Phosphatidylinositol 3-Kinase Class II α-Isoform PI3K-C2α Is Required for Transforming Growth Factor β-induced Smad Signaling in Endothelial Cells*

Background: TGFβ receptor signals through Smad phosphorylation, which is dependent on endocytosis of TGFβ receptors and the Smad anchor protein SARA localized on endosomes. Results: Class II PI3K-C2α is necessary for TGFβ receptor endocytosis into SARA-containing endosomes, SARA-Smad complex formation, and Smad phosphorylation. Conclusion: PI3K-C2α serves endosomal TGFβ receptor signaling. Significance: PI3K-C2α is a key molecule that is generally engaged in endosomal receptor signaling. We have recently demonstrated that the PI3K class II-α isoform (PI3K-C2α), which generates phosphatidylinositol 3-phosphate and phosphatidylinositol 3,4-bisphosphates, plays crucial roles in angiogenesis, by analyzing PI3K-C2α knock-out mice. The PI3K-C2α actions are mediated at least in part through its participation in the internalization of VEGF receptor-2 and sphingosine-1-phosphate receptor S1P1 and thereby their signaling on endosomes. TGFβ, which is also an essential angiogenic factor, signals via the serine/threonine kinase receptor complex to induce phosphorylation of Smad2 and Smad3 (Smad2/3). SARA (Smad anchor for receptor activation) protein, which is localized in early endosomes through its FYVE domain, is required for Smad2/3 signaling. In the present study, we showed that PI3K-C2α knockdown nearly completely abolished TGFβ1-induced phosphorylation and nuclear translocation of Smad2/3 in vascular endothelial cells (ECs). PI3K-C2α was necessary for TGFβ-induced increase in phosphatidylinositol 3,4-bisphosphates in the plasma membrane and TGFβ receptor internalization into the SARA-containing early endosomes, but not for phosphatidylinositol 3-phosphate enrichment or localization of SARA in the early endosomes. PI3K-C2α was also required for TGFβ receptor-mediated formation of SARA-Smad2/3 complex. Inhibition of dynamin, which is required for the clathrin-dependent receptor endocytosis, suppressed both TGFβ receptor internalization and Smad2/3 phosphorylation. TGFβ1 stimulated Smad-dependent VEGF-A expression, VEGF receptor-mediated EC migration, and capillary-like tube formation, which were all abolished by either PI3K-C2α knockdown or a dynamin inhibitor. Finally, TGFβ1-induced microvessel formation in Matrigel plugs was greatly attenuated in EC-specific PI3K-C2α-deleted mice. These observations indicate that PI3K-C2α plays the pivotal role in TGFβ receptor endocytosis and thereby Smad2/3 signaling, participating in angiogenic actions of TGFβ.

coupled receptor S1P 1 (8,15,16). Signaling of VEGFR2 and S1P 1 was defective in PI3K-C2␣-depleted EC: the receptor endocytosis was inhibited, and the signaling on endosomes, particularly Rho GTPase activation, was impaired. These defects result in impaired migration, proliferation, and intercellular junction formation in EC. It is unknown whether and how PI3K-C2␣ regulates signaling of other angiogenic receptors. In addition to our studies, a general regulatory role for PI3K-C2␣ in endocytosis through the generation of PtdIns(3,4)P 2 in the plasma membrane was recently reported (14).
TGF␤ is involved in the regulation of migration and proliferation of EC, production of basement membrane, and differentiation and recruitment of mural cells, thus being essential for normal vascular formation (17)(18)(19)(20). TGF␤ signals through type I and type II TGF␤ receptors, which are both serine/threonine transmembrane kinases (21)(22)(23). TGF␤ binds to type II receptor, which phosphorylates and activates type I receptors, activin receptor-like kinase (ALK) 1, and ALK5. ALK1 and ALK5 in turn phosphorylate the receptor-regulated Smads, Smad1 and Smad5 (Smad1/5) and Smad2 and Smad3 (Smad2/3), respectively. Phosphorylated receptor-regulated Smads form complexes with the common mediator Smad4 and the Smad complexes translocate into the nucleus to regulate gene transcription. It was proposed that TGF␤ signaling pathways via ALK1 and ALK5 in EC may play a balancing role for controlling proliferation and migration of EC during angiogenesis (24,25). Of the two TGF␤ signaling pathways, EC-specific gene ablation of either ALK5 or Smad2/3 resulted in the similar vascular abnormalities, indicating a pivotal role of endothelial ALK5-Smad2/3 pathway in the angiogenic effect of TGF␤ (19,20,26,27). SARA (Smad anchor for receptor activation) protein contains the binding domains for both Smad2/3 and the TGF␤ receptor complex and is localized in the early endosomes through its FYVE domain, which specifically recognizes and binds to PtdIns(3)P (28). Previous studies (28 -31) demonstrated that upon TGF␤ stimulation, the TGF␤ receptor complex undergoes clathrin-dependent endocytosis into the early endosomes containing SARA and that the proper localization of SARA in the early endosomes and the TGF␤ receptor internalization into the SARA-containing endosomes are the events necessary for TGF␤-induced phosphorylation of Smad2/3 and the following nuclear translocation of the Smad complexes. It is likely that PI3Ks are involved in TGF␤ receptor internalization, the endosomal localization of SARA, and thus TGF␤ signaling. However, it is unknown which isoform of PI3K is engaged in the processes of TGF␤ signaling.
