The target of rapamycin (TOR), as part of the rapamycin‐sensitive TOR complex 1 (TORC1), regulates various aspects of protein synthesis. Whether TOR functions in this process as part of TORC2 remains to be elucidated. Here, we demonstrate that mTOR, SIN1 and rictor, components of mammalian (m)TORC2, are required for phosphorylation of Akt and conventional protein kinase C (PKC) at the turn motif (TM) site. This TORC2 function is growth factor independent and conserved from yeast to mammals. TM site phosphorylation facilitates carboxyl‐terminal folding and stabilizes newly synthesized Akt and PKC by interacting with conserved basic residues in the kinase domain. Without TM site phosphorylation, Akt becomes protected by the molecular chaperone Hsp90 from ubiquitination‐mediated proteasome degradation. Finally, we demonstrate that mTORC2 independently controls the Akt TM and HM sites in vivo and can directly phosphorylate both sites in vitro. Our studies uncover a novel function of the TOR pathway in regulating protein folding and stability, processes that are most likely linked to the functions of TOR in protein synthesis.
The target of rapamycin (TOR) is an evolutionarily conserved Ser/Thr protein kinase that forms two distinct protein complexes that regulate cell growth in response to a variety of environmental cues (Wullschleger et al, 2006). In mammals, in addition to mammalian TOR (mTOR), the conserved components of these complexes consist of mLST8 and raptor for mTOR complex 1 (mTORC1), and mLST8, rictor and SIN1 for mTORC2. In the presence of growth‐promoting signals such as nutrients and growth factors, mTOR controls growth by upregulation of protein synthesis (Mamane et al, 2006; Wullschleger et al, 2006). TORC1 regulates translation initiation by phosphorylating the translational regulator and AGC (protein kinases A/G/C) kinase family member, S6K (yeast SCH9), at the conserved hydrophobic motif (HM) site (Mamane et al, 2006; Jacinto and Lorberg, 2008). Mammalian and yeast TOR also regulates the biogenesis of ribosomes, which constitute the basic protein synthesis machinery (Martin et al, 2006). The highly conserved function of TOR in protein synthesis, primarily attributed to TORC1, is inhibited by the immunosuppressive and potential anti‐cancer drug rapamycin. The function of TORC2 is less well‐defined, but in both yeast and mammals, TORC2 is involved in actin cytoskeleton reorganization and cell survival (Loewith et al, 2002; Jacinto et al, 2004, 2006; Sarbassov et al, 2004). These TORC2 functions may also be linked to the regulation of other AGC kinases (Sarbassov et al, 2004; Kamada et al, 2005; Jacinto et al, 2006). Mammalian (m)TORC2 is required for phosphorylation of the HM site of the AGC kinases, Akt and protein kinase C‐α (PKCα) (Hresko and Mueckler, 2005; Sarbassov et al, 2005; Jacinto et al, 2006; Shiota et al, 2006; Yang et al, 2006). Phosphorylation of the HM site increases their catalytic activity (Newton, 2003) and may also affect kinase interaction with substrates (Holz et al, 2005; Jacinto et al, 2006). The HM is part of the carboxyl‐terminal (C) tail of AGC kinases. This kinase family is characterized by the presence of a unique interaction between the C‐tail and the catalytic domain to further regulate kinase activity (Kannan et al, 2007). The mammalian AGC kinases become optimally activated by phosphorylation at the activation loop (A‐loop) in the catalytic domain by PDK1 and at the HM site of the carboxyl‐terminus (Newton, 2003). Accumulating evidence indicates that mTOR, either as part of mTORC1 or mTORC2 mediates phosphorylation of several AGC kinases at the HM site (Jacinto and Lorberg, 2008). HM phosphorylation of S6K and Akt is mediated by mTORC1 and mTORC2, respectively, and occurs in response to growth stimuli (Hara et al, 1998; Hresko and Mueckler, 2005; Sarbassov et al, 2005; Jacinto et al, 2006). HM phosphorylation of PKCα and the yeast YPK1/2, both shown to be mediated by TORC2, does not seem to be responsive to growth stimuli, however (Kamada et al, 2005; Guertin et al, 2006).
