The mechanistic Target of Rapamycin: The grand conducTOR of metabolism and aging

Summary

Since the discovery that rapamycin, a small molecule inhibitor of the protein kinase mTOR (mechanistic Target Of Rapamycin), can extend the lifespan of model organisms including mice, interest in understanding the physiological role and molecular targets of this pathway has surged. While mTOR was already well known as a regulator of growth and protein translation, it is now clear that mTOR functions as a central coordinator of organismal metabolism in response to both environmental and hormonal signals. This review discusses recent developments in our understanding of how mTOR signaling is regulated by nutrients and the role of the mTOR signaling pathway in key metabolic tissues. Finally, we discuss the molecular basis for the negative metabolic side effects associated with rapamycin treatment, which may serve as barriers to the adoption of rapamycin or similar compounds for the treatment of diseases of aging and metabolism.

Introduction to the mechanistic Target of Rapamycin signaling pathway

The mechanistic (previously referred to as mammalian) Target Of Rapamycin (mTOR) is a serine/threonine protein kinase that is inhibited by rapamycin, a compound produced by bacteria originally isolated from the soil of Easter Island that inhibits the proliferation of eukaryotic cells (Vezina et al., 1975). The mTOR protein kinase is found in two discrete complexes, mTOR complex 1 (mTORC1) and mTORC2, each of which contains distinct protein components and phosphorylates different substrates. mTORC1 is acutely inhibited by rapamycin, which has been used therapeutically and as a probe to gain insight into mTORC1 regulation and function. mTORC2 is significantly less sensitive to rapamycin and is only inhibited by chronic treatment in certain cell types and tissues. As a result of its relative insensitivity to rapamycin, the physiological function and molecular targets of mTORC2 have been harder to decipher.

The two complexes formed by mTOR contain both unique and shared components. mTORC1 is defined by interaction of the mTOR protein kinase with the scaffold protein Raptor, as well the Akt substrate PRAS40. Both mTORC1 and mTORC2 include mLST8/GβL, which is required for complex assembly and stability, and the regulatory protein DEPTOR. Finally, mTORC2 is defined by the interaction of mTOR with the scaffold protein Rictor, and other mTORC2-specific protein subunits include mSin1 and Protor-1/2. Cryo-electron microscopy structural studies have determined that mTORC1 is an obligate dimer (Yip et al., 2010). A recent higher-resolution structure shows that mLST8 and the amino-terminal domain of Raptor limit access to the active site (Aylett et al., 2016). While the structure of mTORC2 remains unknown, the structure of the homologous yeast TORC2 structure was recently solved; as with mTORC1, the complex has two-fold symmetry (Gaubitz et al., 2015). In TORC2, the Rictor homolog AVO3 masks the rapamycin-interacting domain of TOR; deletion of this masking domain of AVO3 sensitizes TORC2 to rapamycin (Gaubitz et al., 2015).

The mTOR complexes regulate numerous processes required for cell growth and metabolism, functioning as a signaling node that integrates cellular nutrient and stress status and induces appropriate cellular responses. mTORC1 controls ribosomal biogenesis, protein translation and autophagy, mediated by substrates that include S6K1, 4E-BP1, ULK1 and others (Figure 1). mTORC2 has primarily been characterized as a downstream effector of the insulin/IGF-1 signaling pathway, and several of its substrates have been characterized (Figure 2). This includes three distinct sites on AKT that are important for its activation – AKT T450, the turn-motif, AKT S473, the hydrophobic motif, and AKT S477/479 at the C-terminal end (Ikenoue et al., 2008; Liu et al., 2014a; Sarbassov et al., 2005). mTORC2 also phosphorylates SGK1 (serum- and glucocorticoid-induced protein kinase 1) (Garcia-Martinez and Alessi, 2008), as well as several members of the protein kinase C (PKC) family, including PKCα (Sarbassov et al., 2004), PKCδ (Gan et al., 2012), PKCε (Ikenoue et al., 2008), and PKCζ (Li and Gao, 2014). mTORC2 was recently found to also phosphorylate MST1, a kinase in the Hippo signaling pathway (Sciarretta et al., 2015).

Figure 1. mTORC1, a central regulator of metabolism.

Regulation of metabolic processes downstream of mTORC1, with major known substrates and metabolic processes highlighted. Protein kinases, including mTORC1, are boxed in red. The core proteins of mTORC1 – mTOR, Raptor and mLST8 – are depicted as interacting directly with the mTOR kinase, whereas proteins with nutrient sensitive or transient interactions with the mTORC1 core – DEPTOR, PRAS40, and Tel2/Tti1 – are depicted in a separate, adjacent box. Proteins which regulate the localization of mTORC1 to the lysosome – e.g., the Rag GTPases and the Ragulator and GATOR complexes – are not depicted.

Figure 2. mTORC2, a major effector of the Insulin/IGF-1 signaling pathway.

Regulation of metabolic processes downstream of mTORC2, with major known substrates and metabolic processes highlighted. Protein kinases, including mTORC2, are boxed in red. The core proteins of mTORC2 – mTOR, Rictor, mLST8, mSin1, and Protor – are depicted as interacting directly with the mTOR kinase, whereas proteins with nutrient sensitive or transient interactions with the mTORC2 core – DEPTOR, IKK, Sestrin3, Tel2/Tti1, and Xpln – are depicted in a separate, adjacent box. The TSC complex – in particular, TSC2 - interacts with and regulates mTORC2 (Huang et al., 2008), but it is not clear how or if this interaction is regulated. Dashed Box: Insulin stimulates mTORC1 through the AKT-mediated phosphorylation of specific sites on TSC2 and PRAS40 (Menon et al., 2014; Sancak et al., 2007), but the requirement for mTORC2 in the insulin-mediated activation of mTORC1 remains unclear. Extracellular signals in addition to insulin, IGF-1, and leptin that regulate PI3K or PIP₃ may also regulate mTORC2 activity, but these are not depicted here.

Regulation of mTORC1 and mTORC2 by nutrient and endocrine signaling

The mTOR pathway integrates numerous environmental signals that indicate if conditions are favorable for anabolic processes. Cells in a multicellular organism must integrate information about both the local availability of nutrients and resources that can serve as building blocks, as well as the needs of the entire organism as signaled by circulating factors such as hormones. The mTOR signaling pathway responds to stimuli including the growth factors insulin and IGF-1, the levels of amino acids and glucose, the cellular energy status, and the level of oxygen.

