Research ArticleCell death

Ca2+-dependent demethylation of phosphatase PP2Ac promotes glucose deprivation–induced cell death independently of inhibiting glycolysis

See allHide authors and affiliations

Science Signaling  09 Jan 2018:
Vol. 11, Issue 512, eaam7893
DOI: 10.1126/scisignal.aam7893

A glucose-calcium connection in cell death

Glucose is a critical nutrient for cell survival, particularly in neurons and some types of cancer cells. Lee et al. found that some cancer cells are particularly sensitive to glucose loss but not because of starvation as one might expect. Loss of glucose triggered the influx of calcium across the plasma membrane, which activated a protein that demethylated (and inactivated) the phosphatase PP2A, leading to cell death through the activity of the kinase RIPK1. RIPK1 triggered cell death through a pathway that is unlike the currently recognized apoptosis, necroptosis, and necrosis mechanisms. A nonmetabolizable analog of glucose did not promote, but rather prevented, cell death by inhibiting cell membrane depolarization, hence blocking calcium influx. This knowledge might be used to therapeutically induce cell death in tumors or perhaps even prevent cell death in other cell types, such as neurons.

Abstract

Cancer cells increase glucose metabolism to support aerobic glycolysis. However, only some cancer cells are acutely sensitive to glucose withdrawal, and the underlying mechanism of this selective sensitivity is unclear. We showed that glucose deprivation initiates a cell death pathway in cancer cells that is dependent on the kinase RIPK1. Glucose withdrawal triggered rapid plasma membrane depolarization and an influx of extracellular calcium into the cell through the L-type calcium channel Cav1.3 (CACNA1D), followed by activation of the kinase CAMK1. CAMK1 and the demethylase PPME1 were required for the subsequent demethylation and inactivation of the catalytic subunit of the phosphatase PP2A (PP2Ac) and the phosphorylation of RIPK1. Plasma membrane depolarization, PP2Ac demethylation, and cell death were prevented by glucose and, unexpectedly, by its nonmetabolizable analog 2-deoxy-d-glucose (2-DG), a glycolytic inhibitor. These findings reveal a previously unknown function of glucose as a signaling molecule that protects cells from death induced by plasma membrane depolarization, independently of its role in glycolysis. Components of this cancer cell death pathway represent potential therapeutic targets against cancer.

INTRODUCTION

One of the hallmarks of cancer is an increase in cellular glucose uptake and dependence (1), which is thought, in part, to support aerobic glycolysis (2). Rapidly dividing cancer cells use aerobic glycolysis for the production of metabolic intermediates, amino acids, nucleic acids, and energy (3). This phenomenon is known as the Warburg effect (2, 4). This addiction to glucose distinguishes cancer cells from normal cells and often increases their sensitivity to glucose deprivation (5, 6). Following this logic, glucose deprivation would be predicted to selectively target cancer cells, without adversely affecting normal cells. Thus, targeting glucose metabolism and uptake in cancer cells, for example, by blocking glucose metabolism with nonmetabolizable analogs of glucose, such as 2-deoxy-d-glucose (2-DG) (7), or by blocking glucose uptake through inhibition of glucose transporters such as GLUT1 (SLC2A1) (8), could be therapeutic. However, not all cancer cell types are sensitive to glucose deprivation–induced cell death (9). Identifying the pathways that sensitize cancer cells to glucose depletion will facilitate the development of new therapies.

Protein phosphatase 2 (PP2A) is a serine/threonine phosphatase that exists as a heterotrimer composed of a core heterodimer of the scaffolding “A” subunit (PP2Aa) and the catalytic “C” subunit (PP2Ac), as well as one of various regulatory “B” subunits (10). The interaction between the regulatory subunits and the core heterodimer determines the subcellular localization, substrate specificity, and downstream signaling of the PP2A complex (1012). A critical posttranslational modification of PP2Ac is the methylation/demethylation of its last amino acid, leucine 309 (Leu309). Leu309 is methylated by leucine carboxyl methyltransferase 1 (LCMT1) (13), which promotes PP2A holoenzyme assembly (1417) and PP2A activation (18, 19). PP2Ac is demethylated by PPME1 (protein phosphatase methylesterase 1) (20, 21), which is associated with the accumulation of inactive PP2A (19, 21).

The PP2Ac demethylase PPME1 can be regulated by calcium/calmodulin-dependent protein kinase I (CAMK1) (22), a kinase that is activated in response to increased cytosolic calcium concentration (23). Cytosolic calcium can increase because of an influx either of extracellular calcium after opening the membrane calcium channels or of intracellular calcium released from endoplasmic reticulum/sarcoplasmic reticulum stores (24). An increase in the cytosolic calcium concentration triggers diverse signaling pathways and biological processes, including cell death (25, 26).

There are various calcium channels that are regulated by different pathways. Voltage-sensitive L-type calcium channels open rapidly upon plasma membrane depolarization (27, 28). The resultant influx of calcium into the cytoplasm leads to the activation of multiple calcium-dependent pathways, including the activation of CAMK1 (23). Here, we aimed to identify the pathways that sensitize cancer cells to glucose depletion. Identifying the cell death pathways and revealing molecular targets led us to investigate a potential strategy to develop new therapeutic applications.

RESULTS

Cancer cell lines display different sensitivities toward glucose deprivation–induced cell death

Glucose deprivation is known to induce cell death in some cancer cell lines (9). We characterized a panel of cancer cell lines based on their sensitivity to glucose starvation. Standard culture media contain 25 mM glucose, and 10% fetal bovine serum (FBS) contains about 0.2 to 0.6 mM glucose. We incubated cells in media with or without both glucose and serum and evaluated cell death by propidium iodide (PI) exclusion, which was detected by flow cytometry. Five of the seven cancer lines tested displayed substantial cell death in less than 10 hours of glucose- and serum-starvation (SW480, U2OS, U251MG, SaOS2, and U87MG; collectively referred to as “sensitive cell lines” hereafter in this manuscript) (Fig. 1A and fig. S1A). By contrast, two cancer cell lines (A549 and H1299), two normal primary fibroblasts (WI-38 and IMR-90), and a normal immortalized breast epithelial cell line (MCF-10A) remained impermeable to PI in serum- and glucose-free media (referred to as “insensitive cell lines” hereafter) (Fig. 1A and fig. S1A). Readdition of as little as 0.025 mM glucose was sufficient to prevent glucose- and serum starvation–induced cell death in U2OS cells (fig. S1B), suggesting that the lack of glucose triggers cell death. However, it is known that the absence of serum has effects on cell viability (29, 30). To rule out the potential effects of serum removal on cell death, we cultured cells in media with DFBS (dialyzed fetal bovine serum), which should retain the necessary growth factors but have undetectable levels of glucose. The glucose concentration in the media was also lowered to 1 mM to mimic the physiological tumor environment. Glucose concentrations in human blood are around 5 to 7 mM, and concentrations in cancers are frequently 3- to 10-fold lower than in normal tissues (9, 3133). SW480, U2OS, and U251MG cells grown in DFBS-containing media remained sensitive to glucose deprivation, whereas A549, H1299, IMR-90, WI-38, and MCF-10A cells remained insensitive (Fig. 1B). We found that glucose deprivation of U2OS cells grown in the presence or absence of DFBS triggered similar cell death with similar kinetics (Fig. 1C). On the basis of these data, we concluded that the lack of glucose triggers cell death in a subset of cancer cells. We focused our subsequent analyzes on U2OS cells because they displayed the most rapid cell death after glucose removal.

Fig. 1 A subset of cancer cell lines is sensitive to glucose deprivation–induced cell death.

(A) Propidium iodide (PI) exclusion assay using the indicated cell lines cultured in the presence or absence of both 10% serum and 25 mM glucose for 4 hours (U2OS), 6 hours (U251MG), or 8 to 10 hours (remaining cell lines). A representative analysis from three independent experiments is shown. The percentage of PI-positive dead cells is shown. Asterisk (*) indicates cell lines sensitive to glucose deprivation. (B) PI exclusion assay using the indicated cell lines cultured in media containing 10% dialyzed fetal bovine serum (DFBS) with or without 1 mM glucose for 16 hours. Shown is the mean percentage of dead cells ± SD from three independent experiments. (C) Time course of PI exclusion assay in U2OS cells cultured with or without 10% DFBS and 1 mM glucose for the indicated periods. Shown is the mean percentage of dead cells ± SD from three independent experiments. (D) Representative images from phase-contrast microscopy from three independent experiments. Cells were cultured in media containing 10% DFBS with or without 1 mM glucose for 10 hours. Scale bars, 50 μm.

We performed phase-contrast microscopy to evaluate the morphology of the cells after 10 hours of glucose deprivation. Under glucose deprivation, more of the cells were round and loosely attached to the plates in the five sensitive cell lines, whereas the morphology of the insensitive cell lines was not affected (Fig. 1D and fig. S1C). We also evaluated the kinetics of U2OS cell death upon glucose deprivation with PI staining using the IncuCyte ZOOM System (movies S1 and S2), which revealed that the cells exhibited the typical rounded morphology starting 3 hours after glucose removal and that most cells died within 16 hours.