Another highly conserved region in the carboxyl‐terminus of AGC kinases is the turn motif (TM). Most AGC kinases are phosphorylated at a conserved residue in this motif (Hauge et al, 2007). Structural and in vitro studies done on PKC and PKA suggest that phosphorylation at the TM site stabilizes the kinase core (Bornancin and Parker, 1996; Yonemoto et al, 1997). Whether this is also true for Akt is unclear (Alessi et al, 1996; Bellacosa et al, 1998; Toker and Newton, 2000). The nature of the TM kinase and the biological significance of this phosphorylation are poorly understood. Here, we address the importance of TM phosphorylation in vivo. We demonstrate that phosphorylation of Akt and conventional (c)PKC TM site is mediated by mTORC2 and is crucial for proper carboxyl‐terminal folding and stability of Akt and cPKC. In the absence of TM site phosphorylation, undegraded Akt and cPKC depend on Hsp90 for stability. Our results suggest that mTORC2 controls both the growth factor‐dependent HM phosphorylation and the growth factor‐independent TM phosphorylation of Akt. Our findings reveal a novel function for mTORC2 in protein folding that may be closely linked to the conserved function of mTORC1 in protein synthesis.
The TM phosphorylation in Akt and PKCα/β is a novel SIN1‐controlled, growth factor‐independent event
We recently found that in SIN1‐deficient murine embryonic fibroblasts (MEFS) in which mTORC2 is disrupted, Akt remains active and the A‐loop site (Thr308) is phosphorylated normally in a growth factor‐dependent manner (Jacinto et al, 2006). However, the growth factor‐induced HM phosphorylation of Akt at Ser473 was abolished and this defect correlated with decreased cell survival upon stress induction (Jacinto et al, 2006). Careful examination of Akt from SIN1−/− cells indicated an additional post‐translational modification that could not be explained by the lack of Akt HM site phosphorylation. Akt from SIN1−/− cells had faster electrophoretic mobility than in wild‐type cells even without A‐loop and HM phosphorylation, suggesting that this mobility shift is not due to their absence (Figure 1A). This mobility shift was not altered using different growth conditions. To determine if this modification is phosphorylation, wild‐type cellular extracts were phosphatase‐treated. To exclude phosphorylation of the A‐loop and HM sites that normally occurs in the presence of serum, extracts from starved cells were used. As a control, cellular extracts from SIN1−/− cells were also phosphatase‐treated. In the absence of Thr308 and Ser473 phosphorylation, phosphatase treatment led to increased Akt mobility in wild‐type cells that paralleled the migration pattern in SIN1−/− cells (Figure 1B). Thus, defective phosphorylation accounts for the increased Akt mobility in SIN1−/− cells.
In addition to the A‐loop and HM site phosphorylation, Akt is phosphorylated at other sites, although their significance is less well understood (Alessi et al, 1996; Bellacosa et al, 1998). By mass spectrometry, we found that phosphorylation of Ser124, a reported constitutively phosphorylated site in Akt (Alessi et al, 1996), was not abolished in SIN1−/− cells (data not shown). We therefore examined the TM site, Thr450, which was also shown to be constitutively phosphorylated (Alessi et al, 1996; Bellacosa et al, 1998) and highly conserved from yeast to mammals (Figure 1C) (Newton, 2003; Jacinto and Lorberg, 2008).
We examined the phosphorylation of the TM site in Akt and cPKC, which display very similar sequence around the TM. Indeed, we found that TM site phosphorylation of Akt and PKCα/βII are abolished in SIN1−/− cells (Figure 1D and Supplementary Figure 1). The phospho‐PKCα/βII TM (Ser638/641) antibody also recognized a 60 kDa protein, which we confirmed to be Akt (Supplementary Figure 1). Unlike Akt HM phosphorylation, which is induced upon insulin stimulation, TM phosphorylation is present even under starved conditions. Upon re‐expression of SIN1, the phosphorylation of the TM sites of both Akt and PKCα/βII in SIN1−/− MEFs was restored (Figure 1D). These results reveal that SIN1 is essential for both Akt and PKCα/βII TM site phosphorylation and provide the first genetic evidence for a mechanism that controls TM site phosphorylation.
The TM phosphorylation of Akt and PKCα/β requires mTORC2 components
Turn motif phosphorylation of Akt and PKCα/βII is insensitive to various growth factors and stress inducers such as tunicamycin, heat shock, etoposide, peroxide, UV and gamma irradiation (data not shown). Pharmacological inhibition of the PI3K pathway (by wortmannin) or mTORC1 (by rapamycin) up to 8 h also had no significant effect on the TM site phosphorylation (Figure 2A and data not shown). Likewise, the PI3K inhibitor Ly294002 and MAPK inhibitors U0126, SB203580, SP600125, CDK inhibitor roscovitine, cyclosporin A and FK506 had no obvious effect on TM phosphorylation (data not shown). Taken together, these results suggest that the TM phosphorylation is stable and insensitive to transient inhibition of a number of kinases including PI3K, MAPKs, CDKs and mTORC1. The TM site phosphorylation of novel PKCδ/θ and PKA was not abolished in SIN1−/− cells (Figure 2A and data not shown); it is noteworthy that these kinases do not have the consensus Thr‐Pro‐sequence, which is found in Akt (Drosophila and mammals), mammalian cPKC and yeast PKC1 (Figure 1C) in their TM. These results suggest that the SIN1‐mediated TM site phosphorylation is specific to a subset of the AGC family kinases.
