Research ArticleBiochemistry

TGF-β promotes PI3K-AKT signaling and prostate cancer cell migration through the TRAF6-mediated ubiquitylation of p85α

See allHide authors and affiliations

Sci. Signal.  04 Jul 2017:
Vol. 10, Issue 486, eaal4186
DOI: 10.1126/scisignal.aal4186

The TGF-β–PI3K connection

TGF-β signaling stimulates various intracellular pathways that can promote migration in tumor cells. These pathways are generally thought to be either dependent or independent of transcription factors called SMADs. One of the SMAD-independent pathways (PI3K-AKT) is mediated by a direct interaction between PI3K and the TGF-β type I receptor. However, Hamidi et al. found that the TGF-β–induced activation of PI3K depends on another ubiquitin ligase–mediated mechanism and a SMAD protein but is independent of the kinase function of TβRI. The binding of TGF-β to its receptor triggered the recruitment of PI3K and the ubiquitin ligase TRAF6, which polyubiquitylated the regulatory PI3K subunit p85α, thus enabling phosphorylation of the catalytic PI3K subunit p110, but only in the presence of SMAD7. The abundance of ubiquitylated p85α correlated with migration in cultured cells and prostate tumor grade in patient samples. TRAF6 mediates activation of the other “SMAD-independent” (JNK) pathway. These data suggest that, although distinct, the TGF-β signaling pathways are not as insulated from each other as was once thought.

Abstract

Transforming growth factor–β (TGF-β) is a pluripotent cytokine that regulates cell fate and plasticity in normal tissues and tumors. The multifunctional cellular responses evoked by TGF-β are mediated by the canonical SMAD pathway and by noncanonical pathways, including mitogen-activated protein kinase (MAPK) pathways and the phosphatidylinositol 3′-kinase (PI3K)–protein kinase B (AKT) pathway. We found that TGF-β activated PI3K in a manner dependent on the activity of the E3 ubiquitin ligase tumor necrosis factor receptor–associated factor 6 (TRAF6). TRAF6 polyubiquitylated the PI3K regulatory subunit p85α and promoted the formation of a complex between the TGF-β type I receptor (TβRI) and p85α, which led to the activation of PI3K and AKT. Lys63-linked polyubiquitylation of p85α on Lys513 and Lys519 in the iSH2 (inter–Src homology 2) domain was required for TGF-β–induced activation of PI3K-AKT signaling and cell motility in prostate cancer cells and activated macrophages. Unlike the activation of SMAD pathways, the TRAF6-mediated activation of PI3K and AKT was not dependent on the kinase activity of TβRI. In situ proximity ligation assays revealed that polyubiquitylation of p85α was evident in aggressive prostate cancer tissues. Thus, our data reveal a molecular mechanism by which TGF-β activates the PI3K-AKT pathway to drive cell migration.

INTRODUCTION

Transforming growth factor–β (TGF-β) family members are important regulators of normal epithelial cell differentiation, cytostasis, and apoptosis but can also promote tumorigenesis (1, 2). TGF-β binds to type II and type I serine and threonine kinase receptors (TβRII and TβRI, respectively) and induces the formation of a heterotetrameric complex of two TβRIs and two TβRIIs. In the receptor complex, the constitutively active TβRII phosphorylates TβRI resulting in activation of the TβRI kinase, leading to phosphorylation and nuclear translocation of receptor-associated SMADs (R-SMADs), where they, in complex with SMAD4, regulate the expression of certain target genes (3, 4).

TGF-β also signals through non-SMAD pathways, such as the extracellular signal–regulated kinase 1 and 2 (ERK1/2), c-Jun N-terminal kinase (JNK), p38 mitogen-activated protein kinase (MAPK), and phosphatidylinositol 3′-kinase (PI3K)–protein kinase B (AKT) pathways (5, 6). Both TβRI and TβRII are required for TGF-β–induced activation of class IA PI3K (7). The PI3K family is divided into three classes based on their substrate specificities and structures. Class IA members (hereafter referred to as PI3K) are heterodimers of a regulatory subunit (p85α, p85β, p55α, p55γ, or p50α) and a catalytic subunit (p110α, p110β, or p110γ) (8). PI3K generates phosphatidylinositol 3,4,5-trisphosphate (PIP3) from phosphatidylinositol 4,5-bisphosphate (PIP2) (9). PIP3 interacts with and thereby translocates AKT to the plasma membrane, where AKT becomes activated by phosphorylation at Thr308 by phosphoinositide-dependent kinase 1 (PDK1) and at Ser473 by mTORC2 (mammalian target of rapamycin complex 2) and other kinases (10). There are three isoforms of AKT (AKT1, AKT2, and AKT3); primarily, AKT1 has been linked to oncogenesis (11).

An important mechanism of the regulation of signaling pathways involves posttranscriptional modifications via ubiquitin chains, which are bound covalently to an acceptor lysine of the target protein (12). Whereas Lys48-linked polyubiquitylation of proteins targets its substrates for proteasomal degradation (13), Lys63-linked polyubiquitylation often regulates the function and/or localization of proteins. The RING E3 ligase tumor necrosis factor receptor–associated factor 6 (TRAF6), which is known to induce Lys63-linked polyubiquitylation of its substrates, interacts with a consensus motif present in TβRI. TGF-β induces autoubiquitylation and activation of TRAF6 and Lys63-linked polyubiquitylation and activation of TGF-β–associated kinase 1 (TAK1), which lead to the activation of downstream p38 MAPK (14). TRAF6 also promotes proteolytic cleavage of TβRI, releasing the intracellular domain of TβRI, which, after translocation to the nucleus, induces a transcriptional program (15).

Here, we aimed to elucidate the mechanisms by which TGF-β induces the activation of PI3K and AKT. Our results showed that TRAF6 is required for TGF-β–induced Lys63-linked ubiquitylation of the p85α subunit of PI3K, which leads to PI3K activation, promoting the activation of AKT and its recruitment to the plasma membrane. Moreover, we found that TGF-β–induced cell migration is dependent on PI3K and TRAF6. Notably, by analyzing tissue sections by an in situ proximity ligation assay (PLA), we observed a correlation between polyubiquitylated p85α and tumor aggressiveness in patients with prostate cancer. The current study provides evidence for the molecular mechanisms whereby TGF-β promotes activation of the PI3K-AKT pathway and demonstrates its correlation with aggressive disease for prostate cancer patients.

