The TOR (target of rapamycin), an atypical protein kinase, is evolutionarily conserved from yeast to man. Pharmacological studies using rapamycin to inhibit TOR and yeast genetic studies have provided key insights on the function of TOR in growth regulation. One of the first bona fide cellular targets of TOR was the mammalian protein kinase p70 S6K (p70 S6 kinase), a member of a family of kinases called AGC (protein kinase A/protein kinase G/protein kinase C-family) kinases, which include PKA (cAMP-dependent protein kinase A), PKG (cGMP-dependent kinase) and PKC (protein kinase C). AGC kinases are also highly conserved and play a myriad of roles in cellular growth, proliferation and survival. The AGC kinases are regulated by a common scheme that involves phosphorylation of the kinase activation loop by PDK1 (phosphoinositide-dependent kinase 1), and phosphorylation at one or more sites at the C-terminal tail. The identification of two distinct TOR protein complexes, TORC1 (TOR complex 1) and TORC2, with different sensitivities to rapamycin, revealed that TOR, as part of either complex, can mediate phosphorylation at the C-terminal tail for optimal activation of a number of AGC kinases. Together, these studies elucidated that a fundamental function of TOR conserved throughout evolution may be to balance growth versus survival signals by regulating AGC kinases in response to nutrients and environmental conditions. This present review highlights this emerging function of TOR that is conserved from budding and fission yeast to mammals.
Nutrients fuel the growth and development of living organisms. The eukaryotic machinery that allows unicellular and multicellular organisms to utilize nutrients efficiently involves the highly conserved protein TOR (target of rapamycin). The presence of abundant nutrients allows cells to undergo synthesis of proteins and other macromolecules that are critical for increase in cellular mass and number. In unicellular eukaryotes, such as yeast, the cellular responses triggered by nutrients are coupled to TOR signalling. Under nutrient limitation or other adverse environmental conditions, these responses are down-regulated by TOR to favour energy-saving mechanisms and cell survival. Nutrient starvation, which can be mimicked by the inhibition of TOR, down-regulates translation initiation, up-regulates genes involved in starvation, leads to accumulation of storage carbohydrates and induces autophagy. In multicellular organisms multiple inputs, in addition to nutrients, converge to regulate cellular responses to growth signals in order to accommodate the needs of different tissues and organs of the body. mTOR (mammalian TOR) integrates these multiple growth cues to orchestrate cell growth. Recently, mTOR has also been linked to the regulation of cell survival in response to stress by mechanisms that remain to be elucidated [40,222].
Rapamycin, a bacterial macrolide that potently inhibits TOR, was instrumental in the initial identification of downstream effectors of the TOR pathway that are critical for growth regulation both in yeast and mammals. S6K (S6 kinase), a protein kinase belonging to the AGC (protein kinase A/protein kinase G/protein kinase C) kinase family and a critical regulator of protein synthesis, was first to be identified as a rapamycin-sensitive target of mTOR [84,149]. Since structural and biochemical studies have hinted at a common regulatory mode for regulation of the AGC family kinases, the identification of S6K as an mTOR target also suggested that TOR/mTOR may also regulate other members of this family. Genetic and biochemical studies in budding (Saccharomyces cerevisiae) and fission (Schizosaccharomyces pombe) yeast confirmed that a conserved signalling module consisting of TOR–PDK1 (phosphoinositide-dependent kinase 1)–AGC kinase exists. More importantly, the discovery that TOR/mTOR exists as a part of distinct protein complexes with different rapamycin sensitivities catalysed a new line of inquiry on this conserved signalling mechanism.
TOR STRUCTURE AND COMPLEX COMPONENTS
TOR belongs to a family of protein kinases, termed the PIKKs [PI3K (phosphoinositide 3-kinase)-related kinases], a subgroup of the atypical protein kinases . This family comprises a small subset of protein kinases that do not share sequence homology with conventional protein kinases, but has been demonstrated experimentally to possess protein kinase activity [2,3]. TOR consists of conserved domains that are believed to mediate protein–protein interactions. Its N-terminus is made up of HEAT repeats  [Huntingtin, Elongation Factor 3, A subunit of PP2A (protein phosphatase 2A), TOR1], which adopt a large extended superhelical structure . Electron microscopy and structural analysis revealed that these HEAT repeats form a curved tubular-shaped domain that associates with the C-terminal WD40 repeat domain of KOG1 (kontroller of growth 1), a TORC1 interactor . The HEAT repeats are followed by the FAT [FRAP (FKBP12-rapamycin-associated protein)/TOR, ATM (ataxia-telangiectasia), TRRAP (transactivation/tranformation-domain-associated protein)] region, a domain that is conserved among PIKK proteins. The FAT domain occurs in tandem with the FATC domain at the tail end. Together, the FAT and FATC (FRAP, ATM, TRRAP C-terminal) domains of mTOR are important for its catalytic activity, as mutations in these domains abolish mTOR autophosphorylation activity as well as mTOR-dependent phosphorylation of 4E-BP (eukaryotic initiation factor 4E-binding protein) and S6K [6–8]. These domains were proposed to play a role in the folding or proper organization of the kinase domain . Consistent with this proposal, solution structure and mutational studies on the conserved cysteine residues that form a redox-sensitive disulfide-bonded loop on the TOR1 FATC domain revealed that the protein levels of TOR can be diminished by mutation of the conserved cysteine residue . Adjacent to the kinase domain of TOR is the FRB (FKBP12/rapamycin-binding) domain, where rapamycin, in complex with the peptidyl prolyl isomerase FKBP12, binds and inhibits TOR. The FRB can bind phosphatidic acid and binding of this lipid has been speculated to mediate membrane localization of mTOR [11–14].
In budding yeast, TOR is encoded by two genes, TOR1 and TOR2 (Table 1). TOR1 is part of TORC1 (TOR complex 1) whereas TOR2 is a component of TORC2 in budding yeast. In the absence of TOR1, TOR2 can substitute for its function. Under these conditions, TOR2 forms a complex with the TORC1 components, which may explain why TOR1 deletion is not lethal. However, TOR2 performs an essential function that cannot be replaced by TOR1 . In fission yeast, there are also two TORs, Tor1 and Tor2. In this organism, Tor1 is part of TORC2. Tor2 may form a complex with both TORC1 and TORC2 . Speculatively, the presence of two TOR genes in both the budding and fission yeast may have occurred by gene duplication. Eventually, the two TORs may have acquired specific functions. In other yeast genera and organisms including mammals, there is only one TOR gene. Nevertheless, mammalian TOR can form two distinct protein complexes, mTORC1 and mTORC2, similar to its yeast counterparts.
|S. cerevisiae||S. pombe||H. sapiens|
|AGC kinase target||SCH9||Sck1/2 (?)||S6K|
|Other targets||NPR1||Oca2 (?)|
|S. cerevisiae||S. pombe||H. sapiens|
|AGC kinase target||SCH9||Sck1/2 (?)||S6K|
|Other targets||NPR1||Oca2 (?)|
In budding yeast and mammals, TORC1 signalling regulates the response to nutrients . Budding yeast TORC1 forms a complex with the conserved KOG1 and LST8 (lethal with sec thirteen; in fission yeast Mip1 and Wat1) (Figures 1 and 2), whose mammalian counterparts are raptor (regulatory-associated protein of mTOR) and mLST8 (mammalian LST8; initially called GβL) respectively (Figure 3). Budding yeast and mammalian TORC1 mediate the rapamycin-sensitive functions of TORC1. In fission yeast, the analogous TORC1 is also responsive to nutrient conditions [16,17]. As in mammals, fission yeast TORC1 is controlled by Tsc1/Tsc2 (where Tsc is tuberous sclerosis complex) via negative regulation of the GTPase Rhb1 [18,19]. How TSCs are regulated in fission yeast are yet to be addressed. In mammals, several pathways convey the presence of growth signals such as insulin to mTORC1 via regulation of TSC1/TSC2 . The TSCs do not have orthologues in budding yeast, although a Rheb orthologue, RHB1, is present [21,22]. Whether RHB1 regulates TOR in this organism, despite the absence of the TSCs, remains to be investigated. Due to the amenability of yeast to genetic studies, we will probably gain important insights into the relationship between the two TORCs and the significance of TSC/Rheb and FKBP/rapamycin signals on the TORC pathway using these simpler organisms.
TOR complex signalling to AGC kinases in the budding yeast S. cerevisiae
Tor complex signalling to AGC kinases in the fission yeast S. pombe
mTOR complex signalling to AGC kinases in mammals
KOG1, a product of an essential gene in budding yeast, co-purifies with TOR1 and contains four internal HEAT repeats and seven C-terminal WD40 repeats . Fission yeast Mip1 forms a complex with and displays a genetic interaction with Tor2 (the functional counterpart of TOR1 in S. cerevisiae), as part of TORC1 [16,17].
