Examination of a second node of translational control in the unfolded protein response

Summary

The unfolded protein response (UPR) is a largely cytoprotective signaling cascade that acts to re-establish homeostasis of the endoplasmic reticulum (ER) under conditions of stress by inducing an early and transient block in general protein synthesis and by increasing the folding and degradative capacity of the cell through an extensive transcriptional program. It is well established that the mechanism for the early translational attenuation during ER stress occurs through phosphorylation of eukaryotic initiation factor 2 α (eIF2α) by activated PERK. Our data demonstrate that when eIF2α is dephosphorylated translation is not fully restored to pre-stressed levels. We found that this correlates with reduced mTOR activity and as a result decreases phosphorylation of 4E-BP1, which negatively regulates assembly of the eIF4F complex and cap-dependent translation. The decrease in mTOR activity and 4E-BP1 phosphorylation is associated with activation of AMP kinase, a negative regulator of mTOR, and in the case of some stress conditions, downregulation of signaling through key components of the PI3K pathway. Furthermore, we show that there is a subset of mRNAs that does not recover from UPR-induced translational repression, including those whose translation is particularly sensitive to loss of eIF4F, such as cyclin D1, Bcl-2 and MMP-9. Together these data implicate reduced mTOR activity and 4E-BP1 hypophosphorylation as a second, more restricted mechanism of translational control occurring somewhat later in the UPR.

Introduction

The majority of all secreted and membrane-bound proteins of higher eukaryotic cells are co-translationally translocated into the ER lumen, where they acquire their native conformation with assistance from ER-resident folding factors, chaperones and enzymes. Unfavorable conditions (sub-optimal pH, low ATP levels, and hypoxia), high client protein load, or expression of mutant proteins can perturb ER function and result in the accumulation of unfolded proteins in the ER lumen. In response to this stress, the unfolded protein response (UPR) is triggered; a largely cytoprotective signaling pathway that attempts to ease the effects of ER stress and restore homeostasis to this organelle through both transcriptional and translational programs (Ron and Walter, 2007). If these protective measures are unsuccessful, apoptotic pathways are activated to destroy the compromised cells.

The UPR transcriptional program serves in part to upregulate resident molecular chaperones and folding enzymes, thereby preventing the aggregation of unfolded proteins and bolstering ER folding capacity, while the translational program acts to ease the load of new client proteins entering the organelle by transiently inhibiting protein synthesis. Of note, the decrease in protein synthesis is not restricted to ER proteins. The loss of cytosolically translated cell cycle proteins causes cells to arrest in the G1 phase of cell cycle, ensuring that cells experiencing ER stress are not replicated (Brewer et al., 1999). The UPR is signaled through three ER-localized transmembrane proteins, inositol-requiring enzyme 1 (IRE-1), activating transcription factor 6 (ATF6), and protein kinase RNA (PKR)-like ER kinase (PERK), which collectively monitor folding conditions in the ER through their luminally oriented domains and signal the downstream response through their cytosolic domains (reviewed by Walter and Ron, 2011). While the cellular outcomes of ATF6 and IRE1 activation are largely transcriptional, the PERK arm of the UPR primarily exerts its effects through transient alterations in translation. In response to ER stress, PERK dimerizes, leading to trans-autophosphorylation (Liu et al., 2000) and phosphorylation of the α subunit of eukaryotic initiation factor 2 α (eIF2α) at Ser51 (Harding et al., 1999), resulting in a reduction in global protein translation.

However, certain mRNAs, such as that of ATF4, are preferentially translated under these conditions by virtue of the structure of their 5′ regions, which contain multiple small interfering ORFs that inhibit its translation under non-stress conditions (Harding et al., 2000a). The ATF4 transcriptional target GADD34 together with protein phosphatase 1 (PP1) acts to de-phosphorylate eIF2α, providing an elegant negative feedback loop and allowing general translation to resume (Novoa et al., 2001; Ma and Hendershot, 2003). The beneficial aspects of this PERK-mediated translational block are demonstrated by studies showing that cultured cells (Harding et al., 2000b) and endocrine and exocrine pancreas cells (Harding et al., 2001) lacking PERK are less likely to survive when subjected to ER stress. However, global translation must resume so that proteins essential for survival and cellular maintenance can be restored and that mRNAs upregulated as part of the transcriptional UPR response can be translated. Indeed, cells expressing a mutant GADD34 that lacks eIF2α phosphatase activity are unable to restore translation after PERK activation and exhibit decreased viability when compared to control cells in response to ER stress (Novoa et al., 2003). Thus, accrued experimental evidence suggests that translation must be balanced carefully to maximize cell viability under ER stress conditions (Walter and Ron, 2011).

Recent studies have focused on another translational control axis centered on mTOR. 4E-BP1, a target of mTOR and negative regulator of eIF4F, was found to be a target of the UPR-induced transcription factor ATF4 (Yamaguchi et al., 2008), and several other studies provide evidence that ER stress can impact various elements of the mTOR pathway (Ozcan et al., 2004; Salazar et al., 2009; Di Nardo et al., 2009; Qin et al., 2010), and that alterations in the mTOR pathway can induce ER stress (Ozcan et al., 2008; Qin et al., 2010) suggesting that translational control in the UPR is more complicated than previously indicated.

