Elongation Factor 2 Kinase Is Regulated by Proline Hydroxylation and Protects Cells during Hypoxia

Abstract

Protein synthesis, especially translation elongation, requires large amounts of energy, which is often generated by oxidative metabolism. Elongation is controlled by phosphorylation of eukaryotic elongation factor 2 (eEF2), which inhibits its activity and is catalyzed by eEF2 kinase (eEF2K), a calcium/calmodulin-dependent α-kinase. Hypoxia causes the activation of eEF2K and induces eEF2 phosphorylation independently of previously known inputs into eEF2K. Here, we show that eEF2K is subject to hydroxylation on proline-98. Proline hydroxylation is catalyzed by proline hydroxylases, oxygen-dependent enzymes which are inactivated during hypoxia. Pharmacological inhibition of proline hydroxylases also stimulates eEF2 phosphorylation. Pro98 lies in a universally conserved linker between the calmodulin-binding and catalytic domains of eEF2K. Its hydroxylation partially impairs the binding of calmodulin to eEF2K and markedly limits the calmodulin-stimulated activity of eEF2K. Neuronal cells depend on oxygen, and eEF2K helps to protect them from hypoxia. eEF2K is the first example of a protein directly involved in a major energy-consuming process to be regulated by proline hydroxylation. Since eEF2K is cytoprotective during hypoxia and other conditions of nutrient insufficiency, it may be a valuable target for therapy of poorly vascularized solid tumors.

INTRODUCTION

Many cells require aerobic metabolism to generate energy, necessitating an adequate supply of oxygen. Protein synthesis, especially translation elongation, is a major energy-consuming process, and translation elongation uses both ATP (for aminoacyl-tRNA charging) and GTP (at least two GTP equivalents are used during each round of the elongation process). Overall, at least four ATP equivalents are used for each amino acid added to the growing chain during elongation. Elongation rates can be regulated through the phosphorylation of eukaryotic elongation factor 2 (eEF2) (1). Phosphorylation of eEF2 on Thr56 by eEF2 kinase (eEF2K) inhibits its ability to interact with ribosomes (2), thereby impairing translation elongation. Indeed, a range of studies has shown that increased phosphorylation of eEF2 is associated with slower ribosomal movement along the mRNA (e.g., see references 3 to 5).

eEF2K interacts with calmodulin (CaM) through a binding site which lies almost immediately N terminal to its catalytic domain (6, 7). The catalytic domain belongs to the small group of (six) mammalian α-kinases, rather than the main protein kinase superfamily; α-kinases show no sequence homology and only limited three-dimensional structural homology to other protein kinases (8, 9). eEF2K activity is regulated through several signaling pathways linked, e.g., to nutrient availability; these include signaling through the mammalian target of rapamycin complex 1 (mTORC1), which represses eEF2K activity, and the AMP-activated protein kinase (AMPK), a key cellular energy sensor (10) which causes activation of eEF2K (11, 12), probably in part by inhibiting mTORC1 signaling. Both inputs operate such that nutrient starvation activates eEF2K to inhibit eEF2 and slow down elongation. This, in turn, helps conserve ATP (and GTP; ATP are GTP are interconverted by nucleoside diphosphate kinase) and amino acids, key precursors for protein production. Indeed, recent studies show that eEF2K plays a key role in the ability of cancer cells to cope with nutrient starvation and that they adapt to poor nutrient availability by switching on eEF2K (likely via AMPK) (4). To date, no substrates for eEF2K other than eEF2 have been reported.

Oxygen starvation (hypoxia) also imposes a stress on many cells, e.g., by impairing ATP production by mitochondria (and other effects). Hypoxia is especially important in highly oxidative tissues, such as heart muscle and brain, e.g., during cardiac ischemia or stroke. One important mechanism by which cells can respond to inadequate oxygen (hypoxia) involves the regulation of proteins by proline hydroxylation. Proline hydroxylation is catalyzed by proline hydroxylases (PHDs), which require oxygen as a cosubstrate (13). The best-known example of control of an intracellular protein by proline hydroxylation is the transcription factor hypoxia-inducible factor 1α (HIF1α). During normoxia, proline hydroxylation of HIF1α renders it a substrate for the E3 ubiquitin ligase von Hippel-Lindau, leading to its proteasome-mediated destruction (13). Hydroxylation of HIF1α is impaired during hypoxia, allowing its stabilization and increasing its levels. This enhances the transcription of HIF1α target genes, which encode proteins that help cells withstand hypoxia, e.g., the glucose transporter Glut1 (14). Identifying proteins that are subject to proline hydroxylation is challenging, and very few other intracellular proteins are so far known to be regulated by this modification. In particular, no PHD targets that regulate energy-demanding processes have previously been discovered.

Previous studies in cardiomyocytes and in breast cancer cells have shown that the phosphorylation of eEF2 increases during hypoxia and contributes to cell survival under these conditions (15, 16). However, it was unclear whether eEF2K is actually activated under these conditions. More recently, it has been shown that inhibition of prolyl hydroxylases increases eEF2 phosphorylation (17), but again, the mechanism remained unclear. Here we show that eEF2K is activated during hypoxia or upon inhibition of prolyl hydroxylases. We show that eEF2K is inhibited by its hydroxylation on a highly conserved proline residue, restricting its activity during normoxia. During hypoxia, when proline hydroxylation is impaired, eEF2K becomes more active to inhibit protein synthesis, thus protecting cells from hypoxia. This is the first example of the direct regulation by proline hydroxylation of an enzyme involved in damping down a major energy-consuming process.

MATERIALS AND METHODS

Antibodies and chemicals.

The mTOR inhibitor rapamycin was purchased from Calbiochem (Nottingham, United Kingdom), and the mTOR inhibitor AZD8055 was purchased from Tocris. FG-4497 was synthesized at FibroGen, Inc. (San Francisco, CA). Where indicated, cells were exposed to the PHD inhibitor FG-4497 at the concentrations and periods of time indicated below. Control conditions included exposure to equivalent concentrations of vehicle. All other chemicals and biochemicals were from Sigma, unless otherwise indicated. Antibodies for anti-human eEF2K and anti-mouse eEF2K were generously provided by the Division of Signal Transduction Therapy, University of Dundee, Dundee, United Kingdom. Anti-FLAG was from Sigma, and the anti-phospho-eEF2(Thr56) antibody was generated by Eurogentec; all other antibodies were from Cell Signaling Technology. eEF2 for kinase assays was purified from HeLa cell cytoplasm (essentially as described in reference 18).

Cell culture and treatment.

Animals were bred and maintained in accordance with the UK Animals (Scientific Procedures) Act 1986 and with the approval of the United Kingdom Home Office. Mouse embryonic fibroblasts (MEFs) from eEF2K^−/− (eEF2K knockout [eEF2K-KO]) mice and matched wild-type counterparts were prepared from embryos at embryonic day 13.5. MEFs from eEF2K (wild-type [WT]) and eEF2K^−/− mouse embryos were cultured in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) supplemented with 10% fetal calf serum (FCS), 100 units/ml penicillin, and 0.1 mg/ml streptomycin (Invitrogen) at 37°C in a humidified atmosphere of 5% CO₂. TSC2^−/− MEFs and the corresponding TSC2^+/+ cell line were generously provided by David Kwiatkowski (Harvard University, Boston, MA). HeLa cells were grown in DMEM supplemented with 10% (vol/vol) FBS, 2 mM glutamine, and 1% penicillin-streptomycin. Colorectal adenocarcinoma HCT116 cells were generously provided by Janssen Pharmaceuticals and cultured using standard procedures in a 37°C humidified incubator with 5% CO₂ in high-glucose McCoy's 5A modified medium (Invitrogen) supplemented with 10% (vol/vol) fetal calf serum, 100 units/ml penicillin, and 0.1 mg/ml streptomycin. HEK293 human embryonic kidney cells were cultured and transfected as described previously (19). WT and AMPKα1/α2 double-knockout (DKO) MEFs (20) were generously provided by Benoit Viollet (Institute Cochin, University of Paris). Cells were grown in DMEM supplemented with 10% (vol/vol) fetal bovine serum, 2 mM glutamine, and 1% penicillin-streptomycin.

