A Selective Inhibitor of eIF2α Dephosphorylation Protects Cells from ER Stress

See allHide authors and affiliations

Science  11 Feb 2005:
Vol. 307, Issue 5711, pp. 935-939
DOI: 10.1126/science.1101902


Most protein phosphatases have little intrinsic substrate specificity, making selective pharmacological inhibition of specific dephosphorylation reactions a challenging problem. In a screen for small molecules that protect cells from endoplasmic reticulum (ER) stress, we identified salubrinal, a selective inhibitor of cellular complexes that dephosphorylate eukaryotic translation initiation factor 2 subunit α (eIF2α). Salubrinal also blocks eIF2α dephosphorylation mediated by a herpes simplex virus protein and inhibits viral replication. These results suggest that selective chemical inhibitors of eIF2α dephosphorylation may be useful in diseases involving ER stress or viral infection. More broadly, salubrinal demonstrates the feasibility of selective pharmacological targeting of cellular dephosphorylation events.

All eukaryotes respond to ER stress through a set of pathways known as the unfolded protein response (UPR) (1, 2). The UPR is essential for cellular homeostasis, and its dysregulation has been implicated in many important pathologies, including diabetes, Alzheimer's disease, and viral infection (3). However, several aspects of the mammalian UPR remain obscure. In particular, little is known about the control of the mammalian apoptotic pathways activated by excessive, uncorrected ER stress (4) or how these pathways might be manipulated for therapeutic benefit. We used a chemical biology approach to study apoptosis induced by ER stress in mammalian cells.

In screening ∼19,000 chemicals for compounds that protect the rat pheochromocytoma cell line PC12 from ER stress-induced apoptosis, we identified a small molecule we termed salubrinal (sal) (Fig. 1A). Sal inhibited ER stress-mediated apoptosis induced by the protein glycosylation inhibitor tunicamycin (Tm) in a dose-dependent manner, with a median effective concentration (EC50) ∼ 15 μM (Fig. 1B). Cytoprotection by sal exceeded that by zVAD.fmk, a pan-inhibitor of the caspase family of apoptotic proteases (Fig. 1B). Sal also suppressed Tm-induced DNA fragmentation (Fig. 1C), the processing of caspase-7, a caspase activated by ER stress (Fig. 1D) (5), and the activity of caspase-3- and caspase-7-type enzymes (fig. S1A), confirming its anti-apoptotic effect. Cytoprotection was not specific to Tm, because sal protected cells from brefeldin A, which causes ER stress by blocking ER-to-Golgi vesicle transport (fig. S1B). However, sal is not a general apoptosis inhibitor, because it did not protect against apoptotic stimuli unrelated to ER stress (fig. S1, C and D). Importantly, sal was nontoxic at concentrations required for maximal cytoprotection (TD50 > 100 μM) (Fig. 1, B and C), making it a useful tool for studying ER stress.

Fig. 1.

Sal protects cells against ER stress-induced apoptosis. (A) Sal. (B) Dose-dependent protection by sal of PC12 cells treated with Tm and concentrations (conc.) of sal as indicated for 37 hours, assessed by cellular adenosine triphosphate (ATP) content. 100 μM zVAD.fmk serves as a positive control. Error bars represent standard deviation here and throughout unless otherwise indicated. (C) Sal treatment reduced apoptosis in PC12 cells treated with Tm for 36 hours, assessed by propidium iodide staining and fluorescence-activated cell sorting of cells with subdiploid (< 2N) DNA content. (D) Sal treatment reduced the Tm-induced processing of caspase-7 from its zymogen (C7) to active (arrow) form in PC12 cells treated for 36 hours, assessed by immunoblot. Positions of molecular mass markers (in kD) are indicated at left.

To determine how sal protects cells from ER stress, we examined its effects on known components of the UPR. Unlike Tm, sal did not up-regulate the canonical UPR targets Xbp-1, Grp78/Bip, and Grp94 (68) (fig. S2A), indicating that sal does not cause ER stress or activate the transcription-dependent branch of the UPR (2). A second branch of the UPR involves phosphorylation of eIF2α on Ser51 by the ER-localized kinase PERK (9, 10). Phosphorylated eIF2α mediates both a transient decrease in global translation and the translational up-regulation of selected stress-induced mRNAs (11). Like Tm, sal induced rapid and robust eIF2α phosphorylation (Fig. 2A and fig. S2, B to D) and its downstream effects, including down-regulation of cyclin D1 (12) and up-regulation of GADD34 (13) and CHOP (11), two proteins whose expression is induced by eIF2α phosphorylation (fig. S2C). Thus, sal selectively engages the translational control branch of the UPR by inducing eIF2α phosphorylation without affecting the transcription-dependent component of the UPR.