Regulation of the eIF4F complex formation by the 4E-BPs plays a major role in reducing protein synthesis under conditions of reduced growth factor signaling, nutrient availability, and ATP levels; largely via decreased signaling through the PI3K and mTOR pathways (Hay and Sonenberg, 2004). 4E-BP1 is a direct target of mTOR (Hay and Sonenberg, 2004; Proud, 2004), and as a result decreases in mTOR signaling lead to the accumulation of hypophosphorylated 4E-BP1, which in turn results in decreased formation of the eIF4F complex and reduced cap-dependent translation. Although eIF4F is important for all cap-dependent translation, a subset of mRNAs have been identified whose translation is particularly affected by a reduction in eIF4F complex formation (Hay and Sonenberg, 2004). The characteristics that cause the translation of these various mRNAs to be particularly dependent on eIF4F are not fully understood, but include the presence of long, highly structured 5′ untranslated regions (Pelletier and Sonenberg, 1985; Gingras et al., 1999), a terminal oligopyramidine tract immediately adjacent to the 5′ cap (Hsieh et al., 2012; Thoreen et al., 2012), or a regulatory element in the 3′UTR that marks certain mRNAs for eIF4E-facilitated nuclear export (Culjkovic et al., 2006). Of note, cyclin D1 transcripts are highly structured and are known to be dependent on mTOR signaling (De Benedetti and Graff, 2004). Thus, we designed experiments to more carefully examine translational control during an ER stress response and determine how reduced signaling through the mTOR pathway contributes to this.

Results

ER stress induces an early and transient decrease in global translation that is associated with eIF2α phosphorylation, however not all proteins recover equally

It is well established that ER-stress-induced phosphorylation of eIF2α results in a decrease in global protein translation, and that this block is subsequently relieved by the dephosphorylation of eIF2α by the ATF4 transcriptional target, GADD34 (Novoa et al., 2001; Ma and Hendershot, 2003). To further investigate ER-stress-induced translational changes associated with the UPR, NIH3T3 cells were pre-treated with the ER-stress inducing drug thapsigargin, which inhibits SERCA pumps in the ER membrane, rapidly depleting its calcium stores and broadly affecting protein folding in this organelle. At the indicated times after thapsigargin treatment, cells were pulse-labeled with [³⁵S]methionine and [³⁵S]cysteine for a short period of time to monitor translation. Cell lysates were directly examined on SDS-polyacrylamide gels, and as expected the decrease in protein synthesis at early time points appeared to be global (Fig. 1A). Similar results were obtained with tunicamycin, although the kinetics of translation inhibition was slower and the magnitude was not as great (data not shown). With both stresses, the inhibition of translation correlated with the phosphorylation of eIF2α. After 2 hours, eIF2α phosphorylation decreased (Fig. 1B) and protein translation began to resume (Fig. 1A). Intriguingly, closer inspection of the autoradiograph revealed that not all proteins appeared to recover equally. Although the majority of proteins were translated at levels comparable to pre-stressed cells following dephosphorylation of eIF2α (Fig. 1A, indicated with a circle), other proteins were translated at a higher level and correspond to the size of known ER chaperones that are targets of the response (Fig. 1A, marked with an asterisk). Translation of a subset of proteins did not appear to recover to pre-stress levels (Fig. 1A, marked with a diamond). At least one of the proteins that did not recover was the short-lived cyclin D1 protein, whose translation has been shown to be blocked by eIF2α-dependent translational repression (Brewer and Diehl, 2000; Stockwell et al., 2012). Indeed, in our experiments cyclin D1 protein levels began decreasing within 0.5 hours of thapsigargin treatment and were undetectable by 2 hours (Fig. 1B). However, even though eIF2α was dephosphorylated at later time points (Fig. 1B) and general protein translation had resumed, cyclin D1 protein levels did not recover within the time frame of this experiment (Fig. 1B). The sustained loss of cyclin D1 protein expression was not limited to a single cell line or method of induction of ER stress. HeLa cells, HEK-293T and NB1691 cells all showed decreases in cyclin D1 protein expression when they were examined after 16 hours of treatment with either thapsigargin or tunicamycin (Fig. 1C).

Fig. 1.

Not all proteins recover equally from the early and transient block in protein translation induced by ER stress. (A) NIH3T3 cells were treated with thapsigargin (Tg) for the indicated times. Cells were pulse-labeled with [³⁵S]methionine and [³⁵S]cysteine during the last 5 minutes of treatment. Lysates were prepared, equalized by protein concentration and analyzed by SDS-PAGE. Coomassie Blue staining served as a loading control (upper panel), and ³⁵S signal was detected by audioradiography (lower panel). Examples of stress-induced changes in levels of various proteins are indicated with symbols: diamonds indicate several proteins whose translation rate is less than pre-stressed levels at later time points, whereas asterisks indicate several proteins that are synthesized above pre-stressed levels, and circles indicate several proteins whose levels appear to return to pre-stressed levels. (B) NB1691 cells were treated with Tg for the indicated times, lysates were prepared and analyzed by SDS-PAGE followed by immunoblotting with phospho-eIF2α (Ser51), cyclin D1 and β-actin, as indicated. (C) HeLa and HEK-293T cells were treated with Tg and Tunicamycin (Tm) for 16 hours, lysates were prepared and subjected to SDS-PAGE followed by immunoblotting with cyclin D1 and β-actin antibodies.

