TOR Signaling

See allHide authors and affiliations

Science's STKE  09 Dec 2003:
Vol. 2003, Issue 212, pp. re15
DOI: 10.1126/stke.2122003re15


The mammalian target of rapamycin, mTOR, is a protein Ser-Thr kinase that functions as a central element in a signaling pathway involved in the control of cell growth and proliferation. The activity of mTOR is controlled not only by amino acids, but also by hormones and growth factors that activate the protein kinase Akt. The signaling pathway downstream of Akt leading to mTOR involves the protein products of the genes mutated in tuberous sclerosis, TSC1 and TSC2, and the small guanosine triphosphatase, Rheb. In cells, mTOR is found in a complex with two other proteins, raptor and mLST8. In this review, we describe recent progress in understanding the control of the mTOR signaling pathway and the role of mTOR-interacting proteins.


TOR (target of rapamycin) is a central element in a signaling pathway involved in the control of cell growth and proliferation. In all organisms expressing the protein, TOR function is controlled by nutrient availability, which ensures that protein synthesis is repressed when the supply of precursor amino acids is insufficient. In simple organisms, nutrient availability appears to be the major factor influencing TOR activity. In metazoans, TOR integrates signals arising from the amino acid supply, cellular energy state, and receptors for various hormones and growth factors. The mechanisms through which TOR signals and how the activity of TOR is controlled have been mysterious. Within the past year, a spate of reports from several laboratories has dramatically advanced our understanding of the TOR signaling pathway. In this review, we describe some of the important earlier discoveries in the TOR field and the recent discoveries of TOR-associated proteins and effectors.

Discovery of TOR Proteins

Rapamycin is a potent antifungal agent originally isolated from a strain of Streptomyces hygroscopicus derived from a soil sample collected in Rapa Nui (Easter Island) (1). Rapamycin inhibits functions of TOR proteins in yeast and humans, as well as other organisms. In addition to being a valuable experimental tool for investigating TOR, rapamycin is used clinically to inhibit host rejection of transplanted organs, the growth of tumor cells, and the occlusion of coronary arteries after angioplasty (2-4).

The antifungal action of rapamycin provided the basis of selection in genetic screens that led to the identification of both the intracellular rapamycin receptor and the TOR proteins (5-8). The most frequently observed mutations enabling Saccharomyces cerevisiae to grow in the presence of rapamycin were in FPR1, which encodes the yeast equivalent (FPR1) of a ubiquitously expressed peptidyl-prolyl cis-trans isomerase, FKBP12 (FK506 binding protein, Mr = 12,000) (6, 7). Deleting FPR1 conferred complete resistance to rapamycin, confirming that FPR1 was essential for the action of the drug. However, yeast lacking FPR1 were viable, whereas wild-type yeast treated with rapamycin died, proving that loss of the prolyl isomerase function of FPR1 did not account for the toxic effects of rapamycin.

Mutations in two other genes, designated TOR1 and TOR2, also allowed growth in rapamycin (6). Deleting TOR1 decreased the rate of proliferation but was nonlethal (8, 9). Yeast lacking TOR2 were not viable; however, unlike rapamycin-treated cells, which were arrested specifically in G1, TOR2-deficient yeast was arrested randomly in the cell cycle (8, 9). Thus, even though the two proteins were homologous (~70% identical), it was clear that their functions were not entirely redundant. When both TOR1 and TOR2 were deleted, the cells were arrested in G1 as large, unbudded cells. The G1 block was reversed by reintroducing either TOR1 or TOR2, although the lethality resulting from a TOR2 knockout was not rescued by TOR1 (9). These results indicated that the yeast TOR proteins have overlapping roles in cell cycle progression, and that TOR2 has a function essential for viability that is not shared with TOR1. These elegant studies in yeast established TOR proteins as the targets of rapamycin. Subsequently, TOR2 was found to have effects on cytoskeletal organization that are not shared with TOR1 or inhibited by rapamycin (10, 11).

The dependence on FPR1 for the inhibitory effects of rapamycin was explained by the discovery that TOR proteins bound with high affinity to the rapamycin-FPR1 complex, but not to either rapamycin or FPR1 alone (6). FK506 is structurally similar to rapamycin and competes with rapamycin for binding to FKBP12 (FPR1 in yeast). However, the FK506-FKBP12 complex does not bind TOR. Thus, FK506 acts as a competitive inhibitor of the effects of rapamycin that result from inhibition of TOR. The mechanism of TOR inhibition by rapamycin is conserved from yeasts to humans (Fig. 1) (1).

Fig. 1.

The immunosuppressant macrolides, rapamycin and FK506. Both rapamycin and FK506 bind to the prolyl-isomerase, FKBP12. The rapamycin-FKBP12 complex binds and inhibits mTOR, but not PP2B, whereas the FK506-FKBP12 complex binds and inhibits PP2B, but not mTOR.

Domain Structure of TOR Proteins

In metazoans, a single TOR gene is the rule rather than the exception. The mammalian TOR protein, mTOR (also known as FRAP, RAFT1, or SEPT), was independently identified by four groups (12-15). The human mTOR gene encodes a protein of 2549 amino acids with 42% and 45% sequence identity to yeast TOR1 and TOR2, respectively. Except for the putative regulatory domain (RD), the general domain structure of mTor is similar to that of yeast TOR proteins (Fig. 2).

Fig. 2.

Domains in mTOR. The 2549-amino acid mTOR protein is depicted above a scale indicating amino acid residue number. The relative sizes and positions of various domains discussed in the text are shown.

The C-terminal region of mTOR contains a catalytic domain with sequence similarity to that found in phosphatidylinositol 3-kinase (PI3K). However, mTOR functions as a protein kinase (16, 17). Indeed, it is a founding member of a family of Ser-Thr proteins kinases that are more similar to the lipid kinase PI3K than to members of the larger family of Ser-Thr or Tyr protein kinases. The PIKK (PI3K-related kinase) family also includes DNA-dependent protein kinase, ATM, ATR, and SMG-1 (18). TOR proteins harboring mutations in residues required for kinase activity are functionally inactive. For example, neither TOR1 nor TOR2 bearing such mutations is capable of restoring growth of S. cerevisiae having disrupting mutations in TOR1 or TOR2 (10, 11, 19).

The binding site for rapamycin-FKBP12 is located just upstream of the catalytic domain in a region referred to as the FRB (FKBP12-rapamycin binding domain) (5, 9). Except for mTOR, none of the PIKK family binds rapamycin-FKBP12 with high affinity, although a region in SMG-1 does have limited sequence similarity to the FRB (20). The FRB is the only region of TOR for which definitive structural information exists. In the crystal structure of a TOR fragment containing the FRB complexed with rapamycin-FKBP12 solved by Choi et al. (21), there are extensive interactions between the FRB and rapamycin, but fewer interactions between the FRB and FKBP12 (21). Thus, FKBP12 appears to present rapamycin to TOR in a conformation favorable for interaction with the FRB. The original mutation in TOR2 allowing yeast to grow in rapamycin changed a highly conserved Ser in the FRB to an Arg. This Ser (position 2035 in mTOR) is located within the relatively hydrophobic rapamycin-binding pocket. Replacing this Ser with any residue larger than Ala blocks binding of FKBP12-rapamycin, thus generating a rapamycin-resistant form of TOR (22, 23).

Just upstream of the FRB is a region referred to as the FAT domain (for FRAP, ATM, TRAPP2), which is also found in other PIKKs, although the sequences diverge much more than those in the FATC (for FAT, C-terminal) domain (24). It has been suggested that because the FATC and FAT domains occur together in the PIKKs, the two domains are needed for proper folding or organization of the kinase domain, perhaps by interacting with each other or by mediating other necessary protein interactions. The FAT domain has also been termed the toxic-effector domain, because its overexpression in yeast leads to a G1 cell-cycle arrest (25). This effect was hypothesized to be due to sequestration of necessary interacting proteins away from TOR1.

The extreme C-terminal regions of PIKKs are similar and form a domain designated FATC (24, 25). Deletion of even a single amino acid from the FATC domain, or the addition of an epitope tag to the C-terminus, effectively eliminates mTOR activity (26). Why mutations in this region have such a major effect on mTOR function is unknown. Between the kinase domain and FATC is a putative regulatory domain (RD) that is phosphorylated in response to insulin and growth factors (27-29).

