Redox-dependent AMPK inactivation disrupts metabolic adaptation to glucose starvation in xCT-overexpressing cancer cells

Cancer cells exhibit a metabolic reprogramming to support their increased proliferation (Vander Heiden, 2011). Accelerated aerobic glycolysis is a distinctive metabolic property of cancer cells, which is also known as the Warburg effect. Given that this altered metabolism renders cancer cells highly dependent on glucose for viability, it provides unique vulnerabilities for the selective targeting of such glucose-dependent cancer cells (Hay, 2016).

Previously, we and others have shown that, although some cancer cells demonstrate strict dependency on glucose for their survival, others can adapt to glucose starvation, similar to what is seen with untransformed cells (Jeon et al., 2012; Lee et al., 2018). Tissue origin and oncogenic mutations generate a variety of metabolic reprogramming changes by different means and accordingly confer different dependence on glycolysis. Owing to the lack of known markers to indicate the degree of glucose dependence, glucose restriction strategies for cancer therapy have not been successful. Therefore, establishing biomarkers for the prediction of treatment sensitivity for glycolysis-targeting cancer therapy is necessary.

Several studies have shown that glucose deprivation triggers apoptosis and/or necrosis, or autophagy-related cell death (Jin et al., 2007). However, the underlying mechanism of glucose deprivation-induced cell death is context dependent owing to the differences in cell type and experimental conditions in those studies. Although this mechanism has been suggested to involve depletion of ATP or damage from accumulated reactive oxygen species (ROS), recent evidence indicates the involvement of signaling transduction to regulate cell survival and death in response to metabolic stress (Graham et al., 2012). Therefore, understanding the crosstalk between the metabolic perturbation and signal transduction that leads to cancer cell death is essential for improving therapeutic efficacy.

AMP-activated protein kinase (AMPK) is a serine/threonine kinase that plays a central role in maintaining cellular metabolic balance. AMPK regulates cellular adaptation to energetic stress, and its activation requires Thr172 phosphorylation in the kinase domain of the catalytic α subunit (Hindupur et al., 2015). This is mediated by upstream kinases, such as liver kinase B1 (LKB1; also known as STK11) (Shaw et al., 2004), Ca²⁺/calmodulin-dependent protein kinase 2 (Hawley et al., 2005) and TGFβ-activated kinase 1 (Xie et al., 2006). Upon activation by starvation, AMPK restores metabolic homeostasis by stimulating ATP-regenerating catabolic processes, such as induction of fatty acid oxidation (FAO) (Jeon et al., 2012), autophagy (Kim et al., 2011) and mitochondrial homeostasis (Toyama et al., 2016), while attenuating anabolic ATP-consuming biosynthetic processes, such as suppression of the mammalian target of rapamycin complex 1 (mTORC1) signaling-mediated protein synthesis (Gwinn et al., 2008; Inoki et al., 2003) and fatty acid synthesis (FAS) (Jeon et al., 2012). Therefore, proper AMPK activation under nutrient starvation is essential for overcoming metabolic stress. Understanding AMPK responses under metabolic stress across different cancer cells is necessary for gaining insights into glucose metabolism targeting therapeutic intervention.

Here, we identified that cancer cell lines highly sensitive to glucose deprivation express high levels of cystine/glutamate antiporter xCT (xCT; also known as SLC7A11). The expression levels of this protein determined the sensitivity to glucose deprivation by affecting intracellular NADPH levels under glucose deprivation. Collapse of intracellular redox system upon NADPH depletion caused AMPK inactivation via inhibitory oxidation, which rendered these glucose-dependent cancer cells unable to adapt to glucose deprivation. Therefore, we suggest that xCT could be used as a biomarker for glucose metabolism-targeting cancer therapy and that NADPH depletion is the metabolic determinant for glucose deprivation-induced cell death.

To investigate the signaling mechanisms underlying glucose deprivation-induced cell death, we screened multiple cell lines to determine their sensitivity to glucose deprivation. We incubated cells with glucose-free medium for the indicated period (0, 9 or 18 h) and measured cell death with a propidium iodide (PI) exclusion assay (Fig. 1A). As serum contains a very low amount of glucose, we cultured cells in medium with a dialyzed serum that has the necessary growth factors but retains undetectable levels of glucose. Two cancer cell lines (U2OS and U251MG; which will be referred to hereafter as ‘sensitive cell lines’) showed the highest sensitivity to glucose deprivation by showing rapid loss of cellular adhesion within 3 h, followed by PI-positive cell death within 9 h. The other two cancer cell lines (SW480 and MCF7) showed a modest response to glucose deprivation with cell death within 18 h (which will be referred to hereafter as ‘intermediate sensitive cell lines’). The remaining cancer cell lines (A375 and H1299), kidney embryonic epithelial cell line (HEK293T) and normal fibroblast cell lines (WI-38 and IMR-90) remained viable under glucose deprivation (which will be hereafter referred to as ‘resistant cell lines’; Fig. 1A; Fig. S1A). Next, we investigated whether glucose deprivation-induced cell death is a regulated form of cell death. The medium for culturing U2OS cells was changed from one with 25 mM glucose to one with 0 mM or 1 mM, which mimics the low glucose concentration in the physiological tumor environment (Birsoy et al., 2014; Hirayama et al., 2009). Although U2OS cells cultured in 1 mM glucose maintained viability, cells under glucose deprivation lost viability at 9 h after glucose withdrawal (Fig. 1B). Interestingly, the cell death in U2OS cells was neither apoptosis, necroptosis nor ferroptosis, as none of the inhibitors of regulated cell death [apoptosis inhibitor (Z-VAD-FMK), necroptosis inhibitors (necrostatin-1 and necrosulfonamide) and iron chelator (deferoxamine); Armenta and Dixon, 2020] were able to prevent glucose deprivation-induced cell death (Fig. 1B; Fig. S1B).

