Mammalian TOR: A Homeostatic ATP Sensor

See allHide authors and affiliations

Science  02 Nov 2001:
Vol. 294, Issue 5544, pp. 1102-1105
DOI: 10.1126/science.1063518


The bacterial macrolide rapamycin is an efficacious anticancer agent against solid tumors. In a hypoxic environment, the increase in mass of solid tumors is dependent on the recruitment of mitogens and nutrients. When nutrient concentrations change, particularly those of essential amino acids, the mammalian Target of Rapamycin (mTOR) functions in regulatory pathways that control ribosome biogenesis and cell growth. In bacteria, ribosome biogenesis is independently regulated by amino acids and adenosine triphosphate (ATP). Here we demonstrate that the mTOR pathway is influenced by the intracellular concentration of ATP, independent of the abundance of amino acids, and that mTOR itself is an ATP sensor.

The survival of organisms is dependent on their ability to maintain cellular homeostasis. Environmental cues are deciphered by cellular regulatory elements, which adjust an organism's metabolic state to reflect external conditions. The phosphatidylinositide kinase– related family of protein kinases contains a number of critical effectors that sense environmental factors that control the ability of an organism to survive. One member of the family is mTOR, which resides at the interface between nutrient sensing and the regulation of major metabolic responses (1–3). Depending on mitogen and amino acid availability, mTOR positively regulates translation and ribosome biogenesis while negatively controlling autophagy (1), suggesting that mTOR sets protein synthetic rates as a function of the availability of translational precursors (4,5). In response to mitogens and amino acids, mTOR phosphorylates and controls the activities of two key translational regulators, S6 kinase 1 (S6K1) and initiation factor 4E binding protein (4E-BP1) (2). However, changes in mTOR activity in vitro, after either mitogen or amino acid treatment, have been small and controversial (2, 3). The importance of understanding the molecular mechanisms that control mTOR function is underscored by recent phase 1 clinical trials showing that rapamycin is efficacious in the treatment of solid tumors in patients with metastatic renal cell carcinoma and non–small cell lung, prostate, and breast cancer (6).

Protein synthesis is the major energy-consuming process in the cell (7). In bacteria, protein synthesis and cell growth are linked by ribosome biogenesis, which is independently controlled by amino acid and adenosine triphosphate (ATP) availability (8). Because mTOR is sensitive to amino acids and regulates ribosome biogenesis (1–3), we tested its activity for sensitivity to alterations in intracellular ATP concentrations. We used insulin-induced S6K1 activation and 4E-BP1 phosphorylation as reporters for mTOR function in the presence of glycolytic or mitochondrial inhibitors (9). The glycolytic inhibitor 2-deoxyglucose (2-DG) was more effective in inhibiting S6K1 Thr389 phosphorylation and S6K1 activation than was the mitochondrial inhibitor rotenone (Fig. 1A). Similar results were obtained when an mTOR-mediated phosphorylation site in 4E-BP1 (Ser65) was measured (Fig. 1B) and when iodoacetic acid and dinitrophenol were used as glycolytic and mitochondrial inhibitors, respectively (10). The effectiveness of each agent in inhibiting S6K1 activation paralleled its ability to lower ATP concentrations, as measured in a luciferase reporter assay (Fig. 1C) (11). The glycolytic inhibitors may be more effective in reducing ATP concentrations in transformed HEK293 cells because those cells predominately use anaerobic respiration to produce ATP (12, 13). The modest effect of rotenone in reducing ATP concentrations, as compared with its effects on S6K1 and 4E-BP1 phosphorylation, suggested that the inhibitors were not generally toxic. To verify this, we tested the effect of 2-DG on protein kinase B (PKB) and mitogen-activated protein kinase (MAPK). Like rapamycin, 2-DG did not inhibit insulin-induced activation of PKB, which was sensitive to the phosphatidylinositide-3-OH kinase (PI3K) inhibitor wortmannin, as judged by Ser473phosphorylation and in vitro phosphorylation of histone 2B (H2B) (Fig. 1D). Also, 2-DG had no effect on TPA-induced MAPK activation, with myelin basic protein (MBP) as substrate (Fig. 1E). Hence, the 2-DG–induced reduction in ATP concentrations appears to selectively influence signaling to S6K1 and 4E-BP1.

