 to a range of doses of Tm. MEFs of both genotypes showed a dose-dependent reduction in cell viability after Tm treatment but, at all Tm doses tested, the eEF-2K−/− MEFs retained significantly higher viability than did the wild-type cells (Figure 5b and Supplementary Figure 6). The loss of eEF-2K also reduced the activity of caspase-3/7-type enzymes50 (Figure 5c) and annexin V positivity (Supplementary Figure 6) observed in response to Tm treatment, indicating that the cell death we observed was apoptotic and promoted by eEF-2K. The partial cytoprotection due to eEF-2K loss was consistent over a prolonged period of ER stress (Figure 5d), and it was comparable to the previously observed protection conferred by the deletion of known proapoptotic ESR genes such as CHOP,51 ASK152 and caspase-12.53 We concluded that the genetic ablation of eEF-2K partially protects against ER stress-induced apoptosis.

Finally, our model for eEF-2K signaling in response to sal and ER stress predicts that sal should protect both wild-type and eEF-2K−/− MEFs from ER stress-induced apoptosis because sal is cytoprotective,20 while eEF-2K is proapoptotic under these circumstances (Figure 5). We tested this prediction of our model in two experiments. First, we asked whether sal showed dose-dependent protection of both wild-type and eEF-2K−/− MEFs. We found that sal protected MEFs of both genotypes from Tm in a dose-dependent manner, as measured by cellular ATP content and mitochondrial activity (Figure 6a). Furthermore, while cells of both genotypes were significantly (P<0.05) protected by sal, eEF-2K−/− MEFs were intrinsically more resistant to Tm than were wild-type cells (Figure 6a), consistent with our earlier results (Figure 5). Second, we asked whether a given dose of sal could protect both wild-type and eEF-2K−/− MEFs over a range of Tm concentrations (Figure 6b). We found that sal treatment inhibited Tm-induced apoptosis in MEFs of both genotypes, as measured by ATP and mitochondrial activity (Figure 6b), consistent with our model.

Figure 6
figure 6

Sal protects both wild-type and eEF-2K−/− MEFs from Tm-induced apoptosis. (a) Transformed wild-type (black diamonds) or eEF-2K−/− (gray squares) MEFs were treated with 200 ng/ml Tm plus the indicated sal concentrations for 48 h. Cell viability was assessed by ATP content (top) and MTS assay to measure mitochondrial function (bottom) and was normalized to matched vehicle-only controls. (b) Transformed wild-type (black) or eEF-2K−/− (gray) MEFs were treated as indicated for 48 h. DMSO serves as a vehicle control for sal. Cell viability was assessed by measuring ATP content (top) and mitochondrial function (bottom) and was normalized to matched vehicle-only controls. T-test analysis showed that sal protects both wild-type and eEF-2K−/− cells from Tm (versus wild-type or eEF-2K−/− cells, respectively, treated only with Tm plus DMSO vehicle control; P<0.05)

Discussion

We have shown that eEF-2 phosphorylation occurs in response to both sal treatment and ER stress and have identified eEF-2K as the kinase responsible for this signal (see Figure 7 for an illustration of our proposed model). Because eEF-2 phosphorylation in response to Tm is independent of PERK but stringently dependent on eEF-2K, our results suggest that eEF-2K may mediate a previously undescribed translational control pathway during ER stress. Furthermore, the genetic loss of eEF-2K partially protects cells from ER stress, indicating that eEF-2K activity promotes apoptosis in this context. In addition, sal and ER stress induce eEF-2 phosphorylation differently, since the former requires eIF2α phosphorylation, whereas the latter does not. Therefore, the use of sal enabled us to demonstrate that there is a novel signaling pathway connecting eIF2α phosphorylation and eEF-2 phosphorylation.

Figure 7
figure 7

Proposed signaling model. The multiple arrows from phosphorylated eIF2α to eEF-2K represent a signal or pathway resulting in the translation of an unknown factor that promotes eEF-2K activation. The arrow from Tm to eEF-2K represents an eIF2α phosphorylation-independent mechanism leading to eEF-2K activation, perhaps including calcium flux. See text for details