AMPK and mTORC1 are the central regulators of the cellular response to metabolic stress (Fig. S1C) (Hindupur et al., 2015). AMPK senses energy depletion and is activated to maintain energy homeostasis during glucose deprivation (Xiao et al., 2011; Zhang et al., 2017). mTORC1 alone without AMPK activation can also sense glucose availability (Orozco et al., 2020). Therefore, we asked how these central metabolic checkpoints are regulated in sensitive and resistant cell lines. We incubated multiple cell lines in media with different glucose concentrations [25 mM (standard glucose concentration in Dulbecco's modified Eagle's medium (DMEM), 1 mM (low glucose) or 0 mM (no glucose)] for 3 h and performed a western blotting analysis (Fig. 1C). In U2OS cells, Thr172 phosphorylation of AMPK, which is required for AMPK activation (Shaw et al., 2004), increased when incubated in 1 mM glucose compared to 25 mM glucose. However, western blotting of total AMPK and Thr172 phosphorylation under 0 mM glucose gave rise to smeared bands. In contrast, the intermediate sensitive and resistant cell lines did not show a clear electrophoretic mobility shift of total AMPK in the glucose-free condition. Of note, all the cell lines used in this study expressed LKB1, the upstream kinase responsible for AMPK Thr172 phosphorylation, which means that these cell lines have an intact LKB1–AMPK axis. To further investigate the activation state of AMPK, we examined mTORC1 signaling. When nutrient availability is limited, mTORC1 activity is decreased, in part through AMPK activation, which controls cell growth, proliferation, and autophagy (Gwinn et al., 2008; Inoki et al., 2003). mTORC1 phosphorylates 4EBP1 (also known as EIF4EBP1) at Thr37 and Thr46 to limit the function of 4EBP1 in reducing protein synthesis and cell growth (Fig. S1C). Interestingly, although the intermediate sensitive and resistant cell lines showed decreased mTORC1 activity under glucose deprivation (0 mM glucose), as indicated by the band pattern shift of 4EBP1 total and Thr37 and Thr46 phosphorylation to lower molecular mass as described elsewhere (Kang et al., 2013; Kim et al., 2021), the sensitive cell lines (U2OS and U251MG) maintained high mTORC1 activity, as indicated by a sustained high molecular mass of total and phosphorylated 4EBP1 (Fig. 1C). These data suggest that sensitive cell lines might maintain high mTORC1 and low AMPK activities even under glucose deprivation, whereas resistant cell lines might increase AMPK and/or reduced mTORC1 activities to adapt to the metabolically challenging conditions.

To gain a better insight into how the AMPK and the processes downstream of AMPK signaling are dysregulated in sensitive cell lines upon glucose deprivation, we examined 4EBP1 and acetyl-CoA carboxylase (ACC, also known as ACACA) phosphorylation in a time-dependent manner in U2OS cells (Fig. 1D). ACC is well-known as a direct substrate of AMPK. Under glucose deprivation, Thr172 phosphorylation of AMPK was increased almost immediately after glucose withdrawal but was diminished at around the 1–2 h time point. Similarly, ACC was also phosphorylated at Ser79 immediately upon glucose deprivation, which subsequently diminished, while total ACC amounts were unchanged. Furthermore, total and phosphorylation patterns of 4EBP1 was restored at later time points. The LC3B-II/LC3B-I ratio [between lipidated LC3 proteins (LC3-II) and non-lipidated LC3 proteins (LC3-I), a high value of which is one of the markers of autophagy (Klionsky et al., 2021); LC3 proteins are also known as MAP1LC3 proteins], which is positively regulated by AMPK and negatively regulated by mTORC1 (Kim et al., 2011), was also recovered at later time points. All those changes were not observed when cells were cultured with 1 mM glucose. These results indicate that after experiencing the transient AMPK activation upon glucose withdrawal, U2OS cells maintained low AMPK and high mTORC1 activities. A similar trend was observed in another sensitive cell line, U251MG cells (Fig. S1D). These data suggest that the sensitive cell lines display dysregulated AMPK-mTORC1 signaling under glucose deprivation.

Next, we evaluated AMPK activity in U2OS cells under glucose deprivation by using an ELISA assay, which measures phosphorylation on a direct substrate peptide of AMPK (Niopek et al., 2017; Zadra et al., 2014). AMPK activity was indeed significantly decreased under glucose withdrawal (Fig. 1E).

To investigate whether the AMPK-mTORC1 signaling dysregulation was due to altered energy status, we measured intracellular ATP levels in U2OS cells in a time-dependent manner (Fig. 1F). Although AMPK activity under 0 mM glucose condition had dropped by 90 min (Fig. 1E), ATP levels in U2OS cells cultured in 1 mM and 0 mM glucose were similar up to 3 h, when the initial stage of cell death (cell detachment) occurred in 0 mM glucose, but not in 1 mM glucose (Fig. S1A). These data suggest that another factor rather than reduced ATP levels caused dysregulation of AMPK-mTORC1 signaling under glucose deprivation in sensitive cells.

Given that AMPK activation is essential for overcoming metabolic stress (Jeon et al., 2012; Shackelford et al., 2013), we hypothesized that dysregulated AMPK signaling in sensitive cell lines underlies the failure to adapt to glucose deprivation. Treatment with A769662, an AMPK allosteric activator (Göransson et al., 2007), prevented glucose withdrawal-induced cell death in U2OS cells (Fig. 2A,B), suggesting that AMPK activation is sufficient to rescue cell death induced by glucose deprivation. Similar results were obtained in U251MG cells (Fig. S2A,B). High AMPK activity by A769662 treatment under glucose deprivation was confirmed by western blotting analysis, showing increased Thr172 phosphorylation of AMPK and phosphorylation of ACC and decreased mTORC1 activity as indicated by the band pattern shift of total and phosphorylated 4EBP1 to a lower molecular mass (Kang et al., 2013; Kim et al., 2021) both in U2OS (Fig. 2C) and U251MG cells (Fig. S2C). Knocking down of AMPKα1/2 (i.e. the two subunits encoded by PRKAA1 and PRKAA2) with siRNA accelerated glucose deprivation-induced cell death in U2OS cells (Fig. S2D–F). These data suggest that reduced AMPK activation in the sensitive cell lines contributes to the failure of metabolic adaptation, leading to glucose deprivation-induced cell death.

AMPK maintains catabolic and anabolic metabolic homeostasis by regulating protein synthesis, energy generation, autophagy and redox status (Hindupur et al., 2015). We investigated which functional pathway downstream of AMPK is critical for preventing cell death caused by glucose deprivation. One of the major pathways downstream of AMPK is the mTORC1 pathway, which promotes protein synthesis and cell growth (Fig. S1C). We first tested whether AMPK protected cancer cells from metabolic stress through mTORC1 inhibition. We treated U2OS cells with rapamycin to inhibit mTORC1 signaling because a previous study suggests that mTORC1 inhibition rescues metabolic stress-induced cell death (Choo et al., 2010). However, rapamycin treatment did not affect glucose deprivation-induced cell death (Fig. S3A,B), although it completely inhibited mTORC1 signaling as shown by the band pattern shift of total and phosphorylated 4EBP1 to a lower molecular mass as reported elsewhere (Kang et al., 2013; Kim et al., 2021) (Fig. S3C). These data suggest that glucose deprivation-induced cell death is not due to the lack of mTORC1 inhibition mediated by the reduced AMPK signaling.

