Regulation of TORC1 in response to amino acid starvation via lysosomal recruitment of TSC2

Delineation of interacting regions of TSC2 and the Rag proteins

We next aimed to characterize in more detail the binding between TSC1/2 and the Rag complex. We first asked which component of the Rag complex is binding TSC1/2. Immunoprecipitation of tagged TSC2 showed that it binds RagA significantly more strongly than the other Rag proteins (Figure 1D), suggesting RagA is a likely binding partner for the TSC1/2 complex. To study the residues in RagA involved in TSC2 binding, we exploited the fact that RagA has significantly stronger binding to TSC2 compared to RagB despite the two proteins being almost identical. Compared to RagB, RagA is lacking an N-terminal extension of 33 amino acids, and has 5 amino acid substitutions at the C-terminus of the protein (Figure 2A). We asked which of these differences are important for RagA binding to TSC2. Both removal of the N-terminal extension of RagB, as well as introduction of the five C-terminal amino acid changes into RagB caused improved binding to TSC2 (Figures 2A-B) indicating that both the N-terminal and the C-terminal residues affect TSC2 binding. We then introduced the five RagA-specific amino acid substitutions individually into RagB and found that each of them improves the binding of RagB to TSC2 (Figures 2A and 2C). Hence, each of these five amino acids of RagA contributes towards TSC2 binding, and indeed these 5 amino acid differences between RagA and RagB are evolutionarily conserved.

To identify the binding partner in the TSC1/2 complex, we coIPed epitope tagged TSC1 or TSC2 individually with RagA+C, revealing that TSC2 but not TSC1 binds the Rag complex (Figure S2A). To study which residues in TSC2 are responsible for Rag binding, we generated a series of C-terminal truncations of TSC2 and tested them by coIP (Figure S2B). This analysis revealed that the N-terminal 424 amino acids of TSC2 bind very strongly the Rag proteins, but that nonetheless the remaining amino acids 425-1784 also have Rag binding capacity (Figure 2D). In sum, binding between the two complexes appears to be mediated via TSC2 binding to RagA. Although the interaction surfaces are complex, binding depends strongly on five individual amino acids in RagA identified in this analysis.

Binding between Rag proteins and TSC2 increases upon amino acid removal

We asked whether binding between the Rag proteins and TSC2 depends on cellular amino acid signaling, since our initial findings suggested that the Rag complex might be recruiting a TOR inhibitor upon amino acid removal. Indeed, a complex consisting of human RagA and RagC (“RagA+C”) was able to coIP endogenous TSC2 (lanes 1 and 2, Figure 2E) and this interaction became stronger upon amino acid removal (lanes 2 versus 4, Figure 2E). One mechanistic explanation could be that binding between TSC2 and the Rag GTPases depends on the state of activation of the Rag proteins, which switch to the ‘inactive’ RagA^{low-nucleotide}/RagC^GTP conformation upon amino acid removal. To test this, we assayed binding between endogenous TSC2 and mutant variants of the RagA/RagC complex, locked into either the ‘active’ (RagA^GTP/RagC^{low-nucleotide}) or ‘inactive’ (RagA^{low-nucleotide}/RagC^GTP) states (Sancak et al., 2008), and found that indeed TSC2 binds the Rag proteins much more strongly when they are in the ‘inactive’ state (Figure 2F). Binding of endogenous TSC1 to the Rag proteins closely paralleled binding of TSC2 (Figure 2F), indicating that TSC2 is mediating binding of the entire TSC1/2 complex since TSC1 cannot bind the Rag proteins by itself (Figure S2A). As expected (Sancak et al., 2008), Raptor binding was strongest when the Rag proteins were in the active state (Figure 2F), therefore anti-correlating with TSC2 binding. As a negative control endogenous PTEN did not show any binding to the Rag proteins (Figure 2F). This preference for binding the inactive Rag complex cannot be ascribed to the state of activation of any one of the two Rags, but rather appears to depend on the conformation of both RagA and RagC, since states of intermediate activation (RagA^GTP/RagC^GTP or RagA^{low-nucleotide}/RagC^{low-nucleotide}) also showed intermediate levels of TSC2 binding (Figure S2C). In sum, binding of TSC2 to the Rag complex appears to anti-correlate with binding of mTORC1 to the Rag complex, with mTORC1 binding predominantly when the Rag complex is active, in the presence of amino acids, and TSC2 binding predominantly when the Rag complex is inactive in the absence of amino acids.