Cells of a human lung carcinoma A549 cell line containing a plasmid carrying IPTG (isopropyl-β-d-1-thiogalactopyranoside)-inducible short hairpin RNA (shRNA) directed toward the eEF2K mRNA were generously provided by Janssen Pharmaceuticals, and the cells were cultured using standard procedures in a humidified incubator at 37°C with 5% CO₂ in DMEM supplemented with 10% (vol/vol) fetal bovine serum (FBS), 1.5 mM glutamine, 100 units/ml penicillin, and 0.1 mg/ml streptomycin. To induce the knockdown of eEF2K, cells were cultured for 5 days with 1 mM IPTG prior to experimentation.

Primary cultures of cortical neurons were isolated as previously described (21). Cortices were isolated in ice-cold Hanks' balanced salt solution from postnatal day zero (P0) or P1 eEF2K-KO or WT mice. Tissue was gently minced and digested in neurobasal medium (NM) containing papain (20 U/ml) and 0.32 mg/ml of l-cysteine at 37°C for 20 min, followed by 31°C for an additional 20 min. Tissue was then washed once in NM containing 1 mg/ml each of bovine serum albumin (BSA) and soybean trypsin inhibitor (STI) and then incubated at 37°C for 2 min in NM containing 10 mg/ml each of BSA and STI. The dissociated tissue was then washed once in NM and triturated with fire-polished glass pipettes. Cells were then washed in NM and passed through a 40-μm-pore-size cell strainer, after which they were counted and plated onto poly-d-lysine-coated culture dishes at 1,500 to 2,000 cells/mm². After 1 h, the overlying medium was removed and replaced with NM containing 2% B27 supplement, 0.5 mM l-glutamine, and 100 U/ml penicillin-streptomycin. One-half of the medium was replaced with fresh medium every 3 to 4 days. Cells were used in experiments at in vitro day 7.

Hypoxic culture conditions (1% or 0.1% [vol/vol] O₂) were achieved in a custom-designed hypoxic flush chamber (Billups and Rothenberg, Inc.) by infusion of a preanalyzed gas mixture (1% oxygen, 5% CO₂, 94% N₂ or 1,000 ppm oxygen, 5% CO₂, 2 × 10⁷ Pa nitrogen; BOC Ltd.). All experiments were performed with exponentially growing cells that had been plated at an approximately 60% cell density and then made hypoxic 18 to 24 h later.

Cell lysis and analyses of samples.

For Western blot analyses, cells were extracted into buffer consisting of 50 mM β-glycerophosphate, pH 7.5, 1 mM EGTA, 1 mM EDTA, 1% (vol/vol) Triton X-100, 1 mM Na₃VO₄, 0.1% (vol/vol) β-mercaptoethanol, and protease inhibitors (leupeptin, pepstatin, and antipain, each at 1 μg/ml). Lysates were centrifuged at 13,000 rpm to remove debris. The protein concentrations in the resulting supernatants were determined as described previously (22). In all cases, the same amount of total protein was applied to each lane of a given gel.

Analysis of binding of eEF2K to CaM-Sepharose.

To study further the interaction of eEF2K with CaM, CaM-Sepharose-binding experiments were performed using wild-type glutathione S-transferase (GST)–eEF2K and GST-eEF2K with the W99A, W99L, and D97A mutations. The expressed proteins were treated overnight at 4°C with PreScission protease (20 U/ml) to cleave the GST tag. Cleaved proteins were isolated by adding glutathione resin (GE Healthcare) to remove the PreScission protease and the cleaved GST tag. One microgram of each sample in 50 mM (pH 7.5) and 150 mM NaCl was applied to CaM-Sepharose 4B (GE Healthcare) preequilibrated in 50 mM MOPS [3-(N-morpholino)propanesulfonic acid; pH 7.5], 150 mM NaCl, and either 2 mM CaCl₂ or 4 mM EDTA and incubated for 30 min at 4°C. The resin was washed three times with the appropriate buffer. Samples were analyzed by SDS-PAGE and Western blotting using anti-GST.

Protein synthesis measurements.

Cells were transferred to medium lacking Met and Cys for 1 h prior to labeling. They were incubated in the presence of [³⁵S]methionine/cysteine to a final (radioactive) concentration of 10 μCi/ml for 30 min. After incubation, the medium was removed completely and the cells were washed with ice-cold phosphate-buffered saline and lysed using a standard procedure (as described in “Cell culture and treatment” above). The protein concentrations in the extracts were measured using the Bradford method. Samples of lysate were applied to 3MM filter papers (Whatman) and allowed to dry at room temperature. After three brief washes with 5% (wt/vol) trichloroacetic acid, two at 100°C and one in ethanol, the filters were again dried. The incorporated radioactivity was measured by scintillation counting.

ATP measurements.

To quantitate the ATP present, a CellTiter-Glo luminescent cell viability assay (Promega) was used according to the manufacturer's instructions.

Assays for eEF2 kinase.

eEF2 kinase was assayed as described earlier (23), with the following modifications. HEK293 cells overexpressing eEF2K were harvested in high-Ca²⁺ buffer (50 mM HEPES, pH 7.4, 50 mM NaCl, 50 mM β-glycerophosphate, 0.3% [wt/vol] CHAPS {3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate}, 1 mM CaCl₂) in order to maintain the eEF2 kinase-CaM interaction. Ten nanograms of cell lysate was used in each reaction mixture with a total volume of 40 μl containing 20 mM MOPS, pH 7.0, 5 mM MgCl₂, 2 mM EDTA, 1 mM dithiothreitol, 2% (vol/vol) glycerol, and 4.14 mM CaCl₂–5 mM EGTA (with the concentration of free Ca²⁺ being 1 μM at pH 7) in the presence of purified eEF2 (2 μg) and [γ-³²P]ATP. Activity was assayed at 30°C. The reaction was stopped by the addition of SDS-PAGE sample buffer, and the incorporation of ³²P into eEF2 was determined by SDS-PAGE, followed by staining of the gel with Coomassie brilliant blue and fluororadiography using a Typhoon phosphorimager (GE Healthcare). To measure the activity of endogenous eEF2K, 1 μg of cell lysate was used in each reaction mixture. The eEF2 kinase was immunoprecipitated with antibodies raised against either eEF2 kinase or the FLAG epitope, as appropriate, and the beads were washed 4 times in extraction buffer containing 1 mM Ca²⁺ ions.

To assess the activity of recombinant eEF2K, GST-tagged eEF2K or point mutants were expressed in Escherichia coli Rosetta(DE3)pLysS and purified as previously described (24).

Assays of eEF2K activity were performed using buffer B (50 mM MOPS, pH 7.0 [unless stated otherwise], 20 μg/ml CaM [where present], 5 mM MgCl₂, 14 mM 2-mercaptoethanol, 0.67 mM CaCl₂, 2 mM EDTA, 0.4 mM EGTA, 1 mM benzamidine, 1 mM [each] leupeptin and pepstatin). Reaction mixtures, which typically had a total volume of 20 μl, contained 1 μg of purified eEF2, and the reactions were initiated by adding [γ-³²P]ATP (final concentration, 0.1 mM; 1 μCi in each reaction mixture). The reaction mixtures were incubated at 30°C for times up to 30 min, and then SDS-PAGE sample buffer was added. Samples were immediately heated at 95°C for 5 min to denature the proteins and stop the reaction. Products were analyzed by SDS-PAGE (10% gel), and after staining with Coomassie brilliant blue, the gels were placed into destaining/fixing solution (50% [vol/vol] methanol, 10% [vol/vol] acetic acid). The gels were then placed on Whatman 3MM paper, covered with Saran Wrap, and dried on a vacuum gel dryer. Radioactivity was detected using a phosphorimager (Typhoon; GE Healthcare).

