PTEN negatively regulates mTORC2 formation and signaling in grade IV glioma via Rictor hyperphosphorylation at Thr1135 and direct the mode of action of an mTORC1/2 inhibitor

To investigate the role of PTEN (phosphatase and tensin homolog) in mammalian target of rapamycin complex 2 (mTORC2) signaling in glioblastoma multiforme (GBM), we found higher activation of mTORC2 in PTENmu cells, as evidenced by enhanced phosphorylation of mTOR (Ser2481), AKT (Ser473) and glycogen synthase kinase 3 beta (GSK3β) (Ser9) as compared with PTENwt cells. In addition, PTENwt cells upon PTEN depletion showed mTORC2 activation. The reduced mTORC2 signaling in PTENwt cells was related to higher Rictor phosphorylation at Thr1135 residue. Phosphorylation of Rictor at Thr1135 inhibited its association with mTORC and thus there was a reduction in mTORC2 complex formation. In addition, PTENwt cells expressing mutated Rictor in which Thr1135 was substituted with alanine, showed enhanced mTORC2 formation and signaling. This enhanced mTORC2 signaling promoted inactivation of GSK3β. Thus, we established the reciprocal activation of mTORC2 and GSK3β in GBM. To the best of our knowledge, this is the first report describing role of PTEN in mTORC2 formation by promoting Rictor phosphorylation (Thr1135) in GBM. Furthermore, the drug sensitivity of mTORC2 was evaluated. A newly identified carbazole alkaloid, mahanine, showed cytotoxicity in both PTENmu and PTENwt cells. It inhibited both mTORC1/2 and AKT completely in PTENmu cells, whereas it inhibited only mTORC1 in PTENwt cells. Cytotoxity and AKT-inhibitory activity of the mTORC1/2 inhibitor was increased either by depleting PTEN or in combination with phosphatidylinositol 3 kinase inhibitors in PTENwt cells. In contrast, depletion of Rictor decreased the cytotoxicity of the mTORC1/2 inhibitor in PTENmu cells. Thus, PTEN has an important role in mTORC2 formation and also influences the effectiveness of an mTORC1/2 inhibitor in GBM.

PTEN (phosphatase and tensin homolog), a tumor suppressor protein, is often inactivated in cancers. 21 Cellular cross-talk between PTEN and mTORC1 via the phosphatidylinositol 3 kinase (PI3K)/AKT/mTOR pathway is often deregulated and enhances the malignancy. 22,23 Although the role of mTORC1 is well characterized, the function and regulation of mTORC2 is still poorly understood.
Glioblastoma multiforme (GBM) is a grade IV brain tumor with higher mortality rate. PTEN, EGFR, TP53, CDKN2A, CDK4 and LOH are several frequently mutated genes. 24 PTEN mutations are frequently involved with this aggressive cancer, our initial aim was to decipher the regulation of mTORC2 with respect to PTEN wild-type (PTEN wt ) vs mutated conditions (PTEN mu ; Figure 1a). This information may help us to search for an effective therapeutic strategy for disease management. Earlier, we identified a non-toxic novel carbazole alkaloid (mahanine), which induced apoptosis in several cancers including GBM. [25][26][27][28][29][30][31][32] Therefore, our next aim was to identify the mode of activity of mahanine as a possible mTOC1/2 inhibitor and enhance the sensitivity based on cellular existence of mTORC1/2.
Here we provide evidence for PTEN-mediated regulation of mTORC2 in GBM. We showed that PTEN mutations lead to reduced phosphorylation of Rictor at Thr1135, which in turn promotes enhanced mTORC2 formation and downstream signaling. However, higher phosphorylation of Rictor at Thr1135 leads to the reduction of mTORC2 formation in PTEN wt cells. Furthermore, we confirmed a positive correlation between enhanced mTORC2 formation and sensitivity toward mahanine in PTEN-mutated cells. Mahanine also inhibited mTORC1 activity and may thus be considered as a potential mTORC1/2 inhibitor.