In the present study, we studied a role for PI3K-C2␣ in TGF␤-induced Smad2/3 signaling in EC. We found that TGF␤induced Smad2/3 phosphorylation, Smad2/3-dependent gene expression, and angiogenic responses were strongly dependent on PI3K-C2␣. PI3K-C2␣ was required for TGF␤ receptor internalization but not the endosomal localization of SARA. These observations suggest that PI3K-C2␣ plays an indispensable role in endosomal TGF␤ receptor signaling.
Immunoblotting and Immunoprecipitation Analysis-At 48 h after siRNA transfection, the cells were serum-starved with M199 (Invitrogen Gibco) containing 0.5% fatty acid-free BSA (catalog no. A6003; Sigma-Aldrich) for 4 h and then stimulated with 5 ng/ml TGF␤1 (catalog no. 240-B; R&D Systems, Minneapolis, MN). The cells were washed in PBS and lysed in the cell lysis buffer (20 mM Tris-HCl, pH 7.2, 150 mM NaCl, 1 mM CaCl 2 , 0.5% Triton X-100, 100 mM NaF, 1 mM Na 3 VO 4 ) supplemented with Complete Protease inhibitor mixture (Roche Applied Science) by scraping, followed by centrifugation for 15 min at 16, 000 ϫ g at 4°C. The resultant supernatants were taken, electrophoresed on 8% SDS-PAGE, and transferred onto PVDF membrane (Millipore, Billerica, MA). The membranes were blocked in PBS containing 5% BSA and incubated with respective antibodies overnight. A1978; Sigma-Aldrich). The membranes were incubated with alkaline phosphatase-conjugated secondary antibodies (antimouse IgG antibody, catalog no. 7056; anti-rabbit IgG antibody, catalog no. 7054) (Cell Signaling) and visualized by color reaction using 5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium (Wako, Osaka, Japan). The band intensities were determined using Image Gauge (Fuji Film, Tokyo, Japan). The values were normalized for the value of ␤-actin as a loading control and expressed as multiples over the normalized values of untreated controls.
For immunoprecipitation assay, HEK293T cells were cotransfected with the expression vectors for FLAG-Smad3, either FLAG-SARA or FLAG-⌬SBD-SARA, and either FLAG-wtALK5 or FLAG-constitutively activated ALK5 (caALK5) and 72 h later were lysed in IP buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS) supplemented with Complete Protease inhibitor cocktails. The lysates were incubated with anti-SARA antibody for 1 h at 4°C with rocking, followed by the incubation with protein G-agarose beads (catalog no. 1-719-416; Roche) for 1 h at 4°C. After the beads were washed five times, they were mixed with 2ϫ Laemmli's SDS sample buffer and boiled. The resultant samples were analyzed with immunoblotting using respective antibodies.
