Site Specific Activation of AKT Protects Cells from Death Induced by Glucose Deprivation

Abstract

The serine/threonine kinase AKT is a key mediator of cancer cell survival. We demonstrate that transient glucose deprivation modestly induces AKT phosphorylation at both Thr308 and Ser473. In contrast, prolonged glucose deprivation induces selective AKTThr308 phosphorylation and phosphorylation of a distinct subset of AKT downstream targets leading to cell survival under metabolic stress. Glucose deprivation-induced AKTThr308 phosphorylation is dependent on PDK1 and PI3K but not EGFR or IGF1R. Prolonged glucose deprivation induces the formation of a complex of AKT, PDK1, and the GRP78 chaperone protein, directing phosphorylation of AKTThr308 but AKTSer473. Our results reveal a novel mechanism of AKT activation under prolonged glucose deprivation that protects cells from metabolic stress. The selective activation of AKTThr308 phosphorylation that occurs during prolonged nutrient deprivation may provide an unexpected opportunity for the development and implementation of drugs targeting cell metabolism and aberrant AKT signaling.

Introduction

Cancer cells frequently encounter metabolic stress, including an insufficient supply of nutrients and oxygen because of inadequate neovascularization or during metastasis and therapy (1, 2). The serine/threonine kinase AKT (aka protein kinase B) mediates survival of both normal and malignant cells (3, 4). Aberrant AKT activation is prevalent across multiple cancer lineages as a consequence of amplification or mutation of receptor tyrosine kinases (RTKs), PI3K, or AKT and inactivation of the phosphatase and tensin homolog (PTEN) or inositol polyphosphate-4-phosphatase (INPP4B) (5, 6).

Glucose deprivation, a common form of metabolic stress, can induce AKT activation (7, 8). We thus investigated the effects of prolonged glucose deprivation on AKT phosphorylation and activation and the role of AKT activation on subsequent cell survival.

Results

Transient and Prolonged Glucose Deprivation Induces AKT Phosphorylation and Activation via Distinct Mechanisms

We investigated the effects of glucose deprivation mimicking metabolic stress cancer cells encounter in vivo on AKT phosphorylation and activation. In HeLa cells, transient glucose deprivation (2–4 h) induced a modest increase in both pAKTThr308 (5 to 6-fold) and pAKTSer473 (2 to 5-fold) (Figure 1A). Unexpectedly, prolonged glucose deprivation (16h) induced a marked increase in pAKTThr308 (20- to 30-fold) but only a modest increase in pAKTSer473 (2- to 3-fold) (Figure 1A). pAKTThr308 increased over 16h of glucose deprivation, whereas pAKTSer473 peaked at about 6h and subsequently declined. Thus, AKTThr308 and AKTSer473 phosphorylation were uncoupled, contrasting with growth factor stimulation wherein AKTThr308 and AKTSer473 phosphorylation were induced in parallel (Figure S1A). Apparently, two independent processes were activated during glucose deprivation: one wherein transient glucose deprivation induces coordinate AKTThr308 and AKTSer473 phosphorylation and a second delayed process that results in selective pAKTThr308 accumulation. Phosphorylation of ribosomal protein S6, which is downstream of AKT and AMP-activated protein kinase (AMPK), a sensor of cellular AMP levels (9, 10), markedly decreased within 1h of glucose deprivation, reaching a nadir at about 2h, and returning to greater than basal levels by 10h (Figure 1A), a time course distinct from that of AKT phosphorylation. Phosphorylation of AMPK and its substrate ACC, indicative of AMPK activity (11, 12), increased markedly after 3h of glucose deprivation (Figure 1B), consistent with transient glucose deprivation activating the AMPK pathway and, in turn, inhibiting mTORC1 as indicated by transient decrease in phosphorylation of p70 ribosomal S6 kinase (p70S6K) and S6 (Figure 1A and 1B). Inhibition of mTORC1 can activate AKT via a p70S6K-mediated feedback loop (13, 14). Our results are consistent with a model wherein transient glucose deprivation increases the AMP/ATP ratio activating AMPK, which inhibits mTORC1 and p70S6K, and subsequently induces AKTThr308 and AKTSer473 phosphorylation via the mTORC1 mediated feedback loop.

Figure 1. AKT phosphorylation and activation induced by glucose deprivation.

A, AKTThr308, AKTSer473, and S6 phosphorylation under glucose-deprivation conditions. HeLa cells were serum-starved for 16h and glucose-deprived for the indicated periods. Cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors. Lysates (50 μg/lane) were resolved on 10% SDS-PAGE gels. Erk immunoblotting was used as a loading control. Scanning densitometric values for Western blots were obtained using the ImageJ software program (version 1.63.1; National Institutes of Health, Bethesda, MD). The pAKTThr308, pAKTSer473, and pS6 levels were normalized to the loading control and presented as relative conversion to values in control cells without glucose deprivation (first lane) in the bar graphs. B, phosphorylation of AKT and AMPK substrates under glucose-deprivation conditions. HeLa cells were serum-starved for 16h and deprived of glucose for 0, 3, or 16h. Scanning densitometric values of the bands are presented as relative conversion to values in control cells without glucose deprivation. Intervening lanes not relevant to this study were removed on the pAMPK and p70S6K blots. C, schematic diagram of signaling upstream of and downstream from AKT. D, generalization of prolonged glucose deprivation-induced selective AKTThr308 phosphorylation. HEK293 and NCI-H358 cells were serum-starved for 16h and glucose-deprived for the indicated periods. MDA-MB-468, Skov-3, 786-0, and HCT116 cells were serum-starved and glucose-deprived for the indicated periods. GAPDH was used as a loading control.