Assays of eEF2K activity against the MH-1 peptide were performed in buffer B essentially as described previously (25) using 300 μM peptide and 100 ng of recombinant eEF2K. The filters were immediately placed in approximately 300 ml (for up to 25 filters) of 75 mM orthophosphoric acid and washed thrice more in similar volumes of orthophosphoric acid. They were then rinsed in ethanol and dried in an oven at 100°C. Radioactivity was determined using the Čerenkov method.

ITC.

Isothermal titration calorimetry (ITC) measurements were carried out using an iTC₂₀₀ microcalorimeter (GE Healthcare Bio-Sciences) at 25°C. eEF2K peptides corresponding to the wild-type eEF2K sequence (residues 78 to 100) were synthesized and purified to >95% purity (China Peptides, Shanghai, China). Calmodulin and all peptides were prepared and dialyzed in the same buffer [20 mM piperazine-N,N′-bis(2-ethanesulfonic acid), 150 mM KCl, 10 mM CaCl₂, pH 7.5]. Ligand was titrated into the protein solution at molar ratios of 10:1, corresponding to approximately 220 μM ligand (peptide) and 19 μM protein (CaM). Each experiment consisted of a first injection of 0.3 μl, followed by 39 injections of 1 μl of peptide solution into the cell while stirring at 800 rpm. Control titrations (peptide into buffer) were measured. Data acquisition and analysis were performed using the Origin scientific graphing and analysis software package (OriginLab). Data analysis was performed by generating a binding isotherm and best fit using the following parameters: stoichiometry of the interaction determined in the experiment (N), change in enthalpy (ΔH; cal/mol), change in entropy (ΔS; cal/mol/degree), and the binding constant (K; M⁻¹).

Protein digestion, analysis by MS, and data processing.

Proteomic analyses were carried out by the University of Leicester Proteomics Facility (Protein Nucleic Acid Chemistry Laboratory). Bands of interest were excised from the gel, and in-gel trypsin digestion was carried out upon each excised band (26). Each slice was destained using 200 mM ammonium bicarbonate–20% acetonitrile, followed by reduction (10 mM dithiothreitol; Melford Laboratories Ltd., Suffolk, United Kingdom), alkylation (100 mM iodoacetamide; Sigma, Dorset, United Kingdom), and enzymatic digestion with trypsin (sequencing grade-modified porcine trypsin; Promega, Southampton, United Kingdom) in 50 mM triethylammonium bicarbonate (Sigma) using an automated digestion robot (Multiprobe II Plus EX; PerkinElmer, United Kingdom). After overnight digestion, samples were acidified using formic acid (final concentration, 0.1% [vol/vol]).

Liquid chromatography-tandem mass spectrometry (MS/MS) was carried out using an RSLCnano high-pressure liquid chromatography system (Dionex, United Kingdom) and an LTQ-Orbitrap-Velos mass spectrometer (Thermo Scientific). Samples were loaded at a high flow rate onto a reverse-phase trap column (0.3 mm [inside diameter] by 1 mm) containing 5-μm-particle-size C₁₈ 300-Å Acclaim PepMap medium (Dionex) maintained at a temperature of 37°C. The loading buffer was 0.1% formic acid, 0.05% trifluoroacetic acid, and 2% acetonitrile.

Peptides were eluted from the trap column at a flow rate of 0.3 μl/min and passed through a reverse-phase capillary column (75 μm [inside diameter] by 250 mm) containing Symmetry C₁₈ 100-Å medium (Waters, United Kingdom) that was manufactured in-house using a high-pressure packing device (Proxeon Biosystems, Denmark). The output from the column was sprayed directly into the nanospray ion source of the LTQ-Orbitrap-Velos mass spectrometer.

The LTQ-Orbitrap-Velos mass spectrometer was set to acquire a 1-microscan Fourier transform MS (FTMS) scan event at 60,000 resolution over the m/z range of 350 to 1,250 Da in the positive ion mode. Accurate calibration of the FTMS scan was achieved using a background ion-lock mass for polydimethylcyclosiloxane (445.120025 Da). Subsequently, up to 10 data-dependent higher-energy collisional dissociation (HCD) MS/MS scans were triggered from the FTMS scan. The isolation width was 2.0 Da, the normalized collision energy was 40.0, and the activation time was 10 ms. Dynamic exclusion was enabled.

Mutagenesis of eEF2K.

The cDNA encoding human eEF2K was cloned into a pcDNA3.1 vector for expression of FLAG-tagged eEF2K in HEK293 cells. Mutagenesis of proline 96 or proline 98 in human eEF2 kinase to alanine was performed by PCR using a QuikChange mutagenesis kit (Stratagene). The forward primer for P96A was 5′-GAAGGCCAAGCACATGGCCGACCCCTGGGCTG-3′, and the reverse primer for P96A was 5′-CAGCCCAGGGGTCGGCCATGTGCTTGGCCTTC-3′. The forward primer for P98A was 5′-GCACATGCCCGACGCCTGGGCTGAGTTC-3′, and the reverse primer for P98A was 5′-GAACTCAGCCCAGGCGTCGGGCATGTGC-3′.

Reproducibility and statistical analysis of data.

All experiments were conducted at least three times, with similar data being obtained in each case. For the results of the immunoblotting assays, a typical set of data is shown in each case. Quantitation of immunoblotting data was achieved using LI-COR Odyssey software.

Numerical data are expressed as the mean ± standard error of the mean (SEM) for the number of individual experiments indicated below. Wherever appropriate, the statistical significance of the data was assessed using a Student's t test, a one-way analysis of variance (ANOVA) followed by Dunnett's posttest, or a two-way ANOVA followed by a Bonferroni post hoc test, as described in the figure legends, using GraphPad Prism (version 6) software.

RESULTS

Hypoxia elicits a delayed increased in eEF2 phosphorylation.

Subjecting HCT116 colorectal carcinoma cells to hypoxia caused a small but consistent increase in eEF2 phosphorylation at 2 h and a gradual but more pronounced increase by 8 to 16 h (Fig. 1A). Our findings are distinct from those of an earlier study, which reported a rapid increase in eEF2K phosphorylation during hypoxia (17); our data also revealed the existence of a second, slower, and much more marked rise in eEF2 phosphorylation. The rapid changes in eEF2 phosphorylation observed in the previous study likely reflect different regulatory inputs into eEF2 (17).

FIG 1.

Hypoxia or DMOG treatment induces phosphorylation of eEF2 in diverse cell types. (A to F) The indicated cells were subjected to hypoxia or treated with DMOG (1 mM) for the indicated times. Cell lysates were subjected to Western blot analysis using the indicated antibodies. The graph in panel A (right) shows the quantitation of data from multiple experiments; data are given as the mean ± SEM (n = 3; the value for control cells without hypoxia was set equal to 1). (G) HeLa cells were exposed to FG-4497 at the indicated concentrations for 6 h. Cell lysates were subjected to Western blot analysis using the indicated antibodies. The graph in panel G (bottom) shows quantitation of the data from three experiments; data are given as the mean ± SEM (the value for control cells without treatment was set equal to 1). The positive control for phosphorylated p38 mitogen-activated protein kinase was a sample from bone marrow-derived macrophages treated with lipopolysaccharide. P-eEF2, phosphorylated eEF2; P-rpS6, phosphorylated ribosomal protein S6; MAPK, mitogen-activated protein kinase; P-p38, phosphorylated p38; P-ACC, phosphorylated ACC; P-4E-BP1, phosphorylated 4E-BP1; +ve, positive. Data were analyzed using a one-way ANOVA. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Hypoxia also caused an increase in eEF2 phosphorylation in mouse embryonic fibroblasts (MEFs), HeLa cells, or primary cortical neurons by similar times (Fig. 1B to D). This was not observed in cells from mice in which the EEF2K gene had been disrupted (Fig. 1B; see also Fig. 2A and B), indicating that hypoxia-induced eEF2 phosphorylation requires eEF2K (and is not catalyzed by AMPK, which has been reported to phosphorylate eEF2 [27]).