RESULTS

AKT interacts with TβRI in a TRAF6-dependent manner

AKT is activated upon stimulation by TGF-β (16). To elucidate the molecular mechanism of the TGF-β–induced activation of AKT, we performed coimmunoprecipitation and coimmunostaining analyses to determine whether a complex containing AKT and TβRI is formed. In transfected human prostate cancer (PC-3U) cells, we observed an interaction between hemagglutinin (HA)–tagged TβRI TD [a TβRI mutant that is partly constitutively active (14)] and endogenous AKT that was increased by TGF-β stimulation (Fig. 1A). TRAF6, which induces Lys63-linked polyubiquitylation and activation of AKT upon stimulation by insulin-like growth factor 1 (IGF-1), interleukin-1β (IL-1β), and lipopolysaccharide (17), was also pulled down by the TβRI antibody, as reported previously (14), and AKT activation correlated with its interaction with TβRI (Fig. 1A). To explore whether TRAF6 is necessary for the interaction between HA–TβRI TD and AKT, cells were transiently transfected with HA–TβRI TD or the E161A mutant HA–TβRI TD, in which the TRAF6 binding motif is mutated (14). The interaction of AKT with HA–TβRI TD–E161A was decreased compared to HA–TβRI TD (Fig. 1B). Furthermore, an interaction between endogenous TβRI and AKT in wild-type mouse embryonic fibroblasts (MEFs) that was enhanced upon TGF-β stimulation was significantly reduced in TRAF6−/− MEFs (Fig. 1C). Consistent with the coimmunoprecipitation results, a coimmunostaining analysis showed that activated AKT, TRAF6, and TβRI colocalized in cell membrane ruffles upon TGF-β stimulation (Fig. 1D).

Fig. 1 The interaction between AKT and TβRI depends on TRAF6.

(A) Immunoblotting (IB) for the indicated proteins after immunoprecipitation (IP) for the HA tag from whole-cell lysates (WCL) of PC-3U cells transfected with HA–TβRI TD and treated with TGF-β. S473, Ser473; T308, Thr308; p, phosphorylated. (B) PC-3U cells were transfected with Flag-AKT1 and HA–TβRI TD or the HA–TβRI TD–E161A mutant. After treatment with TGF-β, cell lysates were subjected to immunoprecipitation with a polyclonal rabbit HA antibody, followed by immunoblotting using Flag antibody. (C) Cell lysates prepared from wild-type (WT) and TRAF6−/− MEFs treated or not with TGF-β were subjected to immunoprecipitation with a polyclonal rabbit AKT antibody or control (ctrl) immunoglobulin G (IgG) and then immunoblotted with a polyclonal goat TβRI antibody (V-22). (D) Immunofluorescence staining for the indicated proteins in HA–TβRI TD–transfected PC-3U cells, either untreated or after stimulation with TGF-β. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole. Scale bars, 20 μm. (E) PC-3U cells transfected with Flag-AKT1 and HA–TβRI TD were treated with dimethyl sulfoxide (DMSO) or 5 μM TβRI kinase inhibitor SB431542 (added 1 hour before TGF-β stimulation). Immunoprecipitation of cell lysates was performed, as described in the legend for (B). Blots and microscopy images are representative of three independent experiments. Graphs are means ± SEM from three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 within one set of samples (TGF-β–stimulated versus unstimulated cells); §P < 0.05 between different sets of samples [mutant complementary DNA (cDNA)–transfected versus wt cDNA–transfected or deficient versus wt cells]. a.u., arbitrary units.

Because TRAF6 is required for the interaction between TβRI and AKT, we further examined whether TRAF6 interacts with AKT upon TGF-β stimulation by transfecting human embryonic kidney (HEK) 293T cells with HA-tagged AKT and Flag-tagged wild-type TRAF6. Immunoprecipitation of cell lysates with an HA antiserum, followed by immunoblotting with a Flag antiserum, revealed a band with the size of TRAF6; the density of that band was increased in cells stimulated with TGF-β. In contrast, no coimmunoprecipitation was seen in lysates from cells transfected with the enzymatically inactive C70A mutant TRAF6 (fig. S1A). Activation of AKT, as monitored by phosphorylation of Thr308 and Ser473, correlated with its interaction with TRAF6 (fig. S1A). Immunofluorescence imaging of endogenous TRAF6 and phospho-AKT (pAKT) (Ser473) in PC-3U cells showed that these proteins colocalized in membrane ruffles. Both the colocalization and the number of membrane ruffles were enhanced in cells stimulated with TGF-β (fig. S1B). In cells depleted of TRAF6 by small interfering RNA (siRNA), no Ser473-pAKT was observed at the cell membrane in response to TGF-β stimulation (fig. S1B).

The fact that the C70A mutant TRAF6 did not coimmunoprecipitate with HA-AKT suggests that Cys70 mediated the interaction between TRAF6 and its substrate, in line with the report from Yin et al. (18), or that ubiquitin chains mediated the interaction between TRAF6 and AKT. Therefore, we performed in vivo ubiquitylation assays (14) using PC-3U cells. After preparing lysates from TGF-β–stimulated or nontreated cells, endogenous AKT was immunoprecipitated and subjected to immunoblotting with antibodies against polyubiquitin. TGF-β induced robust polyubiquitylation of AKT, which correlated in time with the phosphorylation and activation of AKT (fig. S2A). Double immunoprecipitation of AKT was performed, followed by immunoblotting with antibodies against Lys63-linked polyubiquitin (fig. S2B), and suggested that the polyubiquitin signal was not from another interacting protein.

To examine the nature of the polyubiquitin chains, we transfected PC-3U cells with HA-tagged wild-type ubiquitin (wild-type 3×HA-Ub) or mutant ubiquitins in which all lysine residues, except for Lys48 (Lys48-only 3×HA-Ub) or Lys63 (Lys63-only 3×HA-Ub), were mutated to arginine residues. TGF-β–induced ubiquitylation of AKT was seen when wild-type or Lys63-only Ub plasmids, but not Lys48-only Ub plasmid, were expressed, indicating that TGF-β induced Lys63-linked polyubiquitylation of AKT (fig. S2C). We further examined whether phosphorylated AKT undergoes Lys63-linked polyubiquitylation by immunoprecipitating Thr308-pAKT and immunoblotting for Lys63-linked polyubiquitin. The same AKT molecules that underwent activation by phosphorylation also displayed Lys63-linked polyubiquitylation upon TGF-β stimulation (fig. S2D).

Because the E3 ligase activity of TRAF6 was required for AKT binding to TRAF6 (fig. S1A) and AKT underwent Lys63-linked polyubiquitylation upon TGF-β stimulation (fig. S2, A to D), we explored whether TRAF6 is responsible for AKT ubiquitylation. TGF-β–induced phosphorylation (fig. S3A) and ubiquitylation (fig. S3B) of AKT were seen in wild-type but not in TRAF6−/− MEFs. Moreover, the phosphorylation of AKT increased with increasing amounts of transfected TRAF6 (fig. S3C), and both phosphorylation (fig. S3D) and ubiquitylation (fig. S3E) of AKT were seen after transfection of wild-type but not that of C70A mutant TRAF6. Thus, our observations suggest that TGF-β causes Lys63-linked polyubiquitylation and activation of AKT in a TRAF6-dependent manner.