Incomplete recovery from translational repression under chronic ER stress is associated with 4E-BP1 hypophosphorylation

To further explore translational control during ER stress, we measured global translation rates in NB1691 cells over a longer time course (Fig. 2A). Once again, we observed an early and dramatic decrease in protein synthesis when thapsigargin was used. However, even after 24 hours of treatment, total TCA-precipitable counts only recovered to ∼80% of pre-stressed levels (Fig. 2A), even though there was an increase in the synthesis of molecular chaperones (data not shown) and UPR target genes such as CHOP. We investigated the phosphorylation status of both eIF2α and 4E-BP1 under these longer periods of ER stress. NB1691 cells were treated with thapsigargin for the indicated times, and cell lysates were subjected to western blotting analysis (Fig. 2B). As expected, ER stress caused an initial increase in the phosphorylation of eIF2α, which decreased to basal levels after 2 hours. When the same membrane was blotted for 4E-BP1, we found that the majority of 4E-BP1 appeared to be hyperphosphorylated (the γ-form) in the absence of ER stress, a state which allows translation to proceed (Fig. 2B). As the thapsigargin treatment time increased, there was a modest but progressive shift to the hypophosphorylated α-form of 4E-BP1. To extend our findings to other cell types and different methods of ER stress induction, we also examined phosphorylation of 4E-BP1 protein in HeLa and HEK-293T cells treated with either thapsigargin or tunicamycin for 16 hours. In both cell lines, the appearance of the hypophosphorylated form of 4E-BP1 could be readily observed, although the relative amounts varied somewhat (Fig. 2C). Of note, we observed an increase in total 4E-BP1 levels in HEK-293T cells (Fig. 2C) and mouse embryonic fibroblasts (data not shown) similar to previous studies in pancreatic cells (Yamaguchi et al., 2008) but not in HeLa or NB1691 cells, suggesting that the upregulation of this protein during UPR activation may be cell-type specific.

Fig. 2.

Incomplete translational recovery with prolonged ER stress is associated with 4E-BP1 hypophosphorylation. (A) NB1691 cells were treated with thapsigargin (Tg) for the time periods indicated and cells were incubated with ³⁵S-labeled methionine and cysteine during the last 5 minutes of treatment. Lysates were prepared and equal amounts of total protein was spotted and dried onto filter paper. The radioactive signal was quantified using a scintillation counter and expressed in graphical form as the percentage of protein synthesis remaining. The short heavy bar indicates the time of maximal eIF2α phosphorylation, whereas the thin dotted line represents decreased translation rate after eIF2α phosphorylation returns to pre-stressed levels. (B) NB1691 cells were treated with Tg for the indicated times. Lysates were analyzed by SDS-PAGE followed by immunoblotting using antibodies directed against 4E-BP1, phospho-eIF2α (Ser51) and β-actin. The phospho-forms of 4E-BP1 are indicated with Greek letters with γ representing the most phosphorylated form and α the least. (C) HeLa and HEK-293T cells were treated with Tg and tunicamycin (Tm) for 16 hours. Lysates were prepared and subjected to SDS-PAGE followed by immunoblotting with 4E-BP1 and β-actin antibodies. The phosphorylated forms of 4E-BP1 are indicated.

ER-stress-induced 4E-BP1 hypophosphorylation results in decreased eIF4F formation and the loss of translation of eIF4F target mRNAs

To assess whether the changes in the phosphorylation status of 4E-BP1 observed under ER stress conditions was sufficient to lead to the binding of 4E-BP1 to eIF4E and a concomitant loss of eIF4G on capped mRNA, we used 7M-GTP beads, as a mimetic of the 5′ cap structure, in a pull down assay. The binding of 4E-BP1 and eIF4G was determined, and as both bind to eIF4E, blotting for this protein served as a control. NB1691 cells were treated with either tunicamycin or thapsigargin for 16 and 24 hours or left untreated (Fig. 3A). Lysates were incubated with 7M-GTP beads and bound proteins were detected by western blotting for eIF4E, eIF4G and 4E-BP1 (Fig. 3A). Although the amount of eIF4E bound to the 7M-GTP beads was comparable between all groups, there was an increase in 4E-BP1 bound to the 7M-GTP beads with either ER-stress-inducing drug (consistent with decreased cap-dependent translation), which corresponded to a reciprocal decrease in the binding of eIF4G (Fig. 3A). These results indicate that the ER-stress-induced reduction in 4E-BP1 phosphorylation is likely to be functionally significant.

Fig. 3.

