Amino acids regulate signalling through the mTORC1 (mammalian target of rapamycin, complex 1) and thereby control a number of components of the translational machinery, including initiation and elongation factors. mTORC1 also positively regulates other anabolic processes, in particular ribosome biogenesis. The most effective single amino acid is leucine. A key issue is how intracellular amino acids regulate mTORC1. This does not require the TSC1/2 (tuberous sclerosis complex 1/2) complex, which is involved in the activation of mTORC1, for example, by insulin. Progress in understanding the mechanisms responsible for this will be reviewed.
Approximately half the different types of amino acids are essential in humans, as we cannot manufacture them ourselves. Since amino acids are required for protein biosynthesis, it is logical that amino acids should control components involved in the regulation of protein synthesis (mRNA translation).
mRNA translation is a complex process requiring a set of proteins (‘translation factors’) that are extrinsic to the ribosome. The functions of several translation factors are indirectly controlled by amino acids. In eukaryotic cells, two main mechanisms have been identified by which amino acids can control translation factor function, although our recent results point to the existence of a third distinct mechanism.
The first mechanism involves the phosphorylation of the α-subunit of eIF2 (eukaryotic initiation factor 2) and will be discussed in another article in this issue  (see pp. 1205–1207). Phosphorylated eIF2 inhibits the activity of the cognate GEF (guanine-nucleotide-exchange factor), eIF2B, which converts inactive eIF2·GDP into active eIF2·GTP. Phosphorylation of eIF2 thus inhibits its function, which is to recruit the initiator tRNAMet (methionyl tRNA) to the ribosome to recognize the start codon [1a].
Four distinct eIF2α kinases are known. They are activated in response to differing stress conditions. In yeast, Gcn2p is switched on in response to amino acid starvation, leading to phosphorylation of eIF2 and repression of general protein synthesis . This will conserve scarce amino acids. The role of mammalian Gcn2 in controlling protein synthesis in response to changes in amino acid availability has not yet been studied very intensively and will not be discussed in further detail here, except to note that our recent results indicate that amino acid deprivation can regulate eIF2B activity independently of changes in eIF2α phosphorylation, probably via changes in the phosphorylation of eIF2B itself (X. Wang and C.G. Proud, unpublished work).
The second mechanism by which amino acids regulate the translational machinery of mammalian cells involves mTOR [mammalian TOR (target of rapamycin)]. The rest of this article will be devoted to a discussion of mTOR and its regulation by amino acids.
Signalling through mTOR controls the protein synthetic machinery in multiple ways
mTOR is a large polypeptide with several distinct functional domains: these include (i) a kinase domain, which resembles lipid kinases of the phosphoinositide kinase family, but which actually phosphorylates proteins; (ii) several HEAT (Huntingtin, elongation factor 3, the PR65/A subunit of protein phosphatase 2A and the lipid kinase Tor heat-treatment) repeats, which are likely to be involved in protein–protein interactions, and (iii) a domain that binds the immunophilin FKBP12 (FK506-binding protein 12; FK506 is an immunosuppressant macrolide) when it is complexed with the immunosuppressant rapamycin . mTOR binds several other proteins and forms two major types of complex, mTORC1 (mTOR, complex 1) and mTORC2 . The function of mTORC1, but not mTORC2, is inhibited by rapamycin. Since mTORC1 is regulated by amino acids, whereas mTORC2 is not known to be controlled by them, the remainder of this discussion focuses on mTORC1.
mTORC1 controls the phosphorylation of several components of the translational machinery  and also regulates ribosome biogenesis  (Figure 1). The mechanisms involved in the former effects are much better understood than the latter. The control of ribosome biogenesis by mTORC1 includes effects on the synthesis of ribosomal RNA and ribosomal proteins . Cytoplasmic ribosomal proteins are encoded by mRNAs that belong to a set of mRNAs that possess 5′-TOP (5′-tracts of oligopyrimidines) (5′-TOP mRNAs) . The 5′-TOP confers control of the translation of these mRNAs, such that they are recruited into larger polysomes in response to stimulation of cells with, e.g. serum, an effect that is inhibited by rapamycin. However, it is so far unclear how the 5′-TOP modulates the translation of these mRNAs or how their translation is controlled by mTORC1.
Amino acids activate mTORC1 to promote protein synthesis and other anabolic processes
It was formerly thought that the ribosomal protein S6Ks (S6 protein kinases) regulated 5′-TOP mRNA translation. The activation of these kinases (S6K1 and S6K2) by agents such as insulin is completely blocked by rapamycin . Results from cells in which both S6K genes have been knocked out  or cells in which all five phosphorylation sites in S6 have been changed to alanine  convincingly show that neither S6Ks nor S6 phosphorylation is required for control of 5′-TOP mRNA translation.
