The mammalian target of rapamycin (mTOR) is a protein kinase that plays key roles in cellular regulation. It forms complexes with additional proteins. The best-understood one is mTOR complex 1 (mTORC1). The regulation and cellular functions of mTORC1 have been the subjects of intense study; despite this, many questions remain to be answered. They include questions about the actual mechanisms by which mTORC1 signaling is stimulated by hormones and growth factors, which involves the small GTPase Rheb, and by amino acids, which involves other GTPase proteins. The control of Rheb and the mechanism by which it activates mTORC1 remain incompletely understood. Although it has been known for many years that rapamycin interferes with some functions of mTORC1, it is not known how it does this, or why only some functions of mTORC1 are affected. mTORC1 regulates diverse cellular functions. Several mTORC1 substrates are now known, although in several cases their physiological roles are poorly or incompletely understood. In the case of several processes, although it is clear that they are regulated by mTORC1, it is not known how mTORC1 does this. Lastly, mTORC1 is implicated in ageing, but again it is unclear what mechanisms account for this. Given the importance of mTORC1 signaling both for cellular functions and in human disease, it is a high priority to gain further insights into the control of mTORC1 signaling and the mechanisms by which it controls cellular functions and animal physiology.
The mammalian target of rapamycin (mTOR) is a protein kinase that is involved in the control of a diverse range of cellular processes, including protein synthesis, ribosome biogenesis, the cell cycle, cell growth, gene transcription, autophagy and metabolism. Activation of mTOR signaling is implicated in diseases such as cancer and tissue hypertrophy; inhibitors of mTOR signaling are used clinically to prevent graft rejection and restenosis after angioplasty, as well as for advanced renal cell carcinoma. Recent data from various model organisms point to a key role for mTOR (and its orthologs in other species) in controlling lifespan/ageing (Proud, 2009; Blagosklonny, 2010). The goal of this review is not to provide a comprehensive account of current knowledge of signaling through mTOR, as there are several recent excellent reviews in this area, a number of which are cited here. The aim of this article is, rather, to try to highlight important outstanding questions about the regulation and function of mTOR complex 1 (mTORC1), one of two types of the multiprotein complexes in which mTOR participates.
The mTORC1 protein
mTOR complexes 1 and 2 (mTORC1/2) differ in their composition, regulation and functions (Figure 1). Both types of complex contain mTOR and mLST8 (also called GβL), while only mTORC1 also contains raptor, which is involved in recruiting substrates for phosphorylation by the kinase domain of mTOR. mTORC2 contains several proteins not found in mTORC1, i.e. rictor, Sin1 and protor. mLST8 and Sin1 each help to maintain the integrity of mTORC2 (Guertin et al., 2006; Jacinto et al., 2006). However, mLST8 does not seem to play the same role for mTORC1 (Guertin et al., 2006) and questions remain about the roles of protor and rictor in mTORC2.
The first part of this review will cover the regulation of mTORC1 signaling (Figure 2). This is a complex area reflecting the multiplicity of inputs that affect mTORC1 signaling in cells. This already introduces our first question: what is meant by activation of mTORC1, what does ‘activation’ mean here? Many studies show that the phosphorylation of substrates of mTOR, such as the S6 kinases (S6Ks) or their own substrate, ribosomal protein (rp) S6 (a component of the small subunit), changes in response to a range of stimuli. Very few studies report changes in the activity of isolated mTOR complexes measured in vitro: Mothe-Satney et al. (2004) have reported that mTOR complexes isolated from hepatocytes preincubated with amino acids and/or insulin showed enhanced kinase activity against recombinant 4E-BP1 (eukaryotic initiation factor 4E-binding protein 1) in vitro.Sancak et al. (2007) showed that insulin pretreatment of cells enhanced the activity of mTORC1 isolated at low, but not at high, salt. This led to the identification of PRAS40 as a binding partner for raptor and an inhibitor of mTORC1 (Vander Haar et al., 2007). PRAS40, which is also a substrate for mTORC1, is discussed in more detail below. So it is unclear what activation of mTORC1 actually means: what effects do agents that stimulate mTORC1 actually have on the complex? Do they increase its intrinsic catalytic activity? Do they cause some rearrangement of its components which enhances its ability to phosphorylate substrates? Interestingly, one recent study reports that activation of mTORC1 increased the binding of substrate (in that case, a protein called 4E-BP1, which is described in detail below; Sato et al., 2009).
Where in the cell is mTORC1 located?
There are numerous reports pertaining to the subcellular localization of mTOR/mTORC1 (some using the alternative names FRAP or RAFT1), which place it in various locations including mitochondria (Desai et al., 2002), neuronal cell membranes (Sabatini et al., 1999), nucleus (Zhang et al., 2002; Bachmann et al., 2006), endoplasmic reticulum (‘microsomes’) (Drenan et al., 2004) and/or Golgi (Withers et al., 1998).
Recent data indicate that it is activated on the surface of lysosomes (Sancak et al., 2010). However, its diverse functions suggest that activated mTORC1 migrates elsewhere. For example, there are now reports that mTORC1 functions in the nucleus (Cunningham et al., 2007; Wei and Zheng, 2009; Kantidakis et al., 2010) including at the RNA polymerase I promoter which functions in the nucleolus. This may well be linked to the ability of mTORC1 to regulate transcription including the synthesis of ribosomal RNA (rRNA) which occurs in the nucleolar compartment (see also below). It has also been suggested that mTOR shuttles between nucleus and cytoplasm and that this shuttling is crucial for mTOR's function in regulating proteins involved in mRNA translation (a cytoplasmic process) (Kim and Chen, 2000). The reasons for this remain obscure. Further work is clearly required to examine the intracellular location and trafficking of mTORC1, both in activated and less active states.
How does Rheb, a small GTPase, activate mTORC1?
Rheb can activate mTORC1 signaling when in its GTP-bound state (Huang and Manning, 2008). However, it remains unclear how Rheb–GTP stimulates mTORC1. Data from Long et al. (2005a) indicate that Rheb can bind to mTOR but, and puzzlingly, Rheb can do so whether it is bound to GTP or GDP. Even a mutant that cannot bind guanine nucleotides interacted with mTOR. These studies were performed with overexpressed proteins and it is important to establish whether Rheb can bind to mTOR at physiological levels, and whether such binding is truly direct (or mediated by another protein, which is possible since these studies used proteins expressed in mammalian cells which might thus contain a protein that mediates Rheb/mTOR binding).
It has also been reported that Rheb binds to FKBP38 (Figure 2), an immunophilin related to FKBP12 (the partner for rapamycin) (Bai et al., 2007). FKBP12 binds to mTOR as a complex with rapamycin and inhibits mTORC1 activity. Association of FKBP38 with Rheb–GTP resulted in the release of FKBP38 from mTORC1 leading to the proposal that Rheb–GTP disinhibits mTORC1 by eliciting its disengagement from FKBP38. However, several lines of evidence are inconsistent with this model. For example, the interaction between FKBP38 and mTOR did not change under conditions where mTORC1 signaling is markedly altered and overexpression of FKBP38 did not, as predicted by this model, inhibit mTORC1 signaling (Wang et al., 2008b). Furthermore, Uhlenbrock et al. (2009) were unable to detect binding between mTOR and FKBP38, placing a question mark against FKBP38's proposed role in linking Rheb to the control of mTORC1.
