Where is mTOR and what is it doing there?

Target of rapamycin (TOR) forms two conserved, structurally distinct kinase complexes termed TOR complex 1 (TORC1) and TORC2. Each complex phosphorylates a different set of substrates to regulate cell growth. In mammals, mTOR is stimulated by nutrients and growth factors and inhibited by stress to ensure that cells grow only during favorable conditions. Studies in different organisms have reported localization of TOR to several distinct subcellular compartments. Notably, the finding that mTORC1 is localized to the lysosome has significantly enhanced our understanding of mTORC1 regulation. Subcellular localization may be a general principle used by TOR to enact precise spatial and temporal control of cell growth.

Introduction

Rapamycin, an antifungal, anticancer, and immunosuppressive compound produced by a soil bacterium from Rapa Nui (better known as Easter Island), was discovered in 1975 (Vézina et al., 1975; Benjamin et al., 2011). The isolation of yeast mutants resistant to the growth-inhibitory properties of rapamycin led to the discovery of TOR (target of rapamycin; Heitman et al., 1991; Kunz et al., 1993). It was subsequently found that TOR is a highly conserved controller of cell growth and that mammalian TOR (mTOR) is implicated in human disease (Menon and Manning, 2008; Dazert and Hall, 2011; Laplante and Sabatini, 2012).

The protein kinase TOR functions in two structurally and functionally distinct multiprotein complexes termed TOR complex 1 (TORC1 in yeast, mTORC1 in mammals) and TOR complex 2 (TORC2 in yeast, mTORC2 in mammals; Wullschleger et al., 2006; Loewith and Hall, 2011; Laplante and Sabatini, 2012). mTORC1 is composed of mTOR, raptor, and mLST8 (Saccharomyces cerevisiae orthologues are TOR1, Kog1, and LST8, respectively). mTORC1 regulates cell growth (accumulation of cell mass) through coordination of protein anabolism (Averous and Proud, 2006; Ma and Blenis, 2009), nucleotide biosynthesis (Ben-Sahra et al., 2013; Robitaille et al., 2013), lipogenesis, glycolysis (Laplante and Sabatini, 2009; Peterson et al., 2011), and autophagy (Ganley et al., 2009; Hosokawa et al., 2009). mTORC2 is composed of mTOR, rictor, SIN1, and mLST8 (S. cerevisiae orthologues are TOR2, Avo3, Avo1, and LST8, respectively). mTORC2 controls growth by regulating lipogenesis, glucose metabolism (García-Martínez and Alessi, 2008; Hagiwara et al., 2012; Yuan et al., 2012), the actin cytoskeleton (Cybulski and Hall, 2009; Oh and Jacinto, 2011), and apoptosis (Datta et al., 1997).

TOR has been found at several cellular locations (Tables 1 and 2; Malik et al., 2013), which has brought cell biology to the forefront of the TOR signaling field. In this review, we discuss the subcellular localization of the TOR complexes vis-à-vis their function and regulation. However, before starting our discussion, it is important to note several caveats in determining the subcellular location of a protein or complex. Antibodies can be nonspecific, overexpressed or tagged proteins can exhibit aberrant localization, different fixation or lysis methods can influence localization, and isolated organelles can be contaminated with other organelles. Additionally, detection of one component of a TOR complex does not necessarily reflect localization of an entire complex. Furthermore, especially when dealing with highly regulated pathways, it is essential to take nutrient, stress, and cell cycle status into account and to consider that commonly used cell lines present mutations that might affect subcellular localization. Thus, when evaluating localization of TOR, or any other protein, it is advisable to consider several complementary approaches because no single technique is without weakness.

Table 1.