TSC2 is recruited to the lysosome in a Rag-dependent manner upon amino acid removal

Preferential binding of TSC2 for the ‘inactive’ Rag complex, which is localized on the lysosome, raised the possibility that TSC2 might be recruited to lysosomes upon amino acid removal. Using an antibody that specifically recognizes endogenous TSC2 ((Dibble et al., 2012) and Figure S3A), we found that in the presence of amino acids TSC2 is diffusely cytoplasmic in HEK293FT cells, in a pattern that is not concentrated on lysosomes/late-endosomes (marked by LAMP2, henceforward referred to as ‘lysosomes’) relative to elsewhere in the cytoplasm (+aa, Figure 3A). Upon removal of amino acids, however, TSC2 quickly accumulates on lysosomes (15min −aa, Figure 3A), and remains associated with lysosomes for hours (240min −aa, Figure 3A). Upon re-addition of amino acids, TSC2 quickly becomes diffusely cytoplasmic again (Figure 3B). Similar relocalization of TSC2 to lysosomes could also be observed in mouse embryonic fibroblasts (MEFs) (Figure 3C), and it was confirmed at the ultrastructural level by anti-TSC2 immuno-electron microscopy (Figure S3B). Furthermore, TSC1 also relocalizes to lysosomes upon amino acid removal, suggesting that the whole TSC1/2 complex relocalizes (Figure S3C).

Open in a separate window

Figure 3

Amino acid starvation causes TSC2 lysosomal localization in a Rag-dependent manner

(A,C) TSC2 accumulates on lysosomes upon amino acid deprivation. Confocal micrograph of HEK293FT cells (A) or MEFs (C) treated +/− all amino acids in the presence of dialyzed FBS for the indicated times, stained for TSC2 (green) and the lysosomal marker LAMP2 (red). LAMP2 aggregates do not show TSC2 accumulation in the +aa condition.

(B) TSC2 quickly relocalizes away from lysosomes upon amino acid readdition. HEK293FT cells, starved for amino acids in the presence of dialyzed serum for 4h, then re-supplied with amino acid-containing media for 5 minutes. Cells stained and imaged as in (A).

In agreement with the biochemical data presented above, lysosomal accumulation of TSC2 upon amino acid withdrawal is mediated via the Rag proteins, since it is strongly blunted upon amino acid starvation in MEFs lacking p14/LAMTOR2 in which the Rag proteins are not lysosomally localized (Sancak et al., 2010) (Figure 4A-A’), in MEFs or HEK293FT cells transiently lacking expression of LAMTOR components due to siRNA-mediated knock-down (data not shown), or in MEFs in which the Rag proteins have been knocked-down (Figure 4B). As expected (Sancak et al., 2008), Rag knock-down caused reduced TORC1 activation upon amino acid re-addition (Figure S4A) as well as reduced TOR lysosomal localization (Figure S4B), confirming successful knock-down of the Rag proteins. In agreement with the biochemical data showing preferential binding of TSC2 for the inactive Rag complex, recruitment of TSC2 to lysosomes in the absence of amino acids was also abrogated upon knock-down of the Rag GAP GATOR1 complex components DEPDC5, NPRL2 and NPRL3 (Figures 4C and S4E). In this setup, the Rag complex is present on lysosomes but remains in the activated state (Bar-Peled et al., 2013; Panchaud et al., 2013), leading, as previously shown, to increased TORC1 activity and lysosomal localization in the absence of amino acids (Figures S4C-D). These data suggest that recruitment of TSC2 to the lysosome upon amino acid withdrawal is due to the Rag proteins changing to the inactive conformation.

