Endosomal mTORC2 Is Required for Phosphoinositide-Dependent AKT Activation in Platelet-Derived Growth Factor-Stimulated Glioma Cells

Simple Summary The full activation of AKT, which is necessary for cell physiological changes, is achieved through the phosphorylation of Thr308 and Ser473 in human AKT. Here, we have addressed how AKT activation at early endosomes occurs during growth factor stimulation and how mTORC2 is recruited into endosomes and associated with AKT. The explanation comes from the discovery of three important events: (1) the physical association of mSIN and Rictor, critical components for mTORC2 assembly and activity, with early endosomes; (2) the control of the recruitment of mSIN to endosomes by PtdIns(3,4)P2; and (3) the PtdIns(3,4)P2-mediated endosomal AKT activation through phosphorylation at Ser473 to control a subset of AKT substrates. Abstract The serine/threonine kinase AKT is a major effector during phosphatidylinositol 3-kinase (PI3K)-driven cell signal transduction in response to extracellular stimuli. AKT activation mechanisms have been extensively studied; however, the mechanism underlying target of rapamycin complex 2 (mTORC2) phosphorylation of AKT at Ser473 in the cellular endomembrane system remains to be elucidated. Here, we demonstrate that endocytosis is required for AKT activation through phosphorylation at Ser473 via mTORC2 using platelet-derived growth factor-stimulated U87MG glioma cells. mTORC2 components are localized to early endosomes during growth factor activation, and the association of mTORC2 with early endosomes is responsible for the local activation of AKT, which is critical for specific signal transduction through glycogen synthase kinase-3 beta and forkhead box O1/O3 phosphorylation. Furthermore, endosomal phosphoinositide, represented by PtdIns(3,4)P2, provides a binding platform for mTORC2 to phosphorylate AKT Ser473 in endosomes through mammalian Sty1/Spc1-interacting protein (mSIN), a pleckstrin homology domain-containing protein, and is dispensable for AKT phosphorylation at Thr308. This PtdIns(3,4)P2-mediated endosomal AKT activation provides a means to integrate PI3K activated by diverse stimuli to mTORC2 assembly. These early endosomal events induced by endocytosis, together with the previously identified AKT activation by PtdIns(3,4,5)P3, contribute to the strengthening of the transduction of AKT signaling through phosphoinositide.


Introduction
AKT/PKB is a central signaling effector that phosphorylates specific serine or threonine residues of core proteins that control cell physiological processes, such as cell survival, metabolism, growth, proliferation, and migration [1,2]. Since dysregulated AKT signaling is a major cause of human metabolic diseases, such as type II diabetes, obesity, and cancers, there are significant ongoing efforts to understand the upstream regulators of AKT.
Herein, we show that AKT activation via Ser473 phosphorylation by mTORC2 requires endocytosis in U87MG glioma cells during platelet-derived growth factor (PDGF) stimulation. We found that mTORC2 is recruited into the early endosome using confocal imaging and biochemical approaches with endosomal fractionation in PDGF-activated cells. We also demonstrate that mSIN1-an essential contributor for mTORC2 activity and integrity-is localized to the early endosome in a PtdIns(3,4)P 2 -dependent manner and that endosomal mTORC2 is responsible for AKT activation through Ser473 phosphorylation. This study highlights the signaling role of PtdIns(3,4)P 2 molecules in early endosomes promoted by endocytosis for mTORC2-mediated AKT activation.

Cell Culture, Transfection, and Reagents
U87MG cells (ATCC, Manassas, VA, USA) were cultured in DMEM (Welgene, Gyeongsan, South Korea, and PAN Biotech, Aidenbach, Germany) supplemented with 10% fetal bovine serum (Gibco-BRL, Grand Island, NY, USA and PAN Biotech) and 1% penicillinstreptomycin (HyClone, Logan, UT, USA, and PAN Biotech). Cells were grown at 37 • C in a humidified atmosphere of 5% CO 2 . Mycoplasma testing was conducted periodically. Cells were subjected to transient transfection with expression plasmids using the Effectene Reagent (QIAGEN, Hilden, Germany) or the Neon transfection method (Thermo Fisher Scientific, Waltham, MA, USA), according to the manufacturers' manual. The following reagents were used at doses indicated in the figure legends: Rapamycin were obtained from Invivogen, Hong Kong; recombinant rat PDGF-BB (520-BB-050) was purchased from R&D Systems, Minneapolis, MN, USA; chlorpromazine was purchased from Sigma-Aldrich, St. Louis, MO, USA; Pitstop 2 was purchased from Abcam, Cambridge, United Kingdom; and CellMask Plasma Membrane Stain was purchased from Life Technologies, Carlsbad, CA, USA.