Regulation of mTORC2

As highlighted in Figure 2, mTORC2 activity is stimulated by insulin, IGF-1 and leptin via the activation of PI3K (Park and Ahima, 2014). The molecular mechanism by which PI3K stimulates mTORC2 is still unclear, although the GTPase Rac1, a downstream effector of PI3K, has been implicated (Saci et al., 2011). It has also recently been proposed that Akt itself may be “upstream” of mTORC2, through the activating phosphorylation of mSIN1 T86 by Akt (Humphrey et al., 2013). However, the precise role of this phosphorylation is unclear, and it may actually negatively regulate mTORC2 function (Liu et al., 2013). It was recently reported that phosphatidylinositol (3,4,5)-trisphosphate (PIP₃) can directly activate mTORC2 through relieving an inhibitory interaction with mSIN1 (Liu et al., 2015). Other signaling pathways that regulate PI3K activity or levels of PIP₃ may also therefore regulate mTORC2.

A growing body of recent literature suggests that mTORC2 (and perhaps also mTORC1) may be regulated by lipids. mTORC2 activity in hepatocytes is inhibited by overexpression of glycerol-3-phosphate acyltransferase-1 (Gpat1), and mTORC2 can be disrupted in vitro by treatment with preferred Gpat1 substrates (Zhang et al., 2012). Similar recent studies have found that lipolytic products produced either in vivo or in vitro can disrupt mTORC2, although the precise lipid or lipids that mediate this effect have not been identified (Mullins et al., 2014). Conversely, mTORC2 activity can be stimulated by expression of fatty acid elongase-5 (Elovl5), or by treatment of cells with the Elovl5 substrate palmitoleic acid (Tripathy and Jump, 2013).

mTORC2 activity may depend in part upon PI3K-stimulated interactions with ribosomal protein subunits (Zinzalla et al., 2011); mTORC2 was recently shown to localize to the ribosome-rich mitochondria-associated endoplasmic reticulum (Betz et al., 2013). The abundance of ribosomal subunits and activity of the ribosome is largely controlled by mTORC1, thus indirectly linking mTORC2 to the stimulation of mTORC1 by amino acids, glucose, cellular energy, and oxygen. Other positive and negative regulators of mTORC2 activity have been described over the past several years, and some of these are highlighted in Figure 2. These positive regulators include IKK, which interacts with mTORC2 and promotes the association between mTOR and Rictor (Xu et al., 2013). Sestrin3, a member of an evolutionarily conserved family of stress response proteins, was also recently reported to interact with mTORC2 and stimulate phosphorylation of AKT S473 (Tao et al., 2014). Two additional proteins that interact with mTORC2 as well as mTORC1 are Tel2 and Tti1, and are important in the assembly and stability of both mTORC1 and mTORC2 as well as related kinases (Kaizuka et al., 2010). Negative regulators of mTORC2 include DEPTOR, which inhibits both mTOR complexes (Peterson et al., 2009), and XPLN (Arhgef3), which specifically interacts with and inhibits mTORC2 (Khanna et al., 2013). More recently, the first plasma membrane transmembrane inhibitor of mTORC2 function, UT2, was described in hematopoietic progenitor cells, potentially linking extracellular signals to control of this complex (Lee et al., 2014).

mTORC2 regulates mTORC1 through phosphorylation of AKT at three sites, which controls not only the activity of AKT, but also its abundance (Hagiwara et al., 2012). AKT promotes mTORC1 activity via two mechanisms: first, AKT phosphorylates PRAS40 (Sancak et al., 2007), which when unphosphorylated binds to and inhibits mTORC1; and second, via inhibitory phosphorylation of TSC2 (Figure 1) (Inoki et al., 2002; Menon et al., 2014). TSC1 and TSC2, in complex with a third protein, TBC1D7, form the tuberous sclerosis complex (TSC), which functions as a mini-regulatory hub upstream of mTORC1 signaling, and TSC functions as a GTPase-activating protein (GAP) toward Rheb, a small GTPase that interacts with and activates mTORC1 (Dibble et al., 2012). In addition to AKT, many different kinases, including AMPK, ERK, GSK3 and IKKβ converge to phosphorylate distinct residues of TSC1/2 and thereby inhibit or activate its GAP activity towards Rheb. Intriguingly, deletion of TSC2 impairs mTORC2 activity independently of its GAP activity, and TSC2 physically interacts with mTORC2 (Huang et al., 2008), suggesting that TSC may play a role in regulating both mTOR complexes.

mTORC1: Location, location, location

One of the primary stimuli to which mTORC1 is responsive is the availability of amino acids, particularly the branched-chain amino acids (BCAAs) - leucine, isoleucine, and valine. However, the mechanism by which mTORC1 senses amino acid levels has remained mysterious until recently. In short, it is now clear that the activation of mTORC1 requires the harmonious interaction of many proteins at an unexpected location: the lysosome. While the lysosome is often conceived of as a simple sack that contains a wide range of digestive enzymes and functions as the recycling center of the cell, it is now clear that this organelle serves as a platform for coordinating amino acid recycling with mTORC1 activation.

We will not discuss the many recent discoveries regarding the regulation of mTORC1 by amino acids in detail, as these have been the subject of a recent comprehensive review by Goberdhan and colleagues (Goberdhan et al., 2016). However, in brief, mTORC1 is recruited to the lysosome by the Rag family of small GTPases; the nucleotide loading state of the Rags regulate their function and ability to recruit mTORC1 to the lysosome (Kim et al., 2008; Sancak et al., 2008). Regulators of the nucleotide loading state of the Rag GTPases include the Ragulator, which functions as a guanine nucleotide exchange factor (GEF) for RagA and RagB, and GATOR1 which functions as a GAP for RagA and RagB (Bar-Peled et al., 2013; Bar-Peled et al., 2012).

Until very recently, the ultimate molecular mechanism by which amino acids regulate mTORC1 activity has remained elusive, but over the last year multiple distinct molecular mechanisms have been identified. SLC38A9, a lysosomal protein that functions as a low-affinity amino acid transporter for arginine (and also glutamine, asparagine, histidine, and lysine), is required for the normal activation of mTORC1 by amino acids (Jung et al., 2015; Rebsamen et al., 2015; Wang et al., 2015). SLC38A9 interacts with both the Ragulator and the vacuolar H(+)-adenosine triphosphatase ATPase (v-ATPase), and it is possible that SLC38A9 may regulate the GEF activity of Ragulator or the interaction of the Rags with mTORC1. GATOR2, which functions as a negative regulator of GATOR1, was recently found to be negatively regulated by both the Sestrin family of stress response proteins as well as by CASTOR1/2 (Chantranupong et al., 2016; Chantranupong et al., 2014; Parmigiani et al., 2014). Sestrin2 directly binds to leucine, which impairs the inhibitory action of Sestrin2 upon GATOR2, allowing mTORC1 to be activated (Saxton et al., 2016; Wolfson et al., 2016). CASTOR1 and its close homolog CASTOR2 similarly regulate GATOR2, but are arginine-dependent (Chantranupong et al., 2016). Other proteins recently identified as critical to the regulation of mTORC1 by amino acids are the leucine transporter LAT1-4F2hc (SLC7A5-SLAC3A2), the adenosine diphosphate ribosylation factor-1 (Arf1) GTPase, and the RagA K63 E3 ubiquitin ligases Skp2 and RNF152 (Deng et al., 2015; Jewell et al., 2015; Jin et al., 2015; Milkereit et al., 2015).