Strikingly, under prolonged glucose-deprivation conditions (16h), pAMPK and pACC returned to basal levels, indicating that energy-sensing signaling was turned off, despite continued low ATP levels (see figure 3A) (Figure 1B). Lack of LKB1 in HeLa may contribute to low levels of AMPK activation under prolonged glucose-deprivation conditions. p70S6K and S6 phosphorylation increased to higher than basal levels consistent with relief of AMPK-mediated mTORC1 inhibition (Figs. 1A and 1B). This suggests that under prolonged glucose deprivation conditions, AKT phosphorylation as a consequence of mTORC1 inhibition is no longer active and that a mechanism distinct from the mTOR feedback loop selectively increases AKTThr308 but not AKTSer473 phosphorylation.

Figure 3. Prolonged glucose deprivation causes energy depletion and enhances AKT association with GRP78 and PDK1 contributing to selective AKTThr308 phosphorylation.

A, energy depletion is responsible for AKTThr308 phosphorylation induced by prolonged glucose deprivation. HCT116 cells were serum-starved and glucose-deprived with or without the addition of MP (10 mM) to cell culture medium for 24h. ATP levels in the cells were measured using a Biovision ATP assay kit and presented as relative conversion to values in control cells without glucose deprivation and MP addition (first lane). Three independent experiments were performed. The error bars indicate standard deviations. B, AKT activation induced by prolonged glucose deprivation is independent of autophagy. ATG5+/+ and ATG5−/− MEFs were serum-starved for 16h with or without glucose deprivation. Intervening lanes not relevant to this study were removed. C, prolonged glucose deprivation enhances AKT association with PDK1. HeLa cells were transfected to express GFP-PDK1 and then serum-starved for 16h with or without glucose deprivation. Co-IP of AKT and GFP-PDK1 was performed with a rabbit antibody against total AKT, and Western blotting was performed with a mouse antibody against GFP. Data are representative of three independent experiments. D, prolonged glucose deprivation induces GRP78 protein levels and enhances GRP78 association with AKT and PDK1. For Co-IP of AKT and GRP78, HeLa cells were serum-starved for 16h with or without glucose deprivation (lane 1 and 2) or with insulin stimulation (20μg/ml) for 10 minutes (lane 3). IP of AKT was performed with a mouse antibody against AKT, and Western blotting was performed with a goat GRP78 antibody and a rabbit AKT antibody. For Co-IP of PDK1 and GRP78, HeLa cells were transfected to express GFP-PDK1 and then serum-starved for 16h with or without glucose deprivation (lane 1 and 2) or with insulin stimulation (20μg/ml) for 10 minutes (lane 3). GFP-PDK1 was immunoprecipitated by GFP-Trap. The arrow designates GRP78. The arrowhead designates likely nonspecific bands. Data are representative of three independent experiments. E, knockdown of GRP78 decreases AKTThr308 phosphorylation induced by prolonged glucose deprivation. HeLa cells were infected to stably express control or GRP78 shRNA and then serum-starved for 16h with or without glucose deprivation. The arrow designates GRP78. The arrowhead designates likely nonspecific bands. Data are representative of three independent experiments. F, inhibition of GRP78 abolishes AKTThr308 phosphorylation and activation induced by prolonged glucose deprivation. HeLa cells were serum-starved for 16h with or without glucose deprivation then treated with DMSO control or VER-155008 (100μM) for 3h before harvesting. Intervening lanes not relevant to this study were removed.

Due to the discrepancy between the time course and magnitude of AKTThr308, AKTSer473 and downstream S6 phosphorylation, we sought to determine whether selective AKTThr308 phosphorylation during prolonged glucose deprivation activated AKT. As indicated in Figure 1B, several time dependent phosphorylation patterns emerged. Forkhead box O3A (FOXO3A), glycogen synthase kinase 3 (GSK3), mTOR, and Y-box binding protein 1 (YB1) phosphorylation time courses were consistent with AKTThr308 phosphorylation with a gradual increase over time. Phosphorylation of several other AKT substrates, including proline-rich AKT/PKB substrate 40 kDa (PRAS40) and BCL-2 antagonist of cell death (BAD), was not altered by transient or prolonged glucose deprivation (Figure 1B). Phosphorylation of a subset of proteins, as detected by an anti-phospho-AKT substrate antibody, was markedly decreased whereas phosphorylation of others was increased with prolonged glucose deprivation. These findings indicated that prolonged glucose deprivation induces AKTThr308 phosphorylation leading to subsequent selective phosphorylation of a specific group of downstream substrates.

