|Home | About | Journals | Submit | Contact Us | Français|
Small molecule inhibitors that selectively target cancer cells and not normal cells would be valuable anti-cancer therapeutics. The mammalian target of rapamycin complex 2 (mTORC2) is emerging as a promising candidate target for such an inhibitor. Recent studies in cancer biology indicate that mTORC2 activity is essential for the transformation and vitality of a number of cancer cell types, but in many normal cells, mTORC2 activity is less essential. These studies are intensifying interest in developing inhibitors that specifically target mTORC2. However, there are many open questions regarding the function and regulation of mTORC2 and its function in both normal and cancer cells. Here, we summarize exciting new research into the biology of mTORC2 signaling and highlight the current state and future prospects for mTOR-targeted therapy.
The mammalian target of rapamycin (mTOR) is a serine/threonine kinase and catalytic subunit of two biochemically distinct complexes called mTORC1 and mTORC2. mTORC1 controls cell autonomous growth in response to nutrient availability and growth factors, whereas mTORC2 is thought to mediate cell proliferation and cell survival. As its name implies, mTOR is the intracellular target of the small molecule inhibitor rapamycin, a naturally occurring compound that, in association with the FKBP12 protein, binds mTOR adjacent to its kinase domain. Importantly, rapamycin-FKBP12 is not a catalytic site inhibitor, but rather allosterically inhibits mTOR and only partially inhibits mTOR function. Rapamycin is a clinically valuable molecule currently used as an immunosuppressant and in treating some cancers.
Discovering that mTOR nucleates two multi-subunit complexes revolutionized mTOR research by expanding a framework for characterizing mechanisms that regulate its activity. The almost simultaneous discovery that AKT signaling directly activates mTORC1 signaling sparked a tidal wave of interest in mTOR biology. The PI3K-AKT pathway is commonly activated in human cancer and active AKT promotes mTORC1 signaling by phosphorylating and inhibiting the TSC1/TSC2 negative regulatory complex (Figure 1). When active, mTORC1 promotes cell growth in part by directly phosphorylating the translational regulators S6K1 and 4E-BP1, by inhibiting autophagy, and by facilitating lipid and mitochondrial biogenesis. mTORC1 is also stimulated by amino acids through an independent and less well-defined mechanism involving the Rag GTPases. Scores of reviews discuss the biology of mTORC1 and its implications in cancer and we refer the readers to some of the most recent (Guertin and Sabatini, 2007; Huang and Manning, 2008; Efeyan and Sabatini, 2009; Kim and Guan, 2009; Wang and Proud, 2009; Laplante and Sabatini, 2009a, b).
Linking mTORC1 regulation to oncogenic PI3K activity provides strong rationale for targeting mTORC1 in cancer and this propelled rapamycin into clinical trials. Unfortunately, the effectiveness of rapamycin as a single agent therapy is stifled in part by strong mTORC1-dependent negative feedback loops that become inactive on mTORC1 inhibition. The best-described negative feedback mechanism occurs through S6K1, which directly phosphorylates insulin receptor substrate 1 causing its mislocalization and degradation (Figure 1). The importance of negative feedback loops to PI3K signaling is evident from clinical trials that find rapamycin increases AKT activation in certain malignancies (O’Reilly et al., 2006). This paradoxically promotes survival and partly explains the general ineffectiveness of rapamycin in the clinic. The platelet-derived growth factor receptor and ERK/MAPK pathways are also targets of S6K1-mediated negative feedback and additional feedback mechanisms likely exist (Zhang et al., 2007; Carracedo et al., 2008; Efeyan and Sabatini, 2009).
Although mTORC1 is in the vanguard of mTOR research, its younger sibling mTORC2 is emerging as a pivotal player in many cancers. Unlike the case for mTORC1, rapamycin does not directly bind and inhibit mTORC2, which initially stymied the identification of its cellular targets. A breakthrough came in 2005 with the discovery that mTORC2 directly phosphorylates AKT on a critical regulatory site required for maximal AKT kinase activity (Sarbassov et al., 2005). This prompted efforts to develop mTOR ATP-competitive inhibitors with the expectation that by targeting both complexes, these inhibitors will outperform rapamycin as anti-cancer drugs. Excitingly, prototype mTOR kinase inhibitors (TKIs) are just now becoming available; however, evaluating this new inhibitor class is in its early days and their efficacy remains to be seen. Several recent studies now provide rationale for developing inhibitors that specifically target mTORC2. Such inhibitors could have a more acceptable therapeutic window and would not perturb the mTORC1-dependent negative feedback loops. Currently, we can only speculate in part because we are just beginning to understand the function and regulation of this complex.