The interaction between TβRI and AKT was not dependent on the kinase activity of TβRI because it also occurred in the presence of TβRI kinase inhibitor SB431542, evident by immunoblotting for pSMAD2 as a control (Fig. 1E). Furthermore, neither the TβRI kinase inhibitors (SB431542 and SB505124; fig. S4, A to C) nor the TβRI and TβRII kinase inhibitor LY2109761 (fig. S4, B and D) inhibited AKT phosphorylation. In addition, the activation of AKT by TGF-β did not depend on the kinase activity of TAK1, which is downstream of TRAF6, because the TAK1 inhibitor chloro-radicicol A (19) did not abolish AKT phosphorylation (fig. S5A). Moreover, AKT activation by TGF-β did not involve activation of the platelet-derived growth factor receptor (PDGFR); PDGFR efficiently activates AKT (20) and interacts with TGF-β receptors (21), but the PDGFR kinase inhibitor AG1296 did not inhibit TGF-β–induced AKT phosphorylation (fig. S5B). We have previously reported that inhibitory SMAD7, which binds to TβRI where it competes with R-SMADs and prevents their activation (4), is required for the TGF-β–induced activation of AKT (22). Consistent with this finding, we found that AKT phosphorylation and Lys63-linked ubiquitylation of AKT and the PI3K regulatory subunit p85α were increased by TGF-β stimulation in wild-type but not in Smad7−/− MEFs (fig. S6).

TGF-β–induced polyubiquitylation and activation of AKT require PI3K

The phosphorylation of AKT upon TGF-β stimulation involved the PI3K pathway because it was inhibited by the PI3K inhibitors LY294002 and wortmannin (fig. S4A). Given that the p85α regulatory subunit of PI3K is acting upstream of PI3K, we hypothesized that p85α would be required for the activation of AKT and in the TGF-β signaling pathway. We thus examined whether the p85α regulatory subunit of PI3K is needed for AKT ubiquitylation and activation. Ubiquitylation (Fig. 2A) and phosphorylation (Fig. 2B) of AKT did not occur after knocking down p85α, confirming that p85α was required for the activation of AKT. In addition, in wild-type MEFs, TGF-β induced polyubiquitylation and phosphorylation of AKT, neither of which was seen in p85−/− MEFs (Fig. 2C).

Fig. 2 TGF-β–induced polyubiquitylation and activation of AKT require p85.

(A) PC-3U cells treated with control or p85α siRNA were stimulated or not with TGF-β for different time periods and then lysed. Lysates were subjected to immunoprecipitation with rabbit AKT antibody, followed by immunoblotting using mouse monoclonal P4D1 to detect ubiquitylated AKT [AKT-(Ub)n]. The immunoprecipitation filter was reblotted with antibodies against AKT. The corresponding whole-cell lysates were subjected to immunoblotting for p85. (B) PC-3U cells treated with control or p85α siRNA were stimulated or not with TGF-β and then lysed after different time periods. Cell lysates were immunoblotted for pAKT Ser473 and pAKT Thr308, AKT, p85, and β-actin. (C) Ubiquitylation and activation of AKT were examined in WT and p85−/− MEFs treated or not with TGF-β for 30 min. Blots are representative of three independent experiments. Graphs are means ± SEM from three independent experiments. *P < 0.05, ***P < 0.001 within one set of samples (TGF-β–stimulated versus unstimulated); §P < 0.05, §§P < 0.01, §§§P < 0.001 between different sets of samples (specific siRNA–transfected versus control siRNA–transfected cells, deficient versus wt cells).

TGF-β induces polyubiquitylation of p85α in a TRAF6-dependent, but receptor kinase–independent, manner

Because p85α was required for the activation and ubiquitylation of AKT, we hypothesized that TRAF6 could, by polyubiquitylation of p85α, revert the inhibitory effect of p85α on p110α (the catalytic subunit) in the PI3K pathway. We therefore next examined whether p85α is modified by TGF-β stimulation. The polyubiquitylation of p85α was induced by TGF-β stimulation but abolished by transfection with TRAF6 siRNA (Fig. 3A). Moreover, polyubiquitylation of p85α was enhanced by TGF-β stimulation in wild-type but not in TRAF6−/− MEFs (Fig. 3B), and p85 underwent Lys63-linked ubiquitylation in a TRAF6 dose-dependent manner (Fig. 3C). TGF-β induced p85α ubiquitylation in PC-3U cells (Fig. 3D), as well as p85α ubiquitylation and AKT activation in RAW264.7 macrophages transfected with wild-type but not C70A mutant TRAF6 (Fig. 3E), demonstrating that the E3 ligase activity of TRAF6 is needed for polyubiquitylation of p85α in response to TGF-β stimulation. Blotting the resulting solutions of an in vitro ubiquitylation assay, where polyubiquitylation of p85α was observed upon incubation with glutathione S-transferase (GST)–TRAF6 (E3), E1, and the E2 enzyme Ubc13-Uev1A (Fig. 3F) (14, 23), suggested that p85α is a direct substrate for TRAF6. Detection of TGF-β–induced polyubiquitylation of p85α after double immunoprecipitation of p85α (Fig. 4A) suggested that the signal was not the polyubiquitylation of an interacting protein. Furthermore, PC-3U cells transfected with wild-type or Lys63-only Ub plasmids, but not with Lys48-only Ub plasmid, showed TGF-β–induced ubiquitylation of p85α (Fig. 4B). Last, an in vivo ubiquitylation assay in the presence of SB505124 (Fig. 4C) or LY2109761 (Fig. 4D), which catalytically inhibit TβRI or both TβRI and TβRII, respectively, revealed that the kinase activities of TβRI and TβRII were dispensable for Lys63-linked polyubiquitylation of p85α.

Fig. 3 Polyubiquitylation of p85α requires TRAF6.

(A) PC-3U cells treated with control or TRAF6 siRNA were stimulated or not with TGF-β, and the ubiquitylation of p85α was examined. (B) Cell lysates prepared from WT and TRAF6−/− MEFs treated or not with TGF-β were examined for p85α ubiquitylation. (C) Lys63-linked (K63 clone Apu3) polyubiquitylation of p85 was examined in PC-3U cells, transfected with different amounts of Flag-TRAF6 cDNA, and treated or not with TGF-β for 30 min. (D) PC-3U cells transfected with Flag-tagged WT or C70A mutant TRAF6 cDNA were treated or not with TGF-β, and cell lysates were prepared. Ubiquitylation of p85α was monitored by immunoblotting with ubiquitin (P4D1) antibody after immunoprecipitation of p85α. (E) RAW264.7 cells were transfected with Flag-tagged WT or C70A mutant TRAF6 cDNA, and ubiquitylation of p85α and activation of AKT were examined. (F) Recombinant His-p85α proteins were incubated in the presence or absence of GST-TRAF6 protein (E3) in a reaction mixture containing E1, Ubc13-Uev1A (E2), ubiquitin, and adenosine triphosphate (ATP). After incubation at 30°C for 1 hour, reaction products were analyzed by immunoblotting with antibodies against ubiquitin and p85 to visualize synthesized polyubiquitin chains. Blots are representative of three independent experiments. Graphs are means ± SEM from three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 within one set of samples (TGF-β–stimulated versus unstimulated); §P < 0.05, §§P < 0.01, §§§P < 0.001 between different sets of samples (specific siRNA–transfected versus control siRNA–transfected cells, deficient versus wt cells, or mutant cDNA–transfected versus wt cDNA–transfected cells).

