The mammalian target of rapamycin (mTOR) signaling pathway is a master regulator of cell growth throughout eukaryotes. The pathway senses nutrient and other growth signals, and then orchestrates the complex systems of anabolic and catabolic metabolism that underpin the growth process. A central target of mTOR signaling is the translation machinery. mTOR uses a multitude of translation factors to drive the bulk production of protein that growth requires, but also to direct a post-transcriptional program of growth-specific gene expression. This review will discuss current understanding of how mTOR controls these mechanisms and their functions in growth control.
In simple terms, cell growth is the accumulation of mass. But this description short changes a process that is vastly more complex and interesting. Cells that are induced from quiescence into growth undergo extreme changes in the demand for biomolecule synthesis, energy production, and nutrient uptake that necessitate the reprogramming of nearly all aspects of cellular metabolism. Throughout eukaryotes, the mammalian target of rapamycin (mTOR) signaling pathway acts a master regulator of this switch. The pathway integrates growth signals (e.g. nutrients and growth factors) and then transmits them to the growth machinery . The biological outputs of mTOR signaling are diverse, ranging from cell and tissue growth to stem cell function and aging. mTOR deregulation is also increasingly linked to human diseases, including a variety of cancers, but also a collection of neurodevelopmental disorders linked to autism and severe epilepsies . This review will examine the relationship between mTOR and the mRNA translation machinery, a system that is essential for the production of new cellular material but also used to evoke a program of gene expression that orchestrates the growth process and other mTOR functions.
The mTOR signaling pathway
The discovery of the mTOR pathway began with the isolation of rapamycin, an antifungal macrolide produced by the Streptomyces hygroscopicus bacteria. Genetic screens in yeast soon identified the targets of this cytostatic compound as the TOR1 and TOR2 kinases [3,4]. Biochemical studies identified the mammalian homolog, mTOR, not long after [5–7]. Since these seminal studies, a more complete understanding of upstream and downstream architecture of the pathway has emerged . These findings are discussed in depth elsewhere , but principal insights are described here. mTOR itself is a large serine/threonine kinase that functions as the catalytic subunit of two distinct multiprotein complexes called mTOR complexes 1 and 2 (mTORC1/2). mTORC1 is defined by the large scaffolding protein Raptor (regulatory-associated protein of TOR), whereas mTORC2 is defined by the protein Rictor (rapamycin-insensitive companion of TOR). Both complexes additionally contain the small protein mLST8 (mammalian lethal with sec-13 protein 8). Only mTORC1 is acutely inhibited by rapamycin, whereas mTORC2 can be impaired following chronic treatment through interference with complex assembly . The activity of both complexes depends on growth factor signaling through the phosphoinositide 3-kinase pathway, while mTORC1 is additionally controlled by nutrient and stress signals. Both mTORCs function in growth control, but the connections to the protein synthesis machinery are primarily transmitted through mTORC1. This review will focus on the translation-related functions of mTORC1, although there is also evidence of communication between ribosomes and mTORC2 [9,10].
Clues that functions of TOR signaling might be intertwined with the translation machinery were evident early on. In yeast, for instance, rapamycin treatment was found to cause general protein synthesis to plummet by nearly 85% within 60 min and impair ribosome biogenesis at multiple levels, including rRNA processing and transcription of ribosomal protein mRNAs [11,12]. At first, it was unclear whether mTOR influence over the translation machinery was as significant in mammalian cells, where the effect of rapamycin on protein synthesis is often mild and highly variable. This inconsistency now appears to be an artifact of the unusual allosteric mechanism that rapamycin employs. Previous work has revealed that phosphorylation of many bona fide mTORC1 substrates is strikingly resistant to rapamycin, including important translation regulators [13–15]. Accordingly, mTOR inhibition with either a newer class of ATP-competitive inhibitors (e.g. Torin1, PP242, Ku0063794, and WYE-354) or by genetic depletion causes complete dephosphorylation of rapamycin-resistant substrates and an ∼60% decrease in general protein synthesis [14,16]. It is not entirely understood why mTOR substrates vary in their sensitivity to rapamycin, but may reflect differences in the ability of mTORC1 to phosphorylate residues in different sequence contexts . Control of protein synthesis has nonetheless emerged as a highly conserved feature of the pathway.