After the discovery of mTOR, researchers struggled explaining why treatments that activate or inhibit the downstream mTOR effectors S6K1 and 4E-BP1 in cells do not affect mTOR in vitro kinase activity. This conundrum cultivated the idea that mTOR functions in a complex with regulatory proteins that are lost during purification, prompting a redesign of the mTOR purification scheme and subsequent discovery that mTOR binds Raptor and mLST8 (KOG1 and LST8, respectively, in yeast) in a rapamycin-sensitive complex now called mTORC1 (Hara et al., 2002; Kim et al., 2002; Loewith et al., 2002). Unexpectedly, these seminal studies found that mTOR additionally interacts with a protein called Rictor (AVO3 in yeast) independently of Raptor in a distinct mTOR complex (now called mTORC2) that also contains mLST8 (Figure 2) (Jacinto et al., 2004; Sarbassov et al., 2004). As rapamcyin does not directly bind mTORC2, it has been dubbed the ‘rapamycin-insensitive’ complex. Earlier studies in Saccharomyces cerevisiae had identified rapamycin-insensitive functions of TOR, which are now known to be mediated by yeast TORC2 (reviewed in Wullschleger et al., 2006). However, more recent studies in mammalian cells reveal that prolonged exposure to rapamycin inhibits mTORC2 activity by disrupting the assembly of new mTORC2 complexes (Sarbassov et al., 2006). All known mTORC2 functions require Rictor, but because Rictor contains few conserved regions and lacks any obvious functional domains, its exact mechanism of regulating mTOR activity is unclear.
mTORC2 contains at least three additional subunits; mSIN1 (AVO1 in yeast), Protor, and Deptor (Figure 2) (Loewith et al., 2002; Jacinto et al., 2004, 2006; Sarbassov et al., 2004; Frias et al., 2006; Yang et al., 2006a; Pearce et al., 2007; Thedieck et al., 2007; Woo et al., 2007; Peterson et al., 2009). mTORC2 integrity requires both Rictor and mSIN1. However, most mSIN1 orthologs contain a Ras-binding domain and a pleckstrin homology (PH)-like domain, suggesting that mSIN1 may have functions in addition to maintaining structural stability (Schroder et al., 2007). Interestingly, alternative splicing generates five mSIN1 isoforms and at least three of them, mSIN1.1, mSIN1.2, and mSIN1.5 (Sin1, Sin1β, and Sin1α in Schroder et al.) assemble independently with mTOR and Rictor possibly defining three distinct mTORC2 complexes with unique intracellular functions (Frias et al., 2006). Protor exists in two isoforms, Protor-1 and Protor-2, which share homology in their amino-terminal halves (Pearce et al., 2007; Thedieck et al., 2007; Woo et al., 2007). Immunopurification studies indicate that Protor tightly binds to Rictor; however, neither mTORC2 assembly nor activity requires Protor and the protein structure lacks any obvious domains to provide clues to its function.
mLST8 and Deptor are the only known mTOR-interacting proteins common to both mTORC1 and mTORC2 (Jacinto et al., 2004; Sarbassov et al., 2004). mLST8, which is almost entirely composed of WD repeats, tightly binds mTOR near the kinase domain and is required for mTORC2 integrity (Kim et al., 2003; Guertin et al., 2006). Deptor (also known as DEPDC6) also binds mTOR near the kinase domain and contains N-terminal DEP (disheveled, egl-10, pleckstrin) domains and a C-terminal PDZ (postsynaptic density 95, discs large, zonula occludens-1) domain (Peterson et al., 2009). Experiments overexpressing recombinant Deptor or depleting it by RNAi indicate that it negatively regulates the kinase activity of both mTORC1 and mTORC2. Interestingly, endogenous Deptor is overexpressed in some multiple myeloma cells and paradoxically this promotes their survival. This is explained by the fact that overexpressing Deptor partially inhibits both mTORC1 and mTORC2, but by inhibiting mTORC1, it relieves strong negative feedback loops to PI3K. Loss of feedback inhibition overrides the partial inhibitory effect of Deptor overexpression on mTORC2 and promotes cell survival. Interestingly, Deptor is found only in vertebrates, suggesting that its regulatory function is a recent adaptation. Defining the in vivo significance of the Deptor–mTOR interaction should be interesting.
As the dynamics of mTORC2 assembly and its 3D structure are unknown, we understand little about the molecular interactions within the complex. Gel filtration and co-immunoprecipitation experiments suggest mTORC2 functions as an oligomer, but how this influences its activity is unclear (Wullschleger et al., 2005; Zhang et al., 2006). An outstanding question is whether any of the mTORC2 subunits have functions independent of their functions in the complex. Interestingly, fission yeast and mammalian SIN1 are implicated in JNK signaling and the stress response pathway (Schroder et al., 2005). Are there additional mTORC2 subunits to be discovered? Biochemical-based searching for mTOR complex-interacting proteins has been rigorous and the most stable-interacting subunits may have been discovered. However, the extent to which unidentified regulatory factors transiently interact with the complex is unknown.