To further determine if TM phosphorylation is mTORC2 dependent, we examined if rictor also mediates the TM phosphorylation of Akt and PKCα/βII. Using rictor−/− cells, essentially the same defects in Akt and PKCα/βII TM phosphorylation were observed similar to SIN1−/− cells (Figure 2B). TM phosphorylation of novel PKCδ/θ and PKA was also not affected in rictor−/− cells (Figure 2B and data not shown), further suggesting that mTORC2 components specifically mediate TM phosphorylation of Akt and cPKC.
Next, we silenced mTOR expression with siRNA and found that knockdown of mTOR decreased the TM site phosphorylation of both Akt and PKCα/βII (Figure 2C). As it was reported that prolonged rapamycin treatment could inhibit the assembly of mTORC2 in many cell types (Sarbassov et al, 2006), we examined if TM phosphorylation is affected under this condition. Treatment of MEFs, PC3 and Jurkat cells with rapamycin for 24 h or longer, which dissociated mTOR binding from SIN1, caused a dramatic decrease in TM phosphorylation of Akt and PKCα/βII (Figure 2D and Supplementary Figure 2). The faster Akt migration in SDS–PAGE, which accompanies the loss of TM phosphorylation, is also evident upon prolonged rapamycin treatment (Figure 2D). HM phosphorylation of Akt was also diminished under these conditions. As previously reported for the HM site (Sarbassov et al, 2006), the TM phosphorylation of Akt in HEK293T and HeLa cells was also resistant to the prolonged rapamycin treatment most likely due to the remaining intact mTORC2 found in these cells (Supplementary Figure 2). Taken together, these results further indicate that mTORC2 is required for TM phosphorylation.
The TM phosphorylation mediated by TORC2 is highly conserved
As Akt and PKC have orthologues in lower organisms, we examined if the TORC2‐dependent TM site phosphorylation is conserved. We conditionally depleted the SIN1 orthologue AVO1 in Saccharomyces cerevisiae (Loewith et al, 2002). TM phosphorylation of PKC1 was significantly decreased upon depletion of AVO1, but not of KOG1, a TORC1 component (Supplementary Figure 3A). These results indicate that similar to mammalian PKCα/βII, the TM phosphorylation of yeast PKC1 requires TORC2. To further address the conserved function of TORC2 in Akt and PKC TM site phosphorylation, we also knocked down Drosophila SIN1 (dSIN1) by siRNA. Knockdown of dSIN1, but not dRaptor, a key component of dTORC1, decreased the phosphorylation of dAkt at Thr484, the equivalent TM site in dAkt (Supplementary Figure 3B), further demonstrating that the TORC2‐dependent TM phosphorylation of dAkt is also conserved in Drosophila. Collectively, these data demonstrate that SIN1‐mediated TM phosphorylation is conserved in yeast, flies and mammals.
Protein levels of Akt and PKCα are diminished in the absence of SIN1 or rictor
Previous studies indicated that TM site phosphorylation of cPKC is important for protein maturation, and in its absence, cPKCα was unstable in vitro (Bornancin and Parker, 1996; Edwards et al, 1999). As the mTOR function in protein synthesis is evolutionarily conserved, we postulated that mTORC2 may regulate events that are closely linked to protein synthesis, such as folding and stabilization of newly synthesized proteins. Since improperly folded proteins are known to be unstable and prone to degradation (Hartl and Hayer‐Hartl, 2002), we determined the protein levels of Akt and PKCα in several independently generated MEF cell lines from either wild‐type or from the SIN1−/− littermate embryos. To appreciate subtle changes in protein expression, we loaded less but equal amounts of lysates from wild‐type and SIN1−/− cells. Compared with the wild‐type cell lines where the levels of Akt and PKCα were highly uniform, the levels of Akt and PKCα in all the SIN1−/− cell lines varied considerably but were consistently reduced compared with wild‐type cells (Figure 3A). A significant decrease in PKCα protein level was also observed in rictor−/− cells (Figure 2B) consistent with previous studies (Guertin et al, 2006), suggesting that the decreased Akt and cPKC protein level is an mTORC2‐related phenotype.
To further examine whether the absence of TM phosphorylation may lead to degradation, we determined if Akt in SIN1−/− cells could have increased ubiquitination, a mechanism that leads to proteasome turnover. Indeed, there was increased ubiquitination in SIN1−/− cells of exogenous Akt, which was more evident in the presence of the proteasome inhibitor MG132 (Figure 3B). Endogenous Akt ubiquitination also became more pronounced in SIN1−/− cells upon MG132 pretreatment (Figure 3C). Taken together, these results indicate that the lack of TM phosphorylation predisposes Akt towards degradation through ubiquitination and proteasome‐mediated turnover.