In mammals, raptor also binds to mTOR [23–25], but this association, unlike the yeast counterpart, is sensitive to rapamycin and perhaps nutrients as well [24,26]. Raptor is thought to serve as a scaffold to present substrates to mTOR, but does not alter the intrinsic catalytic activity of mTOR. No viable null mice were obtained from heterozygous intercrosses of raptor+/− mice, similar to mTOR+/− mice, indicating that raptor is an essential gene [27,28].
mLST8 (LST8; Wat1)
LST8 from budding yeast is an essential, Gβ-like protein made up entirely of seven WD40 repeats. It is found in both TORC1 and TORC2 and binds to the kinase domain of TOR2 . LST8 is required for amino acid permease transport from the Golgi to the cell surface [30,31]. Consistent with these findings, LST8 localizes as a peripheral membrane protein to endosomal or Golgi membranes . Overexpression of CCT6 (chaperonin containing TCP-1), a chaperonin complex subunit, suppresses the abnormal growth phenotype of an LST8 mutant (lst8-2). Interestingly, this chaperonin can also suppress the growth defect of a tor2 mutant or a strain overexpressing the TORC1-regulated phosphatase SIT4 . The mechanism for this suppression is not understood, but supports a role for the TORCs in protein folding and stabilization (see PKC below).
In fission yeast, the LST8 orthologue Wat1 binds both Tor1 and Tor2 [16,17] and is required for proper localization of F-actin , although another study found that neither TORC1 nor TORC2 play a significant role in actin organization in fission yeast . mLST8 binds to mTOR and raptor when both are overexpressed [23,35]. mLST8 binds to the kinase domain of mTOR to stimulate mTOR kinase activity . Its association with mTOR is insensitive to nutrients, but is required for the nutrient-sensitive interaction of raptor to mTOR. How mLST8 can promote mTOR kinase activity is still unclear, but it has been speculated that it could contribute to the stability and folding of the mTOR kinase domain or could also play a role in the recognition and recruitment of substrates to mTOR . Knockdown of mLST8 leads to defects in actin cytoskeleton reorganization upon serum restimulation of fibroblasts, indicating that mLST8 may also function as part of mTORC2 . More recent studies using mLST8-knockout cells indicate that mLST8 is required for mTORC2 but not mTORC1 functions .
Budding yeast TOR2 forms a complex with the conserved proteins AVO1, AVO3 and LST8 to comprise TORC2. In fission yeast, TORC2 consists of Tor1, Ste20, Sin1 and Wat1. Despite the presence of only one mTOR in mammals, mTOR also forms a separate complex with rictor (rapamycin-insensitive companion of mTOR; AVO3 orthologue), SIN1 (AVO1 orthologue) and mLST8. Budding yeast and mammalian TORC2 signalling is not acutely sensitive to rapamycin. Curiously, in fission yeast, rapamycin does not inhibit TORC1 during vegetative growth, but appears to affect some of the TORC2-dependent functions [37,38]. Two-hybrid results also indicate that Fkh1 (FKBP12) can bind to Tor1, a component of TORC2, in the presence of rapamycin . Additionally, genetic data reveal that a TORC2-dependent function is inhibited by rapamycin , suggesting a more complicated mechanism for rapamycin action. In mammals, growth-factor or hormonal signals appear to regulate mTORC2 although the mechanism is not understood [39,40]. Since these signals are not conserved in yeast, mTORC2 may be modulated by signals that similarly regulate yeast TORC2. These signals are yet to be identified, although studies from yeast suggest that they are required for cell integrity and viability under stress conditions [41,42]. In fission yeast, TORC2 is also required for responses to starvation, sexual development and stress conditions [43,44].
AVO3 in budding yeast is required for TORC2 integrity, but is dispensable for the in vitro kinase activity of TOR2 . Mutations in AVO3 (formerly named TSC11) suppress csg2 mutants that accumulate sphingolipids . In this screen, mutants of tor2 were also pulled out. It is currently unclear how TORC2 may be involved in sphingolipid metabolism or signalling, although newly identified TORC2 effectors SLM1/2 (synthetic lethal with MSS4 1/2) are targets of sphingolipid signalling during heat stress [42,46–48].
S. pombe Ste20 (different from S. cerevisiae STE20) shares homology with rictor . Ste20 is probably part of TORC2, since it binds Tor1 more strongly than Tor2 upon overexpression . Fission yeast ste20 mutants are resistant to amiloride, which inhibits sodium uptake and cell growth in S. pombe . In mammals, knockdown of rictor leads to defects in both actin cytoskeleton organization and in the phosphorylation of cPKCα (conventional protein kinase Cα) and Akt at the hydrophobic motif sites [36,39,50]. The most compelling evidence for a role of rictor (TORC2) in phosphorylation of these AGC kinases was demonstrated in studies using rictor-knockout cells (as discussed below). Rictor knockout is embryonic lethal [27,51,52]. MEFs (murine embryonic fibroblasts) isolated from null embryos did not display any remarkable defects in actin cytoskeleton nor morphology, although the growth rate of rictor-null cells was slower than that in the wild-type MEFs .
A truncated form of SIN1 (JC310) was originally identified as a human protein that suppressed the heat-shock-sensitive phenotype of a S. cerevisiae mutant expressing the hyperactive RAS2 version, RAS2Val19 . Sin1 binds S. pombe SAPK (stress-activated protein kinase) whereas, in mammals, it binds JNK (c-Jun N-terminal kinase) and MEKK2 [MAPK (mitogen-activated protein kinase)/ERK (extracellular-regulated-protein kinase) kinase kinase 2] [54–56]. Hence this protein participates in Ras and/or MAPK signalling in yeast and mammals by mechanisms that await further investigation. In budding yeast, AVO1 binds to yeast TORC2. Decreasing expression of AVO1 in yeast leads to actin depolarization, a phenotype that is displayed by tor2 mutants . Mammalian SIN1 is found in the rapamycin-insensitive mTORC2 [40,52]. Knockout of SIN1 in mammals leads to embryonic lethality, loss of TORC2 complex formation and kinase activity [40,52]. In mammals, at least five alternatively spliced isoforms could be generated [55,57]. Three of these isoforms were reported to form three distinct mTORC2 complexes . Most SIN1 orthologues contain a Raf-like RBD (Ras-binding domain) and PH (pleckstrin homology) domain, but the function of these domains in vivo awaits further examination .
Other TORC interactors
Other non-conserved proteins can form a complex with either TORC1 or TORC2. TCO89 (TOR complex one 89 kDa subunit), the product of a yeast-specific non-essential gene in S. cerevisiae, originally identified as a protein involved in glycerol uptake under osmotic stress conditions , associates with TORC1 . Loss of TCO89 leads to rapamycin hypersensitivity and defects in CWI (cell-wall integrity). However, TCO89 does not exclusively associate with TORC1. It has distinct localization in vacuolar structures and associates with the vacuolar armadillo repeat protein VAC8 [62–64], and is therefore thought to perform an additional cellular role that is independent of TORC1 [61,63]. Since TORC1 was recently found at the vacuolar membrane , it remains to be examined whether TCO89 functions together with TORC1 in this compartment. TORC2 also physically interacts with the yeast-specific proteins AVO2 and BIT61 [23,61], both encoded by non-essential genes. Although their function is unknown, BIT61 can associate with SLM1 and SLM2, two proteins that mediate TORC2 function in actin cytoskeleton organization and responses to heat and oxidative stress [41,46–48,65]. The SLMs are phosphorylated in a sphingolipid-dependent manner that is counteracted by dephosphorylation by calcineurin [47,48]. TOR2 has been linked to the phosphorylation of the SLMs in vivo and in vitro, but it is unclear if this is via direct regulation [41,46].
PRAS40 (proline-rich Akt substrate of 40 kDa) can bind and negatively regulate mTORC1 [66–71]. It interferes with binding of substrates to mTOR and itself is phosphorylated by mTORC1 in a rapamycin-sensitive fashion . Akt phosphorylates PRAS40 to negatively regulate this protein, thereby promoting mTORC1 signals. Unlike TSC2, which antagonizes the positive role of Rheb on mTORC1, PRAS40 can directly inhibit mTORC1 . Another proline-rich protein, Protor (protein observed with rictor), also called PRR5 (proline-rich protein 5) binds to rictor, independently of mTOR [72,73]. The function of PRR5 is unknown, but it may play a role in PDGFR (platelet-derived-growth-factor receptor) signalling . Knockdown of PRR5/Protor does not significantly reduce phosphorylation of Akt at Ser473 . A PRR5L (PRR5-like) protein, which regulates apoptosis, was also identified to bind to mTORC2 . Both PRR5 and PRR5L are not required for binding of rictor and SIN1 to mTOR. Whether these mammalian TORC2 interactors could mediate specific functions of mTORC2 remains to be investigated.
