In the present study, we demonstrate that, in pancreatic β-cells, eIF2α (eukaryotic initiation factor 2α) phosphorylation in response to a decrease in glucose concentration is primarily mediated by the activation of PERK [PKR (protein kinase RNA activated)-like endoplasmic reticulum kinase]. We provide evidence that this increase in PERK activity is evoked by a decrease in the energy status of the cell via a potentially novel mechanism that is independent of IRE1 (inositol requiring enzyme 1) activation and the accumulation of unfolded nascent proteins within the endoplasmic reticulum. The inhibition of eIF2α phosphorylation in glucose-deprived cells by the overexpression of dominant-negative PERK or an N-terminal truncation mutant of GADD34 (growth-arrest and DNA-damage-inducible protein 34) leads to a 53% increase in the rate of total protein synthesis. Polysome analysis revealed that this coincides with an increase in the amplitude but not the number of ribosomes per mRNA, indicating that eIF2α dephosphorylation mobilizes hitherto untranslated mRNAs on to polysomes. In summary, we show that PERK is activated at low glucose concentrations in response to a decrease in energy status and that this plays an important role in glucose-regulated protein synthesis in pancreatic β-cells.
Glucose-stimulated insulin secretion from pancreatic β-cells is mediated by a series of steps initiated by the metabolism of glucose, which results in an increase in the cellular ATP/ADP ratio. This increase in the ATP/ADP ratio causes the closure of ATP-sensitive potassium channels, leading to membrane depolaization and the subsequent opening of L-type voltage-gated calcium channels. The resultant influx of calcium activates the exocytotic machinery, resulting in the secretion of insulin (for a review, see ). Glucose also stimulates a rapid increase in the rate of proinsulin synthesis (up to 20-fold within 1 h), the co-ordinate increase in the rate of synthesis of a large subset of proteins and a 2-fold increase in the rate of total protein synthesis [2–5]. These acute changes in the rate of protein synthesis are almost entirely regulated at the post-transcriptional level [2–6] and, like insulin secretion, require the metabolism of glucose . However, the metabolic signals emanating from the mitochondrial metabolism of glucose that stimulate secretion are distinct from those that stimulate protein synthesis [7,8].
An important rate-limiting step in protein synthesis is the assembly of the translational ternary complex, made up of the tRNAiMet (methionyl initiator tRNA) attached to eIF2 (eukaryotic initiation factor 2) in its GTP-bound state (eIF2–GTP:tRNAiMet) (for a review, see ). This ternary complex binds to the 40S ribosomal subunit, which, in association with additional initiation factors, delivers the tRNAiMet to the mRNA translational start site. Following translational start site recognition, the GTP bound to eIF2 is hydrolysed to GDP, rendering eIF2α unable to bind tRNAiMet. Therefore, in order to take part in further rounds of initiation, eIF2-GDP must be recycled back to eIF2–GTP. The ‘recycling’ of eIF2-GDP to eIF2–GTP is catalysed by the GEF (guanine-nucleotide-exchange factor) eIF2B. Since active eIF2–GTP is required for the recruitment of tRNAiMet, changes in the activity of eIF2B are important in regulating the rate of translation initiation . eIF2B activity is regulated by a number of mechanisms, including competitive inhibition by phosphorylation of the α-subunit of its substrate eIF2 (eIF2α) on Ser51 [11,12]. eIF2α phosphorylation occurs in response to a wide array of cellular stresses mediated by one of four eIF2α kinases: protein kinase RNA activated (PKR), mGCN2 [mammalian orthologue of the yeast GCN2 (general control non-derepressible 2)], HRI (haem-regulated eIF2α kinase) and PERK [PKR-like endoplasmic reticulum kinase]. PKR expression is induced by IFN (interferon) and is activated in response to dsRNA [13,14]. In yeast, GCN2 phosphorylates eIF2α in response to uncharged tRNAs (amino acid deprivation) and glucose starvation [9,15,16]. However, the role of the mammalian orthologue (mGCN2) is less clear. HRI is activated in response to haem deprivation in erythroid cells  and PERK is activated in response to protein malfolding in the ER (endoplasmic reticulum) as part of the UPR (unfolded protein response) [18–20]. It has been speculated that the UPR may be activated in response to high glucose concentration due to increased rates of protein synthesis exceeding the folding capacity of the ER and resulting in the accumulation of unfolded proteins and the activation of PERK [18,21–23].