Fig. 2.

Sal induces eIF2α phosphorylation. (A) Dose-dependent phosphorylation of eIF2α in PC12 cells after 36 hours treatment with sal, detected by immunoblot. (B) A derivative of sal, 2, that protected PC12 cells from Tm-induced apoptosis (assessed by ATP assay) also induced eIF2α phosphorylation in PC12 cells (assessed by immunoblot). All treatments were 36 hours. (C) A derivative of sal, 3, that failed to protect PC12 cells from Tm-induced apoptosis did not induce eIF2α phosphorylation in PC12 cells. Treatments and assays as in (B).

EIF2α phosphorylation is cytoprotective during ER stress, because cells are sensitized when this pathway is genetically ablated (14, 15) and protected when it is ectopically enforced (16, 17). We therefore asked whether the induction of eIF2α phosphorylation by sal was related to its anti-apoptotic effects. Sal-induced eIF2α phosphorylation was dose-dependent (Fig. 2A) and correlated well with its protection from Tm (Fig. 1B). Furthermore, chemical derivatives of sal that protected cells from Tm, such as 2, also induced eIF2α phosphorylation (Fig. 2B), whereas derivatives that failed to protect cells, such as 3, did not (Fig. 2C). Taken together with previous work establishing the protective function of eIF2α phosphorylation, these results suggest that eIF2α phosphorylation induction by sal accounts at least in part for its anti-apoptotic activity during ER stress.

To determine how sal induces eIF2α phosphorylation, we asked whether sal activates one of the four mammalian eIF2α kinases: PERK, GCN2, RNA-activated protein kinase (PKR), and HRI (9, 1820). Consistent with previous reports (14), PERK–/– mouse embryo fibroblasts (MEFs) did not phosphorylate eIF2α (21) or induce the eIF2α phosphorylation–dependent expression of CHOP in response to Tm treatment (fig. S3). However, sal induced eIF2α phosphorylation (21) and CHOP expression (fig. S3) regardless of PERK genotype, demonstrating that sal action does not depend on PERK. Similarly, sal induced eIF2α phosphorylation in MEFs deficient in GCN2, PKR, or HRI (21), indicating that sal action does not depend on any single eIF2α kinase.

Next, we asked whether sal inhibited eIF2α dephosphorylation. During ER stress, eIF2α is dephosphorylated by a complex containing the serine/threonine phosphatase PP1 and its nonenzymatic cofactor GADD34 (13). In an in vitro dephosphorylation assay (13), lysate from cells overexpressing an active fragment of GADD34 (GADD34C) efficiently dephosphorylated eIF2α, whereas lysate from vector-only control cells did not (fig. S4A). However, when GADD34C-overexpressing cells were treated with sal, eIF2α dephosphorylation by the resulting lysate was inhibited (Fig. 3A). CReP, a recently discovered homolog of GADD34, is constitutively expressed and mediates eIF2α dephosphorylation in unstressed cells (16). Sal also inhibited in vitro dephosphorylation of eIF2α by lysate from CReP-overexpressing cells (fig. S5A) and induced eIF2α phosphorylation in GADD34–/– cells (fig. S5B), where CReP is active (16). Therefore, sal inhibits eIF2α dephosphorylation mediated by both constitutive and ER stress-induced phosphatase complexes.

Fig. 3.

Sal selectively inhibits eIF2α dephosphorylation. (A) Cell lysates from GADD34C-expressing CHO cells dephosphorylated 32P-eIF2α labeled in vitro with γ-32P-ATP and glutathione S-transferase (GST)–PERK (left). Lysates from GADD34C-expressing cells preincubated for 24 hours with sal were partially inhibited in their ability to dephosphorylate eIF2α (right). Calyculin A (CA) (100 nM) serves as a positive control for inhibition of dephosphorylation. Reactions were analyzed by SDS–polyacrylamide gel electrophoresis (PAGE), and phosphorylation levels were assessed by autoradiography. (B) Tm and sal induce an eIF2α phosphorylation–dependent CHOP-GFP reporter to a similar extent (top). In cells also overexpressing GADD34 (bottom), Tm is unable to induce the CHOP-GFP reporter, whereas induction by sal is less affected. (C) Sal treatment abrogates the co-elution of GADD34 and PP1 in high molecular mass fractions (21 and 22) on gel filtration. PC12 cells were treated for 36 hours with DMSO or sal, and lysates were separated by gel filtration. PP2A is detected in the same fraction numbers in both samples. Proteins were detected by immunoblot. Molecular mass standards eluted as indicated (above fraction numbers). (D) Phosphorylation of eIF2α but not histone H3 was induced in PC12 cells after treatment with sal for 36 hours (top). Phosphorylation of both eIF2α and histone H3 was induced by treatment with CA for 2 hours (bottom). Phosphorylation and total protein levels assessed by immunoblot.