The eIF4F cap-binding complex
Within the translation machinery, the most prominent mTORC1 target is the eIF4F cytoplasmic cap-binding complex. This multiprotein complex consists of the eIF4E cap-binding protein, the large eIF4G scaffold, and a small DEAD-box helicase called eIF4A . Its defining feature is its capacity to bind the mRNA 5′ 7-methylguanosine (m7GTP) cap, a structure that is co-transcriptionally appended to essentially all mRNAs. Once bound, eIF4F uses eIF4A to unwind local mRNA secondary structure. The 43S preinitiation complex (PIC) is then recruited to this RNA ‘landing pad’ through a mechanism that is still incompletely understood . Two features of eIF4F give it unique influence over translational fate of mRNAs. First, the cytoplasmic concentration of eIF4F is thought to be quite low . Under non-stress conditions, eIF4F binding to mRNA and recruitment of the PIC are considered rate-limiting steps, so that changes in eIF4F activity can, in principle, affect the translation of all mRNAs. Secondly, the mRNA cap structure is a privileged binding site that can only accommodate one protein at a time. eIF4F is therefore in competition with other cap-binding factors. For example, the eIF4E homologous protein (4E-HP/eIF4E2), which also specifically recognizes the cap, interacts with several distinct binding partners (Bicoid and GIGYF2) to control the translation of specific mRNAs [20,21]. eIF4F may also compete with abundant general RNA-binding proteins . Thus, the phenotypic outcomes of eIF4F regulation may vary depending on the constellation of other cap-binding factors that are also expressed.
mTORC1 regulation of eIF4F
mTORC1 can independently control eIF4F through several distinct but overlapping mechanisms. One of these is mediated by a family of translation repressors called eIF4E-binding proteins (4E-BPs). Active mTORC1 phosphorylates a series of residues that maintain 4E-BPs in an unbound and unstructured state . Inactivation of mTOR, however, causes rapid dephosphorylation of 4E-BPs, which then bind to eIF4E in a configuration that prevents its interaction with eIF4G. Loss of the eIF4E–eIF4G interaction impairs recruitment of the PIC and also reduces eIF4E affinity for the cap structure . The result is a severe reduction in the cellular concentration of eIF4F available to initiate translation. Each of the 4E-BPs is thought to function identically, but have different although overlapping expression patterns. 4E-BP1 is expressed ubiquitously. 4E-BP2 is the predominant isoform in the brain, although it is expressed elsewhere as well . 4E-BP3 is also widely expressed, but can also be transcriptionally induced by mTOR inhibition to maintain eIF4F inhibition under conditions of chronic mTOR inhibition . Dephosphorylated 4E-BPs can also bind to 4E-HP, albeit with lower affinity. The functional significance of this interaction is not yet clear .
mTORC1 additionally controls eIF4A helicase activity through a separate set of mechanisms that are mediated by family of mTOR substrates called S6 kinases (S6Ks). These kinases directly phosphorylate eIF4B and PDCD4, two proteins that independently alter eIF4A activity. Phosphorylation of eIF4B increases its capacity to stimulate eIF4A function . PDCD4 is an eIF4A inhibitor that, when phosphorylated, is targeted for ubiquitin-dependent degradation . During severe mTORC1 inhibition, these mechanisms may be redundant with 4E-BP activation, which dissociate eIF4G and eIF4A from the mRNA 5′ terminus altogether. However, the activity of S6Ks is more sensitive to mTORC1 activity than 4E-BPs, such that these mechanisms may be important mediators of more modest fluctuations in mTORC1 activity [14,16].
mRNA targets of eIF4F
While control of eIF4F can affect the general rate of protein synthesis, it can also modulate the translation of specific mRNAs. One line of evidence indicates that mRNAs with highly structured 5′ UTRs are particularly dependent on eIF4F for efficient translation . According to this model, these mRNAs require helicase activity of eIF4A to unravel secondary structure near the 5′ cap, which would otherwise interfere with PIC recruitment. Classical examples of these ‘eIF4E-sensitive’ mRNAs include G1/S cyclins, c-myc, and ornithine decarboxylase, all of which serve central functions in cell proliferation [29,30].