The control of mTORC1 by Rheb may be linked to the effects of phosphatidic acid (PA) on mTORC1 (Avruch et al., 2009). PA can bind directly to mTOR, apparently activating it (Figure 2). Consistent with this, inhibition or knockdown of phospholipase D (PLD, which generates PA) impairs mTORC1 signaling (Fang et al., 2001, 2003). It has been suggested that Rheb may activate mTORC1 indirectly, by stimulating PLD (Sun et al., 2008) and thus increasing levels of PA. The binding of PA to mTOR is competitively inhibited by FKBP12/rapamycin; while this provides a possible way that rapamycin inhibits mTORC1, it is also puzzling, since some effects of mTORC1 that are stimulated by Rheb are insensitive to rapamycin, e.g. the phosphorylation of the N-terminal sites in 4E-BP1 (Wang et al., 2005). Since mitogenic stimuli can activate PLD, this could provide a further link between mitogens and the activation of mTORC1. However, further corroborating evidence is needed to establish PA as a physiological modulator of mTORC1—one concern here is that flies lacking PLD show no overt defects (LaLonde et al., 2006), indicating it plays no key role in TOR signaling at least in adult insects.
What are the upstream steps in the regulation of mTORC1 activity?
Control of the GTP-binding state of Rheb will be regulated by the opposing effects of GTP hydrolysis (to GDP) and the replacement of the bound GDP by GTP to recreate active Rheb–GTP.
Reading of the mTORC1-related literature yields the impression that the mechanism by which hormones, growth factors and other agents control mTORC1, through the TSC1/2 complex (Figure 2), is mechanistically well-established. The link between TSC proteins and (m)TORC1 signaling first emerged from investigations into the functions of these proteins in mammalian cells, where they were found to suppress the phosphorylation of targets for mTORC1 signaling (Inoki et al., 2002). These data were reinforced by elegant studies in Drosophila (Saucedo et al., 2003; Stocker et al., 2003). It is very clear that TSC1/2 do act to restrain mTORC1 signaling (downstream targets are hyperphosphorylated in cells lacking one or other of these two proteins, or harboring certain mutations in them) and that TSC1 and TSC2 are both phosphoproteins. Indeed, numerous kinases and signaling cascades are reported to act on TSC2, and to a lesser extent, TSC1 (Huang and Manning, 2008). What is much less clear is how these various phosphorylation events affect the activity of the TSC1/2 (Rheb-GAP) complex. As far as these authors are aware, there is no data directly showing, e.g. an impairment of TSC1/2 activity upon phosphorylation of the complex by kinase(s) linked to the activation of mTOR signaling. Instead, it has been proposed that phosphorylation of TSC2 by PKB/Akt (Figure 2) allows it to bind 14-3-3 proteins (which interact with certain phosphoproteins in a site-specific manner), resulting in movement of TSC2 to the cytoplasm and away from membranes where its substrate Rheb is located (Cai et al., 2006).
Puzzling points here include the facts that: (i) the multiple reported phosphorylation sites are dispersed across the sequence of TSC2 (Huang and Manning, 2008), begging the question ‘how do these diverse (inhibitory and stimulatory) sites exert the same proposed end-effect, altered hydrolysis of Rheb-bound GTP?’; (ii) several key sites in mammalian TSC2 are either not conserved in the Drosophila ortholog, e.g. the sites in mammalian TSC2 that are phosphorylated by the AMP-activated kinase (AMPK), GSK3 or p90RSK (which lie downstream of Ras/Raf/MEK/ERK signaling) or, even where they are conserved (PKB/Akt sites) they are apparently dispensable for the physiological function/control of TSC2 in that species (Dong and Pan, 2004); and (iii) recent data suggesting that the stoichiometry of phosphorylation of many of the regulated sites in TSC2 is rather low, even in response to stimuli that activate mTORC1 (Ballif et al., 2005; unpublished data). This last point raises the question ‘given that TSC1/2 appear to be a basally-active GAP for Rheb, how can substoichiometric phosphorylation of this complex switch it off sufficiently to allow increases in Rheb–GTP levels?’. Maybe there are (differently localized?) pools of TSC1/2 complexes, some (which are subject to control) acting on Rheb, and others having no effect on Rheb.
Alternatively, control of mTORC1 may (also) involve phosphorylation of raptor, e.g. by AMPK (Gwinn et al., 2008) or p90RSK (Carriere et al., 2008) (Figure 2). However, the corresponding phosphorylation sites are in regions of raptor that are not well-conserved between mammals and, e.g. flies, suggesting that different mechanisms for the control of mTOR, for example by energy stress, operate in flies and mammals.
How do amino acids activate mTORC1?
Amino acids could, in principle, act in any one of several diverse ways to activate mTORC1 signaling. Removing amino acids from the medium in which cells are maintained results in a rapid (15 min or so) dephosphorylation of substrates of mTORC1 such as S6K1 and the 4E-BPs (Hara et al., 1998) and this is quickly reversed when amino acids are resupplied. Removing single amino acids also impairs mTORC1 signaling: the effect is especially strong for the essential amino acids leucine and arginine. Adding back individual amino acids partly stimulates the pathway, although this varies between reports and/or cell types, the most effective single amino acid generally being leucine (a branched-chain amino acid). This primary role for leucine is of particular interest given that this amino acid alone can stimulate muscle protein synthesis (Rennie, 2007). There is limited information about this, but it does seem that mTORC1 is activated by intracellular amino acids (Beugnet et al., 2003a) as derived from import from the medium or from protein degradation, e.g. by the proteasome (Vabulas and Hartl, 2005). Inhibitors of protein synthesis (utilization) can therefore ‘rescue’ mTORC1 signaling in amino acid-starved cells thus helping to explain the long-standing observation that cycloheximide (which inhibits translation elongation) induces the activation of S6K1 (Price et al., 1989).
All of which brings us to the key question: how do amino acids activate mTORC1? Amino acids could activate mTORC1 through a mechanism analogous to that proposed for the activation of the pathway by hormones and growth factors, i.e. by inhibiting the Rheb-GAP, TSC1/2 (Huang and Manning, 2008). However, amino acids still affect mTORC1 signaling in TSC2 null cells (Smith et al., 2005) and elicit only small, if any, changes in Rheb GTP-loading (Long et al., 2005a; Roccio et al., 2006). These and other data suggest that amino acids probably do not regulate mTORC1 by affecting the nucleotide-binding status of Rheb. Nonetheless, it is well-established that overexpressing Rheb can fully activate mTORC1 signaling in cells deprived of external amino acids (Long et al., 2005a), suggesting that the mechanisms by which amino acids activate mTORC1 are somehow linked with the mechanism by which Rheb does this. Furthermore, Long et al. (2005b) have provided evidence that amino acid-starvation interferes with the association of Rheb with mTOR, although the mechanism for this remains unknown. It may be linked to changes in the subcellular localization of mTOR in response to amino acids, as described below.
Recently, there have been important advances in our knowledge of the intracellular components involved in the control of mTORC1 by amino acids. MAP4K3 (Figure 2) was initially identified through a screen for protein kinases involved in TOS in Drosophila (Findlay et al., 2007). Its activity was shown to be promoted by amino acids, and MAP4K3 was found to promote mTORC1 signaling and growth in human cells (Findlay et al., 2007). A recent study in Drosophila demonstrated that MAP4K3 plays an important role in controlling TORC1 signaling and growth, and has shown that such effects are diminished under conditions of starvation, as would be predicted for a protein that positively regulates TORC1 in response to amino acids. Subsequent studies have indicated that amino acids promote MAP4K3 activity by maintaining its phosphorylation at S170 (in its activation loop) and withdrawing amino acids causes the dephosphorylation of this site and inactivation of MAP4K3 through the action of protein phosphatase 2A (PP2A) complexed with a regulatory protein, PR61ε (Yan et al., 2010). Clearly, it is crucial to establish, first, how amino acids act to hinder the dephosphorylation of S170 in MAP4K3—do they regulate the PP2A–PR61ε complex, or is this a substrate-directed effect mediated by an effect on MAP4K3 itself? A second key goal is to identify the relevant substrates for MAP4K3 which are involved in the control of mTORC1 activity.