Immunohistochemistry and Immunofluorescence Staining-HUVECs were plated onto type I collagen-coated glass bottom dishes (MatTek Corporation, Ashland, MA) and allowed to adhere to dishes in EGM-2 growth medium overnight. The cells were rinsed with prewarmed PBS once and fixed in prewarmed 4% fresh paraformaldehyde in PBS for 10 min, washed with PBS, and then permeabilized in 0.2% Triton X-100 in PBS for 15 min when necessary. After the cells were incubated with 5% normal goat serum for 60 min to inhibit nonspecific protein binding, the cells were incubated with rabbit polyclonal anti-p-Smad2 antibody (catalog no. AB3849; Millipore), mouse monoclonal anti-Smad2/3 antibody (catalog no. 610842; BD Biosciences), rabbit polyclonal anti-SARA antibody, or mouse monoclonal anti-EEA1 antibody (catalog no. 610456; BD Biosciences) for 2 h at room temperature or overnight at 4°C. The cells were incubated for 60 min at room temperature with Alexa Fluor 488-conjugated goat anti-mouse (catalog no. A31620; Molecular Probes), Alexa Fluor 488-conjugated goat anti-rabbit (catalog no. A11034), Alexa Fluor 594-conjugated goat antimouse (catalog no. A31624), Alexa Fluor 594-conjugated goat anti-rabbit (catalog no. A31620), and secondary antibodies diluted at 1:1000 in PBS. Where appropriate, the cells were counterstained with DAPI (catalog no. D1306; Molecular Probes) for 5 min. The cells were mounted on Fluoromount (catalog no. K024; Diagnostic BioSystems, Pleasanton, CA) and observed under a custom confocal microscope unit as described in detail previously (8). For immunohistochemistry of the sections of Matrigel plugs, the sections of paraformaldehyde-fixed, paraffinembedded Matrigel plug were deparaffinized and processed in heat-induced target retrieval to unmask the antigen using with target retrieval solution (Dako, Carpinteria, CA) (30). The sections were incubated with Dako blocking solution (catalog no. X0909; Dako) for 10 min to inhibit nonspecific protein binding. After blocking, the sections were stained with rabbit polyclonal anti-von Willebrand factor (vWF) (catalog no. A0082; Dako) for 60 min at room temperature. The sections were incubated with the secondary antibody of an EnVision kit (catalog no. K4002; Dako) for 60 min, and the color reaction was developed. Where appropriate, the sections were counterstained with hematoxylin. The sections were examined using a BX41 inverted microscope (Olympus, Tokyo, Japan), and vWF-positive microvessel numbers were determined with ImageJ software.
Proximity Ligation Assay (PLA) Staining-The cells were fixed in prewarmed 4% fresh paraformaldehyde in PBS for 10 min and permeabilized in 0.2% Triton X-100 in PBS for 15 min when necessary. After the cells were incubated with rabbit polyclonal anti-ALK5 antibody (catalog no. sc-398; Santa Cruz), mouse monoclonal anti-Smad2/3 antibody, and mouse monoclonal anti-SARA (catalog no. sc-133071; Santa Cruz) antibody overnight at 4°C, in situ protein interactions were detected using the Duolink proximity ligation assay kit according to the manufacturer's instructions (Olink Bioscience, Uppsala, Sweden). The cells were stained with anti-EEA1-Alexa Fluor 594 (M176 A59; MBL, Nagoya, Japan).
RNA Isolation and Quantitative PCR Analysis-Total RNA in HUVECs was isolated using TRIzol reagent (Invitrogen). One g of total RNA was reverse-transcribed into the first strand cDNA using QuantiTect RT Kit (catalog no. 205311; Qiagen). Quantitative real time PCRs were performed using FAM-conjugated TaqMan inventoried assay from Applied Biosystems for human PI3K-C2␣ (Hs0090461_m1) and human VEGF-A (Hs00900055_m1). 18 S rRNA (Hs99999901_s1) probe was used as an internal control. The mRNA expression levels were normalized for the expression of 18 S rRNA mRNA, and the results were expressed as multiples over control values. Comparative quantitative analysis was performed using the GeneAmp 7300 system (Applied Biosystems, Foster City, CA) based on the ⌬⌬Ct method.
