Hyperglycemia Mediates a Shift From Cap-Dependent to Cap-Independent Translation Via a 4E-BP1–Dependent Mechanism

Male (RFL12xBLK6) F1 mice expressing a bicistronic Renilla luciferase (LucR)–fibroblast growth factor (FGF-2) IRES–firefly luciferase (LucF) mRNA were treated with 50 mg/mL streptozotocin (STZ) for 5 days to induce diabetes. RFL12 mice have previously been described (24). Diabetes phenotype was confirmed by blood glucose concentrations >400 mg/dL. Four weeks after STZ treatment, phlorizin was administered subcutaneously twice daily for 7 days (11). Procedures were approved by the Pennsylvania State University College of Medicine Institutional Animal Care and Use Committee. Analysis of protein phosphorylation in liver homogenates and immunoprecipitations was performed as previously described (11). LucR and LucF activity was measured by Dual-Glo Luciferase Assay (Promega).

Primary cultures of hepatocytes were prepared from the liver of RFL12x-BLK6F1 and previously described Eif4ebp1 or Eif4ebp2 knockout mice (25) using the Worthington Hepatocyte Isolation System (Worthington Biochemical). Prior to plating, Eif4ebp1 or Eif4ebp2 knockout hepatocytes were transfected with a bicistronic reporter plasmid containing the vascular endothelial growth factor (VEGF) IRES using an Amaxa Hepatocyte Nucleofector kit (Lonza). Hepatocytes were seeded onto plates coated with rat tail collagen in Williams E Medium (Gibco) supplemented with 5% FBS (Atlas), 1% penicillin/streptomycin, 1 μmol/L DMSO, 4 μg/mL insulin, 2 mmol/L GlutaMAX (Gibco), and 15 mmol/L HEPES, pH 7.4. After 6 h, cells were transferred to Williams E Medium supplemented with 0.5% penicillin/streptomycin, 0.1 μmol/L DMSO, 1% ITS+ (Gibco), 2 mmol/L GlutaMAX, and 15 mmol/L HEPES, pH 7.4, containing either 11 or 33 mmol/L glucose.

Cultures of wild-type and Eif4ebp1;Eif4ebp2 double knockout mouse embryo fibroblasts (MEFs) were maintained in Dulbecco’s modified Eagle’s medium lacking sodium pyruvate and containing either 25 or 5 mmol/L glucose plus 20 mmol/L mannitol as an osmotic control supplemented with 10% FBS and 1% penicillin/streptomycin. Transfections were performed using Xtremegene HP (Roche). Where indicated, cells were treated with 3 mmol/L glucosamine (Sigma) or 50 nmol/L thiamet G (Cayman) to enhance protein O-GlcNAcylation.

Cultures of wild-type MEFs were maintained in Dulbecco’s modified Eagle’s medium designed for SILAC (Sigma). The medium was supplemented with 105 mg/L l-leucine, 146 mg/L l-lysine–HCl, 84 mg/L l-arginine, 10% dialyzed FBS (Sigma), and 1% penicillin/streptomycin. For the low-glucose condition, MEF cultures were maintained in 5 mmol/L glucose supplemented with 20 mmol/L mannitol and labeled for 6 h in medium supplemented as described above except that lysine was replaced with 4,4,5,5-D₄- l-lysine and arginine with l-arginine-¹³C₆. For the high-glucose condition, MEF cultures were maintained in 25 mmol/L glucose supplemented with 3 mmol/L glucosamine and labeled in medium containing l-lysine-¹³C₆¹⁵N₂ and l-arginine-¹³C₆¹⁵N₄. Cells were washed with PBS and harvested in 4% sodium dodecyl sulfate, 0.1 mol/L dithiothreitol, and 0.1 mol/L Tris-HCl pH 7.6. Aliquots of lysates (0.1 mg protein) from low and high glucose–treated cells were combined and prepared using filter-aided sample preparation methods as previously described (26) with the following modifications: washes were with 40% 2,2,2-trifluoroethanol (TFE) and 0.05 mol/L ammonium bicarbonate, pH 8.5; cysteine reduction was with 50 mmol/L tris(2-carboxyethyl)phosphine (TCEP), 40% TFE, and 0.05 mol/L ammonium bicarbonate, pH 8.5; and protein digestion was with endoproteinase LysC, trypsin, and 0.05 mol/L ammonium bicarbonate, pH 8.5.