Glucose deprivation–induced cell death is independent of ATP depletion

Cancer cells use glucose to generate adenosine 5′-triphosphate (ATP), and ATP depletion can affect cell survival (34). To determine whether ATP depletion was the cause of cell death induced by glucose deprivation, we blocked the ATP production by inhibiting glycolysis with the glucose analog 2-DG. If cell death was triggered by ATP depletion as a consequence of glucose removal, then we predicted that the addition of 2-DG to the cells would accelerate ATP depletion and hence cell death. As we expected, glucose deprivation reduced the amount of ATP in U2OS cells (Fig. 2A). Supplementing the media with 1 mM 2-DG in the presence of 1 mM glucose was not sufficient to reduce ATP, but 1 mM 2-DG significantly reduced intracellular ATP amounts in cells grown without glucose (Fig. 2A). Unexpectedly, 2-DG prevented cell death induced by glucose deprivation in sensitive cancer cell lines, as determined by PI exclusion assay (Fig. 2B) and analysis of cell morphology (Fig. 2C). Even the addition of as little as 0.025 mM 2-DG attenuated cell death in U2OS cells (fig. S2A). Because pyruvate is the end product of glycolysis and the initiator of ATP synthesis in the tricarboxylic acid cycle, we investigated whether increasing ATP by pyruvate supplementation can prevent cell death by glucose removal. Although the addition of pyruvate increased ATP amounts in cells grown in the absence of glucose (Fig. 2A), it failed to rescue cell viability upon glucose depletion (Fig. 2C). Together, these data suggest that cell death induced by glucose depletion is not due to ATP depletion.

Fig. 2 Glucose deprivation–induced cell death is independent of ATP depletion.

(A) Intracellular adenosine 5′-triphosphate (ATP) amount in U2OS cells treated as indicated for 4 hours. Bars represent the mean (±SD) amount of ATP in each sample relative to the control condition (first bar), quantified from more than three independent experiments. ***P < 0.001, unpaired two-tailed Student’s t test. (B) PI exclusion assay with the indicated cell lines cultured with or without 1 mM 2-deoxy-d-glucose (2-DG) in the absence of glucose for 16 hours. Bars represent the mean percentage of cell death ± SD of three independent experiments. (C) Representative phase-contrast images of U2OS cells from three independent experiments. U2OS cells were treated as indicated for 4 hours. Scale bar, 50 μm.

Reactive oxygen species (ROS) contributes to glucose depletion–induced cell death (35). Although ROS was slightly induced after 4 hours of glucose depletion in U2OS cells (fig. S2B), the inhibition of ROS induction by catalase (fig. S2B) did not prevent cell death induced by glucose depletion in U2OS cells (fig. S2C), suggesting that ROS unlikely contributed to the cell death induced by the 4-hour glucose depletion, at least in our cell culture conditions.

To analyze the shared property of glucose and 2-DG that contributes to protecting cells from cell death, we examined the effect of O-glycosylation. Both glucose and 2-DG are known to increase O-glycosylation (36). We speculated that glucose and 2-DG may protect cells from cell death by enhancing O-glycosylation. However, we did not observe obvious changes in O-glycosylation with or without 1 mM glucose or 2-DG for 4 hours in U2OS cells (fig. S3A), indicating that such a low concentration of glucose or 2-DG does not affect O-glycosylation in a shorter incubation period (4 hours). To further examine the involvement of O-glycosylation in cell survival, we treated the cells with PUGNac that enhances O-glycosylation by inhibiting O-GlcNac-β-N-acetylglucosaminidase that is responsible for removing O-GlcNac from O-linked glycosylated proteins. Although we observed the marked increase in O-glycosylation by adding PUGNac in the presence or absence of 1 mM glucose (fig. S3A), unlike 2-DG, PUGNac could not prevent cell death induced by glucose deprivation (fig. S3B). These results suggest that glucose and 2-DG prevent cell death independently from O-glycosylation.

2-DG is a glucose analog in which only the 2-hydroxyl group is replaced by hydrogen (fig. S3C). After conversion to 2-deoxyglucose-6-phosphate by hexokinase, it cannot undergo isomerization and further glycolysis. To understand whether this structural similarity to glucose is important for preventing cell death, we treated the cells with 6-DG, a glucose analog in which the 6-hydroxyl group is replaced by hydrogen (fig. S3C). Because of the lack of a hydroxyl group at the sixth carbon position, 6-DG cannot be phosphorylated by hexokinase and therefore cannot be metabolized in glycolysis. Notably, 6-DG did not prevent U2OS cell death, unlike glucose and 2-DG (fig. S3D), suggesting that the hydroxyl group on the sixth carbon position is important to protect cells from cell death.

Glucose deprivation–induced cell death is neither apoptosis nor related to autophagy

To gain insight into the mechanism of cell death induced by glucose deprivation, we evaluated markers of apoptosis. In contrast to ultraviolet (UV) light exposure, glucose withdrawal of U2OS cells did not induce apoptotic markers, such as cleaved poly(ADP–ribose) polymerase 1 (PARP1) or cleaved caspase 3 (CASP3) (Fig. 3A). In addition, Z-VAD-FMK (a general caspase inhibitor) could not prevent glucose deprivation–induced cell death in U2OS cells, as indicated by phase-contrast images at 4 hours of glucose deprivation, a time point when most of the cells were rounded and loosely attached to the plates (fig. S4A), and by fluorescent images of PI staining at 16 hours of glucose deprivation, a time point when most of the cells were permeable to PI (Fig. 3B). We also did not detect changes in mitochondrial membrane potential (fig. S4B) and cytochrome c release (fig. S4C) after glucose deprivation. These data suggest that cell death induced by glucose deprivation does not occur through apoptosis.

Fig. 3 Glucose deprivation does not induce apoptosis or autophagy.

(A) Representative Western blotting from three independent experiments of U2OS cells cultured with or without 1 mM glucose for 4 hours. Lysate from ultraviolet (UV)–irradiated U2OS cells serves as a positive control for apoptosis. Z-VAD-FMK was used to inhibit caspase activation. (B) Representative phase-contrast and fluorescent images of three independent experiments. U2OS cells were cultured as indicated for 16 hours and stained with PI before imaging. Z-VAD-FMK was used to inhibit apoptosis. Scale bars, 50 μm. (C) PI exclusion assay with U2OS cells transfected small interfering RNA (siRNA) against the indicated autophagy-associated mRNAs or a control (ctrl) and cultured with or without 1 mM glucose for 4 hours. Shown is the mean percentage of dead cells ± SD from more than three independent experiments. (D) PI exclusion assay with U2OS cells pretreated with or without chloroquine for 12 hours followed by the indicated treatments for 4 hours. Shown is the mean percentage of dead cells ± SD from three independent experiments.

Glucose depletion affected neither the abundance of autophagy marker LC3-II nor the LC3-II/LC3-I ratio (fig. S5A), suggesting that autophagy was not induced, at least by the time when the cells started to undergo rapid death after glucose deprivation. To further confirm this, we inhibited autophagy by knocking down ATG5, ATG7, ATG12, and BECN1 (fig. S5, B to E). Glucose deprivation–induced cell death was not affected by knockdown of these proteins, which are essential for autophagosome formation (Fig. 3C) (37). In addition, we inhibited the late stages of autophagy by adding chloroquine, an agent that prevents the acidification of lysosomes and inhibits autophagic flux by preventing lysosomal protein degradation (38). Addition of chloroquine did not prevent glucose deprivation–induced cell death (Fig. 3D). Adenosine 5′-monophosphate–activated protein kinase (AMPK) is an energy sensor activated in response to nutrient stress, and its activation induces autophagy in a cell type–dependent manner (39). We knocked down AMPK (fig. S6A) or inhibited AMPK activity by adding compound C to observe its effects on cell death. Consistent with our aforementioned autophagy-independent cell death data (Fig. 3, C and D), attenuating AMPK function through small interfering RNA (siRNA) or inhibitor did not affect glucose deprivation–induced cell death (fig. S6, B and C). Together, these data suggest that cell death induced by glucose deprivation is neither through AMPK nor through autophagy.