TOR and its targets
mTOR can phosphorylate the AGC kinase S6K (S6 kinase) and the translational regulator 4E-BP . 4E-BP is a small regulatory protein with no enzymatic activity, but it is phosphorylated at multiple residues in a rapamycin-sensitive fashion to promote cap-dependent translation. A common recognition motif, TOS (TOR signalling), is found in S6K and 4E-BP , which could explain why mTOR can phosphorylate these two structurally diverse targets. mTOR, as part of the rapamycin-insensitive complex mTORC2, was also shown to regulate the phosphorylation of another AGC kinase, Akt [39,76]. In budding and fission yeast, TOR can mediate phosphorylation of yeast AGC kinases as well [62,77]. Other non-AGC kinase proteins in yeast are phosphorylated in a TOR-dependent manner, such as the phosphatase regulator TAP42 (type 2A phosphatase-associated protein 42) and TIP41 [78,79], transcription factor GLN3 , NPR1 (nutrient permease regulator) kinase  and phosphatidylinositol-binding proteins SLM1/2 . Whether TOR can directly phosphorylate these targets via recognition of a common motif would need more rigorous scrutiny. Comparison of all known and putative TOR/mTOR targets indicates that TOR may target a serine/threonine residue followed by a proline residue or by a hydrophobic, aromatic amino acid such as a phenylalanine residue. The TOR target site in AGC kinases lies at the conserved hydrophobic motif located at the C-terminal tail of the kinase domain. There are other conserved motifs in the C-terminal tail among the AGC kinases that contain phosphorylated residues, which can also be potentially regulated by TOR. Phosphorylation at the C-terminal tail critically regulates the catalytic activity, stability and localization of the AGC kinases. The importance of phosphorylation at this segment was revealed in early studies on PKC and S6K [82–84]. The structural divergence of the C-terminal tail of AGC kinases seems to coincide with the emergence of TOR. Interestingly, a parasitic unicellular eukaryote was recently reported to have primordial AGC kinases [related to PKA (protein kinase A) and NDR] but does not have TOR . Hence, the TOR and AGC kinase C-terminal tail could have co-evolved to fulfill a basic function among eukaryotes, perhaps to balance growth versus survival signals in response to nutrients and environmental conditions.
REGULATION OF AGC KINASE DOMAIN IN YEAST AND MAMMALS
The AGC kinases belong to the conventional ePKs (eukaryotic protein kinases) (Table 2) . Members of the AGC family share considerable homology in their kinase domains . The catalytic core of AGC kinases, like other ePKs, has a bilobal composition . The smaller N-terminal lobe binds nucleotides whereas the large C-terminal lobe participates in substrate binding and catalysis. Like other ePKs, the AGC kinases are activated by phosphorylation at the activation loop (also called the T-loop, consensus sequence T*FCGT, where * is a phosphorylated residue) that connects the N- and C-lobes of the kinase domain. Phosphorylation at a conserved serine/threonine residue in the T-loop can occur via autophosphorylation or can be catalysed by another protein kinase. Several members of the AGC family can be phosphorylated at the T-loop by PDK1. The AGC kinases are characterized by the presence of a C-terminal tail that is critical for interactions with the N- and C-lobes of the kinase domain. Three regions mediate these interactions, namely the NLT (N-lobe tether), which includes the conserved HM (hydrophobic motif), the AST (active-site tether) and the CLT (C-lobe tether) . These regions could act as cis-acting regulatory modules to allow binding of trans-acting cellular molecules to further regulate AGC kinase activity .
|Budding yeast AGC kinases||Fission yeast AGC kinases||Mammalian AGC kinases|
|PKC1||Pck1/2||cPKC-α/β/γ, nPKC-η/ϵ/δ/θ, aPKC-ι/ζ, PRK1/2, PKN3|
|ROCK1/2, DMPK, MRCK|
|SCH9||Sck1/2||S6K1/2, RSK1/2/3/4, MSK1/2|
|TPK1/2/3||Pka1||PKAc-α/β/γ, PRKY, PRKX|
|MAST1/2/3/4, MASTL βARK1/2, GPR4/5/6/7, GPRK7|
|Budding yeast AGC kinases||Fission yeast AGC kinases||Mammalian AGC kinases|
|PKC1||Pck1/2||cPKC-α/β/γ, nPKC-η/ϵ/δ/θ, aPKC-ι/ζ, PRK1/2, PKN3|
|ROCK1/2, DMPK, MRCK|
|SCH9||Sck1/2||S6K1/2, RSK1/2/3/4, MSK1/2|
|TPK1/2/3||Pka1||PKAc-α/β/γ, PRKY, PRKX|
|MAST1/2/3/4, MASTL βARK1/2, GPR4/5/6/7, GPRK7|
In addition to phosphorylation at the T-loop, most AGC kinases are phosphorylated at the HM (consensus sequence FXXFS/T*Y/F) . The HM is significant for the following reasons. First, the HM site serves as a docking site for PDK1, which catalyses phosphorylation of the T-loop [89,90]. PDK1, which lacks the HM, docks to the HM site of its substrate via its own hydrophobic pocket [referred to as the PIF (PDK-interacting fragment) pocket]. Secondly, phosphorylated HM folds back and interacts with the hydrophobic pocket of the N-lobe. Occupation of the hydrophobic pocket leads to optimal alignment and stabilization of the αC-helix conformation. Together with T-loop phosphorylation, binding of the phosphorylated HM site to the hydrophobic pocket significantly stabilizes the kinase conformation and allosterically activates the AGC kinase. Several kinases have been proposed as a possible ‘PDK2’, the HM kinase . Whereas most AGC members share the same T-loop kinase, PDK1, it seems that the PDK2 is variant among the AGC kinases because there are subtle differences in the mechanism of this phosphorylation. In both yeast and mammals, accumulating data indicate that TOR/mTOR is the ‘PDK2’ and mediates the phosphorylation of the HM site of several AGC kinases. We will discuss the evidence that supports a role for TOR/mTOR as the PDK2 and alternative mechanisms as to how it can mediate or allow HM phosphorylation.
Apart from the HM site, most of the AGC kinases are phosphorylated at several sites in the C-terminal tail. Between the AST and NLT, there is a highly conserved region named the TM (turn motif). In PKA this is at the apex of a tight turn of the tail, hence the name . Most of the AGC kinases are phosphorylated at this residue presumably by TOR/mTOR. S6K and budding yeast SCH9 are additionally phosphorylated at several sites in the C-terminal tail adjacent to the HM [62,93,94]. Phosphorylation of one or more of these sites is rapamycin-sensitive, and most probably directly regulated by TOR.
Budding yeast AGC kinases
The budding yeast (also known as baker's yeast or S. cerevisiae) genome codes for 117 ePKs and 10 aPKs (atypical protein kinase) , with approx. 20 of the ePKs belonging to the AGC kinase family (Table 2 and Figure 4A) [3,95]. We will focus our discussion on the AGC kinases that have genetic or functional linkage to TOR (Figure 4A).
Yeast AGC kinase conserved regions and phosphorylation sites
YPK (yeast protein kinase)
YPK1/2 [also known as YKR2 (yeast-kinase related)] have the highest homology with mammalian SGK (serum- and glucocorticoid inducible kinase) and Akt and can be functionally replaced by SGK, and partially by Akt1 . Deletion of YPK1 results in a slow growth phenotype and in rapamycin hypersensitivity, whereas the loss of YPK2 shows no apparent phenotypic defect [97,98]. The simultaneous deletion of both genes is lethal, suggesting overlapping functions. YPK-deficient cells exhibit translation initiation arrest and in wild-type cells YPK1 itself is rapidly degraded upon nitrogen starvation . These findings suggest that YPK may be a component of a nutrient-sensing pathway that can regulate translation initiation and point to the rapamycin-sensitive TORC1 as a probable YPK regulator. However, rapamycin does not affect the kinase activity and expression levels of YPK . Instead, other evidence points to TORC2 as the upstream regulator of YPK.
In a genetic screen, a 5′-truncated version of the YPK2 gene was identified as a suppressor of the lethality of a strain depleted for TOR2 . Mutation of a conserved sequence (T237FDVT243) in the N-terminal region, resulting in the YPK2D239A allele, suppresses the defective actin organization of a tor2 mutant and the impaired phosphorylation of MPK1, a downstream effector of TOR2 in the cell-integrity pathway. This N-terminal amino acid sequence, which resembles the TOS motif, negatively regulates YPK activity. TORC2 can phosphorylate YPK2 in vitro and this phosphorylation is decreased upon mutation of the HM (Thr659) and the TM (Ser641) sites to an alanine residue, suggesting that these sites are targeted by TORC2 . The TORC2-mediated phosphorylation relieves the inhibition from the TOS-like regulatory region and allows phosphorylation of the activation loop. The kinase activity of YPK2 in a temperature-sensitive tor2-21 mutant was extensively diminished when compared with YPK2 from wild-type cells, confirming that TOR2 as part of TORC2 regulates YPK2 activity . YPK1 and TORC2 act in the same pathway since overexpression of the RHO1-GEF (RHO1-guanine-nucleotide-exchange factor), TUS1, suppresses ypk or tor2 mutants [100,101].