Matrigel Plug in Vivo Angiogenesis Assay-All of the mice used in this study were bred and maintained at the Institute for Experimental Animals, Advanced Science Research Center, Kanazawa University under specific pathogen-free conditions. All procedures were conducted in accordance with the Fundamental Guidelines for Proper Conduct of Animal Experiment and Related Activities in Academic Research Institutions under the jurisdiction of the Ministry of Education, Culture, Sports, Science and Technology of Japan and approved by the Committee on Animal Experimentation of Kanazawa University. Pik3c2a ⌬EC (C2␣ ⌬EC ) (Pik3c2a flox/flox ; Cdh5(PAC)-CreER T2 ) and Pik3c2a ⌬SMC (C2␣ ⌬SMC ) (Pik3c2a flox/flox ; SM22a-Cre) mice were described previously (8). Cre-negative littermates were used as controls. To verify the efficiency of Cre recombination, Cre mice were mated with mice from the Cre reporter transgenic line ROSA26-LacZ (B6.129S4-Gt(ROSA)26Sortm1Sor/J, Jackson Lab). All mice had a C57BL/6J genetic background. For Pik3c2a gene inactivation in adult mice, tamoxifen (10 mg/ml corn oil) (catalog no. T5648; Sigma-Aldrich) was administered seven times by intraperitoneal injection of 100 l of tamoxifen solution. For Matrigel plug assay (32,36), recombinant VEGF-A (200 ng/ml), FGF2 (400 ng/ml) (catalog no. AF-100-18B; PeproTech), and heparin (100 mg/ml) (Sigma-Aldrich) were mixed with growth factor-reduced Matrigel. The Matrigel solutions (300 l each) were injected subcutaneously into the groin area close to the dorsal midline (most angiogenic portion) of anesthetized mice. Matrigel plugs were harvested on day 10 and fixed overnight in 4% paraformaldehyde for paraffin embedding and the following immunohistochemistry.
VEGF-A ELISA Assay-Human VEGF-A protein levels in the conditioned medium of HUVEC cultures were determined using human VEGF-A ELISA immunoassay (catalog no. DVE00; R&D Systems, Minneapolis, MN) according to the manufacturer's protocol. Optical density was measured at 450 nm using a 540-nm correction with a Multiskan GO (Thermo Fisher Scientific, Walyham, MA).
Statistical Analysis-The data are presented as means Ϯ S.E. and expressed as the percentages or multiples relative to the values in control cells. Statistical significance was analyzed using Prism 5 software (GraphPad Software Inc., San Diego, CA). Statistical significance was analyzed either by one-way or two-way analysis of variance followed by Bonferroni test as appropriate. Results with p Ͻ 0.05 were considered statistically significant.

TGF␤1-induced TGF␤ Receptor Internalization into SARAcontaining Endosomes Is Dependent on PI3K-C2␣-TGF␤1
stimulation induced an increase in serine phosphorylation of ALK5 in HEK293T cells transfected with wtALK5 (Fig. 4A). In HEK293T cells, we observed PI3K-C2␣ dependence of TGF␤1induced Smad2/3 phosphorylation (data not shown). PI3K-C2␣ knockdown did not inhibit TGF␤-induced serine phosphorylation of wtALK5, implying that PI3K-C2␣ is necessary for the TGF␤ receptor signaling step, which is distal to phosphorylation of type I TGF␤ receptor. Previous studies (28 -31) showed that TGF␤1 stimulation triggered clathrin-dependent endocytosis of TGF␤ receptor and that TGF␤ receptor endocytosis was required for TGF␤ activation of Smad2/3 signaling. We tested the effect of dynasore, an inhibitor of dynamin that is necessary for clathrin-dependent endocytosis, in EC. Treatment of HUVECs with dynasore abolished TGF␤1-induced phosphorylation of Smad2/3, like an ALK inhibitor (Fig. 4B). Likewise, the expression of the dominant negative dynamin2 mutant but not wild-type dynamin2 inhibited nuclear translocation of Smad2/3 (Fig. 4C). These observations suggested that TGF␤ receptor endocytosis was required for Smad2/3 signaling. TGF␤1 stimulation promoted the internalization of type I TGF␤ receptor ALK5 into the intracellular compartment, which was prevented by PI3K-C2␣ knockdown (Fig. 5A). Likewise, dynasore