Research Article

Sestrin2 is a leucine sensor for the mTORC1 pathway

See allHide authors and affiliations

Science  01 Jan 2016:
Vol. 351, Issue 6268, pp. 43-48
DOI: 10.1126/science.aab2674

From sensing leucine to metabolic control

The mTORC1 protein kinase complex plays central roles in regulating cell growth and metabolism and is implicated in common human diseases such as diabetes and cancer. The level of the amino acid leucine tells an organism a lot about its physiological state, including how much food is available, how much insulin is going to be needed, and whether new muscle mass can be made (see the Perspective by Buel and Blenis). Wolfson et al. identified a biochemical sensor of leucine, Sestrin2, which connects the concentration of leucine to the control of organismal metabolism and growth. When leucine bound to Sestrin2, it was released from a complex with the mTORC1 regulatory factor GATOR2, activating the mTORC1 complex. Saxton et al. describe the crystal structure of Sestrin2 and show how it specifically detects leucine. Aylett et al. determined the structure of human mTORC1 by cryoelectron microscopy and the crystal structure of a regulatory subunit, Raptor. The results reveal the structural basis for the function and intricate regulation of this important enzyme, which is also a strategic drug target.

Science, this issue p. 43, p. 48, p. 53; see also p. 25

Abstract

Leucine is a proteogenic amino acid that also regulates many aspects of mammalian physiology, in large part by activating the mTOR complex 1 (mTORC1) protein kinase, a master growth controller. Amino acids signal to mTORC1 through the Rag guanosine triphosphatases (GTPases). Several factors regulate the Rags, including GATOR1, aGTPase-activating protein; GATOR2, a positive regulator of unknown function; and Sestrin2, a GATOR2-interacting protein that inhibits mTORC1 signaling. We find that leucine, but not arginine, disrupts the Sestrin2-GATOR2 interaction by binding to Sestrin2 with a dissociation constant of 20 micromolar, which is the leucine concentration that half-maximally activates mTORC1. The leucine-binding capacity of Sestrin2 is required for leucine to activate mTORC1 in cells. These results indicate that Sestrin2 is a leucine sensor for the mTORC1 pathway.

It has long been appreciated that in addition to being a proteogenic amino acid, leucine is also a signaling molecule that directly regulates aspects of animal physiology, including satiety (1), insulin secretion (2), and skeletal muscle anabolism (3, 4). Because the liver has a low capacity to metabolize leucine, its concentrations in the blood fluctuate according to its consumption, so that dietary leucine can directly affect physiology (57). A key mediator of the effects of leucine is the mTOR complex 1 (mTORC1) protein kinase (8, 9), which regulates growth by controlling processes such as protein and lipid synthesis, as well as autophagy.

In addition to leucine, many environmental signals regulate the mTORC1 pathway, including other amino acids such as arginine, as well as glucose and various growth factors and forms of stress (10, 11). How mTORC1 senses and integrates these diverse inputs is not well understood, but it is clear that the Rheb and Rag guanosine triphosphatases (GTPases) have necessary but distinct roles. Rheb is a monomeric GTP-binding protein, and the Rags function as obligate heterodimers of RagA or RagB bound to RagC or RagD (1214). Both the Rheb and Rag GTPases localize, at least in part, to the lysosomal surface (1518), which is an important site of mTORC1 regulation (19). In a Rag-dependent manner, amino acids promote the translocation of mTORC1 to the lysosome, where Rheb, if bound to GTP, stimulates its kinase activity. Growth factors trigger the GTP-loading of Rheb by driving its GTPase activating protein (GAP), the tuberous sclerosis complex, off the lysosomal surface (18).