Defining mTORC2’s cellular functions has been more challenging compared with mTORC1 because of its insensitivity to acute rapamycin treatment. Studies in model organisms such as S. cerevisiae, Dictyostelium discoideum, and Trypanosome brucei suggest that TORC2 regulates cytoskeleton dynamics (Lee et al., 1999, 2005; Audhya et al., 2004; Fadri et al., 2005; Kamada et al., 2005; Barquilla et al., 2008; Kamimura et al., 2008). In mammalian cells, transiently depleting mTOR, Rictor, or mLST8 by RNAi causes actin organization defects, suggesting that this function might be conserved (Jacinto et al., 2004; Sarbassov et al., 2004). However, challenging the case for mTORC2 being a cytoskeletal regulator is the surprising discovery that MEFs deleted for rictor, mSIN1, or mLST8 have no obvious cytoskeletal defects (Guertin et al., 2006; Jacinto et al., 2006; Shiota et al., 2006). Thus, the function of mammalian TORC2 in cytoskeletal regulation needs to be clarified.
The discovery that mTORC2 directly phosphorylates AKT provided a leap forward in understanding mTORC2 function (Sarbassov et al., 2005). AKT (also known as PKB) is the major downstream effector in the insulin/PI3K pathway and regulates cell survival, proliferation, growth, metabolism, angiogenesis, and glucose uptake (reviewed in Manning and Cantley, 2007). The mammalian genome contains three AKT isoforms (AKT1/PKBα, AKT2/PKBα, and AKT3/PKBα) that have both overlapping and distinct functions. AKT is recruited to the plasma membrane after PI3K activation through association of its amino-terminal PH domain with PIP3. At the membrane, two phosphorylation events activate AKT: one occurring in the T-loop by the PDK1 kinase (T308 in AKT1) and one occurring in a C-terminal HM site (S473 in AKT1) by a long sought after, but earlier unknown ‘PDK2’ kinase. In both Drosophila and mammalian cultured cells, silencing Rictor, mSIN1, or mTOR expression, but not Raptor, reduces HM phosphorylation of AKT, and in an in vitro kinase assay, only mTORC2 and not mTORC1 phosphorylates AKTS473 (Hresko and Mueckler, 2005; Sarbassov et al., 2005; Frias et al., 2006; Yang et al., 2006a). Importantly, genetic and pharmacological studies confirm that mTORC2 is the major AKTS473 kinase, although it is still debated as to whether other HM kinases exist (Guertin et al., 2006; Jacinto et al., 2006; Shiota et al., 2006; Yang et al., 2006a; Bozulic and Hemmings, 2009; Feldman et al., 2009; Garcia-Martinez et al., 2009; Thoreen et al., 2009; Yu et al., 2009).
In vitro, maximal AKT activity requires phosphorylation at both T308 in the T-loop and S473 in the HM (Alessi et al., 1996). These phosphorylation events can occur independently, which is evident in rictor, mLST8, and mSIN1 knockout MEFs in which AktT308 phosphorylation is intact despite the absence of phosphorylation at AKTS473 (Guertin et al., 2006; Jacinto et al., 2006; Shiota et al., 2006). Moreover, in PDK1-null cells that lack AKTT308 phosphorylation, S473 phosphorylation still occurs (Alessi et al., 1996; McManus et al., 2004). However, depleting Rictor by RNAi simultaneously decreases AKT phosphorylation at both T308 and S473, revealing an inconsistency between knockout and knockdown cells (Hresko and Mueckler, 2005; Sarbassov et al., 2005). One possibility explaining the RNAi results is that PDK1-dependent T308 phosphorylation requires earlier HM phosphorylation at S473. That HM phosphorylation of AKT stabilizes phosphorylation at T308 is clear from structural studies and is the type of co-regulation observed with S6K1, in which HM phosphorylation by mTORC1 is a prerequisite for T-loop phosphorylation by PDK1 (Yang et al., 2002a, 2002b). Perhaps in the knockout cells PDK1-dependent T308 phosphorylation is up-regulated by an unknown compensatory mechanism. Solving this puzzle is important because studies of downstream AKT signaling in mTORC2-deficient MEFs indicate that AKTT308 phosphorylation alone empowers AKT with enough activity to phosphorylate many of its substrates (including TSC2, GSK3β, and partially FoxO1/3) (Guertin et al., 2006; Jacinto et al., 2006; Hietakangas and Cohen, 2007). Thus, targeting mTORC2 in cancer cells might be most effective when both T308 and S473 phosphorylation are inhibited.
Are there additional mTORC2 substrates? Both AKT and S6K belong to the protein kinase A/protein kinase G/protein kinase C (AGC) family of protein kinases, which additionally includes serum- and glucocorticoid-induced protein kinase (SGK) and ribosomal S6K (reviewed in Jacinto and Lorberg, 2008). This subgroup is defined by structurally similar kinase domains requiring PDK1-dependent phosphorylation in the T-loop and C-terminal phosphorylation in an HM site for full activation. The fact that mTOR phosphorylates the HM site in both AKT and S6K suggests that mTOR may have evolved to cooperate with PDK1 to activate AGC kinases. Exploring this possibility led to the recent discovery that mTORC2 additionally phosphorylates the S422 HM site of SGK1 (Garcia-Martinez and Alessi, 2008).