Our work has implications for both the phenotypic consequences of eEF-2K activation during ER stress and for the mechanistic models of how eEF-2 phosphorylation can be induced by multiple cell signaling pathways. With respect to phenotypic consequences, our results suggest that eEF-2K promotes cell death and may contribute to translational control, probably at the level of elongation, during ER stress (Figure 5). It is already well established that ESR regulates translation at the level of initiation.42, 43, 45 Why might it be advantageous to regulate translation elongation during ER stress as well? It is likely that eEF-2 phosphorylation affects ribosomal elongation along all mRNAs to a similar extent,54 implying that eEF-2 phosphorylation, in contrast to eIF2α phosphorylation, probably does not activate or inactivate a particular subset of genes during ER stress but instead globally downregulates translation. This important observation may explain why eEF-2K activation is proapoptotic (Figure 5) during ER stress, because a persistent block in all translational elongation would be expected to promote cell death. Consistent with this model, inactivating eEF-2 in other ways, such as doxorubicin-induced phosphorylation55 or diphtheria toxin-mediated ADP ribosylation56, 57 of eEF-2, also causes cell death, whereas potentiating eEF-2 activity can protect cells from some apoptotic stimuli.58 Therefore, eEF-2K may promote cell death during ER stress via its effect on translation elongation.

With respect to the mechanistic models of eEF-2 phosphorylation induction, our results show that, in contrast to sal (discussed below), Tm induces eEF-2 phosphorylation in eIF2αA/A MEFs (Figure 4a) but not in eEF-2K−/− MEFs (Figure 3b), demonstrating that eEF-2K is necessary for this signaling event during ER stress but eIF2α phosphorylation is not. Therefore, an eIF2α-independent mechanism for activating eEF-2K during ER stress must exist. One attractive candidate for this mechanism is calcium flux into the cytoplasm, which occurs during some types of ER stress5, 11, 59, 60 and is known to promote eEF-2K activation.26 Importantly, however, Tm induces relatively little calcium flux in most cell types, including PC12, used in our system.61 Indeed, we relied primarily on Tm in our experiments, to induce robust ER stress while minimizing calcium flux, enabling us to discriminate between the phenotypic effects of ER dysfunction per se and the more pleiotropic downstream effects of calcium flux. However, it is possible that other ER stress stimuli would impact differently on the eEF-2K pathway.62 Additional future studies will be necessary to understand fully how ER stress induces eEF-2 phosphorylation.

In contrast to bona fide ER stress inducers such as Tm, sal cannot induce eEF-2 phosphorylation in either eEF-2K−/− (Figure 3b) or eIF2αA/A MEFs (Figure 4a). Therefore, a previously unrecognized eIF2α phosphorylation-dependent mechanism of eEF-2K activation must also exist. What might mediate this mechanism? Our experiments with CHX demonstrated that protein synthesis is required for eEF-2 phosphorylation induced by sal (Figure 4c). Although CHX reduces the protein level of eEF-2K somewhat, control experiments using RNAi-mediated eEF-2K knockdown have shown that full eEF-2 phosphorylation occurs even at very low levels of eEF-2K (data not shown), indicating that the effect of CHX is not due merely to reduced eEF-2K levels. Because ATF4 and CHOP are dispensable for eEF-2 phosphorylation by sal (Figure 4b and Supplementary Figure 5) but protein synthesis is required (Figure 4c), it may be that other unidentified mRNAs besides ATF4 and CHOP can be translationally upregulated by eIF2α phosphorylation and can directly or indirectly activate eEF-2K. The identification of these putative mRNAs will be an important goal of future studies.

Our results provide an example of the utility of novel bioactive small molecules in studying a complex process such as ER stress in mammalian systems. The use of sal as a tool to probe biochemical pathways led us to discover the phosphorylation of eEF-2 as a hitherto unknown component of the endogenous ESR and also to identify the kinase that mediates that phosphorylation, eEF-2K. Furthermore, we have shown that eEF-2K plays a previously unrecognized, proapoptotic role during ER stress. The unique activity of sal also enabled us to show that eIF2α phosphorylation can generate a signal that leads to the induction of eEF-2 phosphorylation, a result we could not have achieved using only previously available pharmacological inducers of ER stress. In the future, other small molecules like sal that affect specific ER stress pathways may reveal new biological information about the ESR and provide lead molecules for the development of therapeutics to manipulate the ESR in rational, clinically useful ways. Specifically, we note that eEF-2K−/− mice suffer no obvious developmental abnormalities,25 and so small molecule inhibitors of eEF-2K may be useful in inhibiting unwanted apoptosis without toxic side effects in diseases associated with pathological ER stress, including diabetes and neurodegeneration.

Materials and Methods

Chemicals and reagents

Sal was purchased from Chembridge or synthesized as described.63 Tm was obtained from Calbiochem. Tg and CHX were obtained from Sigma. Cy3- and Cy5-based dyes for DIGE experiments (Amersham) were used as described.35 IEF focusing strips, ceramic troughs, sample cups, swelling buffer, and IEF apparatus (IPGphor) were from Amersham and used according to the manufacturer's instructions. 2D gels were stained with GelCode Blue (Pierce). 35S-methionine/cysteine protein labeling mix was obtained from New England Nuclear.