The maturation of PKCα is defective in the absence of SIN1
We next examined if maturation of cPKC, which was reported to require TM phosphorylation (Edwards et al, 1999), is defective in SIN1−/− cells. By pulse‐chase analysis, we found that, whereas newly synthesized PKCα from wild‐type cells underwent maturation as depicted by the appearance of a slower migrating band with a half‐life time of 10 min, PKCα maturation in SIN1−/− cells was severely impaired with a much slower kinetics (Figure 3D). The slower migrating form of PKCα corresponds to phosphorylated TM and HM sites, while the fast‐migrating species is unphosphorylated at the carboxyl‐terminus (Keranen et al, 1995). The TM phosphorylation of PKCα was also proposed to induce a closed conformation, such that in its absence, PKCα was rapidly degraded in vitro (Bornancin and Parker, 1996). To further investigate how lack of TM phosphorylation renders Akt and PKCα unstable in vivo, we determined the half‐life of Akt and PKCα in wild‐type and SIN1−/− cells. Surprisingly, we found that upon inhibition of translation using cycloheximide, Akt and PKCα in either SIN1−/− or rictor−/− cells had a half‐life comparable with wild‐type cells (Supplementary Figure 4). These results led us to speculate that some of the TM site unphosphorylated Akt and PKCα may elude the proteasome pathway and are protected in cells under normal growing conditions.
Akt and PKCα in SIN1−/− cells are protected by Hsp90
As heat shock proteins such as Hsp90 can protect newly synthesized and/or improperly folded kinases (Caplan et al, 2007), we examined if disrupting Hsp90 function would increase susceptibility of Akt and PKCα to proteasome degradation in SIN1−/− cells. We used three different strategies to disrupt Hsp90 function. First, we used a well‐characterized pharmacological inhibitor of Hsp90, 17‐(Allylamino)‐17‐demethoxygeldanamycin (17‐AAG). Although the 17‐AAG treatment of wild‐type cells had no significant effect on Akt and PKCα levels for up to 8 h, there was a dramatic decrease in Akt and PKCα levels in SIN1−/− and rictor−/− cells as early as 2 h following treatment (Figure 4A and 4B; data not shown). Akt and PKCα protein levels were rescued upon expression of HA–SIN1 in SIN1−/− MEFs, indicating that the 17‐AAG‐induced reduction is specific to the absence of SIN1 (Figure 4A). Second, we used a cell‐permeable peptide inhibitor of Hsp90 that could disrupt Hsp90 interaction with Akt (Miao et al, 2008). Third, we used a dominant‐negative Hsp90 mutant to inhibit the Hsp90 pathway (Miao et al, 2008). The inhibitory peptide or expression of the dominant‐negative Hsp90 both significantly decreased Akt levels in the SIN1−/− MEFs (Supplementary Figure 5). As Akt degradation is mediated through the ubiquitin pathway (Figure 3B and C), we would anticipate that Hsp90 inhibition should further augment Akt ubiquitination in SIN1−/− cells. Indeed, Akt ubiquitination was significantly increased in SIN1−/− cells upon 17‐AAG treatment (Figure 4C). Taken together, these results demonstrate that Akt and PKCα from SIN1−/− cells, which lack TM phosphorylation, are highly dependent on Hsp90 for stability and to prevent proteasome‐mediated degradation.
To further address how the absence of TM phosphorylation may affect Akt and PKCα stability, we induced protein destabilization and misfolding by heat shock treatment. We reasoned that under sustained heat stress, the Hsp90 pathway becomes overwhelmed and thus may lose its ability to rescue Akt/cPKC in SIN1−/− cells. Heat shock attenuated Akt and PKCα levels significantly in SIN1−/− cells but only marginally in wild‐type cells (Figure 4D). We also observed that both the endogenous Akt and exogenous HA–Akt bound more Hsp90 upon heat shock in SIN1−/− cells (Supplementary Figure 6). Despite increased Hsp90 binding, endogenous Akt underwent degradation when heat shock was sustained (Supplementary Figure 6). These results further suggest that the TM site phosphorylation‐defective Akt that escaped proteasome degradation was protected by Hsp90 in SIN1−/− cells under normal growing conditions.