The activation of mTORC1 at the lysosomal surface requires not only the presence of amino acids, but also Rheb-GTP – which is normally inhibited by TSC. While many posttranslational modifications of TSC1/2 have been shown to be inhibitory, it has only become clear in the last year that the activity of TSC is – like mTORC1 itself – regulated by localization. Insulin stimulates the acute dissociation of TSC and Rheb; it was subsequently realized that TSC is localized to the lysosome in the absence of insulin, and departs in response to insulin, permitting Rheb to activate mTORC1 (Menon et al., 2014). The departure of TSC from the lysosome in response to insulin requires the Akt-dependent phosphorylation of TSC2 (Inoki et al., 2002; Menon et al., 2014). This suggests a model in which mTORC2 functions upstream of mTORC1 in the insulin/IGF-1 signaling pathway; however, as mTORC1 activity is largely normal in mice and tissue culture cells with reduced expression of Rictor (e.g., (Lamming et al., 2014b; Sarbassov et al., 2005)), further research will be required to clarify the role of mTORC2 in the regulation of mTORC1 (Sparks and Guertin, 2010).

Many intracellular and extracellular signals regulate mTORC1 via inhibitory or activating phosphorylation of specific residues on TSC1 or TSC2 (reviewed in (Laplante and Sabatini, 2009)). Recent work has found that other cellular stresses, including a lack of amino acids, similarly promote lysosomal accumulation of TSC through a mechanism that may be partially dependent on the Rag GTPases (Demetriades et al., 2014; Demetriades et al., 2016). Interestingly, the lysosomal recruitment of the Rheb GTPase itself is also regulated by amino acids, which stimulate the binding of Rheb to microspherule protein 1 (MCRS1) (Fawal et al., 2015). The localization of Rheb-MCRS1 to the lysosome depends upon both amino acids and the presence of an active Ragulator and v-ATPase. In addition to its localization function, MCRS regulates the association of Rheb with TSC1/2 in an amino acid dependent manner (Fawal et al., 2015). The intricate ballet by which mTORC1 and its activators are localized to the lysosome, while its inhibitors depart (Figure 3), highlights the necessity for precise coordination of upstream signals with downstream cellular responses.

Figure 3. Regulation of mTORC1 activity by amino acids and insulin.

A) In the absence of amino acids and insulin, mTORC1 is found in the cytoplasm while TSC localizes to the lysosome. B) Amino acids result in the inhibition of GATOR1 by GATOR2, the activation of the Rag GTPases, the localization of mTORC1 to the lysosome, and the recruitment of Rheb to the lysosome by MCRS1. TSC continues to inhibit Rheb and thus mTORC1 remains inactive. C) Insulin induces TSC to leave the lysosome, permitting Rheb to bind to GTP; insulin also stimulates the disassociation of PRAS40 from mTORC1. mTORC1 can then interact with GTP-bound Rheb and becomes active.

mTORC1 is also sensitive to the availability of glucose, which like amino acids induces localization of mTORC1 to the lysosome (Efeyan et al., 2013). While it has been assumed that glucose mediates mTORC1 via the regulation of AMPK, mTORC1 signaling is still responsive to glucose in cells lacking AMPK, and it was recently shown that constitutive activation of the Rag proteins makes cells insensitive both to amino acid and glucose withdrawal; moreover, glucose regulates the binding of the v-ATPase to Ragulator (Efeyan et al., 2013). A more recent study found that hexokinase-II binds to and inhibits mTORC1 in the absence of glucose, linking glucose levels to autophagy (Roberts et al., 2014). A role for hexokinase-II in mediating the localization of mTORC1 or the v-ATPase/Ragulator interaction has not been identified, suggesting that glucose may regulate mTORC1 through three separate mechanisms: AMPK, v-ATPase/Ragulator/Rag interaction and localization, and direct inhibition of mTORC1 by hexokinase-II. A complex containing Tel2 and Tti1 (TTT-RUVBL1/2) has also been implicated in controlling the lysosomal localization and dimerization of mTORC1 in response to glucose and other energetic stresses (Kim et al., 2013).

Newly discovered mTORC1 and mTORC2 substrates

While a comprehensive overview of mTOR substrates is beyond the scope of this review, the past several years have led to the identification of several important mTOR substrates and links to new metabolic processes. First, mTORC1 (Figure 1) is well characterized as a regulator of mRNAs that contain a terminal oligopyrimidine (TOP) motif (Thoreen et al., 2012). While the 4E-BPs are responsible for many of the effects of mTOR on protein translation, the La related protein 1 (LARP1) is a recently discovered mTORC1 effector which also represses the translation of TOP mRNAs (Fonseca et al., 2015). Another novel mTORC1 substrate discovered recently is the phosphatidylinositol-5-phosphate 4-kinase γ (PIP4Kγ) (Mackey et al., 2014). Finally, a major new transcriptional effector of TORC1 signaling was identified this year in Drosophila melanogaster, REPTOR (Tiebe et al., 2015). Much like FOXO, REPTOR is phosphorylated by TORC1 and thus sequestered in the cytoplasm by a 14-3-3 protein. Upon TORC1 inhibition, REPTOR is liberated by a phosphatase and translocates to the nucleus, where it binds target genes together with another protein, REPTOR-BP (Tiebe et al., 2015). Mammalian equivalents of REPTOR and REPTOR-BP have not yet been characterized, but Crebrf and Crebl2 are homologous human proteins (Tiebe et al., 2015).

mTORC2 (Figure 2) is a key effector of the insulin signaling pathway, regulating several substrates downstream of the insulin/IGF-1 receptor. mTORC2 phosphorylates AKT on residues T450, S473, and a recently discovered site on the C-terminal tail, S477/T479 (Liu et al., 2014a). mTORC2 is also a well-characterized activator of SGK and PKC-α, but recently mTORC2 has been demonstrated to regulate control of other PKC family members, including PKCδ and PKCζ (Gan et al., 2012; Li and Gao, 2014). Finally, mTORC2 regulates the stability of insulin receptor substrate 1 (IRS1) via phosphorylation of the ubiquitin ligase subunit Fbw8 (Kim et al., 2012). mTORC2 is more than just a simple effector of insulin/IGF-1 signaling. mTORC2 phosphorylates the Ste20 family protein kinase oxidative stress-responsive 1 (OSR1) to promote its activity and response to osmotic stress (Sengupta et al., 2013). Very recently, mTORC2 was discovered to negatively regulate the activity of MST1, a key protein component of the Hippo signaling pathway (Sciarretta et al., 2015). Despite the extensive work already completed on mTOR, the known substrates of both mTORC1 and mTORC2 continue to expand. Knowledge of the full breadth of mTOR functions and, in particular, how these functions are coordinated awaits extensive further analysis.