We have previously demonstrated, in islets of Langerhans and in the pancreatic β-cell line MIN6 (mouse insulinoma cell line 6), that the availability of the translational ternary complex is increased in response to glucose  and that this was probably mediated through the dephosphorylation of eIF2α. In MIN6 cells, changes in both the availability of ternary complex and the phosphorylation status of eIF2α paralleled changes in protein synthesis, leading us to hypothesize that eIF2α phosphorylation is likely to play a key role in glucose-stimulated protein synthesis . Moreover, in MIN6 cells, increases in eIF2α phosphorylation in response to a temporally acute decrease in glucose concentration parallel the recruitment of mRNAs encoding integrated stress response genes on to polysomes . Recently, it has been reported that, in MIN6 cells, eIF2α dephosphorylation in response to increasing glucose concentrations is mediated, at least in part, by an increase in PP1 (protein phosphatase 1) activity directed towards eIF2α . However, the eIF2α kinase that phosphorylates eIF2α at low glucose concentration in pancreatic β-cells and its role in glucose-regulated protein synthesis is unknown.
MATERIALS AND METHODS
Chemicals and materials
FCS (foetal calf serum) was purchased from Invitrogen. 35S-labelled methionine was obtained from MP Biomedicals UK. All other chemicals were purchased from Sigma unless otherwise stated.
Cell culture and treatment
In the present study, MIN6 cells (kindly provided by Professor Jun-Ichi Miyazaki, Division of Stem Cell Regulation Research, Osaka University, Graduate School of Medicine, Osaka, Japan) were used between passages 25 and 40 at ∼80% confluence. MIN6 cells were grown in DMEM (Dulbecco's modified Eagle's medium) containing 25 mM glucose supplemented with 15% (v/v) heat-inactivated FCS, 100 μg/ml streptomycin, 100 units/ml penicillin sulfate, 75 μM 2-mercaptoethanol and 40 mM sodium bicarbonate, equilibrated with 5% CO2 and 95% air at 37 °C. Prior to treatment, the medium was removed and the cells were washed twice with KRB (Krebs–Ringer bicarbonate buffer; 115 mM NaCl, 5 mM KCl, 10 mM NaHCO3, 2.5 mM MgCl2, 2.5 mM CaCl2 and 20 mM Hepes, pH 7.4, supplemented with 0.5% BSA). The cells were then incubated for 1 h at 37 °C in KRB (unless otherwise stated) prior to incubation in KRB or KRB containing 20 mM glucose for a further hour at 37 °C (unless otherwise stated). Full details of treatments are provided in the Figure legends. After treatment, the cells were washed with ice-cold PBS and then lysed by the addition of ice-cold lysis buffer containing 1% Triton, 10 mM 2-glycerophosphate, 50 mM Tris/HCl (pH 7.5), 1 mM EDTA, 1 mM EGTA, 1 mM Na3VO4 (sodium orthovanadate), 1 mM benzamidine/HCl, 0.2 mM PMSF, 1 μg/ml each of leupeptin and pepstatin, 0.1% 2-mercaptoethanol and 50 mM sodium fluoride (unless otherwise stated). The lysates were then centrifuged for 10 min at 16000 g. The supernatants were kept and total protein concentrations were determined by the Bradford assay (Bio-Rad) using BSA as a standard. The protein lysates were stored at −80 °C until further analysis.
Determination of cellular ATP content
ATP content was determined using a bioluminescent assay (Sigma) following the manufacturer's instructions. Briefly, following treatments, as described above and in the Figure legends, cells were lysed by the addition of somatic-cell ATP-releasing reagent. Luminescence was then read immediately in a Novostar (BMG Labtech) 96-well plate reader with injectors.