Analysis of PDGFR Endocytosis
PDGFR endocytosis was assessed as described previously [32], with some modifications. Serum-starved U87MG cells were incubated with bispecific scFv-Cκ-scFv fusion proteins (10 µg/mL) for 5 min at 37 • C for internalization. Cells were washed with cold PBS three times, then incubated with acidic buffer (0.2 M acetic acid, pH 2.7, 0.5 M NaCl) for 3 min at room temperature to remove antibodies bound to the cell surface. Cells were then washed with cold PBS twice, fixed in 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) in PBS for 10 min, and subjected to immunofluorescence staining analysis. Briefly, cells were incubated with PBS containing 5% horse serum (Gibco-BRL, USA) and 0.1% Triton X-100 for 30 min to block nonspecific antibody binding, and then with 2 µg/mL of a fluorescein isothiocyanate (FITC)-conjugated anti-human Cκ antibody (TB28-2, BD Biosciences, Franklin Lakes, NJ, USA) for 2 h at room temperature. Cells were also stained with 0.2 µg/mL 4 ,6-diamidino-2-phenylindole (DAPI, Roche, Basel, Switzerland) to detect DNA. Confocal images were acquired with an LSM 880 microscope (Carl Zeiss, Oberkochen, Germany) at Ewha Fluorescence Core Imaging Center (Seoul, South Korea), and images were processed with Zen software (Carl Zeiss, Thornwood, New York, NY, USA).

Rapamycin (RAPA)-Mediated Depletion of Phosphoinositide in Endosomes
For depletion of each phosphoinositide from Rab5-GTP endosomes, U87MG cells were transiently transfected with iRFP-FRB-Rab5 (an early endosome anchoring protein), mCherry-FKBP-MTM1 (a RAPA-responsive targeting 3-phosphatase), mCherry-FKBP-INPP4B (a RAPA responsive targeting 4-phosphatase), or CFP-FKBP-Inp54p (a RAPA responsive targeting 5-phosphatase) for 24 h. Cells were serum starved for 5 h and then incubated with PDGF (50 ng/mL) in the presence of 40 nM RAPA for 5 min. To monitor the recruitment of each lipid phosphatase to the endosomes in the presence of RAPA, livecell confocal microscopy was performed according to the following method. To monitor the amount of pAKT and PtdIns(3,4)P 2 , immunofluorescence analysis was carried out according to the following method with fixed cells.

Immunofluorescence and Live-Cell Confocal Imaging
Immunofluorescence analysis and live-cell confocal microscopy were performed as described previously, with some modifications [33]. Cells were cultured in 12-well dishes containing coverslips (diameter, 18 mm) coated with poly-L-lysine (Sigma-Aldrich, St. Louis, MO, USA) for both live-cell imaging and immunofluorescence staining. The cells were fixed for 10 min in 4% paraformaldehyde (Electron Microscopy Sciences) in PBS and incubated with a blocking solution containing 5% normal horse serum (Gibco-BRL, USA) and 0.1% Triton X-100 (Duchefa Biochemie, Haarlem, the Netherlands) in PBS for 30 min at room temperature. The cells were labeled with the primary antibody in a blocking solution for 30 min at room temperature or overnight on ice. After three washes with PBS for 5 min, the cells were incubated for 1 h at room temperature with a 1:1000 dilution of the fluorescence conjugated secondary antibody (Invitrogen) in blocking solution. After three washes with PBS, DAPI (0.2 µg/mL) was used for 2 min to detect DNA. Cells were mounted onto glass slides with Fluoromount-G (Southern Biotech, Birmingham, AL, USA). For coimmunostaining of other proteins, the cells were washed three times with PBS for 5 min and the blocking procedure was repeated. Phosphoinositide staining was performed as described previously [34]. Cells were fixed for 20 min in 2% paraformaldehyde and permeabilized with PBS containing 0.5% saponin (Sigma-Aldrich) and 1% bovine serum albumin (BSA, Bovogen, East Keilor, Australia) for 30 min at room temperature. Cells were incubated with lipid-specific antibodies for 2 h in PBS containing 1% BSA. After three 5-min washes with PBS, the cells were incubated with a fluorescent secondary antibody in PBS containing 1% BSA and washed three times for 5 min each with PBS. Confocal images were acquired using an LSM 880 (Carl Zeiss) or A1R (Nikon, Tokyo, Japan) confocal microscope. Live-cell images were obtained using the A1R microscope equipped with an incubation chamber (Live Cell Instrument, Namyangju, South Korea) at 37 • C in an atmosphere of 5% CO 2 . Images were processed using NIS-Elements (Nikon) or Zen (Carl Zeiss) software. The following antibodies were used for immunofluorescence staining:

