We investigated whether Akt inhibits autophagy by directly regulating the core autophagy machinery independently of mTOR. Expression of constitutively active myristoylated (5) and tagged Akt1 (Flag-tagged myr-Akt) in HeLa cells inhibited autophagy during growth in normal medium, in response to serum and amino acid starvation (a physiological inducer of autophagy), in response to treatment with an ATP-competitive inhibitor of mTOR, Torin1 (6), and in response to both starvation and Torin1 treatment (Fig. 1A and 1B). In all conditions, cells expressing myr-Akt1 had decreased numbers of puncta upon transfection with a fusion protein of green fluorescent protein with LC3 (GFP-LC3), a fluorescent marker of autophagosomes; increased amounts of p62 (a substrate that is degraded by autophagy); and increased amounts of the cytosolic non-lipidated form of LC3, LC3-I, and of total LC3 (7). Amounts of phospho-4E-BP1, a phosphorylation target of mTOR, were decreased in Torin1-treated cells, including those expressing myr-Akt1. Thus, myr-Akt1 suppresses basal autophagy, starvation-induced autophagy, and Torin1-induced autophagy, indicating that active Akt can inhibit autophagy through mTOR-independent mechanisms.

We examined whether autophagy execution proteins could be targets of Akt. We focused on Beclin 1 because of its role in autophagy and tumor suppression (4). Endogenous Akt co-immunoprecipitated with endogenous Beclin 1 in HeLa cells, and this interaction was weakened by starvation (Fig. 1C). In contrast, the interaction of myr-Akt1 with a Flag epitope-tagged construct of Beclin 1 was not affected by starvation (Fig. S1). Kinase prediction algorithms (8, 9) showed Beclin 1 to contain a motif (R-X-X-R-X-X-S²⁹⁵) that resembles the consensus Akt phosphoryation motif (R-X-R-X-XS/T) (10) and another sequence (R-X-X-S²³⁴) that corresponds to a 14-3-3 protein binding motif (which may be generated by Akt phosphorylation) (Fig. S2A). Phosphospecific antibodies against these two candidate phosphorylation sites in Beclin 1 (S²³⁴ and S²⁹⁵) recognized wild-type Flag-Beclin 1 expressed in HeLa cells and immunoreactivity was decreased with the corresponding Flag-Beclin 1 ala-nine substitution mutant (Fig. S2B). GST-Akt1 phosphorylated Beclin 1 S²⁹⁵ but not Beclin 1 S²³⁴ in vitro (Fig. S2C), and this was partially blocked by treatment with two Akt inhibitors, MK-2206 and Akt inhibitor X (Fig. 1D). Expression of active Akt1 (myr-Akt1) increased and expression of a catalytically inactive, non-phosphorylatable Akt1 mutant (K179M/T308A/S473A; DN-Akt1) decreased, respectively, phosphorylation of Flag-Beclin 1 S²⁹⁵ and Flag-Beclin 1 S²³⁴ in HeLa cells (Fig. S2D). Expression of myr-Akt1 also led to phosphorylation of endogenous Beclin 1 S²⁹⁵ and endogenous Beclin 1 S²³⁴ which was not reversed by mTOR inactivation with Torin1 (Fig. 1E). Endogenous Beclin 1 S²⁹⁵ phosphorylation increased when starved HeLa cells were fed with normal medium (Fig. S2E). Together, these studies demonstrate that Beclin 1 is phosphorylated by Akt on residue 295 (and possibly 234) in an mTOR-independent manner.

We compared the phosphorylation of Beclin 1 S²⁹⁵ in three paired sets of tumor cell lines with and without Akt activation (Fig. 1F). Melanoma cells with mutant PTEN (WM793) had more phosphorylation of Beclin 1 S²⁹⁵ than did those with wild-type PTEN (451Lu) (11). U87-MG glioblastoma cells with high Akt activity due to inactivating mutations in PTEN showed more phosphorylation of Beclin 1 S²⁹⁵ than did U87-MG cells in which wild-type PTEN was reintroduced (12). In breast carcinoma cells, S²⁹⁵ phosphorylation was detected in MCF10A-DCIS cells with an activating H1047R mutation in PIK3CA but not in MDA-MB231 cells lacking constitutive Akt activation (13). Thus, in three different tumor types, activation of Akt is associated with phosphorylation of Beclin 1 S²⁹⁵, indicating that phosphorylation of Beclin 1 S²⁹⁵ may be common in human tumors with activated Akt.