Adenoviral construction and infection
Recombinant adenoviruses expressing amino acids 1–583 of PERK [i.e. dominant-negative PERK (Ad-PERKΔC)] and GFP (green fluorescent protein; Ad-Empty) have been previously described . In order to generate pAdTrack-GADDΔN, GADD34 (growth-arrest and DNA-damage-inducible protein 34) cDNA was amplified by PCR from a mouse cDNA library generated from MIN6 cells using the primers GADD34-N-F721 5′-AGCGAATTCTCTAGAGAGAAGCCTAAG-3′ and GADD34R-UTR (untranslated region) 5′-TATCTCGAGGGAAACTACTCAGGCTTAGCC-3′. The resultant cDNA was digested with EcoRI and XhoI and subcloned into EcoRI- and XhoI-digested pCMV-Tag3b, creating pCMV-Tag-GADDΔN, encoding the C-terminal end of GADD34 fused in frame with an N-terminal Myc epitope tag. Myc-tagged GADDΔN was amplified by PCR using primers GADD34-FMYC 5′-AGCGGTACCATGGAGCAGAAACTCATC-3′ and GADD34R-3UTR 5′-TATCTCGAGGGAAACTACTCAGGCTTAGCC-3′. The PCR product was digested with KpnI and XhoI and subcloned into KpnI- and XhoI-digested pAdTrack-CMV (CMV refers to cytomegalovirus). The resulting plasmid pAd-GADDΔN was used to generate the recombinant adenovirus Ad-GADDΔN as previously described . For infection, MIN6 cells were incubated in the presence of the virus for 48 h prior to treatments . Under these conditions, >90% of cells were infected, as determined by measuring EGFP (enhanced green fluorescent protein) expression by using fluorescence microscopy.
Silencing of PERK expression using siRNAs (small interfering RNAs)
PERK expression was specifically silenced in MIN6 cells using 25 nt prevalidated siRNA duplexes (Stealth™ Select RNAi purchased from Invitrogen) directed against mouse PERK (siRNA20: sense 5′-UAGAGGAGUUCAAACAGAAUCAAGC-3′ and antisense 5′-CGUUGAUUCUGUUUGAACUCCUCUA-3′; and siRNA21: sense 5′-UCUUUGAACCAUCAUAUGCUCUUGGG-3′ and antisense 5′-CCCAAGAGCAUAUGAUGGUUCAAGA-3′). An siRNA that showed no significant homology to any known protein was used as a control (Con siRNA: sense 5′-CGUGAUUGCGAGACUCUGATT-3′ and antisense 5′-UCAGAGUCUCGCAAUCACGTT-3′). siRNAs were introduced into MIN6 cells by electroporation using a pipette-type electroporator (Microporator MP-100 from Labtech International). Transfections were performed according to the manufacturer's instructions. Typically, transfection >90% was obtained, as assessed by monitoring the transfection efficiency of a fluorescent siRNA (Block-it™ Fluorescent Oligo from Invitrogen).
Protein synthesis measurements
Protein synthesis measurements were performed essentially as described previously . Briefly, cells were incubated in the presence of [35S]methionine for the times indicated in the text. Cell lysates (as described in the subsection ‘Cell culture and treatment’) were prepared and equal amounts of protein (approx. 20 μg) were spotted on to 3MM Whatman filters. The filters were then washed by boiling them for 1 min in 5% (v/v) trichloroacetic acid containing 0.1 g/l of methionine. This was repeated three times. The filters were then dried and immersed in scintillant before determining radioactivity by scintillation counting.
SDS/PAGE and Western blotting
SDS/PAGE and Western blotting were performed as described previously . Anti-phospho-eIF2α (Ser51) antibody was purchased from Biosource. Anti-phospho-AMPK (AMP-activated protein kinase) (Thr172) was purchased from Cell Signaling Technology. Anti-IRE1 (inositol requiring enzyme 1) and anti-phospho-IRE1 (Ser724) were purchased from Novus Biologicals. Anti-PERK antibody was kindly provided by Professor Ronald Wek (Department of Biochemistry and Molecular Biology, Indiana University School of Medicine, Indianapolis, IN, U.S.A.) and anti-eIF2α antibody was a gift from Professor Chris Proud (Department of Biochemistry and Molecular Biology, Columbia University, Vancouver, BC, Canada).