The proteolytic digest was subjected to two-dimensional liquid chromatography–tandem mass spectrometry separation and analysis using an ABSciex 5600 TripleTOF mass spectrometer with a Nanoflow source and ProteinPilot 4.2 Beta analysis software. Initial SCX fractionation was performed as previously described (27,28). Each strong cation exchange (SCX) fraction was dried down and resuspended in 2% (v/v) acetonitrile and 0.1% (v/v) formic acid prior to reverse-phase C18 nanoflow–liquid chromatography separation on a cHiPLC Nanoflex system equipped with a trap column (200 µm × 0.5 mm Reprosil-Pur C18-AQ 3 µm 120 Å) and a separation column (75 µm × 15 cm Reprosil-Pur C18-AQ 3 µm 120 Å), using a 185-min gradient of 0.1% formic acid with increasing percentages of acetonitrile containing 0.1% formic acid, delivered into a 5600 TripleTOF mass spectrometer using a NanoSpray III source (ABSciex) with a 10-μm ID nanospray tip (New Objective, Woburn, MA). Combined MS and MS/MS spectra were analyzed by ProteinPilot 4.2 Beta software searched against the complete Mouse RefSeq database from NCBI (27,303 sequences total, from 4 February 2012, concatenated to a reversed sequence Decoy database derived from the same RefSeq database, plus a list of 389 common laboratory contaminants). Protein identifications were accepted if they had an estimated local false discovery rate of <5%, using the Proteomics System Performance Evaluation Pipeline algorithm to calculate the false discovery rates (29).

The raw data obtained from the pulsed stable isotope labeling by amino acids in cell culture (pSILAC) analysis was uploaded and analyzed using proteome software Ingenuity Pathways Analysis (IPA) (www.ingenuity.com). A medium:high (M:H) cutoff of 0.5–1.5 was set to identify proteins whose synthesis rates were differentially regulated.

The data are expressed as means + SE. Statistical analysis is described in the figure legends.

In the current study, we used a transgenic mouse model that expresses a bicistronic mRNA containing two open reading frames (ORFs) encoding two distinct luciferase enzymes: LucR and LucF. Translation of the first ORF (LucR) occurs in a cap-dependent manner, whereas translation of the second ORF (LucF) is driven by the FGF-2 IRES that is located between the two cistrons (Fig. 1A). Similar transgenic mice have been used to demonstrate that diabetes leads to upregulated expression of FGF-2 by increased utilization of the FGF-2 IRES in the aorta (30). Four weeks after the induction of diabetes, blood glucose levels were raised from 189 ± 12 to 540 ± 26 mg/dL and the activity of the LucR reporter was reduced by 62% in the liver of mice with STZ-induced type 1 diabetes compared with nondiabetic littermates (Fig. 1B), whereas LucF activity was elevated by 192% (Fig. 1C). As a result, the relative LucF-to-LucR activity ratio was enhanced by 633% (Fig. 1D). To investigate a potential mechanism for mediating the shift from cap-dependent to cap-independent translation in the liver of STZ-treated mice, the interaction of 4E-BP1 and eIF4G with eIF4E was evaluated. The amount of 4E-BP1 bound to eIF4E was elevated in the liver of diabetic compared with control mice (Fig. 1E), whereas the amount of eIF4G bound to eIF4E was reduced (Fig. 1F). These changes were accompanied by decreased 4E-BP1 phosphorylation (Fig. 1G) and increased O-GlcNAcylation (Fig. 1H). Whereas the expression of eIF4E in the liver of STZ-treated and control mice was not different (Fig. 1I), total hepatic 4E-BP1 expression was increased (Fig. 1J) and likely contributed to increased interaction with eIF4E.