Another important downstream target of AMPK is ACC, an enzyme that provides the malonyl-CoA substrate for the biosynthesis of fatty acids. AMPK inhibits ACC activity by direct phosphorylation at Ser79, leading to the inhibition of FAS and stimulation of FAO under metabolic stress (Fig. S1C). Therefore, we next determined whether regulation of fatty acid metabolism is sufficient to protect cancer cells from glucose deprivation. We treated U2OS cells with C75, a de novo FAS inhibitor that also facilitates FAO (Zhou et al., 2003). C75 treatment prevented glucose deprivation-induced cell death (Fig. 2D,E), suggesting that dysregulation of fatty acid metabolism contributes to cell death induced by glucose withdrawal, at least partly. Interestingly, C75 treatment activated AMPK signaling in the absence of glucose, as indicated by increased phosphorylation of AMPK at Thr172 and ACC, although its effect on mTORC1 signaling was marginal as shown by phosphorylation patterns of 4EBP1 (Fig. 2F). To further investigate the dysregulation of fatty acid metabolism, we employed etomoxir, a FAO inhibitor (Zhou et al., 2003). Etomoxir treatment accelerated glucose deprivation-induced cell death both in the absence and presence of C75 (Fig. 2G,H), suggesting that FAO contributes to protection against glucose deprivation-induced cell death. However, etomoxir did not completely antagonize the rescue effect of C75 on cell death. As C75 is a FAS inhibitor and FAO promoter, these data suggest that proper regulation of both FAS and FAO contributes to prevention of cell death under glucose withdrawal.

AMPK maintains redox balance by upregulating levels of NADPH, an essential electron donor providing the reducing power in cells, through the inhibition of FAS and promotion of FAO, the processes that consume and generate NADPH, respectively via phosphorylating ACC (Jeon et al., 2012). Glucose is the main source of NADPH, which is generated through the pentose phosphate pathway (PPP). However, under glucose deprivation, NADPH from the PPP is very limited. Therefore, we hypothesized that AMPK activation rescues glucose withdrawal-induced cell death by restoring redox balance through replenishing NADPH via fatty acid metabolism. To test this hypothesis, we first measured the NADPH levels in U2OS cells treated with or without A769662 under glucose deprivation. Glucose deprivation led to a significant depletion of NADPH that was partially rescued upon A769662 treatment (Fig. 3A). Similarly, C75 treatment partially restored NADPH depletion (Fig. 3B). Etomoxir, a FAO inhibitor, partially prevented the rescue effect of C75 on glucose deprivation-induced depletion of NADPH (Fig. 3C). These data suggest that dysregulation of fatty acid metabolism caused by AMPK inactivation contributes to the failure of replenishing NADPH levels under glucose deprivation, at least partly.

Since NADPH was drastically depleted upon glucose deprivation, we next measured the levels of mitochondrial and cytosolic ROS. Previous studies have shown that glucose deprivation induces oxidative stress through an accumulation of cytosolic and mitochondrial ROS, which was considered as a metabolic determinant of glucose deprivation-induced cell death in previous studies (Ahmad et al., 2005; Graham et al., 2012; Jeon et al., 2012). We also observed that glucose deprivation increased both mitochondrial ROS (Fig. 3D) and cytosolic ROS (Fig. 3E). However, the A769662 treatment only prevented the accumulation of mitochondrial ROS and not cytosolic ROS (Fig. 3D,E), which is contrary to the previously suggested causality between cytosolic ROS and cell death under glucose deprivation (Graham et al., 2012). Similarly, C75 treatment prevented the accumulation of mitochondrial ROS but not cytosolic ROS (Fig. 3F,G). Together, these data suggest that dysregulation of fatty acid metabolism caused by AMPK inactivation contributes to the increase in mitochondrial ROS under glucose deprivation.

As we observed accumulation of mitochondrial ROS and depletion of NADPH under glucose deprivation, we next investigated the effect of this collapse of the redox system on mitochondrial function. We assessed the mitochondrial oxygen consumption rate (OCR) and mitochondrial metabolic parameters by performing a Seahorse assay. Compared to U2OS cells cultured in 1 mM glucose, cells under glucose deprivation showed a decrease in basal mitochondrial OCR, mitochondrial ATP production, maximal respiration, spare respiratory capacity (SRC) and coupling efficiency, and an increase in proton leak, all of which could be a sign of mitochondrial damage under glucose deprivation (Fig. 3H,I). Interestingly, A769662 treatment under glucose deprivation rescued these mitochondrial metabolic parameters (Fig. 3H,I), implying that maintaining redox homeostasis through AMPK activation is important for retaining mitochondrial fitness. Given that SRC represents mitochondrial fitness, as it correlates with bioenergetics adaptability in responding to metabolic stress (Marchetti et al., 2020), SRC depletion under glucose deprivation suggests that mitochondria become dysfunctional under glucose withdrawal and that the mitochondrial ROS accumulation could be due to a defect in mitochondrial functions, rather than a promotion of mitochondrial oxidative phosphorylation, which is a major site of mitochondrial ROS production.

We next examined whether we could rescue glucose deprivation-induced cell death by directly restoring mitochondrial function. Cell-permeable methyl-pyruvate (Me-pyruvate) or dimethyl α-ketoglutarate (DMKG) was added in U2OS cells to directly replenish the mitochondrial tricarboxylic acid (TCA) cycle. These are the entry molecules of the TCA cycle, which can produce NADPH through two different reactions catalyzed by mitochondrial malic enzyme (ME) and isocitrate dehydrogenase (IDH) (Ju et al., 2020). Me-pyruvate or DMKG treatment partially rescued NADPH depletion (Fig. S4A) and prevented accumulation of mitochondrial ROS but not cytosolic ROS under glucose deprivation (Fig. S4B,C). Restoring this redox homeostasis resulted in partial rescue of cell death (Fig. S4D,E) and activation of AMPK signaling as indicated by increased phosphorylation of AMPK at Thr172 and ACC, although mTORC1 signaling, as indicated by the band pattern of 4EBP1, was not affected (Fig. S4F). These data suggest that mitochondrial dysfunction might play a role in glucose deprivation-induced cell death.