A major recent development in understanding how amino acids control mTORC1 came from the identification of the Rag GTPases as essential components in the signaling events that couple amino acids to control of mTORC1 and its Drosophila counterpart (Figure 2). Two studies, using quite different approaches, identified the Rag GTPases as being required for the control of (m)TORC1 by amino acids. Sancak et al. (2008) found RagC as an interaction partner for raptor, while Kim et al. (2008) came across the orthologs of the Rag GTPases by conducting a screen for small GTPases that control TOS in Drosophila, where these proteins play key roles in cell growth and cell viability under conditions of amino acid starvation (Bryk et al., 2010). There are four Rag proteins in mammals (RagA∼D): they are unusual GTPases firstly because they act as heterodimers (containing RagA or B with RagC/D) and second because they function with one partner bound to GTP and the other to GDP. For example, the RagB–GTP/RagC–GDP pairing was found to be most effective at binding mTORC1 (Sancak et al., 2008).
The data from Sabatini's lab indicate that the Rag complex resides on the lysosomal surface (due to its interaction with a further, heterotrimeric protein complex termed the ‘ragulator’), and that activation of the Rag complex by amino acids allows it to bind mTORC1. This recruits mTORC1 to the lysosomal surface, where Rheb–GTP is also situated, leading to activation of mTORC1 (Sancak et al., 2008, 2010). Thus, signals from amino acids and hormonal/growth factors stimuli would be integrated via pathways involving distinct GTPases (Rags and Rheb, respectively). However, it is still not conclusively established where Rheb is located within the cell, different studies having placed it on various endomembranes including lysosomes and also endosomes, where it likely interacts with mTOR (Sancak et al., 2010).
Important outstanding questions include: (i) how are intracellular amino acids (leucine) ‘sensed’ (i.e. what is the detector?); (ii) how do the Rag GTPases and MAP4K3 work together to regulate mTORC1; and (iii) how are the Rag GTPases themselves controlled? As for other GTPases, the likeliest mechanisms are through GTPase-activator proteins and GEFs, although the fact that the active Rag complex contains one GTP-bound and one GDP-bound Rag protein does add an extra level of potential complexity.
Interestingly, the closest yeast orthologs of the Rag GTPases are Gtr1 and Gtr2, which also heterodimerize. They are involved in intracellular protein trafficking and the control of microautophagy (Dubouloz et al., 2005; Gao and Kaiser, 2006). Gtr1 has recently been shown to bind to and activate yeast TORC1 in a study which also identified Vam6, a GEF, as regulating the GTP-binding state of Gtr1 (Binda et al., 2009). It remains to be established whether, and through what mechanisms, amino acids modulate the guanine nucleotide-binding status of the mammalian Rag GTPases.
Is there a Rheb-GEF?
By analogy with the control of other GTPases such as Ras, a Rheb-GEF could provide an important input to the regulation of Rheb (Figure 2). Several reports indicate that the levels of Rheb–GTP in cells are rather high even in cells starved of serum or of serum and amino acids (Nobukuni et al., 2005; Smith et al., 2005; Roccio et al., 2006). A caveat here is that most studies involve overexpression of tagged Rheb as antibodies for Rheb suffered from technical problems, and overexpressed Rheb may behave differently from the endogenous protein (Flinn et al., 2010). This implies that the activity of any Rheb-GEF may not be limiting, which also raises an important question about why mTORC1 signaling activity seems to change so markedly even when Rheb–GTP levels are high and/or vary relatively little.
The Drosophila ortholog of the human protein TCTP (translationally controlled tumor protein, also termed P23 and Q23) was shown to positively regulate cell growth, cell proliferation and dTOR signaling (S6K) (Hsu et al., 2007). It was also reported to accelerate the release of GDP from (human) Rheb–GDP complexes (Hsu et al., 2007), indicating it might act as a GEF for Rheb. However, the basal rate of such release is already high (at least in vitro, e.g. Rehmann et al., 2008), making it unclear how important a Rheb-GEF would be to maintain Rheb–GTP levels. Furthermore, two studies (Rehmann et al., 2008; Wang et al., 2008b) found that knocking down TCTP did not impair mTORC1 signaling and overexpressing TCTP did not activate it. These authors were also unable to detect an interaction between TCTP and Rheb. Other data show that TCTP does not stimulate the release of bound GDP from Rheb (Rehmann et al., 2008). In contrast, Dong et al. (2009) provided evidence that human TCTP interacts with Rheb and can promote the release of bound GDP in vitro, as well as activating mTORC1 signaling in cells. The existence of a Rheb-GEF, and whether TCTP performs that function physiologically, remain unresolved issues.
How does rapamycin interfere with mTORC1 signaling?
Although it has been known for around 20 years that rapamycin interferes with cellular functions associated with TOR (or mTOR), and for almost 10 years that mTORC1 is the relevant mTOR complex, it remains unclear how rapamycin actually exerts its effects on mTORC1. This is central to our understanding of mTORC1 and the action of rapalogs. Puzzlingly, much (50×) higher concentrations of rapamycin are needed to inhibit mTOR kinase activity in vitro than to impair mTORC1 signaling in vivo. What is clear is that rapamycin inhibits mTORC1 by binding to the immunophilin FKBP12 and that the interface between mTOR and the FKBP12–rapamycin complex involves elements of both partners, likely accounting for the high specificity of rapamycin (Choi et al., 1996).
FKBP12–rapamycin destabilizes, but does not abolish, the binding of mTOR to raptor (Kim et al., 2002). Other data show that mTOR and raptor can be co-purified on immobilized FKBP12 in the presence of rapamycin (Jacinto et al., 2004), again indicating that the mTORC1 complex remains intact in the presence of rapamycin and implying that FKBP12–rapamycin and raptor can bind simultaneously to mTOR. A recent cryo-EM microscopy study of mTORC1 suggests that (at least in vitro) relatively long-term addition of FKBP12–rapamycin does cause the disintegration of mTORC1 (Yip et al., 2010). The findings of that study have important implications for understanding the action of rapamycin.
It is also unclear why some outputs from mTORC1 are insensitive to rapamycin, or relatively so. The clearest example of this is provided by one of the most intensively studied mTORC1 substrates, the protein 4E-BP1 (Gingras et al., 1999; Wang et al., 2005; Choo et al., 2008), which is phosphorylated at multiple sites. Two sites lie near its main point of interaction with its partner eIF4E (eukaryotic initiation factor 4E; Figures 3 and 4), the protein that binds to the 5′-cap structure found on all cytoplasmic mRNAs in mammalian cells. In human 4E-BP1, these sites are Ser65 and Thr70. Phosphorylation of Ser65 is low in serum-starved cells but is increased, e.g. by insulin, and this is inhibited by rapamycin. Phosphorylation of Ser65 requires the TOR signaling motif (TOS) motif in the C-terminus of 4E-BP1 (Beugnet et al., 2003b). In contrast, the phosphorylation of Thr37 and Thr46 appears to be largely independent of the TOS motif but instead requires an N-terminal feature termed (from its sequence) the RAIP motif (Beugnet et al., 2003b). Phosphorylation of these two sites is already high in serum-starved cells (unlike Ser65), is dependent upon amino acids (Wang et al., 2005) and generally insensitive to rapamycin (Beugnet et al., 2003b; Wang et al., 2005). This can be interpreted as indicating that rapamycin interferes with mTORC1-TOS motif recognition, although other possibilities exist. The RAIP motif also appears to be involved in binding to raptor (Lee et al., 2008): this suggests that two modes of interaction between raptor and 4E-BP1/2 operate, one involving the RAIP motif, which is mainly regulated by amino acids in a way that is affected little by rapamycin, and a second via TOS motifs which mediates rapamycin-sensitive phosphorylation and seems to require additional inputs, e.g. from insulin. These effects likely reflect poorly understood aspects of the regulation of the presentation of substrates or their phosphorylation by mTORC1.