Fig. 4 TGF-β–induced polyubiquitylation of p85α is Lys63-linked and does not depend on TβRI kinase activity.

(A) PC-3U cells were stimulated or not with TGF-β, and cell lysates were prepared and immunoprecipitated with rabbit p85 antibody. Beads were boiled in 1% SDS, and the supernatant was diluted in 0.5% NP-40 in phosphate-buffered saline (PBS) and thereafter subjected to immunoprecipitation with rabbit p85 antibody (double IP), followed by immunoblotting using rabbit monoclonal Apu3 against Lys63 (K63)–linked ubiquitin. (B) PC-3U cells transiently transfected with HA-tagged WT or mutated ubiquitin were treated or not with TGF-β, lysed, and subjected to immunoprecipitation using p85α antiserum and immunoblotting using an antiserum against HA. (C and D) Lysates of the cells treated or not with 10 μM SB505124 (C) or 5 μM LY2109761 (D) before TGF-β stimulation for 1 hour were exposed to immunoprecipitation with rabbit p85α antibody and immunoblotting with mouse ubiquitin antibody (P4D1). The corresponding whole-cell lysates were subjected to immunoblotting for pSMAD2, pAKT Ser473, pAKT Thr308, and p85. Blots are representative of three independent experiments. Graphs are means ± SEM from three independent experiments. *P < 0.05, **P < 0.01 within one set of samples (TGF-β–stimulated versus unstimulated); §P < 0.05 between different sets of samples (mutant cDNA–transfected versus wt cDNA–transfected cells).

TGF-β induces interaction between p85α and TβRI in a TRAF6-dependent manner

To examine whether TRAF6 and p85α interact, we performed coimmunoprecipitation assays in which we pulled down endogenous TRAF6 from lysates of PC-3U cells and then immunoblotted for endogenous p85. We barely detected p85 in TRAF6 immunoprecipitates, but the addition of TGF-β stimulated their interaction (Fig. 5A). p85α consists of an Src homology 3 (SH3) domain, a breakpoint clustered homology (BH) domain, two Src homology 2 (nSH2 and cSH2) domains that bind phosphotyrosine in activated receptors and adaptor proteins, and an inter-SH2 (iSH2) domain that binds the p110-binding domain (24). The association of the nSH2 and cSH2 domains of p85α with p110 has an inhibitory function (25). To further investigate which part of p85α associates with TRAF6, we repeated the coimmunoprecipitation assays in HEK293T cells transfected with Flag-tagged full-length p85α, a deletion mutant containing nSH2, iSH2, and cSH2 domains (3SH2), or a deletion mutant containing SH3 and BH domains (SH3-BH). Immunoblotting with a Flag antibody revealed that full-length p85α and the 3SH2 fragment but not the fragment with SH3 and BH domains interacted with TRAF6 in response to TGF-β stimulation (Fig. 5B).

Fig. 5 TGF-β induces the interaction of p85α with TβRI in a TRAF6-dependent manner, which is mediated by binding of p85α to ubiquitin chains.

(A) Lysates from PC-3U cells were subjected to immunoprecipitation with a TRAF6 antibody and immunoblotted for endogenous p85α. (B) HEK293T cells transfected with Flag-tagged full-length (FL) p85α, nSH2-iSH2-cSH2 (3SH2), or SH3-BH plasmids were subjected to immunoprecipitation with monoclonal mouse TRAF6 antibody and immunoblotting for Flag. (C) PC-3U cell lysates were subjected to immunoprecipitation with a TβRI antibody (goat V-22) and immunoblotting for endogenous p85. (D) HEK293T cells transfected with HA–TβRI TD and Flag-tagged full-length p85α, 3SH2, or SH3-BH plasmids were subjected to immunoprecipitation with polyclonal rabbit HA antibody or control IgG, followed by immunoblotting for Flag. (E) HEK293T cells transfected with Flag-p85α and HA–TβRI TD or HA–TβRI kinase-deficient (KR) plasmids were treated or not with TGF-β and subjected to immunoprecipitation with a polyclonal rabbit HA antibody or control IgG, followed by immunoblotting for Flag. (F) Cell lysates prepared from WT and TRAF6−/− MEFs were subjected to immunoprecipitation with a rabbit antibody against TβRI (V-22) and immunoblotting for mouse p85α. (G) Lysates of HEK293T cells transfected with Flag-tagged full-length p85α, del iSH2, del cSH2, SH3-BH, or iSH2-cSH2 plasmids were incubated with biotinylated Lys63-linked polyubiquitin chains and incubated with streptavidin-agarose beads, which were boiled in a sample buffer, and supernatants were subjected to immunoblotting with a Flag antibody. Blots are representative of three independent experiments. Graphs are means ± SEM from three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 within one set of samples (TGF-β–stimulated versus unstimulated); §P < 0.05, §§§P < 0.001 between different sets of samples (mutant cDNA–transfected versus wt cDNA–transfected cells or deficient versus wt cells).

Because TβRI reportedly forms a complex with TRAF6 (14) and AKT (Fig. 1) in response to TGF-β, we investigated whether TβRI interacts with p85α by examining coimmunoprecipitation of endogenous TβRI and p85α in PC-3U cell lysates. TβRI and p85 associated in TβRI precipitates upon TGF-β stimulation (Fig. 5C). We did not detect any association between p85 and TβRII (fig. S7A). Coimmunoprecipitation assays in lysates of HEK293T cells transfected with HA–TβRI TD (the partly constitutively active mutant) and Flag-tagged full-length or deletion mutants of p85α indicated that TβRI interacted with the full-length and 3SH2 fragment of p85α (Fig. 5D), suggesting that the interaction between p85, TβRI, and TRAF6 is mediated by the nSH2, iSH2, and cSH2 domains in p85. Consistent with the coimmunoprecipitation results, a coimmunostaining analysis showed that p85α, activated AKT, TRAF6, and TβRI colocalized in cell membrane ruffles upon TGF-β stimulation (fig. S7B).

In addition, consistent with the kinase activity of TβRI being dispensable for Lys63-linked polyubiquitylation of p85α (Fig. 4, C and D), the interaction of p85α and TβRI was not dependent on the kinase activity of TβRI, as evident from coimmunoprecipitation assays using cells transfected with Flag-p85α and HA-tagged TβRI TD or a kinase-deficient mutant receptor (Fig. 5E). Immunoprecipitation assays using wild-type or TRAF6−/− MEF cells (Fig. 5F) suggested that TRAF6 is necessary for the interaction between TβRI and p85α. Our data suggest that TRAF6 mediates the interaction between TβRI and PI3K; thus, the previously reported interaction between TβRI and p85 (16) may be indirect.