We transfected MCF7 human breast carcinoma cells [which express low amounts of endogenous Beclin 1 (14)] with GFP-LC3 and wild-type Beclin 1 or Beclin 1 S295A or AA mutants (Fig. S3A). Cells transfected with S295A and AA mutants had increased basal (but not starvation-induced) autophagy (Fig. S3B and C). Inhibition of Akt by MK-2206 (Fig. S3D) increased basal autophagy in MCF7 cells to a lesser extent in cells transfected with Beclin 1 AA than in cells transfected with wild-type Beclin 1 (Fig. S3E). Conversely, expression of active Akt decreased basal autophagy in all MCF7 cells, but cells expressing Beclin 1 AA showed more autophagy than cells expressing wild-type Beclin 1 or vector alone (Fig. S3F and S3G). Thus, Akt appears to inhibit basal autophagy both through Beclin 1 phosphorylation-dependent and independent mechanisms.

To examine the role of Akt-mediated Beclin 1 phosphorylation in Akt-driven tumorigenesis, we transduced Rat2 fibroblasts with myr-Akt1 (which transforms rat fibroblasts (15)) and either wild-type Beclin 1 or Beclin 1 phosphorylation site mutants. Myr-Akt1 suppressed autophagy in Rat2 fibroblasts (Fig. 2A and B), reduced co-immunoprecipitation of Class III PI3K Vps34 with Beclin 1 (Fig. S4A), and decreased Beclin 1-associated lipid kinase activity (Fig. S4B). The autophagy suppressive effects of active Akt were largely prevented by expression of the Beclin 1 AA mutant and mildly decreased by the Beclin 1 S295A mutant (Fig. 2A and B). Expression of myr-Akt1 had minimal effects on the interaction of Beclin 1 AA and Vps34 or on the amounts of Beclin 1 AA-associated Vps34 activity (Fig. S4A and S4B). Thus, myr-Akt1 suppresses Beclin 1-associated Vps34 activity and autophagy in a manner that is partially reversed by a Beclin 1 mutant resistant to Akt-mediated phosphorylation.

In an anchorage-independence growth assay, shRNA depletion of Beclin 1 (Fig. S5) or myr-Akt1 expression caused Rat2 fibroblasts to form numerous colonies in soft agar (Fig. 2C). Both the number and size of colonies formed by myr-Akt1-expressing cells were significantly reduced by co-expression of Beclin 1 AA (Fig. 2C, Fig. S6A and S6B). Rat2 cells expressing myr-Akt1 also had higher amounts of endogenous Beclin 1 S²⁹⁵ phosphorylation than control cells (Fig. 3B). Thus, active Akt1 promotes Beclin 1 S²⁹⁵ phosphorylation, and expression of a non-phosphorylatable mutant of Beclin 1 suppresses Akt-mediated transformation in vitro. Inactivation of Beclin 1 by Akt-mediated phosphorylation may therefore contribute to Akt’s transforming properties.

Beclin 1 AA also suppressed myr-Akt1-driven tumorigenesis in vivo. All myr-Akt1-expressing Rat2 cells formed tumors in immunodeficient NOD SCID® mice, but tumor growth rate and tumor mass upon necropsy were less for cells expressing Beclin 1 AA than for cells expressing wild-type Beclin 1 or myr-Akt1 alone (Fig. 2D and 2E, Fig. S6C). Tumors expressing myr-Akt1 alone had numerous cells displaying p62 immunoreactivity (indicating autophagy suppression) whereas very few p62 immunoreactive cells were detected in tumors from cells expressing both myr-Akt1 and Beclin 1 AA (Fig. 2F). Tumors expressing myr-Akt1 alone or active Akt and Beclin 1 resembled fibrosarcomas (Fig. 2F and table S1); 94% (17/18) showed tissue invasion and 50% (9/18) exhibited nuclear pleomorphism and decreased nucleus:cytoplasmic ratios. Tumors expressing myr-Akt1 and Beclin 1 AA formed less cellular, more disorganized tumors with limited or no tissue invasion; 60% (6/10) displayed degenerative features with numerous pyknotic nuclei. Tumors expressing active Akt alone or with Beclin 1 had higher mean mitotic counts (n=8.39; n=8.37, respectively) than tumors expressing myr-Akt1 and Beclin 1 AA (n=3.67) (P<0.0001). They also had increased K_i-67 labeling (Fig. S6D), and decreased TUNEL staining (Fig. S6E). Thus, the expression of a mutant of Beclin 1 that cannot be phosphorylated by Akt increases tumor cell autophagy, decreases tumor growth rate and size, decreases tumor cellular proliferation, and increases tumor cell death. These results suggest a role for Akt-mediated Beclin 1 phosphorylation in the tumorigenic effects of Akt.