Our data showed that glucose deprivation caused NADPH depletion, mitochondrial ROS accumulation and cell death, all of which were rescued at least partly by treatment with the AMPK activator A769662. Therefore, our next question was how glucose deprivation inhibits AMPK activity in sensitive cell lines. Given that redox status correlates well with cell survival under glucose withdrawal (Fig. 3), we hypothesized that glucose removal-induced oxidative stress dysregulates AMPK signaling, leading to cell death. Treatment with the anti-oxidants N-acetyl-cysteine (NAC) or reduced glutathione (GSH) prevented glucose deprivation-induced cell death (Fig. 4A,B), mitochondrial and cytosolic ROS accumulation (Fig. 4C; Fig. S5A) and NADPH depletion (Fig. 4D) in U2OS cells. NAC or GSH treatment enabled U2OS cells to activate AMPK signaling and inhibit mTORC1 signaling under glucose deprivation, as indicated by increased Thr172 phosphorylation of AMPK and phosphorylation of ACC, and the band pattern shift of total and phosphorylation of 4EBP1 (Fig. 4E). A similar trend was observed in U251MG cells (Fig. S5B–D). To determine whether oxidative stress upon glucose deprivation contributes to AMPK inhibition, we investigated whether NAC treatment rescues AMPK activity under glucose deprivation by performing an ELISA assay. The treatment of NAC completely rescued the AMPK activity that was inactivated by glucose deprivation (Fig. 4F), suggesting that oxidative stress induced by glucose withdrawal perturbs AMPK activity, leading to a failure in metabolic adaptation.

Although ROS-induced oxidative stress was thought to cause nonspecific damage to cellular components, such as lipids, DNA and proteins, recent evidence suggests that ROS can modulate protein function by oxidizing specific cysteine residues on target proteins (Dickinson and Chang, 2011; Finkel, 2011). Therefore, we hypothesized that rapid NADPH depletion under glucose deprivation causes oxidative modification of AMPK, which inhibits AMPK activity (Shao et al., 2014). To examine whether AMPK is directly oxidized under glucose deprivation, we measured the electrophoretic mobility shift of AMPK in the non-reducing SDS-PAGE conditions, which does not contain reducing agents, such as dithiothreitol (DTT) and β-mercaptoethanol (β-ME). Glucose deprivation in U2OS cells induced a mobility shift of AMPK only in the non-reducing SDS-PAGE conditions (Fig. 4G). This mobility shift was prevented by NAC or A769662 treatment. Different levels of mobility shift of AMPK were observed under glucose deprivation. This might be due to the different extent of oxidation of protein. Similar results were obtained in U251MG cells (Fig. S5E). In addition, in the timecourse analysis of AMPK oxidation, we observed an inverse correlation between decreasing Thr172 phosphorylation with increasing oxidation of AMPK in a time-dependent manner (Fig. 4H). Furthermore, the timing of AMPK oxidation (Fig. 4H) was also correlated with the decrease in AMPK activity upon glucose removal (Fig. 1E). These data suggest that oxidation could negatively regulate the function of AMPK.

To confirm whether the mobility shift of AMPK under glucose deprivation in the non-reducing SDS-PAGE conditions is due to disulfide bond formation, we treated the cell lysates with DTT, a reducing agent. DTT treatment removed the mobility shift of AMPK in the non-reducing SDS-PAGE conditions (Fig. 4I). This phenomenon was not observed in reducing SDS-PAGE conditions, suggesting that disulfide bond formation contributes to the mobility shift of AMPK under glucose deprivation. We investigated whether disulfide bond formation of AMPK inhibits its kinase activity under glucose withdrawal. DTT treatment restored AMPK activity under glucose deprivation (Fig. 4J). These data suggest that glucose deprivation-induced AMPK oxidation inhibits AMPK activity.

Next, we asked whether glucose deprivation alone is sufficient to induce cell death or whether other nutrients, such as amino acids, are involved. We cultured U2OS cells in DMEM or amino acid-free DMEM (DMEM-AA) with and without 1 mM glucose and evaluated PI-positive cell death. Glucose deprivation in DMEM-AA did not induce cell death. However, when essential amino acids (EAA), but not non-essential amino acids (NEAA), were supplemented into DMEM-AA, glucose deprivation was able to induce cell death. To further examine which essential amino acid was required for glucose deprivation-induced cell death, we supplemented DMEM-AA with individual EAAs separately, measured cell death and found that cystine was required (Fig. 5A).

Cystine is uptaken through xCT, also known as SLC7A11, the light chain of the cystine/glutamate antiporter system x_c⁻. xCT exchanges intracellular glutamate for extracellular cystine, which is rapidly converted into cysteine using NADPH and serves as a precursor for GSH generation (Lewerenz et al., 2013). Because GSH is a powerful antioxidant, cancer cells often upregulate xCT to maintain high GSH levels, which contributes to chemoresistance, tumor invasion and poor survival (Huang et al., 2005; Polewski et al., 2016). To determine whether cystine uptake through the xCT is required for glucose deprivation-induced cell death, we treated U2OS cells with sulfasalazine (SAS) or extracellular glutamate to inhibit xCT activity (Dixon et al., 2012, 2014). Both xCT inhibitors prevented glucose deprivation-induced cell death (Fig. 5B,C). A similar result was obtained in the U251MG cells (Fig. S6A,B), indicating that the activity of xCT is required for glucose withdrawal-induced cell death.

We proceeded to examine whether cystine uptake through xCT is responsible for the glucose deprivation-induced redox system collapse. Inhibition of xCT with SAS prevented mitochondrial ROS accumulation (Fig. 5D) and partially rescued NADPH depletion in U2OS cells upon glucose deprivation (Fig. 5E). However, SAS treatment did not prevent cytosolic ROS accumulation (Fig. S6C), consistent with our data indicating that cytosolic ROS was not correlated with glucose deprivation-induced cell death (Fig. 3E,G; Fig. S4C). When cultured in DMEM-AA, glucose deprivation alone was not sufficient to induce mitochondrial ROS accumulation and NADPH depletion (Fig. 5F,G). Cystine supplementation in the medium was required for these phenotypes which were prevented by SAS treatment (Fig. 5F,G). However, unlike mitochondrial ROS, cytosolic ROS accumulation was not induced by glucose deprivation in DMEM-AA even with cystine supplementation (Fig. S6D).