Next, we examined whether p85α binds free Lys63-linked ubiquitin chains by transfecting PC-3U cells with full-length or deletion mutants of p85α (del iSH2, del cSH2, SH3-BH, or iSH2-cSH2); cell lysates were prepared and incubated with recombinant biotinylated Lys63-linked ubiquitin chains followed by pulldown with streptavidin-agarose beads and immunoblotting by a Flag antibody. Fragments containing SH3 domain bound to Lys63-linked ubiquitin chains (Fig. 5G), suggesting that the SH3 domain is responsible for the interaction of p85α with Lys63-linked ubiquitin chains. Because our data thus far suggest that p85α binds Lys63-linked ubiquitin chains and TRAF6 mediates Lys63-linked polyubiquitylation of p85α (Fig. 3, D and E), it is possible that Lys63-linked polyubiquitylation of p85α is needed for the interaction between TβRI and PI3K.

Lys513 and/or Lys519 in the iSH2 domain is a major site of Lys63-linked polyubiquitylation of p85α

To identify the domain in p85α, which is the acceptor of polyubiquitylation, we performed ubiquitylation assays of lysates of PC-3U cells transfected with full-length or deletion mutants of p85α [del iSH2, del cSH2, SH3-BH, or iSH2-cSH2 (24)] and HA-ubiquitin (HA-Ub). The SH3-BH fragment did not undergo any detectable ubiquitylation (Fig. 6A). This finding is consistent with the fact that p85α interacts with TβRI and TRAF6 through nSH2-iSH2-cSH2 and not SH3-BH. The fragment with iSH2-cSH2 domains showed a ladder-like ubiquitylation similar to full-length p85α, in contrast to del iSH2 and del cSH2, which showed mono-ubiquitylation (Fig. 6A). These findings suggest that the acceptor lysine residue may be present in the iSH2 and/or cSH2 domains.

Fig. 6 Ubiquitylation of p85α occurs on Lys513 and/or Lys519 in the p85α iSH2 domain.

(A) Lysates of PC-3U cells transfected with HA-tagged WT ubiquitin and Flag-tagged full-length p85α or del iSH2, del cSH2, SH3-BH, or iSH2-cSH2 plasmids were subjected to immunoprecipitation with Flag or mouse IgG antiserum and immunoblotted with HA antibody. (B) Lysates of HEK293T cells transfected with HA-Ub and full-length Flag-p85α were subjected to SDS-PAGE. The protein bands used for mass spectrometry are marked by an arrowhead (nonubiquitylated) and arrows (polyubiquitylated). (C) Two detected branched peptide sequences in iSH2 domain of p85α. The red letters show the putative lysine residues. (D) Lysates of HEK293T cells transfected with HA-Ub and full-length WT Flag-p85α, K567/575R, or K513/519R Flag-p85α plasmids were examined for p85α polyubiquitylation. (E) Lysates of HEK293T cells transfected with HA-Ub and full-length WT Flag-p85α or K513R, K519R, or K513/519R mutant p85α plasmids were subjected to ubiquitylation assay by blotting with a rabbit antibody against Lys63-linked ubiquitin (K63-linked Ub). Blots are representative of three independent experiments. Graphs are means ± SEM from three independent experiments. *P < 0.05 within one set of samples (TGF-β–stimulated versus unstimulated and in panel), §P < 0.05 between different sets of samples (mutant cDNA–transfected versus wt cDNA–transfected cells).

To determine which lysine residues in the iSH2-cSH2 region of p85α are polyubiquitylated, we used large-scale preparations of ubiquitylated Flag-p85α (Fig. 6B). The preparations were resolved by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and stained with silver nitrate; the polyubiquitylated components within the gel pieces were then digested with trypsin, and the peptide fragments were then eluted and subjected to mass spectrometry to identify branched peptides that corresponded to the covalent bond of the C-terminal ubiquitin glycine to the ε-amino group of a lysine in p85α. Inspection of the detected peptide masses identified two branched peptides in the iSH2 domain; Lys513 and/or Lys519 was identified in one of the peptides, and Lys567 and/or Lys575 was identified in the other (Fig. 6C). These amino acid residues were mutated to arginines to produce two double mutants, K513/519R and K567/575R, which were analyzed in an in vivo ubiquitylation assay. We observed a reduced efficiency of p85α ubiquitylation when Lys513 and Lys519 were mutated to arginine, whereas mutation of Lys567 and Lys575 had no effect (Fig. 6D). To further narrow down which residue of Lys513 and Lys519 was ubiquitylated, we produced single-mutant plasmids, K513R and K519R, and compared the efficiency of their ubiquitylation. The K513R and K519R single mutants did not show decreased ubiquitylation (Fig. 6E), suggesting that both residues are acceptor lysines. However, coimmunoprecipitation assays in HEK293T cells transfected with Flag-tagged wild-type, single-mutant, or double-mutant p85α indicated that the mutation of Lys513 and/or Lys519 to arginine does not affect the binding of p85α to p110α (fig. S8).

To further validate whether TRAF6 and TβRI kinase activities are important for PI3K activation, we measured PIP3 concentration and AKT phosphorylation in wild-type MEFs treated with inhibitors of PI3K or TβRI kinase and in TRAF6−/− or p85−/− MEFs. Lysates of the cells stimulated or not with TGF-β were immunoprecipitated by a p85 antibody and subjected to a kinase reaction containing the PIP2 substrate, and the amount of PIP3 was then measured by an enzyme-linked immunosorbent assay (ELISA). TGF-β–induced PI3K activity in wild-type MEF cells was impaired by the PI3K inhibitor wortmannin but not the TβRI kinase inhibitor SB505124, and PI3K activity was significantly reduced in both p85−/− and TRAF6−/− MEFs relative to wild-type MEFs (fig. S9, A and B). Similar assays in PC-3U cells transfected with wild-type or C70A mutant TRAF6 revealed that the E3 ligase activity of TRAF6 is important for TGF-β–induced increase of PI3K activity (fig. S9, C and D). Together, our data thus far indicate that whereas the kinase activity of TβRI was not needed, TRAF6 and p85 were required for the TGF-β–induced activation of PI3K.