PRAS40, another substrate for mTORC1 (Figure 4), also contains a site that is resistant to rapamycin (Wang et al., 2008a), but since it does not contain a recognizable RAIP-like motif, this does not help understand how and why sensitivity to rapamycin apparently varies between mTORC1 substrates. The behavior of certain phosphorylation sites in 4E-BP1 and PRAS40 contrasts markedly with the phosphorylation of Thr389/Thr388 (Figure 4) in the S6Ks, other direct substrates for mTORC1, which is, almost without exception in the extensive literature relating to these enzymes, completely blocked by rapamycin (Avruch et al., 2001). Again, the phosphorylation of the mTORC1 site in S6K1 (for example), requires a TOS motif (Schalm and Blenis, 2002; Nojima et al., 2003). To date, all the known direct substrates for mTORC1 (except perhaps Maf1) possess a TOS motif that binds to raptor, although it is possible that mTORC1 substrates that lack TOS motifs do exist. HIF1α (hypoxia-inducible factor 1α, a transcription factor) also binds raptor through a TOS motif, but has not been shown to be a substrate for mTORC1. It is possible that some proteins interact with raptor (e.g. via a TOS motif) but are not actually substrates for this complex. To date, no proteins other than 4E-BP1/2 have been found to contain functional RAIP-type motifs.
Oddly, the rapamycin-sensitive phosphorylation of S6K1 in vitro seemingly does not require the presence of raptor (Yip et al., 2010), whereas phosphorylation of 4E-BP1 did require the presence of raptor. All-in-all, given the differences in rapamycin sensitivity of mTOR and/or mTORC1 against differing substrates, or different sites within the same substrate, it is probable that rapamycin affects the presentation of substrates to mTOR's kinase domain, although further studies are certainly required to gain clear insights into the molecular details of these effects.
Does mTORC1 regulate protein phosphatases?
The idea that the phosphorylation of some substrates for mTORC1 is not affected (greatly) by rapamycin, based on earlier studies using rapamycin and other, non-selective, inhibitors of PI3-kinase (PI3K) or amino acid-starvation (Wang et al., 2005), has been greatly strengthened by work using newly developed small molecule inhibitors of the kinase activity of mTOR (Feldman et al., 2009; Thoreen et al., 2009).
Another related point is that while the phosphorylation of Ser65 induced by stimuli such as insulin in amino acid-starved cells is generally sensitive to rapamycin, its phosphorylation in serum-fed cells often does not show this property (Wang et al., 2005). One interpretation of this is that phosphorylation of this site is stable (does not turn over quickly), so that once modified, this residue remains phosphorylated even when mTORC1 is inhibited by rapamycin; however, amino acid withdrawal causes the rapid loss of phosphate from this site (Wang et al., 2005), arguing against this (although amino acid starvation might activate a protein phosphatase which acts on this site).
Another instance where control by mTORC1 via a protein phosphatase may be important is presented by the example of the S6Ks: mTORC1 is only to phosphorylate S6K1 at a single site (the aforementioned Thr389). However, treatment of cells with rapamycin induces dephosphorylation of many sites in S6K1 (Avruch et al., 2001); the effect could be explained if Thr389 played an early ‘priming’ role in their phosphorylation—however, Thr389 is actually modified relatively late in the series of events leading to activation of S6K1 leading to the speculation that the rapamycin-induced dephosphorylation of S6K1 is largely due to activation of a phosphatase. Following the finding that rapamycin regulates members of the PP2A family of protein phosphatases in budding yeast through a phosphatase-associated protein (Tap42) (di Como and Jiang, 2006) in the 1990s and the subsequent discovery of a mammalian homolog (α4) (Murata et al., 1997), there were several reports concerning the control of mammalian protein phosphatases by rapamycin, and of their association with, e.g. S6K1 (Murata et al., 1997; Nanahoshi et al., 1998; Peterson et al., 1999). However, there have been relatively few recent developments in this area and it remains unclear whether and, if so, how mTORC1 regulates mammalian protein phosphatase activity.
Are there multiple mechanisms for the control of mTORC1 by cellular energy status?
mTORC1 promotes energy-requiring anabolic processes such as protein synthesis and ribosome biogenesis; it would therefore make excellent sense for mTORC1 signaling to be impaired under conditions of energy limitation, e.g. due to inadequate supplies of oxidisable nutrients (such as sugars or fats) or oxygen (in aerobically metabolizing cells). Indeed, many studies have observed that targets of mTORC1 become dephosphorylated under such conditions. Four main mechanisms have been proposed to explain this: (i) the regulation of TSC1/2 through the phosphorylation and proposed activation of the latter protein by AMPK, as already discussed (Inoki et al., 2003); (ii) alluded to above, the phosphorylation of raptor by AMPK (Gwinn et al., 2008); (iii) based on the fact that mTOR has a high Km for ATP (in the low millimolar range) (Dennis et al., 2001), it was proposed that depletion of cellular ATP would decrease the activity of mTOR by a simple Michaelis–Menten effect, but such a mechanism would only likely come into play if ATP levels fell drastically (to levels at which cells could very likely not survive), thus severely compromising both the pro-survival effects of impairing mTORC1 signaling (to save energy) and the importance of homeostatic mechanisms to maintain cellular ATP levels; this suggested mechanism is now generally not thought to account for the control of mTORC1 signaling by cellular energy status; (iv) the protein REDD1 is transcriptionally induced under conditions of hypoxia (via HIF1α) and this appears necessary for the impairment of mTORC1 signaling under these conditions (Brugarolas et al., 2004). REDD1 can bind 14-3-3 proteins and it is suggested (DeYoung et al., 2008) that REDD1 ‘removes’ the 14-3-3 proteins from the PKB/Akt sites in TSC2, resulting in activation of TSC2 and inhibition of mTORC1. Two key questions here are: first, could REDD1 actually sequester a high enough proportion of 14-3-3 proteins to affect TSC2 given the myriad of other 14-3-3 partners (MacKintosh, 2004; Morrison, 2009); second, how does REDD1 bind 14-3-3 proteins? Although the identified 14-3-3-binding region in REDD1 does contain a potential phosphorylation site (S137), it is not clear whether it is actually phosphorylated and, if so, whether/how this is regulated.
How does MEK/ERK signaling regulate mTORC1?
As mentioned above, evidence has also been published suggesting that phosphorylation of raptor may regulate mTORC1 function. For example, phosphorylation of raptor by p90RSK (which is activated by ERK) has been suggested to explain the stimulation of mTORC1 signaling through the classical MAP kinase pathway, which is activated by oncogenic mutations in Ras or Raf, or by loss-of-function mutations in the Ras-GAP, NF1 (neurofibromin, a tumor suppressor) (Carriere et al., 2008). However, this link is only reported to operate in certain types of cells (such as some types of human embryonic kidney cells) (Roux et al., 2004). A small molecule inhibitor of p90RSK fails to impair the activation of mTORC1 signaling through MEK/ERK, e.g. in cardiac myocytes (MEK/ERK/mTORC1 signaling drives cell growth associated with cardiac hypertrophy, a life-threatening disorder). Phosphorylation of TSC2 by ERK may play a role here (Ma et al., 2005), although TSC2 appears to be a very poor substrate for ERK, at least in vitro (Rolfe et al., 2005). ERK also phosphorylates TSC1 (unpublished data) and this may regulate the TSC1/2 complex. Thus, it remains to be established what mechanism accounts for the activation of mTORC1 signaling by MEK/ERK in cells where such regulation does not require p90RSK function.