RESULTS
Differential mTORC1 and mTORC2 activity in PTEN wt and PTEN mu cells To understand the impact of functional PTEN in mTORC2 formation and downstream signaling, we selected PTEN mu (U87MG) and PTEN wt (LN229) GBM cells and the status of PTEN was verified (Figure 1b). Next, we examined the status of specific phosphorylation of mTOR complexes. Active mTORC2-specific phosphorylation of mTOR at Ser2481 was higher in PTEN mu cells (Figure 1c). In contrast, mTORC1-specific phosphorylation at Ser2448 was almost comparable in PTEN wt and PTEN mu cells. However, Rictor and Raptor, two major mTORC2 and mTORC1associated proteins were very similar in both the cells. This result indicates probable activation of mTORC2 because of PTEN mutation.
To confirm this upstream activation of mTORC2 in PTEN mu cells, we examined the phosphorylation status of a few downstream molecules. Ser473 of AKT is a selective substrate of mTORC2. We observed enhanced expression of phospho-AKT at Ser473 in PTEN mu cells ( Figure 1d). As mTORC2 is not known to influence the Thr308 of AKT, we detected almost comparable levels of phosphorylation in both PTEN wt /PTEN mu cells (Figure 1e). In addition, no significant change of AKT phosphorylation at the Thr450 was found. Densitometric analysis of three phosphorylation sites of AKT revealed that only Ser473 was 5-fold enhanced in PTEN mu cells relative to PTEN wt (Figure 1f). This observation strongly suggested that PTEN mutation may have a significant role in regulating mTORC2 activity, which in turn enhanced the phosphorylation of its substrate AKT at the Ser473.
To confirm the activation of AKT through its Ser473 phosphorylation, next we focused on GSK3β. Activated AKT phosphorylates GSK3β at the Ser9 position and induces its inactivation. Phosphorylation of GSK3β at Ser9 was higher in PTEN mu cells, indicating the sequential signaling of activated mTORC2 and AKT and subsequent inactivation of GSK3β ( Figure 1g). This result suggested the reciprocal relationship of mTORC2 and GSK3β. In contrast, mTORC1 activity was similar in both PTEN mu /PTEN wt cells as judged by the equivalent phosphorylation of S6K1 at Thr389 and 4E-BP1 at Thr37/46 ( Figure 1h).
Knockdown of PTEN enhances mTORC2 signaling in PTEN wt cells To further confirm the role of PTEN in mTORC2 activity, we transiently knocked down PTEN using targeted short interfering RNA (siRNA) in PTEN wt cells (Figure 2a), which exhibited~4-fold increase in mTORC2-specific phosphorylation of mTOR at Ser2481 (Figures 2b and c), demonstrated by the western blot analysis. This result supported our initial observation in PTEN mu cells (Figure 1), confirming the activation of mTORC2 in the absence of wild-type PTEN. The downstream substrate of mTORC2 (Ser473 of AKT) was also highly phosphorylated upon PTEN depletion (Figures 2d and e) and the inhibitory Ser9 phosphorylation of GSK3β was also enhanced (Figures 2f and g). These findings reconfirmed the reciprocal activation profile of mTORC2 and GSK3β. We have also observed the enhanced phosphorylation at Thr308 (substrate of PDK1) of AKT upon PTEN knockdown (Figures 2d and e). This served as a positive control for PTEN deactivation and subsequent activation of the PI3K pathway, which in turn activated mTORC1 as evidenced by increased mTOR phosphorylation at Ser2448 (Figures 2b and c). As mTORC1 activity is regulated by the PI3K pathway, therefore we next studied the phosphorylation of S6K1 and 4E-BP1. We observed hyperphosphorylation of S6K1 at Thr389 and 4E-BP1 at Thr37/46 after knockdown of wild-type PTEN (Figures 2h and i).
Identification of Rictor as the target molecule of PTEN for mTORC2 signaling As binding of Rictor to mTOR is the major determining factor for mTORC2 signaling, we wanted to address its role in the PTENmutated and/or depleted conditions. Accordingly, we transiently knocked down the expression of Rictor by targeted siRNA in the PTEN mu cells (Figure 3a). Upon Rictor depletion, mTORC2 activation was diminished as evidenced by decreased Ser2481 phosphorylation of mTOR (Figures 3b and c). Subsequently, the specific downstream pathway was also hampered as evidenced by the decreased phosphorylation of Ser473 for AKT (Figures 3d and e) and Ser9 for GSK3β (Figures 3f and g). The relationship between mTORC2 inactivation and corresponding GSK3β activation was also observed after Rictor depletion. This data indicated that upon PTEN deactivation, Rictor is essential for the functional activation of mTORC2 and subsequent downstream signaling.