We tested the impact of sal on eIF2α dephosphorylation inside intact cells with the use of a CHOP–green fluorescent protein (GFP) reporter transgene whose expression is stringently dependent upon eIF2α phosphorylation (22). Both sal and Tm induced the CHOP-GFP reporter to a similar extent (Fig. 3B). However, stable overexpression of GADD34 nearly abrogated CHOP-GFP induction by Tm (Fig. 3B), because increased GADD34/PP1 phosphatase activity removes the Tm-induced eIF2α phosphorylation. In contrast, sal markedly activated CHOP-GFP even in the presence of overexpressed GADD34 (Fig. 3B) or CReP (fig. S5C). Therefore, sal is mechanistically distinct from Tm because sal can counteract enforced eIF2α phosphatase activity. However, CHOP-GFP induction by sal was antagonized by GADD34 or CReP overexpression, consistent with the model that these are the relevant targets of sal.

As a first step to characterize the sal mechanism, we examined its effects on the GADD34/PP1 complex. GADD34 and PP1 were present in the same high molecular mass gel filtration fractions in lysate from dimethyl sulfoxide (DMSO)–treated cells but not in lysate from sal-treated cells (Fig. 3C, fractions 21 and 22). Thus, sal may affect the composition of an active GADD34/PP1 complex inside cells. As a specificity control, the serine/threonine phosphatase PP2A was detectable in the same fractions regardless of sal treatment (Fig. 3C). Along with the in vitro dephosphorylation and CHOP-GFP experiments, these results indicate that sal inhibits eIF2α dephosphorylation by cellular complexes. Whether sal inhibits the GADD34/PP1 and CReP/PP1 complexes via direct binding or an indirect signaling event is currently under investigation.

The known chemical inhibitors of PP1 catalytic activity block the dephosphorylation of all PP1 substrates, including eIF2α (23). We therefore asked whether sal is selective in its inhibition of eIF2α dephosphorylation. Treatment of cells with a range of sal doses caused strong phosphorylation of eIF2α but not of Ser7 on histone H3 (Fig. 3D). In contrast, treatment with calyculin A (CA), an active-site inhibitor of PP1 catalysis, caused parallel increases in both eIF2α and histone H3 phosphorylation (Fig. 3D). CA also induced Thr phosphorylation on many proteins in the cell, whereas sal did not induce Thr phosphorylation levels above vehicle control, even after long incubations (fig. S6). Furthermore, low doses of CA were toxic to PC12 cells (fig. S7), whereas doses of sal that induce similar levels of eIF2α phosphorylation (Fig. 3D) were nontoxic and cytoprotective (Fig. 1), confirming that the cellular effects of CA and sal are distinct. As a specificity control, the phosphorylation of the PP2A substrate extracellular signal–regulated kinase 2 (ERK2) (24) was unaffected by sal (fig. S8). Lastly, we used a proteomics approach (25) to look for global changes in protein level or posttranslational modification upon sal treatment. Of thousands of detectable proteins, only three to four were consistently affected by sal treatment (26). We concluded that sal selectively inhibits the PP1-mediated dephosphorylation of eIF2α and possibly a limited number of other substrates.

Protein phosphatases regulate many physiological and pathological processes. To test whether a selective pharmacological inhibitor of eIF2α dephosphorylation might be useful in a disease model, we turned to herpes simplex virus (HSV). Cellular PKR is activated by HSV infection and phosphorylates eIF2α to slow viral protein synthesis and virion production (27). To counteract this activity, HSV encodes ICP34.5, a protein homologous to GADD34, which binds cellular PP1 and mediates eIF2α dephosphorylation (28, 29). Because ICP34.5 is essential for HSV replication in some cell types (28, 30), we asked whether sal could inhibit ICP34.5/PP1-mediated eIF2α dephosphorylation and block viral replication. Sal induced eIF2α phosphorylation in both mock- and HSV-infected Vero cells (Fig. 4A), where ICP34.5/PP1 or PP1 is active (29). Consistent with previous reports (29), lysate of ICP34.5-transfected cells dephosphorylated eIF2α in vitro (Fig. 4B), whereas those of vector-transfected controls did not (fig. S4B). Sal treatment of ICP34.5-transfected cells inhibited dephosphorylation of eIF2α by the resulting lysate (Fig. 4B). We concluded that sal inhibits viral ICP34.5/PP1-mediated eIF2α dephosphorylation.