As summarized in Figure 1C, S6 phosphorylation is downstream of the tuberous sclerosis protein (TSC) 1/2 complex, which is negatively regulated by AKT and also dominantly and positively regulated by AMPK (9, 10). Due to the inhibition of Ras homolog enriched in brain (Rheb) by the TSC complex, AKT activation increases and AMPK activation decreases S6 phosphorylation through mTORC1 and p70S6K (9, 10). p70S6K activation, in turn, feedback inhibits IRS1 and thus AKT phosphorylation.

Selective AKTThr308 phosphorylation induced by prolonged glucose deprivation was generalizable, as this process was recapitulated in multiple breast (MDA-MB-468 and HCC1806), lung (NCI-H358, H226, and A549), ovarian (Skov-3, A2780, and Caov-3), kidney (786-0, ACHN, and RXF393), and colon (HCT116) cancer cell lines as well as normal HEK293 cells (Figure 1D; Figure S1B). These cell lines have different genetic events targeting the PI3K/AKT and LKB1/AMPK pathways including PTEN loss (MDA-MB-468, 786-0, A2780, and RFX393), PIK3CA-activating mutations (HCT116 and Skov-3), and LKB1 mutations (HeLa, A549, and Caov-3). Therefore, glucose deprivation-induced AKTThr308 phosphorylation represents a generalizable process that occurs independently of PTEN, PI3K, or LKB1 mutation status. However, in a small subset of lines including MDA-MB-231, which has mutations in BRAF, KRAS, and TP53, changes in pAKTThr308 were modest to absent suggesting that the degree of change in AKTThr308 phosphorylation may exhibit context dependent effects based on mutational background. Some lines such as H358 show prolonged but modest AKTSer473 phosphorylation following glucose deprivation. However, in all cases increases in pAKTSer473 were markedly less than the induction of pAKTThr308 at late time points (Figure 1D; Figure S1B).

To determine whether the effects of glucose deprivation were due to alterations in osmotic pressure or other physical properties of the medium, we cultured cells with D-glucose or L-glucose, which cannot be used in glycolysis or as an intracellular building block. D-glucose returned AKT phosphorylation to baseline (Figure S1C). In contrast, L-glucose despite reconstituting osmotic pressure and other physical properties of the medium failed to alter pAKTThr308 indicating that AKT phosphorylation occurs independent of any physical features of the medium that are altered by glucose deprivation.

AKTThr308 Phosphorylation Induced by Prolonged Glucose Deprivation Is Dependent on PI3K and PDK1 but not EGFR or IGF1R activity

We next determined the role of upstream components in the RTK/PI3K/PDK1 cascade in selective AKTThr308 phosphorylation induced by prolonged glucose deprivation. In PDK1−/− HCT116 and DLD1 colon cancer cells (15), glucose deprivation-induced pAKTThr308 was abolished, compatible with PDK1 contributing to AKTThr308 phosphorylation following glucose deprivation (Figure 2A). Elevated pAKTSer473 was observed under both basal and glucose deprivation conditions in DLD1 PDK1−/− cells, likely due to consequence of feedback loops downstream of PDK1. Thus AKTThr308 phosphorylation is not a precondition for AKTS473 phosphorylation in DLD1 PDK1−/− cells. Furthermore, in HeLa cells, GSK2334470, a highly specific and potent PDK1 inhibitor (16), GDC-0941, a dual inhibitor of PI3K and mTORC1 (17), and MK-2206, an allosteric AKT inhibitor (18), all abolished prolonged glucose deprivation-induced AKTThr308 phosphorylation. In contrast, rapamycin had no effect on prolonged glucose deprivation-induced pAKTThr308. AZD8055, an inhibitor of both mTORC1 and mTORC2 (19), abolished pAKTSer473 and significantly inhibited pAKTThr308 induced by prolonged glucose deprivation. U0126, an MEK inhibitor (20), had no effect on pAKTThr308 induced by prolonged glucose deprivation (Figure 2B). Similar effects were seen in HCT116 cells (Figure S2), indicating that inhibitor effects on AKT phosphorylation were not cell line specific. Thus PDK1 and PI3K but not MEK1 are likely required for selective AKTThr308 phosphorylation. The effects of MK2206 likely reflect a lack of association of AKT with membrane PtdInsP3 rather than a requirement for AKT catalytic activity (see below). The lack of effect of rapamycin in the light of the effects of the catalytic TOR inhibitor (AZD8055) will require further investigation but suggests that TORC1 activity may not be required for selective AKTThr308 phosphorylation.

Figure 2. Signal components required or dispensable for AKTThr308 phosphorylation induced by prolonged glucose deprivation.