Lysosome localization of TORC1

Pathway	Mechanism of localization	Cell type	Technique	Reference
mTORC1	Rheb localizes to the late endosome/lysosome	MDCK; HeLa; 1321N1	Confocal live-cell imaging; expression of GFP-Rheb	Saito et al., 2005
mTORC1	The TSC complex is at the lysosome	HeLa	Wide-field imaging; PFA fixation; endogenous TSC2	Dibble et al., 2012
mTORC1	Amino acids stimulate mTOR translocation from the cytoplasm to the lysosome in a Rag- and Ragulator-dependent manner; Rheb, Ragulator, and Rags are at the lysosome independently of amino acids	HEK293T	Confocal imaging; PFA fixation; endogenous mTOR, raptor, RagA/B, RagC; validation of antibody expression of GFP-Rheb, GFP-RagB/C/D, GFP-LAMTOR1, HA-raptor, myc-mTOR, FLAG-raptor, FLAG-LAMTOR4, FLAG-LAMTOR5	Sancak et al., 2008, 2010; Bar-Peled et al., 2012
mTORC1	Glutaminolysis stimulates mTOR translocation from the cytoplasm to the lysosome	U2OS	Wide-field imaging; PFA fixation; endogenous mTOR	Durán et al., 2012b
Yeast TORC1	Tco89 (component of TORC1) is at the plasma membrane and the vacuole/lysosome	S. cerevisiae	Immuno-EM; PFA fixation; genomically tagged Tco89-Myc	Reinke et al., 2004
Yeast TORC1	TORC1 (Tco89; Kog1; Sch9; TOR1) is mainly at the vacuole/lysosome	S. cerevisiae	Live-cell, wide-field, and confocal imaging; genomically tagged Tco89-GFP, Kog1-GFP, Sch9-GFP, TOR1-GFP	Huh et al., 2003; Urban et al., 2007; Sturgill et al., 2008
Yeast TORC1	TORC1 (Tco89; Ego1; Sch9; TOR1) is mainly at the vacuole/lysosome, and this localization is independent of amino acid (Leu) availability	S. cerevisiae	Live-cell imaging; genomically tagged Tco89-GFP, Ego1-GFP, Sch9-GFP, TOR1-GFP	Binda et al., 2009

This table groups the most important reports that link TORC1 and mTORC1 to the lysosome. By using a range of different techniques, distinct components of TORC1 have consistently been identified at the lysosome in different cell types. PFA, paraformaldehyde; TSC, tuberous sclerosis complex.

Table 2.

Localization of TORC1 to other organelles

Pathway	Mechanism of localization	Cell type	Technique	Reference
mTORC1	Rheb is at the ER and the Golgi	HEK293	Confocal imaging; expression of EGFP-Rheb	Buerger et al., 2006
mTORC1	Rheb and the TSC complex are at the peroxisome	FAO; HepG2; MEF; HeLa	Confocal imaging and subcellular fractionation; quantification of colocalization; endogenous TSC1, TSC2, and Rheb	Zhang et al., 2013
mTORC1	The TSC complex is mainly cytoplasmic	COS; HeLa	Wide-field imaging and subcellular fractionation; PFA fixation; expression of TSC1 and TSC2	van Slegtenhorst et al., 1998; Nellist et al., 1999
mTORC1	The TSC complex is cytoplasmic and nuclear; Akt stimulates TSC translocation from the cytoplasm to the nucleus	NIH3T3; HeLa; HEK293; Rat-1	Subcellular fractionation	Rosner and Hengstschläger, 2007; Rosner et al., 2007,
mTORC1	mTOR is at ER, Golgi, and in the nucleus; Rheb is cytoplasmic and nuclear; mTORC1 (mTOR, raptor) is in punctate structures upon amino acid starvation	HEK293; CHO; HeLa	Single photon FRET-FLIM on live cells; expression of mTOR-EGFP, Ds-Red-raptor, and Rheb-EGFP	Yadav et al., 2013
mTORC1	mTOR and raptor are highly abundant in the nucleus but mTORC1 integrity is higher in the cytoplasm	HEK293; NIH3T3; IMR-90; MRC-5; WI-38	Subcellular fractionation	Rosner and Hengstschläger, 2008
mTORC1	mTORC1 (mTOR, raptor) is in the nucleus and the cytoplasm; mTORC1 phosphorylates NFACTc4 in the nucleus; mTOR binds to rDNA, rDNA, and tRNA genes in a rapamycin-sensitive manner; serum stimulates mTOR binding to hnRNPs	COS; HEK293; HeLa	Subcellular fractionation; ChIP	Yang et al., 2008; Goh et al., 2010; Kantidakis et al., 2010; Shor et al., 2010; Tsang et al., 2010
mTORC1	During hypoxia, PML inhibits mTORC1 (mTOR) by sequestering it in the nucleus away from cytoplasmic Rheb	MEF; HEK293	Subcellular fractionation; confocal imaging; PFA fixation	Bernardi et al., 2006
mTORC1/2	mTOR shuttles from the cytoplasm to the nucleus	HEK293; CV-1	Wide-field imaging and subcellular fractionation; PFA fixation; expression of mTOR-FLAG	Kim and Chen, 2000
mTORC1/2	mTOR is predominantly nuclear except in HEK293, where it is excluded from the nucleus	HEK293; Rh30; Rh41; IMR90; HCT8; HCT29; HCT116	Confocal imaging and subcellular fractionation; PFA fixation; endogenous mTOR	Zhang et al., 2002
mTORC1	mTOR and raptor are associated with mitochondria; mTOR is associated with the mitochondrial channel VDAC	Jurkat; HEK293	Subcellular fractionation	Schieke et al., 2006; Ramanathan and Schreiber, 2009
Yeast TORC1	TOR1 is mainly nuclear, and nutrient starvation or rapamycin treatment induce translocation to the cytoplasm	S. cerevisiae	Wide-field imaging and subcellular fractionation; endogenous TOR1	Li et al., 2006

This table groups the most important reports that link TORC1 and mTORC1 to sites apart from the lysosome. By using comparable techniques to those that revealed lysosomal localization of mTORC1 (Table 1), distinct components of TORC1 have also been identified at other sites including the nucleus and mitochondria. ChIP, chromatin immunoprecipitation; PFA, paraformaldehyde.