Next, PTEN wt cells were transfected with siRNA for PTEN or cotransfected with siPTEN and siRictor. Both the single-and doubleknockdown of PTEN/Rictor was confirmed ( Figure 3h). PTEN knockdown enhanced the overall mTORC2 signaling molecules as observed before ( Figure 2). However, when Rictor and PTEN were depleted, the mTORC2 signaling was comparatively reverted. Thus, PTEN knockdown-mediated enhanced phosphorylation of mTOR (Ser2481), AKT (Ser473), and GSK3β (Ser9) were decreased upon Rictor depletion. Therefore, our observation confirmed that Rictor is an essential molecule upon PTEN inactivation for mTORC2 signaling. However, in the absence of Rictor, the basal level phosphorylation of AKT (Ser473) was possibly maintained by AKT autophosphorylation. The literature has shown that AKT can be autophosphorylated (Ser473) by the activation of the AKT kinase domain because of phosphorylation at Thr308. 33 Therefore, even after depletion of Rictor, a basal level of Ser473 phosphorylation was observed.
We next addressed whether this mechanism was consistent across other GBM cells namely T98G (PTEN mu ), U373MG (PTEN mu ) and LN18 (PTEN wt ). Both PTEN mu cells were hyperactive in mTORC2 signaling as reflected by higher phosphorylation of (h) mTORC1 activity readout judged by the phosphorylation of its downstream substrate 4E-BP1 and S6K1. (i) Fold change of p-S6K1 and p-4E-BP1 showed with relative densitometry analysis. β-Actin was served as loading control in all set of experiments. All the densitometric analysis was done by ImageJ software and normalized against the expression of β-actin and respective total protein level. The protein levels from three independent experiments were quantified and presented as mean ± s.d. Asterisk indicates significant difference when compared with the control (P o0.05). mTOR (Ser2481) in contrast to PTEN wt (Figure 3i). Downstream substrates (Ser473 of AKT and Ser9 of GSK3β) were also hyperphosphorylated, suggesting that this may be a general phenomenon of PTEN-dependent mTORC2 activation in GBM.
Functional existence of PTEN negatively regulates mTORC2 complex formation After identifying the importance of Rictor in PTEN-mutated and/or depleted conditions, we asked an obvious question about the interconnection between mTOR complex formation and differential signaling of mTORC2 in PTEN wt /PTEN mu cells. A co-immunoprecipitation experiment showed that mTOR was associated with Raptor almost equivalently, indicating a similar level of mTORC1 formation in both cells (Figure 4a). In contrast, mTOR was predominantly associated with Rictor only in PTEN mu cells, demonstrating differential mTORC2 formation between these two cells.
For further reconfirmation, we depleted PTEN from PTEN wt cells using siRNA targeting PTEN and observed a significant association between mTOR and Rictor in deficiency of PTEN, suggesting formation of mTORC2 ( Figure 4b). This result suggested that PTEN in active form not only suppresses the mTORC2 signaling but also inhibits the Rictor-dependent mTORC2 complex formation. PTEN inhibits mTORC2 formation by stimulating Rictor (Thr1135) phosphorylation We then addressed the cause of reduced recruitment of Rictor to mTORC in the presence of functional PTEN. Phosphorylation of Rictor (Thr1135) is known to inhibit mTORC2 activity as evidenced by decreased AKT Ser473 phosphorylation. 34 Accordingly, we addressed whether this phosphorylation was differentially regulated between PTEN wt and PTEN mu cells. Our results revealed that phosphorylation of Rictor at the Thr1135 were significantly higher in PTEN wt cells ( Figure 4c). To investigate whether the presence of functional PTEN had any relevance for Rictor Thr1135 phosphorylation, we transiently knocked down PTEN in PTEN wt cells. These knockdown cells exhibited decreased phosphorylation at Thr1135 of Rictor (Figures 4d and e). This result suggested an involvement of PTEN with Rictor hyperphosphorylation at the Thr1135, which in turn possibly inhibits Rictor's association with mTOR, and we observed reduced mTORC2 formation and lower mTORC2 activity. Considering this, we checked the status of Rictor phosphorylation in other PTEN mu /PTEN wt cells, and other PTEN mu cells also showed a lower level of Rictor (Thr1135) phosphorylation compared with PTEN wt (Figure 4f), consistent with Figure 3i. These findings support our hypothesis that higher phosphorylation of Rictor is responsible for reduced recruitment to mTORC.