Nikon-Structured Illumination Microscopy (N-SIM)
U87MG cells expressing GFP-mSIN were fixed with 4% paraformaldehyde and immunostained with antibodies to PtdIns(3,4)P 2 , followed by secondary antibodies conjugated with Alexa Fluor 546. SIM images were obtained using an N-SIM microscope (Nikon) equipped with a 63×/1.4 NA objective (1024 × 1024 pixels). To obtain line intensity histogram images, the acquired U87MG cell images were analyzed with NIS-Elements AR 4.2 software (Nikon).

AKT Kinase Assay
This assay was performed using an AKT kinase activity kit (ADI-EKS-400A; Enzo Life Sciences, Farmingdale, NY, USA) according to the manufacturer's instructions with minor modifications. Serum-starved and PDGF-stimulated U87MG cells (2.5 × 10 6 per 100-mm dish) were incubated in the absence or presence of chlorpromazine (5 µg/mL) or Pitstop2 (5 µM) for 3 min. The cells were then washed with cold PBS, harvested, and sonicated with 1 mL of lysis buffer containing 25 mM HEPES-NaOH, pH 7.0, 2 mM EDTA, 25 mM β-glycerophosphate, 1% Triton X-100, 10% glycerol, a mixture of protease inhibitors (1 mM DTT, 5 mM NaF, 10 µg/mL aprotinin, 10 µg/mL leupeptin, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride), and a phosphatase inhibitor cocktail (Sigma-Aldrich). The supernatants collected by centrifugation at 12,000× g for 20 min at 4 • C were used for column chromatography using desalting columns with a molecular weight cutoff of 40,000 Daltons (87,769, Thermo Fisher Scientific) according to the manufacturer's instructions. The resulting protein mixtures (molecular weight >40,000) were concentrated with an Amicon Ultra-2 centrifugal filter unit (UFC201024, EMD Millipore, USA) and used for the AKT kinase activity assay. The protein mixture (450 µg) was added to 96 wells of a microtiter plate containing the protein kinase B substrate. To initiate the reaction, ATP (final concentration 123 µM) was added to each well and the mixture was incubated for 90 min at 30°C. After three washes with PBS, the phospho-specific substrate antibody solution (40 µL) was added to each well and the mixture was incubated for 60 min at room temperature. After three washes with PBS containing 0.1% Triton X-100, each well containing the bound phospho-specific substrate antibody was incubated for 30 min at room temperature with a horseradish peroxidase-conjugated anti-rabbit IgG solution (40 µL, 1:1000). After three washes with PBS, each well was incubated for 45 min with 3,3 ,5,5 -tetramethylbenzidine substrate solution (60 µL) to visualize the AKT activity signal. The reaction was stopped by adding HCl to a final concentration of 250 mM, and the signal was then measured using a SpectraMax M2 fluorescence microplate reader (Molecular Devices, San Jose, CA, USA) at an absorbance wavelength of 450 nm.

CellMask Plasma Membrane Staining
Serum-starved U87MG cells were treated with PDGF-BB (50 ng/mL) for 5 min and incubated with a 1:1000 dilution of CellMask deep red plasma membrane stain (C10046, Life Technologies) for 5 min at 37 • C. Cells were washed with PBS three times to remove the staining solution and fixed in 4% paraformaldehyde (Electron Microscopy Sciences) in PBS for 10 min. The cells were then subjected to immunofluorescence staining analysis.

Statistics and Reproducibility
All experiments were repeated independently at least three times. Quantitative data are provided as the means ± SEM of triplicate determinations from representative experiments. Quantitative data were analyzed with Student's two-tailed t-test in Sigma Plot 10.0 and Excel in which a p-Value < 0.05 was considered statistically significant.