ER-stress-induced 4E-BP1 hypophosphorylation results in decreased eIF4F formation and the loss of translation of known eIF4F target mRNAs. (A) NB1691 cells were either left untreated (control conditions; NT) or treated with Tg or Tm for the indicated times. Lysates were prepared, total protein for each sample was equalized and samples were incubated with 7M-GTP–Sepharose beads. Bound protein was eluted and subjected to SDS-PAGE followed by immunoblotting using antibodies directed against eIF4G, 4E-BP1 and eIF4E. (B) NB1691 cells were treated with Tg for the indicated times, lysates were prepared and subjected to SDS-PAGE followed by immunoblotting with antibodies directed against cyclin D1, Bcl2, MMP-9, and for each blot, β-actin as a loading control. (C) NB1691 cells were treated with Tg or Tm for either 16 hours (open bars) or 24 hours (shaded bars) as indicated. Total RNA was extracted and qRT-PCR was performed using specific primers to detect levels of mRNA of cyclin D1 (CCND1), MMP-9 and Bcl-2. Experiments were performed in triplicate and results presented as fold change compared to untreated cells. Values are means ± s.d.

We next examined the steady state levels of several proteins known to be sensitive to eIF4F loss, over a time course of thapsigargin treatment spanning 24 hours. NB1691 cells were treated with this ER-stress-inducing agent for the indicated time points, and subjected to western blot analyses. As expected, the short-lived cyclin D1 protein levels decreased at the earliest time points examined (Fig. 3B), which is consistent with its loss being due to the eIF2α-mediated translational block (Brewer and Diehl, 2000) and remained undetectable even after global translation had been restored. We next examined the expression of Bcl-2, which is normally a rather long-lived protein (Merino et al., 1994) but is known to be sensitive to eIF4 levels (reviewed by Graff and Zimmer, 2003). Somewhat unexpectedly, Bcl-2 levels decreased in NB1691 cells as early as 2 hours of thapsigargin treatment and remained lower throughout the time course. Although Bcl-2 has been shown to have a half-life of between 10 and 20 hours depending on the cell type (Beck et al., 2002; Gao and Dou, 2000; Brunelle et al., 2009), there are reports of certain stresses such as glutathione depletion (Celli et al., 1998) or alterations in the level of the pro-apoptotic protein Bim (Jorgensen et al., 2007) significantly decreasing its stability. We also examined MMP-9 expression after ER stress, which is also known to be dependent on eIF4F levels (De Benedetti and Graff, 2004). Unlike the two previous proteins examined, it is a secreted protein, so its loss from the cell is a combination of changes in the rate of both translation and transport through the cell. Correspondingly, levels of this protein did not decrease noticeably until after 16 hours of thapsigargin treatment (Fig. 3B). Although we tried several antibodies, we were unable to find one that worked for immunoprecipitation assays to more directly monitor synthesis (data not shown). These more eIF4F-dependent proteins showed a dramatic reduction in expression, but there was a substantial induction of the UPR target CHOP (Fig. 3B), indicating that only a subset of targets are affected, in keeping with the data in Fig. 1A. To ensure that the reduced expression of these proteins was the result of decreased biosynthesis not reduced levels of transcripts available for translation we performed real-time PCR analyses (Fig. 3C). Indeed, within the time frame of our experiments we did not observe any major changes in mRNA transcript levels for any of the proteins examined in Fig. 3B. These data indicate that the UPR can induce not only the global translational changes that are mediated by the phosphorylation status of eIF2α, but also more subtle translational changes that affect at least some eIF4F-sensitive mRNAs.

ER stress treatment results in a decrease in mTOR signaling and growth factor receptor maturation

One of the best-characterized 4E-BP1 kinases is mammalian Target of rapamycin (mTOR), and several groups have reported ER-stress-induced perturbations in signaling upstream of mTOR (Ozcan et al., 2004; Salazar et al., 2009; Di Nardo et al., 2009; Qin et al., 2010). We therefore examined mTOR activation status in response to tunicamycin- and thapsigargin-induced ER stress. Phosphorylation of mTOR at Ser2448, indicative of activation, was very modestly reduced after 6 hours of thapsigargin treatment, and was further reduced at 16 and 24 hours with both ER-stress-inducing drugs (Fig. 4A). The different kinetics of the phosphorylation changes induced by these two pharmacological agents is likely to reflect differences in their mechanisms of action, which result in thapsigargin inducing an earlier UPR than tunicamycin.

Fig. 4.

Prolonged ER stress is associated with a decrease in mTOR signaling. (A) NB1691 cells were left untreated or treated with Tg or Tm for the indicated times. Cell lysates were prepared and normalized samples were analyzed by SDS-PAGE and immunoblotted using antibodies specific for phospho-mTOR (Ser2448), mTOR, and β-actin. (B) NB1691 cells were treated as in A and with rapamycin as a control for mTOR-dependent effects on 4E-BP1. Lysates were analyzed by SDS-PAGE followed by immunoblotting using antibodies directed against 4E-BP1 and β-actin. The phosphorylated forms of 4E-BP1 are indicated with Greek letters with γ representing the most phosphorylated form and α the least. (C) Rh30 cells were left untreated or incubated with rapamycin for 4 hours or tunicamycin for 16 hours and lysates were analyzed as in (B). (D). Rh30 cells were treated with varying concentrations of low glucose (mM), tunicamycin or rapamycin for the indicated times. Lysates were prepared, total protein for each sample was equalized and samples were incubated with 7M-GTP–Sepharose beads. Bound protein was eluted and subjected to SDS-PAGE followed by immunoblotting with antiserum directed against 4E-BP-1. A portion of the equalized samples were directly analyzed for evidence of UPR activation using the CHOP antiserum. β-actin serves as a control for loading.