Mammalian SGK exists in three isoforms (SGK1, SGK2, and SGK3) that have both distinct substrates and substrates that overlap with AKT (reviewed in Tessier and Woodgett, 2006). The best-described function of SGK is in regulating sodium transport, but there is growing evidence that the SGKs contribute to oncogenesis (Tessier and Woodgett, 2006; Vasudevan et al., 2009). In mTORC2-deficient MEFs, SGK1 lacks kinase activity and HM phosphorylation (Garcia-Martinez and Alessi, 2008). Moreover, purified mTORC2 phosphorylates SGK1S422 in vitro, indicating that SGK1 is a bona fide mTORC2 substrate. Importantly, phosphorylation of the SGK1 substrate NDRG1 is also absent in rictor, mLST8, and mSIN1-null MEFs, indicating that SGK1 activity is severely impaired. This is in contrast to AKT, which as discussed above is still active in the same knockout lines. Similar to S6K1, HM motif phosphorylation of SGK1 is a prerequisite for PDK1-dependent phosphorylation in its T-loop (T256)—explaining why SGK1 activity is dead in these cells. Interestingly, many cancer cell lines with PI3K-activating mutations preferentially rely on SGK3 signaling rather than AKT for their oncogenic properties (Tessier and Woodgett, 2006; Vasudevan et al., 2009). Clarifying the contributions of AKT and SGK in cancer cell survival will be important in predicting the effectiveness of future mTORC2 inhibitors.
Early studies of Rictor indicate that mTORC2 regulates PKCα phosphorylation and stability (Sarbassov et al., 2004; Guertin et al., 2006). PKCα along with PKCβ and PKCγ constitute the conventional PKCs, which are calcium- and phospholipid-dependent kinases in the AGC family implicated in tumor progression (Martiny-Baron and Fabbro, 2007). The stability of all three cPKCs requires mTORC2, although evidence that mTORC2 phosphorylates them directly is lacking (Ikenoue et al., 2008). In contrast ribosomal S6K phosphorylation is not blocked by an mTOR kinase inhibitor, indicating that not all AGC kinases are mTOR targets (Garcia-Martinez et al., 2009).
It is speculated that mTORC2 additionally phosphorylates a second less-discussed motif in AKT called the turn motif (TM) (Facchinetti et al., 2008). In AKT and PKC, the TM is located between the kinase domain and HM and is required for protein stability. In mTORC2-deficient MEFs, TM phosphorylation of AKT and PKC is decreased (Facchinetti et al., 2008; Ikenoue et al., 2008). However, unlike the HM site, TM phosphorylation is insensitive to growth factors and while one study finds that immunopurified mTORC2 phosphorylates the AKT TM in vitro, another study failed to detect TM phosphorylation in an in vitro kinase assay in which mTORC2 robustly phosphorylates the HM site (Facchinetti et al., 2008; Ikenoue et al., 2008; Garcia-Martinez et al., 2009). Clearly, the mechanisms by which mTORC2 regulates TM and HM phosphorylation differ and could involve an undefined kinase or phosphatase.
Are there additional mTORC2 substrates? A recent genetic study in Drosophila suggests that dTORC2 targets the AGC kinase NDR1 in controlling dendritic tiling in the sensory neuron, which should prompt investigation of mTORC2-dependent NDR1 regulation in mammalian cells (Koike-Kumagai et al., 2009). Notably, mTOR phosphorylation sites in 4E-EBP1 show no structural similarity to the HM site, suggesting that the list of mTORC2 substrates could extend beyond the AGC family. What determines mTORC2 substrate specificity is also an open question. One possibility is that certain mTORC2 subunits are substrate scaffolds. Precedent for such a model exists from studies of mTORC1 that show Raptor facilitates interaction between mTOR and both 4E-BP1 and S6K1 (Schalm and Blenis, 2002; Nojima et al., 2003; Schalm et al., 2003). Co-localization could also facilitate interactions between mTORC2 and its substrates and as discussed below evidence that mTORC2 associates with cell membranes is accumulating.
Little is known about the upstream mechanisms regulating mTORC2. When purified from insulin or IGF1-stimulated cells, mTORC2 in vitro kinase activity is elevated, suggesting that PI3K signaling activates mTORC2 (Sarbassov et al., 2004, 2005; Frias et al., 2006; Jacinto et al., 2006; Yang et al., 2006a). One obvious possibility is that phosphorylation regulates the complex, and not surprisingly, several mTORC2 subunits are heavily phosphorylated (Sarbassov et al., 2004; Yang et al., 2006a; Akcakanat et al., 2007; Hayashi et al., 2007; Acosta-Jaquez et al., 2009; Dibble et al., 2009; Julien et al., 2009; Treins et al., 2009). For example, Rictor is highly phosphorylated on serine and threonine residues clustering in a carboxy-terminal region conserved only in vertebrate Rictor orthologs (Dibble et al., 2009; Julien et al., 2009; Treins et al., 2009). Interestingly, phosphorylation at one site, located at T1135, is sensitive to rapamycin, growth factors, and amino acids, suggesting crosstalk between the mTORC1 pathway and mTORC2. In support, S6K1 directly phosphorylates RictorT1135 in an in vitro kinase assay and in S6K-null MEFs, RictorT1135 phosphorylation is absent. Could this be part of another negative feedback loop? Possibly, because expressing a recombinant rictorT1135A mutant in rictor-null MEFs restores AKTS473 phosphorylation to a higher degree than expressing recombinant wild-type rictor (Dibble et al., 2009). However, mTORC2 stability and in vitro kinase activity, as well as downstream SGK activity, is unaffected by the T1135A mutation, and the only function ascribed to phosphorylation at this site is in promoting interaction with 14–3–3 proteins for reasons currently unknown (Dibble et al., 2009; Treins et al., 2009).