Downstream targets: both molecular and cellular
There are still surprisingly few proteins that are known to be direct substrates for phosphorylation by mTORC1 (Figure 4). The best-studied ones are the S6Ks and the 4E-BPs. Interestingly, the phosphorylation sites for mTORC1 in these proteins lie in contrasting contexts—in the S6Ks, they are in a ‘hydrophobic motif’ (FLGFTYV in human S6K1/2) while the four mTORC1 sites in the 4E-BPs are all followed by prolines. This marked difference is surprising, since many protein kinases recognize rather well-defined, or even strict, consensus sequences in their substrates. This may reflect differences in the way these classes of substrates are presented to the active site of mTOR, although both possess TOS motifs (Schalm and Blenis, 2002; Schalm et al., 2003).
Only a few other mTOR substrates are known or suspected: these include PRAS40 and STAT3 (Yokogami et al., 2000) (the latter protein has not formally been shown to be a direct substrate for mTORC1). As noted above, HIF1α contains a TOS motif and binds raptor (Land and Tee, 2007) but has yet not been shown to be a substrate for mTORC1. A search for proteins that contain TOS-like motifs may reveal other substrates, although in at least one case, it may be a partner for the protein that contains the TOS motif which is actually the substrate for mTORC1 (e.g. Maf1) (Kantidakis et al., 2010).
S6Ks: what are their physiological roles?
Of the small number of reported types of direct substrates for mTORC1 (Figure 4), the two best-characterized ones are functionally linked to protein synthesis (mRNA translation) and its control—i.e. S6Ks and 4E-BPs (Figure 4). S6Ks were the first to be discovered to be controlled in a rapamycin-sensitive manner (Chung et al., 1992). S6Ks comprise four proteins encoded by two genes each of which gives rise to two mRNAs and protein isoforms. S6Ks are activated by phosphorylation at several sites, many of which are dephosphorylated in response to rapamycin (Weng et al., 1998; Avruch et al., 2001) (although only Thr389 in the shorter form of S6K1 is a direct substrate of mTORC1).
Data from flies (Montagne et al., 1999) and mice (Shima et al., 1998) show that S6Ks play an important role in cell and animal growth, but it is not clear how. The original concept that they regulate rp synthesis would have provided a nice explanation—as ribosome levels determine the protein synthetic capacity of cells which is intimately linked to cell growth. rps are encoded by mRNAs that contain a stretch rich in pyrimidines at their 5′ ends. The so-called 5′-TOP (5′-terminal tract of oligopyrimidine) acts to repress the translation of these mRNAs, this inhibition being overcome by stimulating starved cells with, serum or amino acids. This stimulation of their translation is inhibited (but often only partially) by rapamycin (Tang et al., 2001; Stolovich et al., 2002). This feature and other data have led to the suggestion that 5′-TOP mRNA translation may be regulated by an mTOR complex distinct from mTORC1/2 (‘mTORC3’), although the nature of such a complex remains obscure (Patursky-Polischuk et al., 2009).
However, neither S6Ks (Pende et al., 2004) nor S6 phosphorylation (Ruvinsky et al., 2005) are necessary for the regulation of the translation of rp mRNAs, which remains sensitive to rapamycin when S6 phosphorylation is abolished. The mechanism by which mTORC1 signaling stimulates the translation of the 5′-TOP mRNAs remains an important unanswered issue.
S6Ks may regulate ribosome production by controlling the synthesis of rRNAs, which are transcribed by RNA polymerases I and III (Mayer and Grummt, 2006). However, if they do, it is puzzling that the growth (hypertrophy) of cardiac muscle cells, which involves a marked increase in ribosome numbers (Morgan et al., 1985) does not require S6Ks (McMullen et al., 2004).
S6Ks have been shown to phosphorylate about a dozen different substrates. These include rp S6, eIF4B (Raught et al., 2004; Shahbazian et al., 2006) and Pdcd4 (Dorrello et al., 2006) (both modulate the function of eIF4A, an RNA helicase implicated in helping unravel secondary structures which inhibit the translation of certain mRNAs), eukaryotic elongation factor 2 (eEF2) kinase (eEF2K) (Wang et al., 2001), which inhibits the function of eEF2, a key translation elongation factor, and SKAR (S6K1 Aly/REF-like substrate), a protein that may link the process of pre-mRNA splicing and to efficient mRNA translation (Richardson et al., 2004; Ma et al., 2008). However, the contribution of the S6K-catalyzed phosphorylation of these proteins to the overall control of protein synthesis remains an important question.
Recent data show that S6K1 is activated in response to DNA damage and both binds to and phosphorylates mdm2 (a negative regulator of p53) resulting in the induction of p53 (Lai et al., 2010). This provides a potential link between nutrients and growth signals and the control of cell survival, which may be important in oncogenesis. Conversely, there is evidence that p53 negatively regulates mTORC1 signaling, perhaps through sestrins, which are encoded by p53 target genes (Horton et al., 2002; Budanov and Karin, 2008).
The data from mouse cells in which all the phosphorylation sites in S6 have been mutated are of particular interest. Embryonic fibroblasts from these animals actually showed faster rates of protein synthesis and division than control cells, but were also smaller (Ruvinsky et al., 2005). Unlike control cells, rapamycin did not cause a decrease in the size of the S6 mutant cells. Thus, S6 phosphorylation does indeed appear to be a key link between mTORC1 and cell size. But this leaves open the key issue of the molecular mechanisms involved. It is important to note that the S6 mutant mice have smaller β-cells in their islets, and display low levels of blood insulin. These mice also display muscle weakness (Ruvinsky et al., 2009) which stems both from deficient muscle growth (muscle mass, rather than muscle differentiation) and from altered energy metabolism. These studies prompt key questions about the physiological role of S6 phosphorylation, either within the ribosome in the context of protein synthesis or perhaps, by analogy with some other rps (Warner and McIntosh, 2009), possible extra-ribosomal functions for S6.
It also remains to be established how important the phosphorylation of other substrates for mTORC1 is for the control of protein synthesis. Here, it is relevant to note that the short-term treatment of cells with rapamycin has only small effects on overall protein synthesis (Beretta et al., 1996; Feldman et al., 2009; Thoreen et al., 2009). Since S6K activity is completely blocked under such conditions, this observation suggests that the S6Ks play, at most, only a minor role in the short term control of translation.
Several S6K isoforms (the longer form of S6K1 and both variants of S6K2) possess nuclear localization signals (Avruch et al., 2001). They may therefore regulate nuclear events such as gene transcription, including rRNA synthesis (Mayer and Grummt, 2006), although there a surprising paucity of information on this.
Recent data suggest S6Ks play a role on controlling lifespan, and likely provide a link between mTOR and longevity. Mice in which S6K1 has been knocked out showed extended life spans (Selman et al., 2009). What is the mechanism? Intriguingly, such mice show changes in gene expression similar to those observed in animals subjected to caloric restriction (which can also extend life span) or in response to activation of AMPK (a sensor of cellular energy status). These findings have potentially important implications: they suggest nutritional status may affect lifespan through mTORC1 and its target S6K1, indicate that S6K1 may control the expression of genes involved in the aging process, and imply that pharmacological inhibition of mTORC1 signaling might even prolong life! It should be stressed here that other studies point to roles for additional processes downstream of (m)TORC1 in the regulation of lifespan (e.g. autophagy and translation initiation) (Blagosklonny and Hall, 2009; Hands et al., 2009; Fontana et al., 2010; Katewa and Kapahi, 2010). This is an exciting area, highly relevant to human health, and additional work is needed to unravel how mTORC1 and S6Ks affect aging and longevity.