Next, we monitored effects of UPR activation on the phosphorylation of 4E-BP1, an mTOR target and included treatment with rapamycin, a well-characterized mTOR inhibitor, for comparison. We observed a shift from the hyperphosphorylated to less modified forms of 4E-BP1 after longer times of treatment with both of these UPR inducers, which was similar to what was observed with rapamycin or even slightly greater in the case of tunicamycin (Fig. 4B). As the effects on 4E-BP1 even with rapamycin treatment were not particularly dramatic in this cell line, we examined a second line. Rh30 cells were treated with rapamycin and tunicamycin, and effects on 4E-BP1 isoforms were monitored. In this cell line we observed a more dramatic effect on 4E-BP1 phosphorylation with rapamycin treatment and again observed a similar decrease in the tunicamycin-treated cells (Fig. 4C). We also performed the 7M-GTP bead binding assay as a functional measure of cap binding and included varying concentrations of low glucose, as a more slow acting but also more physiological inducer of the UPR (Fig. 4D). Concentrations of glucose as high as 6.25 mM were sufficient to both induce the UPR and lead to hypophosphorylation to levels comparable with rapamycin, whereas neither thapsigargin or tunicamycin treatment induced hypophosphorylation of 4E-BP1 at the shorter time point used, even though it was sufficient to activate the UPR, in keeping with data obtained with the NB1691 cells (Fig. 4A).

To determine how the UPR was affecting mTOR activation, we first examined the PI3K pathway. We began with the receptor tyrosine kinases (RTK), which are the most up-stream components of this signaling pathway and are synthesized in the ER and transported to the cell surface after proper maturation. We reasoned that it was conceivable that ER stress could interfere with their maturation and thus result in loss of signaling through these receptors. NB1691 cells were treated with tunicamycin or thapsigargin for the indicated times and two well-characterized RTK proteins that signal through this pathway were monitored. We found that treatment with the glycosylation inhibitor tunicamycin resulted in the appearance of an unglycosylated form and a decrease in mature levels of the beta subunit of both the insulin receptor (Irβ; Fig. 5A) and the insulin-like growth factor 1 receptor (IGF1Rβ; Fig. 5B) as early as 16 hours of treatment, which was even more dramatic after 24 hours. When cells were treated with thapsigargin, we found that the maturation IRβ was largely unaffected at all treatment times. Very modest decreases in mature levels of IGF1Rβ were observed after 16 hours, which were greater at 24 hours, but were still less dramatic than observed with tunicamycin. As RTKs are glycoproteins, it is probable that the loss of glycosylation that occurs with tunicamycin treatment has greater effects on their maturation than the loss of calcium from the ER that results with thapsigargin treatment.

Fig. 5.

Pharmaceutical agents used to activate ER stress can adversely affect growth factor receptor maturation. NB1691 cells were left untreated or treated with Tg or Tm for the indicated times. Cell lysates were prepared and normalized samples were analyzed by SDS-PAGE and immunoblotted using antibodies specific for (A) insulin receptor β, (B) IGF1 receptor β and β-actin (A,B). Various states of receptor maturation are indicated.

ER stress activators do not uniformly inhibit PI3K–Akt pathway signaling in NB1691 cells

To more globally monitor growth factor receptor signaling, we examined the activation of Akt, which is downstream of a large number of the RTK, and because changes in Akt activity have been shown to occur by a variety of mechanisms (Ozcan et al., 2004; Salazar et al., 2009). Western blots were performed using antibodies specific for phosphorylated Akt (Thr473), a residue known to be important for full activation of Akt and its downstream signaling (Sarbassov et al., 2005). In keeping with two previous studies (Ozcan et al., 2004; Qin et al., 2010), we found that tunicamycin induced an observable decrease in phosphorylated Akt after 16 hours of treatment and much greater decrease after 24 hours of treatment (Fig. 6A), suggesting that the maturation of many RTKs is likely to be affected by tunicamycin, in keeping with the fact that they are glycoproteins. Thapsigargin had little effect at any time point on Akt activation in these cells (Fig. 6A), which indicates that similar to IRβ (Fig. 5A) changes in ER calcium levels do not adversely affect the maturation of the RTKs. Activated Akt phosphorylates TSC2, a negative regulator of mTOR at Thr1462 (Manning et al., 2002), resulting in an inhibition of the suppressive affect of TSC2 on mTOR (Inoki et al., 2002). Therefore, we next examined the phosphorylation of TSC2 at the Akt-specific residue Thr1462. We found a very modest decrease in its phosphorylation, but only in response to tunicamycin treatment, which is in keeping with our finding that tunicamycin but not thapsigargin diminished Akt activation (Fig. 6A).

Fig. 6.