Glucose deprivation–induced cell death requires RIPK1

Necroptosis is a recently described additional form of programmed cell death (40) that occurs rapidly in a receptor-interacting protein kinase (RIPK)–dependent manner (41). We examined RIPK1 phosphorylation, a marker of necroptosis (42), and found that glucose withdrawal reduced the electrophoretic mobility of RIPK1, possibly due to phosphorylation (Fig. 4A). RIPK1 phosphorylation upon glucose deprivation was confirmed by Phos-tag Western blotting, which specifically detects phosphorylation as a shift in mobility of the band (Fig. 4B), and by blotting with an antibody, which specifically recognizes one of the known phosphorylation sites in RIPK1, Ser166 (fig. S7A). The intensity of the bands recognized by a RIPK1 Ser166 phosphorylation-specific antibody was reduced by knockdown of RIPK1, confirming that detected bands were RIPK1-specific (fig. S7A). To further confirm phosphorylation of RIPK1 by glucose deprivation, we treated the protein lysates with λ phosphatase, a serine/threonine/tyrosine phosphatase. After digesting protein lysates with λ phosphatase, the RIPK1 mobility shift caused by glucose depletion was attenuated in U2OS cells (Fig. 4C), indicating that the RIPK1 mobility shift was due to phosphorylation. Glucose deprivation–induced RIPK1 mobility shift was also observed in other glucose deprivation–sensitive cancer cell lines (U251MG and SW480), and these shifts were attenuated by λ phosphatase treatment (fig. S7, B and C). These results show that the phosphorylation of RIPK1 by glucose deprivation is conserved in multiple glucose deprivation–sensitive cancer cells.

Fig. 4 Glucose deprivation induces RIPK1-dependent cell death.

(A to C) Representative Western blotting of U2OS cells cultured as indicated for 4 hours. Western blotting analysis were performed by regular SDS–polyacrylamide gel electrophoresis (PAGE) (A), Phos-tag SDS-PAGE (B), or regular SDS-PAGE using λ phosphatase–digested lysates (C). All blots are representative of three independent experiments. (D) PI exclusion assay of siRNA-transfected U2OS cells cultured with or without 1 mM glucose for 4 hours. The mean percentage of dead cells ± SD from more than three independent experiments is shown. ***P < 0.001, unpaired two-tailed Student’s t test. (E) Representative phase-contrast and fluorescent images from three independent experiments of U2OS cells cultured as indicated for 16 hours and stained with PI before imaging. Scale bars, 50 μm. (F) Western blotting analysis for RIPK3 in cells cultured with or without 1 mM glucose for 4 hours. Blots are representative of three independent experiments. (G and H) Cells were treated with tumor necrosis factor–α (TNF-α), cycloheximide (CHX), Z-VAD-FMK, and necrostatin-1 (Nec) as indicated for 14 hours. Phase-contrast images are shown (G), and protein lysates were analyzed by Western blotting (H). Data are representative of three independent experiments. Scale bars, 50 μm.

Treatment with 2-DG ablated the detection of phosphorylated RIPK1 in glucose-deprived conditions (Fig. 4A), consistent with our aforementioned observation that 2-DG prevented cell death induced by the lack of glucose (Fig. 2B). To test whether RIPK1 is required for glucose deprivation–induced cell death, we knocked down RIPK1 expression in U2OS cells using each of two independent siRNAs (fig. S7A). RIPK1 knockdown significantly restored U2OS cell viability at 4 hours of glucose withdrawal (Fig. 4D) and delayed the U2OS cell death induced by glucose deprivation, as indicated by a delay in the induction of morphological changes associated with cell death monitored up to 12 hours of glucose withdrawal (fig. S8A). Similarly, we knocked down RIPK1 in U251MG and SW480 cells (fig. S8, B and C) and observed that knockdown of RIPK1 expression significantly restored cell viability at 6 hours of glucose withdrawal (fig. S8, D and E). These results suggest that glucose deprivation–induced cell death is RIPK1-dependent in multiple glucose deprivation–sensitive cancer cell lines.

RIPK1 is one of the key components for necroptosis. One of the inhibitors for necroptosis is necrostatin-1, which inhibits RIPK1 kinase activity (43). Although the cell death we observed is RIPK1-dependent, the addition of necrostatin-1 did not prevent glucose deprivation–induced cell death, as shown by phase-contrast images at 4 hours (fig. S9A) and by fluorescent images of cells stained with PI at 16 hours (Fig. 4E). These results suggest that this cell death is not necroptosis. Subsequently, we found that the presence of RIPK3, which is an important component in the necroptosis pathway downstream of RIPK1, was undetectable in U2OS cells compared to HT29 cells (Fig. 4F). HT29 cells are known to undergo necroptosis after treatment with a combination of tumor necrosis factor–α (TNF-α), cycloheximide (CHX), and Z-VAD-FMK (Fig. 4G) (44). Undetectable abundance of RIPK3 in U2OS cells suggests that necroptosis may not have been induced in U2OS cells. The combination treatment of TNF-α, CHX, and Z-VAD-FMK was not able to trigger cell death in U2OS cells (Fig. 4G), although it induced RIPK1 phosphorylation (Fig. 4H). The same treatment induced necroptosis in HT29 cells, and HT29 cell death was prevented by a necroptosis inhibitor, necrostatin-1 (Fig. 4G).

Notably, the pattern of phosphorylated RIPK1 upon glucose deprivation was different from the one induced by the combination treatment with TNF-α, CHX, and Z-VAD-FMK (fig. S9B), suggesting that upon glucose deprivation, RIPK1 may undergo different posttranslational modifications, for example, phosphorylation at different sites or a combination of phosphorylation and other modifications. Whereas RIPK1 phosphorylation by a combination treatment of TNF-α, CHX, and Z-VAD-FMK in U2OS cells was prevented by necrostatin-1 (Fig. 4H), RIPK1 phosphorylation by glucose removal was not blocked by necrostatin-1 (fig. S9C). The inability of necrostatin-1 to prevent glucose deprivation–induced RIPK1 phosphorylation and cell death suggests that the glucose deprivation–induced RIPK1 phosphorylation is not due to RIPK1 autophosphorylation and that RIPK1 kinase activity is not required for glucose deprivation–induced cell death. Together, these data suggest that glucose deprivation induces RIPK1-dependent cell death that is not necroptosis.

Glucose deprivation induces PP2Ac demethylation

Next, we investigated the upstream regulation of RIPK1 phosphorylation upon glucose deprivation. We noticed that the RIPK1 electrophoretic mobility shift after glucose deprivation was similar to that of cells treated with an inhibitor of serine/threonine phosphatases calyculin A in the presence of glucose (fig. S10A). We assessed the phosphorylation of other intracellular proteins that have rapid intracellular phosphorylation/dephosphorylation cycles by Phos-tag Western blotting and found that the PP2A substrates casein kinase 1δ (CSNK1D) and dishevelled segment polarity protein 2 (DVL2) (4547) were also phosphorylated after glucose deprivation in U2OS cells (fig. S10B). On the basis of these results, we speculated that there was reduced activity of PP2A (serine/threonine phosphatase) upon glucose deprivation. Because PP2A activity can be reduced by demethylation of its catalytic subunit, PP2Ac (18, 19), we examined the effect of glucose deprivation on PP2Ac demethylation. We observed induction of PP2Ac demethylation after glucose withdrawal (Fig. 5A). NaOH-treated extracts, which contain fully demethylated PP2Ac (48), was used as a positive control (Fig. 5A). These data suggest a role for demethylation of PP2Ac in the accumulation of phosphorylated RIPK1 and cell death. PP2Ac demethylation is downstream of 2-DG, because the addition of 2-DG blocked PP2Ac demethylation caused by glucose deprivation (Fig. 5B). In addition to U2OS cells, the other sensitive cancer cell lines U251MG and SW480 also had PP2Ac demethylation upon glucose withdrawal, which was attenuated by addition of 2-DG (Fig. 5C and fig. S10C). On the other hand, the amounts of demethylated PP2Ac remained low in the cancer cell lines A549 and H1299 and the normal cell lines WI-38, IMR-90, and MCF-10A that were resistant to glucose deprivation–induced cell death (Fig. 5C and fig. S10C). As we expected, supplementation with pyruvate did not inhibit PP2Ac demethylation (Fig. 5D), consistent with our aforementioned findings that glucose deprivation–induced cell death was independent of ATP amount (Fig. 2C). Finally, concentrations of glucose as low as 0.025 mM prevented PP2Ac demethylation (Fig. 5E), consistent with the ability of glucose to prevent cell death (fig. S1B). The prevention of PP2Ac demethylation with this extremely low glucose concentration suggests that the PP2A-related cell death pathway may be distinct from well-known nutrient stress pathways, such as AMPK-mTOR (mechanistic target of rapamycin kinase) signaling, which is triggered by low (~1 mM) glucose (49), as in the physiological tumor environment. Consistent with a previous report (49), shifting from high glucose (25 mM) to low glucose (1 mM) in the culture media induced AMPK activation, as evidenced by Thr172 phosphorylation (Fig. 5F) (50, 51). However, shifting from low glucose (1 mM) to no glucose (0 mM) did not further induce, but rather diminished, AMPK phosphorylation (Fig. 5F). Only the absence of glucose condition could induce PP2Ac demethylation and RIPK1 phosphorylation, whereas the low glucose (1 mM) condition did not (Fig. 5F), suggesting that AMPK activation is not involved in triggering PP2Ac demethylation and RIPK1 phosphorylation. AMPK knockdown also did not affect PP2Ac demethylation and RIPK1 phosphorylation after glucose deprivation (fig. S6A), consistent with an inability to prevent cell death by AMPK knockdown (fig. S6B). Furthermore, we evaluated mTOR kinase pathway activity, which is inhibited by AMPK activation (52). A marker of mTOR activation, namely, phosphorylation of the mTOR substrate S6 did not change regardless of the presence or absence of glucose (fig. S11). In addition, knockdown of ATG5, ATG7, ATG12, and BECN1 did not affect PP2Ac demethylation and RIPK1 phosphorylation after glucose deprivation (fig. S5, B to E). Together, these results indicate that PP2Ac demethylation and RIPK1 phosphorylation induced by glucose deprivation are independent of the AMPK-mTOR-autophagy–regulated nutrient stress pathway.