AKT Activation via Phosphorylation at Ser473 Requires Endocytosis
Endocytosis during receptor activation is classically involved in cell desensitization during chronic receptor activation and also clearly promotes the signal propagating function in acute response to extracellular stimuli [35,36]. We investigated whether endocytosis is required for PDGF-induced AKT activation in U87MG glioma cells, which display hyperactivated AKT signaling caused by phosphatase and tensin homolog (PTEN) deficiency [37,38], using endocytosis inhibitors. The administration of the receptor recycling inhibitor chlorpromazine, which prevents coated pit assembly at the plasma membrane [39], or an inhibitor of clathrin-mediated endocytosis, Pitstop 2 [40], specifically decreased AKT phosphorylation at Ser473 but not at Thr308 (Figure 1). The AKT kinase assay with lysates from PDGF-stimulated cells in the presence of chlorpromazine or Pitstop 2 showed that these endocytosis inhibitors elicited a reduction in AKT activity by~65% or~30%, respectively ( Figure 1E). These results indicate that endocytosis is critical for full AKT activation through Ser473 phosphorylation. To monitor the effect of chlorpromazine and Pitstop 2 on early endosome formation during PDGF-mediated endocytosis, we used bispecific monoclonal antibodies to PDGF receptor (PDGFR)β, as described in our previous work [32]. The inhibitory effect of chlorpromazine and Pitstop 2 on endosome formation was verified using immunofluorescence staining and monitoring endocytosed fluorescently labeled PDGFRs taking advantage of bispecific anti-mPDGFRβ × cotinine scFv-Cκ-scFv fusion protein ( Figure S1). Confocal imaging showed the newly formed endosomes, which come from receptor-mediated endocytosis after removal of fluorescently labeled antibodies against PDGFRs, unendocytosed on the plasma membrane surface by washing with an acetic acid buffer (pH = 2.7) and a reduction in the early endosome numbers by chlorpromazine and Pitstop 2 in a concentration-dependent manner. Growth factor-induced endocytosis plays a critical role in sustained epidermal growth factor (EGF) receptor signaling on AKT [41] and in the establishment and control of specific downstream signal transduction pathways [42]. Based on our results, PDGFRβ-mediated endocytosis is critical to AKT activation via phosphorylation at Ser473 which led us to investigate the mTORC2 assembly at the early endosome during PDGF stimulation.