A, PDK1 is required for AKTThr308 phosphorylation induced by prolonged glucose deprivation. PDK1+/+ and PDK1−/− HCT116 cells as well as PDK1+/+ and PDK1−/− DLD1 cells were serum-starved for 24h with/without glucose deprivation for 24h. B, PI3K but not mTORC1 or MEK is required for AKTThr308 phosphorylation induced by prolonged glucose deprivation. HeLa cells were serum-starved and glucose-deprived and treated with the indicated kinase inhibitors at the indicated concentrations for 16h. DMSO was used as a control. Data are representative of three independent experiments. Intervening lanes on the same blot not relevant to this study were removed. C, EGFR and IGF1R are dispensable for AKTThr308 phosphorylation induced by prolonged glucose deprivation. HCT116 cells were serum-starved and glucose-deprived and treated with the indicated drugs (1μM) for 24h. D, an intact PH domain of AKT is required for AKTThr308 phosphorylation induced by prolonged glucose deprivation. HeLa cells were transfected to express VN-AKT1 or the VN-AKT1 (R25A) mutant and serum-starved and glucose-deprived for 16h. Arrowhead, VN-AKT1; arrow, endogenous AKT. E, prolonged glucose deprivation does not induce detectable AKT translocation to the cell membrane. HeLa cells were transfected to express the AKT1 PH domain fused to enhanced GFP. Cells were then serum-starved, stimulated with insulin (20 μg/ml for 5 min), or glucose-deprived (16h). At least 100 cells in different fields for each sample were examined. Representative images from three independent experiments are shown. F, Prolonged glucose deprivation does not increase cellular PtdInsP3 levels. HeLa cells were serum-starved, stimulated with insulin (20 μg/ml for 5 min), or glucose-deprived (16h). PtdInsP3 levels were examined by immunofluorescence. At least 100 cells in different fields for each sample were examined. Representative images from three independent experiments are shown. G, AKT kinase activity is dispensable for AKTThr308 phosphorylation induced by prolonged glucose deprivation. AKT1−/− and AKT2−/− HCT116 cells were transfected to express the HA-AKT1 (K179M) mutant. pcDNA3 transfection was used as a control. Cells were then serum-starved and glucose-deprived for 24h. Intervening lanes not relevant to this study were removed.

Intriguingly, although EGFR phosphorylation increased following prolonged glucose deprivation, the EGFR inhibitor gefitinib (21) did not decrease pAKTThr308 induced by prolonged glucose deprivation despite completely blocking both Y845 and Y1068 EGFR phosphorylation (Figure 2C). IGF1R phosphorylation was not increased by prolonged glucose deprivation. Furthermore, GSK1838705A, an IGF1R inhibitor, did not inhibit prolonged glucose deprivation-induced pAKTThr308. Thus, prolonged glucose deprivation-induced AKTThr308 phosphorylation appears independent of EGFR and IGF1R activity (Figure 2C).

The PH domain of AKT, which mediates the interaction with cell membrane PtdInsP3, is required for AKT phosphorylation following growth factor receptor activation (22). Indeed a PH domain mutation (R25A) (22, 23), which blocks the interaction of AKT with PtdInsP3, abolished glucose deprivation-induced AKTThr308 phosphorylation, indicating that AKTThr308 phosphorylation in glucose deprived cells is dependent on AKT PH domain interaction with PtdInsP3 (Figure 2D). Overexpressed AKT1 elevated pAKTSer473 without further induction by glucose deprivation.

In the canonical AKT activation model, AKT translocates to the cell membrane, where Thr308 is phosphorylated by PDK1 (24) and Ser473 by TORC2 (25). In HeLa cells GFP-tagged AKT1 PH domain was predominantly cytoplasmic in serum-starved cells and translocated to the membrane in response to insulin (Figure 2E). In contrast, prolonged glucose deprivation did not induce detectable increases in cell membrane fluorescence (Figure 2E), suggesting that glucose deprivation-induced AKTThr308 phosphorylation is not dependent on acute translocation of AKT to the cell membrane. Furthermore, insulin stimulation induced significant acute increases in PtdInsP3 levels in HeLa cells, whereas prolonged glucose deprivation did not (Figure 2F), indicating that selective AKTThr308 phosphorylation was not due to acutely elevated PtdInsP3 levels. However, we cannot eliminate the potential that changes in PtdInsP3 over a prolonged time course induced transient association of AKT with the cell membrane that were not detectable by the assay approach.

We also determined whether AKT kinase activity was required for selective AKTThr308 phosphorylation. HCT116 cells with knockout of both AKT1 and AKT2 lack endogenous AKT as they do not express AKT3 (15). Transfection of kinase inactive AKT (K179M) (23, 26) into the HCT116 knock out cells resulted in marked AKTThr308 phosphorylation under prolonged glucose-deprivation conditions, indicating glucose deprivation-induced AKTThr308 phosphorylation is independent of AKT kinase cross- or auto-phosphorylation (Figure 2G).

We also examined Protein Phosphatase 2 (PP2A) and PH domain and Leucine rich repeat Protein Phosphatases (PHLPP) that dephosphorylate pAKTThr308 and pAKTSer473, respectively (27, 28). Prolonged glucose deprivation did not change phosphorylation of PP2ATyr307, indicative of PP2A phosphatase activity, suggesting that selective AKTThr308 phosphorylation is not due to decreased PP2A phosphatase activity (Figure S3A). Prolonged glucose deprivation modestly decreased PHLPP1 levels with no change in PHLPP2 levels (Figure S3B), suggesting PHLPP is not likely responsible for the markedly lower amounts of AKTSer473 phosphorylation compared with AKTThr308 phosphorylation after prolonged glucose deprivation. PTEN levels were not significantly altered suggesting glucose deprivation-induced AKTThr308 phosphorylation is independent of changes in PTEN activity (Figure S3C). Further studies are needed to elucidate whether accessibility or activity of these phosphatases contributes to selective AKTSer473 dephosphorylation with a consequent apparent increase in pAKTThr308.