We then asked what would happen to protein levels of Akt and PKCα in wild‐type cells if we inhibit both mTOR and Hsp90. Combined 17‐AAG and prolonged rapamycin treatment led to a considerable decrease in Akt protein level at longer time points (Figure 4E). Remarkably, combined treatment with 17‐AAG, rapamycin and knockdown of mTOR expression diminished Akt expression to similar levels seen in 17‐AAG‐treated SIN1−/− cells (Figure 4E). The expression of PKCα was also significantly decreased by combined Hsp90 and mTOR inhibition/knockdown. These results suggest that when mTORC2‐mediated TM phosphorylation is disrupted, Hsp90 becomes essential for Akt and cPKC stability.
The TM phosphorylation is specifically required for Akt and PKCβII protein stability
To further investigate if it is the specific lack of TM phosphorylation in Akt that caused instability of Akt, we determined how mutation of Thr450 to Ala could affect Akt stability in wild‐type cells. Expression of Thr450Ala Akt in wild‐type MEFs phenocopied the wild‐type Akt in SIN1−/− cells by displaying an augmented susceptibility to the 17‐AAG‐induced degradation (Figure 5A). In sharp contrast, wild‐type Akt or Ser473Ala mutant Akt remained stable under the same conditions. Consistently, the TM site mutants were highly ubiquitinated in wild‐type cells resembling that of Akt in the SIN1−/− cells (Figure 5B). Furthermore, the 17‐AAG‐induced and ubiquitin‐mediated Akt degradation could be inhibited by the proteasome inhibitor MG132 (Supplementary Figure 7). As compensating phosphorylation by adjacent Ser/Thr sites was reported to occur in the absence of TM site phosphorylation in PKCβII (Edwards et al, 1999), we also examined the effect of combined mutation of Thr450 with the adjacent Thr443 to alanine in Akt. Although total HA–Akt levels were comparable in the single versus double mutant (Figure 5A), we observed more pronounced ubiquitination of the double mutant (Figure 5B). Similar to Akt from SIN1−/− cells (Figure 5C), the TM site mutants also bound more Hsp90 than wild‐type Akt (Figure 5D). Thus, the lack of TM phosphorylation specifically renders Akt unstable and dependent on Hsp90 for protection.
We also examined if mutation of the TM site in PKCβII can confer instability upon inhibition of Hsp90. As adjacent Ser/Thr sites can compensate for the absence of phosphorylation at the TM site, we used a PKCβII TM site mutant that contains adjacent Ala mutations at the compensating sites (Edwards et al, 1999). Upon 17‐AAG treatment, protein levels of the TM mutant, but not the HM mutant, Ser660Ala, were decreased (Figure 5E). Taken together, these results indicate that phosphorylation at the TM is specifically required for protein stability and that Hsp90 can stabilize the unphosphorylated Akt and cPKC.
Complementary interactions between a conserved basic patch at the catalytic domain and the phospho‐threonine in the TM have an important function in the stabilization of Akt and cPKC
Why did the lack of TM site phosphorylation cause an instability specifically to Akt and cPKC in SIN1−/− and rictor−/− cells? Is there a structural basis for this TM site phosphorylation‐mediated Akt and cPKC stabilization? The TM region is localized in a mobile loop and undetected in the previous crystal structure of PKBβ (Akt2) (Yang et al, 2002). However, electrostatic surface potential calculation revealed that an area at the catalytic domain close to the missing TM loop (marked by a square in Figure 6A and B) contains intensive positive charges. This basic patch is mainly contributed by K165, R184, R224 (human Akt2 numbering). Interestingly, these residues are conserved in several AGC kinases including the conventional and atypical PKCs, but only partially conserved in PKA and novel PKCs (Figure 6C and data not shown).
Structural modelling of the TM loop with phosphorylated T451 (human Akt2 numbering, and equivalent to T450 in human Akt1 used throughout this study) showed that the negatively charged phosphate group forms strong electrostatic interactions with the conserved basic patch in the N‐lobe of the catalytic domain (Figure 6A and B). It is likely that either during or immediately following Akt translation, the newly synthesized Akt is phosphorylated at the TM site, which may tether the entire TM loop to the basic patch through a ‘knot’‐like interaction as described above (Figure 6B), and consequently stabilizes the C‐terminal tail. In the absence of TM phosphorylation, the C‐terminal tail might be more flexible, hence more accessible by the protein degradation machinery, as we have shown in the above analysis. We would therefore expect that abrogation of the interactions between the conserved basic patch and phosphorylated TM site should result in destabilization of Akt protein.