Fig. 5 Glucose deprivation induces PP2Ac demethylation.

(A) Western blotting analysis for PP2Ac in U2OS cells cultured as indicated for 4 hours. NaOH-treated protein extract was used as a positive control for fully demethylated PP2Ac. Blots are representative of three independent experiments. (B and C) Western blotting analysis for PP2Ac in the indicated cell lines treated as indicated for 4 hours. Biological duplicates are shown for each condition. Blots are representative of three independent experiments. (D) Western blotting analysis for PP2Ac in U2OS cells treated as indicated for 4 hours. Blots are representative of three independent experiments. (E) Western blotting analysis for PP2Ac in U2OS cells cultured with various doses of glucose as indicated for 4 hours in the absence of serum. Blots are representative of three independent experiments. (F) Western blotting analysis in U2OS cells cultured with a different dose of glucose as indicated for 4 hours. Blots are representative of three independent experiments.

PP2Ac demethylation is required for cell death induced by glucose withdrawal

Time course experiments after glucose withdrawal revealed that demethylated PP2Ac was reproducibly detectable as a spike in abundance 15 min after glucose deprivation and then increased steadily from 1 hour onward (Fig. 6A). RIPK1 phosphorylation was observed after PP2Ac demethylation (Fig. 6A). These data suggest that PP2Ac demethylation precedes RIPK1 phosphorylation and cell death. To test this, we knocked down RIPK1 and monitored PP2Ac demethylation. Although knockdown of RIPK1 restored cancer cell viability (Fig. 4D), it did not substantially affect the induction of PP2Ac demethylation by glucose deprivation (Fig. 6B), confirming that PP2Ac demethylation precedes RIPK1 phosphorylation and cell death.

Fig. 6 PP2Ac demethylation is required for glucose deprivation–induced cell death.

(A) Time course of Western blotting analysis in U2OS cells cultured in the absence of both serum and glucose for the indicated time. Blots are representative of three independent experiments. (B) Western blotting analysis in siRNA-transfected U2OS cells cultured with or without 1 mM glucose for 4 hours. Blots are representative of three independent experiments. (C) Western blotting analysis. U2OS cells were grown with or without cycloheximide in the absence of glucose and serum, for the indicated time. TP53 was used as a positive control for the effect of CHX. Blots are representative of three independent experiments. (D) Western blotting analysis in siRNA-transfected U2OS cells cultured with or without 1 mM glucose for 4 hours. Blots are representative of three independent experiments. (E) Representative phase-contrast images of siRNA-transfected U2OS cells cultured as in (D). Results are representative of three independent experiments. Images were taken by IncuCyte ZOOM System. Scale bar, 50 μm. (F) PI exclusion assay using siRNA-transfected U2OS cells treated as in (D). Shown is the mean percentage of dead cells ± SD from more than three independent experiments. **P < 0.01, *P < 0.05, unpaired two-tailed Student’s t test.

We reasoned that glucose deprivation could promote PP2Ac demethylation either by inhibiting the methylation of newly synthesized, unmethylated PP2Ac or by demethylating the existing, methylated PP2Ac. To distinguish between these possibilities, we treated the cells with CHX to inhibit new protein synthesis. Glucose withdrawal efficiently induced PP2Ac demethylation, even in the presence of CHX (Fig. 6C). This is consistent with active demethylation of the existing, methylated PP2Ac after glucose removal.

PPME1 is the only known enzyme that demethylates PP2Ac, and no other substrate of PPME1 is known (19, 20). To determine whether demethylation of PP2Ac by PPME1 is required for cell death, we knocked down PPME1 using two independent siRNAs and monitored PP2Ac demethylation and cell death after glucose withdrawal. PPME1 knockdown inhibited PP2Ac demethylation and RIPK1 phosphorylation (Fig. 6D) and significantly prevented cell death at 4 hours of glucose withdrawal in U2OS cells (Fig. 6, E and F). This effect was not cell type–specific, because PPME1 knockdown inhibited PP2Ac demethylation and protected the cells from death in other glucose depletion–sensitive cell lines SW480 and U251MG (fig. S12, A and B). Time-lapse microscopy showed that PPME1 knockdown delayed U2OS cell death after glucose deprivation (movies S3 to S6). Delay, but not complete prevention, of cell death could be due to incomplete knockdown of PPME1 (Fig. 6D). Together, these data suggest that PP2Ac demethylation by PPME1 is necessary for glucose deprivation–induced cell death.

Glucose deprivation triggers a cytosolic influx of calcium

We next investigated the mechanism by which glucose deprivation induced PP2Ac demethylation and cancer cell death. Glucose deprivation activates calmodulin-dependent protein kinases in Schizosaccharomyces pombe (53, 54), and PPME1 is phosphorylated by CAMK1 (22). Therefore, we tested whether CAMK1 is necessary for glucose deprivation–induced PP2Ac demethylation and cell death. We used two independent siRNAs to knock down CAMK1 in U2OS cells (Fig. 7A). Each of these CAMK1 siRNAs attenuated PP2Ac demethylation and RIPK1 phosphorylation (Fig. 7A) and significantly reduced cell death at 4 hours of glucose deprivation (Fig. 7B). CAMK1 knockdown delayed U2OS cell death after glucose deprivation, as shown by delayed morphological changes monitored up to 12 hours (fig. S13A). These results indicate that CAMK1 critically mediates these phenotypes.

Fig. 7 Glucose deprivation induces calcium signaling.

(A and B) siRNA-transfected U2OS cells were cultured with or without glucose for 4 hours. Protein lysates were analyzed by Western blotting (A). The mean percentage of dead cells ± SD analyzed by PI exclusion assay is shown in (B). **P < 0.01, unpaired two-tailed Student’s t test. Data are from three independent experiments. (C to H) U2OS cells were cultured as indicated for 4 hours and stained with Fluo-4AM for intracellular calcium before imaging. Representative phase-contrast and fluorescent images of cells are shown (C, E, and G; scale bars, 50 μm), and intracellular calcium amounts were quantified (D, F, and H) relative to the control condition (first bar). Data are means ± SD from three independent experiments. ***P < 0.001, unpaired two-tailed Student’s t test. (I to J) Western blotting analysis. U2OS cells were treated as indicated for 4 hours. Blots are representative of three independent experiments. (K) PI exclusion assay of U2OS cells treated as indicated for 4 hours. Bars are mean percentage of dead cells ± SD from three independent experiments.

Because CAMK1 is regulated by intracellular calcium concentration (23), we investigated whether glucose deprivation affects the amount of calcium in cells. U2OS cells were incubated with the fluorescent calcium indicator Fluo-4AM. We found that U2OS cells cultured in the absence, but not in the presence, of glucose displayed green fluorescence (Fig. 7, C and D), indicating that glucose withdrawal increased the intracellular calcium concentration significantly. In addition, we found that the other glucose depletion–sensitive cell lines U251MG and SW480 displayed cytosolic calcium staining in the absence of glucose but that the insensitive cancer cell lines A549 and H1299 and the normal cell lines IMR-90 and MCF-10A did not (fig. S13B). Furthermore, 2-DG blocked the increase in intracellular calcium induced by glucose deprivation (Fig. 7, C and D). Thus, a cytosolic calcium influx correlates with the PP2Ac demethylation and cell death phenotypes regulated by glucose and 2-DG. Next, we characterized the calcium signaling induced by glucose deprivation. Time-lapse microscopy was used to evaluate the kinetics of changes in intracellular calcium upon glucose withdrawal. Cytosolic calcium staining became pronounced within 10 min of glucose deprivation (movies S7 and S8), which is earlier than the PP2Ac demethylation (Fig. 6A). Consistent with these data, knockdown of CAMK1, PPME1, or RIPK1 did not affect the calcium influx induced by glucose deprivation (fig. S13C), suggesting that calcium influx precedes CAMK1 activation, PP2Ac demethylation, and RIPK1-dependent cell death.