YPK1 is activated by phosphorylation at the T-loop (Thr504) of the kinase domain by PKH (PKB-activating kinase homologue) 1, the PDK1 orthologue in budding yeast . This phosphorylation is essential for YPK1 function because mutation of Thr504 to an alanine residue cannot complement the sensitivity of a ypk1Δ mutant to myriocin, a sphingolipid biosynthesis inhibitor . An analogous mutant allele of YPK2 also does not complement a ypkΔ mutant, suggesting that T-loop phosphorylation is also essential for YPK2 function . YPK1 was isolated as a high-copy suppressor of the lethal sphingolipid depletion induced by myriocin . This suppression may be accounted for by the sphingolipid-mediated activation of PKH , which in turn can activate YPK. YPK can also be activated by the sphingolipid signalling molecule, phytosphingosine, in a more direct manner via an unknown mechanism, but speculatively due to HM site phosphorylation by TORC2. A phosphomimetic mutant of YPK1 at the HM site T662D shows higher kinase activity than wild-type and restores myriocin resistance [102,105]. Whether YPK may be recruited to a membrane compartment where it binds sphingolipids is not clear, since YPK has no apparent lipid-binding domain, although it localizes to the membrane depending on the sphingolipid level [64,103,105]. Together, these findings corroborate that the YPKs are downstream effectors of TORC2. However, since other findings suggest a possible role of YPK in translation initiation, it is likely that there is significant crosstalk between the two TORC signals.
SCH9 was isolated as a suppressor of the Ras/PKA signalling defects  and is required for longevity and cell size in budding yeast [107,108]. The closest mammalian homologue of SCH9 is Akt, although recent analysis revealed it has functional resemblance to S6K  and is directly targeted by TORC1. Phosphorylation of seven serine/threonine residues in the C-terminus of SCH9 occurs in vivo . Six of them were rapamycin- and nutrient-sensitive, indicating that they are regulated by TORC1. Thr737 lies in a classical HM sequence whereas Ser758 and Ser765 share similarity to the HM due to the presence of hydrophobic residues surrounding the phosphorylation site (Figure 4A). Ser711 is also followed by a hydrophobic residue at position +1 and +2. Thr723 and Ser726 are followed by a proline residue, very similar to phosphorylated residues in the turn motif. TORC1 directly phosphorylates at least five of these sites in vitro. Phosphorylation of these residues is absolutely essential for SCH9 function. A combined alanine mutation of these TORC1 target sites (SCH95A) abolishes in vitro kinase activity of SCH9. Whether the C-terminal tail sites distal to the HM are analogous to the autoinhibitory domain in S6K remains to be examined.
Like other AGC kinases, SCH9 needs to be phosphorylated by PKH on residue Thr570 of the activation loop. Addition of phytosphingosine, which stimulates PKH1, also leads to phosphorylation and activation of SCH9 in the presence of PKH1 and PKH2 in vitro . Like YPK, SCH9 is activated by phytosphingosine indirectly via PKH phosphorylation of the T-loop. Since a PKH kinase-dead allele also increases SCH9 phosphorylation in the presence of phytosphingosine, a more direct effect of this sphingolipid on SCH9 may also occur. This mechanism remains to be characterized, but speculatively involves the phosphorylation of the C-terminal tail sites by TORC1.
Unlike YPK, SCH9 has a predicted C2 lipid-binding domain. SCH9 is predominantly at the vacuolar membrane . The vacuolar surface also contains active TORC1. Consistent with a model that TORC1 may regulate SCH9 in this compartment, vacuolar membrane targeting of SCH9 results in hyperphosphorylation of SCH9 in a rapamycin-sensitive manner . The targets of SCH9 in this cellular compartment remain to be identified. Since the vacuole is an important reservoir of nutrients in yeast, TORC1 and/or SCH9 may play a role in sensing and/or remobilization of intracellular nutrients. Compatible with its structural similarity to S6K, SCH9 also regulates translation initiation, ribosome biogenesis and phosphorylates RPS6 (ribosomal protein S6) in vitro . SCH9 phosphorylation is also reduced following osmotic, oxidative or thermal stress . Another study provided evidence that SCH9 plays a role in the transcriptional activation of genes that are essential for osmotic stress responses . This function of SCH9 is dependent on the MAPK HOG1 (high osmolarity glycerol response 1) and the ATF (activating transcription factor)/CREB (cAMP-response-element-binding protein) transcription factor SKO1, suggesting that SCH9 functions in a similar manner to the mammalian AGC kinase MSK1/2 (mitogen- and stress-activated kinase 1/2), a close homologue of Rsk1/2 (p90 ribosomal S6K 1/2) . SCH9 does not contain the C-terminal kinase domain that is regulated by MAPK in MSK1/2, but the HOG1/SCH9 signalling mechanism may be ancestral to the mammalian p38 MAPK/MSK signalling and should yield insights on how nutrient and stress signals are integrated.
The budding yeast genome encodes only one PKC, namely PKC1. PKC1 is considered an archetypal PKC , since it possesses all the domains found in the various mammalian PKC isotypes, including the PRKs (PKC-related kinases) and ROCKs (Rho-associated kinases). In a similar manner to the ROCK subfamily of AGC kinases, it binds to and is regulated by Rho1 (GTPase) . The pkcΔ mutant can also be complemented by mammalian novel PKC . PKC1 functions in the CWI signalling pathway and in regulating actin polarization via MAPK signalling . TORC2 signals to the actin cytoskeleton via PKC1. Previous studies have demonstrated that the PKC regulation by TORC2 occurs via the activation of the Rho1 GTPase . PKC1 associates with and becomes activated by GTP-bound Rho1 . It is also phosphorylated by PKH1/2 at Thr983 of the activation loop . In vitro, PKC1 phosphorylation by PKH1 is enhanced by the presence of phytosphingosine , suggesting that PKC1 may be activated by sphingolipids similar to YPK2. The phosphatidylinositol lipid PtdIns(4,5)P2 can bind to the Rho-GTPase GEF Rom2 and may also be required to activate Rho1 and promote PKC activation . The role of lipids (sphingolipids and phospholipids) in PKC activation remains to be further examined, although addition of phosphatidylserine, in the presence of GTP-bound Rho can potently activate PKC1 in vitro . Despite the presence of a lipid-binding domain, PKC1 does not seem to be activated by diacylglycerol or Ca2+, unlike conventional mammalian PKC. Whether PKC1 can be directly regulated and activated by TORC2 remains to be examined. Recently, one of us found that phosphorylation of the conserved TM site in PKC1 is mediated by TORC2 (E. Jacinto, unpublished work). Yeast cells depleted for the TORC2 component AVO1, but not the TORC1 component KOG1, had decreased TM phosphorylation. Phosphorylation of the TM site in mammalian cPKC is important for stability of the kinase and in the absence of this phosphorylation, molecular chaperones such as Hsp (heat-shock protein) 70, were shown to bind and stabilize cPKC [118,119]. In line with these findings, the growth defect of yeast strains expressing a defective Hsp40 and temperature-sensitive allele of Hsp90 was suppressed by overexpression of PKC1 . These findings suggest that stabilization of AGC kinases via phosphorylation and chaperone action could play an important role in their signalling function. Other possible phosphorylation at the C-terminal tail of PKC1, such as the HM site, may also be regulated by TORC2, but this remains to be formally demonstrated.
Other yeast AGC kinases: are they regulated by the TORCs?
Several other AGC kinases in budding yeast have been linked to TOR signalling or at least function in a parallel pathway with common downstream targets. Aside from the AGC kinases discussed above, none of the other members of this family in yeast has been shown to be regulated by the TORCs at their conserved C-terminal tail.
PKA (cAMP-dependent protein kinase)
In yeast there are three coding genes for the catalytic subunit of PKA, TPK1, TPK2 and TPK3 and one for the regulatory subunit BCY1. The N-terminal region of these three catalytic isoforms is highly variable, but the conserved kinase domains have more than 75% homology with each other. Like mammalian PKA, cAMP-binding leads to dissociation of BCY1 from the catalytic subunits to activate PKA. Addition of glucose to cells grown on a fermentable carbon source results in rapid and transient cAMP accumulation and PKA activation. Although there is no evidence that either TORC1 or TORC2 directly phosphorylate subunits of PKA, several lines of evidence indicate a functional linkage between TOR and PKA signalling. Both signalling pathways are key regulators of ribosome biogenesis . Mutations that hyperactivate PKA signalling confer rapamycin resistance . Furthermore, constitutive activation of the Ras/cAMP pathway prevents several rapamycin-induced responses such as nuclear translocation of the stress-related transcription factors MSN2/MSN4, induction of stress genes, induction of autophagy, glycogen accumulation and down-regulation of ribosome biogenesis . Although these findings indicate that PKA is an effector of TOR signals, the mechanisms for PKA regulation by TOR is obscure. Upon rapamycin treatment, TPK1 accumulates in the nucleus, a response similar to cAMP deprivation , suggesting that TOR may regulate localization of TPK. BCY1 is mostly nuclear in glucose-growing cells, and localization of TPK to the nucleus upon rapamycin treatment may promote inhibition of PKA activity by association of TPK with BCY1. These findings imply that TOR maintains PKA activity in the cytoplasm. TPK1 undergoes phosphorylation in a glucose-dependent manner, but the sites that undergo phosphorylation have not been identified, although it has been surmised that possible sites other than the T-loop site become phosphorylated . The TPKs do not contain a serine/threonine residue at the homologous turn motif, however. Therefore it is unlikely that TOR regulates phosphorylation of the C-terminal tail of TPKs.