Given that cysteine serves as a precursor for GSH generation and that NADPH is used for recycling oxidized glutathione (GSSG) to GSH, we examined whether GSH metabolism is also affected by glucose availability and xCT activity. Glucose deprivation in DMEM decreased the GSH/GSSG ratio which was partially restored by xCT inhibition (Fig. S6E). When cultured in DMEM-AA, glucose deprivation alone was not sufficient to oxidize GSH completely, and cystine supplementation was required to induce complete oxidation of GSH, which was partially restored by xCT inhibition (Fig. S6F).

Next, we investigated whether xCT-mediated cystine uptake under glucose deprivation is responsible for the dysregulation of AMPK signaling. When U2OS cells were cultured in DMEM without glucose, xCT inhibition with SAS resulted in high AMPK and low mTORC1 activities, as represented by increased Thr172 phosphorylation of AMPK, phosphorylation of ACC, and the band pattern shift of total and phosphorylated 4EBP1 to a lower molecular mass (Fig. 5H). These results suggest that dysregulation of AMPK signaling upon glucose deprivation is rescued by xCT inhibition. Glucose deprivation alone in DMEM-AA activated AMPK signaling, as indicated by increased Thr172 phosphorylation of AMPK and phosphorylation of ACC. Also, it inhibited mTORC1 signaling, as indicated by the band pattern shift of 4EBP1. However, when cystine was supplemented in DMEM-AA under glucose deprivation, AMPK reduced its activity, as indicated by phosphorylation of AMPK and ACC and the band pattern shift of 4EBP1. These changes were attenuated by xCT inhibition with SAS. Similar results were obtained in U251MG cells (Fig. S6G). We also examined AMPK oxidation in the non-reducing SDS-PAGE conditions. AMPK oxidation induced by glucose deprivation in DMEM was prevented by xCT inhibition with SAS (Fig. 5I). Glucose deprivation alone in DMEM-AA did not oxidize AMPK, but supplementation with cystine induced AMPK oxidation. This AMPK oxidation was prevented by xCT inhibition with SAS (Fig. 5I). Consistent with the AMPK oxidation status, we also observed that glucose deprivation-induced AMPK inactivation was completely rescued by xCT inhibition with SAS by using an ELISA assay (Fig. 5J).

Collectively, our data suggest that cystine uptake through xCT contributes to glucose deprivation-induced redox system collapse and AMPK dysregulation.

Given that the activity of xCT contributes to redox system collapse and AMPK dysregulation upon glucose deprivation, we sought to investigate whether there is a correlation between xCT expression and sensitivity to glucose deprivation between various cancer cell lines. Sensitive cell lines (U2OS and U251MG) expressed high levels of xCT compared to intermediate sensitive and resistant cell lines (SW480, MCF7, A375, H1299 and HEK293T), which indicates that xCT could be a potential biomarker to predict sensitivity to glucose deprivation (Fig. 6A). The expression levels of glucose transporter 1 (GLUT1) and glycolytic enzymes such as hexokinase 1 and 2 and glucose-6-dehydrogenase did not correlate to sensitivity to glucose deprivation (Fig. 6A). Next, to determine whether xCT expression contributes to NADPH depletion under glucose deprivation, we measured intracellular NADPH levels in multiple cell lines. Sensitive cell lines (U2OS and U251MG) showed a marked increase of the NADP+/NADPH ratio compared to intermediate sensitive and resistant cell lines (SW480, A375 and H1299) (Fig. 6B). Sensitive cell lines showed rapid NADPH depletion under glucose deprivation, whereas intermediate sensitive and resistant cell lines managed to maintain NADPH levels under glucose deprivation although partial NADPH decrease was observed in some cell lines (Fig. 6C). This indicates a strong correlation between xCT expression, NADPH depletion and sensitivity to glucose deprivation in cancer cells. Furthermore, sensitive cell lines (U2OS and U251MG) showed an increase (4–5 fold) of NADPt (the sum of NADPH and NADP+), which was due to the accumulation of NADP+, whereas intermediate sensitive and resistant cell lines maintained relatively similar levels of NADPt upon glucose deprivation (Fig. 6D,E).

Taken together, these findings indicate that rapid NADPH depletion upon glucose deprivation is a distinguished feature of sensitive cancer cell lines expressing high levels of xCT.

These data prompted us to investigate whether xCT expression levels determine the glucose dependency of cancer cells. We knocked down xCT with two different siRNA in U2OS cells and monitored sensitivity to glucose deprivation. Knocking down xCT prevented glucose deprivation-induced cell death (Fig. 7A,B) and mitochondrial ROS accumulation (Fig. 7C) and partially rescued NADPH depletion (Fig. 7D). Besides, knocking down xCT restored AMPK activation, reduced mTORC1 activity (Fig. 7E) and prevented glucose withdrawal-induced AMPK oxidation under glucose deprivation (Fig. 7F).

Next, we transiently overexpressed FLAG-tagged xCT in H1299 to investigate whether high expression of xCT confers high glucose dependency in the resistant cell line. Overexpression of xCT in H1299 cells potentiated glucose deprivation-induced cell death (Fig. 7G,H), mitochondrial ROS accumulation (Fig. 7I) and NADPH depletion (Fig. 7J), which were prevented by treatment with the xCT inhibitor SAS (Fig. 7G,H,I,J). In non-reducing SDS-PAGE conditions, glucose deprivation-induced AMPK oxidation was only observed in xCT overexpressing H1299 cells, and not in empty vector-transfected H1299 cells (Fig. 7K). This AMPK oxidation was prevented by xCT inhibition by SAS treatment (Fig. 7K).

Taken together, our results suggest that xCT expression levels determine sensitivity to glucose deprivation, at least in the cancers we studied.

Our results revealed that cystine uptake renders cancer cells more susceptible to glucose deprivation by accelerating NADPH consumption. These data prompted us to investigate whether glucose dependency could be therapeutically targeted. Because it is physiologically impossible to completely deplete glucose for cancer cells in vivo, we employed STF-31, an inhibitor of GLUT1 (Chan et al., 2011). We cultured a sensitive cell line (U2OS) and a resistant cell line (H1299) in 2 mM glucose, which is within the physiological concentration range in the tumor microenvironment (Hirayama et al., 2009) and treated each with STF-31 for 24 h and evaluated cell death (Fig. 8A,B). We found that STF-31 treatment induced moderate levels of cell death in U2OS cells. However, when applied in combination with additional cystine, GLUT1 inhibitor-induced cell death was significantly sensitized. Furthermore, inhibition of xCT with SAS prevented STF-31-induced cell death completely, which indicates that cell death induced by glucose uptake inhibition is dependent on cystine uptake through xCT. By contrast, STF-31 treatment with or without cystine did not markedly induce cell death in H1299 cells. These data suggest that glucose dependency of cancer cells is targetable with GLUT1 inhibitor and cystine, and that high uptake of cystine through xCT could be a key predictor of sensitivity to a GLUT1-targeting therapy.