To test our model, we mutated these basic residues to methionine and determined the effect on Akt protein stability. We found that despite the presence of TM site phosphorylation (Supplementary Figure 8), the K163M, K182M or R222M mutants (human Akt1 numbering) expressed in wild‐type cells displayed a very similar stability defect as the Akt TM mutant (Figure 6D). In the presence of 17AAG, all these Akt mutants, but not wild‐type Akt or Akt S473A mutant, exhibited dramatic degradation in wild‐type cells comparable with that of wild‐type Akt in SIN1−/− cells (Figures 5A and 6D). Additionally, each of these mutants bound more Hsp90 compared with wild‐type Akt (Figure 6E). These results strongly suggest that the complementary charge–charge interactions between these conserved basic residues at the catalytic domain and the phospho‐threonine in the TM have an important function in the folding and stability of Akt, and most likely conventional PKC as well, which displays both sequence and structure conservations to Akt.
The Akt TM phosphorylation is controlled independently of its HM phosphorylation
Turn motif site phosphorylation occurs in starved cells when neither A‐loop nor HM sites are phosphorylated, suggesting that TM phosphorylation is not dependent on phosphorylation of either of these two sites. However, as phosphorylation of both the TM and HM sites is defective in the SIN1−/− cells, one possibility for abolished HM phosphorylation in stimulated cells could be due to the lack of TM phosphorylation. We therefore determined if the Akt HM phosphorylation could occur in wild‐type cells expressing TM site mutant Akt. In the Akt TM site mutants, HM phosphorylation was still induced by growth factors, similar to wild‐type Akt (Figure 7A). These results demonstrate that the TM and HM phosphorylation of Akt are two independent events.
A potential function of SIN1 is to bring Akt to mTOR or prolong their interaction to allow TM or HM phosphorylation, as SIN1 and Akt can interact (Jacinto et al, 2006). Akt activation also requires membrane translocation (Bellacosa et al, 1998). We then determined if constitutive membrane targeting of Akt may override the requirement for the mTORC2‐mediated HM and TM phosphorylation. We expressed a myristylated form of Akt (Myr‐Akt) and examined if it can rescue the TM or HM phosphorylation in SIN1−/− cells. Interestingly, we found that membrane targeting of Akt in SIN1−/− cells is clearly unable to restore TM phosphorylation but could partially rescue HM phosphorylation (Figure 7B). As mTORC2 assembly is disrupted in SIN1−/− cells, we asked whether autophosphorylation may account for the observed HM phosphorylation of Myr‐Akt in SIN1−/− cells (Toker and Newton, 2000). Upon expression of a kinase inactive allele of Akt (Myr‐Akt‐KD) in SIN1−/− cells, very little HM phosphorylation was observed, indicating that in the absence of SIN1, the HM site of Myr‐Akt undergoes autophosphorylation (Figure 7B). This residual autophosphorylation is resistant to prolonged rapamycin treatment (Figure 7C). In contrast, the majority of HM phosphorylation of both wild‐type Akt and Myr‐Akt can be inhibited by prolonged rapamycin treatment in wild‐type cells (Figure 7C). As expected, non‐myristylated Akt‐KD mutant expressed in SIN1−/− cells displayed no phosphorylation at the TM and HM sites (data not shown). These results confirm that HM phosphorylation mainly occurs through mTORC2, although residual autophosphorylation could occur when Akt is artificially targeted to the membrane.
mTORC2 phosphorylates the TM and HM sites of Akt in vitro
Finally, we addressed if mTORC2 can phosphorylate the TM site in vitro. We used dephosphorylated GST–Akt1 from HEK293 cells (see Materials and Methods) as the substrate for in vitro kinase assays. We immunoprecipitated mTORC2 with an anti‐SIN1 antibody and used this as the kinase source. In the presence of SIN1 immunoprecipitates, TM phosphorylation of GST–Akt1 was observed (Figure 7D, lane 2). No TM phosphorylation was detected in the absence of either SIN1 immunoprecipitates (Figure 7D, lanes 1 and 4) or GST–Akt1 (lane 3). Confirming previous studies (Sarbassov et al, 2005; Yang et al, 2006), the HM site of Akt can be robustly phosphorylated by a SIN1‐associated kinase (Figure 7D). The HM site appears to be more efficiently phosphorylated than the TM site, suggesting that the assay conditions used are more favourable towards HM site phosphorylation. These results indicate that a SIN1‐associated kinase(s) can phosphorylate these two distinct Akt sites.
As mTOR, but not SIN1 or rictor, possesses kinase activity, we then analysed if mTOR can directly phosphorylate the Akt HM and TM sites. We compared the kinase activity of HA‐tagged mTOR with a kinase‐dead mutant of HA–mTOR (KD). Whereas both wild‐type and KD mTOR pulled down similar amounts of SIN1 and rictor, wild‐type mTOR, in contrast to KD mTOR, can more efficiently phosphorylate the HM and TM sites (Figure 7E). Taken together, these results provide convincing evidence that Akt TM and HM phosphorylation are mediated and likely undergo direct phosphorylation in vivo by mTORC2.