Interestingly, there is an emerging function for the TSC1–TSC2 complex in positively regulating mTORC2 (Huang et al., 2008, 2009). TSC1/2 functions as a GAP for Rheb (Ras homolog enriched in brain), a small GTPase that directly activates mTORC1, and mutation in either tsc1 or tsc2 causes tuberous sclerosis (Figure 1) (reviewed in Huang and Manning, 2008). Recent work finds that TSC1/2 inactivation not only elevates mTORC1 signaling, but suppresses growth factor-stimulated mTORC2 activity (Huang et al., 2008). This could be explained by the fact that TSC1/2 loss promotes feedback inhibition of PI3K signaling. However, knocking down Raptor in TSC1/2-deficient cells, which prevents feedback inhibition, does not restore mTORC2 activity, arguing that TSC1/2 positively regulates mTORC2 in a manner independent of its function in negatively regulating mTORC1. In support, overexpressing or silencing Rheb has no effect on mTORC2 kinase activity (Yang et al., 2006b; Huang et al., 2008; Sancak et al., 2008). The discovery that S6K1 and TSC1/2 negatively and positively regulate mTORC2, respectively, indicates that considerable cross-communication exists between the mTORC1 branch and mTORC2, and similar to the negative feedback loops, this could be part of a mechanism to restrain AKT signaling when mTORC1 is active. Importantly, these findings also provide an additional explanation for the benign nature of tuberous sclerosis.
Growth factors could also potentiate mTORC2 signaling by regulating its intracellular localization. In PTEN-null prostate epithelial cells, mTORC2 strongly associates with the plasma membrane supporting this notion (Guertin et al., 2009). PH-domains facilitate PDK1 and AKT co-localization at the cell membrane, and interestingly, mSIN1 contains a carboxy-terminal domain structurally similar to a PH domain (Schroder et al., 2007). Elegant studies in S. cerevisiae indicate that TORC2 plasma membrane association is essential for viability and is mediated through the PH-like domain in the mSIN1 ortholog Avo1 (Sturgill et al., 2008; Berchtold and Walther, 2009). Remarkably, replacing the Avo1 PH-like domain with a CaaX-box membrane anchor completely rescues viability (Berchtold and Walther, 2009). Whether mTORC2 localizes to membranes through mSIN1 remain to be seen. Studies in T. brucei additionally suggest that the endoplasmic reticulum and mitochondria membranes are potential mTORC2 localization sites (Barquilla et al., 2008). Notably, SGK1 lacks a PH domain arguing that mTORC2 may have plasma membrane-independent functions as well. Interestingly, the mSIN1.5 isoform lacks the PH-like domain and defines a growth factor insensitive version of mTORC2 (Frias et al., 2006). Perhaps mSIN1.5 mediates plasma membrane-independent mTORC2 signaling.
Another mechanism by which cells could regulate mTORC2 activity is by controlling the levels of essential regulatory subunits. For example, a variety of cancer cells exhibit increased Rictor levels correlating with increased AKTS473 phosphorylation (Masri et al., 2007; Guertin et al., 2009; Gulhati et al., 2009). Moreover, overexpressing recombinant Rictor enhances the oncogenic traits of certain glioma cell lines (Masri et al., 2007). Changes in the expression level of Rictor could occur transcriptionally or translationally, but whether such changes represent an intrinsic regulatory mechanism or result from oncogenic mutations is not known.
Little is known about the physiological processes requiring mTORC2 activity. In the developing mouse embryo, globally deleting mTOR or raptor results in early lethality, whereas deleting rictor or mSIN1 results in lethality around embryonic day 10.5 (Gangloff et al., 2004; Murakami et al., 2004; Guertin et al., 2006; Jacinto et al., 2006; Shiota et al., 2006; Yang et al., 2006a). Surprisingly, globally knocking out mLST8 (which tightly associates with both mTOR complexes) phenocopies rictor-null embryos, indicating that in mammalian development, mLST8 is dispensable for mTORC1 functions, but essential for mTORC2 (Guertin et al., 2006; Shiota et al., 2006). Although the exact reason why mTORC2-deficient embryos die is unclear, they seem incapable of expanding and remodeling the fetal vasculature (Guertin et al., 2006; Shiota et al., 2006).