suppressed TGF␤1-induced ALK5 internalization. In these immunostainings, the anti-ALK5 antibody stained nuclei. ALK5 knockdown did not abolish or reduce the nuclear staining in anti-ALK5 immunostaining (Fig. 5B), suggesting that the nuclear staining was nonspecific. In sc-siRNA-transfected HUVECs, the expression of kdC2␣ r partially inhibited TGF␤1-induced ALK5 internalization, differently from that of wtC2␣ r expression (Fig.  5C). In C2␣-siRNA-transfected HUVECs, the expression of kdC2␣ r did not restore TGF␤1-induced ALK5 internalization. Double immunofluorescent staining of ALK5 and the early endosome marker EEA1 showed that TGF␤1 induced the internalization of ALK5 into the EEA1-positive early endosomes (Fig. 5D). PLA staining to detect interaction or close co-localization of two molecules showed that TGF␤1 stimulation induced the close colocalization of ALK5 and EEA1 (Fig. 5E, green dots), which was nearly abolished by PI3K-C2␣ knockdown. Because SARA is located in the early endosomes and acts as a scaffold for Smad2/3 phosphorylation by ALK5 (28 -30), we studied the requirement of SARA for TGF␤/ALK5 signaling, the possible co-localization of TGF␤1 receptors and SARA, and the effect of PI3K-C2␣ knockdown on the co-localization. Knockdown of SARA nearly completely suppressed TGF␤-induced Smad2/3 phosphorylation (Fig. 6A), indicating that SARA is essential for TGF␤/ALK5 signaling in HUVECs. Dou-

JOURNAL OF BIOLOGICAL CHEMISTRY
VOLUME 290 • NUMBER 10 • MARCH 6, 2015 ble immunofluorescent staining using anti-SARA and anti-EEA1 showed that SARA was localized mainly in the EEA1positive early endosome compartment in sc-siRNA-transfected HUVECs (Fig. 6B). TGF␤1 increased the SARA-and EEA1double positive endosomes. PI3K-C2␣ knockdown did not affect the numbers of either SARA-positive or EEA1-positive vesicles in nonstimulated cells but abolished TGF␤1-induced increase in SARA-and EEA1-double positive early endosomes. PLA staining for ALK5 and SARA, combined with anti-EEA1 immunostaining, showed that TGF␤1 stimulation induced the close co-localization of ALK5 and SARA and that a portion of the closely co-localized ALK5 and SARA existed in the EEA1positive early endosomes (Fig. 6C). Knockdown of PI3K-C2␣ inhibited the close co-localization of ALK5 and SARA in EEA1positive early endosomes. These observations indicate that PI3K-C2␣ is involved in TGF␤1-induced ALK5 internalization into the SARA-containing endosomes through its kinase activity.
PI3K-C2␣ Is Required for SARA-Smad2/3 Complex Formation-Because SARA is associated with Smad2/3 and acts as a scaffold for Smad2/3 phosphorylation by type I TGF␤ receptor ALK5, we studied the PI3K-C2␣ dependence of SARA-Smad2/3 complex formation, using anti-SARA immunoprecipitation and the following anti-Smad2/3 immunoblotting. For the experiments, we employed HEK293T cells for efficiency of gene transduction. We co-transfected HEK293T cells with the expression vectors for either wtSARA or the ⌬SBD-SARA and either wtALK5 or caALK5, with or without Smad2/3 expression vectors. We detected co-immunoprecipitation of Smad2/3 in the anti-SARA immunoprecipitates in the cells transfected with Smad2/3, wtSARA, and wtALK5. However, without Smad2/3 transfection, we did not detect Smad2/3 in the anti-SARA immunoprecipitates (Fig. 8A). In sc-siRNA-treated control cells that had been transfected with wtSARA and wtALK5, we detected the association of Smad3 and Smad2 with SARA (Fig. 8, B and C). The expression of caALK5 substantially stimulated the association of Smad3 and Smad2 with SARA and resultant phosphorylation of Smad3 and Smad2, which were both markedly inhibited by the expression of ⌬SBD-SARA. In contrast, in PI3K-C2␣-depleted cells, caALK5 expression barely stimulated the association of Smad3 and Smad2 with SARA and phosphorylation of Smad3 and Smad2. Thus, PI3K-C2␣ is required for ALK5-mediated formation of the SARA and Smad2/3 complex and phosphorylation of Smad2/3.