Recently, several studies have applied transcriptome-scale techniques to comprehensively search for eIF4A-sensitive mRNAs using silvestrol, a member of the rocaglate family of eIF4A inhibitors [31,32]. In general, these studies support an increased requirement for eIF4A activity in the efficient translation of mRNAs with long structured 5′ UTRs and possibly those containing highly stable G-quadruplex structures . However, it now appears that rocaglates interfere with initiation through a more complex mechanism than simple eIF4A inhibition . The interpretation of studies using these inhibitors might therefore be less straightforward than originally appreciated. In contrast, a recent translation profiling study that instead used temperature-sensitive eIF4A alleles in yeast found few mRNAs that were hyper-dependent on eIF4A, and little correlation with 5′ UTR structure or length . eIF4B, the eIF4A activator, does preferentially enhance the translation of mRNA with long 5′ UTRs, but this relationship may reflect eIF4A-independent functions . Thus, there remains some uncertainty over the precise features of endogenous mRNAs that confer hypersensitivity to eIF4A functions.
5′ Terminal oligopyrimidine mRNAs
A second class of mTOR-regulated mRNAs that are at least in part controlled through eIF4F are defined by a 5′ terminal oligopyrimidine (TOP) motif. This interesting motif begins with a C or U immediately following the cap structure, followed by 4–14 pyrimidines . A C in the +1 position is unusual in mRNAs, occurring in only ∼17% of transcripts, whereas ∼50% of transcripts initiate with A . A +1 U is even less common, occurring in only ∼10% of transcripts. While the length of the pyrimidine sequence varies, a C (and possibly a U) in the +1 position appears to be an absolute requirement for translational regulation of these transcripts . It is not certain how many mRNAs possess this motif. Conservative estimates are pegged at ∼100 . These mRNAs almost exclusively encode translation factors, including all ribosomal proteins, elongation factors, and polyA-binding proteins (PABPs), and have mostly been validated by individual inspection. At the opposite extreme are bioinformatic analyses of transcription start sites that suggest up to 10% of the transcriptome might bear these motifs, although these transcripts are mostly untested . The true number is probably somewhere in between (and may vary between cell types), but will require integration of translation measurements with experimentally determined 5′ end sequences to resolve. Other recent studies, including our own, identified many mRNAs with pyrimidine-rich 5′ motifs that may also be recognized by the TOP repressive mechanism and encode other growth-related proteins (e.g. chaperones and cytoskeletal proteins) [15,41].
The most striking feature of TOP mRNAs is that their translation is exquisitely sensitive to growth signals. Nutrient deprivation, various forms of stress (e.g. hypoxia), or cell-cycle arrest rapidly shifts these mRNAs into a translationally repressed state . Two observations indicate that mTORC1 is the primary regulator of TOP mRNA translation: first, ATP-competitive mTOR inhibitors profoundly and specifically repress TOP mRNA translation [15,31,41,42]. Rapamycin also inhibits TOP mRNA translation, but with variable efficacy that probably reflects a role for mTOR-regulated but rapamycin-resistant substrates . Secondly, constitutive activation, either by loss of the upstream repressor TSC1/2 or overexpression of the mTOR activator Rheb, is sufficient to maintain TOP mRNA translation under conditions that would normally cause their repression [44,45]. These effects are almost certainly mediated through mTORC1, as there is scant evidence that mTOR participates in a third Rheb-regulated mTOR complex. Some reports have found that genetic depletion of Raptor has little effect on TOP translation, but these observations are more likely due to incomplete loss of Raptor or feedback compensation that dampens the TOP regulatory mechanism over the relatively long time required to deplete Raptor levels . mTORC1/2-independent functions that perturb TOP mRNA translation may yet be identified, but the current evidence fails to support this unlikely possibility.
What controls TOP mRNA translation?
Precisely how mTORC1 controls TOP mRNA translation remains a mystery. We and others previously showed that 4E-BPs were required for the acute repression of TOP mRNAs following mTOR inhibition with ATP-competitive inhibitors [15,41]. Consistent with these observations, single-cell measurements of TOP mRNA translation show that repression tracks closely with dephosphorylation of 4E-BPs, but not other mTOR substrates . In some respects, this is hardly surprising. eIF4F binds to mRNAs at the exact 5′ location of the TOP motif, and eIF4F components can be cross-linked to cap-proximal nucleotides (Figure 1) . A simple hypothesis is that eIF4F itself detects the TOP sequence. Under this model, eIF4F would somehow interact more weakly with the 5′ ends of TOP mRNAs, whose translation would thus be more sensitive to the cellular concentrations of eIF4F. Indeed, we previously showed that mTOR inhibition causes preferential displacement of TOP mRNAs from eIF4E . Additionally, overexpression of eIF4E increases the translation of TOP mRNAs in NIH 3T3 cells , although it is unable to reverse their repression under conditions that cause a severe growth arrest . Nonetheless, this model is unlikely for several reasons. First, structural studies show little indication of an interaction between eIF4F and nucleotides beyond the first one or two following the cap structure . Secondly, other forms of stress (e.g. hypoxia) continue to repress TOP mRNA translation in 4E-BP-deficient cells, arguing that a downstream repressor can be controlled by mTOR-independent inputs . Thirdly, the addition of a cap competitor to in vitro translation reactions causes similar inhibition of the translation of TOP and non-TOP mRNAs .