FIG 2.

eEF2K knockout strategy involving generation of an eEF2K conditional knockout allele. (A) Murine eEF2K gene map and targeting vector. Rectangles, coding exons of the eEF2K gene as well as the FLP recombination target (FRT) sites and the neomycin resistance cassette (Neo). The targeting vector can be linearized with NotI. (B) Genotyping of littermates by PCR.

Although hypoxia (15, 16) or inhibition of PHDs using dimethyloxalylglycine (DMOG) has previously been shown to increase eEF2 phosphorylation through PHD2 (17), the underlying mechanisms remained unknown. The delayed nature of this effect was of particular interest, as hypoxia might increase the expression of eEF2K, e.g., at a transcriptional level via HIF1α, which mediates the induction of a gene expression program that helps cells adapt to hypoxia (28), or by stabilizing the eEF2K protein itself (analogous to the stabilization of HIF1α). However, there was no significant increase in eEF2K protein levels in cells subjected to hypoxia (Fig. 1A to C); if anything, in some cases, eEF2K levels actually fell (Fig. 1A), as has been seen under other conditions where eEF2K is activated (3, 29). Thus, the hypoxia-induced phosphorylation of eEF2 does not appear to result from stabilization of eEF2K or induction of its expression. The simplest alternative explanation is that the intrinsic activity of eEF2K increases during hypoxia, and this is examined below.

Hypoxia-induced phosphorylation of eEF2 does not require signaling via AMPK or mTORC1.

We first explored whether known upstream regulators of eEF2K were involved in the hypoxia-induced increase in eEF2 phosphorylation. Members of the p38 mitogen-activated protein kinase family are often activated during cell stresses and can phosphorylate eEF2K, although they provide a negative input to eEF2K activity rather than stimulating it (18, 30). Nevertheless, we examined whether p38 mitogen-activated protein kinase was activated (phosphorylated) during hypoxia in HCT116 cells, MEFs, and HeLa cells. A transient increase was seen in HCT116 cells well before the rise in eEF2 phosphorylation, and no change was seen in the other cell types tested (Fig. 1A to C).

Hypoxia and the consequent ATP depletion can stimulate AMPK, which can, in turn, activate eEF2K (11, 12). For example, hypoxia caused a delayed activation of AMPK in HCT116 cells, as judged from the phosphorylation of a well-established AMPK substrate, acetyl coenzyme A (acetyl-CoA) carboxylase (ACC; Fig. 1A), although this appeared to lag behind the increase in phospho-eEF2, suggesting that it may not be a causative event in regulating eEF2 here. An inhibitor of AMPK, MRT199665 (31), did not prevent the hypoxia-induced phosphorylation of eEF2 (Fig. 3A). Furthermore, hypoxia still increased eEF2 phosphorylation in AMPKα1/α2-double-knockout cells (Fig. 3B). Thus, the hypoxia-induced phosphorylation of eEF2 is independent of AMPK, ruling out the possibility that it is directly mediated by AMPK or via, e.g., the AMPK-induced inactivation of mTORC1 signaling (32, 33). We noted that the levels of both eEF2K protein and EEF2K mRNA were lower in AMPKα1/α2-DKO cells (Fig. 3B and C), suggesting that AMPK that may positively regulate EEF2K expression.

FIG 3.

Hypoxia-induced eEF2K activation is independent of AMPK. (A) HCT116 cells were subjected to hypoxia (0.1% oxygen) for 16 h in the presence or absence of the AMPK inhibitor MRT199665 (MRT). (B) WT or AMPKα1/α2-DKO MEFs were subjected to hypoxia (0.1% oxygen) for 16 h. Cell lysates were subjected to Western blot analysis using the indicated antibodies. (C) eEF2K mRNA levels were determined by reverse transcription-quantitative PCR and normalized to the level of 18S rRNA. Data were analyzed using an unpaired t test. (Inset) Western blot analysis of lysates of WT and AMPK-DKO MEFs; equal amounts of total protein were applied to the gel. The blots were developed using the indicated antibodies. ***, P < 0.001. (D) A549 cells were cultured in the presence or absence of 1 mM IPTG for 5 days to induce the knockdown of eEF2K. Cells were then treated with DMOG (1 mM) or AZD8055 (A8; 1 μM) for the indicated times. (E) Wild-type (eEF2K^+/+) or eEF2K knockout (eEF2K^−/−) MEFs were treated with DMOG (1 mM) or with AZD8055 (1 μM) for the indicated times. 0 h, samples from cells that had been preincubated with or without MRT199665 for 30 min. Samples from cells incubated for 8 h without DMOG were also analyzed. Lysates were analyzed by immunoblotting, as indicated. (F) HCT116 cells were treated with DMOG (1 mM) with or without the AMPK inhibitor MRT199665 for the indicated times. Cell lysates were subjected to Western blot analysis using the indicated antibodies. (G) HCT116 cells were treated with DMOG (1 mM) or AZD8055 (1 μM) for the indicated times. Cell lysates were subjected to Western blot analysis using the indicated antibodies.

Inhibition of proline hydroxylases induces the phosphorylation of eEF2 independently of altered mTORC1 or AMPK signaling.

To explore how PHDs regulate eEF2 and eEF2K, we made use of the PHD inhibitor DMOG. Treatment of HCT116 or HeLa cells with DMOG increased the phosphorylation of eEF2, often markedly (Fig. 1E and F and 3D, E, and G). However, eEF2K levels did not increase, similar to the situation during hypoxia (Fig. 1A to C). Since DMOG affects both prolyl hydroxylases and related enzymes, such as asparaginyl hydroxylase, we also tested the effects of a specific PHD inhibitor, FG-4497 (34). FG-4497 also enhanced eEF2 phosphorylation in HeLa cells (Fig. 1G). Since the use of FG-4497 requires that cells be maintained in lower levels of serum (2%) than usual (10%), with serum itself tending to raise phospho-eEF2 levels, the increase caused by FG-4497 was less than that seen with DMOG (which can be used on cells in the presence of higher serum concentrations). However, the final level of phospho-eEF2 was similar with either reagent, confirming that DMOG affects eEF2 phosphorylation by inhibiting PHDs.

DMOG was also unable to increase eEF2 phosphorylation in A549 cells where eEF2K had been knocked down using shRNA (Fig. 3D) or in eEF2K^−/− MEFs (Fig. 3E), showing that here the phosphorylation of eEF2 is also mediated by eEF2K. The effect of DMOG was relatively slow, typically requiring 4 to 6 h. In contrast, AZD8055 (35), which blocks the mTOR pathway, which negatively regulates eEF2K activity (5), induced a faster increase in eEF2 phosphorylation (Fig. 3G). This is consistent with the effect of DMOG not being due to impaired mTORC1 signaling, a point which is explored in detail below, although alternative explanations are possible. It should also be noted that while hypoxia did decrease mTORC1 signaling in HCT116 cells (as judged from the decrease in phosphorylation of ribosomal protein S6 [rpS6] at Ser240/244, well-known substrates for the p70 S6 kinases which are activated by mTORC1; Fig. 1A), this effect was much more rapid than the rise in phospho-eEF2, which is inconsistent with a direct link between them.