mTORC2 Is Recruited into the Early Endosome upon Growth Factor Stimulation
Considering that mTORC2 is responsible for AKT phosphorylation at Ser473 [5,6,24], we next investigated whether key components of mTORC2 are localized in early endosomes. First, we analyzed the localization of mSIN, which is required for mTORC2 assembly and integrity [43], using immunofluorescence imaging. Many spots corresponding to mSIN staining colocalized with early endosomes identified by endocytosed PDGFRβ in U87MG glioma cells (Figure 2A). The number of Rab5-positive puncta indicating early endosomes increased in cells stimulated with PDGF, compared to starved cells ( Figure 2B), similar to the findings of a previous report on EGF-stimulated cells [44]. Colocalization of mSIN and Rab5 demonstrates that mSIN recruitment to early endosomes depends on growth factor activation ( Figure 2B), indicating that PDGFRβ activation results in the promotion of mTORC2 assembly in early endosomes. mSIN contains a Ras-binding domain (RBD, residues 276-354 in SIN1.1) and a PH domain (residues 377-488 in SIN1.1), which provides a putative binding site for phosphatidylinositol polyphosphates [45]. Therefore, we monitored the localization of GFP-mSIN-RBD-PH (residues 276-488 in SIN1.1) in U87MG glioma cells under PDGF-BB activation using confocal live-cell imaging ( Figure 2C). We found that GFP-mSIN-RBD-PH-positive puncta increased by~50% near the plasma membrane in the presence of PDGF ( Figure 2D and Video S1). We verified that GFP-mSIN-RBD-PHpositive puncta correspond to early endosomes containing constitutively active GTPase (Rab5-Q79L) using a colocalization experiment ( Figure 2E). Imaging showed that Rictor, an obligate component of mTORC2, also partially localized to early endosomes (Rab5-positive endosomes [46]), indicating that active mTORC2 originates in early endosomes during growth factor activation ( Figure 2F).
Moreover, we used a biochemical approach for the purification of early endosomes to assess whether components of mTORC2 exist in these cellular organelles in the absence or presence of PDGF-BB. To isolate early endosomes containing Rab5-GTP, we purified recombinant GST-EEA1-Rab5-GTP-binding domain (GST-EEA1-Rab5BD, residues 36-218 in EEA1, comprising C2H2 Zn 2+ Finger (residues 36-69) and coiled-coil domain (residues 74-218)) from bacteria. The human EEA1 C2H2 Zn 2+ finger domain was previously identified as an active Rab5A-binding region [47]. We added GST-EEA1-Rab5BD proteins to the lysates of U87MG cells before and after PDGF activation, for 5 min each time, and isolated active Rab5 endosomes with glutathione (GSH) beads ( Figure 3A). The immunoblot analysis of purified early endosomal fractions revealed that mSIN, Rictor, and mTORCcomponents of mTORC2-were associated with Rab5-GTP-bound endosomes but not Raptor, a component of mTORC1 ( Figure 3B-F). In addition, phosphorylated AKT at Ser473 via mTORC2 was observed in early endosomes. Late endosome-specific Rab7 and lysosome-resident cathepsin D were not detected in early endosomes purified with the Rab5-GTP-binding domain. This result shows that subcellular AKT activation by early endosome-associated mTORC2 is distinct from that by mTORC2 at lysosomes and late endosomes during growth factor stimulation [8] and by lysosome-dependent mTORC1 in response to various stimuli [17,48]. Furthermore, confocal imaging showed that most puncta corresponding to mSIN or Rictor did not colocalize with the plasma membrane ( Figure S2), indicating mTORC2 assembly in early endosomes. Endocytosis is critical for acute phosphorylation of AKT at Ser473 but not at Thr308: (A) serum-starved U87MG cells were pretreated with chlorpromazine (2.5 μg/mL, 5 μg/mL) for 3 min and incubated with platelet-derived growth factor-BB (PDGF-BB) (5 ng/mL) in the presence or absence of chlorpromazine for 3 min. Cell lysates were subjected to immunoblot analysis with antibodies specific to the indicated proteins; (B) the relative immunoblot signal intensity of pAKT(Ser473) and pAKT(Thr308) normalized over total AKT was determined as the mean ± SEM from three independent experiments, and p-values were calculated using Student's two-tailed t-test. * p < 0.05, *** p < 0.001, N.S., not significant; (C) serum-starved U87MG cells were pretreated with Pitstop 2 (2.5 μM, 5 μM) for 5 min and incubated with PDGF-BB (5 ng/mL) in the presence or absence of Pitstop 2 for 3 min. Cell lysates were subjected to immunoblot analysis with antibodies specific to the indicated proteins; (D) the relative immunoblot signal intensity of pAKT(Ser473) and pAKT(Thr308) normalized over total AKT was determined as the mean ± SEM from three independent experiments, and p-values were calculated using Student's two-tailed t-test. ** p < 0.01, *** p < 0.001, N.S., not significant; (E) serum-starved U87MG cells were pretreated with chlorpromazine (5 μg/mL) for 3 min or Pitstop 2 (5 μM) for 5 min and incubated with PDGF-BB (5 ng/mL) in the presence or absence of chlorpromazine (5 μg/mL) or Pitstop 2 (5 μM) for 3 min. Cell lysates were assayed for AKT kinase activity using the AKT Kinase Activity Kit (Enzo Life Sciences). Data are presented as the mean ± SEM from three experiments, and p-values were calculated using Student's two-tailed t-test. *p < 0.05, ** p < 0.01, *** p < 0.001. Figure 1. Endocytosis is critical for acute phosphorylation of AKT at Ser473 but not at Thr308: (A) serum-starved U87MG cells were pretreated with chlorpromazine (2.5 µg/mL, 5 µg/mL) for 3 min and incubated with platelet-derived growth factor-BB (PDGF-BB) (5 ng/mL) in the presence or absence of chlorpromazine for 3 min. Cell lysates were subjected to immunoblot analysis with antibodies specific to the indicated proteins; (B) the relative immunoblot signal intensity of pAKT(Ser473) and pAKT(Thr308) normalized over total AKT was determined as the mean ± SEM from three independent experiments, and p-values were calculated using Student's two-tailed t-test. * p < 0.05, *** p < 0.001, N.S., not significant; (C) serum-starved U87MG cells were pretreated with Pitstop 2 (2.5 µM, 5 µM) for 5 min and incubated with PDGF-BB (5 ng/mL) in the presence or absence of Pitstop 2 for 3 min. Cell lysates were subjected to immunoblot analysis with antibodies specific to the indicated proteins; (D) the relative immunoblot signal intensity of pAKT(Ser473) and pAKT(Thr308) normalized over total AKT was determined as the mean ± SEM from three independent experiments, and p-values were calculated using Student's two-tailed t-test. ** p < 0.01, *** p < 0.001, N.S., not significant; (E) serum-starved U87MG cells were pretreated with chlorpromazine (5 µg/mL) for 3 min or Pitstop 2 (5 µM) for 5 min and incubated with PDGF-BB (5 ng/mL) in the presence or absence of chlorpromazine (5 µg/mL) or Pitstop 2 (5 µM) for 3 min. Cell lysates were assayed for AKT kinase activity using the AKT Kinase Activity Kit (Enzo Life Sciences). Data are presented as the mean ± SEM from three experiments, and p-values were calculated using Student's two-tailed t-test. * p < 0.05, ** p < 0.01, *** p < 0.001.   , and GSH-bead-bound fractions were precipitated. PNS and GSH-bead precipitates were subjected to immunoblot analysis with antibodies specific to the indicated proteins; (C-F) The relative immunoblot signal intensity of GSH-bead-bound mSIN, Rictor, Rab5, and EEA1 (a marker for Rab5-positive endosomal fractions) normalized by that of GST were determined as the mean ± SEM from three independent experiments, and p-values were calculated using Student's two-tailed t-test. * p < 0.05, ** p < 0.01, *** p < 0.001. Two components (mSIN and Rictor) of mTORC2 were associated with early endosomes.