The increase in cytosolic calcium could be due to an influx of calcium from outside of the cell or from calcium storage organelles within the cells. To determine the source of calcium, we compared the effects of nifedipine, an antagonist of L-type voltage-sensitive calcium channels on the plasma membrane (55), and ruthenium red (RR), an inhibitor of calcium release from the intracellular stores, on the increased calcium after glucose deprivation. We found that nifedipine but not RR prevented the glucose deprivation–induced calcium influx (Fig. 7, E to H), PP2Ac demethylation (Fig. 7, I and J), and cell death (Fig. 7K). These data support a model in which glucose deprivation induces an influx of calcium from outside the cell, which in turn triggers CAMK1 activation, PPME1-mediated PP2Ac demethylation/inactivation, and RIPK1-dependent cell death.

Voltage-sensitive calcium channel Cav1.3 (CACNA1D) is required for glucose depletion–induced cell death

Because nifedipine inhibited calcium influx and cell death induced by glucose deprivation, we investigated the calcium channels that were potentially involved in this pathway. Nifedipine affects L-type calcium channels. There are four L-type calcium channels that differ in their α-1 subunits: CACNA1S (also known as Cav1.1), CACNA1C (also known as Cav1.2), CACNA1D (also known as Cav1.3), and CACNA1F (also known as Cav1.4). We examined the mRNA expression of these channels in U2OS cells and found that only CACNA1C and CACNA1D were expressed (Fig. 8A). We depleted CACNA1C and CACNA1D by RNA interference (fig. S14) and found that knockdown of CACNA1D but not CACNA1C prevented calcium influx, PP2Ac demethylation, and morphological changes associated with cell death after glucose deprivation (Fig. 8, B and C), suggesting that CACNA1D is required to trigger this cell death pathway.

Fig. 8 Glucose deprivation induces plasma membrane depolarization.

(A) Representative reverse transcription polymerase chain reaction (RT-PCR) for the expression of L-type calcium channel α-1 subunits in U2OS cells from three independent experiments. PCR products were visualized by agarose gel electrophoresis. PCRs without a cDNA were included as negative controls. Expression of TATA box-binding protein (TBP) is shown as a positive control for amplification. (B and C) siRNA-transfected U2OS cells were cultured with or without 1 mM glucose for 4 hours. Cells were stained with Fluo-4AM for intracellular calcium before imaging. Representative phase-contrast and fluorescent images of cells are shown in (B). Scale bars, 50 μm. Protein lysates were analyzed by Western blotting analysis (C). Data are representative of three independent experiments. (D) Representative fluorescent images of DiBAC4-stained U2OS cells for membrane depolarization from three independent experiments. Cells were placed in media as indicated. KCl serves as a positive control. Scale bar, 50 μm. (E to G) Quantification of membrane depolarization from more than three independent experiments. DiBAC4-stained cells were placed in media as indicated, and fluorescent intensity was measured by a plate reader. Bars represent mean (±SD) DiBAC4 fluorescence in each sample relative to the control (first bar). *** P < 0.001, unpaired two-tailed Student’s t test.

L-type calcium channels are voltage sensitive and open in response to plasma membrane depolarization (27, 28). We therefore investigated whether depolarization occurred after glucose deprivation. We stained U2OS cells with DiBAC4, an oxonol fluorescent dye that detects plasma membrane depolarization (56). U2OS cells displayed depolarization after glucose deprivation (Fig. 8, D and E). KCl (120 mM) was used as a positive control because it is known to induce plasma membrane depolarization (56). As we expected, 2-DG inhibited glucose deprivation–induced depolarization (Fig. 8, D and E), consistent with its ability to prevent PP2Ac demethylation and cell death (Fig. 5B and 2B). We did not observe membrane depolarization of the insensitive cancer cell line H1299 upon glucose deprivation (Fig. 8F).

We also tested whether membrane depolarization was upstream of the induction of calcium influx. We treated U2OS cells with nifedipine in the presence or absence of glucose and stained the cells with DiBAC4. Although nifedipine completely inhibited calcium influx (Fig. 7, E and F), it did not inhibit membrane depolarization, as shown by the DiBAC4 fluorescence (Fig. 8, D and G), suggesting that plasma membrane depolarization precedes calcium influx upon glucose deprivation.

Targeting calcium signaling and glucose transport is a potential therapeutic intervention

Micromolar concentrations of glucose or 2-DG were sufficient to prevent cell death (figs. S1B and S2A). We hypothesized that the cell lines that are resistant to glucose deprivation–induced cell death may have higher intracellular glucose levels that protect cells from death after extracellular glucose withdrawal. We measured intracellular glucose amounts of the cells cultured in the presence or absence of glucose for 4 hours and found that resistant cancer cell lines A549 and H1299, as well as normal primary fibroblasts IMR-90 and WI-38 cells and nontransformed epithelial MCF-10A cells, maintained higher amounts of intracellular glucose compared to sensitive cell lines (SW480, U2OS, U251MG, SaOS2, and U87MG) after glucose withdrawal (Fig. 9A). Thus, intracellular glucose amounts correlated well with the sensitivity to glucose removal.

Fig. 9 Therapeutic intervention targeting calcium signaling and glucose transport.

(A) Intracellular glucose amounts in cells cultured with or without 25 mM glucose for 4 hours in the absence of serum. Intracellular glucose amounts were normalized against the total protein amount. Shown are means ± SD from three independent replicates. Asterisk (*) indicates cell lines that are sensitive to glucose deprivation. (B) Western blotting analysis for PP2Ac in cells cultured as indicated with 1 mM glucose without serum for 1 day. Blots are representative of three independent experiments. (C) Representative fluorescent images of Fluo-4AM–stained cells for intracellular calcium amounts. Cells were cultured with or without STF-31 in the presence of 1 mM glucose and absence of serum for 1 day. Cells were stained with Fluo-4AM before imaging. Images were representative of three independent experiments. Scale bars, 50 μm. (D) A representative analysis of PI exclusion assay in cells cultured as described in (B). Shown is the percentage of PI-positive dead cells from three independent experiments. (E) Schematic model of glucose deprivation–induced cell death. A, B, and C are PP2A complex subunits; Me, methylation; P, phosphorylation.

The inability of these sensitive cancer cell lines to maintain baseline intracellular glucose amounts upon glucose deprivation prompted us to investigate whether this unique metabolic property in cancer cells could be therapeutically targeted without affecting normal cells. Because it is impractical to completely deplete the source of glucose for cancer cells in patients, we evaluated the effects of STF-31, an inhibitor of the glucose transporter GLUT1 (8), on a sensitive cancer cell line U2OS cells and a normal primary cell line WI-38 cells. We reasoned that STF-31 may induce cell death in U2OS cells, whose intracellular glucose was not maintained upon glucose removal, but not in WI-38 cells, which maintained relatively high intracellular glucose upon glucose removal. We found that STF-31 induced slight PP2Ac demethylation and increased intracellular calcium amounts in U2OS cells but not in WI-38 cells (Fig. 9, B and C, and fig. S15). These data suggest that inhibition of glucose transport by STF-31 reduced the intracellular glucose enough to trigger calcium signaling in U2OS cells but not in WI-38 cells.

Given our findings that an influx of calcium promoted PP2Ac demethylation, we enhanced the induction of intracellular calcium by treating cells with a combination of STF-31 and thapsigargin, which increases cytosolic calcium concentrations by inhibiting the sarcoplasmic/endoplasmic reticulum Ca2+–adenosine triphosphatase (SERCA). U2OS cells treated with a combination of STF-31 and thapsigargin had more demethylated PP2Ac than cells treated with either drug alone (Fig. 9B). Treatment with either drug alone or in combination did not increase the amount of demethylated PP2Ac in WI-38 cells (Fig. 9B). We found that a combination of STF-31 and thapsigargin caused enhanced cell death in U2OS cells compared to either drug alone (Fig. 9D). By contrast, WI-38 cells did not undergo cell death after treatment with STF-31 or thapsigargin alone or in combination (Fig. 9D). Therefore, cancer cells that are sensitive to glucose deprivation can potentially be specifically targeted by combining glucose transport inhibition with enhanced cytosolic calcium accumulation without affecting normal cells.

DISCUSSION

Some cancer cell types are dependent on glucose and aerobic glycolysis and subsequent lactic acid fermentation, rather than oxidative phosphorylation, to fuel cell functions and survival, a phenomenon called the Warburg effect (4). However, in this study, we found that the differential sensitivity of some cancer cell lines to glucose deprivation was independent of the Warburg effect and the inhibition of glycolysis and instead correlated with the intracellular amount of glucose and the demethylation status of PP2Ac. In cells that were sensitive to it, glucose deprivation triggered plasma membrane depolarization, which led to an influx of calcium, CAMK1- and PPME1-dependent PP2Ac demethylation, RIPK1 phosphorylation, and the subsequent induction of RIPK1-dependent cell death (Fig. 9E), a process prevented by the glucose, as well as the glucose analog and glycolytic inhibitor 2-DG.