Localization of mTORC1

mTORC1 at the lysosome.

mTORC1 is activated directly by GTP-bound Rheb on the surface of the lysosome (Fig. 1, Table 1). Two conditions need to be fulfilled for mTORC1 to be activated. One is that mTORC1 needs to translocate to the lysosome, a process stimulated by nutrients and the Rags (see next paragraph), where it encounters Rheb. The other is that Rheb needs to be activated, i.e., converted from a GDP- to a GTP-bound form, in response to growth factors. Rheb is a farnesylated GTPase that is anchored to the surface of the lysosome (Saito et al., 2005; Sancak et al., 2008). Rheb is inhibited by its GAP, a heterotrimer of TSC1 (tuberous sclerosis complex 1), TSC2, and TBC1D7 that is also on the lysosome. Growth factor–stimulated Akt phosphorylates and inhibits the TSC complex to activate Rheb, but it is not known where Akt meets the TSC complex. Akt is generally assumed to be activated at the plasma membrane by PDK1 after growth factor stimulation (Hemmings and Restuccia, 2012).

Figure 1.

Localization of mTORC1 signaling. mTORC1 is in the cytoplasm when amino acids levels are low. Addition of amino acids stimulates the recruitment of mTORC1 in a Rag-dependent manner to the lysosomal surface. Upon growth factor stimulation, PI3K produces PIP3 in the plasma membrane, which in turn activates PDK1 and Akt. After phosphorylation by PDK1, Akt phosphorylates and thereby inhibits the TSC complex, possibly at the lysosome and the peroxisome. Reduced TSC complex GAP activity leads to an increase in GTP-bound Rheb. Rheb-GTP at the lysosomal surface directly binds to and activates mTORC1. Osmotic stress induces sequestration of mTORC1 in stress granules. Pools of mTORC1 have also been reported at other sites, including mitochondria and the nucleus.

mTORC1 translocation to the lysosome is controlled by the Rag GTPase and the so-called Ragulator, in response to amino acids (Kim et al., 2008; Sancak et al., 2008, 2010; Jewell et al., 2013). Rag is a heterodimeric GTPase consisting of RagA or B in complex with RagC or D. Ragulator, the GEF and lysosomal anchor for Rag, is a pentameric complex consisting of p14, p18, MP1, HBXIP, and C7orf59 (also known as LAMTOR1–5; Bar-Peled et al., 2012). Amino acids stimulate guanine nucleotide exchange in Rag. How does Rag sense amino acids to activate mTORC1? There may be more than one mechanism. Zoncu et al. (2011) have proposed that amino acids are sensed in the lumen of the lysosome via the v-ATPase, which then activates Rag via Ragulator and possibly the recently described Rag GAPs (Bar-Peled et al., 2013; Panchaud et al., 2013; Petit et al., 2013; Tsun et al., 2013). Activation of Rag requires ATP hydrolysis by the v-ATPase, but the function of the v-ATPase in establishing a lysosomal proton gradient appears to be dispensable. Ögmundsdóttir et al. (2012) have reported that the lysosomal proton-assisted amino acid transporter PAT1, which pumps amino acids out of the lysosomal lumen, interacts directly with Rag to activate mTORC1. However, as part of their argument that amino acids are sensed in the lysosomal lumen, Zoncu et al. (2011) reported that PAT1 overexpression inactivates mTORC1. The mechanism by which the v-ATPase actually senses amino acids to activate Ragulator–Rag remains to be determined. Durán et al. (2012b) have proposed that leucine (and glutamine) is sensed via glutaminolysis. Leucine is a particularly effective activator of Rag. Leucine is also an allosteric activator of glutamate dehydrogenase (GDH) that deaminates glutamate to produce α-ketoglutarate. α-Ketoglutarate is a cofactor for prolyl hydroxylases which may in turn, by an unknown mechanism, activate Rag (Durán et al., 2012a). Interestingly, under prolonged amino acid starvation, mTORC1 is reactivated by amino acids derived from autophagosomal lysis, thus preventing further autophagy (Yu et al., 2010).