mTOR, a metabolic master regulator

Due to its unique capacity to sense and integrate many different environmental and hormonal cues, the mTOR signaling pathway occupies a central role in the regulation of metabolism. A few of the metabolically active tissues in which mTOR plays an important role, and to which we refer the reader to recent reviews, include the heart (Sciarretta et al., 2014), kidney (Grahammer et al., 2014), and pancreas (Kulkarni et al., 2012). In this section, we will review the most recent findings on the role of mTOR signaling in physiology and metabolism in some of the key metabolic tissues: liver, muscle, adipose, and the brain.

mTOR in the liver

The role of mTOR in the liver has been investigated over the past several years, primarily using three different genetic mouse models: mice lacking hepatic TSC1 (L-TSC1^KO), which have hyperactive mTORC1 signaling; mice lacking hepatic Raptor, with decreased mTORC1 signaling (L-Rap^KO); and mice lacking hepatic Rictor (L-Ric^KO), with decreased mTORC2 signaling.

Several groups have examined the role of mTORC2 in the liver utilizing liver-specific Rictor knockout mice (L-RicKO). Loss of hepatic Rictor results in hepatic insulin resistance, and thus L-RicKO mice, which exhibit loss of AKT S473 phosphorylation and decreased AKT activity, have elevated hepatic glucose output and are glucose and pyruvate intolerant despite also being hyperinsulinemic (Lamming et al., 2012). Additional studies have found that L-RicKO mice have reduced hepatic glucokinase and SREBP1c activity, as well as defective glycolysis and lipogenesis (Hagiwara et al., 2012; Yuan et al., 2012). On a high-fat diet, L-RicKO mice are protected from hepatic steatosis and have reduced cholesterol levels (Yuan et al., 2012). It is unclear if all of these effects are mediated by Akt, as Yuan et al. found that expression of constitutively active Akt did not rescue the lipid defect of mice lacking hepatic Rictor (Yuan et al., 2012), while Hagiwara et al. came to the opposite conclusion (Hagiwara et al., 2012).

A recent transcriptional profiling and phosphoproteomic study of L-RicKO mice found the loss of hepatic Rictor caused profound alterations in gluconeogenesis, with increased expression of gluconeogenic genes including G6p and reduced expression of Gck (Lamming et al., 2014a). It is likely that increased utilization of glucose and pyruvate for gluconeogenesis partially explains the decrease in hepatic lipid synthesis. Analysis of L-RicKO mice suggests that many transcription factors act downstream of hepatic mTORC2, including not only the sterol regulatory element-binding proteins (SREBPs) and FOXO1 as first identified by Hagiwara et al. and Yuan et al., but also other Forkhead family transcription factors such as FOXA2 as well as other transcription factors including PPARγ (Lamming et al., 2014a). While a possible explanation for altered SREBP and PPARγ signaling in the absence of Rictor is altered mTORC1 signaling, mice lacking hepatic Rictor have largely normal mTORC1 activity (Lamming et al., 2014b).

Phosphoproteomic analysis of L-RicKO livers identified two potential pathways impacted by mTORC2. First, the phosphorylation of glycogen synthase (Gys2) (Lamming et al., 2014a), a key enzyme in the conversion of glucose to glycogen was altered at two sites, with increased phosphorylation at Gys2 S6/S8/S11. Data from Gys2 phosphosite mutants (Ros et al., 2009) and the L-RicKO phenotype are consistent with these alterations reducing Gys2 activity. Secondly, p38 MAPK signaling pathway activity, which is now known to be important in maintaining glucose homeostasis (Lee et al., 2011), was blunted in the livers of L-RicKO mice. On the basis of the mTOR consensus motif (Kang et al., 2013), it is unlikely these are direct substrates of mTORC2; identifying the mechanism by which hepatic mTORC2 regulates Gys2 and p38 MAPK signaling may provide greater insight into the dramatic phenotypes of L-RicKO mice.

While the exact mechanism by which PI3K signaling activates mTORC2 is unclear, a recent study identified a sestrin, Sesn3, as an mTORC2 activating factor (Tao et al., 2014). Mice in which hepatic Sesn3 was deleted demonstrated hyperglycemia and decreased glucose tolerance, while transgenic expression of Sesn3 increased glucose tolerance. Hepatic Sesn3 deletion significantly impaired AKT S473 phosphorylation, and subsequent experiments demonstrated that Sesn3 (as well as Sesn2) physically interacts with Rictor. The mechanism by which Sesn3 activates mTORC2 is unclear, but it may act in part by physically stabilizing mTORC2 (Tao et al., 2014).

The role of mTORC1 in the liver was explored very comprehensively in 2009 by Sengupta and colleagues, who utilized mice specifically lacking either hepatic TSC1 (L-TSC1^KO) or Raptor (L-Rap^KO) mice to demonstrate that mTORC1 promotes liver weight and hepatocyte size, while repressing fasting-induced ketogenesis and not measurably impacting glucose or insulin levels (Sengupta et al., 2010). Left unexplained was the finding that L-TSC1^KO mice had decreased locomotor activity and body temperature following a prolonged fast, an observation recently explored by Cornu and colleagues, who determined that these mice also have decreased hepatic triglycerides and glutamine levels (Cornu et al., 2014). Interestingly, L-TSC1^KO mice exhibit significantly greater induction of FGF21, which encodes a liver-produced circulating hormone best known for its role as an insulin-sensitizing hormone produced in response to fasting. Cornu et al. showed that the locomotor activity, body temperature and lipid metabolism phenotypes of L-TSC1^KO mice are mediated by FGF21, which is transcriptionally induced following activation of PPARα by glutamine depletion (Cornu et al., 2014). The high mTORC1 activity in the livers of L-TSC1^KO mice promotes glutaminolysis, lowering levels of glutamine in both the liver and blood, and treatment with glutamine blocks the effect of hepatic mTORC1 hyperactivity on body temperature (Cornu et al., 2014).

mTOR in the muscle

mTOR signaling in skeletal muscle has been highly studied and yet is incompletely understood. The refractory nature of mTOR pathway to discovery in this tissue is likely due to its complex roles. Whereas mTORC1 signaling is required for increases in muscle mass in response to exercise or during tissue repair after damage – it was recently shown that mechanical stimuli activate mTORC1 in part through the phosphorylation of Raptor (Frey et al., 2014) – maintaining the basal state of the pathway in a low level may also be important for maintenance of muscle function with aging. Thus, the dynamic range of the pathway may ultimately be the best measure of efficient skeletal muscle function.