Prolonged Glucose Deprivation Causes Energy Depletion and Enhances AKT Association with PDK1 and GRP78 Contributing to Selective AKTThr308 Phosphorylation

In HCT116 cells, prolonged glucose deprivation markedly decreased ATP levels (Figure 3A), as expected. Addition of methylpyruvate (MP), a membrane permeant pyruvate analog, rescued ATP levels to 50% of normal levels and completely abolished pAKTThr308 induced by prolonged glucose deprivation indicating that the increase in pAKTThr308 is dependent on energy depletion rather than other functions of glucose such as building blocks for fatty acid synthesis or the pentose phosphate shunt.

Autophagy plays a critical role in maintenance of ATP levels in cells under metabolic stress (29). We examined AKT phosphorylation in autophagy-deficient ATG5−/− MEFs (30). Although both basal and induced pAKTThr308 were lower than in ATG5+/+ MEFs, prolonged glucose deprivation induced a marked increase in pAKTThr308 in ATG5−/− MEFs independent of ability to enter autophagy (Figure 3B).

We previously demonstrated that the association of AKT with PDK1 was sufficient for AKTThr308 phosphorylation and activation of key signaling pathways (31). Others have also demonstrated that association with PDK1 facilitates AKT activation (32, 33). Following prolonged glucose deprivation, GFP-PDK1 could be co-immunoprecipitated with AKT in HeLa cells (Figure 3C). Thus prolonged glucose deprivation increases association of PDK1 with AKT.

We then sought to identify proteins that bind to AKT under prolonged glucose deprivation conditions by co-immunoprecipitation and mass spectrometry. Five bands co-IP with endogenous AKT from HeLa cells specifically under prolonged glucose deprivation conditions. The proteins were identified with high confidence as 78 kDa glucose-regulated protein (GRP78 aka BIP), Enolase-1, Flotillin-1, β-actin, and Serpin B3 (Figure S4). GRP78/BIP has previously been shown to promote AKT phosphorylation under endoplasmic reticulum stress conditions (34, 35). GRP78 could be co-immunoprecipitated with AKT and PDK1 in HeLa cells under prolonged glucose deprivation conditions but not under serum starvation or insulin stimulation conditions (Figure 3D). Further, GRP78 protein levels were induced by glucose deprivation, consistent with previous studies (34, 35). Knockdown of GRP78 by shRNA significantly reduced pAKTThr308 induced by prolonged glucose deprivation, with no effect on pAKTSer473 (Figure 3E). Strikingly the selective GRP78 inhibitor VER-155008 (36) inhibited pAKTThr308 and substrate-specific phosphorylation induced by prolonged glucose deprivation (Figure 3F). Thus prolonged glucose deprivation increases both GRP78 levels and association with AKT contributing to AKTThr308 phosphorylation.

Selective AKT Phosphorylation and Activation Contribute to Cell Survival Under Prolonged Glucose Deprivation Conditions

We then examined the effect of AKT activation on cell survival during prolonged glucose deprivation. Over the first 10h of glucose deprivation, very few cells (~3%) detached from the substrate, however, about 30% of the cells detached over 16h, 50% detached over 24h, and almost all cells detached during 48h of glucose deprivation (Figure 4A). In the presence of concentrations of the AKT inhibitor MK-2206 sufficient to inhibit AKT phosphorylation, significantly fewer cells remained attached after 16h of glucose deprivation (Figure 4B). MK-2206 increased cleaved caspase-7 and cleaved Poly (ADP-ribose) polymerase (PARP) consistent with AKT inhibition inducing apoptosis. The GRP78 inhibitor VER-155008 significantly decreased the percentage of attached cells after glucose deprivation but not in the presence of glucose (Figure 4C) and significantly decreased the numbers of colonies formed after glucose readdition to cells that had undergone prolonged glucose deprivation (Figure 4D). Thus GRP78 contributes to AKT phosphorylation and subsequent cell survival under glucose deprivation conditions.

Figure 4. AKT activation induced by prolonged glucose deprivation contributes to cell survival under severe metabolic stress.