The robust induction of eEF2 phosphorylation by DMOG in HCT116 cells was not blocked by MRT199665 (31), again indicating that AMPK is not involved (Fig. 3F). However, DMOG did increase the phosphorylation of acetyl-CoA carboxylase, an AMPK substrate, and MRT199665 blocked this (Fig. 3F), implying that DMOG increases AMPK activity, although the mechanism underlying this is unclear. AMPK is activated by phosphorylation (at Thr172) by the protein kinase LKB1, which is not present in HeLa cells (36). DMOG failed to activate AMPK in this cell type, as judged by the phosphorylation status of its substrate, ACC (Fig. 1F), indicating that the effect of DMOG on AMPK probably requires LKB1.

Hypoxia and DMOG can each impair mTORC1 signaling via the HIF1α-mediated induction of REDD1, a positive regulator of the TSC1/TSC2 complex that inhibits mTORC1 signaling (37, 38). Since mTORC1 signaling negatively regulates eEF2K, hypoxia or DMOG might induce eEF2 phosphorylation by relieving this inhibitory input from mTORC1. In HCT116 colorectal carcinoma cells, MEFs, and HeLa cells, hypoxia or DMOG treatment did indeed inhibit mTORC1 signaling, as judged by the decreased phosphorylation of ribosomal protein S6 (Fig. 1A, C, and F), a substrate for the S6 kinases, which are activated by mTORC1, and of 4E-BP1, another mTORC1 substrate (Fig. 1C). Since the effects of hypoxia on mTORC1 signaling are mediated via the TSC1/TSC2 complex (37), we used TSC2-null cells to eliminate this input. Hypoxia or DMOG did not affect S6 phosphorylation in TSC2^−/− MEFs but still induced the phosphorylation of eEF2 (Fig. 4A and B), albeit more slowly, without the initial rise, likely reflecting the high activity of mTORC1 signaling (and the lower eEF2K activity) in these cells. This implies that the slow phase of the effect of DMOG is not mediated via impaired mTORC1 signaling. PHDs also play a role in the activation of mTORC1 signaling by amino acids (38). However, DMOG did not affect mTORC1 signaling (S6 phosphorylation) in TSC2^−/− cells, where this input was still intact; therefore, the increased phosphorylation of eEF2 does not reflect this role of PHDs in cellular regulation. Taken together, these data demonstrate that hypoxia and DMOG treatment enhance eEF2 phosphorylation independently of inhibition of mTORC1 signaling.

FIG 4.

Hypoxia or DMOG induces eEF2 phosphorylation independently of changes in mTORC1 signaling. (A, B) TSC2^+/+ or TSC2^−/− MEFs were treated with DMOG (1 mM) for the indicated times (A) or subjected to hypoxia (0.1%) for 16 h (B). Equal amounts of cell lysates were analyzed by SDS-PAGE and Western blotting using the indicated antibodies. The graph in panel A (right) shows quantitation of data from multiple experiments. Data are given as the mean ± SEM (n = 3; values for control cells not treated with DMOG were set equal to 1). Data were analyzed using an unpaired t test. (C, D) TSC2^+/+ and TSC2^−/− MEFs were treated with 1 mM DMOG for 16 h. EEF2K and GLUT1 mRNA levels were determined by reverse transcription-quantitative PCR following DMOG treatment. Data are shown as the mean ± SEM (n = 3) and are normalized to the level of 18S rRNA. Data were analyzed using a two-way ANOVA. (E) Cells were treated with DMOG (1 mM) in the presence or absence of acriflavine (ACF; 5 μM). Equal amounts of cell lysate were analyzed by SDS-PAGE and Western blotting using the indicated antibodies. *, P < 0.05; **, P < 0.01.

eEF2K mRNA levels can be reciprocally regulated by mTORC1 and hypoxia.

We noted that eEF2K protein levels are lower in TSC2^−/− MEFs than in wild-type cells and that in these cells they did increase following DMOG treatment (Fig. 4A). To study this further, we measured the levels of the EEF2K mRNA. EEF2K mRNA levels were markedly lower in TSC2^−/− cells than wild-type MEFs (Fig. 4D) and increased in response to DMOG treatment in TSC2^−/− but not in WT MEFs (Fig. 4D). These effects, seen only in TSC2^−/− cells, where mTORC1 signaling is very active, suggest that DMOG may counter an inhibitory effect of hyperactivated mTORC1 signaling on eEF2K expression. DMOG induced another HIF1α-regulated mRNA, GLUT1, to a similar degree in both cell lines, i.e., irrespective of their TSC2 status (Fig. 4C). Acriflavine, which inhibits HIF1α dimerization (39), blocked the DMOG-induced rise in eEF2K levels (Fig. 4E), suggesting that this increase is mediated via HIF1α. However, analysis of the promoter region of the mouse EEF2K gene failed to reveal any consensus binding sites for HIF1α.

Thus, the expression of the mRNA for EEF2K is probably regulated in multiple ways, e.g., downstream of AMPK, and in some settings by mTORC1 signaling and HIF1α; further work, beyond the scope of this study, is required to define the mechanisms involved here.

Hypoxia or DMOG treatment enhances the intrinsic activity of eEF2K.

Our data show that in cells other than the single example of TSC2-null MEFs, hypoxia or DMOG treatment enhanced eEF2 phosphorylation without increasing the levels of the eEF2K protein. This suggested that the intrinsic activity of eEF2K may be enhanced under this condition. To test this, we assayed eEF2K activity in lysates of HeLa or HCT166 cells with or without an episode of hypoxia. As shown in Fig. 5A, hypoxia did indeed increase eEF2K activity. Similarly, higher eEF2K activity was observed following treatment of HEK293 cells or MEFs with DMOG, although the total levels of eEF2K were unchanged (Fig. 5B to D). To confirm conclusively that this change did reflect the activation of eEF2K and not, e.g., the stimulation of a (hypothetical) alternative kinase or a change in phosphatase activity against eEF2, eEF2K was immunoprecipitated from MEF lysates prior to the assay under mild conditions and in the presence of Ca²⁺ ions, where CaM remains associated with eEF2K (23). Again, higher activity was observed in samples from DMOG-treated cells than in samples from untreated controls (Fig. 5D, right), confirming that treatment of cells with DMOG does increase the activity of eEF2K. It should be noted that the data for DMOG and for TSC2^−/− MEFs (Fig. 4A and B) rule out the possibility that inputs from other effects of hypoxia or impaired mTORC1 signaling, respectively, increase eEF2K activity.

FIG 5.

Hypoxia or DMOG treatment enhances the intrinsic activity of eEF2K. The indicated cells were treated with DMOG or subjected to hypoxia (16 h), and eEF2K activity was determined either in whole-cell lysates (WCL) (A, C) or after immunoprecipitation of FLAG-tagged eEF2K (B) (in the presence or absence of Ca²⁺, as indicated). The immunoblot at the bottom of panel B shows total eEF2K levels in lysates from control or DMOG-treated cells. (C) MEFs were treated with DMOG for 6 h, and cell lysates were subjected to Western blot analysis using the indicated antibodies. (D) eEF2K activity was determined in whole-cell lysates or, after immunoprecipitation, in the presence or absence of CaM. Shown are phosphorimages of the corresponding SDS-polyacrylamide gels. IP endog. eEF2K, immunoprecipitated endogenous eEF2K.

Finally, it was possible, as is the case in response to elevated cyclic AMP levels (40), that eEF2K activity became independent of Ca²⁺ ions in DMOG-treated cells. However, eEF2K activity still showed a strong requirement for Ca²⁺ ions in samples from DMOG-treated cells (Fig. 5B).

eEF2K is subject to hydroxylation on a highly conserved proline.