Localization of mTORC2 to Early Endosomes Is Mediated by PtdIns(3,4)P 2
Next, we investigated whether phosphatidylinositol polyphosphate mediates mTORC2 translocation into the early endosome. Previous studies proposed the distinct role of PtdIns(3,4)P 2 -dependent AKT phosphorylation at Ser473, in addition to PtdIns(3,4,5)P 3dependent AKT phosphorylation at Thr308 in AKT activation [28,29,49]. PI3K recruited to the plasma membrane by activated receptor tyrosine kinases sequentially generates PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 , and the accumulation of these molecules at the plasma membrane induces AKT phosphorylation at Thr308 and Ser473, respectively [28,29]. In turn, the dual phosphorylation of AKT at these residues contributes to its full activity [28,29,50]. Considering that AKT Ser473 is phosphorylated by mTORC2 [5] and blocking mTORC2 activity by genetic ablation of mSIN1 or Rictor-both of which are mTORC2-specific components-affects AKT substrate specificity [43,51], the physiological consequence of AKT activity regulation by mTORC2 must be further investigated. Cytosolic AKT proteins are recruited to and activated near the microdomain of the plasma membrane via the interaction of the AKT-PH domain with PtdIns(3,4,5)P 3 or PtdIns(3,4)P 2 [25,52]. To examine how mTORC2 is recruited into early endosomes, we compared the localization of mSIN-RBD-PH with that of AKT-PH using live-cell confocal imaging ( Figure 4A and Video S2). The fluorescence intensity of AKT-PH, which can detect both PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 , increased both at the plasma membrane and in the cytosol close to the inner leaflet of the plasma membrane Intriguingly, mSIN-RBD-PH signals colocalized with some AKT-PH signals only, exhibiting a spotty pattern inside the plasma membrane in a representative image of a live U87MG cell during PDGF-BB stimulation ( Figure 4A). Given that PtdIns(3,4)P 2 generation lags behind PtdIns(3,4,5)P 3 during extracellular stimulation [53,54] and that PtdIns(3,4)P 2 is produced by SH2-containing inositol phosphatase (SHIP) family proteins from PtdIns(3,4,5)P 3 after PI3K activation [55], we next investigated whether mSIN colocalizes to the early endosomes containing PtdIns(3,4)P 2 in PDGF-activated cells using an antibody against PtdIns(3,4)P 2 . The specificity of the antibody to PtdIns(3,4)P 2 was verified by an increase in the amount of PtdIns(3,4)P 2 around the plasma membrane in a PDGF-dependent manner and by the overlapping of the signal of the antibody with that of a known fluorescent PtdIns(3,4)P 2 indicator, GFP-TAPP1-PH [56] (Figure S3A,B). PtdIns(3,4)P 2 colocalized with mSIN, a component of mTORC2, and Rab5, an early endosomal marker, as puncta inside the plasma membrane of PDGF-activated U87MG cells ( Figure 4B-D). Structured illumination microscopy (SIM) revealed that mSIN colocalized with PtdIns(3,4)P 2 at a distance of 30-60 nm, thereby indicating the regions of mSIN that lack PtdIns(3,4)P 2 ( Figure 4E). We rarely observed PtdIns(3,4,5)P 3 in Rab5-positive endosomes and their colocalization with mSIN in PDGF-BB-stimulated cells ( Figure S3C,D).