The region of mTOR beginning just upstream of the FAT domain and extending to the NH2 terminus contains 20 HEAT [Huntington-elongation factor 1A-protein phosphatase 2A (PP2A) A subunit-TOR] repeats (30). Each ~39-amino acid repeat contains several conserved hydrophobic residues and three highly conserved positions occupied by Pro, Asp, and Arg. Based on the crystal structure of the A subunit of PP2A, which contains multiple HEAT repeats, each repeat forms a pair of antiparallel α helices that stack in parallel with other repeats to form an ordered array (31). The stacked helices form an extensive surface for potential protein-protein interactions, which is the presumed function of the HEAT motifs in TOR proteins. The subcellular localization of the yeast TOR protein to membrane-associated clusters is dependent on intact HEAT repeats (32). One of the HEAT motifs in mTOR interacts with gephyrin, a protein necessary for glycine receptor clustering in neurons, although the functional significance of this interaction is not known (33).

Subcellular Localization of TOR

Relatively few studies have addressed the issue of the subcellular distribution of TOR proteins. Only recently was the intracellular location of TOR in yeast described (32, 34). By immunofluorescence studies, TOR1 and TOR2 appeared peripherally associated with the plasma membrane, as well as intracellular vesicular structures (32). Using the higher-resolution technique of immunogold electron microscopy, the TOR proteins were observed in punctuate clusters in regions adjacent to both the plasma membrane and membranous structures within the cell (34). Thus, TOR appears to be associated with membranes in yeast. In contrast, there are huge discrepancies in the reported locations of mTOR in mammalian cells. An initial study using subcellular fractionation of 3T3-L1 adipocytes indicated that most mTOR was associated with the microsomal fraction (35), a finding that is partially consistent with the membrane association of the yeast TOR. A study of several human cell lines placed mTOR at the periphery of the endoplasmic reticulum (ER) and Golgi (36). These findings (35, 36) are partially consistent with the membrane association of the yeast TOR (32, 34). In other experiments in which both subcellular fractionation and immunocytochemical methods were applied with 3T3 fibroblasts and Jurkat cells, some cytosolic mTOR was detected, but most of the protein was associated with mitochondria (37). A markedly different picture emerged when confocal immunofluorescence microscopy was used to localize mTOR in murine myoblasts, human fibroblasts, and several malignant cell lines (38). Most of the mTOR in these cells appeared to be nuclear (38). If it is assumed that the localizations reported in the different studies are correct, then the distribution of mTOR must change dramatically in response to unknown or uncontrolled experimental variables. Additional studies are clearly needed to resolve this issue.

TOR Proteins in Normal Growth and Development

Experimentally depleting TOR proteins in yeast, worms, and fruit flies has provided insight into the function of TOR. Disrupting both the TOR1 and TOR2 genes in S. cerevisiae leads to phenotypic changes resembling rapamycin-treatment of yeast (6, 8, 9). In addition to G1 arrest, the cells exhibit reduced protein synthesis, increased autophagy, decreased amino acid import, and characteristic changes in transcription, resulting from increased activities of the transcription factors GLN3, MSN2, and MSN4 (39, 40). A TOR2-specific mutation in S. cerevisiae disrupts the normal architecture of actin distribution, an effect that can be rescued by components of the RHO1 guanosine triphosphatase (GTPase) switch (41). The effects of disrupting TOR genes in yeast have been reviewed in detail elsewhere (40, 42).

Drosophila melanogaster larvae deficient in TOR fail to pupate and are arrested at a stage similar to that of amino acid-deprived larvae (43, 44). Reducing TOR in Drosophila also causes a reduction in the size of the nucleolus and lipid vesicle aggregation in the larval fat body, and decreased TOR promotes cell-cycle arrest in endoreplicative cell types (43). In Caenorhabditis elegans, deleting TOR leads to developmental arrest and a pronounced intestinal phenotype, characterized by an increase in the gut lumen size and a decreased ability of the intestine to digest and absorb nutrients; both effects apparently due to an inhibition of global protein synthesis (45).

mTOR is ubiquitously expressed in mammalian tissues, with highest levels of mRNA in skeletal muscle and brain (46). mTOR knockout animals have not been described and are unlikely to be viable. Mice harboring a nonfunctional allele of mTOR were identified after ethylnitrosourea-induced mutagenesis. This allele contains an intronic mutation that leads to inappropriate splicing, which then generates messages encoding two protein products--one corresponding to mTOR truncated at amino acid 385, the other to mTOR having a three-amino acid insertion beginning at amino acid 172. Mice homozygous for the mutant allele are termed flat-top because of defects in forebrain development (47). Signaling to targets downstream of mTOR is disrupted in fibroblasts derived from flat-top embryos, presumably because of the reduction in mTOR activity. Although the mice die around embryonic day 9, embryos are smaller than wild-type embryos at the same stage, consistent with the role of mTOR in controlling growth.

Downstream Effectors and Targets of TOR

Identifying processes in cells affected by rapamycin has been a valuable strategy for identifying downstream targets of TOR proteins. Some proteins identified in this manner include CLIP-170 (cytoplasmic linker protein-170) (48, 49), eukaryotic elongation factor 2 (eEF2) kinase (50), glycogen synthase (51, 52), hypoxia-inducible factor 1α (HIF-1α) (53, 54), lipin (55), eukaryotic initiation factor 4E (eIF4E)-binding protein (4EBP1, also known as PHAS-I) (56, 57), protein kinase C (PKC)δ and PKCϵ (58), PP2A (59, 60), p21 and p27 cyclin-dependent kinase inhibitors (61, 62), Rb (63), ribosomal S6 kinase 1 (S6K-1, also known as p70s6k) (64, 65), and STAT3 (66). The best characterized of these targets are the translational regulators, S6K-1 and 4EBP1 (Fig. 3).

Fig. 3.

Phosphorylation sites and domains in the mTOR targets, 4EBP1 and S6K-1. The kinase domain in S6K-1, the eIF4E binding domain (4EBD) in 4EBP1, and the phosphorylation sites in the two proteins are shown. Residues adjacent to the phosphorylated Ser or Thr residues are also indicated. The RAIP motif in 4EBP1 and the TOS motifs in 4EBP1 and S6K-1 (FEMDI and FDIDL, respectively) are also depicted.


S6K-1 is the major ribosomal protein S6 (rpS6) kinase in mammalian cells (67, 68). Rapamycin inhibits activation of S6K-1 in response to all known stimuli, and the drug reduces rpS6 phosphorylation in multiple cell types [for example, see (64, 65)]. Overexpressing rapamycin-resistant mTOR allows activation of S6K-1 in cells incubated with rapamycin, which inhibits the endogenous wild-type mTOR (69). Depletion of mTOR with short-interfering RNA (siRNA) inhibits activation of S6K-1 (46). Although less well characterized, activity of a related kinase, S6K-2, is also activated by mTOR (70). S6K is unquestionably a key element of TOR signaling. S6K-deficient fruit flies are smaller than wild-type flies and contain smaller cells (71). Moreover, overexpressing S6K rescued the TOR-knockout phenotype, which suggests that the loss of signaling to S6K accounts for the cell size phenotype of the TOR-deficient fruit flies (43). S6K-1-knockout mice are about 20% smaller than their wild-type littermates (72). Reducing mTOR activity by treating mice with rapamycin markedly inhibits growth of the animals (47, 73). In view of the small size phenotype of S6K-1-deficient mice, the effect of rapamycin on growth likely involves inhibition of the activation of S6K-1.

S6K-1 is activated by multisite phosphorylation (Fig. 3). Thr229, which is located in the activation loop, must be phosphorylated for activation to occur (74). The activation loop site is conserved in several other kinases, including Akt, where its phosphorylation is also essential for kinase activity (75). Thr229 is phosphorylated by the "loop kinase" phospholipid-dependent protein kinase 1 (PDK-1) (76). Thr389 and Ser371, two sites that can be phosphorylated by mTOR in vitro (17, 77), must also be phosphorylated for S6K-1 activation (77, 78). These two sites, as well as Ser404, fit a motif in which the phosphorylated Ser or Thr is flanked by hydrophobic residues. Phosphorylation of sites in the C-terminal region of S6K-1, known as the SKAIPS (p70 S6 kinase autoinhibitory pseudosubstrate) domain, has a less critical role in activation of the kinase (67, 68). The SKAIPS domain (amino acids 400 to 436) has limited sequence identity (~28%) to the phosphorylated region of rpS6, and it has been proposed to serve as an autoinhibitory domain (79). Phosphorylation of the SKAIPS domain appears to facilitate the phosphorylation of the other sites (67, 68). Ser411, Ser418, Thr421, and Ser424 in the SKAIPS domain conform to a (Ser or Thr)-Pro [(Ser-Thr)Pro] motif, as does Ser371. The phosphorylation of the three hydrophobic sites (Thr389, Thr229, and Ser404), as well as two of the (Ser-Thr)Pro sites (Ser371 and Ser411), is inhibited by rapamycin (78, 80).