Early studies on PKC and PKA have suggested that phosphorylation at the TM is important for their stabilization (Bornancin and Parker, 1996; Yonemoto et al, 1997; Edwards et al, 1999; Messerschmidt et al, 2005; Grodsky et al, 2006). The regulatory mechanism and biological significance of the TM site phosphorylation in Akt and cPKC were previously only speculative. The function of this phosphorylation in Akt, one of the most widely studied AGC kinases due to its oncogenic potential, has been overlooked because of its apparently constitutive nature (Alessi et al, 1996; Bellacosa et al, 1998). We have now shown that phosphorylation of this site is strictly dependent on the integrity of mTORC2 and is essential for the folding and stability of Akt and cPKC. Without TM phosphorylation, Akt and cPKC are partially protected by Hsp90 for stability. Lack of TM phosphorylation of endogenous Akt and cPKC has not been observed in vivo before. Thus, our present study reveals a new and unique regulatory mechanism for a subset of the AGC family kinases by mTORC2 (see Akt regulation as modelled in Figure 8).
Using molecular remodelling and in vivo studies, we demonstrate that the interaction of the phosphate on the TM site with the three highly conserved basic residues in the catalytic region functions to fold and stabilize Akt (Figure 6). A similar structural model was recently proposed, but TM phosphorylation was suggested to synergize with HM phosphorylation to enhance catalytic activity of a number of AGC kinases (Hauge et al, 2007). This is in contrast to our findings that demonstrate an important function of TM phosphorylation primarily for protein stability. Our results concur with previous studies wherein TM site mutation of Akt and PKC did not affect their in vitro catalytic activity (Bornancin and Parker, 1996; Bellacosa et al, 1998; Toker and Newton, 2000) (data not shown). Further studies should reveal how regulation of Akt stability can affect its activity towards its numerous cellular targets.
The TM and HM phosphorylation of Akt clearly require mTORC2, but necessitate different temporal and spatial conditions. Whereas TM phosphorylation remains stable when growth stimuli are withdrawn, HM phosphorylation requires growth factors. Growth factor stimulation is critical for PI3K activation that leads to Akt recruitment to the membrane where it undergoes HM phosphorylation. Interestingly, artificial membrane targeting of Akt in mTORC2‐disrupted cells could partially rescue HM, but not TM, phosphorylation, highlighting the importance of Akt membrane localization for HM phosphorylation. This partial rescue of HM phosphorylation in SIN1−/− cells was not through mTOR but was actually due to autophosphorylation (Figure 7B and C). However, the observed mTORC2‐independent autophosphorylation in this case was minimal (Figure 7B), and it is unlikely that under physiological conditions it could bypass the requirement for mTORC2. In wild‐type cells, the majority of HM and TM phosphorylation of Myr‐Akt can be inhibited by prolonged rapamycin treatment, further indicating that HM and TM site phosphorylation are primarily mediated by mTORC2. This possibility is further supported by the evidence that mTOR (as part of mTORC2) could exert an effect as a direct kinase for the TM and HM sites (Figure 7D–E).
The consensus sequence surrounding the TM (Thr‐Pro‐Pro‐Asp) of Akt and cPKC is distinct from the HM (Phe‐Ser‐Tyr). The TM consensus sequence is also present in dAkt and partially conserved in the budding yeast PKC1. In addition to a function of budding yeast TORC2 in regulating PKC1 activity through the Rho GTPases, we show here that it may also mediate TM site phosphorylation (Supplementary Figure 3). TORC2 also controls TM phosphorylation of the S. cerevisiae AGC kinase YPK2 in vitro (Kamada et al, 2005) and the S. pombe Gad8p (Matsuo et al, 2003). Our study demonstrates that mTOR kinase activity is required for TM phosphorylation. Thus, mTOR has likely preserved its primitive, non‐growth factor‐inducible function in mediating TM phosphorylation from yeast to mammals. Although we could not rule out the existence of another Pro‐directed kinase for TM site phosphorylation, pharmacological inhibition of other known Pro‐directed kinases such as MAPKs and CDKs did not significantly inhibit TM phosphorylation (data not shown). Supporting mTOR as a kinase for a Thr‐Pro motif, the phosphorylation sites in the bona fide mTOR substrate, 4E‐BP, are also followed by a Pro (Gingras et al, 2001).