To confirm our findings, we next transfected PTEN wt cells with either wild-type pRK-5/Rictor (myc-tag), the T1135A mutant of pRK-5/Rictor (myc-tag) or the empty vector alone and checked the association of exogenous Rictor and endogenous mTOR by co-immunoprecipitation. Cells expressing mutated Rictor (T1135A) showed enhanced association with endogenous mTOR in comparison with wild-type Rictor/empty vector-transfected cells (Figure 4g). In addition, the level of myc-tag Rictor was equivalent in both wild-type and mutated transfections (Figure 4h). Western blot also revealed that phosphorylation of AKT (Ser473), GSK3β (Ser9) and mTOR (Ser2481) were increased in cells transfected with mutant Rictor (T1135A), indicating higher mTORC2 signaling. These data conclusively established that functional PTEN inhibit mTORC2 formation by promoting Rictor (Thr1135) phosphorylation (Figure 4i). However, it is unusual that one phosphatase can induce hyperphosphorylation. It has been reported that S6K1 is responsible for the phosphorylation (Thr1135) of Rictor. Earlier, we found that upon PTEN depletion there was an increase in S6K1 phosphorylation (Thr389), that is, the signature of active S6K1 (Figures 2h and i). Accordingly, we hypothesized that upon PTEN depletion in PTEN wt cells, the decrease in Rictor phosphorylation (Thr1135) may revert back because of the increased activity of S6K1. Therefore, the actual effects of PTEN were masked by the increased activity of S6K1. To reveal the definite effects of PTEN depletion upon Rictor phosphorylation, we first inhibited the S6K1 phosphorylation by disrupting active mTORC1 formation, as mTORC1 is responsible for phosphorylation of S6K1. To inhibit mTORC1 activity, we knocked down Raptor in the PTEN wt cells and observed reduced phosphorylation of both S6K1 (Thr389) and mTOR (Thr2448) (Figure 4j). In consequence, the phosphorylation of Rictor (Thr1135) was decreased and mTOR phosphorylation (Thr2481) was increased, indicating S6K1 mediated possible inactivation of mTORC2 (Figures 4j and k). Simultaneous depletion of both Raptor and PTEN from PTEN wt cells triggered a complete reduction in Rictor Thr1135 phosphorylation as well as increased phosphorylation of mTOR (Thr2481). Thus, PTEN depletion coupled with S6K1 deactivation enhances mTORC2 activity through Rictor dephosphorylation (Thr1135).
Mahanine behaves as a mTORC1/mTORC2 inhibitor, acting differentially in PTEN mu and PTEN wt cells Identification of differential mTORC signaling was the foundation for the development of chemotherapeutic agents. We previously identified mahanine as a potential chemotherapeutic agent for GBM (Supplementary Figure S1). 25 It displayed a cytotoxic effect against PTEN mu and PTEN wt cells (Supplementary Figure S2). Here we observed that mahanine suppressed both Ser2448 and Ser2481 phosphorylation of mTOR in PTEN mu cells, whereas only Ser2448 phosphorylation was inhibited in PTEN wt cells (Figure 5a). Owing to the low level of Ser2481 phosphorylation of mTOR in PTEN wt cells, mahanine-mediated regulation was almost undetectable. This result indicated that mahanine act as both an mTORC1 and mTORC2 inhibitor in PTEN mu /PTEN wt cells, depending on the existence of those complexes.
To confirm this mTOR-inhibitory mechanism of mahanine, we studied its downstream AKT pathway. Mahanine-treated PTEN mu cells exhibited decreased phosphorylation at Thr308 and Ser473 when both mTORC1 and mTORC2 were present ( Figure 5b). However, mahanine behaved similarly in both PTEN mu and PTEN wt by inducing downregulation of 4E-BP1 in total protein level and phosphorylated forms but phosphorylation of S6K1 (Thr389) was decreased only in PTEN mu cells in contrast to PTEN wt , where phosphorylation at this site was increased (Figure 5c).