Given that neither mTORC1 nor AMPK was involved in the effects reported here and that eEF2K levels were not increased by DMOG or hypoxia, we examined whether eEF2K is itself subject to hydroxylation. To do this, eEF2K was immunoprecipitated from normoxic cells and then subjected to tryptic digestion followed by mass spectrometric analysis. This revealed the presence of one peptide, i.e., HMPDPWAEFHLEDIATER, that showed a mass shift of 32 Da, which is equivalent to two additional oxygen atoms (mass of observed peptide, 2,224.8 Da; mass predicted from the sequence, 2,192.35 Da). Inspection of the mass spectrum shows that one of these oxygen atoms corresponds to the oxidation of the only methionine in this peptide (Table 1), which can occur during sample preparation. The hydroxylated peptide was observed in five separate experiments with material from normoxic cells, whereas the nonhydroxylated version of the peptide was not observed under this condition. The peptide identified corresponds to a region between the CaM-binding motif (6, 7) and the catalytic domain of eEF2K. It contains two proline residues, Pro96 and Pro98 (Fig. 6A).

TABLE 1.

m/z values for ions derived from the HMPDPWAEFHLEDIATER peptide and amino acid residues^a

^aData are for the b and y ions for the peptide from eEF2K from normoxic cells (the data for hydroxyproline 98 are shaded). The amino acid residue masses for the HMPDPWAEFHLEDIATER peptide are provided to assist with interpretation of the b and y ion series. Doubly charged ions are indicated by, e.g., b + 2H.

FIG 6.

eEF2K is subject to hydroxylation on a highly conserved proline residue. (A) (Top) Schematic depiction of the overall layout of eEF2K; the sequence of the CaM-binding site and adjacent residues is shown; Pro98 is indicated in red. (Bottom) sequences of the identified CaM-binding site in eEF2K proteins from the indicated species. The positions of Pro98 and Trp99 are indicated by capital letters. (B) A predicted mass spectrum, plotted using the m/z values shown in Table 2, showing two potential sites for proline hydroxylation. (C) A representative MS/MS spectrum shows fragmentation of the HMPDPWAEFHLEDIATER peptide. A mass identical to that of a hydroxyproline-containing fragment representing the HM(OH)PDP(OH)WAEFHLEDIATER peptide is observed. A +32-Da mass shift is observed in the peptide from eEF2K under normoxic conditions, for which m/z was 2,224.8 compared to the mass predicted from the sequence (m/z 2,192.35). The spectrum shows the y ion series appearing at y14 at m/z 865.4 (2+) and y16 at m/z 971.5 (+2), which unambiguously assigns hydroxylation to proline 98. The pertinent peaks are indicated (B and C, arrows).

It was important to pinpoint which proline in this peptide is modified. Figure 6B shows a theoretical spectrum based on this peptide, where either Pro96 (yielding the peptide with the peak labeled y16 + 2H in Fig. 6B) or Pro98 (y14 + 2H) is considered to be hydroxylated. Figure 6C and Table 1 show the results for the y ion series at y14 at m/z 865.4 (2+) and the b ion series at b5 at m/z 305.6 (2+). This assigns the hydroxylation event unambiguously to proline 98 and rules out the hydroxylation of Pro96 or any other similar modification, such as the double oxidation of methionine to the sulfone (Table 1).

Samples of eEF2K derived from cells subjected to hypoxia for 24 h (to allow replacement of hydroxylated eEF2K by the nonhydroxylated protein) showed only a lighter version of this peptide, which, unlike that from normoxic cells, was not hydroxylated on either proline residue. These data show that this peptide and Pro98 in particular are not hydroxylated under hypoxic conditions.

Pro98 and the two adjacent residues are highly conserved among vertebrate eEF2K sequences and, strikingly, given the low overall sequence identity in this region of eEF2K, are even conserved in eEF2K from nematode worms (Table 2). Given that Pro98 is close to the CaM binding site, we asked whether treatment of the cells with DMOG to prevent hydroxylation affected the ability of CaM to activate eEF2K.

TABLE 2.

The sequence adjacent to the CaM-binding region, which contains the hydroxylated residue, Pro98, in eEF2K is highly conserved

^aResidues shaded in gray are conserved between all vertebrate eEF2K sequences shown and in some cases in eEF2K from nematodes. The first tryptophan is essential for binding of eEF2K to CaM (6). Residues in boldface are completely conserved in all species; the proline within this boldface sequence corresponds to Pro98 of human eEF2K.

Cells were transfected with a vector harboring FLAG-tagged eEF2K. To ensure that, where appropriate, none of the eEF2K under study was hydroxylated, cells were treated with DMOG for the entire period during which the ectopic eEF2K was expressed. Lysates were prepared using buffer containing Ca²⁺ ions (where some CaM remains associated with eEF2K). Under these conditions, eEF2K from DMOG-treated cells showed activity higher than that from control cells (Fig. 7A, top); since expression of FLAG-eEF2K is much higher than that of the endogenous protein (Fig. 7B, right), the assay reflects the activity of FLAG-eEF2K.

FIG 7.

Hydroxylation on Pro98 limits eEF2K activity. (A) HEK293 cells were transfected with a vector encoding FLAG-tagged eEF2K and, where shown, treated with DMOG (1 mM) for the entire period from transfection to lysis (20 h). Cells were lysed in extraction buffer containing CHAPS with calcium to maintain the CaM-eEF2K interaction. Cell lysates were assayed with or without CaM for 10 min using eEF2 as the substrate. Phosphorimages of the corresponding SDS-polyacrylamide gels are shown. Total eEF2K levels (in whole-cell lysates) were analyzed by Western blotting using anti-FLAG. (B to D) FLAG-tagged wild-type eEF2K or the P96A or P98A mutants were expressed in HEK293 cells, which were subsequently treated with DMOG (6 h, 1 mM) where indicated. Immunoblot (IB) analyses using anti-FLAG (left; antitubulin was also used) or anti-eEF2K from lysates of cells ectopically expressing FLAG-eEF2K or nontransfected cells (endogenous eEF2K). (C, D) (FLAG)-eEF2K activity against eEF2 was determined with or without added calmodulin; (D) data for assays conducted without added CaM (mean ± SEM; n = 3). (E) FLAG-tagged wild-type eEF2K or the eEF2K P98A mutant was expressed in HEK293 cells, which were subsequently treated with DMOG (6 h, 1 mM) where indicated. Cells were lysed in extraction buffer containing CHAPS and calcium to maintain the CaM-eEF2K interaction. Cell lysates were assayed with or without calcium and CaM for 10 min using eEF2 as the substrate.

This difference between control and DMOG-treated samples could reflect a decreased association of eEF2K with CaM or a lower ability of CaM to activate eEF2K. To evaluate this, excess CaM was added to lysates from control or DMOG-treated cells; this strongly stimulated eEF2K activity toward eEF2 (Fig. 7A, middle). Importantly, even in the presence of excess CaM, the activity of eEF2K from DMOG-treated cells was substantially higher than that of eEF2K from control cells (Fig. 7A), where Pro98 is hydroxylated. This indicates that CaM is less able to stimulate eEF2K from control cells, where Pro98 is hydroxylated.

Similarly, in lysates from MEFs, addition of a large amount of CaM to the assay reaction mixture did not eliminate the difference in activity of eEF2K between control and DMOG-treated cells (Fig. 5D). The clear implication of these data is that hydroxylation of eEF2K likely impairs the extent to which CaM stimulates eEF2K, thus limiting its activity.

To test specifically the role of hydroxylation of Pro98 in controlling eEF2K activity, Pro98 was mutated to alanine, which cannot be hydroxylated and is otherwise probably the residue that is the most similar to proline. We also created the eEF2K Pro96Ala mutant as a (negative) control. The WT and mutant eEF2K proteins were expressed at similar levels in HEK293 cells, and their expression was not affected by DMOG treatment, confirming that PHD inhibition does not alter the stability of eEF2K (Fig. 7B). Since FLAG-eEF2K protein levels were much higher than those of endogenous eEF2K (Fig. 7B, right), we could reliably assess FLAG-eEF2K activity in cell lysates (where sufficient CaM is already present). DMOG pretreatment of the cells caused similar increases in the activity of WT eEF2K and the eEF2K P96A mutant (Fig. 7C and D). In contrast, the activity of the eEF2K P98A mutant was already higher than that of wild-type eEF2K in samples from normoxic cells, and more importantly, DMOG did not increase it further (Fig. 7C and D). The activities of WT eEF2K and the eEF2K P98A mutant were strictly dependent upon Ca and CaM under control and DMOG-treated conditions (Fig. 7E). These data suggest that hydroxylation of Pro98 under normoxic conditions limits eEF2K activity, while the absence of hydroxylation permits a higher level of activity.

