The activation of the AGC (protein kinase A/protein kinase G/protein kinase C)-family kinase SGK1 (serum- and glucocorticoid-induced kinase 1) by insulin via PI3K (phosphoinositide 3-kinase) signalling has been appreciated for almost 10 years. PDK1 (phosphoinositide-dependent protein kinase 1), a kinase that phosphorylates the SGK1 catalytic domain at Thr256, is known to play a critical role in SGK1 activation. However, the identity of the protein kinase(s) responsible for phosphorylation of Ser422, a site outside the catalytic domain (the so-called hydrophobic motif, or HM) that promotes activation of the kinase by PDK1, was unclear. In work reported in this issue of the Biochemical Journal, García-Martínez and Alessi have revealed the identity of a ‘PDK2’ kinase that catalyses Ser422 phosphorylation as mTORC2 (mammalian target of rapamycin complex 2), a multiprotein kinase that phosphorylates a similar site in PKB (protein kinase B).
How does a cell distinguish very similar cues present in its environment and respond appropriately? Since the consequence of misreading such a cue could adversely affect the whole organism, highly elaborate mechanisms have evolved that impart signalling selectivity. We now know that signal propagation involves not only very intricate spatial control of where signals are generated and temporal control of their duration, but also mechanisms that impart selectivity in the relay of signalling information. Post-translational modifications, particularly reversible protein phosphorylation, endow selectivity by transmitting the signalling information into altered function of only particular downstream components. In this issue of the Biochemical Journal, García-Martínez and Alessi investigate this latter issue in studying the regulation of the kinase SGK1 (serum- and glucocorticoid-induced kinase 1) . Using a combination of genetic and biochemical approaches, the authors show that SGK1, which is related to the S6Ks (ribosomal S6 kinases) and PKB/Akt (protein kinase B), is a direct substrate of the mTOR (mammalian target of rapamycin) protein kinase, but only when present in one of the two known mTOR-containing complexes, mTORC2.
Previous work indicated that, in metazoans, the mTOR kinase exists in at least two distinct multiprotein complexes. These two complexes can be distinguished both on the basis of specific associated proteins and by selective inhibition by the immunophilin rapamycin. The first complex, mTORC1, is a direct target of rapamycin (in association with the cellular protein, FKBP12), and contains the substrate-binding scaffold protein raptor, a small WD40-repeat protein, mLst8/GβL and PRAS40. The second mTOR complex, mTORC2, also contains mLst8, but is not inhibited by rapamycin/FKBP12, and instead of raptor contains a distinct scaffold protein, rictor, and two other proteins, Sin1 and protor.
The mTORC1 complex phosphorylates S6K1 at Thr389, a site C-terminal to the catalytic domain [termed the HM (hydrophobic motif)], thereby rendering S6K1 activity acutely sensitive to inhibition by rapamycin. Rapamycin is a highly selective inhibitor of mTORC1 but, despite its widespread use, it is currently unclear how it perturbs mTORC1 activity. Aside from the common presence of mTOR and raptor, it is also not known whether mTORC1 complexes are homogeneous with respect to their composition and are inhibited equivalently by rapamycin/FKBP12. The mTORC2 complex is even less well understood, although, like mTORC1, it also appears to be activated by insulin, and may exist in different subcomplexes, each containing rictor, but with different isoforms of Sin1 . Previous data have indicated one clear functional difference though between raptor–mTORC1 and rictor–mTORC2 kinases: unlike mTORC1, which phosphorylates S6K, mTORC2 acts as an HM kinase for PKB/Akt isoforms . Given the similarity at their C-termini of SGK1 and PKB, SGK1 would appear to be a likely candidate also for HM phosphorylation by mTORC2. So does mTOR phosphorylate SGK1, and, if so, is it mTORC1, mTORC2 or both?
Like S6K and PKB, SGK1 belongs to the AGC (protein kinase A/protein kinase G/protein kinase C) kinase family. However, SGK1 is probably one of the least understood members of this family. Three related kinases, encoded by distinct genes, are present in mammals (SGK1, 2 and 3) and, as suggested by the name, expression of the prototypical member SGK1 is induced by a variety of stimuli, including serum and glucocorticoids . However, the physiological functions revealed by deletion of SGK1 in mice appear to be restricted to a defect in renal retention of sodium ions, a phenotype that may be due to effects upon the ENaC (epithelial Na+ channel) . Fortunately, more evidence is available in considering how SGK1 is regulated. Earlier studies indicated that, like S6K1 and PKB, mammalian SGK is activated by insulin and IGF-1 (insulin-like growth factor 1) and suggested an important role for PDK1 (phosphoinositide-dependent protein kinase 1) and PI3K (phosphoinositide 3-kinase) signalling in kinase activation [6,7]. Interestingly, SGK1 activation was found to be refractory to inhibition of mTORC1 by rapamycin , suggesting that its activation mechanism might be more similar to that of PKB than S6K. In line with this idea, the likely budding yeast orthologue of SGK, Ypk2, plays a role in a TOR2-unique function in regulating the yeast cytoskeleton , a function that may be an analogous to that performed by mammalian mTORC2.
It was surprising then when earlier this year another group reported that SGK1 was phosphorylated by mTORC1 . The authors found that phosphorylation of SGK1 by mTOR was sensitive to rapamycin, and that stimuli that selectively activate mTORC1–S6K1 (such as nutrient amino acids) also promote SGK1 HM phosphorylation at Ser422, indicating that the relevant activating kinase for SGK1 was mTORC1.