The role of mTORC1 and mTORC2 in skeletal muscle has been addressed as in other tissues in large part by characterizing knockouts at various parts of the signaling pathway. Raptor null mice present with muscle atrophy, reduced oxidative capacity and increased glycogen stores that progressive results in a dystrophic phenotype (Bentzinger et al., 2008). A more recent study has linked these phenotypes to altered skeletal muscle excitation-contraction coupling, although the mechanisms underlying this pathology remain unknown (Lopez et al., 2015). Interestingly, neither whole body knockout of S6K1 or activation of 4E-BP1 in skeletal muscle, both of which phenocopy reduced mTORC1 signaling, display this severe phenotype (Selman et al., 2009; Tsai et al., 2015), suggesting that the Raptor knockout phenotype may derive from altered phosphorylation of both S6K1 and 4E-BP1, or possibly other mTORC1 targets. Paradoxically, enhanced mTORC1 activation through muscle-specific knockout of TSC1 also leads to skeletal muscle decline, although this time the deficits were attributed to impaired autophagy (Castets and Ruegg, 2013). These mice also showed a significant decrease in adipose mass, suggestive of a cell non-autonomous effect. Intriguingly, as with mice lacking hepatic TSC1, loss of TSC1 specifically in the muscle also induces production of FGF21. However, instead of being regulated by PPARα, in skeletal muscle increased expression of FGF21 is mediated by increased levels of ATF4 due to significant endoplasmic reticulum (ER) stress (Guridi et al., 2015). Notably, while both decreased and increased mTORC1 signaling in skeletal muscle induce similar changes in body composition and myopathy, the metabolic effects on the whole organism are distinct, with activation of mTORC1 in skeletal muscle reducing liver stores of lipids and glycogen and stimulating the browning of white adipose tissue (Guridi et al., 2016).

The role of mTORC2 in skeletal muscle was at first not clear, as one initial report suggested that muscle-specific Rictor knockout did not result in a detectable phenotype (Bentzinger et al., 2008), suggesting that much of the regulation in skeletal muscle is coordinated through the mTORC1 complex. However, Kumar and colleagues reported that muscle-specific Rictor knockout mice had decreased insulin-stimulated glucose uptake and were glucose intolerant (Kumar et al., 2008), a result also found by a more recent study (Kleinert et al., 2014). In support of this finding, Kleinert and colleagues also demonstrated that loss of Rictor in skeletal muscle resulted in resistance to acute mTOR kinase inhibitor-induced glucose intolerance (Kleinert et al., 2014). A mouse knockout of mTOR has also been generated, with phenotypes very similar to the skeletal muscle Raptor knockout (Bentzinger et al., 2008; Risson et al., 2009), and displaying increased basal glucose uptake into skeletal muscle. In addition, loss of mTOR in skeletal muscle results in reduced muscle dystrophin content, a phenotype not observed by specific knockout of either Raptor or Rictor (Risson et al., 2009).

In addition to differentiated myofibers, muscle tissue contains adult stem cells, termed satellite cells, that exist in a quiescent state and become activated in response to muscle damage to repair and replace damaged tissue (Brack and Rando, 2012; Dumont et al., 2015). These cells are a small fraction of the muscle cells in a tissue, but can be studied in vivo after muscle damage or ex vivo in culture proliferation and differentiation models. The mTOR pathway has a stimulatory role at several steps in the differentiation process of satellite cells and a full description is beyond the scope of this review (Rodgers et al., 2014; Schiaffino et al., 2013; Zhang et al., 2015). Continued studies of the role of mTOR in muscle are important given that the signaling pathway is critical in both the metabolic functions of differentiated fibers, as well as regeneration of new ones. One question for instance is whether long-term rapamycin administration, administered at doses that extend mouse lifespan, affects satellite cell activation and muscle regeneration in response to invoked damage.

mTOR in adipose tissue

mTORC1 has been known for over a decade to play a role in adipogenesis, with early experiments demonstrating that mTORC1 is required for adipogenesis as well as maintenance of fat tissues, and that genetic activation of mTORC1 by deletion of TSC2 promotes adipogenesis. The effects of mTORC1 on adipogenesis are mediated by 4E-BP1, Lipin1, and S6K1, which in turn regulate the activity of PPARγ and SREBPs (reviewed in (Lamming and Sabatini, 2013)). An initial study characterized mice lacking adipose Raptor are lean, resistant to the negative effects of diet-induced obesity on glycemic control, and UCP1 expression is upregulated 25-fold in their white adipose tissue (Polak et al., 2008). However, this study relied upon an Ap2-Cre Cre-recombinase, which is now known to have mosaic expression in adipose tissue and expression in many other cell types (Jeffery et al., 2014; Lee et al., 2013). A more specific Adiponectin-Cre directed deletion of Raptor in adipose tissue found that loss of Raptor led to lipodystrophy and insulin resistance, with a significant increase in liver mass due to hepatic steatosis (Lee et al.). However, both models of adipose Raptor loss promote leanness and resistance to diet-induced obesity. These phenotypes are not phenocopied by activation of 4E-BP1 in adipose, which surprisingly results in an obese phenotype and suggests that other substrates of mTORC1 mediate the Raptor knockout phenotype (Tsai et al., 2015).

The browning or beiging of white adipose tissue is an important physiological adaptation to cold and perhaps other stresses (Wu et al., 2013a), and it has recently become clear that mTORC1 plays an important role in beiging as well as the converse process, in which brown adipose tissue can whiten. It was discovered only last year that activation of mTORC1 via deletion of TSC1 leads to the whitening of brown adipose tissue, upregulating white adipose tissue markers while promoting lipid accumulation (Xiang et al., 2015). This phenotype could be reversed both in vivo and in vitro in cultured brown adipocytes by treatment with rapamycin, which rescued the normal brown phenotype. Intriguingly, rapamycin treatment also promotes a brown phenotype in white adipocytes; treatment with rapamycin blocks white adipocyte beiging in response to acute β3-adrenergic receptor agonist stimulation, and suppresses cold-induced beiging of white adipose tissue, decreasing cold tolerance (Tran et al., 2016). Subsequent work determined that β3-adrenergic signaling stimulates PKA-mediated phosphorylation of both mTOR and Raptor, leading to activation of mTORC1 and expression of uncoupling protein-1 (Ucp1) (Liu et al., 2016).

