Rheb Binds Tuberous Sclerosis Complex 2 (TSC2) and Promotes S6 Kinase Activation in a Rapamycin- and Farnesylation-dependent Manner*

Recently the tuberous sclerosis complex 2 (TSC2) tumor suppressor gene product has been identified as a negative regulator of protein synthesis upstream of the mTOR and ribosomal S6 kinases. Because of the homology of TSC2 with GTPase-activating proteins for Rap1, we examined whether a Ras/Rap-related GTPase might be involved in this process. TSC2 was found to bind to Rheb-GTP in vitro and to reduce Rheb GTP levels in vivo. Over-expression of Rheb but not Rap1 promoted the activation of S6 kinase in a rapamycin-dependent manner, suggesting that Rheb acts upstream of mTOR. The ability of Rheb to induce S6 phosphorylation was also inhibited by a farnesyl transferase inhibitor, suggesting that Rheb may be responsible for the Ras-independent anti-neoplastic properties of this drug.

The Ras subfamily of small GTPases regulates a vast array of biological events that include cell growth, differentiation, and transformation (1). The prototypic Ras proteins, Ha-, K-, and N-Ras, are known to transduce mitogenic and differentiation signals from cell surface receptors to the nucleus (1). However, despite their conservation throughout eukaryotic evolution, the function of Ras-related GTPases such as Rap1, R-Ras, Ral, Rheb, and Rit remains, at best, poorly understood. Ras activity is regulated by a GDP/GTP cycle whereby guanine nucleotide exchange factors promote the release of GDP from inactive Ras, facilitating its loading with the more abundant GTP (2). Binding to GTP induces a conformational change enabling interaction with and activation of downstream effector proteins (1). The termination of this signal is regulated by GTPase-activating proteins (GAPs), 1 which greatly accelerate the intrinsic GTPase activity of Ras, returning it to its inactive state (2). Loss of Ras GAP, as occurs in the genetic disorder neurofibromatosis type 1, results in increased Ras GTP levels, which contribute to tumor development (3,4).
Tuberous sclerosis complex (TSC) and lymphangioleiomyomatosis (LAM) are additional diseases that could be linked to the loss of GTPase regulation. TSC is a genetic disorder that results in the formation of benign tumors known as hamartomas, most typically found in kidney, brain, heart, and lung (3,5). LAM is a devastating lung disease that affects mainly women and is characterized by proliferation of atypical smooth muscle cells within the lung parenchyma (6). The inactivation of two tumor suppressor genes has been associated with TSC and LAM: TSC1 encodes the protein hamartin (TSC1), and TSC2 that encodes tuberin (TSC2) (7)(8)(9). Although TSC1 contains coiled-coil domains that are important for the formation of a functional complex with TSC2, TSC2 shares sequence homology with a family of GAPs that regulate the Ras-related GTPase, Rap1 (5). Accordingly, TSC2 has been reported to increase the intrinsic GTPase activity of Rap1 and also Rab5 in vitro (10,11). However, it is not known whether this activity occurs in vivo or if it contributes to the physiological function of the TSC1-TSC2 complex.
Several recent studies in Drosophila and mammalian cells have indicated that the TSC1-TSC2 complex represses cell growth downstream of insulin through the inhibition of S6K, a key component in the regulation of protein translation (12)(13)(14)(15)(16)(17).
Although not all studies are in full agreement, the TSC complex appears to negatively regulate the S6K activator, mTOR (13,14,16,18). This inhibition is relieved by direct phosphorylation of TSC2 by Akt (13, 19 -21). However, the biochemical function of the TSC complex remains unclear. Over-expression of the C terminus of TSC2, which lacks interaction with TSC1, was sufficient to suppress tumorigenesis in cancer cell lines and in a transgenic rat system, suggesting that it is the active component of the TSC complex (22,23). Intriguingly, this region of TSC2 encompasses the Rap GAP-related domain and is the most highly conserved region of TSC2 in eukaryotic species ranging from yeast to man (5).
In this study, in an effort to understand its mechanism of action, we investigated whether TSC2 could regulate Rap1 or related GTPases in vivo. Our results suggest that TSC2 suppresses mTOR-dependent activation of S6K via the Ras-related GTPase, Rheb.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-Full-length human TSC2 and TSC1 were kindly provided by E. Henske and A. Astrinidis (Fox Chase Cancer Center). TSC1 was subcloned into the NotI site of a modified pFLAG-CMV2 vector (Sigma) in which the FLAG tag had been frame-shifted by cutting, blunting, and religating the HindIII site. pBluescript SK-II(ϩ)- TSC2 was digested with KpnI and BamHI and inserted into the modified pFLAG-CMV2 vector. To clone the C terminus of TSC2 (residues 965-1807), an EcoRI/BamHI fragment from full-length TSC2 was ligated in-frame with the FLAG epitope of the modified pFLAG-pCMV2. Rheb wild type (24) was subcloned from pGEX-2T into the BamHI site of pCGN. GST-Ras constructs were as previously described (25).