Hydroxylation of Pro98 impairs binding of eEF2K to CaM and its activation by CaM.

The proximity of Pro98 to the CaM-binding site suggested that its hydroxylation might affect eEF2K's interaction with CaM. We therefore used isothermal titration calorimetry (ITC) to study the binding of CaM to synthetic peptides, including the CaM-binding site of eEF2K; we used peptides corresponding to residues 78 to 100 of human eEF2K and containing, at the position equivalent to residue 98, either proline or hydroxyproline. The unmodified peptide bound to CaM with an affinity of 66 ± 9 nM, while binding to residues 78 to 100 (with hydroxylated Pro98 [HO-Pro98]) was somewhat weaker (dissociation constant [K_d], 123 ± 3 nM) (Fig. 8A and B; Table 3), showing that the hydroxylation of Pro98 does interfere with CaM binding but does so only modestly. To assess whether hydroxylation affected the overall ability of eEF2K to bind CaM, HEK293 cells were transfected with a vector for FLAG-eEF2K. eEF2K was immunoprecipitated from control or DMOG-treated cells (Fig. 8C). No difference in the amount of CaM which copurified with the FLAG-tagged eEF2K was observed, consistent with the modest change in the affinity for CaM shown by the ITC data.

FIG 8.

Effect of hydroxylation of Pro98 on binding to CaM and of selected mutations at this position on eEF2K activity. (A, B) Isothermal calorimetry experiments were performed as described in Materials and Methods using unmodified peptide (A) and peptide hydroxylated on Pro98 (B). (Top) The original titration curve; (bottom) the resulting binding isotherm. (C) DMOG treatment of cells does not significantly affect the copurification of CaM and eEF2K. HEK293 cells were transfected with a vector encoding FLAG-eEF2K. They were then maintained for 48 h in the presence or absence of 1 mM DMOG from the point of transfection. Cells were lysed, and immunoprecipitations were carried out in a modified extraction buffer containing 1 mM CaCl₂ to maintain eEF2K-CaM interactions, as described in Materials and Methods. The FLAG-tagged eEF2 kinase was immunoprecipitated from 200 μg of lysate protein using immobilized FLAG antibody and then subjected to SDS-PAGE, followed by Western blotting for total eEF2 kinase or bound CaM. Blots for each protein are from nonadjacent lanes of the same gel. (D) The abilities of eEF2K and the indicated mutants to bind CaM were tested by affinity chromatography using CaM-Sepharose and bacterially expressed WT eEF2K or the eEF2K W99A, W99L, or D97A mutant from which the GST tag had been removed by PreScission protease cleavage, as described in Materials and Methods. Assays were conducted in the presence of CaCl₂. One microgram of cleaved WT eEF2K and mutant proteins was used in the pulldown with CaM resin, and samples of the proteins (0.1 μg) were run to display equal input levels. Samples were analyzed by immunoblotting using anti-eEF2K antibodies; the migration positions of GST-eEF2K and cleaved eEF2K are shown. (E) Activity of wild-type recombinant eEF2K or the indicated mutants toward the MH-1 peptide. Data are shown as the mean ± SEM (n = 3). P values compared with the results for the WT were obtained using a two-way ANOVA followed by Bonferroni posttests. *, P < 0.05; ***, P < 0.001; ****, P < 0.0001. (F) Activity of wild-type eEF2K or the eEF2K W99A mutant against eEF2. (Top) A phosphorimage of the Coomassie-stained gel shown below.

TABLE 3.

ITC data for the interaction of Ca²⁺ and CaM with eEF2K-based peptides^a

Peptide	Kd (nM)	ΔH (kcal/mol)	ΔS at 298 K (cal/mol)	−TΔS (kcal/mol)	ΔG (kcal/mol)	N
eEF2K from residues 78 to 100	66 ± 9.38	−15.2 ± 0.70	−18.4	−5.49	−9.76	0.99 ± 0.035
eEF2K from residues 78 to 100 with HO-Pro98	123 ± 2.64	−14.8 ± 0.18	−18.2	−5.43	−9.43	0.99 ± 0.007

^aThermodynamic parameters were obtained by fitting the ITC data to a single-state binding model. N, stoichiometry of the interaction determined in the experiment; K_d, dissociation constant; ΔH, change in enthalpy; ΔS, change in entropy; T, temperature; ΔG, Gibb's free energy, obtained by calculation of ΔG = ΔH − TΔS. Data are expressed as mean values ± standard deviations for three experiments.

If the altered affinity of eEF2K for CaM sufficed to explain its control by proline hydroxylation/DMOG, then addition of additional, saturating levels of CaM should overcome the difference in the affinity for CaM of eEF2K from control cells compared to that of eEF2K from hypoxic or DMOG-treated cells. However, the data reported in Fig. 6D and 8C show that addition of excess CaM did not bring the activity of eEF2K from control cells up to the higher level seen for DMOG-treated cells, indicating that binding of CaM to eEF2K does not limit its activity here. Thus, the hydroxylation of Pro98 probably restricts the activity of eEF2K by decreasing its capacity to be fully activated by CaM and/or its maximal activity.

As noted above, Pro98 lies in the linker region between the CaM-binding and catalytic domains of eEF2K, which is strongly conserved across vertebrate eEF2K sequences, suggesting that it might play an important role (Fig. 6A). Indeed, Pro98 and its flanking residues (Asp97 and Trp99 in human eEF2K) are even conserved in the eEF2K from distantly related nematodes. To test the roles of these residues, we mutated them to alanines and also, more conservatively, mutated W99 to leucine, another hydrophobic residue. The eEF2K D97A, W99A, and W99L mutants all still bound to CaM (Fig. 8D), likely because they lie beyond the identified CaM binding site. In contrast, in the presence of CaM, these mutants each showed substantially lower activity than WT eEF2K against MH-1, an established peptide substrate for eEF2K (7) (Fig. 8E). Indeed, the eEF2K W99A mutant almost entirely failed to undergo activation by Ca²⁺ and CaM when its activity against this peptide or eEF2 was assessed (Fig. 8E and F). Thus, these highly conserved residues in the linker region between the CaM-binding site and the catalytic domain apparently play a key role in coupling CaM binding to the stimulation of eEF2K activity, consistent with a role of hydroxylation of Pro98 in influencing the activation of eEF2K by CaM.

Taken together, our data indicate that hydroxylation of Pro98 both somewhat weakens the binding of eEF2K to CaM and impairs the ability of CaM to fully activate eEF2K (or achieve its maximal activity). Thus, even when eEF2K is assayed in the presence of excess CaM, hydroxylation of Pro98 still limits the maximal activity of eEF2K, thereby restraining eEF2K activity during normoxia.

Since there is no known mechanism to reverse proline hydroxylation (13), the enhancement of eEF2K activity during hypoxia requires the replacement of the preexisting hydroxylated eEF2K by newly made, more active nonhydroxylated eEF2K protein. Importantly, the rate of accumulation of new (in this case, FLAG-tagged) eEF2K is comparable to the rate of the hypoxia- or DMOG-induced increase in eEF2 phosphorylation (Fig. 9; cf. Fig. 1). Previous data indicate that eEF2K has a relatively short half-life (41). The nature of this control mechanism presumably explains why the effects of hypoxia or DMOG on eEF2 phosphorylation are relatively slow.

FIG 9.