Open in a separate window

Figure 4

Lysosomal recruitment of TSC2 depends on the Rag proteins

(A-A’) Lysosomal localization of TSC2 upon amino acid removal requires the LAMTOR complex. Confocal micrograph of p14/LAMTOR2 knock-out MEFs (A), or control p14 knock-out MEFs reconstituted to express an EGFP-p14 fusion (A’), treated +/− amino acids for 1h in the presence of dialyzed FBS. Lysosomes marked either with anti-LAMP2 antibody or with the lysosomaly localized EGFP-p14.

(B) Lysosomal localization of TSC2 upon amino acid removal requires the Rag proteins. MEFs transfected with control siRNA against Renilla luciferase or siRNAs targeting the various Rag proteins, treated +/− amino acids for 1h in the presence of dialyzed FBS. Cells stained and imaged as in (A).

(C) Lysosomal localization of TSC2 upon amino acid removal requires that the Rag proteins change to the inactive conformation. MEFs transfected with control siRNA or siRNAs targeting components of the GATOR1 complex DEPDC5, NPRL2, or NPRL3 required for the Rag proteins to become inactive upon amino acid removal. Cells stained and imaged as in (A).

TSC2 is required for complete inactivation of TORC1 upon amino acid removal

Having seen that TSC2 is recruited to lysosomes upon amino acid removal, we asked whether this has a functional consequence in terms of TORC1 activity. We tested if cells lacking TSC2 are impaired in their ability to turn off TORC1 upon amino acid removal. Unlike control Drosophila S2 cells which shut off TORC1 almost completely upon amino acid starvation, S2 cells with decreased Tsc2 levels show only a partial, initial drop in TORC1 activity, after which they retain a significantly elevated amount of TORC1 activity stably for up to 2 hours (Figure 5A). This cannot be explained simply by elevated starting levels of TORC1 activity in the Tsc2 knock-down cells, because S2 cells lacking PTEN have similarly elevated starting levels of TORC1 activity, but nonetheless efficiently shut off TORC1 upon amino acid removal (Figure 5B). These data are in agreement with previous observations (Gao et al., 2002). As in Drosophila cells, mouse embryonic fibroblasts lacking TSC2 or TSC1 are also unable to completely inactivate mTORC1 upon amino acid starvation (Figures 5C and S5A). Also in mouse cells this cannot be explained by elevated starting levels of mTORC1 activity because PTEN knock-out MEFs also start with elevated mTORC1 activity but efficiently shut off mTORC1 upon amino acid withdrawal (Figure 5D). Treatment of TSC1^−/− or TSC2^−/− MEFs with the TORC1-specific inhibitor rapamycin eliminated the aberrantly elevated S6K phosphorylation, showing that it is reflecting elevated activity of TORC1 and not another kinase (Figures 5C and S5A). An alternate explanation could be that lack of the TSC1/2 complex leads to reduced phosphatase activity towards S6K. To test this, we performed a rapamycin treatment timecourse to compare the rate of S6K dephosphorylation in control and TSC2 knock-down conditions. In both MEFs and Drosophila S2 cells, however, rapamycin treatment caused rapid dephosphorylation of S6K with similar kinetics in cells containing or lacking TSC2 (Figures S5F-G, see quantification), indicating that phosphatase activity is unchanged. We wondered whether cells lacking TSC2 might simply have a delayed response to amino acid starvation, in which case extending the time of incubation in medium lacking amino acids would allow them to more completely shut off mTORC1. Astoundingly, however, TSC2^−/− MEFs retained elevated levels of mTORC1 activity in the absence of amino acids for up to 1.5-2 days, at which point they started dying, leading to concomitant drops in both phospho-S6K and total S6K signals on the immunoblots (Figure S5B). When phospho-S6K signals are normalized to total S6K levels, one can see that TSC2^−/− MEFs retain a consistent level of mTORC1 activity up to their death (lower panel Figure S5B). In sum, TSC2 seems to be intimately linked to amino acid sensing by cells, in that 1) TSC2 localization changes upon amino acid removal, and 2) TSC2 is required for mTORC1 to fully turn off in the absence of amino acids.