Active Rag recruits mTORC1 to Rheb on the surface of the lysosome. Rag was originally suggested to deliver mTORC1 from the cytoplasm to the lysosome, with the implication that Rag itself shuttles on and off the lysosome (Sancak et al., 2008; Zinzalla and Hall, 2008). However, a more recent study that directly examined the cellular location of Rag showed that it appears to be fixed to the lysosome, from where it recruits mTORC1 to the lysosomal surface (Sancak et al., 2010). Most recently, it has again been suggested that Rag captures mTORC1 in the cytoplasm and then shuttles it back to the lysosomal surface (Bar-Peled et al., 2012). Furthermore, the E3 ubiquitin ligase TRAF6 and the signaling adaptor p62 appear to be involved in the recruitment of mTORC1 to the lysosome in response to amino acids (Linares et al., 2013). Thus, the mechanism by which Rag mediates lysosomal localization of mTORC1 remains to be determined. Once on the lysosome, mTORC1 forms a four-component super complex with v-ATPase, Ragulator, and Rag, as suggested by coimmunoprecipitation of mTORC1 with at least Ragulator and Rag (Sancak et al., 2010).

mTORC1 signaling in cells lacking TSC is resistant to growth factor withdrawal but still responsive to amino acids (Smith et al., 2005; Roccio et al., 2006). This indicates that to activate mTORC1, Rheb needs amino acid–induced targeting of mTORC1 to the lysosome. On the other hand, forced localization of mTORC1 to the lysosome renders cells insensitive to amino acids and independent of Rag, but still dependent on Rheb and growth factors (Sancak et al., 2010). Overexpression of Rheb can activate mTORC1 even under amino acid starvation, possibly by enabling interaction between mislocalized Rheb and mTORC1 at a site(s) away from the lysosome (Inoki et al., 2003; Smith et al., 2005). Interestingly, amino acid starvation changes the location of the entire lysosome within the cell (Korolchuk et al., 2011), but the effect of this altered lysosomal localization on mTORC1 activity is unknown. Inhibition of mTORC1 activity with rapamycin does not affect mTORC1 localization to the lysosome (Sancak et al., 2008), and kinase-dead mTORC1 still localizes to lysosomes (Tabatabaian et al., 2010). Finally, energy stress prevents assembly and lysosomal localization of mTORC1, through the Tel2-Tti1-Tti2 (TTT)–RUVBL1/2 complex (Kim et al., 2013). Tel2 and Tti1 are critical factors for the assembly and stability of mTORC1/2 and other PIKK family members (Takai et al., 2007; Kaizuka et al., 2010).

What does mTORC1 do once it is activated on the surface of the lysosome? Does it remain on the lysosome to phosphorylate substrates locally? mTORC1 phosphorylates TFEB, a transcription factor involved in autophagy, at the lysosome (Martina et al., 2012; Martina and Puertollano, 2013). However, presumably active mTORC1 is also found at other cellular locations (see next subheadings), yet Rheb and Rag are required for activation of mTORC1 toward all examined substrates. It is unclear whether mTORC1 active in the cytoplasm (or in any other nonlysosomal location) remains bound to Rheb or, alternatively, it retains a “memory” of Rheb. We note that Rheb or TSC have also been identified on endomembranes other than the lysosome (Table 2; Henry et al., 1998; Plank et al., 1998; van Slegtenhorst et al., 1998; Nellist et al., 1999; Noonan et al., 2002; Murthy et al., 2004; Teleman et al., 2005; Buerger et al., 2006; Dibble et al., 2012; Yadav et al., 2013; Zhang et al., 2013). Of particular interest is the recent finding that both Rheb and TSC are on the peroxisome (Zhang et al., 2013). mTORC1 itself has so far not been found on the peroxisome but is inhibited by radical oxygen species generated in the peroxisome. The significance of TSC and Rheb on the peroxisome and whether Rag (or another recruitment factor) is also on the peroxisome should be investigated further. Future studies may reveal that mTORC1 is independently activated at different locations, possibly in response to different cues, to phosphorylate physically separate substrates. To determine what mTORC1 is doing where will require defining where it phosphorylates particular substrates.

mTORC1 localization to lysosomes is usually monitored by immunofluorescence (IF) using an antibody against mTOR (Table 1). In these experiments, mTOR is dispersed (cytoplasmic) under amino acid–starved conditions, and punctate/lysosomal upon re-addition of amino acids. An mTOR antibody should theoretically detect both mTORC1 and mTORC2, but mTORC2 does not localize to the lysosome and is unresponsive to amino acids. Thus, the mTOR antibody that is commonly used in these studies recognizes mTOR only in mTORC1, possibly because the epitope is masked in mTORC2. Another possibility would be that mTORC1 is much more abundant than mTORC2; however, immunoprecipitation experiments suggest this is not the case (Rosner and Hengstschläger, 2008). Other antibodies that detect mTOR at mitochondria and the ER might detect an epitope exposed only in mTORC2. Finally, immunofluorescence experiments with antibodies against mTOR or raptor have generally detected only cytoplasmic/lysosomal mTORC1, yet other approaches have revealed mTORC1 at other sites (discussed later). This may be due to mTORC1 being mainly cytoplasmic/lysosomal. Alternatively, localization to other sites may require specific conditions such as stress.

mTORC1 in the nucleus.