PCR-based assay for XBP-1 (X-box binding protein-1) splicing
After treatments described in the legend to Figure 5, total RNA was prepared from cells using TRI Reagent (Sigma) and first-strand cDNAs were synthesized from 4 μg of each sample using SuperScript™II reverse transcriptase (Invitrogen) according to the manufacturer's instructions. XBP-1 processing is characterized by excision of a 26 bp sequence from the coding region of XBP-1 mRNA . Thus the cDNAs prepared were used as a template for PCR by using primers flanking the splice site, as described by Marciniak et al. . Unspliced XBP-1 gave a product of 480 bp, and the spliced cDNA was 454 bp. The cleaved 26 bp fragment contains a PstI restriction site, and so the extent of XBP-1 processing can be evaluated by restriction analysis . PCR products were purified and digested with PstI. Restriction digests were separated on 3% agarose gel containing ethidium bromide. PCR products derived from unspliced XBP-1 mRNA (indicating absence of ER stress) were digested into two bands of 290 and 190 bp. In contrast, products amplified from spliced XBP-1 mRNA were resistant to digestion and remained 454 bp long, indicating the presence of ER stress.
Polysome profile analysis
After treatment, cells were lysed in polysome buffer (20 mM Hepes, pH 7.6, 15 mM MgCl2, 300 mM KCl, 1 mg/ml heparin, 0.1 mg/ml cycloheximide, 1 mM dithiothreitol, 1 μl/ml RNAguard™, 1 mM benzamidine/HCl, 0.1 mM PMSF, 1 μg/ml leupeptin and 1 μg/ml pepstatin) supplemented with 1% Triton. Cell lysates were centrifuged at 13000 g for 10 min at 4 °C to remove the nuclei and cell debris. The supernatants were then layered on to 20–50% sucrose gradients (made in polysome buffer) and centrifuged at 39000 rev./min for 2 h at 4 °C in a Sorvall TH64.1 rotor. The gradients were fractionated using an ISCO gradient fractionator that continuously measured the A254.
PERK is activated upon glucose deprivation
To investigate whether the eIF2α kinase PERK is activated by glucose deprivation, MIN6 cells were mock-infected or infected with a recombinant adenovirus overexpressing a dominant-negative mutant of PERK (AdPERKΔC). At 48 h post-infection, the cells were incubated in KRB without glucose or containing 20 mM glucose in the presence or absence of thapsigargin (an ER stress-inducing agent known to lead to the activation of PERK) as a positive control. The activation of PERK was assessed by its migration on SDS/PAGE using antisera against total PERK (Figure 1a). In the mock-infected cells, glucose deprivation and thapsigargin treatment resulted in a decrease in PERK mobility on SDS/PAGE, consistent with PERK phosphorylation, and an increase in the phosphorylation of its substrate eIF2α (Figure 1a). The overexpression of PERKΔC inhibited PERK's mobility shift and eIF2α phosphorylation induced in response to glucose deprivation and thapsigargin. These results demonstrate that PERK is activated upon glucose deprivation and that PERK activation is primarily responsible for eIF2α phosphorylation in glucose-deprived cells. To provide additional evidence that PERK is responsible for the phosphorylation of eIF2α in response to glucose deprivation, the expression of PERK was knocked-down in MIN6 cells using two distinct sets of siRNAs. The expression of PERK was reduced by up to 80%, which resulted in a 41–45% and a 34–41% reduction in eIF2α phosphorylation in response to either glucose deprivation or thapsigargin treatment respectively (Figure 1b). Taken together, these results provide good evidence that PERK phosphorylates eIF2α in glucose-deprived MIN6 cells.
PERK in activated by a decrease in glucose concentration
To determine whether these PERK-dependent increases in eIF2α phosphorylation in response to glucose deprivation also occur at physiologically relevant glucose concentrations, MIN6 cells were mock-infected or infected with the recombinant adenovirus AdPERKΔC and incubated in KRB containing 0, 2, 5, 7.8, 10, 20 or 40 mM glucose and, as a positive control, thapsigargin (Figure 1c). In mock-infected cells, the phosphorylation of eIF2α increased with decreasing glucose concentrations. Importantly, the expression of PERKΔC inhibited the phosphorylation of eIF2α (Figure 1c) at all glucose concentrations tested. This provides evidence that PERK is activated in response to physiologically relevant decreases in glucose concentrations.