Miscellaneous-Ras protein binding assays were performed as described previously (25), except the nonhydrolyzable GTP analog, Gp-pNHp (Sigma), and ϳ5 g of GST-Ras fusion proteins were used. For TSC2 expression, 293T cells were transfected using the calcium phosphate precipitation method. In vivo GAP activity was determined essentially as described (25), except that cells were transfected by calcium phosphate precipitation, and only 150 ng of HA-tagged Rheb plasmid was used for the in vivo Rheb GAP assay.

RESULTS AND DISCUSSION
Several recent studies have demonstrated that the tumor suppressor complex TSC1-TSC2 is an antagonist of mTOR-dependent activation of S6K (13,16,18). Although this inhibition can be relieved by Akt-mediated phosphorylation of TSC2 (13, 19 -21), the mechanism of action of TSC1-TSC2 is poorly understood. Studies in tumor cell lines and a transgenic rat system have demonstrated that the tumor suppressor activity of TSC2 is located within a C-terminal region that encompasses a sequence homologous to that of the catalytic domain of Rap GAPs (22,23). As an initial step to address the functional importance of the GAP-related domain of TSC2, we metabolically labeled the guanine nucleotide pool of 293T cells expressing HA-tagged Rap1A using 32 P i and examined the Rap1 GTP/ (GTP ϩ GDP) ratio in the presence or absence of overexpressed TSC2 or RapGAP1 (26), a known Rap GAP. The basal level of 7-10% Rap1A-GTP was reduced to almost undetectable levels upon expression of RapGAP1 (Fig. 1). However, over-expression of TSC2 did not affect the amount of GTPbound Rap1A. Because TSC2 complexes with TSC1, it was possible that the latter is necessary for the correct localization and interaction of TSC2 with Rap1A in vivo. Thus, we tested the levels of GTP-bound Rap1A when TSC1 and TSC2 are over-expressed together. Although TSC1 stabilized TSC2 ex-pression, we still could not detect any TSC2 GAP activity toward Rap1A (Fig. 1). These results suggest that TSC2 does not act as a Rap1 GAP in vivo but do not exclude a functional activity for the Rap1 GAP homologous domain of TSC2.
To determine whether TSC2 might alternatively regulate other Ras family members, we expressed a FLAG-tagged Cterminal fragment of TSC2 and attempted to precipitate it from 293T cell lysates using a panel of GST-fused Ras proteins that had been loaded with the nonhydrolyzable GTP analog, Gp-pNHp. As shown in Fig. 2, the truncated TSC2 was efficiently precipitated by GST-Rheb but only weakly or not at all by Rap1A, Ral, Ha-Ras, or R-Ras.
To examine whether the binding of TSC2 to Rheb promotes GTP hydrolysis in vivo, we performed experiments similar to those described in Fig. 1 to measure the levels of GTP-bound Rheb following TSC2 over-expression. As reported previously (27), the basal level of GTP-bound Rheb was higher than that of Rap1A (Fig. 3). Over-expression of TSC2 significantly decreased the levels of GTP-bound Rheb (Fig. 3), consistent with TSC2 binding to Rheb and acting as a Rheb GAP in vivo. However, because the GTP-bound state of Rheb is influenced by its expression level (27), the reduced binding of Rheb to GTP in the presence of TSC2 may just be a reflection of its reduced expression (Fig. 3, compare intensities of spots in lanes 1 and 2 with those in lanes 3 and 4).