TGF-β–induced cell migration requires the PI3K pathway and TRAF6 and is not dependent on the kinase activity of TβRI

The activation of PI3K/AKT promotes cell growth, survival, and migration (7, 26, 27). We therefore explored the importance of the TRAF6/PI3K/AKT pathway in TGF-β–induced cell migration using a Transwell migration assay and cell culture–based wound healing assay. Inhibition of PI3K by wortmannin significantly inhibited TGF-β–induced migration of RAW264.7 cells through the membrane of the Transwell chamber, whereas the TβRI kinase inhibitor SB431542 only had a minor inhibitory effect (Fig. 7A). The corresponding cell lysates were analyzed to control for the effect of the inhibitors. This result is consistent with a previous finding that TGF-β regulates macrophage migration (28). In addition, the migration of PC-3U cells in response to TGF-β in the wound healing assay decreased significantly by inhibition of PI3K by wortmannin, LY294002, or NVP-BKM120, but not by inhibition of the TβRI kinase by SB431542 (Fig. 7B). These data and the observations that wild-type but not p85−/− or TRAF6−/− MEFs migrated through the membrane of the Transwell chamber in response to TGF-β (Fig. 7, C and D) indicate that p85, TRAF6, and PI3K/AKT activities are required for TGF-β–induced cell migration.

Fig. 7 TGF-β–induced cell migration requires the PI3K pathway and TRAF6 and is not dependent on the kinase activity of TβRI.

(A) RAW264.7 cells treated with the PI3K inhibitor wortmannin, TβRI kinase inhibitor SB431542, or 0.1% DMSO for 1 hour were subjected to Transwell migration assay. Scale bar, 100 μM. The corresponding whole-cell lysates were subjected to immunoblotting using antibodies against pAKT Ser473, pAKT Thr308, and pSMAD2. (B) PC-3U cells treated with the PI3K inhibitors NVP-BKM120, LY294002, and wortmannin, TβRI kinase inhibitor SB431542, or 0.1% DMSO as control were stimulated or not with TGF-β in a cell culture wound healing assay, as described in Materials and Methods. Scale bar, 100 μM. (C) WT and p85−/− MEF cells were exposed to Transwell migration assay as described for (A). Scale bar, 100 μM. (D) WT and TRAF6−/− MEFs treated or not with TGF-β for 24 hours were exposed to wound healing assay as described for (B). Scale bar, 100 μM. Blots and microscopy images are representative of three independent experiments. Graphs are means ± SEM from three independent experiments. **P < 0.01, ***P < 0.001 within one set of samples (TGF-β–stimulated versus unstimulated); §P < 0.05, §§P < 0.01, §§§P < 0.001 between different sets of samples (inhibitor-treated versus control cells or deficient versus wt cells).

Next, we examined the role of p85α polyubiquitylation in the efficiency of PI3K activity and consequential cell migration. PI3K activity and AKT phosphorylation in cells expressing K513R and K519R mutant p85α were not significantly different than those in cells expressing wild-type p85α, whereas they were reduced in cells expressing double-mutant K513/519R p85α (Fig. 8, A and B). TGF-β did not induce phosphorylation of p85α at Tyr458, and expressing K513/K519R p85α did not affect the amount of Tyr-phosphorylated p85 (Fig. 8B), indicating that the regulation of PI3K by TGF-β occurs through a mechanism that is independent of the phosphorylation of p85 that is mediated by tyrosine kinase receptors (29, 30).

Fig. 8 TGF-β–induced polyubiquitylation of p85α on Lys513 and Lys519 is important for PI3K and AKT activation and for cell migration.

(A) PC-3U cells transfected with WT, K513R, K519R, or K513/519R plasmids of p85α were treated or not with TGF-β for 30 min. Cell lysates were prepared and subjected to a PI3K assay, followed by measurement of PIP3 concentration, as described in Materials and Methods. Data are means ± SEM from three independent experiments; *P < 0.05 within one set of samples (TGFβ-stimulated versus unstimulated), §P < 0.05 between different sets of samples (mutant cDNA-transfected versus wt cDNA-transfected cells). (B) Whole-cell lysates corresponding to (A) were subjected to immunoblotting with antiserum against pAKT Ser473, pAKT Thr308, pp85, p85, Flag, and pS6. (C) PC-3U cells silenced or not for endogenous p85α [by siRNA targeting 3′ untranslated region (3′-UTR)] and transfected with WT, K513R, K519R, or K513/519R plasmids of p85α were subjected to Transwell assay. Scale bar, 100 μM. Bar graphs show the means ± SEM from three independent experiments; *P < 0.05. (D) Cell lysates corresponding to the samples in (C) were subjected to immunoblotting for p85, Flag, and β-tubulin to check the knockdown and transfection efficiency. The corresponding whole-cell lysates were immunoblotted for p85 and Flag. The filters were stripped and reblotted for β-actin. Blots and microscopy images are representative of three independent experiments.

The fact that transfection of K513/519R p85α did not reduce TGF-β–induced phosphorylation of S6 ribosomal protein (Fig. 8B) reinforces the previous finding that the activation of S6 is independent of AKT signaling (31). Because PI3K activity is important for TGF-β–induced cell migration, we examined TGF-β–induced cell migration by a Transwell migration assay using cells that had been silenced for endogenous p85α and transfected with Flag-tagged wild-type, K513R, K519R, or K513/519R mutant p85α. Cells transfected with K513/519R p85α showed decreased migration compared to wild-type Flag-p85α transfected cells (Fig. 8, C and D), suggesting that polyubiquitylation of p85α at Lys513 and Lys519 mediates TGF-β–induced cell migration. The amount of Lys63-linked ubiquitylated p85α (fig. S10A), phosphorylated AKT, and migration was decreased in p85−/− MEFs transfected with K513/519R mutant p85α than in cells transfected with wild-type p85α (fig. S10, A to C), indicating that residual endogenous wild-type p85 did not influence the PC-3U migration assays above.

Polyubiquitylation of p85α correlates with aggressiveness of human prostate cancer

Activation of the PI3K-AKT pathway has earlier been demonstrated to be linked to tumor progression in prostate cancer (32). Using immunohistochemistry and in situ PLAs (33) to examine tissue from patients with prostate cancer, we found an increased abundance of activated SMAD2 and AKT and Lys63-linked polyubiquitylated p85α in prostate tumor tissue that had a high Gleason score (indicating aggressive disease) (Fig. 9, A to C), suggesting that TGF-β signaling—both SMAD- and PI3K/AKT-dependent paths—and the TRAF6-p85α mechanism that we have identified may play a role in the progression of prostate cancer.

Fig. 9 Lys63-linked ubiquitylation of p85α and activation of SMAD2 and AKT in prostate cancer tissues are correlated with aggressiveness.