mTORC1 control of TOP mRNA translation.
A more likely scenario is that eIF4F competes with an unidentified repressor for binding to the 5′ termini of TOP mRNAs. This model remains consistent with a function for 4E-BPs, which reduce eIF4F binding to mRNAs and thereby expose their 5′ ends to potential repressor proteins . The existence of a TOP repressor was also suggested by early studies in various in vitro translation systems that use wheat germ or rabbit reticulocyte extracts. TOP mRNAs are inefficiently translated in these systems, but can be partially rescued by the addition of pyrimidine RNA oligonucleotides [52,53].
Over the last several decades, a series of potential TOP regulators have been proposed. Among the most recent candidates are the T-cell intracellular antigen 1 (TIA1) and TIA1-related (TIAR) RNA-binding proteins. These translation repressors preferentially bind to pyrimidine-rich sequences, including TOP motifs, and are required for TOP mRNA repression during amino acid starvation . However, mTOR-dependent control of TOP mRNAs is retained in cells lacking TIA proteins, suggesting that there are either redundant TOP-regulatory mechanisms or, like 4E-BPs, they act upstream of the ultimate TOP detector [15,54]. More recently, La ribonucleoprotein domain family member 1 (Larp1) has been reported as either a TOP repressor  or a TOP activator . Larp1 is an abundant member of the La family of RNA-binding proteins that has also been identified in phosphoproteomic screens for mTOR substrates [57,58]. Larp1 interacts with RNA directly, but also through an association with PABP and possibly eIF4E . Recent biochemical studies have further shown that a C-terminal domain of Larp1 can specifically recognize some TOP sequences (e.g. RPS6), although it fails to bind others (e.g. PABPC1) . It thus remains unclear whether, in cells, the protein directly detects the TOP motif or instead affects TOP mRNA translation by manipulating eIF4F. Other putative TOP regulators have also been proposed (La, CNBP, and mir-10A), but none appear to be universally required for TOP mRNA control . Conclusive identification of the mechanism that detects and represses TOP mRNAs thus remains a major unanswered question.
Growth functions of eIF4F
How does control of eIF4F contribute to mTORC1 functions? eIF4F and mTORC1 have both been implicated in a similar spectrum of diverse phenotypes, including aging, metabolism, and synaptic function [1,18]. But the most conclusive evidence for mTORC1 functions that are mediated through eIF4F is control of growth and proliferation, particularly in oncogenic contexts. Growth-promoting functions of eIF4F itself were noted soon after its discovery, as overexpression of eIF4E is sufficient to drive quiescent cells to proliferate . More recently, mice engineered to overexpress eIF4E were found to be highly tumor-prone , while deletion of a single copy of eIF4E confers remarkable resistance to tumor formation in cancer PTEN and RAS-driven cancer models . Induced expression of a constitutively active (i.e. non-phosphorylatable) 4E-BP1 allele has the opposite effect, driving cell-cycle arrest and preventing tumor growth [64,65].
Importantly, 4E-BPs are critical mediators of cell-cycle arrest caused by mTOR inhibition. Cells treated with mTOR active-site inhibitors typically enter a quiescent but reversible cell-cycle arrest [14,15]. However, cells that are depleted of 4E-BPs continue to proliferate even in the absence of all detectable mTOR activity . Interestingly, mTOR inhibition continues to reduce cell size in 4E-BP-deficient cells, suggesting a separate S6K-regulated sizing mechanism that has yet to be identified. 4E-BP-dependent inhibition of G1/S cyclin (e.g. Cyclin D1/D3) translation may be sufficient to explain the cell-cycle arrest following mTOR inhibition. However, repression of other transcripts probably compounds the cytostatic effect. For instance, it is hard to imagine how cell growth would proceed without the production of new ribosomal proteins, nearly all of which are encoded by TOP mRNAs. These antiproliferative effects are likely to synergize with other mechanisms that mTOR uses to control ribosome biogenesis (e.g. TIF-IA and Maf1) [66,67].