As mTORC2 activity is essential for development, conditional mouse knockout models are being generated to define the tissue-specific functions of mTORC2 in adult mice. Skeletal muscle is the first tissue in which mTORC2 activity has been conditionally ablated by Cre-mediated deletion of rictor ‘floxed’ alleles (Bentzinger et al., 2008; Kumar et al., 2008). Insulin stimulates glucose uptake and conversion to glycogen in skeletal muscle, which serves as the major mechanism of clearing glucose after a meal. As expected, deleting rictor in muscle greatly reduces insulin-stimulated AktS473 phosphorylation, whereas surprisingly, AktT308 phosphorylation is maintained and even slightly increased perhaps by the same compensatory mechanism operating in rictor knockout MEFs (Kumar et al., 2008). However, rictor-deficient skeletal muscle surprisingly exhibits only mild defects in glucose transport and glycogen synthase activity, and overall, these mice have normal glucose tolerance. In stark contrast, deleting raptor in skeletal muscle causes severe muscle dystrophy and premature death (Bentzinger et al., 2008).
As in muscle, adipose-specific rictor-deletion reduces AKTS473 phosphorylation and not T308 (Cybulski et al., 2009). Adipose-specific rictor knockout mice are slightly larger than wild-type mice due to increases in the lean tissue mass of individual organs (including heart, kidneys, spleen, pancreas, and bone), but curiously not in fat mass. Moreover, feeding mice a high-fat diet exacerbates the mass increase of the non-adipose tissue, disproportionately increasing pancreas, and β-cell mass. Adipose-specific rictor knockout mice are mildly insulin resistant, but more glucose tolerant than wild-type mice because they have elevated levels of insulin (probably because of the increase in β-cell mass) and IGF1. The increase in insulin/IGF1 levels could be compensating for the mild insulin resistance and driving the growth of lean tissues.
One potential downstream physiological process under mTORC2 control inferred from the study of lower eukaryotes is regulation of lipid metabolism. In S. cerevisiae, components of TORC2 are implicated in the sphingolipid biosynthesis pathway (Audhya et al., 2004; Tabuchi et al., 2006; Aronova et al., 2008). TORC2 seems to function early in the pathway by regulating the synthesis of cerimides, which are precursors to more complex sphingolipids (Aronova et al., 2008). In Caenorhabditis elegans, a screen for genes that regulate lipid metabolism identified several loss-of-function rictor mutants that show excess fat stores, indicating that a function for TORC2 in lipid regulation may be conserved (Jones et al., 2009; Soukas et al., 2009). In worms, TORC2-dependent lipid metabolism depends mostly on SGK1 because sgk1 mutants more closely resemble rictor mutants. Whether lipid metabolism is a function of mammalian Rictor/mTORC2 is not apparent from genetic knockout studies described above, but the investigation into mTORC2 signaling in vivo is just beginning.
The discovery that AKT activates mTORC1 by phosphorylating and inhibiting TSC2 provided rationale for mTOR-targeted therapy, propelling rapamycin into clinical trials as the first mTOR-based cancer therapeutic (reviewed in Guertin and Sabatini, 2007). Frequently occurring mutations that activate the PI3K-AKT pathway in cancer include PIK3CA-activating mutations and gene amplification, loss of the PTEN tumor suppressor, AKT mutation and amplification, and receptor tyrosine kinase amplification (Liu et al., 2009). Consequently, there is considerable investment in developing PI3K pathway inhibitors.
Discovering that mTORC2 directly phosphorylates AKT led to speculation that mTORC2-specific inhibitors might also be valuable cancer drugs. In addition to blocking AKT phosphorylation, an mTORC2 inhibitor has the advantage of not disrupting mTORC1-dependent negative feedback loops. As mTORC2-specific inhibitors do not yet exist, this question was addressed by genetically ablating mTORC2 activity (by deleting rictor) in a PTEN-deletion-dependent mouse model of prostate cancer (Guertin et al., 2009). Deleting PTEN specifically in the prostate epithelium induces invasive prostate cancer; however, deleting rictor in combination with PTEN blocks tumor development. Importantly, rictor deletion alone in the prostate epithelium has no deleterious effects, indicating that mTORC2 activity is only required for PTEN-deletion-induced cancer and not for normal prostate function. This is reminiscent of genetic studies in Drosophila, in which drictor is non-essential for fly development, but is required for phenotypes induced by PTEN deletion (Hietakangas and Cohen, 2007). Whether mTORC2 inactivation has deleterious consequences in other cell types remains to be seen; however, the selective requirement for mTORC2 activity in PTEN-deficient prostate cancer cells, but not in normal prostate cells, provides compelling rationale for developing inhibitors specifically targeting mTORC2. RNAi-mediated Rictor silencing is also toxic to several human cancer cell lines with elevated AKT signaling further predicting the potential value of mTORC2 inhibitors (Masri et al., 2007; Hietakangas and Cohen, 2008; Guertin et al., 2009; Gulhati et al., 2009). Currently, three main classes of small molecules capable of inhibiting mTORC2 exist, but none are specific. They include (1) the recently developed ATP-competitive TKIs, which target both mTORC1 and mTORC2; (2) the dual specificity inhibitors, which target PI3K in addition to both mTORC1 and mTORC2; and (3) rapamycin, which can inhibit mTORC2 assembly in addition to directly inhibiting mTORC1.