Although mahanine showed cytotoxicity in PTEN wt cells, we did not observe any inhibition of AKT phosphorylation (Thr308 and Ser473) when mTORC2 was not in significant level and mainly mTORC1 was present (Figure 5b). Therefore, mahanine was possibly inducing cell death in PTEN wt cells by targeting other pathways. Inhibition of mTORC1 releases the feedback inhibition mediated by the S6K1-IRS1-PI3K loop, which leads to activation of the AKT pathway in cancer. 35 Accordingly, we inhibited the PI3K pathway using two specific inhibitors, wortmannin and LY294002, separately to check the status of AKT activation in PTEN regulates mTORC2 formation in GBM K Bhattacharya et al mahanine-treated PTEN wt cells. Mahanine, even at a low dose, exhibited reduced AKT phosphorylation at Thr308 and Ser473 sites in the presence of these inhibitors, suggesting involvement of the PI3K pathway in AKT activation in PTEN wt cells (Figures 5d  and e). We also checked other downstream effector molecules of mTORC1 and observed phospho-S6K1 (Thr389) and 4E-BP1 (Thr37/46) were reduced significantly when the mTORC1/2 inhibitor was used in combination with wortmannin as compared with the inhibitor alone (Figure 5e). Furthermore, the mTORC1/2 inhibitor with LY294002 or wortmannin synergistically enhanced the cell death in PTEN wt cells compared with the inhibitor alone (Figures 5f and g). Overall, we demonstrated that, in the PTEN wt condition, where mTORC1 formation was predominant and mTORC2 was not present in significant amount, this mTORC1/2 inhibitor behaved like an mTORC1 inhibitor and enhanced the AKT-mediated cell protective pathway, which could be reverted by PI3K inhibition.
A therapeutic strategy to treat GBM by suppressing native PTEN to enhance the activity of the mTORC1/2 inhibitor Next, we addressed the role of PTEN in activation of AKT in mahanine-treated PTEN wt cells. We depleted PTEN from the PTEN wt cells and observed the effect of the mTORC1/2 inhibitor. Western blot analysis revealed that the mTORC1/2 inhibitor could inhibit AKT activation, identified by a reduction in phosphorylation in PTEN wt cells when the PTEN was depleted by siRNA ( Figure 6a). The activity of the mTORC1/2 inhibitor in PTEN-depleted PTEN wt cells was enhanced, as identified by the increased propidium iodide (PI) positivity compared with mock-transfected cells (Figure 6b). This is possibly because PTEN depletion induces generation of mTORC2 as a target complex. This result indicated that the mTORC1/2 inhibitor exhibited more effects when PTEN is mutated and subsequently mTORC2 is activated.
To confirm our findings, we next inhibited mTORC2 formation in PTEN mu cells by depleting Rictor and then treated with mahanine. We observed decreased PI positivity when there was a reduction in mTORC2 activity (Figure 6c). Microscopic images also proved that the mTORC1/2 inhibitor has a significantly lowered cytotoxic effect after Rictor knockdown (Figure 6d). Therefore, we concluded that cells having enhanced mTORC2 are more sensitive to the mTORC1/2 inhibitor. Our overall findings established that, as in PTEN wt cells, no significant mTORC2 formation were occurring, so the mTORC1/2 inhibitor-induced AKT activity was driven by the sole inhibition of mTORC1. Upon PTEN depletion, they behave like a dual mTORC1/2 inhibitor because of the formation of enhanced mTORC2 as a target molecule (Figure 6e). On the contrary, Rictor depletion inhibits mTORC2 formation, made PTEN mu cells less sensitive to the mTORC1/2 inhibitor. Thus, PTEN has an important role in mTORC2 formation and also influences the effect of an mTORC1/2 inhibitor in GBM. mTORC2 formation showed reduced sensitivity to a GSK3β inhibitor Along with mTORC1/2, GSK3β has long been a chemotherapeutic target in GBM. We had identified a significant reciprocal activation between mTORC2 and GSK3β (Figures 1g and 2f). In PTEN mu cells, mTORC2 was activated and GSK3β was inactivated, whereas in PTEN wt cells the situation was the reverse. To check the functionality of this observation, we performed a cell death assay and observed that a specific GSK3β inhibitor (SB216763) showed lower PI positivity, indicating lower cell death in PTEN mu cells compared with PTEN wt (Figure 6f). We hypothesized that in PTEN mu cells, GSK3β was already inhibited by a higher level phosphorylation (Ser9) and therefore the addition of a GSK3βinhibitor did not exert any extra effect. In contrast, an mTORC1/2 inhibitor (mahanine) exhibited significantly enhanced cytotoxicity in PTEN mu cells. This was further confirmed by demonstrating higher G0/G1 phase cell cycle arrest (Figure 6g). These results validated that, depending on PTEN status, mTORC2 and GSK3β maintain a reciprocal relationship. Cells, which are susceptible to mTORC2 inhibition, may not be susceptible to GSK3β inhibition and vice versa (Figure 6h).