(A and B) Prostate tumor tissue samples were stained with pSMAD2 (A) and pAKT Ser473 (B) antibodies. Quantification shows the means ± SEM of pAKT Ser473 of 10 patients in each group. (C) The association between p85α and Lys63 (K63) ubiquitin in prostate cancer patients (brown dots) was determined by PLA (300 cells were analyzed in each group). Data are means ± SEM of five patients in each group. **P < 0.01, ***P < 0.001. Scale bars, 50 μm. (D) Proposed model for TGF-β–induced polyubiquitylation of p85α and activation of PI3K and AKT. TGF-β–induced activation of PI3K is initiated by ligand-induced oligomerization of TGF-β receptors, which juxtaposes TRAF6 molecules, constitutively bound to TβRI, that undergo autoubiquitylation and ubiquitylate TβRI. Lys63-linked ubiquitin chains on TRAF6 or TβRI mediate the recruitment of PI3K to the complex by binding the SH3 domain of p85α. Activated TRAF6 then causes Lys63-linked ubiquitylation of p85α and activation of PI3K, which in turn leads to PIP3 production. Whether TRAF6, PI3K, AKT, and SMAD7 simultaneously form a complex or whether interactions occur sequentially remains to be elucidated. AKT is recruited to the complex whereby it is ubiquitylated and activated. The recruitment of p85α and AKT may be facilitated by SMAD7, which has been shown to have an adaptor function in the non-Smad TRAF6 pathway. Attachment of Lys63-linked polyubiquitin chains on p85α possibly favors a conformational change of p85α, thereby removing the inhibitory contacts of the SH2 domains of p85α from p110, which enhances the kinase activity of PI3K. AKT activation promotes cell migration, which has been shown to be dependent on TRAF6-mediated c-Jun activation as well (43). For simplicity, the PI3K-AKT-SMAD7 complex is shown only on one side of the symmetric TGF-β receptor complex.

DISCUSSION

The PI3K-AKT pathway is activated in many types of human malignancies (8). In prostate cancer, increased activation of AKT correlates with aggressive disease, whereas activated AKT is undetectable in normal prostate tissue (34). TGF-β functions as a tumor suppressor in normal cells and in early phases of tumor development but switches to act as a tumor promoter in the late phases of tumor development (35, 36). It has been shown that TGF-β can activate the PI3K pathway directly (6, 16) and indirectly (22, 37). However, no clear, direct mechanistic link between TβRI and PI3K has yet been demonstrated.

Here, we have shown that the E3 ligase activity of TRAF6 is crucial for recruitment of the PI3K regulatory subunit p85α to TβRI and for TGF-β–induced Lys63-linked polyubiquitylation of p85α. The TGF-β–induced TRAF6-mediated Lys63-linked polyubiquitylation of p85α is needed for activation of and does not induce the degradation of p85 in contrast to Lys48-linked polyubiquitylation of p85α, which leads to proteasomal degradation (24, 38). Lys63-linked polyubiquitylation of AKT has been shown to be selectively driven by two different E3 ligases in response to diverse growth factors; thus, IGF-1, epidermal growth factor (EGF), and IL-1 receptors engage TRAF6 (17), whereas signaling through the ErbB family receptors involves the Skp2 SCF (SKP1/cullin/F-box protein) complex to trigger ubiquitylation of AKT (39). TRAF6 also plays a role in integrin α3β1–induced Lys63-linked polyubiquitylation of AKT (40). We found that TRAF6 activity is required for the TGF-β–induced recruitment of AKT to the complex of PI3K and TβRI, where the polyubiquitylation and activation of AKT occur.

The AKT pathway plays a central role for various cellular activities in response to growth factor signaling via activation by receptor tyrosine kinases, such as EGF and PDGF, through direct receptor binding or through tyrosine phosphorylation of scaffolding adaptor proteins, which in turn can bind to and activate PI3K (8, 41). Hyperactivation of AKT is closely related to various forms of cancer. In prostate cancer, loss of the PIP3 phosphatase PTEN (phosphatase and tensin homolog) is frequently found, leading to aberrant activation of AKT. We have also previously reported that knockdown of SMAD7 decreased the TGF-β–induced activation of AKT (22) and found that TRAF6 and SMAD7 are important for activation of the TGF-β–induced activation TAK1–p38 MAPK pathway (14, 42). Here, we demonstrated that SMAD7 was also required for ubiquitylation of AKT and p85α. Exactly how SMAD7 functions together with TRAF6 to promote activation of the PI3K-AKT pathway remains to be elucidated.

Activation of p38 MAPK in prostate cancer cells leads to subsequent phosphorylation of c-Jun on Ser63, thereby promoting its transcriptional activity and resulting in increased expression of the transcription factor Snail1, which in turn promotes migration and invasion of prostate cancer cells (43). Moreover, activation of AKT leads to phosphorylation and inactivation of GSK3β. When GSK3β is active, it phosphorylates transcription factors, such as c-Jun and Snail1, thereby marking them for subsequent proteasomal degradation. Therefore, activation of AKT leads to stabilization of both c-Jun and Snail1 (41, 44). In contrast, in normal cells, AKT is known to counteract the JNK–c-Jun pathway (45), whereas, in PTEN-deficient prostate cancer cells and in glioblastoma cells, AKT and c-Jun have been found to act in parallel pathways, promoting tumor aggressiveness (46, 47). This mode of activation and stabilization of c-Jun through TRAF6-p38 pathway is SMAD-independent; however, Thakur et al. showed that TGF-β induces transcription of c-Jun in a SMAD-dependent manner (43). It is possible that inhibition of TβRI kinase activity did not reduce cell migration significantly because preexisting c-Jun was activated and stabilized by TGF-β, although new transcription of c-Jun was blocked.

SH3 domains were previously reported to bind ubiquitin (48). Here, we found that the fragments of p85α containing SH3 domain bind Lys63-linked ubiquitin chains. It is possible that the SH3 domain of p85α binds ubiquitin chains present on autoubiquitylated TRAF6 or on TβRI (14, 49) after TGF-β stimulation; thereby, p85α is recruited to the complex of TGF-β receptors TRAF6 and SMAD7 at the plasma membrane and polyubiquitylated.

When PI3K resides in the cytosol, nSH2 and cSH2 domains of the regulatory subunit inhibit the enzymatic activity of the catalytic subunit. For PI3K to be activated, conformational changes take place to remodel the enzyme’s interface and open the catalytic loop (50). Lack of the PI3K SH2 binding motif pYXXM (50) in TGF-β receptors and TRAF6 is consistent with the notion that another mechanism is involved in the binding of PI3K to the TGF-β receptor complex than in the binding of the SH2 domain to phosphorylated tyrosine residues, which has been observed after the activation of tyrosine kinase receptors. The viral nonstructural protein 1 from influenza virus activates PI3K by interaction with the iSH2 domain of p85β in a manner that prevents the nSH2 domain from inhibiting p110 (51). Because we found that Lys63-linked ubiquitylation of p85α on Lys513 and Lys519 was needed for enzymatic activity of PI3K, it is possible that ubiquitylation of the iSH2 domain results in a conformational change dislocating of the inhibitory nSH2 and cSH2 domains. This possibility remains to be elucidated by investigation of the crystal structures of ubiquitylated p85α bound to p110. By studying the available crystal structures of PI3K heterodimer (52, 53), we found that the acceptor sites for ubiquitylation, Lys513 and Lys519, are located at the surface of PI3K and thus accessible for the ubiquitylation machinery. On the basis of our previous findings and data reported in this report, we suggest that a multicomponent complex assembles at the activated TGF-β receptors, containing SMAD7, TRAF6, PI3K, and AKT, and that these components cooperatively builds a signaling platform (Fig. 9D).