Forced Depletion of Endosomal PtdIns(3,4)P 2 by Targeted Phosphatases Causes Reduced AKT Phosphorylation at Ser473
The presence of mTORC2 in early endosomes motivated us to test the possibility that the endosomal pool of mTORC2 is required for AKT activation and depends on the accumulation of PtdIns(3,4)P 2 . To this end, we investigated if the specific removal of endosomal phosphoinositide by the rapamycin-inducible targeting of phosphatases into early endosomes affects AKT activation. We selected myotubularin 1 (MTM1) as a PtdIns 3-phosphatase [57], INPP4B as a PtdIns 4-phosphatase [58], and Inp54p as a PtdIns 5-phosphatase [59]. Furthermore, we used the rapamycin-dependent heterodimerization of two components (iRFP-FRB-Rab5 and mCherry-FKBP-MTM1 for 3-phosphatase recruitment; iRFP-FRB-Rab5 and mCherry-FKBP-INPP4B for 4-phosphatase recruitment; iRFP-FRB-Rab5 and CFP-FKBP-Inp54p for 5-phosphatase recruitment) of the system with some modifications, as previously described [60], to selectively deplete each phosphoinositide in endosomes ( Figure 5A). The recruitment of each lipid phosphatase (MTM1, INPP4B, and Inp54p) into early endosomes in the presence of rapamycin was verified using confocal microscopy ( Figure S4 and Video S3). The level of AKT Ser473 phosphorylation in early endosomes was reduced by~50% in cells expressing endosome-targeting MTM1 (3-phosphatase) or INPP4B (4-phosphatase), but in cells expressing Inp54p (5-phosphatase), it remained unaltered in the presence of rapamycin during PDGF-BB stimulation ( Figure 5B,C). Image quantification shows that depletion of PtdIns(3,4)P 2 , an endosomal phosphoinositide containing two phosphates at the 3rd and 4th position of the inositol ring, inhibited local AKT activation through Ser473 phosphorylation. This result indicates that AKT activation via mTORC2 depends on PtdIns(3,4)P 2 , but not PtdIns(3,4,5)P 3 , at the early endosome, which confirms findings from a previous report [27] that mTORC2 is targeted to the cellular membrane compartment in a PtdIns(3,4,5)P 3 -independent manner. Since the level of AKT Thr308 phosphorylation remained consistent in cells expressing each lipid phosphatase in the presence of rapamycin ( Figure 5D), we speculate that AKT activation by Thr308 phosphorylation does not depend on endosomal phosphoinositide and results predominantly from the accumulation of PtdIns(3,4,5)P 3 at the plasma membrane [24,61]. We verified that the rapamycin-induced localization of 3-phosphatase or 4-phosphatase to early endosomes results in reduced endosomal PtdIns(3,4)P 2 but that of 5-phosphatase dose not ( Figure 5E,F). The activity of FKBP-MTM1, FKBP-INPP4B, and FKBP-Inp54p has been well demonstrated in previous works [60,[62][63][64][65]. We investigated if rapamycin treatment (40 nM, 5 min incubation) for lipid phosphatase targeting may indirectly affect AKT activation in our experiments, as it was shown that rapamycin inhibits mTORC1 activity acutely [66] and mTORC2 activity chronically [67]. AKT phosphorylation of both Ser473 and Thr308 remained unaltered by rapamycin treatment during the 5 min PDGF stimulation ( Figure S5). Taken together with the presence of mTORC2 components in early endosomes (Figures 2-4), the targeting of specific phosphoinositide phosphatases with the rapamycin-dependent heterodimerization strongly indicates that AKT activation by Ser473 phosphorylation requires an increase in the amount of PtdIns(3,4)P 2 and mTORC2 in early endosomes of cells activated with growth factors.