DISCUSSION
Mutation of PTEN is associated with high-grade GBM formation. mTORC2 is emerging as a potential drug target in PTEN-mutated cancers. The detailed mechanism of molecular regulation and cross-talk between PTEN and mTORC2 is still not clearly known. Accordingly, we investigated the role of PTEN in mTORC2 regulation and signaling in GBM. The major achievement of this study is providing evidence that PTEN mu cells exhibit enhanced mTORC2 formation and signaling because of the decreased inhibitory phosphorylation of Rictor (Thr1135). This is the first evidence suggesting Thr1135 phosphorylation of Rictor is controlled by PTEN, and demonstrates PTEN as the master regulatory molecule in mTORC2 formation. In addition, we have established mahanine, carbazole alkaloid from an edible plant, as a potential mTORC1/2 inhibitor that showed cytotoxicity both in PTEN mu /PTEN wt cells. mTORC2 activity is required for generation of PTEN deletioninduced prostate cancer. 36 EGFRvIII-mediated activation of mTORC2 was reversed by PTEN activity. 37 A correlation between PTEN and Rictor was also reported in leukemogenesis. 38 Here we PTEN regulates mTORC2 formation in GBM K Bhattacharya et al showed that PTEN mu cells have higher mTORC2 formation and downstream signaling than PTEN wt cells as evidenced by enhanced mTOR Ser2481, AKT Ser473 and GSK3β Ser9 phosphorylations. Although a correlation between PTEN and AKT Ser473 phosphorylation was reported in lung cancer and rhabdomyosarcomas, no cross-linking molecule between these phosphatase and kinase was demonstrated. 39,40 Here, we showed mTORC2 is a connecting molecule between PTEN and AKT.
During exploration of mTORC1/2-associated molecules, we observed that Rictor is differentially phosphorylated (Thr1135) with respect to the functional existence of PTEN. Rictor phosphorylation (Thr1135) is responsible for decreased activity of mTORC2 and reduced AKT (Ser473) phosphorylation. 34 Therefore, reduced mTORC2 activity and AKT Ser473 phosphorylation in PTEN wt cells is a consequence of increased Rictor Thr1135 phosphorylation. This event was confirmed when we knocked down PTEN in PTEN wt cells and observed reduced Rictor phosphorylation, and as a consequence, enhanced mTORC2 signaling. Although Rictor Thr1135 phosphorylation is responsible for decreased activity of mTORC2, reports suggest this had no effect on mTORC2 complex formation. 34,[41][42][43] However, the detailed mechanism behind reduced mTORC2 activity without hampering mTORC2 complex formation was not revealed. Our data revealed that not only was downstream signaling affected, but also there was differential mTORC2 complex formation in PTEN wt and PTEN mu cells. PTEN wt cells lack significant mTORC2 formation, and that was reverted with PTEN knockdown. Therefore, we conclude that in the presence of PTEN wt , there is enhanced Rictor Thr1135 phosphorylation, decreased mTORC2 formation, which connected to lower AKT activity, and increased GSK3β activity. The role of PTEN in regulating Thr1135 phosphorylation and mTORC2 complex formation was further demonstrated by expressing T1135A mutant Rictor in PTEN wt cells. We observed significant mTORC2 complex formation and downstream signaling in the presence of PTEN when there was a mutant Rictor that cannot be phosphorylated on Thr1135. Therefore, we concluded that this may be a unique phenomenon in GBM where phospho-Rictor Thr1135 could not be recruited to mTORC2 in the presence of PTEN wt .