Open in a separate window

Figure 5

TSC2 loss causes insensitivity to amino acid removal in Drosophila and mammalian cells

(A-B) Tsc2 but not PTEN knock-down makes Drosophila cells largely insensitive to amino acid removal. (A) Immunoblots of Drosophila S2 cells treated with no dsRNA, dsRNA against GFP (control), or dsRNA against Tsc2, treated with media containing (+) or lacking amino acids for the indicated times. (B) PTEN knock-down S2 cells efficiently inactivate TORC1 upon amino acid removal, despite starting with elevated TORC1 activity levels.

(C-D) TSC2 knock-out MEFs do not completely shut off TORC1 upon amino acid withdrawal, whereas PTEN knock-out MEFs do. Immunoblots from (C) TSC2^−/− and the respective control TSC2^+/+ MEFs, or (D) PTEN knock-out MEFs, treated with medium +/− amino acids for the indicated times. Lanes 6 and 12 (C), cells treated +20nM Rapamycin for 1h in the presence of amino acids (‘Rapa’). Quantified with a LI-COR imaging system.

TSC2 is required to completely release mTORC1 from the lysosome upon amino acid starvation

mTORC1 activation requires two regulated steps – one is activation of Rheb, and the other is localization of mTORC1 to the lysosomal surface where it meets Rheb, thereby forming an active complex (Inoki et al., 2003; Tee et al., 2003). In particular, the presence or absence of amino acids is known to regulate mTORC1 localization (Kim et al., 2008; Sancak et al., 2008). Since TSC2 appears to be involved in amino acid signaling to mTORC1, we asked whether TSC2 is required for the proper subcellular localization of mTORC1. As previously described (Kim et al., 2008; Sancak et al., 2008), in control cells mTOR accumulates on lysosomes in the presence of amino acids, and becomes diffusely cytoplasmic in the absence of amino acids (Figure 6A). In contrast, in TSC2-null MEFs (which, as expected, are larger than control cells) mTOR localization is very strongly lysosomal, both in the presence and in the absence of amino acids (Figure 6A’ and top panels Figure 6B). This defect can be rescued by transfecting them with a TSC2 expression plasmid, confirming specificity of the phenotype (lower panels in Figure 6B, note lysosomally localized mTOR in the starved non-transfected TSC2^−/− cells, which is rescued in the transfected cell expressing TSC2 and EGFP). As expected from the known effect of TSC1 loss on TSC2 stability, this defect in mTOR localization could also be observed in TSC1-null MEFs (Figure S6A). Therefore, unexpectedly, the TSC1/2 complex is required for mTORC1 to be released completely from lysosomes upon amino acid withdrawal. Consistent with this phenotype being specific for loss of the TSC1/2 complex, and in agreement with the data presented in Figure 5, mTOR was readily released upon amino acid removal from lysosomes in PTEN-mutant MEFs, which also start with elevated mTORC1 activation levels (Figure S6B).

Open in a separate window

Figure 6

TSC2 is required for mTOR to localize away from lysosomes upon amino acid removal in a Rheb dependent manner

(A-B) mTOR is released from lysosomes upon amino acid removal in control (A) but not TSC2-null MEFs (A’). This is rescued by re-expressing TSC2 + EGFP (to mark transfected cells) but not EGFP alone in the TSC-null MEFs (B). Note that the cell expressing TSC2 and GFP no longer has mTOR accumulated on lysosomes, whereas the surrounding, non-transfected cells retain lysosomally localized mTOR.