The role of hyperglycemia in mediating the diabetes-induced shift toward cap-independent translation was assessed using phlorizin treatment, which rapidly lowers blood glucose concentrations by blocking intestinal glucose absorption and producing renal glucosuria (31). Phlorizin treatment reduced the nonfasting blood glucose of mice with STZ-induced diabetes from a concentration of 539 ± 22 to 300 ± 16 mg/dL (Fig. 2A). LucR activity was repressed in the liver of diabetic compared with control mice, and upon phlorizin treatment its activity was elevated by 47% (Fig. 2A). Whereas LucF activity was elevated by 99% in diabetic compared with control mice, after treatment with phlorizin LucF activity was elevated by only 42% (Fig. 2C). Lowering blood glucose levels reduced the relative LucF-to-LucR ratio observed in the liver of diabetic mice by 52% (Fig. 2D). Phlorizin treatment of diabetic mice also reduced the amount of 4E-BP1 in eIF4E immunoprecipitates (Fig. 2E) and concomitantly elevated the interaction of eIF4E with eIF4G (Fig. 2F). The reduced interaction of 4E-BP1 with eIF4E upon phlorizin treatment was likely in part due to lower 4E-BP1 expression (Fig. 2G) as well as increased phosphorylation (Fig. 2H) and decreased O-GlcNAcylation (Fig. 2I) of 4E-BP1 compared with untreated diabetic mice. Similar to previous findings with Ins2^Akita/+ diabetic mice (11), phlorizin treatment attenuated the elevated hepatic expression of glutamine-fructose-6-phosphate amidotransferase (GFAT) (Fig. 2J) but did not alter total eIF4E expression (Fig. 2K).

GFAT is the rate-limiting enzyme of the HBP and as such regulates the production of UDP N-Acetylglucosamine, which serves as the substrate for protein O-GlcNAcylation. Changes in GFAT expression suggest that flux through the HBP plays an important role in mediating the effects of hyperglycemia on 4E-BP1 O-GlcNAcylation and expression. For further evaluation of the role of hyperglycemia and the HBP in cap-dependent and cap-independent translation, hepatocytes isolated from the liver of the transgenic bicistronic reporter mice were incubated in medium containing either 11 or 33 mmol/L glucose to match blood glucose concentrations observed in control and STZ-treated mice. Hyperglycemic conditions dramatically enhanced the relative cap-independent translation (Fig. 3A) and 4E-BP1 expression (Fig. 3B) without altering eIF4E expression (Fig. 3C). Immunoprecipitation of eIF4E revealed elevated interaction of 4E-BP1 with eIF4E in hepatocytes maintained in the presence of high compared with low glucose (Fig. 3D). When hepatocytes maintained in low glucose medium were treated with glucosamine to directly stimulate the HBP or thiamet G, a potent and selective inhibitor of O-GlcNAcase that dramatically elevates protein O-GlcNAcylation (32) by bypassing GFAT, total protein O-GlcNAcylation was enhanced (Fig. 3E). Similarly, both glucosamine and thiamet G enhanced the LucF-to-LucR activity ratio in hepatocytes when maintained in low glucose medium (Fig. 3F). Both glucosamine and thiamet G enhanced expression of 4E-BP1 (Fig. 3G) but not eIF4E (Fig. 3H), and immunoprecipitation of eIF4E revealed elevated interaction of 4E-BP1 with eIF4E (Fig. 3I). Overall, these findings support the conclusion that the hyperglycemia-induced shift from cap-dependent to cap-independent luciferase reporter activity was mediated by changes in protein O-GlcNAcylation that coincide with alterations in the expression of 4E-BP1 and its interaction with eIF4E.