We also studied the interaction of endogenous SARA and Smad2/3 in HUVECs using PLA staining. TGF␤1 promoted SARA-smad2/3 interaction in the endosomes in sc-siRNAtransfected control HUVECs. A portion of the PLA signal was co-localized with EEA1, indicating that SARA and Smad2/3 complex were located in the early endosomes. PI3K-C2␣ knockdown inhibited TGF␤1-stimulated SARA-smad2/3 interaction (Fig.  8D). The observations indicate that TGF␤1 stimulation of the  MARCH 6, 2015 • VOLUME 290 • NUMBER 10 JOURNAL OF BIOLOGICAL CHEMISTRY 6095 interaction of endogenous SARA and Smad2/3 in the endosomes requires PI3K-C2␣ in HUVECs. (38 -41), TGF␤1 increased the expression of mRNA and protein of VEGF-A in control HUVECs (Fig. 9, A-C). The stimulatory effects of TGF␤1 were inhibited by knockdown of the common Smad Smad4, suggesting the involvement of Smad2/3. Furthermore, TGF␤1-induced VEGF-A expression was abolished by the pharmacological blockade of ALK5 (Fig. 9D). These observations together suggested that TGF␤1-induced VEGF-A up-regulation was dependent on the ALK5-Smad pathway. In agreement with the involvement of the canonical Smad pathway in TGF␤1-induced VEGF-A expression, knockdown of PI3K-C2␣, but not PI3K-C2␤, p110␣, or Vps34, inhibited TGF␤1induced VEGF-A expression (Fig. 9, A-C). Moreover, treat-   PI3K-C2␣ Is Required for TGF␤1-induced Endothelial Cell Migration, Tube Formation, and in Vivo Angiogenesis-In a wound healing assay, PI3K-C2␣ knockdown inhibited migration of HUVECs induced by either TGF␤1 or VEGF-A (Fig. 10,  A and B). The ALK5 inhibitor suppressed TGF␤1-induced cell migration. Interestingly, the inhibitor of VEGFR2 suppressed not only VEGF-induced but also TGF␤1-induced cell migration, indicating that TGF␤1-induced cell migration is dependent on VEGFR2. Likewise, PI3K-C2␣ knockdown inhibited tube formation induced by either TGF␤1 or VEGF-A (Fig. 10, C  and D). The inhibition of VEGF-A-induced tube formation by PI3K-C2␣ knockdown is most likely because VEGFR2 signaling is dependent on PI3K-C2␣ as we demonstrated previously (8). The ALK5 inhibitor and the VEGFR2 inhibitor blocked TGF␤1-induced tube formation (Fig. 10, E and F). In addition, dynasore suppressed TGF␤1-induced tube formation. These observations suggest that TGF␤1-induced, ALK5-mediated stimulation of endothelial migration and morphogenesis is dependent on stimulation of VEGF-A expression and VEGFR2 signaling, in which PI3K-C2␣ and the endocytic process are involved.

PI3K-C2␣ Is Required for TGF␤1-induced VEGF-A Expression in EC-Consistent with previous studies
We finally investigated a role for PI3K-C2␣ in in vivo angiogenesis, using a Matrigel plug assay in conditional PI3K-C2␣ knock-out mice. We subcutaneously injected Matrigel plug with or without TGF␤1 in mice with endothelial specific deletion of PI3K-C2␣ (C2␣ ⌬EC ) or smooth muscle-specific PI3K-C2␣ deletion (C2␣ ⌬SMC ) and compared angiogenesis in both mutant mice with that in control mice (C2␣ flox/flox ). The inclusion of TGF␤1 in Matrigel increased the formation of anti-vWF-positive microvessels in Matrigel plugs in control mice, compared with vehicle (Fig. 11, A-F). In contrast, in C2␣ ⌬EC mice TGF␤1 failed to stimulate microvessel formation. In C2␣ ⌬SMC mice, however, TGF␤1 stimulated microvessel formation in Matrigel plugs to the similar extent as in control mice (Fig. 11, D-F). We performed double immunofluorescent staining of p-Smad2 and EC marker CD31 in Matrigels containing either TGF␤ or vehicle that had been implanted in C2a DEC and C2a DSMC mice. C2a DEC mice showed much reduced p-Smad2-and CD31-double positive cells compared with C2a DSMC (Fig. 11G), suggesting that Smad2 activation in EC was attenuated in C2a DEC mice. These observations suggest that TGF␤1-induced microvessel formation in Matrigel plugs is dependent on PI3K-C2␣ that is expressed in EC but not smooth muscle.

DISCUSSION
Accumulated evidence indicates that TGF␤ receptor-Smad2/3 signaling is dependent on the endocytosis of the TGF␤ receptor complex (28 -31). Upon TGF␤ binding, TGF␤ receptors are internalized into early endosomes, where the Smad anchor SARA is enriched through its FYVE domain. SARA interacts with Smad2/3, facilitating Smad2/3 phosphorylation and thereby their nuclear translocation. PI3K may be involved in at least two steps of these TGF␤ receptor signaling processes: TGF␤-induced TGF␤ receptor internalization and the endosomal localization of SARA. In the present study, we identified class II PI3K-C2␣ as PI3K isoform that is engaged in TGF␤-induced activation of Smad2/3 signaling. Our data indicate that PI3K-C2␣ is required for the endocytosis of TGF␤ receptor but not for endosomal localization of SARA.