In the new work reported by García-Martínez and Alessi , strong evidence is presented that supports a different conclusion: that mTOR present in mTORC2 is SGK's ‘PDK2’. Like other AGC kinases, phosphorylation of SGK1 at Ser422 by mTORC2 is likely to act as a docking site allowing PDK1 to bind to the kinase, facilitating phosphorylation of a second critical site (Thr256) in the activation loop of the catalytic domain, thereby allowing full activation of SGK1.
To investigate the issue of which mTOR complex regulates SGK1, the authors took advantage of cells deficient in components of mTORC2, rictor and Sin1 and mLst8. Interestingly, although present in both mTORC1 and mTORC2, mLst8 now appears to play a role only in mTORC2 activity . They found that, in rictor-deficient MEFs (mouse embryonic fibroblasts), SGK1 was essentially inactive and that, in rictor-, Sin1- and mLst8-deficient cells, SGK1 was not phosphorylated at Ser422. Raptor-deficient MEFs cannot be obtained because of an important role for raptor in early embryonic development , so it was not possible to genetically assess the role of mTORC1 and thus to rule out its contribution to SGK1 activity. Nonetheless, using a pharmacological approach, the authors found that rapamycin, although completely blocking S6K Thr389 phosphorylation, had no effect either on SGK1 activity in vitro, phosphorylation at Ser422 or phosphorylation of the SGK1 substrate NDRG1 (N-myc downstream regulated gene 1) in cells. Similarly, immunopurified mTORC2 stimulated phosphorylation of SGK1 at Ser422 butnot S6K1 Thr389in vitro, whereas mTORC1 could phosphorylate S6K1 at Thr389, but could not phosphorylate SGK at Ser422.
As with other important advances, new questions are raised by this study. Both PDK1 and mTORC2 appear to be critical for full activation of SGK1, but even in the absence of mTORC2, addition of PDK1 in vitro still causes substantial SGK1 activation, even if the Ser422 site is mutated to a non-phosphorylatable alanine residue . Since phosphorylation of PKB at Thr308 in its catalytic domain appears to be unaffected by deficiency in mTORC2 , PDK1 should still be fully active in mTORC2-deficient cells, but nonetheless SGK1 activity is negligible. The probable explanation here is that the ability of PDK1 to access and phosphorylate SGK1 is severely compromised in cells, but not in vitro. This then raises the related questions of how mTORC2 and PDK2 co-operate to phosphorylate the two important sites in SGK1, and where in the cell these events occur. PDK1 appears to be constitutively active in the cytoplasm, but probably phosphorylates PKB via transient recruitment of the active kinase to the location of Ser473-phosphorylated PKB at the plasma membrane, via a PH (pleckstrin homology) domain that binds PtdIns(3,4,5)P3. If Ser422 phosphorylation precedes activation by PDK1, one wonders whether activation of SGK1 by mTORC2 employs a similar mechanism of bringing together an already-active mTORC2 kinase to the site of its substrate. The fact that mTORC2 activity does appear to be activated by insulin, at least in some Sin1-containing isoforms of mTORC2, appears to argue against this possibility, although since one mTORC2 complex containing the mSin1.5 isoform is active even in the absence of insulin stimulation , it remains formally possible.
If, then, the mTORC2 kinase is activated first to phosphorylate Ser422, followed by SGK1 activation by PDK1, where do these events occur in the cell? The most obvious possibility is that both events occur at a membrane, thereby allowing a very rapid activation of SGK1 by presenting both activating kinases and the substrate on a surface. Alternatively, if one or both of the phosphorylation events occurred in the cytoplasm, a slower activation of SGK1 might be anticipated. Unfortunately, at this point it is unclear where in the cell the active mTORC2 complex resides. However, previous results have indicated that the PKB Ser473 kinase was probably present at the plasma membrane . Thus, if this kinase is indeed the same mTORC2 that phosphorylates SGK1, the reported slower activation kinetics of SGK1 compared with PKB  might be explained by the phosphorylation of Ser422 occurring (like PKB Ser473) at the plasma membrane, with the PDK1 phosphorylation occurring in the cytoplasm and limited by diffusion. In accordance with this idea, mutation of the PH domain of PDK1 in the livers of knock-in mice, a manipulation that should prevent PDK1 recruitment to the plasma membrane, does not prevent SGK1 activation by insulin , suggesting that both mTORC2 and PDK1 phosphorylation of SGK1 are intact and that PDK1 can activate SGK1 in the cytoplasm.
Since the activation of SGK1 by PDK1 occurs only after phosphorylation by mTORC2, a critical issue now appears to be whether mTORC2 is itself activated by stimuli to stimulate downstream kinases like PKB and SGK1, and if so how this occurs, or whether another mechanism operates that brings mTORC2 together with its downstream substrates. The ability to monitor the localization of mTOR complexes will clearly help to exclude this latter possibility, but the former will need a more complete understanding of the regulation of the activity of mTORC2. This is currently an area of intense activity in the field, with several laboratories seeking to identify stable post-translational changes such as phosphorylation in mTOR or its associated proteins, and attempting to establish their functional significance; clearly, a case then of ‘watch this space’.
We thank David Sabatini (Whitehead Institute, MIT, Cambridge, MA, U.S.A.) for helpful discussion.
Work in the authors' laboratory in funded by Cancer Research UK and the Tuberous Sclerosis Association.