We had previously reported that PERK was unlikely to be responsible for the phosphorylation of eIF2α at low glucose concentration . This conclusion was primarily based on results showing that, at low glucose concentration, PERK was not significantly phosphorylated at Thr980. However, it is likely that the anti-phospho-PERK Thr980 antibody used in our previous study was not sensitive enough to consistently detect changes in PERK phosphorylation at low glucose concentrations.
Decreases in intracellular energy status activate PERK
Given that protein folding is an energy-dependent process, one possible effector of PERK activation is a fall in the energy status of the cell in response to a decrease in glucose concentration. To investigate this possibility, we initially determined whether mitochondrial inhibitors of the electron-transport chain could activate PERK and stimulate the phosphorylation of eIF2α (Figure 2). Treatment of MIN6 cells with three distinct mitochondrial inhibitors of the electron-transport chain (Rotenone, sodium azide and oligomycin) all led to the phosphorylation of PERK and the phosphorylation of eIF2α (Figure 2a). All inhibitors had similar effects on intracellular ATP content, reducing ATP to less than one-sixth of that of the control (Figure 2b). Therefore a decrease in the energy status of MIN6 cells can stimulate the activation of PERK and the phosphorylation of eIF2α. To investigate whether physiological decreases in energy status also led to the activation of PERK and eIF2α phosphorylation, we artificially reduced cellular ATP content and the energy status of the cell to within the range observed in glucose-deprived cells using oligomycin (Figure 3). Increasing oligomycin concentrations led to a dose-dependent decrease in intracellular ATP content (Figure 3a) and decreases in the energy status of the cell, as assessed by the phosphorylation state of AMPK (a sensor of intracellular AMP and hence cellular energy status ) (Figure 3b). Importantly, oligomycin-induced decreases in ATP concentration and cellular energy status that were within the range seen in glucose-deprived cells caused increases in eIF2α phosphorylation (Figures 3b and 3c). These changes in eIF2α phosphorylation paralleled increases in the phosphorylation of PERK (Figures 3b and 3c). Increasing oligomycin concentrations also led to a dose-dependent decrease in the rate of [35S]methionine incorporation into protein, relative to the cells incubated at 20 mM glucose alone (Figure 3d). Interestingly, decreases in energy status/ATP alone are not solely responsible for the dramatic decrease in the rate of protein synthesis observed in glucose-deprived cells, as oligomycin-induced decreases in ATP/energy status similar to that evoked by glucose deprivation led to a comparatively small decrease in the rate of protein synthesis compared with that evoked by glucose deprivation (Figure 3).
PERK is activated by mitochondrial inhibitors
Decreases in energy status parallel increases in PERK and eIF2α phosphorylation
To confirm that oligomycin stimulated eIF2α phosphorylation by activating PERK, MIN6 cells were mock-infected or infected with AdPERKΔC. At 48 h post-infection, the cells were incubated in KRB or KRB plus 20 mM glucose in the presence or absence of increasing concentrations of oligomycin (Figure 3e). In the mock-infected cells, increasing concentrations of oligomycin led to a dose-dependent decrease in the mobility of PERK on an SDS/polyacrylamide gel, consistent with PERK phosphorylation. This phosphorylation of PERK paralleled increases in the phosphorylation of eIF2α (Figure 3e). Overexpression of PERKΔC resulted in an inhibition of PERK's phosphorylation in response to all doses of oligomycin tested and in response to glucose deprivation. These decreases in PERK phosphorylation correlated with a decrease in eIF2α phosphorylation (Figure 3e), demonstrating that oligomycin evokes the phosphorylation of eIF2α via the activation of PERK.