Also potentially relevant for human health, S6Ks can negatively regulate insulin signaling through the phosphorylation of insulin receptor substrate 1 (IRS1) and which impairs PI3K/Akt signaling downstream of the insulin receptor (Harrington et al., 2004; Um et al., 2004; Shah and Hunter, 2006), although there is disagreement over which phosphorylation site(s) in IRS1 are involved (Shah and Hunter, 2006). Since mTORC1 signaling and the S6Ks are stimulated by nutrients, this loop may contribute to diet-induced obesity and insulin resistance. Consistent with this, S6K1 knockout mice show decreased susceptibility to obesity and insulin resistance when fed a high-fat diet (Um et al., 2004). S6K is a potential candidate for therapies directed at restoring insulin-sensitivity, and this merits further exploration.
4E-BPs: regulators of the translation of certain mRNAs
By binding to eIF4E, 4E-BPs inhibit its function in mRNA translation by preventing it interacting with eIF4G1/eIF4G2 (Figure 4), closely related scaffold proteins that also interact with additional proteins involved in efficient recruitment of ribosomes to the 5′ end of the mRNA such as eIF4A and eIF4B, and eIF3, which also binds the 40S ribosomal subunit, as well as the PABP, which binds the 3′ end of the mRNA, effectively circularizing it. 4E-BPs are thus translational repressors: but do they inhibit protein synthesis in general, or the translation of specific mRNAs? The available data suggest they primarily affect specific mRNAs, perhaps ones that contain a structured 5′-untranslated region (UTR) and require the helicase activity of eIF4A in order to be translated.
This array of interactions and their control by mTORC1 seem to provide a sensitive and sophisticated way to control protein synthesis. However, several observations suggest this is not a major way of controlling general protein synthesis—for example, mice knocked out for 4E-BP1 (Blackshear et al., 1997) or 4E-BP1/2 are viable (Le et al., 2007); rapamycin has little effect on protein synthesis in serum-fed cells in the short term (see above) and ablation of stimulated eIF4G/eIF4E-binding does not diminish the mTORC1-dependent activation of protein synthesis in cardiac myocytes (Huang et al., 2009). Nonetheless, 4E-BP knockout mice do show interesting phenotypes. The double knockout mice show increased body weight, in part due to increased fat and an enhanced tendency to diet-induced obesity (Le et al., 2007). However, this may reflect enhanced activation of S6Ks rather than changes in the expression of specific proteins whose translation is normally controlled by the 4E-BPs. The enhanced S6K activity may result from the absence of 4E-BPs which normally compete (through their common TOS motifs) for binding to raptor.
Interestingly, 4E-BP1 knockout mice actually show a decrease in adipose (fat tissue) (Tsukiyama-Kohara et al., 2001), likely via enhanced translation of the mRNA for the transcription factor PGC1, which in turn upregulates expression of uncoupling protein 1 (UCP1). UCP1 enhances metabolic rate, which may account for the decreased fat storage in these mice. Loss of 4E-BP1/2 also enhances the innate immune response, likely through increased translation of the mRNA for interferon regulatory factor 7 (Colina et al., 2008).
These findings from transgenic mice raise a key question: which (other) mRNAs are translationally controlled by 4E-BPs? Interest in this question is very high given the body of data showing that overexpression of eIF4E (which may then ‘escape’ control by 4E-BPs) can transform cells, create tumors in mice and is prevalent in human tumors (Lazaris-Karatzas et al., 1990; Ruggero et al., 2004; Wendel et al., 2004, 2007; Fischer, 2009). The hyperactivation of mTORC1 in tumors harboring certain oncogenic mutations or lacking functional tumor suppressors such as PTEN or NF1 can also free eIF4E from inhibition by 4E-BPs. Proliferation of certain tumor cells or growth of certain tumors can be inhibited by rapamycin and related compounds, and this effect is overcome by increasing levels of eIF4E, pointing to eIF4E as a key target for the (indirect) action of rapamycin. To identify mRNAs whose translation is controlled by eIF4E, Mamane et al. (2007) used a system for the inducible expression of eIF4E, and studied which mRNAs shifted into polysomes (i.e. underwent more efficient initiation) as eIF4E levels were increased. The translationally upregulated mRNAs included several involved in cell survival or cell proliferation.
This set of mRNAs included also many belonging to the group of mRNAs that contain a 5′-TOP (Meyuhas and Hornstein, 2000). However, it remains to be established whether this control really reflects regulation by eIF4E and the 4E-BPs—for example, by no means all these mRNAs appeared among the set upregulated by eIF4E and many have short, unstructured 5′-UTRs that would not seem to need high levels of eIF4G/eIF4E binding and the attendant recruitment of the eIF4A helicase to the mRNA. As noted above, this set of mRNAs is of particular interest not only because they have been known for more than 15 years to be controlled in a rapamycin-sensitive manner, but also because they encode all the rps and several other components of the translational machinery, so that upregulation of their translation by mTORC1 (along with increased rRNA synthesis, see below) provides a way to enhance the cell's overall capacity for protein synthesis, i.e. its ribosome levels. This is likely of crucial importance for cell proliferation and cell growth.
Santhanam et al. (2009) performed an analysis of the non-coding UTRs of the mRNAs whose translation was upregulated by increasing eIF4E levels. Their 5′-UTRs showed only modest propensities for base-pairing; given that eIF4E helps recruit the eIF4A helicase to the translation initiation complex, and this enzyme can unwind secondary structure that inhibits ribosome movement along the 5′-UTR, this is perhaps surprising. The analysis further revealed that the 3′-UTR of many of this set of mRNAs show high G + C content, especially just after the stop codon. The authors suggest that increased internal base pairing may decrease the susceptibility of mRNAs to miRNA-mediated translational repression. Further work is clearly needed to study how mTORC1 signaling, through the control of eIF4E, may modulate miRNA-mediated translational control.
PRAS40: substrate and regulator?
PRAS40 (Nascimento and Ouwens, 2009) was first identified as a protein that undergoes phosphorylation by PKB/Akt (Harthill et al., 2002; Kovacina et al., 2003; hence PRAS40 is proline-rich Akt substrate; Figures 2 and 4), which binds 14-3-3 proteins when it is phosphorylated by Akt. It was subsequently ‘rediscovered’ as a binding partner for mTORC1 (Sancak et al., 2007; Vander Haar et al., 2007), being released at high salt, apparently conferring the salt-sensitive regulation of mTORC1 referred to above (Sancak et al., 2007). This led to the conclusion that PRAS40 is a physiological regulator of mTORC1, being released in response to signals that activate Akt (such as insulin), allowing the activation of mTORC1 (Sancak et al., 2007; Vander Haar et al., 2007).
Around the same time, other studies (Fonseca et al., 2007; Oshiro et al., 2007; Wang et al., 2007) showed that PRAS40 contains a TOS motif (required for its binding to raptor) and is actually a substrate for mTORC1, which phosphorylates several sites in PRAS40. Rapamycin affects some, but not all, these sites in vivo (Oshiro et al., 2007; Wang et al., 2008a).