ATP-competitive inhibitors of mTOR are the newest generation of mTOR inhibitors to emerge (Table 1) (reviewed in Guertin and Sabatini, 2009; Shor et al., 2009). We will refer to this inhibitor class collectively as TKIs as suggested by Shor et al. (2009). It is anticipated that TKIs will have broad application because they target the catalytic site of both mTORC1 and mTORC2. Several groups independently developed prototype TKIs, including Torin1, PP242, PP30, Ku-0063794, WAY-600, WAY-687, WAY-354, and AZD8055 (Feldman et al., 2009; Garcia-Martinez et al., 2009; Thoreen et al., 2009; Yu et al., 2009; Chresta et al., 2010). Common to all is their remarkable selectivity toward mTOR, exhibiting IC50 values in the low nanomolar range (as determined by S6K1T389 and AKTS473 phosphorylation, respectively), inhibiting related kinases such as PI3K only at much higher concentrations. In an exciting recent study, one group finds that PP242, but not rapamcyin, induces death of both mouse and human leukemia cells harboring the Philadelphia chromosome translocation (Janes et al., 2010).
Compared with rapamycin, TKIs more dramatically inhibit protein synthesis and induce G1 cell cycle arrest in several cancer cell lines (Feldman et al., 2009; Garcia-Martinez et al., 2009; Shor et al., 2009; Thoreen et al., 2009; Yu et al., 2009). At first glance, one might expect that this is because TKIs target both mTORC1 and mTORC2. Surprisingly, this is not necessarily the case and may reflect more profound inhibition of mTORC1. For example, TKIs more completely inhibit 4E-BP1 phosphorylation by mTORC1 compared with rapamcyin (Feldman et al., 2009; Thoreen et al., 2009). This explains why rapamycin has little impact on capdependent protein translation in many cancer cells and at least partly explains the marginal effectiveness of rapamcyin as a single agent therapy.
One caveat concerning the application of TKIs in cancer therapy is their potential toxicity. However, one study finds that genetically ablating mTOR in prostate epithelial cells has no deleterious effect, whereas another study finds that TKIs perform well in a short-term xenograft study and are well tolerated (Nardella et al., 2009; Yu et al., 2009). A second caveat is that any inhibitor of mTORC1 will disengage the negative feedback loops, which could be problematic even in the context of mTORC2 inhibition. Peterson et al. (2009) provide proof of principal by showing that low concentrations of Torin1 (50 nm) causes loss of feedback inhibition and AKT activation—presumably because uninhibited mTORC2 complexes become hyperactive—and only at high concentrations of Torin1 (250 nm) is mTORC2 more completely inhibited. Finally, mTORC1 inhibition activates autophagy, which could promote cancer cell survival (reviewed in White and DiPaola, 2009). In support, TKIs are far more formidable activators of autophagy in mammalian cells compared with rapamycin (Thoreen et al., 2009; Chresta et al., 2010).
A related class of inhibitors is the dual specificity inhibitors, including wortmanin, LY294002, PI-103, BGT226, XL765, and NVP-BEZ235, which target the structurally related kinase domains of both PI3K and mTOR (reviewed in Liu et al., 2009). By additionally targeting PI3K, loss of feedback inhibition might be less problematic with this inhibitor class. Similar to the TKIs, the dual specificity inhibitors are more effective than rapamycin at inhibiting cancer cell proliferation. Moreover, in mouse cancer models, they are particularly effective against oncogenic PI3K-driven tumors, but are only effective against KRAS-driven tumors if combined with an MEK inhibitor (Engelman et al., 2008). Clearly, for all PI3K pathway inhibitors, defining the molecular pathology of specific cancer cells is critical to finding patients with the best chance of responding.