Regulation of the Rag GTPases by amino acids is complex, and many distinct factors have important roles (20). A lysosome-associated supercomplex containing Ragulator, SLC38A9, and the vacuolar adenosine triphosphatase (v-ATPase) interacts with the Rag GTPases and is necessary for the activation of mTORC1 by amino acids (2124). Ragulator anchors the Rag heterodimers to the lysosome and has nucleotide exchange activity for RagA and RagB (21, 25). SLC38A9 is an amino acid transporter and a potential lysosomal arginine sensor (23), but the function of the v-ATPase in mTORC1 activation is unclear. Two GAP complexes stimulate GTP hydrolysis by the Rag GTPases, with GATOR1 acting on RagA and RagB (26), and folliculin–folliculin interacting protein 2 (FLCN-FNIP2) acting on RagC and RagD (27). The separate GATOR2 complex negatively regulates GATOR1 through an unknown mechanism and is necessary for mTORC1 activation (26). Last, the Sestrins are GATOR2-interacting proteins that inhibit mTORC1 signaling but whose molecular function is not known (28, 29). Previous reports argue that the Sestrins inhibit mTORC1 signaling through AMPK and TSC (30), but recent studies question this mechanism (28, 31).

The amino acid sensors upstream of mTORC1 have eluded researchers for many years. Whereas SLC38A9 is a strong candidate for sensing arginine at lysosomes (23), the long-sought sensor of leucine has thus far been unknown. Here, we present evidence that Sestrin2 is a leucine sensor for the mTORC1 pathway.

Leucine directly regulates the Sestrin2-GATOR2 interaction

Activation of mTORC1 by amino acids requires the pentameric GATOR2 complex (26). Although its molecular function is unknown, epistasis-like experiments suggest that GATOR2 suppresses GATOR1, the GAP for and inhibitor of RagA and RagB (26). Within cells, Sestrin2 binds to GATOR2 in an amino acid–sensitive manner (28, 29); removal of all amino acids from the culture media induces the interaction, and readdition of the amino acids reverses it (28). Although several amino acids can regulate mTORC1 signaling, arginine and leucine are the best-established, and deprivation of either strongly inhibits mTORC1 in various cell types (8, 32, 33).

In human embryonic kidney–293T (HEK-293T) cells, removal of either leucine or arginine from the cell medium inhibited mTORC1 signaling to similar extents, as indicated by ribosomal protein S6 kinase 1 (S6K1) phosphorylation. However, only leucine depletion caused Sestrin2 to bind to GATOR2, inducing the interaction as effectively as complete amino acid starvation did (Fig. 1A). Leucine readdition rapidly reversed the binding, and amino acids did not affect the interaction between WD repeat–containing protein 24 (WDR24) and Mios, two core components of GATOR2 (Fig. 1A and fig. S1A).

Fig. 1 Leucine, but not arginine, disrupts the Sestrin2-GATOR2 interaction in cells and in vitro.

(A) Binding of Sestrin2 to GATOR2 in HEK-293T cells stably expressing FLAG-WDR24 (a component of GATOR2). Cells were deprived of leucine, arginine, or all amino acids for 50 min. Where indicated, cells were restimulated with leucine, arginine, or all amino acids for 10 min, and FLAG immunoprecipitates (IPs) were prepared from cell lysates. Immunoprecipitates and lysates were analyzed by immunoblotting for the indicated proteins and phosphorylation states. FLAG–methionine aminopeptidase 2 (metap2) served as a negative control. (B) Effects of leucine and arginine on the Sestrin2-GATOR2 interaction in ice-cold detergent lysates of amino acid–starved cells. HEK-293T cells stably expressing FLAG-metap2 or FLAG-WDR24 were deprived of all amino acids for 50 min. Leucine or arginine was added to the culture media or to cell lysates, and FLAG immunoprecipitates were prepared and analyzed as in (A). (C) Effects of individual amino acids on the purified Sestrin2-GATOR2 complex. FLAG immunoprecipitates were prepared from HEK-293T cells stably expressing FLAG-metap2 or FLAG-WDR24 and deprived of all amino acids for 50 min. Indicated amino acids (300 μM) were added directly to the immunoprecipitates, which, after re-washing, were analyzed as in (A). (D) Disruption of the purified Sestrin2-GATOR2 complex by leucine. The experiment was performed and analyzed as in (C), except that indicated concentrations of leucine (Leu) or arginine (Arg) were used. (E) Disruption of the Sestrin2-GATOR2 interaction by isoleucine (Ile) and methionine (Met). The experiment was performed and analyzed as in (C), except that the indicated concentrations of isoleucine, methionine, leucine, or arginine were used.