2-DG has been tested as a potential therapeutic agent because it inhibits glycolysis (57, 58). It also inhibits N-glycosylation (59) to induce cell death. However, we found that as little as 0.025 mM 2-DG (or glucose) was protective rather than inhibitory of cell survival in a subset of cancer cell lines. At this low concentration, it is unlikely that glycolysis is fully functional. The Km (Michaelis constant) of glycolytic enzymes for glucose, 2-DG, and glucose metabolites in glycolysis range from 0.1 to 5 mM (60, 61), and thus, lower concentrations are unlikely to be sufficient to generate metabolites for the entire glycolytic pathway. Although our data do not exclude the possibility that metabolites derived from glucose or 2-DG may be involved, we think that the role of the glycolytic pathway in cell survival is minimal in the cell culture conditions used in this study.

Although both glucose and 2-DG enhance O-glycosylation in some cell types (36, 62), our data suggest that O-glycosylation is not involved in protecting cancer cells against cell death. We hypothesize that low concentrations of glucose, 2-DG, or its phosphorylated counterpart glucose-6-phosphate directly or indirectly regulate the plasma membrane potential to limit the influx of Ca2+ to the cytoplasm. Thus, in their absence, Ca2+ influx triggers PP2A-mediated RIPK1-dependent cell death. Notably, these findings also indicate that some glucose analogs (including 2-DG) prevent, rather than promote, cancer cell death and could support cancer proliferation in some contexts.

Glucose deprivation induces various types of cell death, including apoptosis (63), necrosis (6466), necroptosis (67, 68), and autophagic cell death (69, 70). Our data suggest that RIPK1, although known to regulate apoptosis (71) and necroptosis (72), is a key mediator of glucose deprivation–induced cell death that is neither necroptosis nor apoptosis but might be a previously unknown type of programmed necrosis. Relative to our understanding of the signaling pathway downstream of RIPK1, the upstream kinase(s) and phosphatase(s) of RIPK1 are unknown. Although more work is needed to firmly establish mechanistic links, our findings reveal that PP2A is a possible direct or indirect upstream phosphatase of RIPK1. Upon glucose deprivation, RIPK1 underwent posttranslational modifications—namely, phosphorylation—that was apparently mediated differently from that induced by necroptosis inducers, suggesting that there are kinases and phosphatases for RIPK1 that specifically respond to glucose deprivation. It will be interesting to explore any changes in or effects of other posttranslational modifications of RIPK1, such as ubiquitylation (71, 73), alone or in combination with phosphorylation, in the cellular response to glucose deprivation.

The C-terminal methylation of PP2Ac promotes the function of the phosphatase by facilitating heterotrimer formation (74). Induction of PP2Ac demethylation may therefore inactivate a subset of these PP2A complexes (14, 21, 75), thereby increasing the phosphorylation of targets. Such a mechanism would be distinct from others that inactivate PP2A heterotrimers (76), such as DNA tumor virus small T antigens (77), and mutations in the genes encoding the scaffolding PP2Aa subunit, which are found in lung and colon cancers and alter the formation of the heterotrimer (7880). These findings suggest that the loss of specific subsets of PP2A promotes tumor formation. In contrast to some studies showing that PP2A activation is associated with the induction of cancer cell death (76, 81), but consistent with others that suggest PP2A inactivation has pro–cell death roles (8286), our data show that demethylation of a subset of PP2Ac can trigger cell death in some cancer cell lines. Thus, the effect of PP2A on tumor suppression, as well as its role in cell death, is likely influenced by cell context and the specific B subunits involved.

Although we found that glucose depletion induces membrane depolarization, it is not clear precisely how it does so. Plasma membrane potential is determined by relative ion concentration, mainly of Na+ and K+, across the membrane (87). Upon membrane depolarization (that is, increased positive charge inside cells), voltage-sensitive L-type calcium channels rapidly open to trigger calcium influx to the cytoplasm, which regulates various biological processes, such as cardiac musculoskeletal contraction, neuronal transmission, and nutrient absorption (8895). Thus, potentially, glucose deprivation–induced depolarization involves an influx of Na+ into the cells (by Na+ channel opening), a rise of intracellular K+ (by K+ channel closure), or a combination of both, which triggers calcium influx through L-type Ca2+ channels. The identification of the precise mechanism might enable therapeutic development to trigger cell death in glucose-dependent tumors. Furthermore, how glucose regulates the plasma membrane potential may be relevant to understanding the function and survival of other human cell types, particularly those that are highly dependent on glucose, such as neurons.

It will also be important to understand how increased Ca2+ concentrations and consequently CAMK1 activity regulates PPME1 to demethylate PP2Ac. Although PPME1 is phosphorylated by CAMK1 (22), it is unclear how this phosphorylation specifically affects its function. Our data and the mechanism we propose for Ca2+-CAMK1-PP2A–mediated cell death in cancer are distinct from, and in some cases are in contrast to, previously published reports related to the interactions between glucose, CAMK kinase (CAMKK), CAMK1, CAMK2, and PP2A in cell survival and cell death (53, 54, 9698). There is also a report, albeit unrelated to glucose, showing that PP2A can inhibit CAMK1 (22), potentially hinting at an unknown feedback or reciprocal mechanism of regulation. These mechanistic differences may be due to differences in the cell type, species of origin, or the extent or nature of nutrient depletion. Whether calcium influx– and PP2Ac demethylation–mediated cell death are conserved among other species remains to be elucidated. Increases in cytoplasmic calcium have been associated with the induction of cell death by TNF (99) or viral infection (100). It would be interesting to know whether PP2A is also involved in these processes.

Although we referred to a subset of cancer cells that undergo cell death by glucose deprivation as “sensitive” and to the others as “insensitive,” an exact classification is difficult because cells’ responses to glucose deprivation vary. Even in cells that are vulnerable to glucose deprivation, the latency to cell death varies. Thus, the classification is relative and can be affected by numerous factors. A cell’s ability to salvage and retain intracellular glucose, such as through breakdown of intracellular glycogen or a reverse reaction of PPP, may enable compensation for a low extracellular supply of glucose. Conversely, some cancer cells that are addicted to glycolysis, such as through overexpression of glycolytic enzymes (101), and/or are less efficient at gluconeogenesis may tend to have lower amounts of intracellular glucose. Because these phenomena are not seen in normal/healthy cells, this inability of some tumors to maintain intracellular glucose levels could be an unappreciated Achilles’ heel that might be therapeutically targeted, such as by enhancing the Ca2+-PP2A-RIPK1–cell death pathway, in tumors while sparing healthy tissue. A molecular signature of cancers with reduced ability to maintain intracellular glucose would help develop that strategy for clinical application.

In addition, varied sensitivity to glucose deprivation could also reflect dependency on alternate nutrient sources, such as glutamine (102105), fatty acids (106, 107), lactate (108), or acetate (109), or reliance on other metabolic pathways, such as oxidative phosphorylation (9) and autophagy (103, 110). This metabolic reprogramming can be caused by intrinsic factors, such as the deregulation of the proteins MYC (105), AKT (111), and RAS (104, 112), or by input from the tumor microenvironment, such as a supply of fatty acids from adipocytes (113) or lactate from cancer-associated fibroblasts and neighboring tumors (108, 114). Future investigations will determine how to manipulate these factors to make tumors sensitive to glucose depletion–induced cell death.

Inhibitors of glucose metabolic flux, such as 2-DG, have been developed to treat cancer, but they have had limited success (57, 58). This may be partially because, as our study here suggests, 2-DG may prevent RIPK1-dependent cell death in the context of limited glucose. Another approach is to block glucose uptake with glucose transporter inhibitors. GLUT1 is frequently overexpressed in many types of cancers (115); however, normal cells also require GLUT1 to function. Thus, identifying the therapeutic window and those patients who are most likely to benefit is critical for the clinical success of GLUT1 inhibitors. For example, it may be possible to screen patient tumor samples for PP2Ac methylation status and/or cell death after brief culture in the absence of glucose. In addition, a glucose uptake inhibitor may be more efficacious in combination with another drug, as we found with the GLUT1 inhibitor STF-31 and thapsigargin. Both drugs have been tested in preclinical or clinical trials (8, 116), respectively, and their combination may increase the therapeutic window for clinical use. These hypotheses remain to be assessed, but together, our study reveals an avenue to explore for targeting glucose-sensing pathways in cancer.