The role of hyperglycemia in modulating the shift from cap-dependent to cap-independent translation was further assessed using MEFs transiently transfected with bicistronic luciferase constructs in which the activity of LucR occurs in a cap-dependent manner and the activity of LucF is regulated by either the FGF-2 or the VEGF. Under high- compared with low-glucose conditions, the ratio of LucF to LucR activity increased by 20% for the FGF-2 IRES and 77% for the VEGF IRES (Fig. 4A). Further, the ratio of LucF to LucR activity with the VEGF IRES also increased by 79% when cells maintained in low-glucose medium were treated with glucosamine (Fig. 4B). When wild-type and 4E-BP1/2 knockout (Eif4ebp1;Eif4ebp2) MEFs were analyzed after transient transfection with the bicistronic luciferase constructs containing the VEGF IRES, the expression of 4E-BP1 was increased by 72% in cells maintained in medium containing high glucose and treated with glucosamine compared with the low-glucose condition (Fig. 4C). There was no detectable 4E-BP1 under either condition in knockout cells. Exposure to high glucose and glucosamine increased the ratio of LucF to LucR activity in wild-type cells to 333% that of the low-glucose condition. In contrast, there was no effect on the ratio of LucF to LucR activity with treatment of 4E-BP1/2 knockout cells (Fig. 4D). For confirmation of the role of 4E-BP1 specifically, hepatocytes from 4E-BP1/2 knockout mice were transfected with a bicistronic plasmid containing the IRES for VEGF and maintained in medium containing either low glucose or high glucose with glucosamine. There was no significant difference in the ratio of LucF to LucR activity under the high- and low-glucose conditions for 4E-BP1 knockout hepatocytes, whereas high glucose increased the ratio of LucF to LucR activity in 4E-BP2 knockout hepatocytes to 234% that of the low-glucose condition (Fig. 4E). These results demonstrate that expression of 4E-BP1 but not 4E-BP2 is necessary for the hyperglycemia-induced shift from cap-dependent to cap-independent translation.

To identify hyperglycemia-induced changes in the translational control of protein expression on a global scale, we used pSILAC to evaluate newly synthesized proteins and quantitate changes in their accumulation rates in MEFs in the presence of either low glucose or high glucose supplemented with glucosamine, as these conditions produced the largest increase in the LucF-to-LucR ratio (Fig. 4B). In total, 1,355 and 2,905 proteins were identified with high confidence in two independent runs. While the majority of proteins had medium-to-heavy ratios of ~1:1, indicating that their translation was not significantly altered by hyperglycemic conditions, the synthesis of a number of proteins was significantly altered between the two conditions. The top-scoring proteins whose synthesis was either upregulated or repressed under high- compared with low-glucose condition are listed in Tables 1 and 2, respectively.

In the classic model of cap-dependent translation initiation, ribosome scanning of the 5′-UTR is impaired by strong secondary structures. Characteristics that decrease the efficiency of cap-dependent translation include the length of the 5′-UTR and the complexity of secondary structure (33). Thus, we predicted that the high-glucose condition would repress the translation of mRNAs with less complex 5′-UTRs in favor of mRNAs with more structured 5′-UTRs. Most cellular IRES elements are 150–300 nucleotides in length, although some are as short as 22 nucleotides (34). The mRNA-encoding proteins whose synthesis was upregulated by high glucose averaged 185.3 nucleotides in length and were 74% longer than the average number of nucleotides in 5′-UTRs of mRNAs encoding proteins whose synthesis was downregulated by high glucose (Fig. 5A). To investigate the complexity of these 5′-UTRs, we used mfold (http://mfold.rna.albany.edu/?q=mfold) to compute secondary structures and calculate a corresponding free energy change for folding (ΔG). The top-scoring proteins whose synthesis was downregulated by high-glucose conditions have an average ΔG of −41.1 kcal/mol, whereas the top-scoring proteins whose synthesis was upregulated by high-glucose conditions have a significantly lower average ΔG value of −77.5 kcal/mol, indicative of more stable secondary structures (Fig. 5B). To verify that the hyperglycemia-induced enhancement in the synthetic rates of proteins with longer 5′-UTRs was not the result of elevated mRNA transcription, we evaluated the abundance of mRNAs corresponding to six proteins from Table 1 with the longest 5′-UTRs (Fig. 5C). Of the six analyzed, only one exhibited increased abundance, and that was 50-fold less than the accumulation of newly synthesized protein. Changes in protein degradation may also impact the M-to-H ratio observed during pSILAC; however, its contribution is relatively small (35). These findings suggest that hyperglycemia favors the translation of messages with longer/more highly structured 5′-UTRs. Although no proteins encoded by mRNAs containing known IRES structures were identified among the top-scoring upregulated proteins, several proteins were identified whose mRNAs are listed on IRESite, the database of experimentally verified IRES structures (36). Among these, cellular inhibitor of apoptosis 1 (c-IAP1) (gi|133922596), heat shock 70-kDa protein 1A (HSPA1A) (gi|40254361), cold shock domain containing E1 (UNR) (gi|240255574), and utrophin A (UTRA) (gi|110431378) were all elevated by nearly twofold in cells exposed to high-glucose conditions in two independent runs, while BCL2 (gi|6753200) increased by nearly 80-fold in the second run but was not detected in the initial run.