Prolonged ER stress treatment is associated with a decrease in PI3K–Akt pathway signaling and an increase in AMPK signaling. NB1691 cells were left untreated or treated with Tg or Tm for the indicated times. Lysates were prepared and analyzed by SDS-PAGE. Immunoblotting was performed using antibodies raised against (A) phospho-Akt (Thr473), pan-Akt, and phospho-TSC2 (Thr1462) or (B) phospho-AMPK (Thr172), pan AMPK and phospho-TSC2 (Ser1387), with β-actin serving as a loading control.

Our finding that thapsigargin was equally effective as tunicamycin in reducing 4E-BP1 phosphorylation led us to examine other proteins that impinge on the mTOR/4E-BP1 axis. mTOR activity can also be negatively regulated by phosphorylated AMP-activated protein kinase (AMPK) (Hardie et al., 2012), which was recently shown to be activated in response to some ER stress conditions (Pereira et al., 2010). To examine the possible involvement of this kinase in the reduced mTOR signaling observed with longer treatments with ER stressors, NB1691 cells were treated with tunicamycin and thapsigargin for the indicated times and the proteins were assessed via western blot analyses. Thapsigargin treatment resulted in an increase in AMPK phosphorylation at Thr172 within 6 hours of treatment, and was sustained through 16 and 24 hours (Fig. 6B). In keeping with the somewhat slower kinetics of UPR activation by tunicamycin, there was no observable increase in Thr172 phosphorylation at 6 hours, but after 16 and 24 hours there was a substantial increase in Thr172 phosphorylation. To further explore the significance of this activation to mTOR signaling, we examined the phosphorylation of the AMPK target TSC2 at Ser1387, which enhances the inhibitory effects of TSC on mTOR (Huang and Manning, 2008; Inoki et al., 2003). After 24 hours of treatment with either stressor, phosphorylation of TSC2 at this residue was increased above control levels (Fig. 6B), indicating that the observed ER-stress-induced AMPK phosphorylation contributes to inhibitory TSC signaling and thus probably to the decrease in mTOR phosphorylation as shown in Fig. 4A.

Discussion

The ability of cells to modify translation in response to both intracellular cues and extracellular conditions is critical to cell survival, especially under conditions of stress. It is well established that mammalian cells experiencing ER stress can induce a transient and global block in translation via phosphorylation of eIF2α, which is achieved by the activation of the ER transmembrane protein PERK (Harding et al., 1999). We have examined a second mechanism by which cells experiencing ER stress can fine-tune their translational programs. We found that translation rates after dephosphorylation of eIF2α only returned to ∼80% of that occurring before thapsigargin-induced stress. Based on the inspection of proteins synthesized after a short metabolic pulse-labeling, this did not appear to represent a general decrease in the translational capacity of the post-stressed cells, but instead seemed to be due to changes in the expression of a specific subset of proteins. The continued translational repression of this group of proteins correlated with increased hypophosphorylation of 4E-BP1 leading to its increased binding to eIF4E, which would interfere with the formation of the eIF4F complex at the mRNA cap. Indeed, we found that the expression of several proteins known to be sensitive to eIF4F availability were decreased after longer periods of ER stress, including cyclin D1, Bcl-2 and MMP-9. These data are in keeping with a previous study demonstrating that deletion of 4E-BP1 is required to allow translation to be fully restored after ER stress (Yamaguchi et al., 2008).

There is a growing body of literature describing interactions or crosstalk between the mTOR pathway and the UPR. Several studies have demonstrated a decrease in Akt signaling in response to ER stress leading to reduced mTOR activity. In one case the UPR was induced in human glioma cells by Δ⁹-tetrahydrocannabinol (THC) leading to a CHOP-dependent induction of the tribbles homologue 3,TRB3, which inhibited Akt and as a result mTOR (Salazar et al., 2009). Alternatively, another study using Fao liver cells found that Ire1-dependent activation of JNK led to phosphorylation of insulin receptor substrate 1, IRS-1, which reduced insulin receptor signaling and activation of Akt (Ozcan et al., 2004). Of note, the decrease in Akt activity in this and another study (Qin et al., 2010) occurred with both tunicamycin and thapsigargin, suggesting that our failure to observe decreases in Akt activation with thapsigargin might be somewhat cell-type specific. Our study does suggest that the proper maturation of RTKs, which are glycosylated on multiple residues, are more adversely affected by some forms of ER stress, such as tunicamycin and probably the more physiological UPR activator, low glucose. These data provide yet another means of reduced Akt activity during UPR activation and furthermore highlight the often overlooked fact that the agents used to induce ER stress have cellular effect that lie outside the signaling pathways that comprise the UPR.