Sestrin2 is homologous to two other proteins, Sestrin1 and Sestrin3 (3436); when overexpressed, all three can interact with GATOR2 (28). As with Sestrin2, leucine starvation and stimulation strongly regulated the interaction of endogenous Sestrin1 with GATOR2 (fig. S1B). In contrast, endogenous Sestrin3 bound to GATOR2 irrespective of leucine concentrations (fig. S1B), suggesting that this interaction is constitutive or regulated by signals that remain to be defined.

Although enzymatic events triggered by leucine may mediate the effects of leucine on the Sestrin2-GATOR2 interaction, it was tempting to consider the possibility that leucine might act directly on the complex. Consistent with this possibility, the addition of leucine, but not arginine, to ice-cold detergent lysates of cells deprived of all amino acids abrogated the interaction to the same extent as leucine stimulation of live cells did (Fig. 1B). Leucine also disrupted the interaction when added directly to immunopurified Sestrin2-GATOR2 complexes that were isolated from amino acid–deprived cells. Of the 18 amino acids tested at 300 μM each, only those most similar to leucine—methionine, isoleucine, and valine—had any effect on the Sestrin2-GATOR2 interaction in vitro (Fig. 1C).

When added to the purified complexes, leucine dose-dependently disrupted the Sestrin2-GATOR2 complex, with the half-maximal effect at about 1 μM (Fig. 1D). Methionine and isoleucine were considerably less potent, acting at concentrations about 10- and 25-fold greater than leucine, respectively (Fig. 1E). These values reflect only the relative potencies of these amino acids, owing to the fact that equilibrium conditions were not attained because the large assay volume precluded Sestrin2 from rebinding to GATOR2 once it had dissociated.

Sestrin2 binds leucine with a dissociation constant (Kd) of 20 μM

Given that leucine disrupts the purified complex, we reasoned that leucine might bind directly to Sestrin2 or GATOR2. To test this, we developed an equilibrium binding assay in which purified proteins immobilized on agarose beads were incubated with radioactive amino acids, and the bound amino acids were quantified after washing. Radiolabeled leucine bound to Sestrin2 but not to WDR24, the GATOR2 complex, or the control protein Rap2A (Fig. 2A), in a manner that was fully competed by excess nonradiolabeled leucine. In contrast, arginine did not bind to either Sestrin2 or Rap2A (fig. S2A). Consistent with the differential sensitivities to leucine of the Sestrin1- and Sestrin3-GATOR2 complexes, Sestrin1 bound leucine to a similar extent as did Sestrin2, whereas Sestrin3 bound leucine very weakly (Fig. 2B and fig. S2A). Drosophila dSestrin (CG11299-PD) also bound leucine, albeit at lower amounts than the human protein did (fig. S2, B and C).

Fig. 2 Sestrin2 binds leucine with a Kd of 20 μM.

(A) Binding of radiolabeled leucine to Sestrin2 but not to WDR24, GATOR2, or the control protein Rap2A. FLAG immunoprecipitates prepared from HEK-293T cells transiently expressing the indicated proteins or complexes were used in binding assays with [3H]leucine, as described in the supplementary materials. Unlabeled leucine was added where indicated. Values are means ± SD for three technical replicates from one representative experiment (n.s., not significant). SDS–polyacrylamide gel electrophoresis, followed by Coomassie Blue staining, was used to analyze immunoprecipitates that were prepared in parallel to those included in the binding assays. Asterisks in the right panel indicate breakdown products in the WDR24 and GATOR2 purifications. (B) Leucine-binding capacities of Sestrin1 (two isoforms), Sestrin2, and Sestrin3. FLAG immunoprecipitates were prepared and binding assays were performed and analyzed as in (A). (C) Leucine binds to bacterially produced Sestrin2 but not to the RagA/RagC heterodimer. Leucine binding assays were performed as described in the supplementary materials and analyzed as in (A) with polyhistidine (His)–mannose binding protein (MBP)–Sestrin2 or His–RagA/RagC bound to the Ni–NTA (nitrilotriacetic acid) resin. (D) Effects of leucine and arginine on the melting temperature of bacterially produced Sestrin2 in a thermal shift assay. His-MBP-Sestrin2 was incubated with Sypro Orange dye, with or without leucine or arginine. When the sample was heated, the change in fluorescence was captured and used to calculate melting temperatures (Tm) under the indicated conditions. Values are means ± SD for three replicates. (E) Sestrin2 binds leucine with a Kd of 20 μM. FLAG-Sestrin2 immunoprecipitates, prepared as in (A), were used in binding assays with 10 or 20 μM [3H]leucine and indicated concentrations of unlabeled leucine. In the representative graph, each point represents the normalized mean ± SD for n = 3 experiments in an assay with 10 μM [3H]leucine. Kd was calculated from the results of six experiments (three with 10 μM and three with 20 μM [3H]leucine). (F) Methionine can compete with the binding of leucine to Sestrin2. (G) Isoleucine can compete in the binding of leucine to Sestrin2. In (F) and (G), FLAG-Sestrin2 immunoprecipitates, prepared as in (A), were used in binding assays with 10 μM [3H]leucine and indicated concentrations of unlabeled methionine and isoleucine, respectively. In the graphs, each point represents the normalized mean ± SD for n = 3 experiments Ki values were calculated using data from the three experiments.