The outcome of clinical trials targeting cancer cell metabolism has not been successful due to our limited understanding of glucose deprivation-induced cell death. Here, we extended our understanding of glucose deprivation-induced cell death with multiple pieces of evidence: (1) glucose deprivation-sensitive cancer cell lines failed to activate AMPK due to redox-dependent inhibitory oxidation, leading to failure to adapt to metabolic stress; (2) xCT expression is a potential biomarker which determines and predicts sensitivity to glucose deprivation; and (3) NADPH depletion upon glucose deprivation is a key metabolic determinant of glucose deprivation-induced cell death (Fig. 8C).

AMPK is a major metabolic checkpoint that senses energy or nutrient depletion to maintain metabolic homeostasis. Therefore, tight regulation of the AMPK pathway is essential for cancer cells to overcome a limited nutrient environment. For instance, AMPK protects leukemia-initiating cells from metabolic stress (Saito et al., 2015), whereas LKB1-null cancers are highly vulnerable to energetic stress due to lack of an activating kinase for AMPK (Shackelford et al., 2013). In the present study, despite having the intact LKB1-AMPK axis, the cell lines most sensitive to glucose deprivation (U2OS and U251MG cells) displayed failure of AMPK activation under metabolic stress.

The inhibition of mTORC1-dependent global mRNA translation is an example of AMPK-mediated resistance to nutrient starvation (Inoki et al., 2003). In the present study, although highly sustained mTORC1 activity was observed under glucose deprivation in sensitive cancer cell lines due to defective AMPK activation, the mTORC1 inhibitor rapamycin did not rescue glucose deprivation-induced cell death, suggesting high ATP expenditure or mRNA translation due to the high mTORC1 activity is not the main cause of glucose deprivation-induced cell death, at least in the cell lines we studied. Instead, we observed that C75, a FAO activator and FAS inhibitor, rescued cell death under glucose deprivation by replenishing NADPH levels. In the absence of glucose supply, the primary NADPH source is the mitochondrial TCA cycle, supported by FAO, which produces citrate and malate, the substrates for the NADPH-producing enzymes IDH and ME, respectively. On the other hand, FAS consumes two NADPH molecules. Thus, it is reasonable that C75, a FAO activator and FAS inhibitor, rescued cell death by increasing NADPH levels. AMPK phosphorylates ACC to inhibit its functions, and ACC is a key enzyme that regulates fatty acid metabolism by inhibiting FAO and promoting FAS. ACC produces malonyl-CoA, a precursor for FAS. Malonyl-CoA is also a negative regulator of carnitine palmitoyl transferase 1, which catalyzes the transport of long-chain fatty acids into mitochondria for promoting FAO. In fact, glucose deprivation has been shown to activate AMPK for maintaining NADPH levels by inhibiting ACC to increase FAO and decrease FAS (Jeon et al., 2012), which is consistent with our data showing that xCT overexpressing cancers failed to activate AMPK and phosphorylate ACC to maintain NADPH levels under glucose deprivation. Although we cannot exclude the possibility that AMPK activation facilitates the TCA cycle and provides NADPH via other means (Blagih et al., 2015; Cai et al., 2020), AMPK activation improves cancer cell viability by regulating redox homeostasis at least in part via FAO- and FAS-mediated NADPH maintenance.

The intracellular oxidizing condition can induce covalent disulfide bond formation within and between proteins via cysteine residues, which alters biological functions, subcellular localization and protein–protein interactions (Dickinson and Chang, 2011; Finkel, 2011). Recent evidence suggests that only specific cysteine residues are susceptible to oxidative stress-induced modification, indicating protein oxidation is a tightly regulated post-translational modification (Xiao et al., 2020). Here, we found that glucose deprivation in the sensitive cancer cells dampened AMPK-mediated adaptive response by a redox-dependent mechanism. AMPK has been reported to be susceptible to oxidative stress (Byrne et al., 2020; Shao et al., 2014). Direct AMPK oxidation by ROS occurs at residues Cys130 and Cys174, which can then form intermolecular disulfide bonds, leading to AMPK aggregation, which inhibits AMPK function (Shao et al., 2014). In our study, glucose deprivation induced AMPK electrophoretic mobility shift in non-reducing SDS-PAGE, which was rescued by the treatment of anti-oxidant or reducing agent. Furthermore, AMPK electrophoretic mobility shift was inversely correlated with AMPK activity measured by ELISA-based assay. These phenomena were only observed in highly sensitive cancer cell lines to glucose starvation. These data suggest that AMPK oxidation via the formation of intra- and intermolecular disulfide bonds inactivates AMPK upon glucose deprivation. Furthermore, the extent of AMPK oxidation correlated to redox collapse by NADPH depletion, rather than elevation of cytosolic ROS. This was supported by lines of evidence: SAS, A769662, C75, Me-pyruvate and DMKG treatments which ameliorated mitochondrial ROS accumulation and NADPH depletion but not cytosolic ROS, prevented glucose deprivation-induced AMPK oxidation and cell death, at least, in the experimental context in this study.

ROS might affect AMPK activity by oxidation differently depending on levels of ROS and cell types. It has been shown that temporal ROS elevation can activate AMPK (Rabinovitch et al., 2017; Zmijewski et al., 2010). However, severe oxidative stress induced by NADPH depletion by glucose deprivation in cells with high levels of xCT such as U2OS and U251MG cells can collapse thiol anti-oxidant systems, such as thioredoxin and GSH, which require NADPH for their recycling. Therefore, this condition can disrupt the protein dithiol/disulfide balance, thereby inducing AMPK disulfide bond formation (Shao et al., 2014). Why the increase of mitochondrial ROS contributes to AMPK dysregulation and cell death under glucose deprivation remains to be elucidated. Investigating the subcellular localization of NADPH and AMPK under glucose deprivation would help understanding on this issue.