Five residues (Phe-Asp-Ile-Asp-Ile) near the N-terminus of S6K-1 are necessary for efficient activation of the kinase in cells (81). This amino acid sequence, which consists of an aromatic residue followed by alternating hydrophobic and acidic residues, forms the TOS (TOR signaling) motif. Deletion or mutation of the TOS motif markedly decreased the phosphorylation of Thr389 and Thr229 in response to insulin (81). Earlier studies had shown that deleting residues from the N-and C-termini of S6K-1 resulted in a form of S6K-1 that was not inhibited by rapamycin (82, 83).

S6K-1 has been proposed to control cell size by increasing mRNA translation, particularly of those messages containing the tract of pyrimidines (TOP) motif (84). The TOP motif, which is located immediately downstream from the m7G cap, consists of a cytosine followed by 4 to 14 pyrimidines. All known ribosomal proteins and several abundant elongation factors are encoded by TOP messages (84). Hence, increasing translation of TOP messages not only increases protein content directly, but also increases the translational capacity of cells. It has been known for many years that insulin, which activates S6K-1, preferentially increases the synthesis of ribosomal proteins (85).

In normal cells, S6K-1 activity correlates well with translation of TOP messages. Thus, insulin, as well as serum and amino acids that activate S6K-1, shift TOP mRNAs into polysomes (86, 87). Rapamycin partially inhibits translation of TOP mRNAs (86, 87). More directly implicating S6K-1 in TOP translation was the finding that overexpression of a Ser229 to Ala mutant of S6K, which acts as a dominant-negative, decreased TOP mRNA in polysomes (88). In view of the evidence implicating S6K-1, it came as a surprise that TOP messages were found to be controlled normally in response to mitogenic stimulation or amino acid starvation in embryonic stem cells lacking S6K-1 (89, 90). In addition, the major known target of S6K-1, rpS6, was not phosphorylated in S6K-1-deficient cells (89, 90). Despite demonstrating that neither S6K-1 activation nor rpS6 phosphorylation is required for translation of TOP messages, these findings do not prove that S6K-1 is not involved, because redundant control mechanisms may exist. Indeed, as ribosome biogenesis may account for up to 80% of the metabolic output in actively growing cells (91), multiple mechanisms would be expected to be involved in controlling translation of TOP messages.


4EBP1 was first identified as an adipocyte protein that underwent increased phosphorylation in response to insulin (92, 93). In unstimulated cells, the protein functions as a translational repressor, and inhibition of its activity contributes to the effects of mTOR on cap-dependent mRNA translation (57, 94). Nonphosphorylated 4EBP1 binds tightly to eIF4E, the mRNA cap-binding protein, and it represses cap-dependent translation by blocking the binding of eIF4E to eIF4G, apparently in a competitive manner (95). The eIF4E-binding motif in 4EBP1, Tyr-X-X-X-X-Leu-Φ (where X is any amino acid and Φ is a hydrophobic amino acid), is also found in the scaffolding protein, eIF4G (96). 4EBP1 does not inhibit binding of eIF4E to the cap, but instead appears to actually strengthen this interaction (97). Phosphorylation of 4EBP1 triggers dissociation of the 4EBP1 and eIF4E complex (92, 93), allowing eIF4E to engage eIF4G. eIF4G also binds an mRNA helicase (eIF4A), poly(A)-binding protein (Pab1), and eIF3, which links the eIF4G-associated factors to the small ribosomal subunit (98). Thus, 4EBP1 phosphorylation allows several important initiation factors, as well as the 40S ribosomal subunit, to be positioned at the 5′ end of the mRNA to begin the process of scanning.

The findings that rapamycin attenuated the phosphorylation of 4EBP1, as well as partially prevented dissociation of the 4EBP1-eIF4E complex in response to insulin and growth factors, first implicated mTOR in the control of the 4EBP1 (56, 57, 94). The conclusion that mTOR controlled 4EBP1 phosphorylation was solidified by the demonstration that overexpressing rapamycin-resistant mTOR increased 4EBP1 phosphorylation in cells incubated with rapamycin (16).

The following phosphorylation sites have been identified in 4EBP1: Thr37, Thr46, Ser65, Thr70, Ser83, Ser101, and Ser112 (residue numbers are for the human protein, subtract 1 for sites in the rat and mouse proteins) (99-101). As will be discussed later, several of these sites can be phosphorylated by mTOR in vitro. 4EBP1 lacks the hydrophobic class of phosphorylation sites found in S6K-1, and except for Ser112, which is followed by Gln, all of the sites in 4EBP1 conform to a (Ser-Thr)Pro motif. Phosphorylation of the sites in cells occurs in an ordered fashion, with that of Thr37 and Thr46 occurring first, followed by Thr70, and finally Ser65 (102, 103). Ser112 has been proposed to be a necessary priming site for the phosphorylation of other sites in 4EBP1 (99, 101). However, recent studies indicate that Ser112 can be mutated to Ala without affecting either the phosphorylation of other sites (99) or the release of 4EBP1from eIF4E (104). In vitro, phosphorylation of Ser65 has the most dramatic effect on decreasing the affinity of 4EBP1 for eIF4E (92), but replacing Ser65 with Ala had no effect on the amount of 4EBP1 bound to eIF4E in HEK293 cells, in either the absence or presence of insulin (102, 103). Phosphorylation of Thr46 also decreased the affinity of 4EBP1 for eIF4E in vitro (105, 106). Presumably, phosphorylation of some combination of the three (Ser-Thr)Pro sites leads to dissociation of 4EBP1 from eIF4E. The sites in 4EBP1 differ in sensitivity to rapamycin-treatment of cells: Phosphorylation of Thr70 and Ser65 is generally more reduced by rapamycin than is the phosphorylation of Thr37 and Thr46 (102, 103).

Two motifs in 4EBP1 are essential for the efficient phosphorylation in cells. The first is a TOS motif formed by the last five C-terminal amino acids, Phe-Glu-Met-Asp-Ile (81). The second motif was discovered by characterizing a hypophosphorylated 4EBP1 fragment that bound tightly to eIF4E in cells undergoing apoptosis (107). Amino acid sequencing revealed that the fragment was generated by caspase-cleavage of the Asp24-Gly25 bond in 4EBP1. This second critical region needed for efficient phosphorylation of 4EBP1 has the sequence Arg-Ala-Ile-Pro, and is called the RAIP motif (107).

α4 and PP2A

In S. cerevisiae TOR controls the function of the protein phosphatases, SIT4, PPH21, and PPH22 (collectively referred to as PPH21/22) (108). In the presence of adequate nutrients TOR stimulates the association of these enzymes with the regulatory subunit, TAP42 (108). Rapamycin promotes dissociation of TAP42, thereby increasing phosphatase activity. Mutations in TAP42 provide partial resistance to rapamycin (108), which is additional evidence that TAP42 has an important role in TOR signaling.

Two models have been put forth to explain how TOR promotes the association of TAP42 with the phosphatases. In one model, phosphorylation of TAP42 is proposed to increase binding to SIT4 and PPH21/22 (109). TOR has been reported to phosphorylate TAP42 directly (109), although this finding has not been confirmed. In the other model, the association of TAP42 with SIT4 is determined by the phosphorylation state of the protein, TIP41 (110). Nonphosphorylated TIP41 binds TAP42, blocking binding of TAP42 to SIT4. Through an undefined mechanism, TOR promotes the phosphorylation of TIP41, which then triggers dissociation of the TIP41-TAP42 complex, allowing TAP42 to engage and inhibit SIT4 (110).

TAP42 and PPH21/22 are homologs of the mammalian proteins, α4 and the PP2A catalytic subunit (C), respectively. A mammalian homolog of TIP41 appears to exist (for example, accession no. AAH09506), but its function has not been assessed. A fraction of PP2A C is found in a heterodimeric complex with α4 (111), although most C is complexed with regulatory A subunits as AC dimers or heterotrimers of A, C, and a third subunit (B, B′, or B′′) (112). α4 also associates with PP4 and PP6, two protein phosphatases that are less well characterized than PP2A (113). Recombinant α4 was found to inhibit PP2A activity markedly in vitro (114), which suggested that α4 may function in a manner similar to TAP42. However, rapamycin does not cause dissociation of the α4-PP2A complex (113). Moreover, there is evidence that instead of acting as an inhibitory subunit, α4 may function to change the substrate specificity of the catalytic subunit (111). Thus, there are clearly some differences between the yeast and mammalian systems.

Rapamycin has been reported to activate PP2A in Jurkat cells (60) and to prevent the inactivation of PP2A by insulin in rat skeletal muscle cells (59). Whether these effects of rapamycin involve modification of the α4 or TIP41-like proteins is not known, but the effects of rapamycin suggest that mTOR controls PP2A activity. Given the relatively broad substrate selectivity of PP2A (112), the control of this enzyme could link mTOR to a myriad of cellular processes, including other signaling pathways. Alternatively, targeting subunits may limit PP2A activity to specific proteins.