Control of translation factors by mTORC1
mTORC1 regulates proteins involved in the initiation or elongation stages of mRNA translation [3,4]. The best understood of these is eIF4E, the protein that binds the 5′-cap structure of the mRNA and co-operates with its partner, the scaffold protein eIF4G, to bring the 40 S subunit to the mRNA [1a,10]. eIF4E also binds small heat-stable proteins termed 4E-BPs (eIF4E-binding proteins). They inhibit the binding of eIF4E to eIF4G and thus the formation of the eIF4F initiation complex, which contains eIF4E, eIF4G and the RNA helicase eIF4A. eIF4A is thought to unwind secondary structure in 5′-UTRs (5′-untranslated regions) of mRNAs to allow unimpeded scanning by the 40 S subunit and associated proteins to locate the start codon. Such unwinding would be especially important for translation of mRNAs whose 5′-UTRs contain extensive secondary structure. eIF4B, which assists eIF4A, is a substrate for the S6Ks. Phosphorylation of eIF4B increases its interaction with eIF3, which is involved in recruiting the 40 S subunit to the mRNA .
The best understood of the 4E-BPs, 4E-BP1, undergoes phosphorylation at several sites, at least four of which are controlled by mTOR. These are (in human 4E-BP1) Thr37, Thr46 and Thr70, and Ser65. The latter two sites inhibit the association of 4E-BP1 with eIF4E, while phosphorylation of Thr37/Thr46 is a prerequisite for the phosphorylation of Ser65 and Thr70 . mTOR-mediated phosphorylation of 4E-BP1 makes eIF4E available to bind eIF4G and promotes formation of functional initiation complexes. Interestingly, phosphorylation of Thr37/Thr46 is controlled primarily by amino acids rather than by agents such as insulin [13,14], and is not affected much by rapamycin. In contrast, the phosphorylation of Ser65 requires both amino acids and insulin and is blocked by rapamycin. This may reflect the operation of different types of output from mTORC1 .
The S6Ks and 4E-BPs possess short sequences termed TOS (TOR-signalling) motifs, which interact with the mTORC1 component raptor and recruit these mTORC1 substrates to mTORC1 [15,16]. 4E-BP1 also contains a distinct ‘RAIP’ motif (named for its sequence). This is essential for phosphorylation of 4E-BP1 at all the sites mentioned here and apparently mediates the amino acid-dependent input to 4E-BP1 [14,17].
mTORC1 also modulates the translation elongation machinery through the control of the phosphorylation of eEF2 (eukaryotic elongation factor 2). When phosphorylated at Thr56, eEF2 is inactivated and cannot perform its normal function of facilitating the translocation stage of elongation (during which the ribosome moves by one codon relative to the mRNA). mTORC1 signalling promotes the dephosphorylation of eEF2, thereby activating the elongation machinery . This effect involves the inactivation of eEF2 kinase, a very unusual protein kinase . eEF2 kinase is regulated by phosphorylation. mTORC1 controls the phosphorylation of at least three inhibitory sites in eEF2 kinase (Ser78, Ser359 and Ser366 in human eEF2 kinase) . One (Ser366) is phosphorylated by S6K but it is not known how the others are linked to mTORC1.
mTORC1 signalling is regulated by amino acids
Almost 10 years ago, it became clear that signalling through mTOR is regulated by amino acids. (The existence of two types of mTOR complex was only recognized later, and it now appears that amino acids regulate mTORC1.) Starving cells for amino acids quickly (minutes) impairs mTORC1 signalling and makes it refractory to stimulation by agents such as insulin [20,21].
In most cell types, the essential amino acid leucine appears to be the most important one for control of mTORC1 although other amino acids can also affect mTORC1 signalling. It appears that mTORC1 signalling is regulated by intracellular amino acids , implying the existence of a mechanism for sensing intracellular levels of amino acids such as leucine (rather than one that detects amino acids in the medium).
We have seen that a major function of mTORC1 signalling is to promote the activation of the initiation and elongation phases of mRNA translation. Since translation consumes amino acids (nearly all of which are used in elongation), it makes physiological sense for amino acids to positively regulate a major signalling pathway that controls this process.
How do amino acids regulate mTORC1?
There is substantial evidence suggesting that agents such as insulin activate mTORC1 via a signalling pathway that involves the TSC (tuberous sclerosis complex) (which contains two proteins TSC1 and TSC2; ). TSC2 acts as the GAP (GTPase-activating protein) for the small G-protein Rheb (Figure 2). Rheb-GTP can activate mTORC1's kinase activity . TSC2 therefore negatively regulates mTORC1. Its phosphorylation by PKB (protein kinase B; also called Akt) inhibits its GAP activity against Rheb-GTP (see ).
Components implicated in the upstream control of mTORC1
Overexpressing Rheb switches on mTORC1 signalling in amino acid-starved cells and allows its further stimulation by agents such as insulin, suggesting that inactivation of Rheb, or events downstream of it, underlies the inhibition of mTORC1 signalling seen upon amino acid starvation. This could in principle be due to the activation of TSC1/2 by amino acid starvation: perhaps leucine (etc.) inhibits TSC1/2?