Recently, multiple studies have demonstrated that xCT activity is required for glucose deprivation-induced cell death. The various underlying mechanisms have been proposed, such as mitochondrial dysfunction after glutamate export (Koppula et al., 2017; Shin et al., 2017), NADPH depletion and ROS accumulation upon cystine uptake (Goji et al., 2017; Joly et al., 2020) and disulfide stress (Liu et al., 2020). Consistent with these reports, we found that cystine was required for glucose deprivation-induced NADPH depletion, mitochondrial dysfunction and cell death. In particular, our data suggest that NADPH consumption is a key metabolic determinant of sensitivity to glucose deprivation, and that the expression levels of xCT determine NADPH consumption and cell death under glucose deprivation. Cystine imported through xCT is quickly reduced to cysteine, which consumes NADPH in this conversion process (Pader et al., 2014). Because NADPH is constantly supplied through the PPP, which branches from glycolysis, this NADPH consumption does not cause oxidative stress in the presence of glucose. On the other hand, under glucose deprivation in cells with high xCT expression, high cystine uptake might accelerate NADPH consumption by reducing cystine to cysteine and lead to NADPH depletion because NADPH from the PPP is limited. At the same time, glutamate export through xCT can lead to a reduction of α-ketoglutarate, a metabolite of glutamate entering the TCA cycle, where NADPH is produced. In our study, both cystine uptake and glutamate export seem to contribute to the rapid depletion of NADPH under glucose deprivation because replenishing the TCA cycle with DMKG, a cell-permeable α-ketoglutarate, was only able to partially rescue the NADPH depletion and cell death.

We observed that sensitive cancer cell lines (U2OS and U251MG) expressed high levels of xCT compared to resistant cancer cell lines. High expression of xCT can cause glucose deprivation-induced mitochondrial dysfunction by promoting glutamate export, resulting in a shortage of TCA cycle substrate. High expression of xCT also contributes to NADPH depletion, leading to AMPK dysregulation. Inactivated AMPK fails to mediate the switch from glycolysis to FAO, resulting in a further shortage of NADPH and TCA cycle substrate, redox collapse and cell death. These data provide insight into a previously underappreciated novel crosstalk between metabolic regulation and signaling transduction, which is essential for metabolic adaptation.

In conclusion, our data revealed AMPK inactivation in sensitive cancer cell lines expressing high levels of xCT and a novel link between xCT expression and AMPK dysregulation by its oxidation in response to glucose deprivation. Given that cancer cells expressing high levels of xCT are resistant to conventional radio or chemotherapy due to a high rate of generation of GSH (Huang et al., 2005; Polewski et al., 2016), a glucose uptake inhibition strategy together with elevating tissue cystine concentration could be therapeutically more effective for these types of cancers, and xCT expression could be considered as a biomarker for predicting sensitivity to the therapy.

U2OS, U251MG, SW480, MCF7, A375, H1299 and HEK293T cells were purchased from the American Type Culture Collection (ATCC). WI-38 and IMR-90 cells were purchased from the Coriell Institute. Cells were authenticated by the ATCC or Coriell Institute and tested negative for mycoplasma. All cell lines were cultured in high-glucose DMEM with 25 mM glucose and 4 mM L-glutamine and without sodium pyruvate (#11965, Gibco, Life Technologies) supplemented with 10% fetal bovine serum (FBS; HyClone, GE Healthcare Life Science), penicillin (100 units/ml) and streptomycin (100 µM/ml; Gibco, Life Technologies) in a 5% CO₂-humidified atmosphere at 37°C. For glucose deprivation, cells were washed with phosphate-buffered saline (PBS) three times and cultured in glucose-free DMEM (#10966, Gibco, Life Technologies) with 10% dialyzed FBS. DMEM-AA was generated following the recipe for DMEM (CaCl₂ 200 mg/l, Fe(NO₃)₃•9H₂O 0.1 mg/l, KCl 400 mg/l, MgSO₄ 96.67 mg/l, NaCl 6400 mg/l, NaHCO₃ 3700 mg, NaH₂PO₄•H₂O 125 mg/l) and 4 mM L- glutamine was added.

Glucose, rapamycin, PI, EAA, NEAA, glutamate, oligomycin, FCCP, rotenone, antimycin A, methyl-pyruvate, DMKG, NAC and reduced glutathione (GSH) were purchased from Sigma-Aldrich. SAS and A769662 were purchased from Cayman Chemical. Necrosulfonamide and STF-31 were purchased from Merck Millipore. Deferoxamine, C75, and etomoxir were purchased from MedChemExpress. Z-VAD-FMK was purchased from Santa Cruz Biotechnology. Necrostatin-1 was purchased from Abcam. Arginine, cystine, histidine, isoleucine, leucine, lysine, methionine, phenylalanine, threonine, tryptophan, tyrosine and valine were kindly provided by Dr Jean-Paul Kovalik (Duke-NUS Medical School, Singapore).

The commercial dialyzed serum (HyClone, GE Healthcare Life Science) was further dialyzed using a Slide-A-Lyzer G2 dialysis cassette (MWCO 10,000; Thermo Fisher Scientific). Dialysis was performed in a cold PBS buffer containing 1 mM PMSF for 24 h in a cold room.

siRNAs targeting xCT (SLC7A11) were obtained from Qiagen. Control siRNA (ON-TARGET plus nontargeting pool) was obtained from Dharmacon. The targeted siRNA sequences were: xCT #1, 5′-AACCACCTGTTTCACTAATAA-3′; xCT #2, 5′-TGGGTGGAACTCCTCATAATA-3′.

siRNAs targeting AMPKα1 and AMPKα2 were mixed for use. siRNAs targeting AMPKα1 were obtained from Thermo Fisher Scientific. The targeted siRNA sequences were: AMPKα1 #1, 5′-CCCATCCTGAAAGAGTACCATTCTT-3′; AMPKα1 #2, 5′-ACCATGATTGATGATGAAGCCTTAA-3′; AMPKα1 #3, 5′-CCCTCAATATTTAAATCCTTCTGTG-3′. siRNAs targeting AMPKα2 were obtained from QIAGEN. The targeted siRNA sequences were: AMPKα2 #1, 5′-CCGATTTCGGATTATCTAATA-3′; AMPKα2 #2, 5′-ACGACTAAGCCCAAATCTTTA-3′; and AMPKα2 #3, 5′-CCGAAGTCAGAGCAAACCGTA-3′.

Full-length xCT was amplified from U251MG cells and cloned into vector pcDNA3.1⁽⁺⁾ (Thermo Fisher Scientific) with C-terminal 3xFLAG tagging. The cloned construct was confirmed by direct DNA sequencing. Cells were transiently transfected with siRNA or plasmid using ScreenfectA according to the manufacturer's instructions (FUJIFILM Wako Pure Chemical Corporation). siRNAs were used at 20 nM (xCT) or 40 nM (AMPKα1 and AMPKα2 together) for transfection.