Thr1135 phosphorylation of Rictor is also controlled by S6K1 via mTORC1. 34,43 This downstream kinase is known to be a linker molecule between mTORC1/2. Therefore, we checked mTORC1 activity and found equivalent mTORC1 formation and downstream activity as judged by comparable phosphorylation of downstream signaling molecules (S6K1 and 4E-BP1). This data reconfirmed the involvement of PTEN in regulation of mTORC2 in GBM. Therefore, S6K1 may not be solely responsible for controlling Rictor (Thr1135) phosphorylation. Simultaneous suppression of S6K1 activity and functional PTEN in PTEN wt cells completely abolishes Rictor phosphorylation (Thr1135) and enhances active mTORC2 formation. However, PTEN, being a phosphatase, cannot phosphorylate a molecule directly. We assume that PTEN may be affecting some other kinases by removing inhibitory phosphorylation, and they in turn phosphorylate Rictor at the specific site. Alternatively, it may also be possible that PTEN dephosphorylates some other site of Rictor that enables the molecule to change its conformation in such a way so that the Thr1135 site is exposed for further phosphorylation by known kinases like S6K1. Experiments to distinguish between these possibilities are ongoing.
Basic understandings at the cellular and molecular level are successful when a translational or therapeutic approach works out positively. We have observed mahanine exerted cytotoxicity both in PTEN mu and PTEN wt cells thus behaved as an mTORC1/2 inhibitor. However, in PTEN wt cells, where mTORC2 was not significantly active and mTORC1 is predominant, mahanine behaves like an mTORC1 inhibitor. Here we observed an activation of the AKT pathway in PTEN wt cells. Several reports indicated that mTORC1 inhibition could lead to activation of AKT by other pathways. 35,44,45 This was due to release of the negative feedback loop via S6K1 leading to the activation of PI3K/AKT pathway. 44 mTORC1 inhibition also leads to MAPK pathway-dependant AKT activation. 45 This activation of AKT was reversed in PTEN wt cells in the presence of specific PI3K inhibitors. We also observed a synergistic effect when a combination of mTORC1/2 and PI3K inhibitors were used to overcome this AKT activity in PTEN wt cells. These data suggest that for an effective therapeutic strategy both mTORC1/2 and PI3K inhibitors may be necessary for a beneficial effect in PTEN wt GBM.
Our findings demonstrated that PTEN behaved like a physiological inhibitor of mTORC2 formation. Generation of mTORC2 upon PTEN depletion provided the specific target of mahanine and hence cell death was increased in PTEN wt cells.
We have observed a reciprocal relationship between mTORC2 and GSK3β in GBM. Therapeutically, we also proved this reciprocal relationship. In PTEN mu cells where mTORC2 and AKT are active and GSK3β is inactive, treatment with GSK3β inhibitors did not add any extra benefit. In contrast, a GSK3β inhibitor exerted a significant effect on PTEN wt cells where GSK3β is active.
Studies have suggested that mTORC1/C2 inhibitors have more effect on PTEN-mutated cancers. 46,47 Our results convincingly showed the mechanism whereby PTEN mutations increase the sensitivity of cancers to mTORC1/2 inhibitors. We have established loss of PTEN induces more mTORC2 formation, and more susceptible to mTORC1/2 inhibitors. This hypothesis was further proven when we depleted Rictor from PTEN mu cells, which had abundant mTORC2 formation. Depletion of Rictor leads to decreased mTORC2 formation, and therefore they behaved like PTEN wt cells, with predominant mTORC1. Accordingly, mTORC1/2 inhibitors acted as only mTORC1 inhibitors and we observed a reduction in cell death. Therefore, mTORC2 is the prime factor that is responsible for differential sensitivity to mTORC1/2 inhibitors. PTEN showed its effect by inhibiting Rictor and thereafter mTORC2 kinase activity in PTEN wt cells.
This observation was additionally confirmed by using a wellknown mTORC1 inhibitor, rapamycin. Although rapamycin shows cytotoxicity in PTEN wt /PTEN mu cells, however, these cells activate AKT as evidenced by the enhanced Thr308 and Ser473 phosphorylation. Phosphorylation of 4E-BP1 remained unchanged whereas S6K1 was decreased (Supplementary Figure S5). Therefore, these data indicate again that not only mahanine but other mTORC1 inhibitors also activate AKT in PTEN wt cells. Most importantly, a sole mTORC1 inhibitor can hyperphosphorylate AKT irrespective of PTEN status.