mTOR control of elongation
While the ties between mTORC1 and the translation initiation machinery have received the most scrutiny, signals are also relayed to regulators of elongation. This occurs through S6K-dependent phosphorylation of the elongation factor 2 kinase (EEF2K), an atypical calcium/calmodulin-dependent kinase . Under growth-promoting conditions, S6Ks phosphorylate EEF2K at sites that inhibit its activity . Dephosphorylation of those sites occurs when mTOR (and S6Ks) is inhibited, freeing EEF2K to phosphorylate a key regulatory residue (T56) on eukaryotic elongation factor 2 (EEF2), a core translation factor that shifts the nascent polypeptide chain from the ribosomal A site to the P site . Phosphorylation of EEF2 inhibits its activity, slowing elongation. Loss of EEF2K in mice appears to have relatively mild effects under normal growth conditions, most notably a reduction in quality control of germline cells . However, more striking EEF2K functions emerge under stress. Loss of EEF2K is highly sensitizing to both ionizing radiation  in vivo and nutrient deprivation in cultured cells . Conversely, loss of EEF2K protected tumor cells from the effects of rapamycin in an APC-driven model of colorectal cancer .
Does EEF2K activation control the translation of specific mRNAs? Initiation is generally considered the rate-limiting step of the translation process . Consequently, elongation becomes rate-limiting only once ribosome collisions begin to limit the initiation frequency. Because mTORC1 inhibition also decreases the initiation rate, it is difficult to predict a priori whether elongation may become rate-limiting for some mRNAs but not others. This may, however, be more likely to occur on mRNAs enriched for suboptimal codons, which are decoded more slowly. Interestingly, it was recently shown that the elongation rate is also a major determinant of mRNA stability . EEF2K may therefore affect gene expression by modulating mRNA stability, independent of changes in overall translation rates. It remains to be determined whether EEF2K-dependent phenotypes reflect changes in the translation of specific mRNAs, or whether they are a consequence of more general changes in the translation process.
The complexity of the relationship between mTORC1 and the translation machinery should come as no surprise. Translation is, after all, uniquely situated to influence the growth process: it generates the bulk protein required for growth, but can also tune the translation of specific mRNAs so as to regulate the process. But many basic questions remain. What is the mechanism that detects and represses TOP mRNA translation? How does control of eIF4F contribute to mTOR-regulated phenotypes, including non-growth phenomena such as aging? Do the functions of EEF2K depend on changes in the translation of specific mRNAs? Answers to these questions will guide a deeper understanding of mTOR functions in the many normal and disease contexts where it is known to have important roles.
Communication between mTORC1 and the translation machinery probably extends beyond the mechanisms described here. Recent phosphoproteomic analyses have identified many RNA-binding proteins whose phosphorylation is affected by mTOR signaling [57,58]. While many of these are largely uncharacterized, it is plausible that they might be used to target the translation of specific mRNAs. At the same time, various studies have proposed new mRNA targets of mTORC1-regulated translation. A few recent examples are the senescence-promoting kinase MAPKAP2 , the inflammatory cytokinase IL-1A, and the stress-induced transcription factor ATF4 . Whether these are targeted by known mTORC1-regulated mechanisms (e.g. 4E-BPs) or through currently uncharacterized mTORC1 substrates remains to be determined.
- APC, adenomatous polyposis coli; ATF4, activating transcription factor 4; CNBP, CCHC-type zinc finger nucleic acid binding protein; 4E-BPs
eIF4E homologous protein
eukaryotic elongation factor 2
elongation factor 2 kinase
- eIF4B, eukaryotic initiation factor 4B; Larp1
La ribonucleoprotein domain family member 1
- MAPKAP2, mitogen-activated protein kinase-activated protein kinase 2; mTOR
mammalian target of rapamycin
mTOR complex 1 and 2
- PDCD4, programmed cell death protein 4; PIC
- PTEN, phosphatase and tensin homolog; Raptor
regulatory-associated protein of TOR
rapamycin-insensitive companion of TOR
T-cell intracellular antigen 1
terminal oligopyrimidine; TSC, tuberous sclerosis complex; UTRs, untranslated regions.
The Author declares that there are no competing interests associated with this manuscript.