Functional relationships between genes that were differentially expressed in low and high glucose were investigated with IPA. IPA showed that the proteins whose accumulation was upregulated are associated with pathways that regulate eIF2, eIF4F, and S6K1 signaling; protein ubiquitination; and aminoacyl-tRNA biosynthesis (Table 3). Further, the most significant toxicologic list pathways were hypoxia-inducible factor (HIF) signaling, oxidative stress mediated by nuclear factor erythroid 2–related factor 2 (Nrf2), mitochondrial dysfunction, and DNA damage checkpoint regulation of cell cycle phase G2/M (Table 3). Both HIF-1 and Nrf2 contain IRES sequences in their 5′-UTR that allow their translation to be maintained under stress conditions that are inhibitory to cap-dependent translation (6,37). IPA was also used to predict which mRNAs encoding transcription factors were most likely responsible for the altered pattern of expression under low- and high-glucose conditions. The top transcription factor altered by high glucose was the proto-oncogene myc (Table 3), whose 5′-UTR contains an IRES. Activation of myc was predicted under high-glucose conditions based on a broad network of interacting proteins (Supplementary Fig. 1).

The findings of the current study provide insight into a novel mechanism through which diabetes-induced hyperglycemia alters the selection of mRNA for translation. In the liver of diabetic mice, 4E-BP1 exhibited reduced phosphorylation, elevated O-GlcNAcylation, and enhanced interaction with eIF4E compared with nondiabetic mice. These alterations in 4E-BP1 were associated with a shift from cap-dependent to cap-independent translation using bicistronic luciferase reporter assays. For demonstration of the component of the diabetic state responsible for this shift, diabetic mice were treated with phlorizin to reduce blood glucose concentrations. When the blood glucose concentration was lowered, the interaction of 4E-BP1 with eIF4E returned to nondiabetic levels and the shift from cap-dependent to cap-independent reporter activity was reversed. A similar shift toward cap-independent mRNA translation was observed with cells in culture upon exposure to hyperglycemic conditions or under conditions that promoted protein O-GlcNAcylation. O-GlcNAcylation of 4E-BP1 correlated with the hyperglycemia-induced shift from cap-dependent to cap-independent translation, and expression of 4E-BP1 was necessary for this effect. We extended these findings using pSILAC to identify novel proteins that undergo altered rates of synthesis in high- versus low-glucose conditions. Examination of the expression pattern of these proteins revealed a hyperglycemia-induced shift toward proteins with more complex 5′-UTRs. Overall, the results are consistent with a model wherein the O-GlcNAcylation of 4E-BP1 results in elevated expression and promotes its interaction with eIF4E (11), thereby altering gene expression in response to hyperglycemia and diabetes.

Regulation of the initiation complex eIF4F by posttranslational modification is of critical importance in the selection of mRNAs for cap-dependent translation initiation. Binding of hypophosphorylated 4E-BP1 to eIF4E, which is mutually exclusive of interaction with eIF4G, prevents assembly of functional eIF4F complexes to repress loading of ribosomes onto the mRNA 5′-cap. The interaction of 4E-BP1 with eIF4E is not only regulated by phosphorylation; O-GlcNAcylation of 4E-BP1 also enhances its binding to eIF4E independent of 4E-BP1 phosphorylation status (10). Thus, hyperglycemia-induced sequestration of eIF4E potentially downregulates the synthesis of a broad array of proteins. However, under conditions where eIF4E is limiting, the same canonical initiation factors can facilitate the cap-independent loading of ribosomes onto mRNA that contain IRES elements (8). Therefore, by reducing the competition from capped mRNAs, hyperglycemia-induced sequestration of eIF4E potentially acts as a positive regulator of cap-independent translation in response to conditions of stress.