To correlate our observation, we found that mahanine inhibited AKT in two PTEN mu prostate cancer cells. 26,32 Mahanine, in contrary, prevented palmitate-induced inhibition of the insulinstimulated phosphorylation of AKT in a PTEN wt cell. 48 According to our hypothesis and findings, in PTEN mu prostate cancer cells, mahanine might inhibit AKT completely by suppressing both  Therefore, it is envisaged that PTEN-dependent mTORC1 and mTORC2 dynamics are not restricted only to the GBM system and could be a global phenomenon. Our findings will open up a new avenue of thought for designing modified and efficient therapy against malignancy. However, more in-depth studies are needed to uncover the molecular function of PTEN in regulation of mTORC2.
Cell lines and culture condition PTEN mu (U87MG, T98G and U373MG) and PTEN wt (LN229 and LN18) human GBM cell lines (Supplementary Figure S4) were from ATCC (Manassas, VA, USA) and mycoplasma contamination was checked. They were cultured in Iscove's modified dulbecco's medium (IMDM) supplemented with 10% heat inactivated fetal bovine serum and 1% antibioticantimycotic, and maintained at 37 o C with 5% CO 2 . For specific experiments, cells (1 × 10 4 − 1 × 10 6 ) were treated with various inhibitors/ drugs under identical conditions. Transfection, siRNA and plasmids siRNA against human PTEN, Rictor and Raptor were from Ambion Life Technologies, Carlsbad, CA, USA (Supplementary Figure S3). Vector encoding myc-tag wild-type Rictor (pRK-5/Rictor) or T1135A mutated form of Rictor (pRK-5/Rictor T1135A) was kind gift from Dr Philippe P Roux (Faculty of Medicine, Universite´de Montre´al, Montreal, Quebec, Canada). PTEN mu /PTEN wt cells were transiently transfected with human PTEN/Rictor/ Raptor targeting siRNAs (Supplementary Figure S3) or pRK-5/Rictor or pRK-5/Rictor T1135A or empty vector alone using lipofectamine LTX and plus reagent following the manufacturer's instruction. PTEN-depleted/ Rictor-depleted cells were further treated with inhibitors for 24 h and processed for western blot and flow cytometric analysis. 25 Immunoblotting Drug/inhibitor treated or siRNA/plasmid transfected or untreated/untransfected cells were harvested and lysed and processed for western blot analysis. 25 Target proteins were identified by the West Pico chemiluminiscence detector system or Biorad ChemiDoc MP System (Bio-Rad, Hercules, CA, USA). Developed bands were densitometrially analyzed by ImageJ software (NIH, Bethesda, MD, USA). 25

Co-immunoprecipitation and immunoprecipitation
For detection of the mTOR complexes, cells were lysed in ice by sonication (Qsonica-LLC, XL-2000 series, Newtown, CT, USA). Protein (200 μg) was incubated with anti-mTOR antibody (1:100) overnight at 4 o C. Immunocomplex was incubated with protein A-sepharose 4B, washed with ice-cold phosphate-buffered saline, resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and subsequently identified using the respective antibodies.
Detection of cell death and cell cycle analysis by MTT assay and flow cytometry Mahanine was purified as described previously. 31 Cell viability of mahanine (0-30 μM)-treated cells (1 × 10 4 ) was checked after 48 h. They were further incubated with MTT (100 μg/ml) for 3 h. Formazan crystals were dissolved in dimethylsulfoxide and optical density was taken at 550 nm. 25 Cells (1 × 10 6 ) were treated with mahanine, SB216763, Wortmanin and LY294002 alone or in combination for 24 h and processed using Cell Cycle Test Plus kit. 25 At least, 20 000 cells were acquired and analyzed by flow cytometry using CellQuest Pro software (BD FACSCalibur, San Jose, CA, USA). 49 Morphology of mahanine-treated cells were imaged by transmitted light digital inverted microscope (EVOS XL Core, Life Technology, Carlsbad, CA, USA).

Statistical analysis
All the data were from at least three independent experiments and statistical analysis was performed using Graph Pad Prism 5 (La Jolla, CA, USA). The differences between the groups were analyzed by two tail Student's t-test or Mann-Whitney U-test. Standard error bars represent the s.d. of the mean (± s.d.) and *Po 0.05, denoted the significant differences between the means of the untreated and treated cells or two test groups.