In contrast, mTORC2 was initially believed to play at most a minor role in adipogenesis, since mice lacking adipose Rictor showed metabolic and physiological phenotypes but normal adipogenesis (Cybulski et al., 2009; Kumar et al., 2010). However, this is now believed to be an effect of timing, as Rictor null mouse embryonic fibroblasts (MEFs) cannot be differentiated into adipocytes (Yao et al., 2013). Interestingly, in adipocytes it has been suggested that Akt may be “upstream” effector of mTORC2, through the activating phosphorylation of mSIN1 T86 by Akt (Humphrey et al., 2013). In such a model, PI3K would activate Akt through phosphorylation of T308 by PDK1; Akt would then phosphorylate mSin1, activating mTORC2, which would then fully activate AKT through the phosphorylation of S473. It remains to be determined if this model is correct, and if it is specific to adipocytes or more generalizable.

Deletion of mTORC2 in adipose tissue would naturally be expected to impair glycemic control due to peripheral insulin resistance. However, Cybulski et al. reported that although adipose Rictor knockout mice are insulin resistant, they are also glucose tolerant due to high levels of insulin and had a significant enlargement of both the pancreas and the pancreatic islets, suggesting a non-cell autonomous phenotype (Cybulski et al., 2009). Examining slightly older mice, Kumar et al. observed a different phenotype, with glucose intolerance and hepatic steatosis despite hyperinsulinemia (Kumar et al., 2010). The reason for the different results of these two studies is not clear, but the age of the mice examined, diet, and genetic background are possible explanations. Intriguingly, these adipose Rictor knockout mice are also hypothermic under normal housing conditions, with an inability to maintain their body temperature when exposed to cold stress despite an increase in shivering thermogenesis (Albert et al., 2016). Albert and colleagues determined that cold normally activates mTORC2 in brown adipose tissue; mice lacking adipose mTORC2 have impaired glucose uptake into brown adipose tissue following cold exposure, and are unable to increase glycolysis (Albert et al., 2016).

A complication of these studies is that they relied upon the non-specific Ap2-Cre Cre-recombinase (Jeffery et al., 2014; Lee et al., 2013). A new study in which the more adipose-specific Adiponectin-Cre was used to delete Rictor reports reduced de novo lipogenesis in adipose tissue due to reduced expression of the lipogenic transcription factor ChREBPβ (Tang et al., 2016), as well as reduced glucose uptake into adipocytes. This model of adipose Rictor deletion resulted in mice that fail to appropriately suppress hepatic gluconeogenesis in response to insulin; while the mechanism behind this is unclear and will require future study, altered hepatic lipid composition may contribute to these phenotypes.

Specific roles for mTORC1 and mTORC2 in brown adipose tissue were recently elucidated by Hung et. al, who deleted Raptor or Rictor specifically in the Myf5 lineage (Hung et al., 2014). As with white adipose tissue, Raptor was found to be extremely important for the establishment of brown adipose tissue precursors; deletion of Raptor in the Myf5 lineage leads to embryonic lethality. Deletion of Rictor in vitro blocks the differentiation on brown preadipocytes, an effect that is mediated by AKT1 and BMP7 (Hung et al., 2014). Deletion of Myf5 lineage Rictor in vivo does not compromise viability, but does lead to a very significant decrease in brown adipose tissue mass. While Rictor null brown adipose ceases to grow after 6 weeks of age, these mice were protected from high fat diet-induced weight gain, hepatic steatosis, and metabolic dysfunction when housed in a thermoneutral environment (Hung et al., 2014). This effect is likely mediated in part by the significantly increased mitochondrial activity and decreased lipogenesis observed in brown adipose tissue lacking Rictor. However, some white adipose depots as well as skeletal muscle cells are generated from the Myf5 lineage, and Rictor deletion in these other locations may partially contribute to the organismal phenotypes.

mTOR in the brain

Within the brain, it has become apparent over the last decade that the hypothalamus is not only a key regulator of metabolism, but also a key tissue from which mTOR regulates organismal physiology (Garelick and Kennedy, 2011). mTORC1 is activated by feeding in both AgRP/NPY and POMC neurons by hormones including insulin, leptin, and ghrelin (reviewed in (Haissaguerre et al., 2014; Lamming, 2014)). The hypothalamic expression of the mTOR interacting protein DEPTOR, which is associated with obesity in both mice and humans and promotes adipogenesis and the expansion of white adipose tissue (Laplante et al., 2012), is regulated by nutritional status and obesity, suggesting that hypothalamic mTORC1 is involved in the long-term regulation of energy balance (Caron et al., 2015).

A recent study by Kocalis et al. examined the role of mTORC2 in the brain by deleting Rictor either in all neurons, or by specifically deleting Rictor in either AgRP neurons or POMC neurons (Kocalis et al., 2014). Loss of Rictor in the whole brain resulted in increased food intake and obesity, fasting hyperglycemia and glucose intolerance. A similar phenotype was also observed following Rictor deletion specifically in POMC neurons, while deletion in AgRP neurons had minimal effects. Although the mechanism by which POMC mTORC2 regulates glucose tolerance is not clear, mice with neuronal deletion of IRS2 or the insulin receptor display some overlapping phenotypes with mice lacking Rictor in POMC neurons, suggesting a critical role for mTORC2 in regulating whole body metabolism downstream of neuronal PI3K signaling.

mTOR signaling in the aging process

A role for mTOR signaling in aging was first discovered over 10 years ago, with the demonstration that genetic inhibition of mTOR signaling could extend lifespan in worms and flies (Kapahi et al., 2004; Vellai et al., 2003). Subsequent work demonstrated that rapamycin treatment can extend lifespan in yeast, worms, flies, and even mice (Bjedov et al., 2010; Harrison et al., 2009; Kaeberlein et al., 2005; Powers et al., 2006; Robida-Stubbs et al., 2012). Several mouse models of decreased mTOR signaling – mice deleted for S6K1, expressing a hypomorphic allele of mTOR, or mice heterozygous for both mTOR and mLST8 – likewise possess extended longevity (Lamming et al., 2012; Selman et al., 2009; Wu et al., 2013b).

Conversely, mTORC1 signaling is increased in many age-related diseases and pathologies, including cancer (Bar-Peled et al., 2013; Grabiner et al., 2014), and may be linked to Hutchinson-Gilford progeria syndrome, which is associated with a dominant splice site mutation in the LMNA gene and resembles premature aging (Schreiber and Kennedy, 2013). Cao et al. demonstrated that rapamycin treatment of HGPS patient fibroblasts leads to restoration of nuclear structure, reduced toxic progerin aggregates and enhanced autophagy (Cao et al., 2011). In progeroid Lmna^−/− mice, which develop a range of pathologies including muscular dystrophy, lipodystrophy, dilated cardiomyopathy and peripheral neuropathy (Schreiber and Kennedy, 2013), mTORC1 signaling is elevated in specific tissues and rapamycin can partially rescue disease pathology, at least in the heart and skeletal muscle (Choi et al., 2012; Ramos et al., 2012). While the contribution of increased mTORC1 signaling to the phenotypes of HGPS in humans remains to be determined, these exciting results have spurred the start of a clinical trial of the rapamycin analog everolimus for HGPS.