Components of the mTORC1 signaling pathway, including mTOR, raptor, and S6K, have been detected in the nucleoplasm (Table 2; Kim and Kahn, 1997; Kim and Chen, 2000; Zhang et al., 2002; Kikani et al., 2005; Lian and Di Cristofano, 2005; Rosner et al., 2007; Yadav et al., 2013). However, although mTOR and raptor are detectable inside the nucleus where they may have a role in transcription regulation (Bernardi et al., 2006; Cunningham et al., 2007; Yang et al., 2008; Goh et al., 2010; Kantidakis et al., 2010; Shor et al., 2010; Tsang et al., 2010), these nuclear versions appear not to form an intact mTORC1 (Fig. 1; Rosner and Hengstschläger, 2008).

mTORC1 in mitochondria.

mTOR has been found at mitochondria (Fig. 1), and rapamycin treatment affects mitochondrial function in Jurkat cells (Paglin et al., 2005; Schieke et al., 2006; Ramanathan and Schreiber, 2009). However, it is unclear whether mTORC1 localizes to mitochondria because mTOR, instead of complex-specific rictor or raptor, was probed to investigate mTORC1 localization. Although rapamycin acutely inhibits only mTORC1, prolonged rapamycin treatment can also inhibit mTORC2 in certain cells (including Jurkat cells; Sarbassov et al., 2006). We speculate that at least some studies might have observed mTORC2 rather than mTORC1 at mitochondria (discussed later).

mTORC1 in stress granules.

mTORC1 is sequestered in stress granules in response to stress in an astrin-dependent manner (Fig. 1; Thedieck et al., 2013). This localization is mediated by physical interaction with the kinase DYRK3 but is independent of DYRK3 kinase activity (Wippich et al., 2013). Yeast TORC1 is similarly targeted to stress granules in response to heat stress (Takahara and Maeda, 2012). Localization of mTORC1 to stress granules is a mechanism to sequester mTORC1 and thus to arrest growth in unfavorable conditions.

mTORC1 at the plasma membrane.

A recent report suggests that the WD40 domain in raptor binds the lipid PI(3,5)P2, thereby targeting mTORC1 to the plasma membrane (Bridges et al., 2012). The authors observed that this lipid is generated at the plasma membrane and the lysosome, depending on the cell type, after insulin or amino acid stimulation, and speculate that PI(3,5)P2 binding could also contribute to lysosomal targeting. Interestingly, both PI(3,5)P2 and TORC1 are enriched at discrete sites on the yeast lysosome (known as the vacuole in yeast; Han and Emr, 2011), but it has yet to be determined whether TORC1 colocalizes with PI(3,5)P2.

mTORC1 in the cytoplasm.

mTORC1, in particular under amino acid starvation conditions, exhibits a diffuse, cytoplasmic distribution. However, it may also have a function in the cytoplasm when active. mTORC1 associates with the general translation initiation complex eIF3 and phosphorylates the translation inhibitor 4E-BP upon stimulation by growth factors and nutrients to promote translation initiation (Holz et al., 2005; Harris et al., 2006; Proud, 2009; Sonenberg and Hinnebusch, 2009). These events presumably take place in the cytoplasm.

Yeast TORC1.

TORC1 in S. cerevisiae is on the limiting membrane of the vacuole, the major nutrient reservoir in yeast cells (Reinke et al., 2004; Urban et al., 2007; Berchtold and Walther, 2009; Binda et al., 2009). The yeast vacuole is equivalent to the lysosome in higher organisms. At the vacuole, TORC1 localizes to discrete, PI3P-enriched subdomains (Sturgill et al., 2008). Interestingly, in sharp contrast to mTORC1 at lysosomes, yeast TORC1 localization to the vacuole is independent of nutrient availability (Binda et al., 2009). TORC1 at the vacuole/lysosome in both yeast and higher eukaryotes is the best-characterized localization of TORC1. However, as in mammals, the precise function of TORC1 at the vacuole/lysosome is unknown. A small fraction of yeast TORC1 is near the plasma membrane (Wedaman et al., 2003; Reinke et al., 2004). Curiously, Li et al. (2006) have reported that TORC1 is mainly nuclear.

TORC1 in other organisms.

TORC1 in the alga Chlamydomonas is localized to the ER and regulates Bip (Grp78) phosphorylation (Díaz-Troya et al., 2008, 2011). TORC1 in plants is cytoplasmic and nuclear and regulates development and growth (Ren et al., 2011). In Drosophila, TOR appears to localize to vesicular structures, possibly the lysosome (Hennig et al., 2006), and to a perinuclear structure, possibly the ER (Chang and Neufeld, 2009). Although TORC1 is highly conserved in eukaryotes, subcellular localization in organisms such as worms or fish has to our knowledge not been reported.