MATERIALS AND METHODS

Cell cultures and reagents

U2OS, SaOS2, SW480, U87MG, U251MG, A549, H1299, HT29, IMR-90, WI-38, and MCF-10A were purchased from American Type Culture Collection (ATCC). All cells except U251MG were cultured in high glucose (25 mM) Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, Life Technologies) supplemented with 10% FBS (HyClone, GE Healthcare Life Science), penicillin (100 units/ml), and streptomycin (100 μg/ml; Gibco, Life Technologies) in 5% CO2-humidified atmosphere at 37°C unless otherwise stated. U251MG was cultured under the same condition supplemented with 1× MEM nonessential amino acids (Gibco, Life Technologies) and 1× sodium pyruvate (Gibco, Life Technologies). For the glucose- and serum-deprivation assays, cells were cultured in media without glucose and serum. For the glucose-deprivation assays, cells were cultured in media containing 10% dialyzed serum and no glucose unless otherwise stated. For the induction of apoptosis, cells were irradiated with 60 J/m2 of UV, and protein was extracted 20 hours after irradiation. To inhibit apoptosis, cells were pretreated with 50 μM Z-VAD-FMK for 12 hours, followed by the indicated treatment in media containing 50 μM Z-VAD-FMK. To inhibit caspase activity, we pretreated the cells with 10 μM Z-VAD-FMK for 12 hours, followed by the indicated treatment in media containing 10 μM Z-VAD-FMK. For the induction of necroptosis, cells were treated with a combination of TNF-α (100 ng/ml), CHX (10 μg/ml), and 20 μM Z-VAD-FMK in the presence of glucose and 10% serum. To inhibit necroptosis, we pretreated the cells with 10 μM necrostatin-1 for 10 hours, followed by the indicated treatment in media containing 10 μM necrostatin-1. To inhibit new protein synthesis, we treated the cells with 100 μM CHX. Glucose, 2-DG, 6-DG, PUGNac, sodium pyruvate, CHX, chloroquine, dorsomorphin (compound c), TNF-α, rapamycin, nifedipine, RR, calyculin A, thapsigargin, PI, catalase, and hydrogen peroxide were purchased from Sigma-Aldrich. Necrostatin-1 (Abcam), STF-31 (Merck Millipore), and Z-VAD-FMK (Santa Cruz Biotechnology) were also purchased.

Dialyzed serum

The commercial dialyzed serum used (HyClone, GE Healthcare Life Science) contains less than 0.1 mM glucose. To further remove glucose, Slide-A-Lyzer G2 dialysis cassette (MWCO 10,000) (Life Technologies) was used. Dialysis was performed in a cold phosphate-buffered saline (PBS) buffer containing 1 mM phenylmethylsulfonyl fluoride for 24 hours in a cold room. PBS buffer was exchanged once during the dialysis.

siRNA silencing experiments

siRNAs targeting a specific gene were obtained from Ambion, Invitrogen, or Qiagen. Control siRNA (ON-TARGET plus nontargeting pool) was obtained from Dharmacon. The targeted siRNA sequences were: AMPK_1, CCCATCCTGAAAGAGTACCATTCTT; AMPK_2, CCCTCAATATTTAAATCCTTCTGTG; AMPK_3, ACCATGATTGATGATGAAGCCTTA; ATG5, CCUUUGGCCUAAGAAGAAAdTdT; ATG7_1, CCAAAGUUCUUGAUCAAUA; ATG7_2, GAAGAUAACAAUUGGUGUA; ATG12_1, GGGAAGGACUUACGGAUGU; ATG12_2, GCAGUAGAGCGAACACGAA; BECN1_1, GAUACCGACUUGUUCCUUA; BECN1_2, CUAAGGAGCUGCCGUUAUA; PPME1_1, GAAGGAAGUGAGUCUAUAAdTdT; PPME1_2, GGAAGAAAGCGGGACUUUUdTdT; RIPK1_1, GCAAAGACCUUACGAGAAUdTdT; RIPK1_2, CCACUAGUCUGACGGAUAAdTdT; CAMK1_1, AUACAGCUCUAGAUAAGAAdTdT; CAMK1_2, CCAUAGGUGUCAUCGCCUAdTdT; CACNA1C_1, CTGGTTTGGTTCGGTTATCTAdTdT; CACNA1C_2, TCCAGGGATGTTAGTCTGTATdTdT; CACNA1D_1, CACGCGAACGAGGCAAACTATdTdT; and CACNA1D_2, CCGGAACACGATACTGGGTTAdTdT. Cells were transiently transfected with siRNA using RNAiMAX according to the manufacturer’s instructions (Invitrogen, Life Technologies). All siRNAs were used at 20 nM for transfection.

RT-PCR experiments and analysis

Total RNA was extracted using RNeasy kit (Qiagen). cDNA was synthesized using iScript cDNA Synthesis kit (Bio-Rad). Reverse transcription polymerase chain reaction (RT-PCR) was performed with KAPA SYBR fast qPCR kit (KAPA Biosystems) using the CFX96 System (Bio-Rad). For assessing gene expression (Fig. 8A), the following primers were used: CACNA1C, TGATTCCAACGCCACCAATTC (forward) and GAGGAGTCCATAGGCGATTACT (reverse); CACNA1D, CGCGAACGAGGCAAACTATG (forward) and TTGGAGCTATTCGGCTGAGAA (reverse); CACNA1F, GATCCAGGAGTATGCCAACAA (forward) and GAAGGAAGACACATAGGCAGAG (reverse); CACNA1S, TTGCCTACGGCTTCTTATTCCA (forward) and GTTCCAGAATCACGGTGAAGAC (reverse); and TBP (TATA-binding box), CGCCGAATATAATCCCAAGC (forward) and TCCTGTGCACACCATTTTCC (reverse). PCR products were visualized by DNA agarose gel electrophoresis. The expected sizes of PCR products were 103 bp (base pair) (CACNA1C), 81 bp (CACNA1D), 100 bp (CACNA1F), 105 bp (CACNA1S), and 103 bp (TBP). For quantitative RT-PCR (fig. S14), these primers were used, corresponding to the respective siRNA: CACNA1C_#1, CCATTGTGTATGCCCAATAATTTGT (forward) and CAAACCCACCTGTACACCCA (reverse); CACNA1C_#2, ATGGGATCATGGCTTATGGCG (forward) and CCAGGTTGTCCACAGCAATG (reverse); CACNA1D_#1, GCAGCATCAACGGCAGC (forward) and CGGCTGAGAAGTTGGTCCTT (reverse); and CACNA1D_#2, AGAGGACCCCATCCGCA (forward) and GGCCCCTTTGTGGAGGAAA (reverse). Relative expression was calculated with Bio-Rad CFX manager software using the expression of TBP as an internal control. PCR products were verified by sequencing.

Protein analysis

For Western blotting analysis, all cells (including attached cells and floating cells) were lysed with 2% SDS lysis buffer (50 mM tris-HCl, pH 6.8, 10% glycerol, and 2% SDS). Lysates were separated by 8% SDS–polyacrylamide gel electrophoresis (PAGE) to detect phosphorylated RIPK1 (except in Figs. 4H and 6A, which were 11% SDS-PAGE), where 11% SDS-PAGE was performed for other protein detections. Detection was done by incubation with horseradish peroxidase–conjugated anti-mouse, anti-rabbit, or anti-goat immunoglobulin G (IgG) (Jackson ImmunoResearch), or anti-mouse IgM (Santa Cruz Biotechnology) secondary antibody followed by the reaction for chemiluminescence (SuperSignal, Pierce). Infrared fluorescence–conjugated anti-mouse, anti-rabbit, or anti-goat IgG secondary antibody (DyLight, Jackson ImmunoResearch) was also used for infrared fluorescence detection (LI-COR Odyssey). The following antibodies were used: anti-AMPK (Santa Cruz Biotechnology, clone 71.54, catalog no. sc-130394), anti-phospho-AMPK T172 (Cell Signaling, catalog no. 2531), anti-ATG5 (Cell Signaling, clone D5F5U, catalog no. 12994), anti-ATG7 (Cell Signaling, clone D12B11, catalog no. 8558), anti-ATG12 (Cell Signaling, clone D88H11, catalog no. 4180), anti-BECN1 (Cell Signaling, clone D40C5, catalog no. 3495), anti-CAMK1 (Santa Cruz Biotechnology, clone H-125, catalog no. sc-33165), anti-CASP3 (Cell Signaling, catalog no. 9661), anti-Dvl2 (Cell Signaling, clone 30D2, catalog no. 3224), anti-LC3 (Cell Signaling catalog no. 2775), anti-total mTOR (Cell Signaling, clone 7C10, catalog no. 2983), anti-RIPK3 (Cell Signaling, clone E1Z1D catalog no. 13526), anti-HSPA9 (Santa Cruz Biotechnology, clone H-155, catalog no. sc-13967), anti-C1QBP (Santa Cruz Biotechnology, D-19, catalog no. sc-10258), anti-TP53 (Santa Cruz Biotechnology, clone DO1, catalog no. sc-126), anti-demethylated PP2Ac (Santa Cruz Biotechnology, clone 4B7, catalog no. sc-13601), anti–cytochrome c (Thermo Fisher Scientific, clone 7H8.2C12), anti-RIPK1 (BD Biosciences, clone 38/RIP), anti–phospho-RIPK1 Ser166 (Cell Signaling, clone D1L3S, catalog no. 65746), anti-PARP (BD Biosciences, clone C2-10, catalog no. 556362), anti-actin (Merck Millipore, clone C4, catalog no. MAB1501), anti-PPME1 (Merck Millipore, catalog no. 07-095), anti-total PP2Ac (Merck Millipore, catalog no. 07–324), anti-S6 (Cell Signaling, clone 54D2, catalog no. 2317), anti–phospho-S6 Ser235/236 (Cell Signaling, clone D57.2.2E XP, catalog no. 4858), anti-methylated PP2Ac (custom-made antibody from Egon Ogris, Max F. Perutz Laboratories), anti-PP2Aa (custom-made antibody from D.M.V., Duke-NUS), anti-CSNK1D (custom-made antibody from Eli Lilly Laboratories), and anti-tubulin (Abcam, clone no. DM1A+DM1B). The alkaline demethylation assay was performed as previously described (48). Briefly, cells were lysed with 2% SDS lysis buffer and incubated with NaOH at 1 mM final concentration for 30 min at room temperature. Extracts were then neutralized with HCl for SDS-PAGE.