Because all of these proteins were expressed in and purified from HEK-293T cells, the above results did not rule out the possibility that an unidentified protein that copurifies with Sestrin2 (and Sestrin1) is the actual receptor for leucine. To address this possibility, we prepared human Sestrin2 in bacteria, which is a heterologous system that does not encode a Sestrin homolog or even a TOR pathway. Consistent with the results obtained with Sestrin2 prepared in human cells, radiolabeled leucine bound to bacterially produced Sestrin2 but not to the RagA-RagC heterodimer, which was used as a control (Fig. 2C). Furthermore, in a thermal shift assay, leucine, but not arginine, shifted the melting temperature of bacterially produced Sestrin2, but not of two control proteins, by up to 8.5° C (Fig. 2D and fig. S2, E and F). Collectively, these data strongly suggest that leucine binds directly to Sestrin2.

Although the thermal shift assay is valuable for assessing the capacity of a protein to bind a ligand, this method is not suitable for obtaining an accurate Kd (37). Therefore, we used a competition binding assay with increasing amounts of unlabeled leucine to determine that leucine has a Kd for Sestrin2 of 20 ± 5 μM (Fig. 2E). In comparison, methionine and isoleucine competed with leucine binding with inhibitory constants (Ki) of 354 ± 118 μM and 616 ± 273 μM, respectively (Fig. 2, F and G). These values are respectively about one-eighteenth and one-thirtieth the affinity of leucine for Sestrin2, and they correlate well with the relative potencies of leucine, methionine, and isoleucine in disrupting the Sestrin2-GATOR2 interaction in vitro (Fig. 1, D and E).

Sestrin2 regulates mTORC1 through GATOR2

Consistent with leucine regulating mTORC1 by modulating the binding of Sestrin2 to GATOR2, 20 to 40 μM leucine had half-maximal effects on both the Sestrin2-GATOR2 interaction and mTORC1 activity in HEK-293T cells (Fig. 3, A and B). This concentration range encompasses the Kd of leucine for Sestrin2, indicating that the affinity of Sestrin2 for leucine is physiologically relevant.

Fig. 3 Sestrin2 regulates mTORC1 through GATOR2.

(A) Effects of varying leucine concentrations on mTORC1 activity, as measured by the phosphorylation of S6K1. HEK-293T cells were deprived of leucine for 50 min and restimulated with leucine at the indicated concentrations for 10 min. Cell lysates were analyzed via immunoblotting for the indicated proteins and phosphorylation states. (B) Effects of varying leucine concentrations on the Sestrin2-GATOR2 interaction. HEK-293T cells stably expressing the indicated proteins were starved as in (A) and FLAG immunoprecipitates were collected. The immunopurified complexes were treated with the indicated concentrations of leucine and then analyzed as in Fig. 1C. (C) Decreased GATOR2-binding capacity of the Sestrin2 S190W mutant. FLAG immunoprecipitates were prepared from HEK-293T cells transiently expressing the indicated proteins and were analyzed by immunoblotting for the indicated proteins. (D) Determination of the leucine-binding capacity of Sestrin2 S190W. Assays were performed and immunoprecipitates were analyzed as in Fig. 2A. (E) In Sestrin1, -2, and -3 triple-null cells expressing Sestrin2 S190W, the mTORC1 pathway cannot sense the absence of leucine. Wild-type HEK-293T cells and Sestrin1, -2, and -3 triple-null HEK-293T cells, generated with the CRISPR/Cas9 (clustered regularly interspaced short palindromic repeat/CRISPR-associated nuclease 9) system, were used to express the indicated FLAG-tagged proteins. Cells were starved of leucine for 50 min and, where indicated, stimulated with leucine for 10 min; lysates were analyzed via immunoblotting.