DBF (dumbbell forming)
DBF2/20 are part of the NDR subfamily of AGC kinases. A unique feature of this subfamily is the presence of a 30–60-amino-acid insert between subdomains VII and VIII of their kinase domain. The function of this insert is poorly understood, but it is known in mammalian NDR1 to contain a nuclear localization signal . Members of this kinase subfamily are mainly nuclear and regulate cell cycle and morphogenesis. Although so far no direct link between DBF2/DBF20 and the TOR complexes has been observed, a possible connection between the DBF2/20 and TOR could exist on the basis of genetic data. dbf2 mutants show increased sensitivity to rapamycin and wortmannin in comparison with wild-type cells . Interestingly, they also show increased glycogen accumulation in a similar manner to tor1 mutants . The T-loop site (Ser281 of NDR1) of this family of kinases becomes autophosphorylated in mammalian NDR1, whereas the HM site (Thr444) could be phosphorylated by the STE20 family of kinases. In budding yeast, the DBF2 upstream kinase is the STE20-like CDC15 . The amino acids surrounding the HM site diverge from the consensus sequence (Figure 4A), hence TOR may only indirectly regulate DBF.
RIM15 (regulator of IME2)
RIM15 is a distant member of the NDR family of AGC kinases. Regulation of RIM15 occurs via the regulation of its subcellular distribution. RIM15 seems to be a common target of TORC1 and PKA . A PKA-dependent phosphorylation leads to nuclear exclusion of RIM15 because of binding to BMH1 and BMH2 [95,130]. In parallel, TOR also negatively regulates nuclear accumulation of RIM15 via unknown mechanisms .
PKH1/2 encode orthologues of mammalian PDK1 since the lethality of a pkh1Δpkh2Δ double mutant is suppressed by expression of human PDK1 [96,116]. PKH1/2 have no apparent PH domains and they are unlikely to be regulated by PtdIns(3,4,5)P3 since budding yeast does not produce PtdIns(3,4,5)P3 or PtdIns(3,4)P2 . In addition, PDK1 lacking a PH domain could rescue pkh1Δpkh2Δ growth defects . In contrast with mammalian PDK1, which is regulated by 3-phosphoinositides, PKHs need sphingolipids for full activation . Deletion of PKH1, but not PKH2, also results in rapamycin hypersensitivity . Although TOR and PKH share some overlapping functions, there is no evidence that PKH is directly regulated by TOR.
Fission yeast AGC kinases
The fission yeast (S. pombe) genome codes for 106 protein kinases . Of these, 20 were classified as potential AGC kinases (Table 2 and Figure 4B) . There is very little known on how these AGC kinases may be regulated at their kinase domain. Remarkably, the existence of a conserved TOR–PDK1–AGC kinase signalling module was first recognized in S. pombe studies .
Gad8 (G1-arrest defective 8)
Gad8 is the fission yeast orthologue of budding yeast YPK and mammalian SGK, although rescue experiments with the heterologous proteins have yet to be performed. gad8+ was first isolated as a high-copy suppressor of the sterility and the temperature-sensitive growth phenotype of a mutant expressing a C-terminally truncated Tor1 protein . Site-directed mutagenesis of each of the AGC kinase conserved phosphorylation sites (Thr387, Ser527 and Ser546) to an alanine residue revealed that phosphorylation of Ser527 in the turn motif and Ser546 in the HM is Tor1-dependent. However, evidence that Ser527 and Ser546 are directly phosphorylated by Tor1 is missing. As in mammals and budding yeast, Thr387 within the activation loop of Gad8 is phosphorylated directly by the PDK1 orthologue, Ksg (kinase responsible for sporulation and growth) 1, in vitro. Analysis of gad8 mutant alleles demonstrated that phosphorylation of all three residues (Thr387, Ser527 and Ser546) is essential for Gad8 activity and in vivo function. Gad8 also bears a C2 lipid-binding domain of unknown function.
Sck1/2 (suppressor of loss of cAMP-dependent protein kinase)
Sck1/2 represent the fission yeast orthologues of budding yeast SCH9 [133,134]. As in budding yeast, loss of PKA activity can be suppressed by overexpression of sck1+ or sck2+ . Interestingly, overexpression of sck2+ leads to decreased mating under starvation conditions . Inhibition of sexual differentiation and mating also occurs in cells overexpressing tor2+, a component of TORC1 . The role of Ksg1 and Tor2 in phosphorylating the conserved sites in the T-loop and C-terminal tail respectively, remains to be demonstrated. Like SCH9, Sck1 has a C2 lipid-binding domain of unknown function.
Ksg1 is an essential protein and shows the typical domain structure known for PDK1-like enzymes, including a C-terminal PH domain . This PH domain may bind lipids since S. pombe can synthesize PtdIns(3,4,5)P3 . ksg mutants show impaired mating and low sporulation efficiency [135,137]. Mutants also show defects in G1 arrest in response to nitrogen starvation . Ksg1 also controls CWI, since growth defects of ksg1 mutants could be suppressed by the addition of osmotic stabilizers . All phenotypic defects could be suppressed by the expression of human PDK1, confirming that Ksg1 is the fission yeast PDK1 orthologue .
Ksg1 interacts with the fission yeast orthologues of PKC, Pck1 and Pck2, in a two-hybrid assay. This interaction depends on the presence of the PH domain of Ksg1 . The interaction of Ksg1 and Pck1 was also confirmed biochemically. Genetic evidence supports the view that Pck1 acts downstream of Ksg1, since ksg1 mutants are suppressed by overexpression of Pck2. The PKA orthologue in fission yeast, Pka1, is also regulated by Ksg1. Pka displayed less phosphorylation in ksg1 mutants. Phosphorylation of Thr356 of Pka within the activation loop is Ksg1-dependent . However, evidence for direct regulation of Pka1 by Ksg1 is not available. In a similar manner to yeast PKH and mammalian PDK, there is currently no evidence linking fission yeast Tor in the regulation of Ksg1.
Mammalian AGC kinases
Out of approx. 500 protein kinases in the human genome, there are 63 kinases grouped into the AGC family (Table 2 and Figure 5). Apart from the catalytic domain, the AGC kinases show little similarity and not surprisingly, are controlled via distinct mechanisms. Nevertheless, the structural similarity of their catalytic domain suggests overlapping substrate specificity and regulation of this domain.
Mammalian AGC kinases conserved regions and phosphorylation sites
Initial studies identified S6K as insulin- or interleukin 2-activated protein kinase that is potently inhibited by rapamycin [139,140]. S6K is encoded by two genes, S6K1 and S6K2 . Active S6K phosphorylates the ribosomal protein S6 to promote protein synthesis. S6K can be divided into four modular domains (Figure 5): the N-terminal domain, which contains the TOS motif, the catalytic domain, which includes the activation loop, a linker domain, which contains two important regulatory phosphorylation sites namely the TM and HM sites, and the autoinhibitory pseudosubstrate domain that is unique to S6K among AGC kinases and is phosphorylated at multiple sites . In S6K2, a proline-rich region is also present following the autoinhibitory domain . Phosphorylation of S6K occurs at multiple sites in a hierarchical manner and is sensitive to nutrient or growth-factor withdrawal and rapamycin treatment. The TOS motif in the N-terminus facilitates association of S6K1 with the mTORC1 partner raptor to allow efficient phosphorylation of S6K1 . Truncation of the N-terminus including the TOS motif inhibits S6K activation [144,145]. This motif consists of a five-amino-acid sequence (FDIDL) that is conserved among S6K1 and S6K2 across species . A similar motif (FEMDI) is also found at the C-terminus of the translational regulator 4E-BP, another well-characterized mTORC1 substrate. This motif may not exclusively be a TORC1-recognition motif, since a similar sequence is found in the yeast YPK2, which is regulated by TORC2 .
The phosphorylation of Thr229 at the activation loop by PDK1 is mandatory for S6K activation. In PDK1−/− ES (embryonic stem) cells, Thr229 phosphorylation and kinase activity are abrogated . PDK1 docks to the phosphorylated HM site (Thr389) of S6K. In cells with a knockin mutation of the PH domain of PDK1 that abolishes PtdIns(3,4,5)P3 binding, S6K can be activated and phosphorylated at Thr389 by expression of Rheb or by the addition of amino acids, but not by IGF-1 (insulin-like growth factor 1) [147,148]. However, the PIF pocket of PDK1 was found to be required for the amino-acid-induced activation. These findings support the model in which PtdIns(3,4,5)P3 binding to PDK1 is not required for the nutrient-induced S6K activation but PDK1 docking to the HM site is critical for S6K phosphorylation. The failure of IGF-1 to stimulate S6K in the PH domain mutant knockin cells could be explained by the defective Akt activation in these cells, which leads to down-regulation of Rheb and mTORC1  (Figure 3).