Protein expression controlled by TOR

Application of gene chip technology has demonstrated that the abundance of many mRNAs is changed in response to rapamycin or TOR depletion in yeast (115) and mammalian cells (116). A large-scale analysis of genetic interactions between TOR and yeast genes has also been described (117). Proteomic analyses using mass spectrometric sequencing identified many proteins whose levels were changed in response to rapamycin (118). Interpreting the data and investigating the leads generated by these screens represents a challenging endeavor that should lead to a clearer understanding of the control of gene expression by mTOR.

mTOR as a Ser-Thr Protein Kinase

Despite its sequence similarity to PI3K, there is no direct evidence that TOR phosphorylates lipids. The idea that mTOR was a protein kinase was kindled when cDNA cloning revealed that the catalytic subunit of DNA-dependent protein kinase, a bona fide Ser-Thr kinase, also shared sequence similarity with PI3K (119). Shortly thereafter, mTOR, which had been immunoprecipitated from extracts of cells or tissues was shown to autophosphorylate (69, 120) and to phosphorylate recombinant 4EBP1 and S6K-1 (16, 17). It has not been possible to produce homogeneous recombinant mTOR, and the reliance on immune complex assays to assess mTOR activity has hindered investigations of the kinase. However, mTOR harboring a point mutation changing Asp2338 to Ala exhibited no kinase activity, which confirmed that the activity measured in immune complexes was due to mTOR (16, 69). This Asp residue is conserved in both protein and lipid kinases, where it is required for kinase activity (121). mTOR autophosphorylation, which occurs on Ser2481, is also blocked by mutating Asp2338 (122).

Incubating cells with serum (17), insulin (27), or certain growth factors (29, 66) increases the protein kinase activity measured in immune complexes containing mTOR; however, such changes in mTOR activity have not been detected in other studies. Thus, although there is general agreement that insulin and growth factors increase mTOR function, there are still questions as to whether these agents produce stable changes in the protein kinase activity of mTOR.

There are some unusual aspects to the mTOR-mediated phosphotransferase reaction. The Km of mTOR for adenosine triphosphate (ATP) (~1 mM) is high relative to that of most other protein kinases (<50 μM) (123). Most investigators have observed that MnATP is preferred over MgATP (16, 35, 124, 125), a preference also noted with certain other PIKKs, such as ATM (126, 127). The phosphorylation of full-length forms of 4EBP1 and S6K-1 by mTOR is inhibited by the nonionic detergents, Triton X-100 and NP-40 (46, 124, 128), which promote dissociation of raptor, an important regulator of mTOR function [see below; (46, 128, 129)].

In vitro, Thr389 is the preferred site in S6K-1 phosphorylated by mTOR (17), but Ser371, Thr421, Ser424, and possibly Ser411 are also phosphorylated (77, 130). In vitro, mTOR phosphorylates Thr37, Thr46, Thr70, and Ser65 in 4EBP1 (100, 131, 132). Phosphorylation of Thr70 and Ser65 occurs more slowly than that of the other two sites, but it is readily detected after incubating mTOR with an activating antibody (133) or with physiological concentrations of ATP (123). As in intact cells, the 4EBP1 sites phosphorylated by mTOR in vitro differ in sensitivity to rapamycin. In the presence of FKBP12, rapamycin essentially abolishes phosphorylation of Ser65 and Thr70 by mTOR, but only partially inhibits phosphorylation of Thr37 and Thr46 (103, 132). The doses of rapamycin that inhibit mTOR in vitro are about 10 times those required to inhibit phosphorylation of the mTOR targets in cells (23). This discrepancy suggests that the inhibitory effect of rapamycin in cells may not be directly linked to inhibition of the kinase function of mTOR. Alternatively, mTOR might associate with proteins or other factors in cells that enhance rapamycin sensitivity.

In contrast to rapamycin, the two PI3K inhibitors, wortmannin and LY294002, completely abolish mTOR activity (23, 120). By inference to the x-ray crystallographic studies of PI3Kγ (134), wortmannin and LY294002 most likely bind in the active site of mTOR, which would be consistent with the complete inhibition of kinase activity produced by these agents. Rapamycin is clearly not an active site inhibitor, because the rapamycin-FKBP12 binding site is well removed from the catalytic domain.

The inhibition of mTOR complicates the use of PI3K inhibitors in implicating PI3K in a response. There is selectivity with wortmannin, as the concentration of this drug that is needed to inhibit mTOR is about 10 times that required to inhibit PI3K activity (120); however, the dose responses for inhibition of PI3K and mTOR by LY294002 are almost identical (23, 120). The incomplete inhibition of mTOR by rapamycin also raises a cautionary note, because there may be mTOR-mediated responses in cells that are not blocked by rapamycin.

Methylxanthines also block the phosphorylation of 4EBP1 and S6K-1 in response to insulin (56, 135). These agents inhibit adenosine 3′,5′-monophosphate (cAMP) phosphodiesterase and may increase intracellular cAMP, which is associated with attenuation of mTOR signaling in some cell types (56, 135). However, the inhibition of mTOR by methylxanthines may be direct. Caffeine, theophylline, and isobutylmethylxanthine directly inhibit mTOR activity in vitro (23, 136). The site of action is probably the catalytic domain, because the agents also inhibit other members of the PIKK family (137).

The FRB is essential for the kinase activity of mTOR (138). Note that a fragment of mTOR containing the FRB was found to bind tightly to phosphatidic acid, which is generated in cells through the action of phospholipase D (139). Incubating cells with n-butanol, which competes with water for the hydrolysis of phospholipids by phospholipase D, decreases cellular phosphatidic acid and attenuates signaling by mTOR (139). mTOR harboring point mutations in the FRB that decreased the affinity for phosphatidic acid exhibited reduced kinase activity in vitro (139). In view of these intriguing results, additional studies are needed to investigate the role of phospholipids in the control of mTOR activity.

Mutations in the FRB may influence substrate selectivity (23). Mutating Ser2045 to Ile, which renders mTOR resistant to inhibition by FKBP12-rapamycin, markedly decreased phosphorylation of Thr37 and Thr46 in 4EBP1 by mTOR, but had little effect on the phosphorylation of Thr70 (23). These findings indicate that mutations used to produce rapamycin-resistant mTOR are not silent with respect to mTOR function. Findings with rapamycin-resistant forms of TOR protein in Schizosaccharomyces pombe (140) and mammalian cells (53) reinforce this point.

The distinct character of the hydrophobic motif, represented by Thr389, and the (Ser-Thr)Pro motif of the sites in 4EBP1 indicate that determinants for phosphorylation by mTOR are more complicated than the sequence of amino acids in the phosphorylation site. In this respect, the RAIP, TOS, or both motifs appear to be important, because disrupting these motifs markedly decreased the rate of phosphorylation of S6K-1 and 4EBP1 by mTOR in vitro (81, 129, 141, 142).

Control of TOR Function by Nutrient-Sensing Pathways

In S. cerevisiae, inhibiting TOR with rapamycin mimics nitrogen starvation, which produces a characteristic response that includes G1 arrest, reduced protein synthesis, and increased autophagy (143). Supplying yeast with a rich nitrogen source promptly reverses these effects. Levels of glutamine in particular appear to link TOR to nitrogen availability (144), although mechanisms involved are still not fully defined.

TOR was recently implicated in the control of GCN2, the amino acid sensor in the general control response system in yeast. GCN2 is an eIF2α kinase that contains a regulatory domain with sequence similarity to histidyl-tRNA synthetase (145). Uncharged transfer RNAs (tRNAs), which are abundant during amino acid starvation, activate GCN2 by binding to this domain. The resulting phosphorylation of eIF2α represses translation initiation, paradoxically increasing the translation of GCN4 through a mechanism involving upstream open reading frame (uORF), bypassing and translation reinitiation (145). GCN4 is a transcription factor that promotes transcription of genes encoding various enzymes and proteins needed to adapt to starvation conditions. GCN2 is phosphorylated in Ser577 by another kinase, which inhibits GCN2 activation by uncharged tRNA (146). Inhibiting TOR with rapamycin promotes the dephosphorylation of Ser577 through a mechanism involving TAP42. The finding that TOR influences the activity of GCN2 raises the intriguing question of whether GCN2 might influence TOR activity.

In mammalian cells, mTOR activity is controlled by amino acid availability, although it is not clear that mTOR senses the same signals as yeast TOR. In mammals, the primary indicator of amino acid supply is leucine instead of glutamine (147, 148). Abundant leucine increases mTOR activity, both in animals and in various cell types in culture (147, 149, 150). The effects of leucine do involve changes in the activities of either PI3K or AKT, and leucine's stimulation of mTOR is not blocked by wortmannin at concentrations that selectively inhibit PI3K (147-150). Thus, leucine does not use upstream signaling elements in the insulin or growth factor pathways. Just how leucine increases mTOR activity has not been established, although there is evidence implicating the mTOR-associated protein raptor [see below; (46)] and changes in mitochondrial metabolism (150).