Since PRAS40 binds quite tightly to mTORC1 and is released upon phosphorylation, it is suggested that it negatively regulates mTORC1 by impeding the binding of other TOS motif-containing substrates (Wang et al., 2007). There remain a number of questions about the regulation and function of PRAS40. For example, some studies show that 14-3-3 binding to PRAS40 is not required for activation of mTORC1 (Sancak et al., 2007; Fonseca et al., 2008), and even that phosphorylation of PRAS40 on the Akt site is not required for this. For example, phorbol esters, which activate protein kinase C (PKC) and the ERK pathway, stimulate mTORC1 without affecting PRAS40. How do they do this? How do they evade the inhibitory action of PRAS40? Although knocking down PRAS40 expression is reported (as expected from the above model) to promote mTORC1 signaling in serum- or amino acid-starved cells (Oshiro et al., 2007; Sancak et al., 2007; Wang et al., 2007), another study reported that this manoeuvre actually impairs mTORC1 signaling suggesting PRAS40 plays a positive role in the pathway (Fonseca et al., 2007). Interestingly, reducing the levels of Drosophila Lobe, a close relative of PRAS40, also led to impaired TORC1 signaling (Wang and Huang, 2009).
Does PRAS40 function only as a regulator of mTORC1 or does it have additional roles? Several lines of evidence suggest that PRAS40 or Drosophila Lobe affect apoptosis, although reports vary as to whether PRAS40 is pro-apoptotic (knocking down its expression reduced cell death) (Thedieck et al., 2007) or anti-apoptotic (i.e. knockdown promoted cell death) (Madhunapantula et al., 2007). Consistent with the latter, overexpressing PRAS40 had neuroprotective effects (Saito et al., 2004). Lobe also appears to exert anti-apoptotic effects (Singh et al., 2006). PRAS40's effects on apoptosis may be independent of its effect on mTORC1 signaling (Thedieck et al., 2007).
It thus remains to be established whether PRAS40 is a general regulator of mTORC1 signaling and what its (other) physiological roles are, and how it performs them (through additional-binding partners?).
Rapamycin blocks G1-S progression, showing that mTORC1 controls this stage of the cell cycle. This is exerted partly through regulation of the levels of the cell cycle inhibitor p27, although additional effects are also involved (Leung-Pineda et al., 2004). These include on the expression of proteins such as cyclin D that are needed for passage into S-phase. mTORC1, and its downstream effector eIF4E, seem to control cyclin D levels through several mechanisms including effects on the levels, stability and translation of its mRNA (Rosenwald et al., 1995; Hashemolhosseini et al., 1998; Averous et al., 2008), and the stability of cyclin D protein (Hashemolhosseini et al., 1998).
A full discussion of the control of the cell cycle is beyond the scope of this review, but it is worth mentioning that several recent studies indicate that mTORC1 (or its orthologs in other species) also controls mitotic entry (Wang and Proud, 2009). Intriguingly, mTORC1 signaling seems to promote progression in some species (e.g. human cells) but retard it in others (fission yeast). Further work is needed to achieve a more complete understanding of the mechanisms by which mTORC1 signaling controls the cell cycle.
Recent data indicate that the activation state of mTORC1 signaling plays a role in the processes which lead cells in which p53 has been activated to become quiescent or to senesce (Demidenko et al., 2010; Korotchkina et al., 2010). For example, treating cells prone to senescence with rapamycin caused them to become quiescent instead. However, the molecular mechanisms involved in this interplay between mTOR and p53 in modulating of cell cycle progression and senescence remain to be elucidated.
The control of cell proliferation by mTORC1 requires the 4E-BPs, indicating it is regulated via eIF4E (in line with a role for eIF4E, coupled to mTORC1 signaling, in tumorigenesis and in regulating the translation of mRNAs for proteins involved in cell cycle progression) (Dowling et al., 2010). On the other hand, cell growth (size) was unaffected by loss of the 4E-BPs; this study also corroborated other work (discussed above) which indicates that the S6Ks are an important regulator of cell and animal growth (Shima et al., 1998; Montagne et al., 1999). This prompts the question: which substrates for S6Ks are the key ones for the control of cell growth?
Recent studies show a role for mTORC1 signaling in the control of lipid metabolism and adipogenesis. Sterol response element-binding protein 1c (SREBP1c) is a transcription factor that regulates the expression of proteins involved in lipid metabolism. A key study (Porstmann et al., 2008) showed that the activation of SREBP1c by insulin is controlled by mTORC1 (Figure 4), which promotes the cleavage and subsequent nuclear accumulation of SREBP1c and the expression of genes that it regulates. It will be important to identify how mTORC1 controls the trafficking of SREBP1c. mTORC1 also controls the expression of the mRNA for SREBP1c (Li et al., 2010), but again the mechanism is not known—it appears not to involve the S6Ks.
Another very recent report, using an adipocyte-like cell line (3T3-L1) shows that activating mTORC1 (by overexpressing Rheb) suppresses the levels of lipolytic enzymes and represses lipolysis, while activating lipogenesis and lipid storage (Chakrabarti et al., 2010), consistent with the data just mentioned. In contrast to its effects on lipogenesis, mTORC1 does not appear to regulate gluconeogenesis (Li et al., 2010).
These findings are important for elucidating the links between nutrition, signaling and conditions such as type II diabetes and obesity; the use of animals models in which components of the mTORC1 pathway have been modified will be instructive both in terms of understanding the underlying physiology and for the development of new therapeutic approaches.
mTORC1 appears to control gene-specific transcription in multiple ways. We have just seen that it activates SREBP1c. It can also potentially control PGC1α through the control of the translation of its mRNA by 4E-BP1 (Tsukiyama-Kohara et al., 2001). PGC1α interacts with another transcriptional regulator, YY1 (Figure 4), acting as a transcriptional co-activator for genes encoding proteins important in mitochondrial functions including respiration. The co-activation of YY1 by PGC1α is inhibited by rapamycin (Cunningham et al., 2007), leading to repression of many mitochondrial genes. Furthermore, these authors showed that mTOR/raptor interact with YY1 at the promoters for such genes. This provides a mechanism by which mTORC1 can promote energy (ATP) supply presumably to support energy-requiring processes such as lipogenesis, protein synthesis, cell growth and cell proliferation. This study also implies that mTORC1 functions within the nucleus to regulate gene transcription, although it is not known whether mTORC1 directly phosphorylates YY1 or PGC1α (Figure 4).
mTOR can also control the expression of HIF1α (Bernardi et al., 2006), while this pro-angiogenic transcription factor also interacts with mTORC1 (Figure 4). Thus, mTORC1 can act in multiple ways to control the transcription of specific mRNAs.
Most RNA synthesis in dividing cells is the production of rRNA. mTORC1 also controls this process (Mayer and Grummt, 2006) (Figure 4), which involves two distinct RNA polymerases, Pol I (which makes the large rRNAs, 28S, 18S and 5.8S) and Pol III, which produces the 5S rRNA. Together with the mTORC1-dependent translation of 5′-TOP mRNAs encoding rps (Figure 4), this allows coordinated control of ribosome biogenesis by mTORC1 signaling. It should be noted that, in contrast to the situation in budding yeast, mTORC1 does not control the transcription of rp genes. How does mTORC1 regulate Pol I and Pol III?
Pol I requires the accessory proteins SL1, UBF and TIF-1A (also called Rrn3). Early data suggested TIF-1A played a key role in the control of Pol I by mTORC1 (i.e. in its inhibition by rapamycin) (Mayer et al., 2004). Rapamycin causes inhibition of the interaction between TIF-1A and Pol I/SL1, leads to its exit from the nucleolus to the cytoplasm and alters the phosphorylation of two sites in TIF-1A in different directions. It is not known which kinases/phosphatases link the reciprocal control of these sites to signaling from mTORC1. UBF1 is also a target for mTORC1 signaling, being a substrate for S6K1, which positively regulates its function (Hannan et al., 2003). The relatively mild phenotype of S6K double-knockout mice suggests that S6Ks do not play an essential role in ribosome biogenesis.