AMPK activation does not stimulate eIF2α phosphorylation or protein synthesis in pancreatic β-cells
As the activation of AMPK parallels eIF2α phosphorylation (see Figure 3), we investigated whether AMPK plays a role in the regulation of eIF2α phosphorylation or indeed glucose-regulated protein synthesis. MIN6 cells were pre-incubated for 1 h in KRB at 0 mM glucose before treatment for a further hour at either 0 or 20 mM glucose, or 20 mM glucose in the presence of increasing concentrations of AICAR (5-amino-4-imidazolecarboxamide riboside), a pharmacological activator of AMPK (Figure 4a). As expected, glucose deprivation led to a significant increase in the phosphorylation of eIF2α (Figure 4a) that correlated with an increase in AMPK activation, as assessed using a phospho-specific antibody to Thr172 of AMPK. However, when cells were treated with increasing concentrations of AICAR, AMPK phosphorylation was not accompanied by an increase in eIF2α phosphorylation (Figure 4a). Additionally, AICAR had no inhibitory effect on glucose-stimulated protein synthesis (Figure 4b). These results provide evidence that neither changes in eIF2α phosphorylation nor protein synthesis are mediated via changes in the activity of AMPK.
AMPK activation does not stimulate eIF2α phosphorylation
Investigation into the mechanism of PERK activation in response to glucose deprivation
Transducers of the UPR include two related ER transmembrane serine-threonine kinases: PERK and IRE1 (an RNase required for the splicing and activation of XBP-1. PERK and IRE1 are thought to be activated by a similar mechanism requiring homo-oligomerization, possibly initiated by the dissociation of BiP (immunoglobulin heavy-chain-binding protein) from their ER luminal domains [19,32,33]. As glucose deprivation leads to the activation of PERK, we investigated whether glucose deprivation also resulted in the activation of IRE1 (Figure 5a). Initially, we investigated the autophosphorylation state of IRE1 on Ser724, whose phosphorylation is required for IRE1 activation (Figure 5a, panel i). MIN6 cells were incubated in KRB or DMEM containing 0 or 20 mM glucose or treated with thapsigargin as a control (Figure 5a, panel i). IRE1 was more phosphorylated on Ser724 in MIN6 cells incubated at 20 mM glucose compared with that seen in glucose-deprived cells, indicating that IRE1 was possibly inactivated in glucose-deprived cells. To assay IRE1 activity, IRE1-dependent splicing of its substrate, the XBP-1 mRNA, was determined by RT (reverse transcription)–PCR (Figure 5a, panel ii). In this case, MIN6 cells were maintained in complete medium or incubated in either KRB (0 mM glucose), KRB supplemented with 0.5 mM glucose or KRB supplemented with 20 mM glucose for 2, 16 and 24 h. As a positive control, cells were also treated with thapsigargin for 2 or 4 h. As anticipated, treating cells with thapsigargin led to a large increase in the spliced form of the XBP-1 mRNA (Figure 5a, panel ii). The transfer of cells from complete medium (Figure 5a, panel ii, ‘c’) containing 25 mM glucose to KRB or KRB supplemented with 0.5 or 20 mM glucose led to a decrease in the amount of the spliced form of XBP-1 (Figure 5aii). However, in cells incubated in KRB supplemented with 20 mM glucose, an increase in the spliced form of XBP-1 was observed at 16 and 24 h. Therefore IRE1 is inactive in glucose-deprived cells or cells incubated at low glucose concentrations, whereas IRE1 is activated at high glucose concentrations. These results are in agreement with previous reports investigating IRE1 activation in response to differing glucose concentrations in islets . As IRE1 is not activated by glucose deprivation, whereas PERK is, the mechanism or threshold of activation of PERK must differ from that of IRE1.