Time course for the synthesis of new eEF2K molecules. HEK293 cells were transfected with a vector encoding FLAG-eEF2K; 24 h later, cells were treated with cycloheximide (CHX; 10 μg/ml) for 16 h, followed by release into cycloheximide-free medium for the indicated times. Cell extracts were prepared and run on SDS-polyacrylamide gels, followed by Western blotting using anti-FLAG and antitubulin antibodies.

eEF2K aids the survival of neuronal cells during hypoxia.

The activation of eEF2K that occurs during hypoxia is expected to restrain the rate of elongation and thereby help cells cope with the available energy supply. We therefore tested whether eEF2K plays a role in helping cells to withstand hypoxia. This condition has a particularly pronounced impact on cells or tissues which depend heavily upon oxidative metabolism (e.g., cardiomyocytes, neurons). Cardiac muscle and brain undergo oxygen deprivation during ischemia or stroke. eEF2K has previously been shown to protect cardiomyocytes against hypoxic damage (16). We therefore tested its importance in another oxygen-dependent cell type, primary neuronal cultures from wild-type or eEF2K^−/− mice. Following hypoxia, eEF2K^−/− neurons showed greater depletion of ATP and increased cleavage of poly(ADP-ribose) polymerase (PARP) (42) compared to the findings for wild-type cells (Fig. 10A and B). This indicates that eEF2K is cytoprotective in primary neurons, as reported previously for cancer cell lines (15) (Fig. 10B). These data are consistent with the finding that, in neuronal cells subjected to hypoxia, inhibition of protein synthesis using carbimazole (which induces eEF2 phosphorylation) or other agents preserved the ATP content and prevented cell damage (43). These data underline the importance of eEF2K for resistance to oxygen deprivation. Interestingly, recent work has revealed that eEF2K is cytoprotective during nutrient starvation (4). This key role for eEF2K in helping cells withstand nutrient deficiency likely reflects the facts that protein synthesis consumes a large proportion of the energy generated by oxidative metabolism (44) and that eEF2 phosphorylation inhibits translation elongation, reducing the energy demands.

FIG 10.

The loss of eEF2K compromises the ability of primary neuronal cells to withstand hypoxia. (A, B) Primary neuronal cultures from control or eEF2K knockout mice were maintained in culture for 7 days and then subjected to hypoxia (0.1% O₂, 20 h). ATP levels were measured using a CellTiter-Glo luminescent cell viability kit (A), or Western blots were performed for the indicated proteins (B). The graph in panel B (right) shows the levels of cleaved PARP (mean ± SEM; n = 3; the level of cleaved PARP in wild-type cells subjected to hypoxia was set equal to 1). Data were analyzed using a two-way ANOVA. RLU, relative light units. (C) TSC2^−/− MEFs were treated with DMOG (1 mM), rapamycin (Rapa; 200 nM), or cycloheximide (10 μg/ml) for 6 h and lysed. Samples were analyzed by immunoblotting using the indicated antibodies. (D) TSC2^−/− MEFs were treated with DMOG (1 mM, 6 h) and then incubated with [³⁵S]methionine/cysteine (for the final 30 min). Samples were processed to measure the incorporation of radiolabel into protein; data are shown as the mean ± SEM (n = 3). Data were analyzed using an unpaired t test.

It was important to assess whether the observed phosphorylation of eEF2 was associated with inhibition of protein synthesis. It is not possible to do this for hypoxia, since opening the hypoxic chamber to add the radiolabel admits oxygen, abrogating the hypoxia. It was therefore more appropriate to study the effect of DMOG; however, this generally impairs mTORC1 signaling, which in turn controls other components of the translational machinery, making it impossible to interpret data from wild-type (or even eEF2K^−/−) MEFs. To avoid this complication, we used TSC2^−/− MEFs, where DMOG affects neither mTORC1 signaling (as assessed by the phosphorylation of S6) nor the phosphorylation of eIF2, another key regulator of protein synthesis (Fig. 10C). DMOG substantially inhibited protein synthesis in TSC2^−/− MEFs (Fig. 10D), consistent with the concept that the phosphorylation of eEF2 serves to restrict ATP consumption, thereby favoring cell survival. It is important to note that the complete inhibition of protein synthesis during hypoxia is inappropriate, given that it is crucial that for some proteins to continue to be made to allow cells to adapt to this condition (e.g., HIF1α [45]).

DISCUSSION

Here we describe the first example of oxygen-dependent proline hydroxylation regulating a protein (eEF2K) that is directly involved in regulating a major energy-consuming process (protein synthesis). Hydroxylation of eEF2K during normoxia restrains its activity, such that its activity is enhanced during hypoxia, a response that will serve to decrease the demand for ATP and GTP for translation elongation. Importantly, Pro98 is not hydroxylated in eEF2K from hypoxic or DMOG-treated cells. eEF2K is also the first protein kinase whose activity is known to be regulated by proline hydroxylation; although IκB kinase-β is also subject to this modification, it primarily appears to regulate the levels of this kinase rather than its intrinsic activity (46). Two very recent studies showed that ribosomal protein S23 is subject to proline hydroxylation (47, 48), but unlike the effects reported here, this modification does not appear to be involved in the overall control of translation during hypoxia.

The fact that the modified residue in eEF2K, Pro98, and neighboring residues are completely conserved through evolution from nematodes to primates suggests that this is an ancient mechanism for regulating protein synthesis in response to oxygen deficiency. Our findings also point to a key role for residues in the linker region between the CaM-binding site in eEF2K and its catalytic domain in its activation by CaM. Given the high demand of protein synthesis for cellular energy (44), it makes clear physiological sense that hypoxia should induce the phosphorylation and inhibition of eEF2 to reduce the energy needs of protein synthesis. Ideally, it would be useful to show that WT eEF2K inhibits protein synthesis to a lesser extent than Pro98Ala mutant eEF2K under normoxic conditions within cells; however, more active mutants of eEF2K inhibit their own synthesis (or that of a cotransfected reporter) and are therefore expressed at levels markedly lower than those for less active variants (25, 49). This unfortunately makes it very hard to interpret the data from such experiments.

Our findings also reveal that the level of expression of the EEF2K mRNA is lower in cells lacking active AMPK or with hyperactivated mTORC1 signaling, pointing to transcriptional control of eEF2K expression by nutrient-sensitive signaling pathways. Interestingly, nutrient deprivation increases the expression of the EEF2K mRNA in mammalian cells and in Caenorhabditis elegans (4), consistent with enhanced expression of this cytoprotective kinase as a widespread response to a lack of nutrients. In contrast, oxygen deprivation or inhibition of PHDs did not increase the levels of eEF2K, except in the single example of TSC2^−/− cells, where constitutive mTORC1 signaling normally appears to repress eEF2K expression. Oxygen availability thus generally appears to promote the activation, rather than the expression level, of eEF2K.

The replacement of preexisting, hydroxylated, and less active eEF2K by more active, nonhydroxylated eEF2K presumably allows cells to adapt to low-oxygen conditions, such that enhancement of eEF2K activity no longer has to rely on inputs from inhibition of mTORC1 or activation of AMPK. This resets the activity of eEF2K and, thus, eEF2 phosphorylation at a higher level, providing a mechanism to allow cells to adapt to reduced oxygen availability by stably enhancing the activity of a protein kinase that slows down translation elongation. The activation of AMPK or inhibition of mTORC1 signaling, each of which can acutely activate eEF2K, likely serves to provide a short-term modulation of eEF2K activity. The long-term adaptive response revealed in this study is distinct from the rapid responses to PHD inhibition reported recently (17).

These data, together with earlier findings showing cytoprotective roles for eEF2K in hypoxic cardiac muscle cells (15) and in response to starvation for amino acids and glucose (see, e.g., reference 4), indicate that eEF2K helps to defend cells against the adverse effects of deficiencies in diverse nutrients. eEF2K may therefore be a useful target in tackling poorly vascularized solid tumors, regions of which may become hypoxic and require eEF2K to allow cell survival (50).