If amino acids did act by restraining the Rheb-GAP activity of TSC1/2, then mTORC1 signalling should not be impaired by withdrawal of amino acids in TSC2−/− cells. However, it is clear that taking away amino acids still impairs mTORC1 signalling in TSC2−/− cells [26,27]. It remains unclear how amino acids actually regulate mTORC1, but a number of candidate components or mechanisms have been identified. For example, Long et al.  showed that amino acid starvation impaired the binding of Rheb to mTORC1. However, it is not yet clear how the Rheb–mTOR interaction is controlled.
For most small G-proteins, their guanine-nucleotide-binding status modulates their binding to downstream effectors, such that only the GTP-bound form interacts with their effectors. However, two lines of evidence suggest that amino acids do not control the Rheb–mTORC1 interaction through changes in Rheb-GTP levels: (i) amino acid starvation has, at most, only a small and transient effect on the GTP loading of Rheb and (ii) the binding of Rheb to mTOR is independent of its GTP/GDP-binding status [26–28].
Activation of mTORC1 signalling by amino acids is blocked by inhibitors of PI3K (phosphoinositide 3-kinase) such as wortmannin. This led Nobukuni et al.  to study the role of the class 3 PI3K Vps34 (vacuolar protein sorting 34) in the control of mTORC1 by amino acids. Vps34 regulates endosomal vesicle trafficking: amino acids promote the accumulation of PI3P (phosphatidylinositol 3-phosphate) in endosomes. Furthermore, ectopic expression or knockdown of Vps34 affects mTORC1 signalling such that Vps34 appears to be a positively acting component in its control by amino acids. However, it remains to be established both how amino acids control Vps34 and how Vps34 (presumably via PI3P) regulates mTORC1. Interestingly, Vps34 is also implicated in the regulation of autophagy, which in turn is controlled by mTORC1 signalling.
A recent RNAi (RNA interference) screen in Drosophila cells identified the homologue of the mammalian protein kinase MAP4K3 (mitogen-activated protein kinase kinase kinase kinase 3) as a positive regulator of TOR signalling . In human cells, knockdown of MAP4K3 impairs mTORC1 signalling, whereas overexpression enhances it. MAP4K3 activity is impaired by starving cells of amino acids and enhanced by adding them back. The latter effect is insensitive to rapamycin, indicating that MAP4K3 is likely to be upstream (not downstream) of mTORC1. Again, it remains to be established how MAP4K3 activity is modulated by amino acids and how it acts to promote mTORC signalling.
Lastly, TCTP (translationally controlled tumour protein) was recently reported to act as a positive upstream regulator of TOR signalling in Drosophila and to catalyse GTP/GDP exchange on Rheb . It is not yet clear whether it is involved in the control of mTORC1 by amino acids.
mTORC1 signalling and other anabolic processes
mTORC1 signalling is positioned to integrate signals from nutrient availability and from anabolic (insulin) or mitogenic agents (growth factors). One might anticipate that mTORC1 would promote other biosynthetic processes in addition to protein synthesis. Indeed, we have already seen that it does: it positively regulates ribosome biogenesis . This is a longer-term way of increasing cells’ capacity for mRNA translation and complements the ability of mTORC1 quickly to activate pre-existing components of the protein synthetic machinery.
GSK3 (glycogen synthase kinase 3) inhibits glycogen synthase, the main regulator of glycogen deposition in mammals. Since GSK3 can be phosphorylated and inhibited by S6Ks [32,33], amino acids could promote glycogen synthesis, although this has not been explored in detail.
The ability of mTORC1 to control rDNA gene transcription has already been mentioned, and it is probable that mTORC1 signalling also controls many other genes. This link would provide an input from amino acids, but the mechanisms by which mTORC1 actually regulates nuclear events such as gene transcription are unclear.
Owing to space constraints, readers have been directed to review articles wherever possible: apologies are due to the authors of the primary research papers that we could not cite directly.
Metabolism: A Focus Topic at Life Sciences 2007, held at SECC Glasgow, U.K., 9–12 July 2007. Edited by M. Case (Manchester, U.K.), S. Eaton (University College London, U.K.), P. Newsholme (University College Dublin, Ireland), S. Shirazi-Beechey (Liverpool, U.K.) and C. Sutherland (Dundee, U.K.).
eukaryotic initiation factor
eukaryotic elongation factor
glycogen synthase kinase 3
mitogen-activated protein kinase kinase kinase kinase 3
mammalian target of rapamycin
- mTORC1 mTOR
S6 protein kinase
translationally controlled tumour protein
5′-tracts of oligopyrimidines
target of rapamycin
tuberous sclerosis complex
vacuolar protein sorting 34
Work in my laboratory on amino acid-regulated signalling is supported by funding from the Ajinomoto Amino Acid Research Program and the Canadian Institutes for Health Research.