Localization of mTORC2

mTORC2 at mitochondria-associated ER membrane.

mTORC2 interacts with the ER proteins Hsp70 and Grp58 (Martin et al., 2008; Ramírez-Rangel et al., 2011), is sensitive to ER stress (Hosoi et al., 2007; Yung et al., 2011; Chen et al., 2011; Appenzeller-Herzog and Hall, 2012), and is localized to the ER (Boulbés et al., 2011). It is not fully understood which steps in mTORC2 signaling occur at the ER and in what chronological order. Akt, a major, direct mTORC2 substrate, is also found at the ER (Hresko and Mueckler, 2005; Boulbés et al., 2011), although the first step in Akt activation—PDK1-mediated phosphorylation of the activation loop in Akt—is generally thought to occur at the plasma membrane. However, PDK1, PTEN (the PIP3 phosphatase), and PI3K have been observed on other organelles, including the ER (Daniele et al., 1999; Lim et al., 2003; Downes et al., 2004). It is thus possible that the first step in Akt activation might also occur at the ER. Alternatively, PDK1-phosphorylated Akt may translocate to the ER to be phosphorylated by mTORC2. It is unclear whether reported cotranslational phosphorylation of Akt and IMP1 by mTORC2 (Oh et al., 2010; Dai et al., 2013) occurs at the ER.

There are longstanding indications that mTORC2 is also associated with mitochondria. Mitochondrial membrane potential, respiration, and the phosphorylation status of several mitochondrial proteins are changed upon rapamycin treatment or rictor knockdown in Jurkat cells (Desai et al., 2002; Schieke et al., 2006). As noted earlier, mTORC2 is highly sensitive to rapamycin in Jurkat cells (Sarbassov et al., 2006). Furthermore, mTOR localizes to structures in close proximity to the mitochondrial outer membrane (OMM; Desai et al., 2002) and interacts with VDAC, an OMM channel (Ramanathan and Schreiber, 2009), although these two studies did not address whether the observed mTOR localization was mTORC1 or mTORC2. We suggest that mitochondrial mTOR may correspond to mTORC2. This would be consistent with other studies linking mTORC2 to mitochondrial function (Wang et al., 2010; Colombi et al., 2011; Murata et al., 2011; Hagiwara et al., 2012; Yuan et al., 2012; Wu et al., 2013). Furthermore, Akt has been detected at mitochondria (Bijur and Jope, 2003; Miyamoto et al., 2008; Antico Arciuch et al., 2009; Su et al., 2012), and another direct substrate of mTORC2, SGK1 (serum- and glucocorticoid-inducible kinase 1), is primarily localized to mitochondria (Engelsberg et al., 2006; Cordas et al., 2007). PTEN has been observed at mitochondria (Zhu et al., 2006; Zu et al., 2011) and mitochondria-associated ER membranes (MAMs; Bononi et al., 2013), although it is unclear whether mitochondrial PTEN functions as a PIP3 phosphatase.

Is there a link between mitochondrial mTORC2 and ER-associated mTORC2? MAM is a subdomain of the ER that is physically tethered to mitochondria (Raturi and Simmen, 2012). The main function of MAM is to facilitate the transfer of lipids and calcium between the two organelles. MAM thereby controls mitochondrial metabolism and apoptosis (Rizzuto et al., 1998; Csordás et al., 1999). Several observations have linked mTORC2 to MAM. First, mTOR and mLST8 interact with VDAC1 (Ramanathan and Schreiber, 2009) and Grp75 (Behrends et al., 2010), respectively. VDAC1 together with Grp75 and IP3R tethers mitochondria to the ER to form MAM (Szabadkai et al., 2006). Second, the MAM resident proteins IP3R, PACS2, and the VDAC-interacting protein hexokinase 2 (HK2) are Akt substrates (Simmen et al., 2005; Khan et al., 2006; Szado et al., 2008; Aslan et al., 2009; Marchi et al., 2012), and Akt itself is associated with MAM (Giorgi et al., 2010). Third, MAM appears to have a role in modulating ER stress (Simmen et al., 2010; Malhotra and Kaufman, 2011; Verfaillie et al., 2012), which in turn can inhibit mTORC2 (Appenzeller-Herzog and Hall, 2012). Fourth, liver-specific knockout of Mfn2 (Sebastián et al., 2012), a key MAM protein, confers a whole body phenotype strikingly similar to that of liver-specific rictor knockout (Hagiwara et al., 2012). Fifth, a recent proteomic study on MAM detected mTOR (Poston et al., 2013). These results suggest that the previously described mitochondrial and ER-associated mTORC2 might actually be at MAM.