The cellular function of Rheb in mammals is unknown (24,27,28). However, genetic studies in fission yeast have implicated it in amino acid uptake and metabolism. Loss of Rhb1, a Rheb-related GTPase in Schizosaccharomyces pombe, causes growth arrest and a cellular morphology similar to that generated by nitrogen starvation (29). Additionally, the Saccharomyces cerevisiae Rheb was shown to regulate the uptake of basic amino acids (30). Mutation of the S. pombe TSC1-TSC2 complex also disrupts nutrient uptake, supportive of a functional link between Rheb and TSC2 (31). The mammalian TSC1-TSC2 complex seems to regulate cell growth by inhibiting the mTOR-dependent activation of S6K1. Indeed, overexpression of TSC2 (12) or the TSC1-TSC2 complex (Fig. 4A  (13, 16)) significantly reduced basal-and insulin-stimulated phosphorylation of ribosomal protein S6. Because TSC2 decreases the level of active, GTP-bound Rheb in vivo, it seemed likely that Rheb is the cellular target of TSC1-TSC2 in the mTOR signaling pathway. Therefore, we explored whether over-expression of wild type Rheb or a GTPase-defective Rheb(64L) mutant affects endogenous S6 phosphorylation. Both proteins induced strong phosphorylation of S6 at serines 235 and 236, whereas constitutively activated Rap1A(63E) or Rap1A(64A) mutants were ineffective (Fig 4B). The strong activity of the wild type Rheb was consistent with the high levels of active GTP-bound Rheb detected under basal conditions (see Fig. 3), whereas the weaker activity of Rheb(64L) was likely because of its lower expression.
Phosphorylation of the ribosomal protein S6 occurs upon S6K activation. Because S6K is activated by phosphorylation on several residues (32), we analyzed S6K activity using phospho-specific antibodies. Over-expression of wild type Rheb in- FIG. 1. TSC2 is not a Rap1A GAP in vivo. 293T cells were transfected with plasmid encoding epitope-tagged wild type Rap1A along with pCMV2, pCMV2-TSC1, TSC2, TSC2 plus TSC1, or pCMV2-Rap-GAP1 as indicated. The guanine nucleotide pool was metabolically labeled in the absence of serum using 32 P i . Rap1A-bound GTP and GDP levels were determined following immunoprecipitation, thin layer chromatography, and quantification of spots. %GTP indicates the average (GTP cpm/(1.5 ϫ GDP) cpm ϩ GTP cpm) ϫ 100 ratio from three independent experiments.
FIG. 2. TSC2 efficiently binds to Rheb. GST-Ras proteins were loaded with GppNHp and used to precipitate a FLAG-tagged C-terminal TSC2 fragment (FLAG-⌬N-TSC2) from 293T cell lysates in the presence of MgCl 2 . Bound TSC2 was detected by Western blotting with M2 anti-FLAG antibody. The experiments were performed at least two times, giving identical results. duced strong phosphorylation of endogenous S6K at residues Thr-389, Thr-421, and Ser-424, which correlated with an increase in S6 phosphorylation (Fig. 5). Although it has been shown that TSC1-TSC2 appears to regulate S6K phosphorylation primarily at residue 389 (12,13,19), cells derived from LAM lesions and TSC2Ϫ/Ϫ cells exhibit constitutive phosphorylation of S6K at residues Thr-389, Thr-421, and Ser-424 (12).
Because the effect of TSC1-TSC2 on S6K activity is mTORdependent, we next determined whether Rheb affects S6 phosphorylation in an mTOR-dependent manner. Treatment with rapamycin, which inhibits mTOR activity, completely blocked the increase in S6 phosphorylation induced by wild type Rheb, whereas the MEK inhibitor U0126 was ineffective (Fig. 5). Overall, these results are consistent with TSC1-TSC2-mediated inhibition of the mTOR pathway occurring at the level of Rheb. Although these data do not conclusively establish whether Rheb acts directly on mTOR or via a parallel pathway to coordinate S6 phosphorylation, in favor of TSC1-TSC2 and Rheb acting upstream of mTOR, it has been shown that dTSC1-TSC2 interact with dTOR in Drosophila melanogaster cells (18), and coexpression of TSC1-TSC2 with mTOR in mammalians cells inhibits mTOR kinase activity (13).
On the basis of current knowledge, we propose that Rheb-GTP is an upstream activator of mTOR. In this model, TSC2 acting as a Rheb GAP would prevent Rheb from activating mTOR. After mitogen stimulation, Akt phosphorylates TSC2, reducing its Rheb GAP activity, thus increasing the amount of Rheb-GTP and consequently mTOR kinase activity. However, this model is not consistent with some studies that failed to detect significant changes in mTOR activity upon mitogenic stimulation (33)(34)(35). An alternative possibility is that Rheb might alleviate the activity of a negative regulator of mTORdependent S6K activation. In this regard, it has been shown that protein phosphatase 2A-mediated regulation of S6K is controlled by mTOR (36). A model in which Rheb regulates accessibility of mTOR to such a phosphatase would be consistent with our findings. During the review process, additional studies confirmed that Rheb is a component of the insulin/TOR pathway and that TSC2 possesses Rheb GAP activity (37)(38)(39)(40). However, the specific mechanism by which Rheb regulates the mTOR pathway remains to be established.