PERK is activated by a mechanism independent of IRE1 activation and the accumulation of nascent unfolded proteins
PERK and IRE1 are classically activated by the accumulation of unfolded proteins within the ER, where ER client load exceeds the folding capacity of the ER. Therefore, to investigate whether the accumulation of nascent unfolded/incompletely processed proteins within the ER led to PERK activation under conditions of glucose deprivation, MIN6 cells were incubated in the presence of cycloheximide for 1 h (Figure 5b, Preinc) to inhibit protein synthesis and allow nascent proteins undergoing processing within the ER to be chased into mature processed proteins (Figure 5b). The cells were then further incubated in the presence of cycloheximide for 1 h in the absence or presence of 20 mM glucose and the levels of eIF2α phosphorylation were compared with cells in which cycloheximide had not been added. To ascertain that nascent proteins within the ER were fully processed during the 1 h pre-incubation with cycloheximide and that cycloheximide had effectively blocked further protein synthesis, [35S]methionine was added to the cells 15 min prior to the pre-incubation (P) and left on the cells during the pre-incubation period. The incorporation of [35S]methionine into proinsulin, insulin and total protein was assessed by SDS/PAGE (Figure 5bi). In the presence of cycloheximide, no increase in protein synthesis was detected (compare lanes 1–3). Additionally, the labelled proinsulin was no longer detectable after 1 h incubation with cycloheximide (compare lanes 1–3), indicating that the proinsulin had been fully processed to insulin and presumably chased out of the ER/cell (Figure 5b, panel i). At high glucose concentration, cycloheximide reduced eIF2α phosphorylation (Figure 5b, panel ii), indicating that ER client load and the accumulation of unfolded proteins may contribute to ER stress and the activation of PERK, as previously predicted . In contrast, in the absence of glucose, cycloheximide treatment had no significant effect on eIF2α phosphorylation (Figure 5b, panel ii). This indicates that the accumulation of unfolded/unprocessed nascent proteins is unlikely to account for the activation of PERK under conditions of glucose deprivation.
PERK-dependent eIF2α phosphorylation contributes to the suppression of protein synthesis at low glucose concentrations
To investigate the role of PERK-dependent eIF2α phosphorylation in the suppression of protein synthesis and proinsulin synthesis at low glucose concentrations, MIN6 cells were infected with an empty adenovirus (Ad-Empty), AdPERKΔC or an N-terminal truncation mutant of GADD34 (AdGADDΔN), which constitutively directs protein phosphatase-1 to eIF2α, leading to eIF2α dephosphorylation. At 48 h post-infection, the cells were incubated in KRB without glucose or containing 20 mM glucose in the presence or absence of thapsigargin as a positive control. The phosphorylation state of PERK and eIF2α was assessed by Western blotting (Figure 6a, panel i). In the control cells infected with Ad-Empty, glucose deprivation and thapsigargin treatment resulted in a decrease in PERK mobility on SDS/PAGE, consistent with PERK phosphorylation. This phosphorylation of PERK by either glucose deprivation or thapsigargin paralleled increases in the phosphorylation of eIF2α and decreases in the rate of protein synthesis compared with cells incubated in 20 mM glucose (Figure 6a, panels i and ii). Overexpression of PERKΔC or GADDΔN significantly inhibited eIF2α phosphorylation induced in response to glucose deprivation and thapsigargin treatment (Figure 6a, panel i) and, as expected, significantly overcame the inhibition of protein synthesis induced by thapsigargin (Figure 6a, panel ii). Importantly, overexpression of PERKΔC or GADDΔN increased protein synthesis by an average of 53% in glucose-deprived cells but still only partially overcame the inhibitory effects of glucose deprivation on protein synthesis. The overexpression of PERKΔC or GADDΔN had similar effects on the synthesis of proinsulin, indicating that eIF2α phosphorylation does not play a role in the specific up-regulation of proinsulin synthesis induced by glucose (Figure 6a, panel iii). Analysis of the polysome profiles from glucose-deprived MIN6 cells overexpressing PERKΔC or GADDΔN demonstrate that the inhibition of eIF2α phosphorylation leads to an increase in the amplitude of the polysomes but not in the number of ribosomes loaded on to mRNA (Figure 6b). One possible explanation is that the inhibition of eIF2α phosphorylation in glucose-deprived cells stimulates the recruitment of otherwise dormant mRNAs into polysomes.
The role of eIF2α phosphorylation in glucose-stimulated protein synthesis
In the present study, we present evidence that PERK is activated in pancreatic β-cells in response to a decrease in glucose concentration. This increase in PERK activity is evoked by a decrease in the energy status of the cell through a potentially novel mechanism, independent of the accumulation of nascent unfolded proteins within the ER and the activation of IRE1. We also demonstrate that the phosphorylation of eIF2α at low glucose concentrations plays an important role in suppressing global rates of protein synthesis in pancreatic β-cells.