In addition to reduced signaling through Akt, activation of the Tsc1/2 complex is another mechanism of inhibiting mTOR. A recent study demonstrated LPS-induced differentiation of mouse splenic B cells led to a reduction in mTOR activity, which could be inhibited by genetic ablation of TSC1 (Goldfinger et al., 2011). Importantly, terminal plasma cell differentiation relies on the activation of a partial UPR (Iwakoshi et al., 2003; Shaffer et al., 2004), arguing for a link between the UPR and TSC1/2 regulation of mTOR. Under conditions of decreased availability of energy, AMP kinase (AMPK) is activated (Hardie et al., 2012), leading to phosphorylation of TSC2 on Ser1387. This modification inhibits the Rheb protein, which is another positive regulator of mTOR (Laplante and Sabatini, 2012). We found that induction of ER stress with either tunicamycin or thapsigargin resulted in the phosphorylation of both AMPK itself and the residue of TSC2 known to be phosphorylated by AMPK. However, another study saw no evidence of AMPK activation by tunicamycin in mouse embryonic fibroblasts (Qin et al., 2010), again suggesting that the mechanisms inhibiting mTOR activity during ER stress might vary by both the stressor used and the type of cells examined. In an attempt to more directly demonstrate that AMPK was the crucial culprit, we used the AMPK inhibitor Compound C. However, the combination of this inhibitor with either tunicamycin or thapsigargin proved to be toxic to our cells (data not shown) preventing us from making this point. Interestingly, recent data show reciprocally that mTOR activity can promote ER stress. Two separate studies (Ozcan et al., 2008; Qin et al., 2010), revealed that hyperactivation of mTOR due to loss of Tsc2 led to UPR activation, probably as a result of increased protein synthesis in the ER. In both cases, UPR activation was accompanied by decreased Akt activity, establishing this aspect of the UPR as a feed back loop to control mTOR activity.

The reduced activity of mTOR during ER stress leads to hypophosphorylation of 4E-BP1, which we and others (Yamaguchi et al., 2008) show reduces the ability of eIF4G to bind to ⁷mGTP beads as a mimetic of the binding of the eIF4F complex to the mRNA cap. Although not examined here, a number of proteins such as c-myc, ornithine decarboxylase and fibroblast growth factor 2 are known to be particularly dependent on eIF4F levels and can be loosely categorized as ‘pro-growth’ proteins (Graff and Zimmer, 2003; Mamane et al., 2004) and are not translated when mTOR is inhibited with rapamycin in a 4E-BP1-dependent manner (Dowling et al., 2010). Therefore, it is plausible to suggest that the ER-stress-induced decrease in 4E-BP1 phosphorylation could be a mechanism by which cells limit cell division and growth during stressful conditions, while at the same time allowing them to continue to produce proteins that maintain cellular fitness. In this way, cellular growth and division would be limited while still allowing for the translation of UPR transcriptional targets and those mRNAs required for cellular maintenance. Indeed, the well-characterized arrest of cells experiencing ER stress in the G1 phase of cell cycle was shown to be due to loss of cyclin D1 protein, even though D1 transcript levels were not altered (Brewer et al., 1999). This was consequently linked to PERK-dependent eIF2α phosphorylation (Brewer and Diehl, 2000), which rapidly led to the depletion of this very short-lived protein. The data presented here are consistent with the PERK/eIF2α axis playing an early role in suppressing cyclin D1 biosynthesis, but further reveal that once most translation has resumed, D1 protein remains very low, even though the transcripts are still present, which is probably regulated by the eIF4 axis of translational control.

By contrast, a number of pro-survival proteins are also among those that are significantly dependent on eIF4F levels, including Bcl2 and survivin (Konicek et al., 2008). The balance between cytoprotective and cytodestructive aspects of the UPR is quite complex and the timing varies by cell type and cellular conditions (Tabas and Ron, 2011). Indeed, decreases in Bcl2 mRNA levels can occur downstream of the CHOP transcription factor (McCullough et al., 2001). However, within the time frame of our experiments Bcl2 transcript levels were not diminished and in fact were very slightly increased by tunicamycin treatment. It is well established that many of these proteins, including MMP-9, are crucial for tumor growth and/or metastasis, and a growing number of studies have documented UPR activation in a variety of tumor types, leading investigators to suggest it could be a target for therapeutic intervention (reviewed by Wang and Kaufman, 2012; Li et al., 2011; Healy et al., 2009). This would suggest that the loss of these proteins in tumors experiencing ER stress would be counterproductive to tumor growth and survival. Thus it is noteworthy that eIF4E, the rate-limiting step in cap-dependent translation, and the target of hypophosphorylated 4E-BP1, is often upregulated in tumors and is considered to be essential for their survival (De Benedetti and Graff, 2004).

Here, we have examined a second node of translational control in the UPR, centered around 4E-BP1, which exerts its effects after eIF2α-mediated translation inhibition. In our model, we observed that translation of eIF4F-sensitive proteins do not recover from the eIF2α-induced translational block upon eIF2α dephosphorylation. Instead, translation of these proteins remains repressed, which is probably due to the UPR-induced negative regulation of the mTOR pathway and the resulting hypophosphorylation of the eIF4F repressor, 4E-BP1. This repression appears to be achieved in our cells, at least in part, through UPR-induced AMPK activation, which is a negative regulator of the mTOR pathway, and reduction in signaling through the PI3K–AKT pathway. Unlike the more global inhibition of protein synthesis mediated by eIF2α, the translational targets of this secondary node appear to be restricted to a sub-group of proteins that are potent activators of growth and survival.