Fig. 4.

Sal inhibits HSV replication by inhibiting eIF2α dephosphorylation. (A) Sal induced eIF2α phosphorylation in both mock-infected and HSV-infected Vero cells. Vero cells were treated with sal for 24 hours, infected with HSV, and harvested 18 hours postinfection (hpi). Protein levels were assessed by immunoblot. (B) Cell lysates made from ICP34.5-transfected CHO cells and treated with DMSO for 24 hours efficiently dephosphorylated 32P-eIF2α labeled in vitro with γ-32P-ATP and GST-PERK (lanes 1 to 6). Lysates from ICP34.5-expressing cells incubated with sal for 24 hours were partially inhibited in their ability to dephosphorylate eIF2α (lanes 7 to 12). CA (250 nM) serves as positive controls for inhibition of dephosphorylation. Reactions were analyzed by SDS-PAGE, and phosphorylation levels were assessed by autoradiography. (C) Vero cells were pretreated with sal as indicated for 24 hours before infection with an average of 150 plaque-forming units HSV per well. After 2 to 3 days in the continued presence of sal, viral plaques were quantitated, showing that sal inhibited HSV replication in a dose-dependent manner. Wells treated with DMSO alone contained ∼150 plaques. (D) Wild-type MEFs (circles) or eIF2αA/A MEFs (squares) were treated with the indicated concentrations of sal for 24 hours and then infected with HSV at multiplicity of infection equal to 10 in the continued presence of sal. After the virus adsorbed to the cells, medium containing the indicated concentrations of sal was added. The virus produced was titered on Vero cells.

To determine whether sal inhibits HSV replication, we infected Vero cells with HSV in the presence of a dose range of sal and measured viral plaque formation. At high doses, sal reduced plaque formation by more than 95% [median inhibitory concentration (IC50) ∼ 3 μM] (Fig. 4C). To examine the role of eIF2α phosphorylation in the anti-HSV activity of sal, we used MEFs homozygous for a Ser51 → Ala51 mutation in eIF2α (eIF2αA/A), which eliminates its phosphorylation site (15). Sal strongly inhibited HSV replication in wild-type MEFs, but inhibition was greatly reduced in eIF2αA/A MEFs (Fig. 4D), demonstrating that induction of eIF2α phosphorylation is important for sal's anti-HSV effect. The residual inhibition of HSV replication in eIF2αA/A MEFs may be due to sal's effect on as-yet unidentified substrates of ICP34.5/PP1 besides eIF2α. Lastly, we asked whether sal could inhibit HSV replication in a mouse cornea infection model (31). Compared to vehicle control, topical sal treatment significantly reduced the viral titer recovered from eye swabs of infected animals (n = 8, P < 0.008) (fig. S9). We concluded that sal can inhibit HSV replication in both cultured cells and an animal model of viral infection.

Because the catalytic subunits of most protein phosphatases show little intrinsic substrate preference, the selective pharmacological inhibition of protein dephosphorylations has generally been regarded as difficult or impossible. We have identified a small molecule with low toxicity that selectively inhibits a known PP1-mediated dephosphorylation in intact cells and used it to manipulate cellular physiology to achieve specific outcomes. Therefore, our results provide a proof-of-principle demonstration that pharmacological intervention against the dephosphorylation of selected substrates is a feasible approach for the development of research reagents and perhaps therapeutic agents. We show that selective small molecule inhibition of eIF2α dephosphorylation effectively protects cells from ER stress, revealing a critical role for eIF2α dephosphorylation in ER stress–induced apoptotic signaling, and blocks the replication of HSV, a widespread human pathogen. Future drugs that target eIF2α dephosphorylation may therefore complement the anti-HSV agents in current clinical use. More broadly, a detailed understanding of sal's mechanism may facilitate the discovery of inhibitors of other selected cell- or pathogen-directed dephosphorylations. Such small molecules could find therapeutic use in diseases as diverse as neurodegeneration, cancer, and viral infection.

Supporting Online Material

Materials and Methods

Figs. S1 to S10

References and Notes

Stay Connected to Science

Navigate This Article