A, prolonged glucose deprivation induces cell detachment from cell culture plates. HeLa cells were serum-starved for 16h and glucose-deprived for 0, 10, and 16h. HeLa cells were also serum starved and glucose deprived for extended 24h and 48h. Detached and attached cells were collected separately, and cells were counted using an automatic cell counter. Three independent experiments were performed. The error bars indicate standard deviations. B, inhibition of AKT increases cell detachment under prolonged glucose deprivation conditions. HeLa cells were serum-starved and glucose-deprived and MK-2206 was added to the medium at the indicated concentrations for 16h. Cell detachment was measured as described above. The error bars indicate standard deviations from three independent experiments. Inhibition of AKT was verified using Western blotting. C, inhibition of GRP78 increases cell detachment under prolonged glucose deprivation conditions. HeLa cells were serum-starved and glucose-deprived for 24h and VER-155008 was added to the medium for the last 3h. Cell detachment was measured as described above. The error bars indicate standard deviations from three independent experiments. D, inhibition of GRP78 decreases the numbers of colonies formed following glucose readdition after prolonged glucose deprivation. HeLa cells were serum-starved and glucose-starved for 24h. DMSO control or VER-155008 was added to the medium for the last 3h. All cells were harvested and same numbers of cells were plated for colony formation assays in complete medium. The error bars indicate standard deviations from three independent experiments. E, dynamics of apoptosis induced by glucose deprivation. HeLa cells were serum-starved for 16h and glucose-deprived for the indicated periods. Detached cells were washed off. Only attached cells were harvested and examined for apoptosis using Western blotting. F, glucose readdition to the cell culture medium reverses AKT phosphorylation induced by prolonged glucose deprivation. HeLa cells were serum-starved and glucose-deprived for 16h, and glucose (25 mM) was added to the cell culture medium. Cells were harvested at 0, 3, 6, and 10h after glucose readdition. Data are representative of three independent experiments.

Within the first 12h of glucose deprivation, cleaved caspase-7, caspase-3, and PARP increased consistent with induction of apoptosis in attached cells (Figure 4E). Intriguingly, caspase and PARP cleavage were significantly lower in attached cells at 16h than at 12h of glucose deprivation, indicating that fewer attached cells were undergoing apoptosis at 16h. However, of note, by 16h about 30% of the cells had already detached likely as the consequence of cell death through apoptosis. Cells that remained attached at 16h were relatively resistant to apoptosis, which is consistent with the marked increase in pAKTThr308 in these cells.

We next determined the effect of readdition of glucose to culture medium on AKT phosphorylation. Strikingly, pAKTThr308 decreased significantly within 3h and returned to close to basal levels by 10h after glucose readdition (Figure 4F). Furthermore, cleaved caspase-7 decreased to basal levels within 10h after glucose readdition, indicating suppression of apoptosis after relief of metabolic stress.

Prolonged Glucose Deprivation Induces Activation of Additional Survival Signaling Networks

We utilized reverse phase protein arrays (RPPA) to explore the effects of glucose deprivation across multiple signaling pathways. As shown in Figure 5A, a small number of proteins exhibited marked upregulation of phosphorylation (to the left of x-axis) or downregulation of protein expression or phosphorylation (to the right of x-axis) under prolonged glucose deprivation conditions (16h). An unsupervised median-centered heat map of HeLa cells with 0, 3, or 16h of glucose deprivation is shown in Figure S5. In terms of proteins showing significant changes (Figure 5B and C), increases in pACCS79, pAKTSer473 and cleaved caspase observed during transient glucose deprivation recapitulated results seen with western blotting. Intriguingly, decreases in c-Myc and Snail, albeit with greater effects at 16h, were observed with both transient and prolonged glucose deprivation. Glucose deprivation for 16h also markedly decreased cyclin E1 and E2 consistent with cell cycle changes as well as tuberin, ATM, and MSH6 that are involved in DNA repair and increased pCHK2 also involved in DNA damage checkpoint control (Figure 5A and C). Strikingly, prolonged glucose deprivation increased phosphorylation of multiple components of the PI3K pathway (pAKTThr308, pNF-κB, pGSK3, pFOXO3A, and pS6) as indicated above. Components of the MAPK pathway including MEK1, Erk1/2, p38, c-Jun and YB1 demonstrated marked increases in phosphorylation. Whether this was consequent to activation of the EGFR or Src, both of which can activate the MAPK pathway, remains to be determined.

Figure 5. RPPA analysis of signaling changes in HeLa cells induced by glucose deprivation.

A, signaling changes induced by prolonged glucose deprivation. HeLa cells were serum-starved and glucose-deprived for 16 hours. Cells were then harvested and signaling in the cells was analyzed using RPPA. The signaling values are presented as relative conversion to values in control cells without glucose deprivation. The data are presented as either fold increase or percentage decrease compared with the values in control cells from high to low. The insets show selected increased and decreased signaling. B, signaling changes induced by transient glucose deprivation. HeLa cells were serum-starved for 16h and glucose-deprived for 3h. The data are presented the same as in A. C, unsupervised median-centered heat map of selected signaling clusters in HeLa cells with transient and prolonged glucose deprivation. HeLa cells were serum-starved for 16h and glucose-deprived for 0, 3, and 16h. Two independent sets of samples were subjected to RPPA analysis with 242 antibodies. Selected signaling clusters are shown.