Metabolic Signaling to mTOR

It has been suggested that its high Km for ATP allows mTOR to respond as a homeostatic sensor of changes in intracellular ATP (123). However, cellular ATP is maintained within a relatively narrow concentration range in cells. Under most circumstances, the concentration of ATP remains well above the reported Km (~1 mM) of mTOR, so that small changes in ATP levels would not significantly affect mTOR kinase activity. In contrast, the concentration of adenosine monophosphate (AMP) increases markedly in response to increased metabolic demand (151). Thus, AMP is a much more sensitive indicator of the energy status of the cell than ATP. Increasing AMP activates energy producing processes, such as glycogenolysis and glucose transport, but inhibits processes that consume energy, such as protein synthesis (151). An important mediator in many of these responses is AMP-dependent protein kinase (AMPK). This kinase is activated by AMP, as well as by the AMP homolog, 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR), which has proven useful for investigating the effects of activating AMPK in cells and tissues (151). Injecting rats with AICAR promoted the dephosphorylation of 4EBP1 in skeletal muscle and decreased the amount of eIF4E bound to eIF4G (152). AICAR blunts phosphorylation of S6K-1 in multiple types of cultured cells (153, 154), as do other treatments that increase AMP concentrations in cells, such as glucose deprivation or incubation with 2-deoxyglucose or metabolic inhibitors (123, 153). AICAR was without effect on the rapamycin-resistant form of S6K-1, supporting the conclusion that the effect of AMPK on S6K-1 resulted from inhibition of mTOR function (153). AICAR was also reported to inhibit the phosphorylation Ser2448 in mTOR (152), but this effect has not been reproduced in other systems.

It is intriguing that a fragment of the AMPK γ1 subunit was identified as an mTOR-interacting partner in a yeast two-hybrid screen (153). Very recent evidence suggests that AMPK suppresses mTOR function by increasing the activity of the TSC-TSC2 complex (described later) (154a).

Other potential mechanisms coupling mTOR activity to metabolism have been described (37, 155). The activity of mTOR was reported to be stimulated several-fold by polyphosphate, a polymer linked by high-energy phosphate bonds (155). Ectopic expression of a yeast polyphosphatase, PPX1, blunted the stimulatory effects of insulin and amino acids on the phosphorylation of4EBP1 in MCF-7 mammary cells (155). It is a fascinating possibility that polyphosphate, a primitive indicator of cellular energy state found in all cells (156), might control TOR function. There is also evidence that the association of mTOR with mitochondria allows mTOR to sense changes in mitochondrial function (37).

Control of mTOR by Insulin and Growth Factors

Sufficient information to formulate a working hypothesis for insulin signaling to mTOR now exists. The initial steps in the pathway in Fig. 4 are unique for insulin, but steps downstream of PI3K are shared by other hormones and growth factors that activate the lipid kinase.

Fig. 4.

A schematic of insulin activation of a signaling pathway to mTOR. Insulin binding activates the insulin receptor (IR) tyrosine kinase, which phosphorylates IRS-1 or IRS-2 (IRS). PI3K binds phosphorylated IRS by Src-homology 2 (SH2) domains in the p85 regulatory subunit. This interaction activates the p110 catalytic subunit, which phosphorylates phosphatidylinositol, generating products that recruit Akt to the membrane, where it is phosphorylated and activated by phospholipid-dependent kinase 1 (PDK-1). Akt phosphorylates mTOR directly. Akt also phosphorylates TSC2 and decreases the Rheb GAP function of the TSC1-TSC2 complex through an undefined mechanism. As a result, the GTP-bound form of Rheb accumulates, leading to activation of mTOR. A Rheb GEF is shown, although this enzyme has not been identified. Also depicted are the mTOR-associated proteins, raptor and mLST8, and the downstream targets of mTOR, 4EBP1 and S6K-1. Amino acids activate the mTOR pathway through an unknown mechanism that might involve dissociation of raptor from mTOR.

IRS-1 and PI3K

The activated insulin receptor phosphorylates members of a family of insulin receptor substrate (IRS) proteins (157). Studies in myeloid progenitor cells place IRS-1, one of four IRS proteins, in the pathway leading to 4EBP1 phosphorylation (158). These cells lack endogenous IRS-1, and insulin is unable to stimulate the phosphorylation of 4EBP1, even after overexpression of the insulin receptor. However, coexpression of the insulin receptor and IRS-1 resulted in insulin-stimulated 4EBP1 phosphorylation. Phosphorylation of IRS-1 in Ser-Thr sites that limit tyrosyl phosphorylation (157), trigger degradation by the proteasome (159), or some combination of these effects leads to an attenuation of insulin signaling. Rapamycin blocks Ser-Thr phosphorylation of IRS-1 (159-162) in several sites including Ser312 (163), which suggests that a feedback loop exists between mTOR and IRS-1; however, there is not complete agreement that mTOR controls IRS-1 degradation (162).

Insulin stimulates the phosphorylation of 18 tyrosyl residues in IRS-1, and IRS-1 harboring Tyr to Phe mutations in all of these sites is functionally inactive (158). Restoring just the three sites involved in recruiting PI3K allowed IRS-1 to support insulin-stimulated phosphorylation of 4EBP1 in myeloid progenitor cells (158). In HEK293 cells, overexpressing the p110 catalytic subunit of PI3K increased the phosphorylation of both 4EBP1 and S6K-1 (164, 165). Inhibitors of PI3K block insulin-stimulated phosphorylation of S6K-1 and 4EBP1 in various cell types (120, 158, 166). The inhibition produced by wortmannin occurred at concentrations well below those needed to directly inhibit mTOR (120). Overexpressing dominant-negative forms of the p85 subunit of PI3K also inhibited insulin-stimulated phosphorylation of 4EBP1 and S6K-1 (164, 167), as does overexpressing the phosphatidylinositol 3 phosphatase, PTEN (phosphatase and tensin homolog on chromosome 10) (168). In contrast, inactivating mutations in PTEN increased basal phosphorylation of the two mTOR targets (169, 170).

The findings described above would certainly seem to place PI3K upstream of mTOR. However, the positioning of PI3K and mTOR has been debated for several years. At the heart of the issue is the finding that growth factor activation of N- and C-terminally modified S6K-1, which is not inhibited by rapamycin, is blocked by PI3K inhibitors (82, 83). Thus, it has been argued that mTOR could not be downstream of PI3K, and that mTOR and PI3K must be on parallel signaling pathways. A cogent counterargument is that mTOR is one of at least two PI3K-dependent inputs to S6K-1.


PI3K inhibitors block activation of the Ser-Thr kinase, Akt [also known as protein kinase B (PKB)], which is recruited to membranes and activated by the phospholipid products of the PI3K reaction (171). Attaching a myristoylation site to Akt results in a form of the kinase that is constitutively targeted to membranes and activated (171). Overexpressing membrane-targeted Akt increased phosphorylation of S6K-1 and 4EBP1 (172-174). Phosphorylation of both proteins was inhibited by overexpressing a kinase-dead form of Akt (172). Overexpressing Akt not constitutively targeted to the membrane increased 4EBP1 phosphorylation, but failed to influence S6K-1 phosphorylation in HEK293 cells (175). Dufner et al. (175) have argued that the increase in S6K-1 activity due to overexpression of membrane-targeted Akt is an artifact, because the resulting increase in Akt activity is much higher than that produced after activation of the endogenous Akt by insulin or growth factors. Thus, Akt activation alone may not be sufficient for activation of S6K-1.

It is not unreasonable to expect that the persistent increase in Akt activity produced by overexpression would produce phenotypic changes not representative of the acute and transient activation of Akt produced by insulin. This concern has been addressed with a conditionally active form of Akt (MER-AKT), which was generated by fusing AKT to a mutant estrogen receptor (173). When overexpressed, MER-AKT remains inactive until cells are incubated with tamoxifen, an estrogen antagonist. Tamoxifen binds MER-AKT and produces conformational changes in the estrogen receptor component of the fusion protein that allow activation of the kinase. In 3T3 fibroblasts stably expressing MER-AKT, tamoxifen activated the kinase with a time course comparable to that with which insulin or growth factors activate endogenous Akt (173). Tamoxifen also rapidly stimulated the phosphorylation of both S6K-1 and 4Ebp1 (173).