Aberrant up-regulation of the mTOR signaling pathway resulting from the loss of either TSC2 or TSC1 function in TSC and LAM patients points to rapamycin as a potential treatment. The present work also suggests that Rheb is a critical component of the mTOR pathway. Like Ras, Rheb is posttranslationally modified by the farnesyl lipid, enabling attachment to membranes for normal cellular function (24). Indeed, farnesylation of Rheb is essential for cell cycle progression in yeast and cannot be substituted by the related geranylgeranyl moiety (41). FTIs have been designed as anti-tumor agents to block the lipid modification of oncogenic Ras. However, the anti-neoplastic activities of these inhibitors appear to be because of their ability to alter the prenylation and function of proteins other than Ras (42)(43)(44). Interestingly, FTIs block S6K activation (45), suggesting that Rheb may be the elusive farnesyl transferase substrate responsible for tumor cell growth arrest. To evaluate whether farnesylation is important for the mTOR-dependent S6 phosphorylation induced by Rheb, we treated cells with FTI 277. As anticipated, FTI 277 dramatically inhibited basal and Rheb-stimulated S6 phosphorylation FIG. 3. TSC2 decreases Rheb-GTP levels in vivo. 293T cells were transfected with epitope-tagged wild type Rheb and TSC2 and Rheb-GTP levels determined as in Fig. 1. %GTP represents the average Rheb-GTP concentration from three independent experiments.

FIG. 4. Rheb induces S6 phosphorylation.
A, 293T cells were transfected with empty vector or vector encoding FLAG-tagged TSC1 plus TSC2 as indicated. Basal or insulin-stimulated (0.5 g/ml, 15 min.) phosphorylation of endogenous ribosomal protein S6 was determined by immunoblotting cell lysates with a phospho-specific antibody for S6 at Ser-235 and Ser-236 (pS6). B, 293T cells were transfected with empty vector, HA-tagged wild type Rheb (wt), Rheb(64L), Rap1A(63E), or Rap1A(64A) as indicated. Mutant Rheb migrated anomalously. The phosphorylation of endogenous S6 was determined as described in A. In A and B, an anti-ribosomal S6 protein antibody was used to confirm unchanged S6 levels and an anti-glyceraldehyde-3-P dehydrogenase antibody (GAPDH) to confirm equal protein loading. Results shown are representative of at least three independent experiments.
FIG. 5. Rheb-induced phosphorylation of S6 correlates with activation of S6K. 293T cells were transfected with empty vector or vector encoding HA-tagged wild type Rheb as indicated. Endogenous S6K activation was examined using phospho-specific antibodies for p70S6K at Thr-389 and at Thr-421 and Ser-424. The same lysates were also analyzed for S6 phosphorylation at Ser-235 and Ser-236. Note that Rheb also induced phosphorylation of p85S6K at Thr-444 and Ser-447 but not at Thr-412. Data are representative of three independent experiments. (Fig. 6). These observations imply that Rheb farnesylation is critical for upstream regulation of S6K activity and may be the cellular target of FTIs, and that these drugs may be useful in the clinical treatment of TSC and/or LAM.
In summary, our present results have demonstrated that TSC2 binds to and decreases the levels of active GTP-bound Rheb in vivo, suggesting a functional role for Rheb in opposing the TSC1-TSC2 signaling pathway. Consistent with this notion, we have demonstrated that Rheb induces mTOR-dependent S6K1 activation and S6 phosphorylation. Importantly, this Rheb activity can be blocked by FTI treatment, suggesting that Rheb may be the missing target of FTIs in cancer treatment.
FIG. 6. Rheb induces phosphorylation of S6 in an mTOR-and farnesylation-dependent manner. A, 293T cells expressing empty vector or wild type Rheb were treated for 1 h with rapamycin (20 nM), MEK inhibitor (U0126, 10 M), or carrier (Me 2 SO). The effect of drugs on S6 phosphorylation was determined as described in Fig. 4. GAPDH, glyceraldehyde-3-P dehydrogenase. B, the importance of farnesylation on Rheb-induced S6 phosphorylation was evaluated using the farnesyl transferase inhibitor FTI 277. Two hours post-transfection, 293T cells were cultured in medium containing carrier or FTI 277 (10 M) until preparation of cell lysates and S6 phosphorylation, determined as above. The lower panel shows HA-Rheb expression. Note the absence of prenylated Rheb in cells treated with FTI 277.