A critical role for 4E-BP1 in upregulating cap-independent translation has previously been demonstrated (38). Viral IRES-mediated translation is promoted when cellular cap-dependent translation is diminished under conditions of cellular stress (38). However, 4E-BP1 knockdown prevents upregulation of viral cap-independent translation under conditions of amino acid starvation (39). Furthermore, 4E-BP1 plays a crucial role in mediating differential protein expression during hypoxia (40). Overexpressed 4E-BP1 and eIF4G mediate a hypoxia-activated switch from cap-dependent to cap-independent mRNA translation that promotes increased tumor angiogenesis and growth (41). It has been previously demonstrated that the presence of 4E-BP1/2 is also necessary for increased expression of VEGF in both the retina during diabetes and cells maintained under hyperglycemic conditions (42). The results of the current study suggest that increased VEGF expression occurs through increased use of internal ribosome entry sites in response to increased flux through the HBP. This finding likely extends to other mRNAs that are translated through an IRES-dependent mechanism, as overexpression of O-GlcNAcylation transferase produces an accumulation of 80S monosomes, suggesting that excessive O-GlcNAcylation suppresses translation initiation (23).

To explore the effect of hyperglycemia on mRNA translation, we used pSILAC to identify proteins that undergo altered rates of synthesis. Hyperglycemia promoted translation of mRNAs that contained long and structured 5′-UTRs, a characteristic associated with mRNAs containing IRES. This observation does not directly indicate that hyperglycemic conditions favor translation of all mRNAs with IRES domains. Instead, it is more likely that hyperglycemia-induced modification of the translational machinery enhances translation of a specific subset of messages, some of which contain IRES domains. In this study, hyperglycemia specifically enhanced IRES-mediated translation from luciferase reporters driven by either the FGF-2 or VEGF IRES sequence. Using pSILAC, we also observed increased translation of the following cellular mRNAs that have been previously shown to contain IRES structures: B-cell lymphoma 2 (BCL2), 78-kDa glucose-regulated protein (BiP), c-IAP1, HSPA1A, Runt-related transcription factor 1 (RunX1), UNR, and UTRA. IRES containing mRNAs often encode proteins with crucial biological functions, such as the major vascular growth factors VEGF, FGF-2, platelet-derived growth factor, and RunX1. Thus, enhanced hyperglycemia-mediated translation of angiogenic proteins potentially disrupts the relative balance of inducers and inhibitors of angiogenesis to promote the neurovascular complications associated with diabetes. Furthermore, our data support a previous microarray analysis of the retinal transcriptome where normalization of systemic glycemia in diabetic rats primarily restored pathways associated with growth factor signaling (43).

A functional link between the metabolic abnormalities associated with disease and regulation of IRES sequences has previously been proposed (44), yet direct evidence remains limiting. In the current study, we find that hyperglycemia and conditions that promote protein O-GlcNAcylation enhance IRES-mediated mRNA translation through a mechanism that is dependent on 4E-BP1. Thus, regulation of 4E-BP1 expression by O-GlcNAcylation represents a novel mechanism for altering the selection of mRNAs for translation under pathophysiological conditions. Elucidation of molecular mechanisms underlying cap-independent translation initiation will not only enhance our understanding of gene expression but also impact the development of treatment strategies for addressing the pathophysiology of diabetes.

The research reported here was supported by National Institutes of Health grants DK088416 (to M.D.D.) and DK13499 (to L.S.J.).

No potential conflicts of interest relevant to this article were reported.

M.D.D. researched data and wrote and edited the manuscript. J.S.S. contributed to discussion and reviewed the manuscript. B.A.S. researched data, wrote the manuscript, contributed to discussion, and reviewed and edited the manuscript. S.R.K. and L.S.J. designed experiments, contributed to discussion, and reviewed and edited the manuscript. M.D.D. is the guarantor of this work and as such had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Parts of this study were presented in abstract form at the 72nd Scientific Sessions of the American Diabetes Association, Philadelphia, Pennsylvania, 8–12 June 2012.

The authors thank Dr. N. Sonenberg (McGill University) for kindly providing 4E-BP1/2 DKO MEFs and Dr. A.-C. Prats (INSERM) for providing the bicistronic reporter containing the FGF-2 IRES. The authors also thank W. Dunton (Pennsylvania State University College of Medicine) and M. Wu (Pennsylvania State University College of Medicine) for assistance with animals.