Results of genetic analyses of the control of cell size and number in Drosophila are consistent with the positioning of TOR downstream of Akt. In fruit flies, as in mammals, body size is determined both by the number of cells and the size of individual cells (176, 177). The PI3K pathway controls both cell size and number, as evidenced by the findings that down-regulating the insulin receptor, IRS, or PI3K (Dp110) results in smaller flies that have fewer and smaller cells (176). Conversely, up-regulating expression of either the insulin receptor or PI3K results in larger flies with more and larger cells (176, 177). Below the level of PI3K, the control of cell number and cell size diverges. Altering the expression of Akt changes cell size but not cell number (176). Down-regulating either TOR or S6K also selectively decreases cell size, consistent with the model placing TOR downstream of Akt (177).

There is biochemical evidence in mammalian cells connecting Akt to mTOR. Insulin stimulation of mTOR phosphorylation has been demonstrated both by 32P labeling in 3T3-L1 adipocytes (136) and with phosphorylation-sensitive antibodies in various cell types (28, 29, 178). Binding of the antibody, mTAb1, is inhibited by phosphorylation of its epitope, which is found in an 18-amino acid sequence (DTNAKGNKRSRTRTDSYS2450) near the C-terminus of mTOR (27). Incubating 3T3-L1 adipocytes with insulin nearly abolished binding of mTAb1 to mTOR, which suggests that this region may be stoichiometrically phosphorylated in cells. Tamoxifen increased the phosphorylation of mTOR in fibroblasts expressing MER-AKT, which implicated Akt in the control of mTOR phosphorylation (27). The mTAb1 epitope contains two sites, Thr2446 and Ser2448, that fit the following consensus for phosphorylation by Akt: Arg-X-Arg-X-X (Ser-Thr)ϕ, where X may be any amino acid, and ϕ is a hydrophobic residue (179). Results with phosphospecific antibodies to the Ser2448 site confirm that insulin stimulates phosphorylation of this site in both cultured cells (28, 29) and intact skeletal muscle (178, 180). The phosphorylation of Ser2448 in HEK293 cells was blocked by overexpressing kinase-dead Akt in cells, which further implicated Akt (29). Phosphorylation of Ser2448 by Akt in vitro has been detected with phosphospecific antibodies (28, 29), although the stoichiometry of phosphorylation achieved in these studies was not reported.

Binding of mTAb1 increased the kinase activity of mTOR activity several-fold (131), as did deleting the 18-amino-acid region in mTOR containing the mTab1 epitope (29). Thus, the mTAb1 epitope appears to form a part of an inhibitory regulatory domain (RD, Fig. 2) in mTOR. Although one might predict that phosphorylation of Ser2448 would activate mTOR; there is no direct evidence to support such an effect, and there is evidence that Ser2448 phosphorylation is not necessary for the activation of mTOR. When overexpressed in HEK293 cells, mTOR harboring Thr2446 and Ser2448 to Ala mutations was no less effective than overexpressed wild-type mTOR in supporting insulin-stimulated phosphorylation of S6K-1 or 4EBP1 (29). Moreover, S6K is activated in response to insulin/IGF-1 signaling in fruit flies (43), even though Drosophila TOR lacks both of the AKT phosphorylation sites.

TSC1 and TSC2

A possible mechanism for the activation of TOR by Akt has been pieced together from genetic and biochemical studies in Drosophila and mammalian cells. This story begins with the identification by positional cloning of the genes that are mutated in tuberous sclerosis, a human genetic disorder causing serious defects in multiple organ systems (181). Mutations in either of two genes, TSC1 (182) or TSC2 (183), may lead to the disorder. The protein products, TSC1 (also known as hamartin, Mr ~130,000) and TSC2 (also known as tuberin, Mr ~200,000) form a heterodimeric complex (TSC1-TSC2), linked by interactions between their respective N-terminal regions (184, 185). The TSC1-TSC2 complex is the functional unit, which explains why mutations in either protein can lead to tuberous sclerosis.

A hallmark of tuberous sclerosis is the development of hamartomas, typically benign tumors that often contain very large cells (186). TSC2 was directly implicated in the control of cell size by the discovery that the mutation in Drosophila responsible for the gigas phenotype occurred in the Tsc2 gene (187). Gigas fruit flies have large eyes, containing both larger bristles and cells that are two to three times larger in area than normal. The large-cell phenotype was suggestive of stimulation of the TOR pathway, but other mechanisms could not be excluded. Subsequent analysis indicated that TSC1 and TSC2 were negative regulators of the insulin and insulin-like growth factor (insulin/IGF) pathway (188). A link to mTOR was made by the findings that S6K-1 was more highly phosphorylated in human and mouse cells lacking functional TSC2 and that phosphorylation of S6K-1 in these cells could be decreased by rapamycin (189, 190). In other experiments, overexpressing the combination of TSC1 and TSC2 in HEK293 cells blocked the increase in phosphorylation of 4EBP1 and S6K-1 in response to amino acids or growth factor stimulation (191-193), whereas overexpression of the proteins individually was without effect. Thus, the TSC1-TSC2 complex suppresses mTOR signaling.

Experiments with cells from patients with tuberous sclerosis, and with mouse cells deficient for TSC1, indicated that the increased activity of S6K-1 was not associated with increased activities of PI3K or Akt (189, 190, 192). Overexpressing TSC1 and TSC2 inhibited the phosphorylation in Ser2448 in mTOR (193), but this did not appear to be due to suppression of Akt activity. A breakthrough in connecting Akt to TSC1-TSC2 was made by Manning et al. (194), who used mass spectrometry to identify cellular proteins that immunoprecipitated with a phosphospecific antibody designed to recognize sites phosphorylated by Akt (194). One of the proteins identified in this manner turned out to be TSC2. Experiments with purified proteins confirmed that Akt phosphorylated TSC2 in several sites, including Ser939, Ser1130, and Thr1462 in the C-terminal region of the protein (193-195). Overexpressing TSC2 mutants with Ala substitutions at Akt phosphorylation sites blocked activation of S6K-1 in response to growth factors (193, 194), which indicated that TSC2 was downstream of Akt in the pathway leading to S6K-1.

TSC2 may also mediate the inhibitory effects of AMPK on mTOR. Phosphorylation of Thr1227 and Ser1345 in TSC2 appears to enhance the suppressive effect of the TSC1-TSC2 complex on mTOR. Thus, findings with Akt and AMPK imply that phosphorylation may either increase or decrease the function of TSC1-TSC2, depending on the sites in TSC2 that are phosphorylated. (154a).

Overexpressed TOR was reported to co-immunoprecipitate with endogenous TSC2 in Drosophila Schneider 2 (S2) cells (196), but endogenous TOR has not been shown to interact with either of the endogenous TSC proteins. Thus, whether TSC1, TSC2, or the complex physically associates with TOR or mTOR is still unclear.

Phosphorylation of TSC2 (195) or substitution of the Akt-phosphorylation sites in TSC2 with Glu (193) has been reported to inhibit the association of TSC1 and TSC2. However, dissociation of the complex in response to Akt phosphorylation was not observed by others (194, 197). Phosphorylation has been reported to accelerate degradation of TSC1 and TSC2 (193, 197). This effect may allow for long-term up-regulation of TOR function, but the response time is too slow to explain the acute and rapidly reversible changes in phosphorylation of 4EBP1 and S6K-1 that occur in response to insulin and growth factors.

Another fly in the ointment (complication) is the finding in insect cells that the activation of S6K in response to insulin was totally resistant to wortmannin at a concentration that essentially abolished the stimulatory effects of insulin on the increase in phosphatidylinositol 3,4,5-trisphosphate (PIP3) levels (the product of increased PI3K activity) and Akt activity (198). Also, Akt overexpression partially reversed a wing-shape phenotype produced by overexpression of S6K, an effect opposite to that which would be expected if S6K were downstream of Akt (198). One interpretation is that, in Drosophila, Akt has nothing to do with the control of TOR. Such a conclusion would be counter to a large amount of evidence obtained from studies in mammalian cells. It may be that Akt signaling to mTOR emerged more recently in evolution than the signaling events leading to activation of Drosophila TOR, a scenario that would also be consistent with the absence of the Akt phosphorylation sites in Drosophila TOR.


After the sequencing of TSC2, it became apparent that the C-terminal half of TSC2 contained a domain homologous to the GTPase-activating protein (GAP) Rap1 (199, 200). This sequence similarity implied that the TSC1-TSC2 complex functioned as a GAP for one or more small GTP-binding proteins. Just which member(s) of this very large family of proteins linked TSC1-TSC2 to TOR was the question. Complementary screens for loss- and gain-of-function mutations affecting cell size in Drosophila provided this piece of the puzzle. The GTP-binding protein, Rheb (Ras homolog enriched in brain), was independently identified in screens conducted by several groups (201-203). Overexpressing Rheb increased cell size and prevented the reduction in cell size caused by nutrient deprivation (201, 202). Rheb depletion not only decreased cell size, but also prevented the larval lethality resulting from TSC1 knockout. These findings, along with results of detailed epistatic analyses using ommatidial size (202) or wing imaginal disc size (201) as the read-outs, provide genetic evidence for placing Rheb downstream of TSC1-TSC2 and upstream of TOR. Biochemical evidence in fruit flies also supports the conclusion that Rheb signals upstream of TOR. Overexpressing Rheb increased S6K activity measured in larval extracts (202). It was interesting that reducing Rheb not only decreased S6K activity but also increased Akt activity (202), which brings to mind earlier evidence of a feedback loop between mTOR and IRS-1 (159-161).