Materials and Methods

Cell culture

NIH3T3, HeLa and HEK-293T cells were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 2 mM glutamine and 1% antibiotic-antimycotic at 37°C in a 5% CO₂ incubator. NB1691 neuroblastoma and Rh30 rhabdomyosarcoma human cell lines were cultured in RPMI 1640 supplemented with 10% fetal bovine serum and 2 mM glutamine. Cells were plated and left untreated (control) or treated with thapsigargin (2 µM; Sigma-Aldrich, St. Louis, MO, USA), tunicamycin (2.5 µg/ml; Sigma-Aldrich, St. Louis, MO, USA) for the indicated times, or rapamycin (100 nM, 4 hours).

Metabolic labeling

Following the indicated experimental treatments, cells were washed with phosphate-buffered saline and then pulsed for 5 minutes in methionine- and cysteine-free DMEM labeling medium containing 10% dialyzed FBS and 100 µCi [³⁵S]methionine and [³⁵S]cysteine (³⁵S-TransLabel; MP Biomedicals, Santa Ana, CA, USA). Cells were lysed in CHAPS lysing buffer [pH 7.4 (40 mM HEPES, 120 mM NaCl, 1 mM EDTA, 10 mM sodium pyrophosphate, 10 mM sodium β-glycerophosphate, 50 mM NaF, 1.5 mM Na₃VO₄, 0.3% CHAPS)], sonicated and centrifuged at 12,000 r.p.m. for 10 minutes at 4°C to remove nuclei and cellular debris. Total protein was quantified using the Bradford assay (Bio-Rad, Hercules, CA, USA), and samples were equalized for total protein. For experiments in which the translation of individual proteins was to be studied, lysates were analyzed by SDS-PAGE and radiographic signals were visualized by incubating gel with Amplify (GE Healthcare, Fairfield, CT, USA) and exposing gel to X-ray film. For experiments in which total incorporation of radiolabeled amino acids was quantified, 2 µl of each lysate was spotted onto filter paper squares, allowed to dry, and then precipitated by boiling in 10% TCA for 10 minutes, followed by water, ethanol and acetone rinses. Dry filters were then subjected to scintillation counting. Counts were normalized to protein quantity and counts calculated as a percentage of control.

7methyl-GTP bead pull down

Following the indicated experimental treatments, cells were lysed in CHAPS lysing buffer, supplemented with complete protease inhibitor tablet (Roche, Pleasanton, CA, USA) and PhoSTOP phosphatase inhibitor tablet (Roche, Pleasanton, CA, USA). Lysates were sonicated, clarified by centrifugation at 14,000 rpm for 10 minutes at 4°C, and protein content was quantified. The 7methyl-GTP bead binding assay was performed as previously described (Inoki et al., 2003). Briefly, lysates were diluted in CHAPS buffer to a concentration of 20 µg total protein in a total volume of 170 µl. 30 µl of a 33% slurry of 7methyl-GTP-conjugated Sepharose beads (GE Healthcare, Fairfield, CT, USA) was added to the diluted lysates. The mixture was rotated for 2 hours at 4°C. Following this incubation, Sepharose bead pellets were washed three times with 500 µl of washing buffer [pH 7.5 (50 mM HEPES, 150 mM NaCl)]. Bound proteins were eluted with 20 µl Laemmli loading buffer and boiled, after which samples were analyzed by SDS-PAGE and western blotting.

Western blotting

For whole-cell western blotting, lysates were diluted to achieve equal protein concentrations and subjected to SDS-PAGE and western blotting. Antibodies used to detect specific proteins for western blots were as follows: phospho-eIF2α (S51), Akt (total), phospho-Akt (S473), AMPKα (total), phospho-AMPK (T172), Bcl-2, eIF4E, eIF4G, 4E-BP1 (total), IGF1 receptor β, insulin receptor β, mTOR (total), phospho-mTOR (S2448), phospho-TSC2 (S1387), phospho-TSC2 (T1462), all from Cell Signaling Technology (Danvers, MA, USA); β-actin antibody from Sigma-Aldrich (St Louis, MO, USA); MMP antibody from Millipore (Billerica, MA, USA); and cyclin D1 was from Santa Cruz Biotechnology (Dallas, Texas, USA). Blots were incubated with the appropriate HRP-conjugated secondary antibody, and proteins were visualized using the Pierce enhanced chemiluminescent substrate (Thermo Scientific, Waltham, MA, USA).

mRNA quantification by qRT-PCR

Total RNA was extracted using the RNeasy Qiagen mini-prep kit according to the manufacturer's protocol. cDNA was produced using 1 µg total RNA and reverse transcriptase reactions were performed using a high capacity cDNA reverse transcription kit (Applied Biosystems, Carlsbad, CA, USA). Amplification of the indicated genes was carried out using Power SYBR Green PCR Master Mix (Applied Biosystems, Carlsbad, CA, USA) with specific primers (Bcl2, 5′-acggggtgaactgggggagga-3′, 5′-tccacaaaggcatcccagcctc-3′; MMP9 5′-agccgggacgcagacatcgt-3′, 5′-ttggaaccacgacgcccttgc-3′; cyclin D1 5′-caagtgtgacccggactgcctc-3′, 5′-cgccctcagatgtccacgtcc-3′) and measured continuously using an ABI 7900 HTI Detection System. The value for untreated cells was set to 1, and the value for the various treatments was calculated as a fraction of this number.