The present observation in EC that TGF␤-induced Smad2/3 phosphorylation is dependent on TGF␤ receptor internalization into the EEA1-positive, SARA-containing early endosomes  is similar to the previous observations in other types of cells including HeLa cells, HepG2 cells, and Mv1Lu cells (28,30,31), although some discrepant results on the necessity of SARA for ALK5/Smad2/3 signaling were reported (42,43). Either PI3K-C2␣ depletion (ϳ80ϳ90%) or the expression of the kinase-deficient C␣ mutant strongly sup- pressed the internalization of TGF␤ receptor, like the dynamin inhibitor dynasore (Fig. 5). Because either PI3K-C2␣ depletion or dynasore markedly inhibited TGF␤-induced Smad2/3 phosphorylation (Figs. 1, B and C, and 4B), there is a good correlation between TGF␤-induced TGF␤ receptor internalization and Smad2/3 phosphorylation. As discussed in detail below, PI3K-C2␣ depletion did not compromise the endosomal distribution of the Smad anchor SARA. Based on these findings, it is reasonable to suggest that PI3K-C2␣ is involved in TGF␤ receptor-activated Smad2/3 signaling largely through regulat- ing TGF␤ receptor internalization. In addition, the overexpression of wtC2␣ did not affect TGF␤-induced Smad2/3 activation or ALK5 internalization, suggesting that the endogenous level of wtC2␣ was sufficient for full activation of TGF␤ receptorinduced Smad2/3 signaling.
We recently demonstrated in EC that ligand binding-triggered endocytosis of two different classes of cell surface receptors, VEGFR2 and S1P 1 , was dependent on PI3K-C2␣ (8,15). Interestingly, PI3K-C2␣ depletion inhibited only a part of multiple signaling pathways activated by VEGF and S1P: Rho activation in VEGF signaling and Rac activation in S1P signaling. We observed using FRET imaging technique that both VEGFinduced Rho activation and S1P-induced Rac activation occurred in PtdIns(3)P-enriched endosomes, as well as the plasma membrane. The present study together with those previous observations indicate that PI3K-C2␣ participates in signaling on the endosomes, upon the activation of different classes of receptors including receptor tyrosine kinases, G protein-coupled receptors, and receptor serine/threonine kinases. Thus, the ability of PI3K-C2␣ to regulate the endocytosis of different classes of cell surface receptors controls endosomal signaling.
A recent study (14) suggested a novel mechanism about a general role of PI3K-C2␣ in clathrin-dependent endocytosis in nonvascular cells; the formation of PtdIns(3,4)P 2 by PI3K-C2␣ at clathrin-coated pits and late endocytic intermediates before dynamin-mediated fission recruited the PtdIns(3,4)P 2 -effector protein SNX9, promoting maturation of clathrin-coated pits toward endocytic vesicles. They suggested that PI3K-C2␣ formed PtdIns(3,4)P 2 from PtdIns(4)P, which was generated through 5Ј-dephosphorylation of PtdIns(4,5)P 2 enriched in the clathrincoated pits. We observed that TGF␤1 induced the rapid and sustained formation of lamellipodia with a local lamellipodial increase in PtdIns(3,4)P 2 (Fig. 7C). In the lamellipodial region of the plasma membrane, endocytosis carries membrane-anchored Rho GTPases and integrins to the cell interior, and these molecules are recycled to the specific regions of the plasma membrane, which promotes lamellipodial protrusion (44). Considering the rapid onset of TGF␤1-induced lamellipodial formation and an increase in PtdIns(3,4)P 2 level, these responses very likely represent nongenomic effects of TGF␤1. It remains to be clarified how TGF␤ induces a rapid increase in the level of PtdIns(3,4)P 2 through a mechanism involving PI3K-C2␣ in EC.