To investigate how Akt phosphorylation of Beclin 1 inhibits autophagy, we determined whether Akt-mediated phosphorylation of Beclin 1 generates a 14-3-3 binding motif (10). Endogenous 14-3-3 proteins and Beclin 1 co-immunoprecipitated in HeLa cells and this interaction decreased during starvation (Fig. 3A). Mutation of predicted Beclin 1 14-3-3 binding sites, S234A or S295A, weakened the Beclin 1/14-3-3 interaction and the double mutation (AA) nearly completely abolished Beclin 1/14-3-3 binding (Fig. S7A). Similar amounts of Atg14, a component of the autophagy-inducing Beclin 1/Class III PI3K complex (16), immunoprecipitated with WT and non-phosphorylatable mutants of Beclin 1, indicating these mutations do not cause major alterations in protein stability or folding. Expression of myr-Akt1 blocked starvation-induced disruption of 14-3-3/Beclin 1 binding and conversely, expression of DN-Akt1 inhibited 14-3-3/Beclin 1 binding during growth in normal medium (Fig. S7B). Thus, Beclin 1 interacts with 14-3-3 proteins through S234 and S295, and this interaction is negatively regulated by starvation and Akt inhibition.

14-3-3 proteins can regulate their binding partners’ functions through interactions with intermediate filaments (17, 18). The intermediate filaments, keratin 18 (K18) and vimentin, immunoprecipitated with Flag-Beclin 1 in HeLa cells and this was reduced by expression of DN-Akt1 (Fig. S7C). Immunoprecipitation of Flag-Beclin 1 was increased by mutants of vimentin and K18 with increased binding to 14-3-3 proteins (K18 R89C (19)) but decreased with mutants with decreased binding to 14-3-3 proteins (K18S33A (19) and vimentin S39A) (Fig. S7D-E). The Flag-Beclin 1 AA mutant did not immunoprecipitate with wild-type vimentin-GFP (Fig. S7F). siRNA targeted against 14-3-3ε abolished the interaction between Flag-Beclin 1 and vimentin-GFP (Fig. S7G). Thus, Beclin 1 interacts with 14-3-3 proteins and intermediate filament proteins through a mechanism involving the S²³⁴ and S²⁹⁵ Akt phosphorylation/14-3-3 binding sites of Beclin 1 and the 14-3-3 binding sites of intermediate filament proteins.

In Rat2 cells, expression of myr-Akt1 increased the interactions of endogenous 14-3-3 proteins and vimentin (the major intermediate filament protein expressed in fibroblasts (20)) with endogenous Beclin 1 in parallel with its increased phosphorylation (Fig. 3B and C). Expression of active Akt had little effect on the binding of Beclin 1 AA with 14-3-3 proteins and vimentin (Fig. 3D). Thus, active Akt may promote the interaction of Beclin 1 with vimentin through phosphorylation of Beclin 1 and generation of 14-3-3 binding sites.

Wild-type Beclin 1 had a diffuse cytoplasmic localization in the absence of active Akt expression, and localization of Beclin 1 with vimentin was observed primarily in a perinuclear pattern (Fig. 3E). Expression of myr-Akt1 redistributed wild-type Beclin 1 into a reticular pattern and increased Beclin 1/vimentin colocalization, but did not have these effects on Beclin 1 AA. Conversely, starvation decreased Beclin 1/vimentin colocalization in a reticular pattern (Fig. S7H). Thus, active Akt enhances the colocalization of Beclin 1 with vimentin in Rat2 cells in a manner that requires the Beclin 1 Akt phosphorylation and 14-3-3 binding sites S²³⁴ and S²⁹⁵.

In Rat2 cells, two different shRNAs that target vimentin (Fig. S8A) increased autophagy (Fig. 4A). Vimentin appears to inhibit autophagy downstream of Beclin 1 phosphorylation because depletion of vimentin also increased autophagy in Rat2 fibroblasts expressing myr-Akt1 (Fig. S8B and Fig. 4B). This increased autophagy was associated with inhibition of Akt-mediated transformation; vimentin shRNAs significantly inhibited the number and size of Rat2 colonies formed in soft agar (Fig. 4C and Fig. S8C). Thus, the regulation of Beclin 1/vimentin interactions by active Akt may be a mechanism for both inhibition of autophagy by intermediate filaments and for Akt-mediated transformation. As in the case of the interaction of AMBRA1 with the dynein moter complex (21), these data indicate that interactions between core autophagy proteins and cytoskeletal elements may regulate autophagy.

Our findings demonstrate a link between oncogenic signaling, the core autophagy machinery, and cytoskeletal proteins in the intermediate filament family. Akt signaling, intermediate filaments and 14-3-3 proteins may be mechanistically linked to autophagy inhibition and tumorigenesis through regulation of the Beclin 1 complex (Fig. 4D). These findings also demonstrate a specific mechanism by which autophagy may be suppressed in human cancer. Cross-talk between oncogenic kinases and autophagy proteins might represent a fundamental mechanism underlying the regulation of mammalian cell growth control and cancer.

2