S6K phosphorylation at Thr389 of the HM and Ser404 at the linker domain is highly rapamycin-sensitive, indicating that S6K is regulated by mTORC1 . Indeed, immunopurified mTOR can phosphorylate S6K1 at Thr389 and Ser412 (Ser411) in vitro [149,150]. Knockdown of the mTORC1 component, raptor, also led to decreased phosphorylation of Thr389  and Ser411  and correlated with decreased cell size and growth rate. However, other kinases were also implicated to phosphorylate Thr389 such as PDK1 , NEK [NIMA (never in mitosis in Aspergillus nidulans)]-related kinase] 6 and 7 , and S6K itself . It is possible that there are several inputs that promote or enhance the phosphorylation of the HM site to prolong the activation state or enhance stability of S6K. S6K is also phosphorylated at seven other residues with a serine/threonine–proline motif . One of these sites, Ser371, is at the TM, located at the linker region, close to the HM site. Phosphorylation at the TM is mitogen-inducible and is required for kinase activity along with HM and T-loop phosphorylation . Mutation of the TM site to an alanine or aspartate residue abolished serum-induced activation of S6K. However, a phosphomimetic mutant of the HM site did not rescue the activity of the TM alanine mutant, indicating that phosphorylation at the TM site plays a more independent role in regulating S6K activity . There are conflicting reports as to whether this phosphorylation is rapamycin-sensitive [93,155,156], although it was shown to be phosphorylated by mTOR in vitro .
Four phosphorylation sites (Ser411, Ser418, Thr421 and Ser424), all of which contain a proline residue at the +1 position and a hydrophobic residue at the −2 position, are found in the autoinhibitory domain. These sites are hypophosphorylated in quiescent cells and undergo hyperphosphorylation upon serum stimulation. Mutation of these sites to an alanine residue suppresses activation of S6K, but phosphomimetic mutants still require phosphorylation at the activation loop and linker region to fully activate the kinase . Deletion of the C-terminus, including the autoinhibitory domain, or substitution of acidic residues for all four residues, reversed the effect of truncation of the S6K1 N-terminus and led to kinase activation [144,145,157]. This active mutant allele of S6K1 is also rapamycin-insensitive, i.e. Thr389 of the HM remains phosphorylated. Surprisingly, this phosphorylation depends on mTORC2 . An extended C-terminal tail containing multiple phosphorylation sites also exists in SCH9, a TORC1 target . These findings suggest that the structure of the C-terminal tail may play an important role in determining the requirement for mTOR partners. Phosphomimetic mutation of the four residues (D3E) in the autoinhibitory domain does not enhance T-loop phosphorylation, although this mutant S6K is catalytically more active. Therefore these four sites co-operate with HM phosphorylation to synergistically activate the kinase. T-loop phosphorylation can occur, albeit weakly, in the absence of Thr389 phosphorylation, suggesting that the HM phosphorylation although not absolutely required, increases the affinity for PDK1 to enhance T-loop phosphorylation. Intriguingly, in PDK1−/− ES cells, Thr389 phosphorylation is also absent , suggesting that PDK1 may influence HM site phosphorylation as well.
How is mTORC1 activity towards S6K regulated? It was previously demonstrated that, depending on the activation state of S6K, it could associate with different regulatory complexes. In the inactive state, it is bound to eIF3 (eukaryotic initiation factor 3) . Upon cell stimulation, mTOR/raptor binds to eIF3 and subsequently phosphorylates S6K at Thr389. The phosphorylation of S6K by mTORC1 leads to its dissociation from the eIF3–mTORC1 complex and it becomes available to phosphorylate its downstream targets. These findings raise the possibility that the activity of mTOR/raptor is dependent on the signalling complex it associates with and most probably on the cellular compartment it localizes to. The nutrient-dependent cellular signals that promote mTORC1 phosphorylation of S6K remain to be characterized. Previously, a class III PI3K, hVps34 (human vacuolar protein sorting 34), was identified to regulate mTOR in the presence of nutrients [159,160]. The product of hVps34 kinase, PtdIns3P, mediates recruitment of proteins containing FYVE or PX (Phox homology) domains to endosomal membranes. S6K does not have discernible FYVE, PX or other lipid-binding domains, but an attractive model would be that these phospholipids may provide a platform for mTORC1 phosphorylation of S6K in a membrane compartment.
Three Akt genes have been identified in mammals: Akt1 (PKBα), Akt2 (PKBβ), and Akt3 (PKBγ) . Akt phosphorylates a number of diverse targets and functions in cell survival, growth, proliferation and migration. A unique feature of Akt not found in other AGC kinases is the presence of a PH domain in its N-terminus. This domain in Akt specifically binds PtdIns(3,4)P2 and/or PtdIns(3,4,5)P3 and plays a negative regulatory role. First, when the Akt PH domain is deleted, PDK1 can phosphorylate this mutant Akt in the absence of phosphoinositides [162,163]. Secondly, the PH and kinase domains were found to interact intramolecularly to prevent PDK1 phosphorylation of the activation loop despite the physical association of PDK1 with Akt . PDK1 also contains a PH domain in its C-terminus and exhibits a high affinity for PtdIns(3,4,5)P3 . Thus the presence of PtdIns(3,4,5)P3 could enhance phosphorylation of Akt by binding of PtdIns(3,4,5)P3 on the PH domain of PDK1. Indeed, knockin mutation of the PH domain of PDK1 that fails to bind PtdIns(3,4,5)P3 abolished Akt activation by IGF-1 . Therefore PtdIns(3,4,5)P3 binding to both Akt and PDK1 PH domains are required for Akt activation. This requirement could co-localize these two kinases at the membrane where they bind the lipids. The finding that membrane-associated Akt mutants become constitutively active and phosphorylated at Thr308 supports this view [166–168]. Since these mutants are also constitutively phosphorylated at Ser473 of the HM site, phosphorylation of the HM is also believed to occur in the membrane.
The kinase that phosphorylates the HM site of Akt has been termed ‘PDK2’ . Among the kinases that were identified as a PDK2 in addition to mTOR are DNA-PK (DNA protein kinase), ILK (integrin-linked kinase), cPKC, ATM and Akt itself [91,170–173]. HM phosphorylation is induced by growth factors such as insulin and is sensitive to inhibitors of PI3Ks such as wortmannin. Previous studies provide compelling evidence that mTORC2 mediates Akt phosphorylation, strongly suggesting it is the bona fide PDK2 [39,76]. mTORC2 promotes phosphorylation of Ser473in vitro in a serum-inducible and wortmannin-sensitive manner . In another study using 3T3 L1 adipocytes, the kinase activity was not induced by serum stimulation but was increased by exogenous addition of PtdIns(3,4,5)P3 , consistent with the previously held proposition that the PDK2 is constitutively active at the membrane and that insulin enhances Ser473 phosphorylation via the effect on Akt itself and not on PDK2 [76,174,175]. It is noteworthy that although Ser473 phosphorylation is highly inducible in vivo upon growth-factor or serum stimulation, its induction in vitro using the immunoprecipitated mTOR complex as a kinase source is not as dramatic. Although there could be several technical explanations for this, such as high basal phosphorylation of exogenous Akt, it is also possible that the spatial context is amiss in these assays. The presence of negative regulators that can associate with the mTOR complex may also inhibit mTOR activity in vitro .
In vivo analyses also support mTORC2 as a critical regulator of HM phosphorylation. Knockdown of the mTORC2 component rictor, but not the mTORC1 component raptor, led to defective phosphorylation of Akt at Ser473 . This decrease in Ser473 phosphorylation was also accompanied by defective Thr308 phosphorylation, in agreement with previous claims that Ser473 phosphorylation is important for subsequent Thr308 phosphorylation [168,176,177]. Genetic studies using MEFs from mice that have deficient mTORC components confirm that mTORC2 is required for HM site phosphorylation. In rictor-, mLST8- or SIN1-knockout cells, Ser473 phosphorylation was abolished. Phosphorylation at Thr308 was not grossly abnormal in rictor−/− cells [51,52] and was normal in mLST8−/−  and in SIN1−/− cells . These findings support the alternative view that PDK1 docking to the phosphorylated HM site of Akt is not critical for T-loop phosphorylation (see also S6K and PKC). Studies from PDK1−/− ES cells have also revealed that in the absence of Thr308 phosphorylation, the HM site phosphorylation at Ser473 still occurs . Thus phosphorylation of the T-loop and HM site of Akt seems to occur independently of each other. Akt remained partly active in mTORC2-disrupted cells and phosphorylated a subset of its known substrates but was defective in the phosphorylation of the apoptosis promoting proteins Foxo1/3a . These findings suggest that the phosphorylation of Ser473 may play a role in cell survival and that HM site phosphorylation may regulate substrate specificity in Akt and possibly other AGC kinases as well.