Discussion

In this study, we demonstrated that transient and prolonged glucose deprivation induce AKT phosphorylation and activation via distinct mechanisms. Transient glucose deprivation modestly induces both pAKTThr308 and pAKTSer473 likely via a feedback loop from p70S6K. In contrast, prolonged glucose deprivation induces selective AKTThr308 phosphorylation through the formation of a complex including GRP78 and PDK1. Selective AKTThr308 phosphorylation targets AKT to a specific group of substrates contributing to cell survival. Further, PDK1 and GRP78 are required for the phosphorylation of AKT and for cell survival under prolonged glucose deprivation.

We further characterized the mechanisms of selective AKTThr308 phosphorylation induced by prolonged glucose deprivation and demonstrated that prolonged glucose deprivation-induced pAKTThr308 is dependent on both PDK1 and PI3K. Intriguingly, EGFR, although activated by prolonged glucose deprivation, and IGF1R kinase activity are dispensable for glucose-deprivation induced AKTThr308 phosphorylation. Prolonged glucose deprivation increases GRP78 levels and its association with AKT, which may facilitate AKT association with PDK1 contributing to selective AKTThr308 phosphorylation and cell survival. Other processes, such as inactivation or sequestration of mTORC2 or activation of specific phosphatases may be responsible for the lower AKTSer473 phosphorylation compared with AKTThr308.

We also showed that an intact PH domain of AKT is required for selective AKTThr308 phosphorylation induced by prolonged glucose deprivation. However, we did not observe localization of AKT at the cell membrane during glucose deprivation. One possible reason for this lack of localization is that cell membrane translocation of AKT is not required for phosphorylation; instead, binding of AKT to PtdInsP3 in other cell compartments facilitates AKT phosphorylation by PDK1 under prolonged glucose deprivation conditions. Another possibility is that in contrast to acute insulin stimulation, we could not detect a transient or low level translocation of AKT to the cell membrane during the prolonged period of glucose deprivation. This latter contention is compatible with the requirement for PI3K for delayed AKTThr308 phosphorylation. Our data support a model of AKT activation induced by prolonged glucose deprivation, wherein PI3K activation is required to produce PtdInsP3 that initiates AKT association with PDK1 on the cell membrane or other membrane systems. Glucose deprivation stabilizes the AKT-PDK1 complex, likely due to GRP78 and potentially other scaffold proteins that bind to the complex, contributing to the selective AKTThr308 but not AKTSer473 phosphorylation.

The selective AKTThr308 phosphorylation and consequent phosphorylation of specific AKT substrates we describe herein are distinct from metabolic stress-induced AKT phosphorylation and activation described previously where in most cases acute effects of glucose deprivation were studied. For example, Zhong et al. (37, 38) reported that 2-deoxyglucose, a glycolysis inhibitor, induced AKTThr308 and AKTSer473 phosphorylation via IGF1R but not EGFR. In addition, Pelicano et al. (39) reported that mitochondrial respiration defects activated AKT via PTEN inactivation. In contrast with previous studies, the present study showed that prolonged glucose deprivation induces selective and marked AKTThr308 phosphorylation and phosphorylation of a subset of AKT substrates independent of PTEN status and mTORC1 inhibition, revealing a novel AKT activation mechanism. Our data suggest that a combination of AKT or GRP78 inhibition with metabolic stress may synergistically kill cancer cells that are resistant to either treatment alone.

We previously showed that AKT and MAPK signaling are usually counterbalanced (40, 41). However, the data in the present study demonstrate that both AKT and MAPK family members, including Erk1/2 and p38, are highly activated during prolonged glucose deprivation, indicating that the signaling network in cells under severe metabolic stress can be rewired and rebalanced to induce effective activation of cell survival machinery.

As shown in the model in Figure 6, when a tumor grows large or is treated with antiangiogenesis drugs to block its nutrient supply, some tumor cells may be subject to glucose deprivation. We propose that when cells encounter metabolic stress, such as glucose deprivation, distinct survival mechanisms are sequentially activated according to the severity and duration of the stress, providing a layered defense against cell death. During transient glucose deprivation, both AKTThr308 and AKTSer473 phosphorylation is modestly induced via a feedback loop (13, 42) as a consequence of activation of energy-sensing AMPK signaling and inhibition of mTORC1, which provides the first line of defense against cell death under mild or transient metabolic stress.

Figure 6. Model of AKT phosphorylation induced by glucose deprivation in vitro and in vivo.

However, during prolonged glucose deprivation, we propose that this first line of defense becomes insufficient to protect cells against death, and, indeed, is no longer activated as indicated by decreased AMPK signaling and relief of mTORC1 inhibition despite continued low ATP levels. The next tier of survival mechanism is subsequently activated, which selectively and markedly increases pAKTThr308, providing a second line of defense against cell death under prolonged metabolic stress. In this tier of protection, AKT phosphorylates a specific subset of substrates that are likely critical for cell survival. In contrast, some AKT substrates that may be dispensable for cell survival or that could increase cell stress by inducing protein synthesis or cell cycle progression are not phosphorylated. The substrate specificity of AKT may result from discordant levels of AKTThr308 and AKTSer473 phosphorylation, consistent with previous reports that differentially phosphorylated AKT possesses activity that selectively targets different substrates (43, 44).