Cells were lysed with 2% SDS lysis buffer (50 mM Tris-HCl, pH 6.8, 10% glycerol and 2% SDS). Equal amounts of proteins (20–30 µg) were subjected to SDS-PAGE. After all proteins were transferred to a nitrocellulose membrane, immunoblots were probed with the indicated antibodies. Detection was performed by incubation of HRP-conjugated anti-mouse or anti-rabbit IgG (Jackson ImmunoResearch) secondary antibody followed by the reaction for chemiluminescence (SuperSignal, Thermo Fisher Scientific). Infrared fluorescence-conjugated anti-mouse or anti-rabbit IgG secondary antibody (Dy-light, Thermo Fisher Scientific) was used for infrared fluorescence detection (LI-COR Odyssey).

For AMPK redox analysis, cells were washed with ice-cold PBS once and lysed with 0.1% NP-40 lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM PMSF, 1× protease inhibitor cocktail without EDTA from Nacalai tesque, 1× phosphatase inhibitor cocktail from Roche) containing 20 mM N-ethylmaleimide to prevent further cysteine oxidation during protein extraction. Equal amounts of proteins (20 µg) were subjected to SDS-PAGE with loading buffer with or without DTT and β-ME. The dotted line in western blotting images is used to demarcate different cell types or treatment groups and does not indicate removal of lanes or consolidation of data from different blots.

The following antibodies were used; anti-AMPK (1:1000, Cell Signaling Technology, #2793), anti-phospho-AMPK Thr172 (1:1000, Cell Signaling, #2535), anti-ACC (1:1000, Cell Signaling, #3676), anti-phospho-ACC Ser79 (1:1000, Cell Signaling, #11818), anti-G6PD (1:1000, Cell Signaling, #8866), anti-LC3B (1:1000, Cell Signaling, #2775), anti-LKB1 (1:1000, Cell Signaling, #3047), anti-xCT/ SLC7A11 (1:1000, Cell Signaling, #12691), anti-phospho-4EBP1 Thr37/46 (1:1000, Cell Signaling, #9459), anti-4EBP1 (1:1000, Cell Signaling, #9644) anti-GLUT1 (1:1000, abcam, #ab115730), anti-HXK1 (1:1000, Santa Cruz, #sc-6517), anti-HXK2 (1:1000, Santa Cruz, #sc-6521), anti-actin (1:10000, Merck Millipore, clone C4, cat #MAB1501) anti-tubulin (1:3000, Abcam, clone no. DM1A+DM1B). The signal intensity of phosphorylated ACC and total ACC were analyzed by ImageQuant software (Molecular Dynamics). Uncropped western blotting images are provided in Fig. S7.

Cells were stained with PI to determine the percentage of cell death. Media containing floating cells were collected, combined with trypsinized cells and centrifuged (300 g for 3 min). The cell pellet was washed once with PBS. After centrifugation, cells were re-suspended and stained with PI (10 μg/ml) for 10 min at room temperature. Data were collected with a MACSQuant analyzer (Miltenyi Biotec). Quantification and analysis of the data were performed with Flowjo software.

Cells (2.5×10⁵/well) were seeded in six-well plates and were lysed with 0.1% NP-40 lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM PMSF, 1× protease inhibitor cocktail without EDTA from Nacalai tesque, and 1× phosphatase inhibitor cocktail from Roche). The relative activity of AMPK in cell lysates was measured using CyLex AMPK kinase assay kit (MBL International Corporation), according to the manufacturer's instructions. Signal intensities were examined by optical density (OD) measurements at 450 nm using an Infinite M200 plate reader (TECAN). Relative AMPK activity was defined as Compound C-sensitive protein kinase activity in cell lysate using this kit. Thus, AMPK activity was normalized to the protein amount, and relative AMPK activity was calculated by subtracting OD measurements at 450 nm for the inhibitor control from that of the test sample.

Cells (2×10⁴/well) were seeded in 96-well plates. Intracellular ATP levels were measured with a CellTiter-Glo luminescent cell viability assay (Promega) according to the manufacturer's guidelines.

Cells (2.5×10⁵/well) were seeded in six-well plates. Cells were deprived of glucose for 3 h. During the last 30 min of glucose deprivation, 2.5 µM DCFDA or 2.5 µM mitoSOX (Thermo Fisher Scientific) was added to the medium for staining for 30 min at 37°C. Medium containing floating cells were collected, combined with trypsinized cells, and centrifuged. The cell pellet was washed once with PBS. Data were collected with a MACSQuant analyzer (Miltenyi Biotec). Quantification and analysis of the data were done with Flowjo software.

Intracellular NADPH/NADP⁺ levels were measured using the NADP/NADPH Quantification Kit (ab65349, Abcam) according to the manufacturer's instructions. Briefly, 2.5×10⁵ cells were lysed with 350 μl of extraction buffer at indicated time points after glucose deprivation. For the reaction, 50 μl of the final sample was used. Signal intensities for NADPH were examined by OD measurements at 450 nm using an Infinite M200 plate reader (TECAN).

Intracellular GSH/GSSG was measured with the GSH/GSSG-Glo™ luminescent assay (Promega) according to the manufacturer's instructions. Briefly, 2×10⁴ cells in 96-well plates were lysed at the indicated time point in the indicated condition.

The oxygen consumption rate (OCR) was measured using a Seahorse Bioscience XF96 Extracellular Flux Analyzer (Seahorse Bioscience). 2×10⁴ cells were plated into Seahorse tissue culture 96-well plates. Cells were cultured in DMEM with or without 1 mM glucose for 2 h. And then cells were cultured in Seahorse assay medium containing 2 mM glutamine with or without 1 mM glucose and incubated in a CO₂-free incubator for 1 h before measurement. An XF Cell Mito Stress Test Kit was used to analyze mitochondrial metabolic parameters by measuring the OCR. Oligomycin (1 μM) was injected to determine the oligomycin-independent lack of the OCR. The mitochondrial uncoupler FCCP (1 μM) was injected to determine the maximum respiratory capacity. Rotenone (1 μM) and antimycin A (1 μM) were injected to block complex I and complex III of the electron transport chain.

Data are presented as mean±s.d. of at least three independent experiments. An unpaired two-tailed Student's t-test was performed for statistical analysis. All statistical analyses were conducted using GraphPad Prism 9.0.2 software.

We thank Drs David M. Virshup, Anne-Claude Gingras, Egon Ogris, Brijesh K. Singh, and Jennifer Alagu for the discussion of the data and the manuscript.