The mechanism by which PERK is activated by decreased glucose concentration/energy status is unknown. Yet, it has previously been hypothesized that glucose starvation may, through a decrease in UDP-N-acetylhexosamines production, inhibit protein glycosylation, resulting in the accumulation of unprocessed proteins within the ER and the activation of PERK . Indeed, in a number of cell types, tunicamycin, an inhibitor of glycosylation, activates PERK and the UPR . However, the mitochondrial metabolism of pyruvate is sufficient to stimulate the dephosphorylation of eIF2α in glucose-deprived MIN6 cells, via a mechanism that is presumably independent of the hexosamine pathway (M. L. Powell and T. P. Herbert, unpublished work). Additionally, up to 4 h incubation of MIN6 cells with tunicamycin is unable to stimulate the phosphorylation of eIF2α (M. L. Powell and T. P. Herbert, unpublished work). Another possibility is that PERK activation is mediated through the inhibition of the proteasome, a high consumer of ATP, which in turn may lead to the inhibition of ERAD (ER-associated protein degradation) and the accumulation of unfolded proteins within the ER. In support of this, proteasome inhibition by MG-132 (the proteasome inhibitor carbobenzoxy-L-leucyl-L-leucyl-leucinal) stimulates eIF2α phosphorylation [35,36], although this has been reported to be via the activation of mGCN2 . Alternatively, energy status may directly lead to the inhibition of protein folding [37,38], as BiP binding to unfolded proteins is dependent on an ADP/ATP cycle in which cellular ATP is consumed . Yet, we provide evidence that PERK activation occurs via a mechanism independent of the accumulation of nascent unfolded proteins and processing intermediates in the ER (Figure 5b). On the other hand, it is possible that PERK is activated by the dynamic folding status of an ER-resident protein(s), a hypothesis that we currently favour. However, it remains unclear why PERK activation occurs independently of IRE1 activation in glucose-deprived β-cells.
Glucose causes a rapid increase in the rate of protein synthesis in pancreatic β-cells. However, the mechanism by which glucose does this is poorly understood. The inhibition of eIF2α phosphorylation stimulates a 53% increase in the rate of protein synthesis in glucose-deprived cells but is unable to restore protein synthesis to the rates seen in glucose-replete cells (Figure 6a, panel ii). Polysome analysis showed that inhibition of eIF2α phosphorylation in glucose-deprived cells led to an increase in the amplitude of the polysomal peaks but not in the size of the polysomes. One possible interpretation of these results is that eIF2α dephosphorylation mobilizes hitherto untranslated mRNAs on to polysomes.
We and others have shown that incubation of MIN6 cells/islets at low glucose concentration leads to an increase in the expression of proteins associated with the integrated stress response [23,25], which has been shown in other cell types to be dependent on eIF2α phosphorylation and important in preconditioning cells in becoming resistant to the effects of various cell stresses, including oxidative stress [18,20,39]. Therefore PERK-dependent phosphorylation of eIF2α at low glucose concentrations may play a role in protecting pancreatic β-cells against oxidative stress mediated by physiological rapid increases in glucose concentration.
The activation of PERK through a decrease in the energy status of the cell is unlikely to be a unique feature of the pancreatic β-cell, as hypoxia and ischaemia/reperfusion, both associated with decreases in intracellular energy status, lead to PERK-dependent phosphorylation of eIF2α in a number of other cell types [40–43].
The present study was supported by a Wellcome Trust Project Grant (awarded to T. P. H.). E. G. was supported by the Wellcome Trust. M. L. P. was supported by an MRC studentship. We also thank Professor Ronald Wek, Professor David Ron (The Kimmel Center for Biology and Medicine at the Skirball Institute, New York University School of Medicine, New York, NY, U.S.A.) and Professor Chris Proud for generously providing reagents.
AMP-activated protein kinase
immunoglobulin heavy-chain-binding protein
Dulbecco's modified Eagle's medium
eukaryotic initiation factor 2
foetal calf serum
growth-arrest and DNA-damage-inducible protein 34
general control non-derepressible 2
haem-regulated eIF2α kinase
inositol requiring enzyme 1
Krebs–Ringer bicarbonate buffer
methionyl initiator tRNA
mammalian orthologue of the yeast GCN2 (general control non-derepressible 2) protein kinase
mouse insulinoma cell line 6
protein kinase RNA activated
PKR-like ER kinase
small interfering RNA
unfolded protein response
X-box binding protein-1
These authors contributed equally to this work.