An interesting question is whether mTORC1 signaling increases during healthy aging. While a number of studies have suggested that mTORC1 substrate phosphorylation increases with age in individual rodent tissues (e.g. (Chen et al., 2009; Leontieva et al., 2014a; Sengupta et al., 2010)), we and others who have performed more comprehensive studies have observed that aging increases mTORC1 signaling in only a few tissues, and in some tissues mTORC1 signaling even decreases with age (Baar et al., 2016; Houtkooper et al., 2011). We also observed sex-based differences in the phosphorylation of mTORC1 targets in the fasted states, as well as sex-based difference in the trajectory of mTORC1 signaling with age (Baar et al., 2016). Human data is limited, but transcriptional profiling of human blood suggests an age-associated decrease in mTOR signaling (Harries et al., 2012), while phosphoproteomic data on skeletal muscle is mixed (Francaux et al., 2016; Sandri et al., 2013). Collectively, these data suggest that while mTORC1 signaling may be elevated in select tissues of fasting aged animals, mTORC1 signaling does not generally become hyperactive with age. The potent effect of rapamycin on lifespan suggests that even normal levels of mTORC1 signaling may be inappropriately high for the maintenance of health in aging cells and tissues (Blagosklonny, 2009).

Excitingly, rapamycin not only extends lifespan, it also prevents or delays the onset of age-related diseases such as cancer and Alzheimer’s disease, rejuvenates the aging mouse heart, and ameliorates age-related cognitive decline (Dai et al., 2014; Flynn et al., 2013; Majumder et al., 2012; Ozcelik et al., 2013; Spilman et al., 2010; Wilkinson et al., 2012). Unfortunately, rapamycin has been proven to have numerous negative side effects that would seem to preclude its wide-scale use as an anti-aging compound. These effects include immunosuppression in humans – leading to an increased incidence of viral and fungal infections – as well as metabolic effects including hyperlipidemia and decreased insulin sensitivity in both mice and humans (reviewed in (Lamming et al., 2013) and (Arriola Apelo and Lamming, in press)).

Most use of rapamycin and its analogs everolimus and temsirolimus have been in transplant recipients or cancer patients, and rapamycin was approved last year for the treatment of lymphangiomyomatosis; a recent short-term trial in the elderly (Mannick et al., 2014) is a rare exception. This has led some to speculate that rapamycin therapy may not be toxic to healthy humans. However, everolimus was recently approved for the treatments of subependymal giant cell astrocytoma and renal angiomyolipoma in patients with tuberous sclerosis complex (TSC), who have generally normal immune systems and metabolism. In a recent study of TSC patients treated daily with everolimus, metabolic dysfunction and other side effects were observed in most individuals; more disturbingly, 2 of the 18 patients followed developed life-threatening infections, and one of these patients died (Trelinska et al., 2015). While this is potentially worrying from the standpoint of a potential anti-aging compound, the doses of rapamycin or its analogs used for cancer therapy or in TSC patients are typically quite high. The severity of side effects for rapamycin and its analogs is thought to be dose-dependent, and few adverse events were observed in elderly patients who took low doses of everolimus for only six weeks (Mannick et al., 2014). In mice, the effect of rapamycin on the immune system is less clear-cut, with rapamycin treatment resulting in negative effects on immune cell number and function, but with certain immunomodulatory effects of rapamycin possibly being beneficial (Goldberg et al., 2015; Goldberg et al., 2014; Hurez et al., 2015).

To elucidate the mechanism behind the effects of rapamycin on glucose tolerance, we treated mice with vehicle or rapamycin for two weeks and performed hyperinsulinemiceuglycemic clamps. We determined that these mice had increased hepatic glucose output and decreased hepatic insulin sensitivity (Lamming et al., 2012). Intriguingly, liver-specific deletion of Raptor had no effect on glucose tolerance, and we hypothesized that rapamycin might instead be acting via disruption of mTORC2. We determined that the effects of rapamycin on hepatic insulin sensitivity are mediated by disruption of mTORC2 utilizing a genetic mouse model of whole-body Rictor depletion (Lamming et al., 2012). While liver-specific deletion of Rictor is sufficient to impair glucose tolerance and increase hepatic gluconeogenesis (Hagiwara et al., 2012; Lamming et al., 2012; Yuan et al., 2012), our work has determined that in vivo, rapamycin inhibits both mTORC1 and mTORC2 in multiple tissues, including liver, skeletal muscle, and adipose tissue (Lamming et al., 2012; Schreiber et al., 2015), although mTORC2 inhibition is more variable than mTORC1 (Schreiber et al., 2015). The possible contribution of extra-hepatic tissues to the effects of rapamycin on hepatic insulin resistance remains an area for future study. While rapamycin impairs glucose tolerance and insulin sensitivity in both inbred and genetically heterogeneous mice (Liu et al., 2014b), rapamycin treatment induces adiponectin, improves insulin sensitivity, and promotes leanness in obese, insulin resistant db/db mice which lack a functional leptin receptor (Deepa et al., 2013). The influence of genetic background on mTORC2 function and the metabolic response to rapamycin also remains an area for future research.

Studies in the last few years have identified many important roles for mTORC2 in physiology and metabolism, and the ability of rapamycin to disrupt mTORC2 signaling in most tissues (Schreiber et al., 2015) suggests that many of the effects of rapamycin on metabolism and immunity may be due in part to disruption of mTORC2. For example, prolonged ex vivo exposure to rapamycin inhibits lipogenesis in rat hepatocytes (Brown et al., 2007), an effect we can now attribute in part to disruption of mTORC2 (Hagiwara et al., 2012; Yuan et al., 2012). Similarly, we now believe that the effect of rapamycin on the insulin content and mass of pancreatic beta cells in vivo is mediated in part by diminished mTORC2 signaling (Barlow et al., 2012; Yang et al., 2012). Recent work has identified many important roles for mTORC2 in the immune system, in T cells, B cells and macrophages (Byles et al., 2013; Festuccia et al., 2014). In T cells, mTORC2 promotes TH2 differentiation via the SGK1-mediated degradation of JunB (Heikamp et al., 2014). Intriguingly, deletion of SGK1 in T cells makes mice resistant to experimentally induced asthma and enhances the TH1-dependant response to viral and tumor challenges. These positive effects of SGK1 inhibition could help to explain why some studies have found a positive effect of rapamycin on the immune system (Hua et al., 2015; Hurez et al., 2015).