Akt is also constitutively phosphorylated at Ser124 and Thr450 . Both residues are immediately followed by a proline residue. Ser124 is located in the linker region between the PH domain and the catalytic domain, but this residue is not conserved in Drosophila Akt . The significance of this phosphorylation site remains unknown. Thr450 is part of the TM. Structural and biochemical studies on cPKC have suggested that the phosphorylation at the TM stabilizes the protein presumably by enhancing the HM interaction at the C-terminal tail. Regulation of the TM site in Akt is quite different from the HM site. Although HM phosphorylation is growth-factor-inducible, TM phosphorylation is constitutive, but is dependent on the integrity of the PH domain . Since TM phosphorylation in S6K and yeast AGC kinases has been shown to be TOR-dependent [62,77,93,99], it is probable that mTOR also regulates this site in Akt. Why the TM and HM sites could be regulated very differently by mTOR is quite intriguing and implies that it is not the direct regulation of mTOR activity itself that leads to phosphorylation of these two sites. Regulation of the mTORC components or compartmentalization of mTORCs are possible mechanisms to distinctly regulate these sites. If mTOR is not the direct TM or HM kinase, mTORC2 may facilitate autophosphorylation, or activate the TM/HM kinases. Other possible mechanisms include inhibition of phosphatases, and/or mediating compartmentalization, both of which are discussed below.
Phosphatases that dephosphorylate and inactivate Akt probably play an equally important role as well. PP2A was proposed to dephosphorylate Akt based on the effect of the PP2A-specific inhibitor okadaic acid . Another type of phosphatase of the PP2C family, PHLPP (PH domain leucine-rich repeat protein phosphatase)  which is okadaic-acid-insensitive, was identified as more specific for the dephosphorylation of Ser473 of the HM site of Akt. The two PHLPP isoforms display preference towards different Akt isoforms to regulate Akt substrate specificity. Thus, in addition to the mTORC2 regulation of Akt substrate specificity, PHLPP also regulates Akt substrate specificity via dephosphorylation of Akt at the HM site. It would be interesting to see whether mTORC2 may also regulate PHLPP. Studies in yeast have emphasized a critical role for TOR in the negative regulation of phosphatases. The yeast phosphatase regulator TAP42 was shown to associate with TORC1 on membrane structures and the TAP42–phosphatase complex is released into the cytosol upon rapamycin treatment and nutrient deprivation . These findings suggest that the phosphatases are kept inactive by TORC1 in the membrane and imply that compartmentalization plays an important role in the regulation of TORC targets.
In mammals, there are three classes of PKC, grouped according to the domains present in the regulatory moiety present in the N-terminus. The conventional type of PKC (cPKC) includes α, β1, βII and γ; the nPKC (novel type) consists of δ, ϵ, θ and η/L; and the aPKC (atypical PKC) consists of ζ and ι/λ . The activation loop of these different PKC classes is phosphorylated by PDK1. T-loop phosphorylation by PDK1 is constitutive and does not require phosphoinositides . Newly synthesized PKC associates with a membrane compartment where it is in an ‘open’ conformation and can be phosphorylated at the T-loop and C-terminal tail. T-loop phosphorylation by PDK1 critically regulates C-terminal tail phosphorylation. In the absence of T-loop phosphorylation, the TM and HM sites remain unphosphorylated and the protein is highly unstable. In PDK1−/− ES cells, protein levels of different PKC isotypes were largely diminished . This is consistent with the requirement for prior T-loop phosphorylation followed by TM and HM site phosphorylation to stabilize newly synthesized PKC. The mRNA level of nPKCδ was also diminished 2-fold in the absence of PDK1 . Hence the decreased protein levels may be a combined defect of the lack of phosphorylation by PDK1 at the T-loop site and defective mRNA processing or stability in the knockout cells.
In earlier studies, phosphorylation of the HM sites of the cPKCα and nPKCδ was reported to be rapamycin-sensitive, since the kinase activity of a membrane-associated protein that phosphorylates the HM site in vitro was suppressed by rapamycin treatment . Purification of this kinase activity identified the aPKCζ as the possible HM kinase, and identified that the activated mutant of aPKCζ phosphorylated the HM site of nPKCδ in a rapamycin-independent manner. Upon overexpression of nPKCδ and nPKCϵ, the HM site phosphorylation of these isoforms was also shown to be rapamycin-sensitive . Furthermore, the serum-induced phosphorylation of this site in nPKCδ is inhibited by amino acid deprivation, suggesting a role for mTOR in phosphorylating the HM site . Using pharmacological inhibitors, the HM site of nPKCϵ was dephosphorylated by a phosphatase that is not mediated through PP1, PP2A or PP2B during cell passage of quiescent cells . The phosphatase may be regulated by mTOR since dephosphorylation occurs upon rapamycin treatment. In nPKCθ, HM site (Ser695) phosphorylation is increased by PMA or anti-CD3/CD28 stimulation of Jurkat T lymphocytes, but is not sensitive to rapamycin . Taken together, the HM site of nPKC may also be regulated by mTOR, but whether this is via a direct mechanism would need to be evaluated.
Genetic studies using rictor- and SIN1-knockdown and knockout cells provide new insights into the regulation of PKC phosphorylation at the HM site. Knockdown of rictor and mTOR expression revealed diminished phosphorylation of cPKCα . In rictor-, SIN1- and mLST8-knockout cells, phosphorylation of cPKCα at the HM site (Ser657) was abolished, but this was also accompanied by a dramatic reduction in cPKCα protein expression . However, the decrease in cPKCα expression was not evident in rictor-knockdown cells . It is possible that the T-loop and TM site phosphorylation in cPKC are also defective. Earlier in vitro studies have demonstrated that, without TM-site phosphorylation, cPKCα is sensitive to oxidation, cleavage and dephosphorylation, whereas kinase activity is intact . Mutation of the TM site, together with adjacent compensating phosphorylation sites in cPKCβII generates an inactive kinase . Phosphorylation at the TM is necessary for subsequent HM site autophosphorylation [188,189] and occurs via an intramolecular mechanism . Whether TM phosphorylation can also occur via an mTORC2-dependent autophosphorylation of PKC is not currently clear. Based on the in vitro studies, TM-site phosphorylation could play a role in stabilization of the protein. Consistent with this interpretation, the absence of TM-site phosphorylation specifically increased the binding of cPKC to the folding chaperone Hsp70. The binding of Hsp70 to TM-site-mutated PKC stabilizes the protein and prolongs the signalling capacity of the kinase [118,191]. The amino acid sequences of all PKCs align very well in the TM and probably adopt the same structure. Analysis of aPKCι and cPKCβII structure implies that TM phosphorylation is critical in stabilizing the kinase domain by fixing the C-terminus at the top of the upper lobe of the kinase core [192,193]. The TM phosphate of PKC contacts at least three basic residues, which are also conserved in Akt . Given the constitutive nature of the phosphorylation of the HM and TM sites in PKC, it is puzzling how mTORC2 may regulate these similar sites in Akt, particularly that of the Akt HM site, which requires induction of PI3K signalling. This paradox may be explained by a model wherein the HM/TM kinase (or mTOR) is constitutively active and that the proper conformation or localization of the AGC kinase determines when it will be phosphorylated.
Other mammalian AGC kinases: evidence of mTOR regulation?
Although evidence is lacking for a direct regulation by mTOR of the C-terminal tail of the mammalian AGC kinases discussed below, these kinases have been linked to mTOR signalling, either functionally or indirectly by regulation of a common downstream target.
In mammals, there are four Rsk (Rsk1–Rsk4) and two related MSK (MSK1/MSK2) genes and, although there are orthologues in flies and worms, there is none in yeast (but see SCH9). Rsk is a downstream effector of ERK in response to numerous cellular stimuli . Among its cellular targets are proteins involved in translation, metabolism and cellular stress responses. Rsk contains two kinase domains connected by a regulatory linker sequence. The N-terminal kinase domain shares homology with other AGC kinases and phosphorylates cytoplasmic and nuclear substrates. The C-terminal kinase domain is similar to the CaMK (Ca2+/calmodulin-dependent protein kinase) family of kinases and functions to activate the N-terminal kinase. The activation of Rsk requires co-ordinate regulation by the Ras/MAPK pathway and PDK1. Four major phosphorylation sites are critical for Rsk activation. Ser369 (human Rsk2 numbering) at the linker, followed by Thr577 of the C-terminal kinase domain are phosphorylated by ERK. Ser386 of the HM site is also phosphorylated in an ERK1/2-dependent manner and, finally, Ser227 of the T-loop is phosphorylated by PDK1 [196,197]. MK2 (MAPK-activated protein kinase 2) was previously shown to phosphorylate the HM site of Rsk via p38 in dendritic cells . Notably, MK2 can also phosphorylate the HM site of Akt in neutrophils . Phosphorylation of the HM site of Rsk lacks rapamycin sensitivity and is therefore unlikely to be regulated by mTORC1, although a role for mTORC2 remains to be examined. The TM site is also phosphorylated indirectly by ERK . C-terminal truncation of Rsk resulting in the loss of the ERK-docking site, abolishes Rsk activity . Intriguingly, upon membrane targeting of this C-terminally truncated form, Rsk becomes activated and phosphorylated at the TM and HM sites. Hence it is possible that ERK may escort or mediate association of Rsk to the membrane, where it becomes phosphorylated at this linker region speculatively by mTOR.