How can mTORC2 similarly mediate phosphorylation of the Akt TM and HM sites that do not respond to the same cellular stimulus? One model would be that the proper localization and/or conformation of Akt, and likely PKC as well, determines when it will be regulated or phosphorylated by mTORC2. A similar model was proposed for the PDK1 phosphorylation of the activation loop of several AGC kinases that are regulated by various stimuli (Biondi, 2004). If mTOR were the kinase for both sites, this would imply that mTOR intrinsic activity does not have to be regulated by these various signals. This also implies that mTORC2 may be distinctly regulated in different cellular compartments. Identification of specific components that may regulate mTORC2 in these compartments would provide a mechanism for this divergent regulation. The immediate appearance (visible at 3 min) of mature, phosphorylated PKC in our pulse‐chase experiment hints that the TM site phosphorylation may occur during or immediately following translation (Figure 3D). Therefore, although mTORC2‐mediated HM phosphorylation occurs post‐translationally as a result of Akt localization to the membrane, it is possible that mTORC2‐mediated TM site phosphorylation of Akt and cPKC may occur at the translational machinery.
Modulation of folding and stability of AGC kinases by mTOR and molecular chaperones is perhaps a fundamental mechanism for AGC kinase function. Hsp90 and Hsp70 can prolong the kinase activity of Akt and PKCβII, respectively (Sato et al, 2000; Gao and Newton, 2002). Whether TM site phosphorylation (Ser371) of the mTORC1‐regulated S6K controls folding and stability of this kinase would need to be evaluated, but this site is also mitogen‐induced and can be phosphorylated by mTOR in vitro (Moser et al, 1997; Saitoh et al, 2002). The co‐regulation of protein synthesis and folding is highly conserved among prokaryotes and eukaryotes and involves association of chaperones with the protein synthesis machinery (Hartl and Hayer‐Hartl, 2002). The TOR complexes may have evolved to co‐regulate the protein synthesis machinery and co/post‐translational folding of eukaryotic AGC kinases. As TM site phosphorylation may stabilize the structure of the majority of AGC kinases (Hauge et al, 2007), future studies should uncover the exact mechanism for the target specificity by mTORC2.
Finally, given that Akt, PKC, Hsp90 and mTOR are tapped as important drug targets, our findings provide rationale for how the control of protein stability of Akt and PKC by the mTOR pathway and molecular chaperones could be exploited to develop more specific and efficacious drugs against cancer and other growth‐related diseases. Combination therapy that target mTOR and the chaperone pathways should lead to more efficient growth inhibition of cancer cells.
Materials and methods
Plasmid constructs, antibodies, reagents
HA‐tagged Akt constructs were obtained from D Alessi (University of Dundee) (wt, S473A, Myr‐Akt) and A Toker (Harvard University) (T450A, T443A/T450A). Wild‐type or mutant HA‐Akt ORFs were each subcloned into the retroviral vector pMIGW at the BglII and ERI sites. GST–Akt was obtained from D Sarbassov (UTMDACC). HA‐Akt K to M mutants and Myr‐Akt‐KD (K179M) were constructed by using a PCR‐directed mutagenesis method. HA‐PKCβIIAAA mutant was obtained by subcloning PKCβIIAAA (Edwards et al, 1999) into pCI‐HA vector. pMIGW‐HA‐SIN1 and siRNA constructs were prepared as described previously (Jacinto et al, 2004, 2006). Myc‐ubiquitin was obtained from X Feng (Baylor). HA‐mTOR WT and KD were prepared as described previously (Dennis et al, 2001). All other reagents were obtained as described in Supplementary data.
Cell culture, stimulation and harvest
Murine embryonic fibroblasts were cultured, stimulated and harvested as described previously (Jacinto et al, 2006). Rictor−/− MEFs were obtained from KL Guan (UCSD). MEFs were infected with retrovirus expressing either HA–Akt or HA–SIN1 as described previously (Jacinto et al, 2006) and in Supplementary data.
Murine embryonic fibroblasts were starved for 30 min in DMEM lacking methionine and cysteine, then pulsed for 7 min with 35S‐methionine/cysteine labelling mix. Cells were chased with normal DMEM supplemented with 5 mM methionine and cysteine. PKCα was immunoprecipitated with PKCα monoclonal antibody from BD Transduction (San Jose, CA, USA), then fractionated by SDS–PAGE.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
We thank Chang‐Chih Wu, Drs Dario Alessi, Alex Toker, Xin‐Hua Feng, Guy Werlen and Hans‐Peter Schmitz for sharing plasmids and reagents; Dos Sarbassov for advise on mTORC kinase assays; Kun‐Liang Guan and Qian Yang for rictor−/− cells and Drosophila siRNA reagents; Terri Goss Kinzy, Anja Lorberg and Arthur Horwich for helpful discussions and critical reading of the manuscript. This work was supported in part by the American Heart Association, New Jersey Commission on Cancer Research and American Cancer Society RSG0721601TBE (to EJ), and by grants HL070225 (NIH) (to BS) and Nuclear Receptor Signaling Atlas (DK62434) (to JQ).
- Copyright © 2008 European Molecular Biology Organization