Figure 1

The effects of metabolic inhibitors. (A) Serum-starved HEK293 cells were extracted directly or stimulated with 200 nM insulin in the presence of 100 mM 2-DG or 20 μM rotenone for 30 min. S6K1 levels and Thr389phosphorylation were measured by Western blot analysis (9). S6K1 activity was assayed as described previously (22). (B) The expression and phosphorylation of transiently transfected HA-4E-BP1 was measured as in (A), with a polyclonal antibody to HA and phosphospecific antibody against Ser65, respectively (9). (C) ATP levels were measured from mock-transfected HEK293 cells treated as in (A), with a luciferase-based assay (11), with the results expressed as a percentage of the insulin-stimulated control. (D) Transiently transfected, HA-tagged PKB activation was measured in vitro with H2B as substrate (22) after immunoprecipitation from serum-starved cells extracted directly or after insulin stimulation with or without the addition of 100 mM 2-DG, 20 nM rapamycin, or 100 nM wortmannin (9). HA-PKB expression and Ser473 phosphorylation were measured as in (B) and (9), respectively. (E) HA-MAPK kinase activity toward myelin basic protein was measured from serum-starved cells extracted directly or after 100 nM 12-o-tetradecanoylphorbol-13-acetate (TPA) stimulation in the presence or absence of 20 nM rapamycin or 100 mM 2-DG for 30 min (22). All results and the SE in (C) are representative of at least three independent experiments.

To assess whether the inhibitory effects of 2-DG on S6K1 activation were mediated by mTOR, we used a rapamycin-resistant allele of S6K1. Rapamycin resistance was conferred by fusing glutathione S-transferase (GST) to the NH2-terminus of S6K1 and truncating the COOH-terminus, creating a construct termed GST-ΔC-S6K1 (14). Both S6K1, having a COOH-terminal GST tag (S6K1-GST), and GST-ΔC-S6K1 were phosphorylated and activated by insulin in a wortmannin-sensitive manner. However, only GST-ΔC-S6K1 was resistant to inhibition by rapamycin (Fig. 2A). 2-DG reduced insulin-induced activation of S6K1-GST (Fig. 2B) in a dose-dependent manner, paralleling its effect on intracellular ATP concentrations (Fig. 2C). In contrast, insulin-induced activation of GST-ΔC-S6K1 was unaffected by 2-DG treatment (Fig. 2B). Thus, 2-DG appears to selectively inhibit signaling to mTOR effectors, supporting a model whereby mTOR is controlled by intracellular ATP concentrations.

Figure 2

Specific inhibition of mTOR-signaling by 2-DG. (A) S6K1-GST and GST-ΔC-S6K1 from transiently transfected, serum-starved HEK293 cells were extracted directly or after insulin stimulation with or without the addition of 20 nM rapamycin or 100 nM wortmannin (9). Kinase activities, expression levels, and T389 phosphorylation were as in Fig. 1A and (22). (B), Expression and activity of S6K1-GST or GST-ΔC-S6K1 were measured after insulin stimulation alone or with increasing amounts of 2-DG as in (A). (C) Mock-transfected HEK293 cells were treated as in (B) and extracted for ATP analysis as in Fig. 1C. Results and SE in (C) are typical of at least two independent experiments.

Because mTOR signaling is dependent on concentrations of aminoacylated tRNAs (5), the effects of ATP on S6K1 and 4E-BP1 may be indirect, occurring through inhibition of tRNA aminoacylation. To examine this possibility, we analyzed total cellular tRNA on acid-urea polyacrylamide gels, which resolve aminoacylated from nonacylated tRNA (15). Neither insulin stimulation nor 2-DG treatment had an effect on total amounts of aminoacylated tRNA (Fig. 3A). Unexpectedly, amino acid deprivation also had no effect (Fig. 3A), even though such treatment was sufficient to completely block phosphorylation of S6K1 and 4E-BP1 (Fig. 3B). To further test this finding, we examined selected tRNAs by Northern blot analysis. As with total tRNA, none of the three treatments had an effect on the aminoacylation status of leucyl, histidyl, or threonyl tRNA (Fig. 3A) (10), indicating that amino acid pools, rather than amounts of aminoacylated tRNA, were important for mTOR signaling (16). Indeed, amino acid deprivation resulted in a decrease in the amounts of essential amino acids, particularly the branched-chain amino acids, whereas 2-DG had little effect (Fig. 3C) (10). Furthermore, amino acid deprivation had no effect on concentrations of ATP (Fig. 3D). Thus, regulation of mTOR by ATP is independent of amino acids pools.

Figure 3

Effect of 2-DG on tRNA aminoacylation and amino acid levels. (A) Total RNA was extracted from serum-starved HEK293 cells (15) directly or after insulin stimulation, with or without 2-DG or amino acid starvation (–AA). After resolution on an acid-urea gel (15), tRNA was visualized with ethidium bromide staining (top) or by Northern blot analysis with probes specific for tRNALeu (middle) and tRNAHis (bottom). DA in vitro deacylated tRNA. (B) Expression levels and phosphorylation states of endogenous S6K1 (left) and 4E-BP1 (right) were measured from serum-starved cells, stimulated with insulin in the presence or absence of amino acids in the media as in Fig. 1A and (9). (C) Levels of individual amino acids were measured in extracts prepared from insulin-stimulated HEK293 cells in the presence or absence of amino acids or with 100 mM 2-DG treatment (24) and expressed as a percentage of the insulin-stimulated control in the presence of amino acids. (D) Cells treated as in (B) were extracted for ATP analysis (11), and the results were expressed as a percentage of the insulin-stimulated control in the presence of amino acids. Results and SE derived in (C) and (D) are typical of at least two independent experiments.