To formally test whether Sestrin2 regulates mTORC1 by interacting with GATOR2, we identified a Sestrin2 mutant (S190W) that binds leucine but has a severely decreased capacity to bind GATOR2 (Fig. 3, C and D). In Sestrin1, -2, and -3 triple-null HEK-293T cells, mTORC1 signaling was active and unaffected by leucine deprivation (Fig. 3E). In these cells, expression of wild-type Sestrin2 restored the leucine sensitivity of the mTORC1 pathway, but expression of Sestrin2 S190W had no effect (Fig. 3E). Thus, Sestrin2 must be able to interact with GATOR2 for the mTORC1 pathway to sense the absence of leucine.

For leucine to activate mTORC1, Sestrin2 must be able to bind leucine

To test whether the leucine-binding capacity of Sestrin2 is necessary for mTORC1 to sense the presence of leucine, we identified two Sestrin2 mutants, L261A and E451A, which do not bind leucine to an appreciable degree (Fig. 4A). Leucine did not meaningfully affect the interaction of the mutants with GATOR2 in vitro, consistent with the role of Sestrin2 in mediating the effects of leucine on the Sestrin2-GATOR2 complex (Fig. 4B). Expression of wild-type Sestrin2 in the Sestrin1, -2, and -3 triple-null cells restored the leucine sensitivity of the mTORC1 pathway in these cells, but expression of either mutant inhibited signaling and rendered the mTORC1 pathway insensitive to leucine (Fig. 4C and fig. S3A). Furthermore, in Sestrin1, -2, and -3 triple-null cells that expressed the mutants, the localization of mTOR to lysosomes in the presence of leucine was decreased, whereas that of RagC was not affected (fig. S4, A to D). Thus, activation of mTORC1 by leucine requires the binding of leucine to Sestrin2.

Fig. 4 The capacity of Sestrin2 to bind leucine is required for the mTORC1 pathway to sense leucine.

(A) The Sestrin2 L261A and E451A mutants do not bind leucine. Binding assays were performed and immunoprecipitates were analyzed as in Fig. 2A. (B) Leucine insensitivity of the interactions of Sestrin2 L261A or E451A with GATOR2. FLAG immunoprecipitates were prepared from cells transiently expressing the indicated proteins. The immunoprecipitates were treated with the indicated concentrations of leucine and analyzed as in Fig. 1C (HA, hemagglutinin epitope). (C) In Sestrin1, -2, and -3 triple-null cells expressing Sestrin2 L261A or E451A, the mTORC1 pathway cannot sense the presence of leucine. Cells were generated and analyzed as in Fig. 3E. (D) Model showing how amino acid inputs from multiple sensors in distinct compartments impinge on the Rag GTPases to control mTORC1 activity.

Conclusions

Sestrin2 has several properties that are consistent with its being a leucine sensor for the mTORC1 pathway: (i) it binds leucine at affinities consistent with the concentrations at which leucine is sensed; (ii) Sestrin2 mutants that do not bind leucine cannot signal the presence of leucine to mTORC1; and (iii) loss of Sestrin2 and its homologs renders the mTORC1 pathway insensitive to the absence of leucine. Although we have not investigated Sestrin1 as thoroughly, it appears to behave similarly to Sestrin2, so we propose that Sestrin1 and Sestrin2 are leucine sensors upstream of mTORC1.

Given that Sestrin2 has appreciable affinity for methionine, it would not be surprising if, in contexts where leucine concentrations are low and methionine concentrations are high, Sestrin2 serves as a methionine sensor for the mTORC1 pathway.