The mTOR inhibitors most developed for cancer therapy are based on the chemical structure of rapamycin (reviewed in Guertin and Sabatini, 2009; Yuan et al., 2009). Most studies have focused on rapamycin’s acute inhibitory effect on mTORC1. However, in a subset of cancer cell lines, prolonged exposure to rapamycin decreases AktS473 phosphorylation in addition to phosphorylation of S6K1T389 and this is attributed to its ability to bind free mTOR and block the assembly of new mTORC2 complexes (Sarbassov et al., 2006). Notably, free rapamycin (not bound to FKBP12) can also inhibit mTORC2 at much higher concentrations (Shor et al., 2008). Interestingly, in the protozoan parasite T. brucei, rapamycin exclusively inhibits TORC2 assembly, whereas TORC1 is insensitive to the drug (Barquilla et al., 2008). Is it possible that some clinical successes with rapamycin result from inhibiting both complexes? This is shown in a recent study comparing rapamycin-sensitive and rapamycin-resistant colorectal cancer cells (Gulhati et al., 2009). In rapamycin-sensitive cells, both S6K1T389 and AKTS473 phosphorylation decrease on rapamycin treatment; in contrast, rapamycin only decreases S6K1T389 phosphorylation in the resistant cells. Rapamycin is particularly effective against Kaposi’s sarcoma (a highly vascular tumor) and against some hematopoietic cancers, and interestingly, AKTS473 phosphorylation is highly sensitive to prolonged rapamycin exposure in endothelial and hematopoietic cells (Recher et al., 2005; Witzig et al., 2005; Phung et al., 2006; Sarbassov et al., 2006; Witzig and Kaufmann, 2006; Zeng et al., 2007; Gulhati et al., 2009).
Why prolonged rapamycin exposure reduces AktS473 phosphorylation in only a subset of cancer cells is under investigation; however, this does not seem to reflect differences in mTORC2 sensitivity to the drug. A recent study finds that one mTOR auto-phosphorylation site located at S2481 specifically associates with mTOR molecules assembled into mTORC2 (Copp et al., 2009). In several cancer cells, exposure to rapamycin for prolonged periods reduces mTORS2481 phosphorylation by disrupting mTORC2 assembly, and importantly, this holds true even in cancer cells in which AktS473 phosphorylation is unchanged or increases after rapamycin treatment. This argues that rapamcyin universally blocks mTORC2 assembly. Why then does prolonged rapamycin exposure not universally inhibit AktS473 phosphorylation? The simplest explanation is that in some cells uninhibited mTORC2 complexes are sufficient to phosphorylate AKT or become hyperactive due to loss of feedback inhibition, although an unknown compensatory kinase has not been ruled out. Regardless, these studies clearly indicate that rapamycin partially inhibits both mTOR complexes, and going forward, it is important that we identify biomarkers predictive of a clinical response to the drug.
With the emergence of mTORC2 as a critical player in cancer, we envision a future generation of mTOR inhibitors specifically targeting mTORC2. How might a theoretical mTORC2 inhibitor function? One possible mechanism is by disrupting important protein–protein interactions required for mTORC2 integrity (Figure 3a). For example, the integrity of mTORC2 requires Rictor, mSIN1, and mLST8, so a small molecule that blocks their assembly or breaks apart their association with mTOR would destabilize the complex (Frias et al., 2006; Guertin et al., 2006; Jacinto et al., 2006; Yang et al., 2006a). A major challenge to rationally designing such a molecule is the lack of a 3D structure. Finding a small molecule that targets the substrate-binding interface is another strategy (Figure 3b). A third possibility is finding a molecule that prohibits mTORC2 localization (Figure 3c). For example, if the mSIN1 PH-like domain localizes mTORC2 to membranes, one could imagine a small molecule inhibitor that targets the lipid-binding interface. However, such a molecule may not inhibit mTORC2 functions occurring elsewhere in the cell. Targeting an upstream-activating kinase is also a possibility; however, critical regulatory enzymes and motifs have not yet been identified (Figure 3d).
The development of TKIs marks the beginning of an exciting new phase in mTOR-targeted therapy. As these compounds target both mTORC1 and mTORC2, and greatly inhibit protein translation, we anticipate that they will have more potent anti-cancer activity than rapamycin. Although we are just beginning to evaluate the potential of TKIs, we are already looking forward to future generation inhibitors that specifically target mTORC2. Several recent reports suggest that mTORC2 inhibitors may be valuable cancer therapeutics. mTORC2 inhibitors might have an advantage over mTOR catalytic site inhibitors by not perturbing mTORC1-dependent negative feedback loops or activating autophagy. Moreover, deleting rictor in muscle, adipose, and prostate has relatively minor consequences in comparison with the severe effects from deleting raptor in muscle, possibly indicating that an mTORC2 inhibitor would be well tolerated. However, many questions regarding the 3D structure, regulation, and function of mTORC2 remain unanswered. Importantly, we have not yet addressed the function of mTORC2 in more advanced cancers. mTORC2 activity is required for tumor initiation induced by PTEN deletion in the mouse prostate, and for the oncogenic characteristics (that is proliferation, growth in soft agar, ability to form tumor xenografts) of many human cancer cell lines. But what is the function of mTORC2 in later, more therapeutically relevant stages of cancer or in the tumor microenvironment? The growing appreciation of mTOR’s vital function in tumorigenesis is driving intense interest in targeting mTOR in cancer and we are optimistic that current and future generation mTOR inhibitors, particularly those specifically targeting mTORC2, will have broad impact in cancer therapy.
DAG is supported by grants from the National Institutes of Health (R00 CA129613), the Charles Hood Foundation, and the UMass Center for Clinical and Translational Sciences.
Conflict of interest
The authors declare no conflict of interest.