Active Rsk is known to phosphorylate a number of substrates, most of which overlap with other AGC kinases such as S6K. Rsk phosphorylates the S6K substrates rpS6 and eIF4B, but in a rapamycin-insensitive manner [201,202]. It is interesting to note that in S6K1−/−/S6K2−/− mice, phosphorylation of rpS6 and eIF4B are normal . Hence in the absence of S6K, Rsk may compensate to allow phosphorylation of these substrates, which are critical for translation.
SGK is highly regulated on different levels: its expression, its kinase activity and subcellular localization . Three SGK paralogues exist in mammals and mainly function to modulate salt balance. SGKs activate ion channels, carriers and the Na+/K+-ATPase. It also phosphorylates other cellular substrates that are involved in stress responses and some of these targets overlap with Akt substrates. The SGKs are structurally highly similar to Akt. Although they do not contain a PH domain, they require PI3K activation for function. SGK3, but not the other two SGK paralogues, contain a PX domain . PX domains are known to bind monophosphorylated lipids and to direct PX domain proteins to endosomal membranes . SGK1 and SGK2 contain a truncated PX domain, although this domain in SGK1 can bind phosphoinositides in vitro and is required for its function . PDK1 phosphorylates SGK on its T-loop [208,209] and this requires the prior PI3K-dependent phosphorylation of SGK1 at the HM site by an unidentified kinase. Studies on SGK3 hint that localization of SGK3 at the endosomes via a PI3K-dependent and staurosporine-sensitive HM kinase is important for SGK3 activation . The hyperphosphorylation of SGK upon serum stimulation is not blocked by rapamycin treatment , suggesting that mTORC2 may mediate phosphorylation of the HM site. SGK is a very attractive candidate for mTOR regulation given the reported functions of SGK in the modulation of nutrient transporters and ion channels.
PKA is the principal target of the intracellular second messenger cAMP. The cAMP/PKA signalling pathway is activated by a number of different receptors that couple to G-proteins upon binding of their respective ligands. Mammalian PKA, like the yeast counterpart, is made up of the regulatory (R) and catalytic (C) subunits and is activated in a similar manner. Phosphorylation at two sites at the C-terminus, Thr197 of the T-loop and Thr338 of the TM, are important for catalysis and stabilization. PKA does not contain an HM. In contrast with other AGC kinases, phosphorylation of the C-subunit occurs before assembly into an inactive holoenzyme complex. PDK1 is not required for T-loop phosphorylation in vivo  although it can recognize it in vitro . Structural and biochemical studies have shown that phosphorylation at Ser338 at the TM stabilizes the kinase core. Phosphorylation of this site is resistant to phosphatases and appears to be required for stabilization . There is currently no evidence linking mTOR to phosphorylation of PKA, but the TM site may be a possible regulatory site. As in yeast, it is not entirely clear whether PKA directly mediates one or more of the functions of mTOR in growth and stress responses. Some findings indicate that PKA can phosphorylate LKB1, a tumor suppressor protein, to suppress cell growth [212,213]. Although PKA activation appears to be associated with decreased mTORC1 signalling , it remains to be examined whether mTORC2 could play a role in PKA regulation.
Unlike other AGC kinases, PDK1 is encoded by a single gene in mammals. It contains an N-terminal kinase domain and a C-terminal PH domain. The PH domain binds to PtdIns(3,4,5)P3 and localizes PDK1 to the membrane. Although binding of PtdIns(3,4,5)P3 does not increase PDK1 autophosphorylation, sphingosine appears to activate PDK1 by promoting its autophosphorylation . However, the mechanism of this is not understood. PDK1 is phosphorylated at several sites but only Ser241, the T-loop phosphorylation site, is conserved in other PDK1 orthologues . As discussed above, the T-loop phosphorylation of AGC kinases is mediated by PDK1. The importance of this phosphorylation is underscored by the finding that knockout of PDK1 or knockin mutations in the PH domain or the PIF pocket is embryonic lethal [147,148,216]. PDK1 hypomorphic mice displayed normal phosphorylation of Akt, S6K and Rsk at the T-loop but these mice had a smaller body and organ size . Hence, in addition, PDK1 could regulate growth independently of Akt, S6K and Rsk. These PDK1 hypomorphic mice also displayed reduced intestinal and renal amino acid transport and could therefore contribute to the growth defects observed in these mice . PDK1 is also required for expression of key nutrient receptors and transporters in T-cell progenitors . Studies on PDK1−/− ES cells also reveal that there is a more general defect in mRNA translation in these cells . Taken together these studies suggest that PDK1, like mTOR, plays diverse functions in growth regulation. There is currently no direct evidence linking mTOR to the regulation of PDK1.
The TOR–AGC kinase signalling module has been conserved during the evolution of eukaryotes to co-ordinately regulate the growth and survival of cells. Regulation of the kinase domain including the C-terminal tail of the AGC kinases are providing clues on the functional diversity of these kinases despite common upstream cellular signals. Although a number of studies have now demonstrated a role for PDK1 in the phosphorylation of the activation loop, we are only beginning to understand how the phosphorylation of the C-terminal tail of the AGC kinases can be regulated by TOR and other protein kinases. Although T-loop phosphorylation is mandatory for kinase activity, C-terminal tail phosphorylation imparts another level of regulation to increase kinase activity, protein stability, control substrate specificity and protein–protein interaction, and other functions that remain to be discovered. How can TOR promote phosphorylation at the C-terminal tail? The emerging picture is that growth signals can regulate the AGC kinases such that by a change in conformation and/or localization, the activation loop and C-terminal tail of most of these kinases become accessible for phosphorylation by PDK1 and by TOR (or TOR-regulated kinase) respectively. The signal allowing changes in conformation and/or localization is variable among AGC kinases, for example PI3K-generated phosphoinositides, membrane localization via a lipid-binding motif or relief from autoinhibition. Whether these signals can also regulate TOR or its complex partners remains to be elucidated. But, given the variability of these signals, it seems more reasonable to speculate that these signals do not regulate TOR kinase activity themselves. The recent discovery of other TORC partner proteins in both yeast and mammals indicates that there are other non-conserved organism-specific proteins that couple TORC signalling to specific cellular pathways and most probably in different cellular compartments, suggesting the presence of several distinct complexes. Thus further identification of TORC components that can specifically mediate phosphorylation of AGC kinases would provide answers as to how TOR can be involved in the phosphorylation of several AGC kinases and in promoting multiple cellular functions. Another outstanding question is why multiple kinases, including TOR, can phosphorylate the HM site, particularly in mammals. It would be of interest to determine in which context TOR can mediate phosphorylation of this site in several AGC kinases and how HM kinases may differentially affect the AGC kinase activity. It is also intriguing why the TORCs are essential for the phosphorylation of specific AGC kinases despite the high conservation of the TM and HM regions. What is the basis of this specificity? Lastly, why are there two distinct TOR core complexes and what is the reason for the difference in rapamycin sensitivities between the complexes? Comparative studies on how the TOR complexes can regulate the AGC kinases in different organisms should yield important answers to this question.
We would like to thank Christian Ungermann, Hans-Peter Schmitz and Alexandra Newton for helpful discussions and critical reading of the manuscript before its submission. This work was funded in part by grants from the American Heart Association, New Jersey Commission on Cancer Research, and American Cancer Society (to E. J.) and by Deutsche Forschungsgemeinschaft Grant SFB 431/P19 (to A. L.).
protein kinase A/protein kinase G/protein kinase C-family
atypical protein kinase C
Ca2+/calmodulin-dependent protein kinase
conventional protein kinase C
eukaryotic initiation factor 4E-binding protein
eukaryotic initiation factor 3
eukaryotic protein kinase
FRAP/TOR, ATM, TRRAP
FRAP/TOR, ATM, TRRAP C-terminal
G1-arrest defective 8
high osmolarity glycerol response 1
human vacuolar protein sorting 34
insulin-like growth factor 1
kontroller of growth 1
kinase responsible for sporulation and growth
(mammalian) lethal with sec thirteen
mitogen-activated protein kinase
murine embryonic fibroblast
mitogen- and stress-activated kinase 1/2
MAPK-activated protein kinase 2
mammalian target of rapamycin
novel protein kinase C
PH domain leucine-rich repeat protein phosphatase
protein kinase A (cAMP-dependent protein kinase)
protein kinase C
PKB-activating kinase homologue
proline-rich Akt substrate of 40 kDa
protein observed with rictor
proline-rich protein 5
regulatory-associated protein of mTOR
rapamycin-insensitive companion of mTOR
regulator of IME2
p90 ribosomal S6K
suppressor of loss of cAMP-dependent protein kinase 1/2
serum- and glucocorticoid inducible kinase
synthetic lethal with MSS4 1/2
type 2A phosphatase-associated protein 42
TOR complex one 89 kDa subunit
target of rapamycin
TOR complex 1/2
yeast protein kinase
Note on nomenclature: for the sake of clarity, we did not conform to the standard nomenclature used by the yeast field to name genes and proteins. Budding yeast gene/protein names are all capitalized, whereas fission yeast gene/protein names are only initially capitalized.