Recently, mTOR has been reported to be present at the Pol I promoter (Tsang et al., 2010). Interestingly, in yeast rapamycin also affects nucleolar epigenetic marks and nucleolar morphology (Tsang et al., 2003). In contrast, very little is known about the role of histone or DNA modifications in the control of nucleolar events in mammalian cells, or about the role of mTORC1 signaling in the processing of rRNA or other aspects of ribosome assembly in which it might be anticipated to play a coordinating role.
mTOR was also found at the Pol III promoter (Tsang et al., 2010). Very recent studies found that the Pol III-specific transcription factor complex, TFIIIC, associates with mTOR, likely through its subunit TFIIIC63, which contains a TOS-like motif (Kantidakis et al., 2010). This suggests that TFIIIC could be a substrate for mTORC1; however, Maf1, rather than TFIIIC63, shows rapamycin-sensitive phosphorylation and is a direct substrate for mTORC1 (Kantidakis et al., 2010; Michels et al., 2010). Maf1 is a repressor of Pol III. They identify one rapamycin-sensitive site in Maf1 (Ser75) but the data suggests there are others. In yeast, activation of TORC1 leads to nuclear exclusion of Maf1 and activation of Pol III (Willis and Moir, 2007). In contrast, mTORC1 does not seem to affect the nuclear localization of mammalian Maf1. Indeed, Maf1 remains associated with the Pol III machinery even after activation of mTORC1 signaling. Thus, although Maf1 performs analogous overall functions in yeast and mammals (repression of Pol III), both its regulation and its mode of action differ.
In addition to activating biosynthetic processes (protein, ribosome and lipid synthesis), mTORC1 signaling also inhibits an important catabolic process, autophagy (Jung et al., 2010). Autophagy is also controlled by TOS in yeast, and the mechanisms involved are currently better understood in that system than in mammalian cells. In yeast, TOR phosphorylates Atg1, a protein kinase which seems to play an upstream role in the control of autophagy. Atg1 binds several other proteins including Atg13, which is required for the induction of autophagy. It seems that phosphorylation of Atg1 (downstream of TOR) decreases the binding of Atg1 to Atg13, impairing autophagy. Atg13 also appears to be phosphorylated by TOR, and this seems to repress its function. Conversely, when TOR is inactivated, the hypophosphorylated forms of Atg1 and Atg13 work, along with Atg17, to promote autophagosome formation.
In mammals, several proteins are related to Atg1, i.e. ULK1-4. On the basis of data from overexpression of wildtype or kinase-dead mutants, ULK1-3 seem to play roles in inducing autophagy (reviewed in Jung et al., 2010). There appears to be an ortholog of Atg13 in mammals, and an additional protein, FIP200, also associates with Atg13 and ULK1. Furthermore, mTORC1 phosphorylates the Atg1 homologs ULK1/2 (Ganley et al., 2009; Hosokawa et al., 2009; Jung et al., 2009) (Figure 4) and this appears to repress their activities. However, other aspects of the regulation of the complex differ in mammals—for example, inhibition of mTORC1 does not affect ULK1/Atg13 binding. Understanding of the control of autophagy in mammalian cells is still very far from complete and, as described by Jung et al. (2010), many questions remain.
There is a further potential link between mTORC1 signaling and autophagy, provided by eEF2K, since mTORC1 negatively regulates eEF2K (Redpath et al., 1996; Proud, 2007) and several reports suggest eEF2K is a positive regulator of autophagy (Wu et al., 2006, 2009; Py et al., 2009; Cheng et al., 2010). However, it remains to be established how eEF2K is linked to the control of autophagy. For example, what are the relevant substrates of eEF2K—or indeed of ULK1/2? Is eEF2K's effect on autophagy direct, or is it somehow connected to the control of translation elongation?
The ever-growing body of information about mTORC1 signaling continues to underscore its importance for cellular physiology and, as a consequence, in disorders such as cancers, diabetes, obesity and perhaps others such as neurodegenerative disease (perhaps through its effects on autophagy) (Meijer and Codogno, 2009). Since rapamycin, and to an even greater extent, the new mTOR kinase inhibitors, block multiple cell functions that are linked to mTORC1 (and in the second case also mTORC2), it will be important to gain a full understanding of the molecular links between mTOR and cellular functions that are defective in human disease. This could allow the development of therapeutic strategies that target the specific processes downstream of mTORC1 that are involved in a given state, rather than using the ‘blunderbuss’ approach of blocking mTORC1 signaling in general.
A brief mention of mTORC2
In comparison to the wealth of information on mTORC1, knowledge of the functions and control of mTORC2 is more limited. This largely reflects the absence of a specific tool like rapamycin with which to study this complex. Embryos lacking rictor stop growing around E11.5 but are viable long enough to create rictor-null cells (embryonic fibroblasts) (Shiota et al., 2006); this contrasts with knocking out raptor, which results in lethality early in development (Guertin et al., 2006). The main substrates identified for mTORC2 are the Akt (protein kinase B, PKB) group of protein kinases and their close relatives the SGKs (Garcia-Martinez and Alessi, 2008).
The PKBs and SGK1 are activated by PI3K signaling. mTORC2 phosphorylates Akts/SGK1 at a threonine within a hydrophobic motif that corresponds to Thr389 in S6K1, a substrate for mTORC1 (Hresko and Mueckler, 2005; Sarbassov et al., 2005). Unlike PKBs, SGK1 does not contain a PH domain which in the case of PKB can bind the product of PI3K, PtdIns(3,4,5)P3, so an alternative mechanism must account for the activation of SGK1 by PI3K. This implies that PI3K somehow activates mTORC2 (Garcia-Martinez and Alessi, 2008) but it is not clear whether this is mediated through a change in mTORC2's activity, through the co-localization of Akt/SGK1 with mTORC2 or another mechanism.
S6Ks, Akts and SGK1 are all members of the AGC family of kinases, all of which contain ‘hydrophobic sites’. mTORC2 also phosphorylates and regulates PKCα (Ikenoue et al., 2008) but it is not yet clear which other AGC kinases mTORC2 phosphorylates physiologically. The availability of rictor-null cells and of specific inhibitors of mTOR's kinase activity (which inhibit mTORC2 and mTORC1) should help reveal the full array of cellular functions of mTORC2.
The last 10 years have witnessed enormous advances in our understanding of mTOR, its function and control. These include the discovery of the mTOR binding partners raptor and rictor, and of the control of mTORC1 by TSC1/2 and Rheb. Interest in this area has also risen dramatically, with almost 5500 articles concerning mTOR having appeared since 2000! Nevertheless, as described above, many important questions remain. For example, it is not yet clear where in the cell mTORC1 is located, how Rheb activates mTORC1 and what the functional meaning of ‘activation of mTORC1’ actually is.
Although recent work has revealed components involved in the control of mTORC1 by amino acids, it is not yet understood how amino acid levels are detected to control mTORC1. Surprisingly, it is still not clear how rapamycin interferes with mTORC1 function, or why it only affects some of the actions of mTORC1. Downstream of mTORC1, we do not yet how the S6Ks affect animal growth, or how mTORC1 controls processes such as 5′-TOP mRNA translation, SREBP1c and gene transcription. Finally, a key issue about mTORC2 concerns how its activity is controlled.
Work in this laboratory on mTOR signaling is supported by grants from the Biotechnology & Biological Sciences Research Council, the Wellcome Trust, the Royal Society, the European Union, and the British Heart Foundation, and by funding from AstraZeneca UK, Janssen Pharmaceutica and Hypha Discovery (all to C.G.P.).
Conflict of interest: none declared.