Reducing ATP concentrations could lead to a stable change in mTOR activity, through a posttranslational modification, or ATP could directly affect mTOR activity. To test the first possibility, we transiently expressed hemaglutinin (HA) epitope–tagged mTOR in HEK293 and measured its activity after treatment of cells with 2-DG (17). Although such treatment blocked insulin-induced activation of S6K1 (Fig. 1A), it had no effect on the kinase activity of mTOR in vitro toward Thr389 of S6K1 (Fig. 4A). Indeed, neither amino acid deprivation, insulin stimulation, nor transient expression of an activated allele of PI3K affected mTOR kinase activity in vitro (Fig. 4, A and B), despite their ability in the intact cell to affect Thr389 phosphorylation and S6K1 activation (Figs. 3B and 4C). Next, we measured mTOR activity in vitro at ATP concentrations that approached physiological levels of 1 to 5 mM in mammalian cells (18). Specific activity of mTOR increased up to ∼1 mM ATP, saturating at around 2 or 3 mM ATP, whereas the catalytically inactive mTOR mutant did not phosphorylate Thr389 (Fig. 4D). On the basis of these values, we calculated a K m (Michaelis constant) for ATP of slightly greater than 1 mM (17). Most protein kinases analyzed to date show an apparent K m for ATP of 10 to 20 μM (19), one-hundredth to one-fiftieth of that observed for mTOR. Because mTOR also phosphorylates several sites in 4E-BP1, including Ser65 (2), we also assayed the ATP requirement of mTOR for Ser65 of 4E-BP1. Using the same assay conditions described for Thr389phosphorylation in S6K1, we obtained almost identical results for Ser65 phosphorylation (Fig. 4D). These findings sustain the role of mTOR as an ATP effector and suggest that it is a direct sensor of ATP in the cell.

Figure 4

mTOR activity requires high ATP concentrations. (A) HA-tagged, wild-type mTOR (HA-mTORwt) was obtained from serum-starved HEK293 cells that were insulin-stimulated with or without 100 mM 2-DG treatment (left) or amino acid starvation (right). HA-mTORwt was assayed for Thr389 kinase activity, with a kinase inactive mutant of S6K1 (D3E,K100Q-GST) as substrate (17). (B) HA-mTORwt was transiently expressed alone or with a constitutively membrane-targeted PI3K(CD2-PI3K) and then extracted directly or stimulated with insulin before extraction. HA-mTOR kinase was measured as in (A). (C) Myc-tagged, wild-type S6K1 was expressed alone or with CD2-PI3K in starved or insulin-stimulated HEK293 cells. S6K1 kinase activity was assayed as in Fig. 1A after immunoprecipitation with an antibody to myc. (D) The expressions and activities of HA-mTORwt (WT) and kinase inactive (KI) were assayed against S6K1 Thr389 (top) or 4E-BP1 Ser65 (bottom) (9). ATP concentrations used in the assay are indicated below each panel. (E) The schematic demonstrates distinct control of mTOR activity by ATP and amino acids.

Our data support the hypothesis that intracellular concentrations of ATP directly regulate mTOR, whereas mTOR regulation by amino acids uses a separate mechanism (Fig. 4E). Likewise, in bacteria, amino acids and ATP are sensed by different mechanisms. Amino acid deprivation in bacteria triggers the “stringent response,” the production of the guanosine triphosphate derivative ppGpp, which blocks ribosome biogenesis by directly binding to rRNA polymerase (8). In contrast, as ATP concentrations begin to decrease in growing bacteria, the rate of initiation at rRNA promoters decreases because of the rapid decay of open promoter complexes (20). Here we propose that, as ATP is used in eukaryotic cells, mTOR functions as a homeostatic sensor, adjusting the rate of ribosome biogenesis to reflect intracellular ATP concentrations (Fig. 4E). Increased ribosome biogenesis is a predictive indicator of solid tumor progression (21), and in such tumors, metabolic flux is redirected to glycolysis, leading to the more rapid production of ATP (12,13). If tumors gain an mTOR-specific growth advantage because of increased production of ATP, they may be more susceptible to the effects of rapamycin.

  • * To whom correspondence should be addressed. E-mail: gthomas{at}


View Abstract

Navigate This Article