A key mechanistic point is that TSC1-TSC2 has Rheb GAP activity. TSC1-TSC2 accelerates GTP hydrolysis by Rheb in vitro (191, 204-206). GTP•Rheb is elevated in TSC2-deficient mouse embryo fibroblasts (205), and overexpressing TSC1-TSC2 in cells decreased the GTP-bound form Rheb (204). Insulin increased the ratio of GTP•Rheb:GDP•Rheb in NIH-3T3 cells (205), presumably by inhibiting the Rheb-GAP function of TSC1-TSC2.

Overexpression studies in mammalian cells support the conclusion that Rheb controls mTOR activity. Expressing a dominant-negative Rheb in HEK293 cells blocked activation of S6K-1 in response to insulin (207). This mutant form of Rheb has much higher affinity for GDP than for GTP and was identified by random mutagenesis of Rheb homolog in S. pombe. Overexpressing Rheb mimicked insulin by increasing the phosphorylation of both S6K-1 and 4EBP1 (191, 205, 206, 208). Mutating a Cys near the C-terminus of Rheb abolished the ability of overexpressed Rheb to increase S6K activity (191, 208). This Cys residue forms part of a CAAX motif and is the site of farnesylation in many small GTP-binding proteins. Thus, blocking farnesylation of Rheb is likely to be the explanation of earlier findings that farnesyltransferase inhibitors blocked the stimulatory effects of insulin or growth factors on the mTOR signaling pathway (209). Overexpressing Rheb did not increase the activity of a rapamycin-resistant form of S6K-1 (191, 206). This finding is perhaps the best evidence that Rheb signaling to S6K-1 is through mTOR and not through a parallel pathway. It is premature to conclude that Rheb directly activates mTOR. Deleting Rheb in S. cerevisiae leads to an increase in the uptake of arginine and lysine by the amino acid permease Can1p (210). This suggests that Rheb may control mTOR indirectly by changing amino acid levels.

TOR Signaling Complexes

Recent findings have confirmed an earlier suspicion that TOR functions as part of a larger signaling complex. By sequencing proteins that copurified with mTOR, two mTOR-associated proteins, raptor (also known as mKOG1) and mLST8 (also known as GβL), have been identified. Raptor (regulatory associated protein of mTOR) was independently identified by three groups (46, 128, 211). This protein (Mr = 150,000), which is approximately half of the size of mTOR, possesses a unique N-terminal region followed by three HEAT motifs and seven WD40 repeats (Fig. 5). mLST8 (Mr = 36,000) consists almost entirely of seven WD40 repeats and is similar in sequence to the β subunits of heterotrimeric G proteins (211, 212). It is striking that mTOR, raptor, and mLST8 have among themselves a total of 23 HEAT motifs and 14 WD40 repeats. Thus, the potential for protein-protein interactions is enormous. Several proteins associating with TOR proteins in S. cerevisiae have also been purified and identified by mass spectrometric analyses (211). The combinations of these proteins with TOR1 and TOR2 define rapamycin-sensitive and rapamycin-insensitive complexes, referred to as TORC1 and TORC2, respectively.

Fig. 5.

Domains in raptor and mLST8. The locations of the HEAT repeats in raptor and the WD40 domains in raptor and mLST8 are shown.

Yeast TORC1

Loewith et al. (211) presented evidence for two forms of TORC1, designated TORC1-A and TORC1-B, containing the proteins LST8 (28% identical to mLST8), KOG1 (27% identical to raptor), and either TOR1 or TOR2, respectively. TORC1 containing TOR2 was not detected in another study (34), raising a question as to whether TORC1-B exists. In any event, KOG1 is an essential gene in S. cerevisiae (211). Reducing KOG1 levels mimics rapamycin treatment by decreasing protein synthesis, increasing expression of genes controlled by GLN3 and RTG1 and RTG3 (collectively referred to as RTG1/3), and increasing both glycogen accumulation and the size of the central vacuole (211). LST8 was originally described as causing synthetic lethality with sec13, a gene involved in protein transport from the endoplasmic reticulum to the Golgi (213). Genetic analyses have implicated LST8 in several functions in S. cerevisiae (213, 214) and in S. pombe (215, 216). Not all of these functions necessarily involve TOR, because not all LST8 is associated with TOR (211). However, reducing LST8 causes defects in actin polarization (211), like that produced by TOR2 mutations. LST8 reduction also induces RTG1/3 expression mimicking mutations in TOR1 (213, 214).

Mammalian TORC1

Depleting 293T cells of raptor by using siRNA caused a decrease in cell size comparable to that produced by using siRNA to deplete cells of mTOR (46). Raptor depletion also decreased the phosphorylation of S6K-1 (46, 128), which indicated that raptor is essential for signaling to this important target of mTOR. The relative levels of raptor mRNA in different mammalian tissues correlate almost exactly with levels of mTOR mRNA (46, 211), as would be expected if raptor functioned as a subunit in an mTOR signaling complex. There is general agreement that raptor is important for TOR signaling in yeast and mammals, but there are differences in opinion as to the actual role of raptor in the process.

Hara et al. (128) noted that the association of raptor with mTOR was disrupted by nonionic detergents, previously shown to inhibit the phosphorylation of 4EBP1 by mTOR kinase. These investigators also found that increasing raptor-mTOR complexes by coexpressing raptor with mTOR enhanced the kinase activity of mTOR measured in vitro (128). Moreover, both S6K-1 and 4EBP1 coimmunoprecipitated with raptor, and raptor was recovered when complexes of 4EBP1 and eIF4E were purified with m7GTP-Sepharose (128). NP-40 did not inhibit binding of 4EBP1 and S6K-1 to raptor (128, 129), which indicated that the proteins interact directly with raptor instead of with mTOR, which dissociates from raptor under these conditions. On the basis of these findings, Hara et al. (128) proposed that raptor functions to position substrates for phosphorylation by mTOR (Model 1, Fig. 6). If the phosphorylation of 4EBP1 and S6K-1 by mTOR is determined by binding to raptor, it becomes easier to understand how the nature of the sites phosphorylated by mTOR in the two proteins could be so different.

Fig. 6.

The two major models for the influence of raptor on mTOR. Model 1, originally proposed by Hara et al. (128), raptor is an essential subunit, which binds and presents substrates to mTOR for phosphorylation. Dissociation of raptor by nonionic detergents in vitro markedly decreases the kinase activity of mTOR. Model 2, proposed by Kim et al. (46), interactions with raptor may enhance (+) or inhibit (–) mTOR function. When tightly bound to raptor, mTOR is relatively inactive. Nutrients such as leucine are proposed to increase mTOR function in vivo by weakening the interaction with raptor. Details of the two models are discussed in the text.

There is evidence that TOS motif-mediated binding to raptor is important for phosphorylation by mTOR. Mutating the TOS motifs in 4EBP1 and S6K-1 abolished binding to raptor and inhibited phosphorylation of the proteins by mTOR in vitro (81, 129, 141, 142). In contrast, the relatively low rate of phosphorylation by mTOR of a full-length S6K-1 protein having a Phe to Ala point mutation in the TOS motif was not decreased by dissociating raptor with NP-40 (129), and the phosphorylation of a C-terminal fragment (332-502) of S6K-1, which lacks the TOS motif, was actually increased by dissociating raptor with Triton X-100 (46). Thus, when bound to mTOR, raptor appears to facilitate the phosphorylation of some substrates, but inhibit the phosphorylation of others.

Attaching the last 20 amino acids of 4EBP1, which includes the TOS motif, to glutathione S-transferase (GST) was sufficient to promote binding to raptor (217). Disrupting the RAIP motif in 4EBP1 also decreased raptor binding (141), although the first 24 amino acids in 4EBP1, which include the RAIP motif, were insufficient to confer raptor binding when fused to GST (217).

The interaction with raptor also suggests a mechanism to explain the ordered phosphorylation of 4EBP1. Phosphorylation of 4EBP1 decreases its affinity for raptor (128), although sites responsible have not been identified. 4EBP1 tightly bound to raptor may be positioned in such a way that mTOR is only able to phosphorylate 4EBP1 on Thr37 and Thr46. Phosphorylation of Thr37 and Thr46 may weaken the interaction between 4EBP1 and raptor and thus allows mTOR to phosphorylate 4EBP1 on Thr70 and Ser65. Alternately, phosphorylation of Thr37 and Thr46 may prime 4EBP1 for phosphorylation by another undiscovered protein kinase, as has been previously proposed (132).