In summary, we demonstrated that prolonged glucose deprivation induces selective AKTThr308 phosphorylation and AKT activation toward a specific group of substrates via, at least in part, enhanced AKT association with PDK1 and GRP78. Small molecule inhibitors of both PDK1 and GRP78 are in clinical development providing a ready approach for translation to the clinic. As a newly identified hallmark of cancer, deregulated metabolism is emerging as a major target in cancer therapy (1). Our data reveal a critical AKT-mediated survival mechanism under prolonged metabolic stress, which is of importance to development and implementation of drugs targeting cell metabolism and AKT signaling.

Materials and methods

Cell Culture and Plasmids

Tet-on derivative HeLa was from BD Clontech (Palo Alto, CA). The human embryonic kidney line HEK293; kidney cancer lines 786-0, ACHN, and RXF393; non-small cell lung cancer line NCI-H358; lung adenocarcinoma cancer line A549; squamous lung cancer line H226; breast cancer line HCC1806, and ovarian cancer lines Skov-3, A2780 and Caov-3 were from American Type Culture Collection (Manassas, VA). The breast cancer line MDA-MB-468 was from Janet Prices (MDACC). HCT116 and DLD1 (wild type, PDK1−/−, and AKT1/2−/−) colon cancer lines were from Dr. Bert Vogelstein (Johns Hopkins University). ATG5+/+ and ATG5−/− mouse embryonic fibroblasts (MEFs) were from Dr. Noboru Mizushima (Tokyo Medical and Dental University, Tokyo, Japan).

HeLa, 786-0, ACHN, RXF393, and HEK293 cells were cultured in DMEM (Invitrogen, Carlsbad, CA), HCT116 and DLD1 in McCoy’s 5A growth medium (Gibco), and MDA-MB-468, HCC1806, Skov-3, A549, A2780, Caov-3, H226, and NCI-H358 in RPMI 1640 medium (Invitrogen) supplemented with 10% (v/v) fetal calf serum (Gibco). Cell line identity was routinely confirmed by STR profiling in the MDACC CCSG core. To initiate coordinate serum starvation and glucose deprivation, complete medium was replaced with serum- and glucose-free medium. To evaluate serum starvation for 16h and glucose deprivation for shorter times (3h), complete medium was replaced with serum-free medium for 13h and then replaced with serum-and glucose-free medium for an additional 3h before harvesting.

pcDNA3.1/Zeo(+) was obtained from Invitrogen. Plasmids expressing the PH domain of AKT fused to GFP and GFP-PDK1 were from Bioimage (Soeborg, Denmark). Venus cDNA was from Dr. Atsushi Miyawaki (RIKEN, Saitama Japan). Venus N-terminal fragment (VN) was PCR-amplified and cloned into pcDNA3.1/Zeo(+) to express VN (amino acids 1-158). Human AKT1 was cloned from OVCAR3 cells using RT-PCR as described previously (45). VN-AKT1 plasmid was constructed by cloning AKT1 into pcDNA3.1/Zeo(+)-VN vector. VN-AKT1 (R25A) mutant was created using QuikChange mutagenesis kit (Stratagene, La Jolla, CA). HA-AKT1 (K179M) mutant was constructed by cloning AKT1 into pcDNA3 followed by site-directed mutagenesis. Lentiviral vectors for control and GRP78 shRNA were from Open Biosystems (Huntsville, AL). Cells were transfected using Lipofectamine 2000 transfectionreagent (Invitrogen) according to manufacturer’s protocol.

Reagents

The AKT inhibitor 8-[4-(1-aminocyclobutyl)phenyl]-9-phenyl-1,2,4-triazolo[3,4-f] [1,6]naphthyridin-3(2H)-one hydrochloride (MK-2206) was synthesized at MD Anderson according to a structure reported previously (46). PDK1 inhibitor GSK2334470 and IGF1R inhibitor GSK1838705A were from GlaxoSmithKline (London, UK) (47). PI3K inhibitor GDC-0941 was from Axon Mechem (Groningen, The Netherlands). MEK inhibitor U0126 was from Promega (Sunnyvale, CA). Rapamycin was from Cell Signaling Technology (Beverly, MA). Dual mTORC1 and mTORC2 inhibitor AZD8055 and EGF receptor (EGFR) inhibitor gefitinib were from AstraZeneca (Wilmington, DE). GRP78 inhibitor, VER-155008, was from Santa Cruz Biotechnology (Santa Cruz, CA). Methylpyruvate (MP), insulin, D-glucose, and L-glucose were from Sigma-Aldrich (St. Louis, MO). GFP-Trap was from Allele Biotechnology and Pharmaceuticals, Inc. (San Diego, CA).

Antibodies

Antibodies used in this study are listed in Supplemental Table 1.

Western Blotting, Co-Immunoprecipitation, and RPPA

For Western blotting, cells were lysed in RIPA buffer. For co-immunoprecipitation (IP), cells were lysed in 0.5% NP-40 buffer. GFP-Trap was performed according to manufacturer’s protocol (Allele Biotechnology and Pharmaceuticals, Inc., San Diego, CA). Co-IP and Western blotting were performed as described previously (31). RPPA was performed as previously described (40, 48).