Discussion
In this study, we first demonstrated that in PDGF-activated cells, endocytosis is required for AKT phosphorylation at Ser473 but dispensable for AKT phosphorylation at Thr308. Given that the sequential phosphorylation of AKT at both Thr308 and Ser473 is critical for the fidelity of AKT signal transduction and the specificity of downstream substrate phosphorylation [28,29,43,50,51], this finding supports the fact that the signaling endosomes originate from endocytosis and provide platforms that enable intracellular signal propagation and maintenance [35,36,41]. Secondly, we showed that PtdIns(3,4)P 2 molecules are key factors for the interaction of AKT with mTORC2 via mSIN in early endosomes. Therefore, local mTORC2 activity in early endosomes is responsible for AKT Ser473 phosphorylation and the activation of a subset of AKT substrates, such as GSK3β and FoxO1/3a. In glioma cells, we demonstrate that an increase in endosomal PtdIns(3,4)P 2 in response to PDGF-BB is a prerequisite for AKT phosphorylation at Ser473 but not at Thr308 using a drug-inducible heterodimerization system and for targeting mTORC2 into endosomes via mSIN, which contains an RBD-PH domain, to specifically bind PtdIns(3,4)P 2 . Thirdly, we highlighted the consequence of PDGF-induced AKT activation in endosomes during endocytosis. Newly formed signaling endosome induced AKT activation results in the increased specificity by inducing the phosphorylation of GSK3β and FoxO1/3a without affecting mTORC1 signaling. Here, we also explore the fact that the control of AKT signaling in subcellular compartments, such as ribosomes, MAMs, and lysosomes, was reported to have advantages to orchestrate growth and metabolism [8,21,22].
A model illustrating the mechanism underlying PtdIns(3,4)P 2 accumulation in early endosomes and its role in AKT signaling is shown in Figure 7.
GSK3β and FoxO1/3a without affecting mTORC1 signaling. Here, we also explore the fact that the control of AKT signaling in subcellular compartments, such as ribosomes, MAMs, and lysosomes, was reported to have advantages to orchestrate growth and metabolism [8,21,22].
A model illustrating the mechanism underlying PtdIns(3,4)P2 accumulation in early endosomes and its role in AKT signaling is shown in Figure 7. Proposed scheme for the mechanism of AKT activation by mTORC2 in early endosomes during growth factor stimulation. PDGF-dependent PI3K activation triggers AKT phosphorylation through PDK1 and mTORC2. PtdIns (3,4,5)P3 molecules produced by class I PI3K at the plasma membrane provide a binding platform for PDK1 and AKT through each protein's PH domain. PtdIns(3,4)P2 molecules produced predominantly by SHIP1/2 5-phosphatase from PtdIns(3,4,5)P3 play a role in connecting mTORC2 to AKT by providing a binding site for mSIN, a component of the complex, and AKT. This local activation of AKT is required for a subset of substrates represented by GSK3β and FoxO1/3a.
In this study, we propose that endosomal PtdIns(3,4)P 2 molecules play a key role in the mTORC2 mediated activation of AKT. AKT is most frequently hyperactivated in human cancers by mutations affecting upstream regulators, but AKT inhibitor treatment use has been restricted as the direct inhibition of AKT activity can lead to high cytotoxicity. PtdIns(3,4)P 2 is predominantly produced by class I PI3 kinase and SHIP 5-phosphatase activated sequentially from PtdIns(4,5)P 2 during growth factor stimulation [62]. Our study provides a rationale for a different approach to inhibit mTORC2 activity via reducing the endosomal PtdIns(3,4)P 2 level for AKT-targeting therapeutic agents with minimal side effects by avoiding the inhibition of the downstream AKT signaling broadly.

Conclusions
In glioma cells, we demonstrate AKT activation in early endosomes during PDGF stimulation. mTORC2, which is associated with AKT activation, is recruited into endosomes by PtdIns(3,4)P 2 . We also show that endocytosis is key to AKT activity and endosomal AKT activation regulates AKT substrates such as GSK3β and FoxO1/3.