The interaction of the FYVE domain in SARA with PtdIns(3)P, a predominant phosphoinositide in the endosomes, serves the endosomal localization of SARA (28,29,45). Previous studies (12,46) showed that PI3K-C2␣ formed PtdIns(3)P in cells. These observations together with the finding of the endosomal localization of PI3K-C2␣ (8) suggested that PI3K-C2␣ might be responsible for enrichment of PtdIns(3)P in the endosomes. However, the present study showed that PI3K-C2␣ depletion did not reduce PtdIns(3)-enriched endosomes (Fig.  7A). Instead, knockdown of either PI3K-C2␤ or Vps34 reduced PtdIns(3)-enriched endosomes. In agreement with these findings, knockdown of PI3K-C2␤ or Vps34 but not PI3K-C2␣ reduced SARA-containing vesicles (Fig. 7B). Our observations of the effects of PI3K-C2␤ and Vps34 depletion on SARA distribution are consistent with the previous reports showing that the nonselective PI3K inhibitor wortmannin totally changed the endosomal localization of SARA to a diffuse cytosolic pattern with inhibition of Smad signaling at the relatively low concentrations of 50 -100 nM (29,45). This range of concentration of wortmannin does not effectively inhibit PI3K-C2␣ because PI3K-C2␣ is relatively resistant to wortmannin compared with the other PI3K (5,7,47). Thus, PI3K-C2␣ is not a major enzyme to be responsible for the accumulation of PtdIns(3)P in the SARA-localized vesicular compartment. Our data indicate that the other PI3K including class II PI3K-C2␤ and class III Vps34 are involved in PtdIns(3)P accumulation in this compartment. Another point of interest is that partial reductions (25ϳ30%) of SARA association with the endosomes by depletion of either PI3K-C2␤ or Vps34 (Fig. 7B) did not inhibit Smad2/3 phosphorylation (Fig. 1C), which suggests that such partial reductions of the SARA association with the endosomes do not compromise Smad signaling.
We observed that PI3K-C2␣ depletion suppressed TGF␤induced Smad2/3 signaling in several different vascular EC of human and mouse origins, but not vascular smooth muscle or epithelial cells (Figs. 1 and 2). In addition, Smad2/3 phosphorylation mediated by caALK5 in HEK293 cells was also PI3K-C2␣-dependent (Fig. 8). Therefore, there appears to be some cell type specificity concerning the PI3K-C2␣ dependence. A few explanations for this may be possible. Another class II PI3K member, PI3K-C2␤, possesses the similarities to PI3K-C2␣, in the distribution of the expression, the structure, the clathrin binding capacity, and the substrate specificity (7). Functionally, PI3K-C2␤ is necessary for growth factor signaling (48) and cell migration with its lamellipodial distribution (49). Therefore, in vascular smooth muscle cells and epithelial cells, PI3K-C2␤ may be able to compensate for PI3K-C2␣ depletion. Alternatively, it might be possible that TGF␤-induced Smad2/3 signaling could be cell type-specific, because it was reported that the dependence of Smad2/3 signaling on endocytosis differed, depending on cell type (50).
In EC, TGF␤-induced activation of Smad2/3 signaling pathway is linked to the up-regulation of VEGF-A gene expression (39,41). Consistent with the essential role of PI3K-C2␣ in TGF␤-induced Smad signaling activation, TGF␤ stimulation of VEGF-A expression was completely and specifically dependent on PI3K-C2␣ (Fig. 9). Interestingly, TGF␤-induced endothelial cell migration and tube formation were both dependent on VEGFR2 (Fig. 10). It is reasonable to conceive that VEGF-Astimulated cell migration and tube formation requires PI3K-C2␣ because VEGFR2 signaling and transport of VE-cadherin and other molecules are dependent on vesicular trafficking (8,51). Hence, TGF␤-induced stimulation of these cellular responses requires PI3K-C2␣ for at least at two steps, i.e. TGF␤-induced Smad2/3 signaling-dependent VEGF expression and VEGF activation of VEGFR2 signaling. PI3K-C2␣ has a significant in vivo functional role in TGF␤-induced neovessel formation at an organismal level as demonstrated by the observations in EC-specific PI3K-C2␣-deleted mice (Fig. 12).
In summary, the present study indicates that PI3K-C2␣ is indispensable for TGF␤-induced Smad signaling through being engaged in the internalization of TGF␤ receptors into the Smad anchor SARA-containing early endosomes. Thus, our study suggests that PI3K-C2␣ is essential for endosomal signaling of TGF␤ receptors. The elucidation of the role for PI3K-C2␣ in TGF␤ receptor signaling opens the new avenue for understanding in more depth normal TGF␤ actions and their derangements in diseases.