(C) Defective release of mTOR from lysosomes upon amino acid withdrawal in TSC2-null MEFs is rescued by knocking down Rheb. TSC2-null MEFs transfected with either Rheb siRNAs (upper panels) or control siRNAs (lower panels) treated 1h with +/− amino acid containing medium in the presence of dialyzed FBS.

Since mTORC1 localization is known to be regulated via the GTP/GDP load of the Rag GTPases (Sancak et al., 2008), and TSC2 has been described as a GAP for other small GTPases (Inoki et al., 2003; Wienecke et al., 1995; Xiao et al., 1997), we hypothesized TSC2 might regulate mTORC1 localization by acting as a Rag GAP. However, we radioactively quantified the relative GTP/GDP load of overexpressed Rag proteins in cells in the presence or absence of co-expressed TSC2, and could find no evidence to support this (data not shown). Another possible explanation could be that TSC2 affects the level of intracellular amino acids, thereby affecting mTORC1 localization. However, this does not seem to be the case: The intracellular levels of 10 amino acids drop upon amino acid starvation. For these, the intracellular levels drop equivalently, or even more dramatically, in the TSC2-null MEFs compared to control cells (Figure S5C). The intracellular levels of 8 other amino acids, unexpectedly, either remain constant or increase in control cells upon amino acid removal from the medium (Figures S5D-E). For these, the intracellular levels either increase less dramatically, or actually decrease, in TSC2-null MEFs compared to controls (Figures S5D-E). Since TSC2 acts via Rheb (Inoki et al., 2003), and Rheb is known to bind the mTORC1 complex (Long et al., 2005), this raised the possibility that Rheb might mediate the effect of TSC2 on mTOR localization by helping to anchor mTOR at the lysosomal membrane in collaboration with the Rag proteins. Indeed, in TSC2-null MEFs, knock-down of Rheb by siRNA allowed mTOR to be released from lysosomes upon amino acid removal (Figure 6C, controls in Figures S6C-D). Equivalent results could be observed in TSC2-null cells stably transfected with inducible shRNA constructs targeting an independent region of Rheb, confirming specificity of the effect (Figure S6E). Furthermore, overexpression of active, but not inactive, Rheb was sufficient to localize mTOR to the lysosome even in the absence of amino acids (data not shown). In sum, these results identify Rheb as the TORC1-anchoring activity in the TSC2-null MEFs.

Cell Imaging

All cell images within one panel were acquired and displayed with the same settings. Details of confocal microscopy in Supplemental Material.

A.T. would like to dedicate this paper to Rich Losick and Steve Cohen, who taught him how to do research. We thank David Kwiatkowski and Michael Hall for the TSC2 mutant MEFs, Lukas Huber for the p14^−/− MEFs, and Hong Wu for the PTEN-mutant MEFs. We would also like to aknowledge David Sabatini for supplying the pRK5-HA-GST-Rag plasmids (Adgene #19298, 19301, 19304, 19307), Kun-Liang Guan for the pcDNA3-HA-RagA and RagC point mutants, Brendan Manning for the pcDNA3-FLAG-TSC2 and pRK7-FLAF-TSC1 expression vectors, and DJ Pan for the pAc5.1-Tsc2-V5-His expression vector. Several antibodies were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa. We thank the Metabolomics Core Technology Platform of the Excellence cluster ‘CellNetworks’ (University of Heidelberg), and the Deutsche Forschungsgemeinschaft (grant ZUK 40/2010-3009262) for support with the HPLC-based intracellular amino acid measurements, the DKFZ Mass Spec core facility for protein identifications, Androniki Kolovou and Jacomine Krijnse Locker and the University of Heidelberg EM Core Facility for support with electron microscopy experiments, and Sibylle Schleich and Katrine Weischenfeldt for technical support. This work was supported by the European Research Council (ERC) under the European Union’s 7^th Framework Programme.