A recent study that directly investigated mTORC2 localization confirmed that mTORC2 indeed localizes to MAM (Fig. 2, Table 3; Betz et al., 2013). mTORC2 localization to MAM is stimulated by growth factors, and mTORC2 at MAM controls Akt and the Akt targets PACS2, IP3R, and HK2 to ultimately control MAM integrity, calcium release, and mitochondrial physiology (Betz et al., 2013). Thus, it appears that MAM is a hub for mTORC2-Akt signaling. MAM regulates many physiological functions including mitochondrial metabolism, the synthesis and maturation of cholesterol, phospho- and sphingolipids, autophagy, and ER stress (Fujimoto et al., 2012; Raturi and Simmen, 2012; Verfaillie et al., 2012; Hamasaki et al., 2013), and is implicated in a number of diseases including diabetes, neurodegeneration, and cancer (Bononi et al., 2012; Leem and Koh, 2012). Interestingly, mTORC2 is also linked to many of these processes and diseases, suggesting that they may be due to mTORC2 function (or dysfunction) at the MAM signaling hub.

Figure 2.

Localization of mTORC2 signaling. mTORC2 interacts with ribosomes in a PI3K-dependent manner. Upon growth factor stimulation, mTORC2 is recruited to MAMs, presumably from the cytoplasm. mTORC2 has also been observed in the nucleus and on lipid rafts at the plasma membrane.

Table 3.

Mitochondria, ER, MAM, and other localization of mTORC2

Pathway	Mechanism of localization	Cell type	Technique	Reference
mTORC2	mTORC2 (mTOR, rictor) is mainly at the ER	MDA-MB-435; A549	Confocal imaging and subcellular fractionation; PFA fixation; endogenous mTOR and rictor	Boulbés et al., 2011
mTORC1/2	mTOR is associated with mitochondria	NIH3T3	Confocal imaging, subcellular fractionation, and immuno-EM; PFA fixation; endogenous mTOR	Desai et al., 2002
mTORC2	mTORC2 (mTOR; Sin1; rictor) is at MAM	HeLa; MEF; mouse liver	Confocal imaging, subcellular fractionation, and immuno-EM; PFA fixation; endogenous mTOR, Sin1, and rictor; validation of antibody	Betz et al., 2013; Poston et al., 2013
mTORC2	A part of mTORC2 (mTOR; rictor; Sin1) is nuclear; rapamycin treatment induces translocation of mTORC2 to the cytoplasm	HEK293; IMR-90; NIH3T3	Subcellular fractionation	Rosner et al., 2007; Rosner and Hengstschläger, 2008, 2011, 2012
Yeast TORC2	TORC2 (TOR2; Bit61; Avo1-3) is at or near the plasma membrane, possibly cortical ER	S. cerevisiae	Subcellular fractionation, immuno-EM, live-cell imaging, and wide-field and confocal imaging; PFA fixation; TOR2-GFP, Bit61-GFP, Avo1-3-GFP	Kunz et al., 2000; Wedaman et al., 2003; Aronova et al., 2007; Sturgill et al., 2008; Berchtold and Walther, 2009

This table groups the most important reports that investigate TORC2 and mTORC2 localization. mTORC2 has consistently been identified in proximity to mitochondria, the ER, and MAM. Other studies suggest that mTORC2 is nuclear or at the plasma membrane. EM, electron microscopy; PFA, paraformaldehyde.

How does mTORC2 localization to MAM relate to activation of mTORC2 by ribosome association? Electron microscopy images (Ruby et al., 1969; Csordás et al., 2006; Lebiedzinska et al., 2009) and proteomic profiling (Poston et al., 2011, 2013; Zhang et al., 2011) of MAM suggest that ribosomes are present at this ER subdomain. Calnexin, a MAM-enriched chaperone (Lynes et al., 2012), interacts with the MAM protein PACS2 and anchors ribosomes at the ER and possibly MAM (Lakkaraju et al., 2012). The role of MAM ribosomes has yet to be defined. However, certain viruses that replicate at MAM, such as CMV, actively recruit the translation machinery to MAM (Zhang et al., 2011), suggesting that ribosomes at MAM are required for translation. Interestingly, CMV infection also increases mTORC2 activity (Clippinger et al., 2011), and a MAM deficiency results in reduced mTORC2 activity (Sebastián et al., 2012). Thus, it appears that ribosomes at MAM and activation of mTORC2 are functionally linked. A recent report suggests that ATP regulates mTORC2 integrity (Chen et al., 2013). At MAM, mTORC2 is in a complex with VDAC, the exit channel for mitochondrial ATP, thus providing more evidence that MAM localization is important for mTORC2 activity. Altogether, the above suggests that MAM is a signaling platform for mTORC2 like the lysosome is for mTORC1 (Fig. 2).