The discovery that rapamycin disrupts mTORC2 in vivo leads to the question of whether rapamycin extends lifespan solely through disruption of mTORC1 signaling. While it is clear that specific inhibition of mTORC1 is sufficient to extend lifespan (reviewed in (Lamming et al., 2013)), the impact of mTORC2 signaling on lifespan is confusing. In support of a model in which reduced mTORC2 signaling is beneficial to longevity, calorie restriction in humans reduces PI3K/Akt signaling (Mercken et al., 2013), and mice heterozygous for Akt1 display a significantly increased lifespan (Nojima et al., 2013). mTORC2 signaling is also significantly elevated in long-lived Snell dwarf mice and Growth Hormone Receptor knockout mice (Dominick et al., 2015). However, in support of a model in which inhibition of mTORC2 is deleterious for lifespan, we recently determined that genetic depletion of Rictor, a key component of mTORC2, severely shortens the lifespan of male, but not female, mice (Lamming et al., 2014b). Inhibition of the insulin/IGF-1/mTOR signaling pathway may be an essential mechanism through which pro-longevity interventions (e.g., rapamycin, calorie restriction) extend lifespan (Blagosklonny, 2009; Lamming, 2014; Lamming and Anderson, 2014). The data from mouse models is consistent with the possibility that decreased Akt signaling may be beneficial for lifespan, but that mTORC2 signaling through other downstream pathways such as SGK and PKC promotes male longevity.

We have suggested that more specific inhibition of mTORC1 is likely to retain the beneficial effects of rapamycin and is likely to result in reduced side effects (Lamming et al., 2013). As disruption of mTORC2 requires prolonged exposure to rapamycin, we recently hypothesized that the differential kinetics of mTORC1 and mTORC2 inhibition by rapamycin might create a therapeutic window through which intermittent administration of rapamycin might more specifically target mTORC1 (Arriola Apelo et al., 2016a). A sustained impact of rapamycin on lifespan would be consistent with an early study in a small number of mice showing that aged mice dosed with rapamycin for only 6 weeks had extended longevity (Chen et al., 2009). We identified an intermittent rapamycin treatment regimen with a reduced impact on the immune system as well as glucose metabolism, and we have now demonstrated that this regimen potently extends lifespan (Arriola Apelo et al., 2016a; Arriola Apelo et al., 2016b). Two other recent studies have reported beneficial results, including increased survival, from weekly administration of rapamycin in the context of diet-induced obesity (Leontieva et al., 2014b; Makki et al., 2014). Another possible dosing regimen, recently explored by Mannick and colleagues, is to deliver rapamycin at very low doses for only a short period time; they observed rejuvenation of the immune system in aged humans with comparatively low levels of serious side effects (Mannick et al., 2014).

Ultimately, the development of mTORC1-specific inhibitors will require a thorough understanding of the molecular mechanisms which underlying the differential sensitivity of cell lines and tissues to the disruption of mTORC2 by rapamycin. We recently took a major step in this direction, discovering that the sensitivity of cell lines and tissues is determined by the relative expression of levels of different FK506-binding proteins (FKBPs) (Schreiber et al., 2015). As outlined in Figure 4, we have discovered that rapamycin complexed with FKBP12 can inhibit mTORC2 when given chronically by preventing the formation of mTORC2. However, FKBP51 complexed with rapamycin can only inhibit mTORC1. The relative expression of FKBP12 and FKBP51 therefore determines the sensitivity of mTORC2 to rapamycin, with expression of FKBP12 correlating with increased sensitivity of mTORC2 to rapamycin, and the expression of FKBP51 inversely correlating with the sensitivity of mTORC2 to rapamycin (Schreiber et al., 2015). Rapamycin analogs with decreased binding affinity to FKBP12, or increased binding affinity to FKBP51, may therefore represent an appealing option for the pharmaceutical development of mTORC1-specific inhibitors.

Figure 4. The action of rapamycin against mTORC1 or mTORC2 is dependent upon FK506-binding proteins.

Rapamycin binding to either FKBP12 or FKBP51 forms a complex that can then act to inhibit mTORC1. However, inhibition of mTORC2 activity by rapamycin is dependent upon a rapamycin-FKBP12 complex that can prevent the formation of mTORC2 by binding to free mTOR, preventing the incorporation of mTOR into mTORC2. The relative expression level of FKBP12 and FKBP51 determines the rapamycin sensitivity of mTORC2 in each cell line or tissue.

The future of mTOR research

The past few years have seen a veritable explosion in the literature on our knowledge of how mTOR signaling is regulated at molecular level and the regulation of metabolism by mTOR signaling at both the molecular, cellular and organismal level. However, there are still major unanswered questions that need to be addressed in order to complete our understanding of the regulation and role of mTOR in the regulation of metabolism, and some of these will need to be addressed in order to successfully translate our new knowledge into effective therapies for age-related diseases. While we now have a very detailed understanding of mTORC1 regulation by many amino acids at the lysosomal surface, new findings demonstrating both amino acid sensitive and insensitive mTORC1 signaling elsewhere in the cell, including the nucleus (Zhou et al., 2015), show that we still have much to learn. While our new understanding of mTORC1 signaling at the lysosome presents novel pharmaceutical targets, the role of mTORC1 signaling at this location vs. elsewhere in the cell is unclear. We still do not know which substrates downstream of mTORC1 mediate the positive effects of rapamycin on lifespan, nor do we know what tissues are critical to enhance longevity and healthspan. Indeed, it is possible that inhibition of mTORC2 in specific tissues may even contribute to the beneficial effects of rapamycin on health and longevity, but as we do not yet possess a comprehensive understanding of the pathways mediated by mTORC2.

From the perspective of using mTOR inhibition as a technique to combat age-related diseases, it is still early days. While a recent short-term study showed that a low-dose, short term course of Everolimus boosted immune response in the elderly (Mannick et al., 2014), an appropriate dosing regimen for fighting or delaying chronic diseases has not yet been determined, and a better molecular understanding of the basis for the beneficial effects of rapamycin will need to be developed in order to determine appropriate biomarkers for such a study. Recent efforts to define the minimal and/or intermittent rapamycin dosing regimen necessary to promote health and longevity in mice will significantly advance the effort to unleash the potential benefits of therapies based on mTOR inhibition. Unrealized side effects may also emerge, such as the recent discovery that mTORC1 inhibition can accelerate the growth of solid tumors in a mouse model of pancreatic cancer by stimulating the catabolism of extracellular proteins (Palm et al., 2015). Ongoing testing of rapamycin in higher organisms, including canines (Kaeberlein, 2015; Kaeberlein et al., 2016) and non-human primates (Tardif et al., 2015) will further enhance our understanding of the likely beneficial effects and possible negative consequences of treating humans with mTOR inhibitors. However, it is likely that the wide-scale use of mTOR inhibition in the clinic will await novel pharmaceutical approaches that take advantage of the recent molecular findings described above to specifically target mTORC1.