DIGE experiments and image analysis

PC12 cells (5 × 105) were seeded in 10 cm plates and treated with either DMSO or 75 μM sal. At the desired time points, cells were trypsinized, washed and collected and whole-cell lysate was prepared essentially as described.34, 35 Each sample was split in half, and one half was labeled with Cy3 and the other half with Cy5, as described.34 Complementary samples from each time point were mixed and run on the same 2D gel. Gels were visualized for Cy3 and Cy5 wavelengths using a custom-built robot as described.34, 35 Protein spots that shifted in a sal-dependent fashion were identified by visual inspection of the images. To facilitate image analysis, the Cy3 and Cy5 images from each single gel (such as the two images in Figure 1a) were combined into a two-frame movie using IPLab Spectrum software (scanalytic) and Quicktime (Apple Computer). Viewing the movie on a continuous loop aids in identifying shifting spots. A movie version of the still images in Figure 1a is available on the Cell Death and Differentiation website as Supplementary Figure 1. After fluorescent image analysis, the gels were stained with GelCode Blue (Pierce) to visualize spots of interest, which were excised by hand for mass spectrometry analysis.

Mass spectrometry

Mass spectrometry and data analysis of eEF-2 samples from DIGE gels were performed by the Taplin Mass Spectrometry facility at Harvard Medical School.

Cell culture

PERK−/− MEFs,40 ATF4−/− MEFs,47, 48 and CHOP−/− MEFs51 and their respective matched wild-type counterparts were kind gifts of D Ron (Skirball Institute, New York, NY, USA). EIF2αA/A MEFs42 and matched wild-type counterparts were a kind gift of Randy Kaufman (University of Michigan Medical School, Ann Arbor, MI, USA). EEF-2K−/− MEFs and matched wild-type counterparts have been described.41 PC12 cells were purchased from the American Type Culture Collection and were maintained in Dulbecco's modified Eagle's Medium (DMEM) (Gibco) with 10% horse serum and 5% fetal bovine serum (FBS) (Sigma), and 1 × penicillin/streptomycin (PS) (Gibco). ATF4−/− and matched wild-type MEFs were maintained in DMEM supplemented with 10% FBS, 1 × essential amino acids (Gibco), 1 × non-essential amino acids (Gibco), 55 μM β-mercaptoethanol (Fisher), 2 mM L-glutamine (Gibco) and PS, as described.48 EIF2αA/A MEFs and matched wild-type controls were maintained in DMEM with PS and 10% FBS, supplemented with additional 1 × essential amino acids and 1 × non-essential amino acids (Gibco). All other MEFs were maintained in DMEM with PS and 10% FBS or newborn calf serum. To transform MEFs, primary eEF-2K−/− and matched wild-type cells were infected with SV40 large T antigen retrovirus. Transformed MEFs grew out after several culture passages.

Immunoblots

Cells were treated as indicated, lysed in 2 × Laemmli sample buffer, analyzed by SDS-PAGE, transferred to PVDF membranes, blocked and probed according to standard procedures.64 The following antibodies and dilutions were used: anti-total eEF-2 (1 : 1000), anti-phospho-Thr56 eEF-2 (1 : 1000) (Cell Signaling).

Metabolic labeling

Equal numbers of wild-type or eEF-2K−/− MEFs were treated with 1 μM Tg for 2 h, washed twice in methionine/cysteine-free DMEM (Sigma), and then incubated in methionine/cysteine-free DMEM supplemented with 35S-methionine/cysteine protein labeling mix (200 μCi/ml) for 15 min. Cells were washed once with 4°C phosphate-buffered saline and lysed in 2 × SDS-PAGE sample buffer. Equal amounts of protein from each sample were analyzed via 4–20% gradient SDS-PAGE (Bio-Rad) and autoradiography.

Apoptosis assays

The Cell Titer-Glo Assay (ATP), Cell Titer 96 Aqueous Non-Radioactive Cell Proliferation Assay (mitochondrial activity) and Caspase-3/7 Glo Assay (caspase activity) were obtained from Promega and performed according to the manufacturer's instructions in a 96-well plate format and analyzed using a Wallac Victor II plate reader (Perkin Elmer). In all cases, the final concentrations of the DMSO vehicle were equal across all experimental conditions to control for solvent effects.

Bright field microscopy

Bright field images were acquired using a Nikon Eclipse TE300 at × 200 magnification and analyzed using Northern Exposure software (Empix Imaging).