mTORC2 at the plasma membrane.

mTORC2 and its substrate Akt have been isolated from lipid rafts traditionally associated with the plasma membrane (Fig. 2; Hill et al., 2002; Partovian et al., 2008; Gao et al., 2011), although lipid raft domains also exist in the ER, where they form MAM (Poston et al., 2011; see previous section). Evaluation of rictor localization by immunofluorescence (Zhang et al., 2010) and subcellular fractionation (Boulbés et al., 2011) suggests that only a minor fraction of mTORC2 is at the plasma membrane.

mTORC2 in the nucleus.

mTORC2 has been reported to shuttle between the cytoplasm and the nucleoplasm (Table 3, Fig. 2; Rosner et al., 2007; Rosner and Hengstschläger, 2008, 2011, 2012). A nuclear function for mTORC2 is so far unknown.

TORC2 in yeast.

TORC2 in yeast is at or near the plasma membrane (Kunz et al., 2000; Wedaman et al., 2003; Aronova et al., 2007; Sturgill et al., 2008; Berchtold and Walther, 2009). TORC2 near the plasma membrane might actually be at cortical ER, as suggested by its association with distinct membranous tracks just under the plasma membrane (Wedaman et al., 2003). It is currently unknown whether TORC2 can also localize to MAM in yeast, as this structure is structurally and functionally different from mammalian MAM (Kornmann et al., 2009; Michel and Kornmann, 2012).

TORC2 in other organisms.

TORC2 in Trypanosomes localizes to both the ER and mitochondria (Barquilla et al., 2008). Although present in Dictyostelium, Drosophila, Caenorhabditis, and zebrafish, TORC2 localization has not yet been investigated in these organisms.

Considerations and future directions

A major question is, why is mTOR at different cellular locations? The initial response to this question is that mTOR phosphorylates functionally distinct substrates at different locations. Another possibility is that mTORC1 and mTORC2 sense different inputs at different locations. For example, mTORC1 at the lysosome senses amino acids, whereas mTORC1 at the peroxisome would sense peroxisome-generated hydrogen peroxide (Benjamin and Hall, 2013). The answer to the above question will be revealed as we learn more about the cellular biology of mTOR.

Several mTOR complex components, including mTOR, rictor, raptor, and Sin1 are phosphorylated in vivo. Although the functionality of most of these modifications is unclear, posttranslational modifications (PTMs), including phosphorylation, are prime suspects in regulating protein targeting. Farnesylation is a key determinant of Rheb subcellular localization to endomembranes (Buerger et al., 2006). Palmitoylation, a PTM that strongly affects protein localization, has not been reported for mTOR components; however, bioinformatic analysis (Ren et al., 2008) predicts at least one strong candidate palmitoylation site (C322) in the Sin1 isoform Sin1.5 (NP_001006619; unpublished data). It is of interest to determine the potential role of PTMs in the localization of TOR pathway components.

TORC1 localization is well characterized and consistent in mammalian and yeast cells. It has been shown by many methods that both yeast TORC1 and mTORC1 are mainly at the vacuole/lysosome. TORC2 localization is more ambiguous. mTORC2 is at MAM, and yeast TORC2 is at the plasma membrane. Although this diversity can be attributed to cell type specificity, there might also be multiple pools of TORC2 (or TORC1), each at a different subcellular location. Furthermore, there are at least three different mTORC2 pools, defined by different Sin1 isoforms (Frias et al., 2006). These alternatively spliced Sin1 isoforms differ in containing PH and RBD (Ras binding) domains, and thus present an interesting mechanism for how different mTORC2 complexes might localize to different compartments. Future studies may reveal the molecular basis of differently localized TORC1 or TORC2 subpopulations.

Another important unresolved issue is how the TORCs at different locations are regulated. For example, is all mTORC1 activated by Rheb and Rag at the lysosome and then distributed to other sites, or are there so far undetected amounts of Rheb and Rag at other locations? The following questions have also not been addressed carefully and are interesting points for future investigation. Is there a specific site inside the cell where TOR complexes are assembled? What role do lipids play in TOR localization? What forces/factors mediate TOR translocation in the cell? How does disease influence mTOR localization and vice versa? Does TOR localization play a role in developmental processes? To answer these questions, new technologies that allow sensitive detection of intact mTOR complexes in live cells, such as FRET-FLIM (Ishikawa-Ankerhold et al., 2012), may be necessary. Other methodological or technological developments such as proximity ligation (Blazek et al., 2013) and super-resolution microscopy, respectively, will also enhance our understanding of TOR localization. The ultimate goal is to obtain a comprehensive, dynamic, and spatial model of TOR signaling.