Sestrin2 binds to and likely inhibits GATOR2, but how this leads to suppression of mTORC1 cannot be determined until the molecular function of GATOR2 has been clarified. In addition, structural studies are needed to understand how the binding of leucine to Sestrin2 disrupts its interaction with GATOR2, and why leucine binds very poorly to Sestrin3.

Because Sestrin1 and Sestrin2 are soluble proteins, it is likely that they sense free leucine in the cytosol. Although these concentrations are unknown, the Michaelis constant of the human leucyl–transfer RNA synthetase (LRS) for leucine has been reported to be 45 μM (38). This value is similar to the affinity of Sestrin2 for leucine, suggesting that cytosolic free leucine concentrations are within this range. Like Sestrin2, LRS binds isoleucine and methionine at lower affinities than it binds leucine (about 30-fold less in the case of LRS) (38). The similarities between the amino acid–binding characteristics of Sestrin2 and LRS support the notion that the affinity and specificity of Sestrin2 for leucine are sufficient for it to serve as a leucine sensor.

Our work suggests a model in which signals emerging from distinct amino acid sensors in different cellular compartments converge on the Rag GTPases at the lysosomal surface to regulate mTORC1 activity (Fig. 4D). The putative arginine sensor SLC38A9 probably monitors lysosomal contents, and Sestrin2 is almost certainly a cytosolic sensor. There must also be an amino acid sensor upstream of the FLCN-FNIP2 complex (the GAP for RagC and RagD), but its identity and cellular localization are unknown. A future challenge is to determine how the Rag GTPases integrate the inputs from the different sensors; this will likely require a much better understanding of the function of each Rag in the heterodimer. Moreover, in vivo characterization of the different sensors will be needed to comprehend how specific tissues adapt the amino acid sensing pathway to their particular needs.

Given that Sestrin2 (and Sestrin1) are likely to have leucine-binding pockets, these proteins may be targets for the development of small-molecule modulators of the mTORC1 pathway. Leucine attenuates the decrease in skeletal-muscle protein synthesis that occurs in the elderly and stimulates satiety (1, 37). Thus, small molecules that potently mimic the effects of leucine on Sestrin2 could have therapeutic value. Furthermore, caloric restriction inhibits mTORC1 signaling (40, 41) and is associated with increased health span and life span in multiple organisms (42, 43). Thus, small molecules that antagonize the effects of leucine on Sestrin2 might have caloric restriction–mimicking properties.

Supplementary Materials

www.sciencemag.org/content/351/6268/43/suppl/DC1

Materials and Methods

Figs. S1 to S4

References (44, 45)

REFERENCES AND NOTES

Acknowledgments: D.M.S. is a founder, a member of the Scientific Advisory Board, a paid consultant, and a shareholder of Navitor Pharmaceuticals, which is targeting for therapeutic benefit the amino acid sensing pathway upstream of mTORC1. The His-MBP-Sestrin2 pMAL6H-C5XT plasmid is available from Navitor Pharmaceuticals under a materials transfer agreement with Navitor Pharmaceuticals. R.L.W., L.C., R.A.S., D.M.S., and the Whitehead Institute have filed two provisional patents that relate to the Sestrin2-GATOR2 interaction. We thank all members of the Sabatini Lab for helpful insights, in particular S. Wang for experimental advice; O. Levsh, from the laboratory of J.-K. Weng, for generously providing the control proteins used in the thermal shift assays; Navitor Pharmaceuticals for providing the His-MBP-Sestrin2 pMAL6H-C5XT plasmid; and Cell Signaling Technology for providing many antibodies. This work was supported by grants from NIH (R01 CA103866 and AI47389) and the Department of Defense (W81XWH-07-0448) to D.M.S.; fellowship support from NIH to R.L.W. (T32 GM007753 and F30 CA189333) and L.C. (F31 CA180271); and a grant from the Paul Gray Undergraduate Research Opportunities Program Fund to S.M.S. (3143900). K.S. is a Pfizer Fellow of the Life Sciences Research Foundation. D.M.S. is an investigator of the Howard Hughes Medical Institute.
View Abstract

Navigate This Article