Phos-tag Western blotting

The preparation of the Phos-tag acrylamide gels (Wako Pure Chemical Industries Ltd.) and gel electrophoresis were done according to the manufacturer’s instructions. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes, and Western blotting was performed by the standard procedure.

λ Phosphatase assay

Cells were lysed with 0.5% NP-40 lysis buffer (0.5% NP-40, 50 mM Hepes, pH 7.5, 100 mM NaCl, 1× protease inhibitor cocktail without EDTA from WAKO), and endogenous phosphatase activity was heat-inactivated at 75°C for 5 min. A total of 400 units of λ phosphatase (New England Biolabs) was added to each lysate and incubated at 30°C for 1 hour. Enzyme activity was heat-inactivated at 95°C for 5 min before analyzing by Western blotting.

ATP measurement assay

Cells were lysed with 0.5% NP-40 lysis buffer (50 mM tris-HCl, pH 7.5, 100 mM NaCl, 0.5% NP-40, 1 mM phenylmethylsulfonyl fluoride, 1× protease inhibitor cocktail). Lysates were used for measuring ATP by CellTiter-Glo luminescent cell viability assay (Promega).

Glucose measurement assay

Cells were lysed with 0.5% NP-40 lysis buffer, and lysates were used for measuring glucose amount by Amplex Red Glucose/Glucose Oxidase Assay kit (Life Technologies). A glucose concentration of each lysate was calculated by comparing to the glucose standard curve as recommended by the manufacturer. Glucose concentration was further normalized against the protein amount of the lysate loaded and plotted in the graph.

PI exclusion assay

Cells were stained with PI to determine the percentage of cell death. Media containing floating cells were collected, combined with trypsinized cells, and centrifuged. The cell pellet was washed once with PBS. After centrifugation, cells were resuspended in PBS containing PI (10 μg/ml) and stained for 15 min. Data were collected with MACSQuant analyzer (Miltenyi Biotec). Quantification and analysis of the data were done with FlowJo software.

TMRE measurement assay

Cells were deprived of glucose for 4 hours. During the last 30 min of glucose deprivation, tetramethylrhodamine, ethyl ester (TMRE) (Life Technologies) was added to media for staining for 30 min at 37°C. Media containing floating cells were collected, combined with trypsinized cells, and centrifuged. Cells were washed once with PBS. TMRE signals were detected by flow cytometry using MACSQuant analyzer (Miltenyi Biotec) and analyzed with FlowJo software.

Calcium flux measurement assay

Cells were placed in phenol-free DMEM containing 10% dialyzed serum and 5 μM Fluo-4AM (Life Technologies) and incubated at 37°C for 15 min in the presence or absence of glucose. Phase-contrast and fluorescent images of cells were taken by Olympus Model IX71 inverted microscope. Quantification of fluorescence intensity was done with ImageJ software. The changes in intracellular calcium amounts were also monitored by IncuCyte ZOOM (Essen BioScience). After cells were placed in phenol-free DMEM containing 10% dialyzed serum and 5 μM Fluo-4AM in the presence or absence of glucose, images were taken every 10 min for 30 min.

Membrane potential depolarization measurement assay

Briefly, cells were stained with 5 μM DiBAC4 for 5 min and washed once with media containing the indicated reagents without DiBAC4 (Enzo Life Science), followed by the indicated treatment with DiBAC4. Images were immediately taken using a Leica fluorescence microscope (within 5 min). For quantification, fluorescence intensity of DiBAC4 was measured by the plate reader (Infinite M200, TECAN) (excitation, 490 nm; emission, 522 nm).

ROS measurement assay

After 3 hours of glucose deprivation, cells were trypsinized and washed once with PBS. Cells were then stained with 10 μM H2DCFDA (Thermo Fisher Scientific) in phenol red–free DMEM without glucose for 1 hour and analyzed by MACSQuant (Miltenyi Biotec) analyzer. FlowJo software was used for data analysis.

Time-lapse microscopy

Cells were placed in DMEM containing the indicated reagents and 10% dialyzed serum with or without glucose, and images were taken by IncuCyte ZOOM (Essen Bioscience). Phase-contrast or fluorescent images were taken every 10 or 30 min for the indicated hours depending on the experiments. We noticed that IncuCyte ZOOM has a tendency to give a slight delay in morphological changes and cell death compared to the condition in the regular 5% CO2 cell culture incubator.

Cytoplasmic fractionation

Cytoplasmic fractionation was performed as previously described (117), with a change in the concentration of the digitonin. Cytoplasmic-enriched fractions were prepared using digitonin solution (75 μg/ml).

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/11/512/eaam7893/DC1

Fig. S1. Glucose deprivation induces cell death in a subset of cancer cells.

Fig. S2. 2-DG rescues glucose deprivation–induced cell death that is independent from ROS induction.

Fig. S3. 2-DG rescues glucose deprivation–induced cell death independently from O-glycosylation.

Fig. S4. Glucose deprivation–induced cell death is not mediated by apoptosis.

Fig. S5. Glucose deprivation–induced cell death is independent from autophagy.

Fig. S6. AMPK is not involved in glucose deprivation–induced cell death.

Fig. S7. Glucose deprivation induces RIPK1 phosphorylation.

Fig. S8. Glucose deprivation induces RIPK1-dependent cell death.

Fig. S9. Glucose deprivation–induced cell death is not mediated by necroptosis.

Fig. S10. Glucose deprivation induces PP2Ac demethylation.

Fig. S11. Glucose deprivation induces PP2Ac demethylation and RIPK1 phosphorylation independently from the mTOR signaling pathway.

Fig. S12. PP2Ac demethylation is required for glucose deprivation–induced cell death.

Fig. S13. Glucose deprivation induces calcium influx into the cytoplasm.

Fig. S14. Knockdown efficiency of CACNA1C and CACNA1D.

Fig. S15. GLUT1 inhibitor increases cytoplasmic calcium concentration in U2OS cells.

Movie S1. PI staining in the presence of glucose.

Movie S2. PI staining in the absence of glucose.

Movie S3. Cell death in the presence of control siRNA and glucose.

Movie S4. Cell death in the presence of PPME1 siRNA and glucose.

Movie S5. Cell death in the presence of control siRNA but in the absence of glucose.

Movie S6. Cell death in the presence of PPME1 siRNA but in the absence of glucose.

Movie S7. Calcium staining in the presence of glucose.

Movie S8. Calcium staining in the absence of glucose.

REFERENCES AND NOTES

Acknowledgments: We thank A. Andersen, Life Science Editors, for editorial assistance. We thank T. W. Soong (NUS) and P. Yen (Duke-NUS) for helpful discussion for the manuscript and P. Xu, J. K. Cheong, K. Iwamoto, A. K. Guo, and Y. Lee at Duke-NUS for the various reagents, help, and suggestions. Funding: This work was supported by Duke-NUS Signature Programme Block Grant and the Singapore Ministry of Health’s National Medical Research Council grants (NMRC/CBRG/0031/2013 and NMRC/OFIRG/15nov049/2016 to K.I.; NMRC/STaR/0017/2013 to D.M.V.; NMRC/CBRG/0075/2014 to H.S.J.), Singapore Ministry of Education Academic Research Fund Tier 2 grants (MOE2013-T2-2-123 to K.I.; MOE2014-T2-2-071 to H.S.J.), and a Khoo Postdoctoral Fellowship Award (Duke-NUS-KPFA/2015/0001 to M.F.). Author contributions: H.Y.L., Y.I., M.F., and K.I. performed the experiments. E.O. and S.S. made the antibody specific to methylated PP2Ac. H.Y.L., Y.I., M.F, H.S.J., D.M.V., and K.I. designed the experiments and analyzed the results. H.Y.L., Y.I., E.O., D.M.V., and K.I. wrote the manuscript. Competing interests: H.Y.L., Y.I., D.M.V., and K.I. are inventors on international patent application no. PCT/SG2017/050208 for “A potential combination therapy using an inhibitor of glucose transport and an intracellular calcium inducer to target cancer metabolism.” E.O. serves as a consultant to Millipore Corporation. All other authors declare that they have no competing interests. Data and materials availability: There is a materials transfer agreement with the Medical University of Vienna (E.O.) for the methylated PP2Ac antibody.
View Abstract

Navigate This Article