There are clearly circumstances in which raptor does not act to facilitate mTOR function. Overexpressing raptor decreased phosphorylation of both 4EBP1 and S6K-1 in HEK293 cells (46, 128). Inappropriately high levels of raptor might prevent substrates or unknown effectors from interacting with the mTOR-raptor complex, as suggested by Hara et al. (128). Another possibility is that raptor acts directly on mTOR to suppress activity.

Kim et al. (46) have presented evidence that the influence of raptor on mTOR is more complicated than that of a substrate-presenting subunit. Incubating cells with leucine, which increases mTOR function in cells, decreased the amount of raptor that coimmunoprecipitated with mTOR (46). On the other hand, if a chemical cross-linker was used to fix the raptor-mTOR complex before immunoprecipitation, then the same amount of raptor was recovered with mTOR after leucine treatment (46). These results imply that leucine weakens the interaction between raptor and mTOR, but by an amount insufficient to dissociate the complex in cells. Several agents that affect mitochondrial function or energy metabolism and inhibit mTOR signaling, including valinomycin, antimycin A, and 2-deoxyglucose, were found to stabilize the raptor-mTOR complex. Kim et al. (46) hypothesize that raptor inhibits mTOR function when it is bound in a high affinity state (Model 2, Fig. 6). Conditions that stimulate the mTOR pathway, such as increased amino acid availability, are proposed to shift the raptor-mTOR complex to a lower affinity, activated state. Raptor appears to form contacts with multiple regions of mTOR. The N-terminal half of mTOR, from amino acids 1 to 1482, bound raptor almost as tightly as full-length mTOR (46). In addition, the C-terminal region of mTOR, encompassing amino acids 1348 to 2549, contains a weak interaction site for raptor. Which regions of interaction are affected by amino acids is not known.

Curiously, Kim et al. (46) also found that rapamycin decreased the amount of raptor that coimmunoprecipitated with mTOR. To be consistent with their model, the decrease in affinity of raptor for mTOR produced by rapamycin, which inhibits mTOR, would have to be different from that of amino acids, which activate. Such a mechanism is conceivable, and the complete dissociation of raptor from mTOR would be expected to inhibit mTOR, because the siRNA experiments indicate that raptor is essential for mTOR function (46, 128). More difficult to explain is the failure of others to detect an effect of rapamycin on the amount of raptor bound to mTOR (128, 211).

An uncontrolled variable in many of the initial experiments with mTOR and raptor was mLST8. By all indications, mLST8 is an essential intrinsic subunit of the mTOR-signaling complex. The amount of mLST8 bound to mTOR does not change when cells are incubated with amino acids, serum, or rapamycin (212). mLST8 also remains bound to mTOR when immune complexes are washed with detergents that promote dissociation of raptor (212). Decreasing mLST8 with siRNA was without effect on mTOR or raptor expression, but blocked stimulation of S6K-1 in response to serum or amino acids. Depleting cells of mLST8 also caused a decrease in cell size comparable to that produced by depletion of mTOR or raptor (212). It was interesting that leucine did not dissociate overexpressed mTOR and raptor unless mLST8 was also overexpressed (212). This could explain why dissociation of overexpressed raptor and mTOR in response to amino acid stimulation was not detected in other studies (128, 211).

mLST8 interacts with the kinase domain of mTOR (212). Overexpressing mLST8 increased the kinase activity of mTOR and enhanced the interaction between raptor and mTOR under basal conditions (212). Also, mLST8 proteins having point mutations that reduced binding to mTOR exhibited a decreased ability to activate mTOR kinase activity in vitro (212). Thus, mLST8 is required for optimal mTOR activity, both in vivo and in vitro. The effect of mLST8 on enhancing raptor binding does seem inconsistent with the model in which tight binding of raptor inhibits mTOR activity. Again, the explanation may relate to the fact that the mTOR lacking raptor is inactive.

Although the models of Hara et al. (128) and Kim et al. (46) differ in important respects, not all aspects of the models are mutually exclusive. It might turn out that raptor both presents substrates and modulates mTOR activity through its binding interactions with mTOR. It is to be hoped that these issues will be resolved soon, because raptor has become a focus of studies in the TOR field.


Our understanding of TORC2 is derived solely from studies in S. cerevisiae. This complex, which does not interact with the rapamycin-FKBP12, contains TOR2, LST8, and the proteins, AVO1, AVO2, and AVO3 (211). AVO1 and AVO3 are essential genes (211, 218). Deleting AVO1 cause defects in actin polarization closely resembling those produced by deleting TOR2 (11, 211). In addition, suppressors of the TOR2 defect in actin polarization (including ROM2 and RHO2) also suppress defects resulting from AVO1 deficiency (211). Thus, TORC2 appears to be responsible for mediating the rapamycin-insensitive component of TOR signaling in S. cerevisiae.

SIN1, the AVO1 counterpart in S. pombe, is an essential gene in fission yeast. It was interesting that the mammalian homolog of AVO1 is able to substitute for Sin1, which indicates that the function of the protein has been conserved (219). Levels of mSin1 among tissues correlated well with levels of mTOR, although direct association between the two proteins has not been demonstrated (211). Very little is known about the biochemical functions of the AVO proteins, but sequence similarities with other proteins point to a role in Ras signaling. AVO1 has a Ras association domain, AVO2 contains five ankyrin repeats in its N-terminal region, and AVO3 contains a Ras guanine nucleotide exchange factor (GEF) domain. Homologs of both AVO1 and AVO3 have been implicated in Ras signaling in Dictyostelium discoideum (220, 221). The potential connection between TOR and Ras signaling is intriguing in view of the Ras homolog, Rheb, which is an upstream regulator of mTOR. However, since AVO1 is the only one of the AVO proteins having a known homolog in mammalian cells, it is not clear that mammals have the equivalent of TORC2.

Concluding Remarks

After years of frustration, there has been remarkable recent progress in defining upstream elements in the signaling pathway through which insulin and growth factors activate mTOR. Genetic, biochemical, and pharmacological evidence support the model in Fig. 4. Although the currently identified pieces of the puzzle have been assembled, there remain many holes that will need to be filled. Because knowing what is not known may sometimes be almost as useful as knowing what is, we conclude by mentioning some of the outstanding questions in the mTOR field.

With respect to the upstream effectors of mTOR, we still do not know the mechanism by which amino acids increase mTOR activity. How Akt-mediated phosphorylation influences the function of the TSC1 and TSC2 complex and how Rheb signals to mTOR are two other major unanswered questions. At the level of mTOR kinase activity, precisely how mTOR signals to its downstream effectors is still not clear. Are the downstream targets, 4EBP1 and S6K-1, directly phosphorylated by mTOR in cells, or does mTOR signal to these effectors by modulating the activities of other protein kinases, phosphatase, or both? The intracellular localization of mTOR is still mysterious. Knowing where mTOR is located will be essential for understanding both the control and function of the protein. The mTOR-associated proteins, raptor and mLST8, have only recently been discovered, and there are almost diametrically opposed opinions as to their roles. Considering the multiple HEAT and WD40 motifs found in the raptor, mTOR, and mLST8 complex, it seems almost certain that other proteins associate with the signaling complex. Is there a mammalian counterpart of the yeast TORC2? Are there mTOR-mediated responses that are not inhibited by rapamycin?

Although the questions posed above do not reflect all that is not known with respect to mTOR, these issues will undoubtedly influence the direction of future mTOR research.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 154a.
  156. 155.
  157. 156.
  158. 157.
  159. 158.
  160. 159.
  161. 160.
  162. 161.
  163. 162.
  164. 163.
  165. 164.
  166. 165.
  167. 166.
  168. 167.
  169. 168.
  170. 169.
  171. 170.
  172. 171.
  173. 172.
  174. 173.
  175. 174.
  176. 175.
  177. 176.
  178. 177.
  179. 178.
  180. 179.
  181. 180.
  182. 181.
  183. 182.
  184. 183.
  185. 184.
  186. 185.
  187. 186.
  188. 187.
  189. 188.
  190. 189.
  191. 190.
  192. 191.
  193. 192.
  194. 193.
  195. 194.
  196. 195.
  197. 196.
  198. 197.
  199. 198.
  200. 199.
  201. 200.
  202. 201.
  203. 202.
  204. 203.
  205. 204.
  206. 205.
  207. 206.
  208. 207.
  209. 208.
  210. 209.
  211. 210.
  212. 211.
  213. 212.
  214. 213.
  215. 214.
  216. 215.
  217. 216.
  218. 217.
  219. 218.
  220. 219.
  221. 220.
  222. 221.
  223. 222.
View Abstract

Stay Connected to Science Signaling

Navigate This Article