Inhibition of the kinase activities of TβRI and TβRII did not prevent TGF-β–induced AKT activation and its interaction with TβRI (Fig. 1E and fig. S4) but inhibited TGF-β–induced SMAD2 activation. These findings suggest that the TGF-β–induced activation of PI3K-AKT occurs in a TGF-β receptor kinase–independent but TRAF6-dependent manner. Moreover, our data indicate that p85 polyubiquitylation and its interaction with TβRI do not require TGF-β receptor kinase activity. This is consistent with the previous findings by Sorrentino et al. (14), who show that TRAF6 induces Lys63-linked polyubiquitylation of TAK1 upon TGF-β stimulation through the non-SMAD pathway, independent of the kinase activity of TGF-β receptor. TGF-β–induced cell migration was dependent on polyubiquitylation of p85α on Lys513 and Lys519 and the activities of PI3K and TRAF6, but not on the kinase activity of TβRI. Our observation of Lys63-linked polyubiquitylation of p85α and activation of AKT in aggressive prostate tumors showing signs of active TGF-β signaling provide support for a protumorigenic role of activation of the PI3K-AKT pathway in prostate cancer. This is in agreement with a previous study of Leight et al. (54), who show that in a high-rigidity matrix, TGF-β–induced AKT activation enhances epithelial-to-mesenchymal transition, invasion, and metastasis in murine mammary gland cells.

On the basis of this current study and our previous findings (14), we propose a model for TGF-β–induced activation of the PI3K-AKT pathway in which the catalytic activity of TRAF6 plays an important role for Lys63-linked polyubiquitylation of p85α at Lys513 and Lys519. Presumably, this leads to a change in the conformation of the protein, thereby releasing the inhibitory effect of p85α on p110 (Fig. 9D), resulting in the generation of PIP3 in the cell membrane. Thereafter, AKT is recruited to the TβRI-TRAF6-p85α complex at the cell membrane, leading to phosphorylation of AKT by the PIP3-activated PDK1.

MATERIALS AND METHODS

Cell lines, reagents, and antibodies

HEK293T cells, HaCaT cells, and wild-type, Smad7−/−, and Traf6−/− MEFs were grown in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS). The human prostate carcinoma cell line PC-3U, originating from PC-3 cells (55), was grown in RPMI 1640 medium containing 10% FBS and 2 mM l-glutamine. p85α−/− p85β−/− MEF cells were a gift from G. Scita (School of Medicine, University of Milan, Italy) (56) and are referred to as p85−/− MEFs throughout the study. p85−/− MEFs were grown in DMEM containing 15% FBS in the presence of 10% CO2. Mouse macrophage RAW264.7 cells, which were purchased from American Type Culture Collection, were cultured in DMEM containing 10% FBS. Transient transfections of HEK293T cells were performed using the calcium phosphate method (19). PC-3U and RAW264.7 cells were transiently transfected using FuGENE HD (Roche), as described previously (42). Cells were starved in serum-free medium at least 20 hours before stimulation with a growth factor. Cells were stimulated by adding TGF-β1 (5 ng/ml) (referred to as TGF-β throughout the study; R&D Systems), EGF (10 ng/ml) (PeproTech Inc.), or PDGF-BB (100 ng/ml) (Amgen) to the starvation medium. For immunoprecipitation experiments, polyclonal rabbit AKT1, AKT2, and AKT3 (referred to as AKT throughout the study; H-136) and mouse p85α (B-9) antiserum (Santa Cruz Biotechnology Inc.) were used. Polyclonal rabbit HA (Y-11), monoclonal mouse HA (F-7), monoclonal mouse ubiquitin (P4D1), polyclonal rabbit and goat TβRI (V-22), and monoclonal mouse TRAF6 antibodies (used for immunofluorescence) were purchased from Santa Cruz Biotechnology Inc. Mouse monoclonal Lys63-linked polyubiquitin antibodies (HWA4C4 and Apu3) were purchased from Enzo Life Sciences Inc. and Millipore, respectively. Monoclonal rabbit pAKT (Ser473 and Thr308), rabbit phospho-PI3K p85 (Tyr458)/p55 (Tyr199), monoclonal rabbit p85 (recognizing both p85α and p85β; 19H8), monoclonal rabbit p110α, and rabbit pS6 ribosomal protein (Ser235/236) antibodies were from Cell Signaling. Mouse monoclonal Penta·His antibody was from Qiagen. Monoclonal mouse β-actin and monoclonal rabbit TRAF6 (used for immunoblotting) were obtained from Abcam. Tetramethyl rhodamine isothiocyanate–labeled phalloidin and monoclonal mouse FLAG (M2) antibodies were obtained from Sigma. Rabbit Alexa Fluor 488 and Alexa Fluor 555 and mouse Alexa Fluor 488 antibodies raised in donkey were purchased from Invitrogen. Light chain–specific peroxidase IgG, fraction monoclonal rabbit IgG, and light chain–specific peroxidase AffiniPure mouse IgG were purchased from Jackson Immunoresearch Laboratories Inc. Biotinylated poly-Ub wild-type chains (2 to 7) (K63-linked) were purchased from Boston Biochem.

Inhibitors

The PI3K inhibitors LY294002, wortmannin (KY12420), and NVP-BKM120 were purchased from Calbiochem. The TβRI kinase inhibitors SB431542 and SB505124 were from Tocris and Sigma, respectively. LY2109761, which inhibits the kinases of both TβRI and TβRII, was purchased from Cayman Chemical. The PDGFR kinase inhibitors imatinib and AG1296 were obtained from Novartis Pharma AG and Sigma, respectively.

Immunoblotting and immunostaining

HEK293 cell lines and PC-3U cells were grown in 10-cm dishes. Forty-eight hours after transfection, cells were starved for 20 hours, stimulated with TGF-β, PDGF-BB, or EGF for the indicated time periods, washed once with ice-cold PBS, and lysed in ice-cold lysis buffer [150 mM NaCl, 50 mM tris (pH 8), 1% Triton X-100, 10% (v/v) glycerol, 1% (v/v) NP-40, 0.5% sodium deoxycholate, 1 mM aprotinin, 1 mM Pefabloc, and 1 mM sodium orthovanadate]. After centrifugation, supernatants were collected, and protein concentrations were measured by using the BCA (bicinchoninic acid) protein assay kit (Thermo Fisher Scientific). Equal amounts of proteins were used for immunoprecipitations. Samples were subjected to SDS-PAGE in 8 or 10% polyacrylamide gels, followed by blotting onto polyvinylidene difluoride membranes and immunoblotting, as described previously (22).

For analysis by immunostaining, PC-3U cells were seeded in six-well plates and grown on coverslips 24 hours before transfection. Immunostaining was performed, as described previously (42). Photomicrographs were obtained using an Axioplan 2 microscope (Carl Zeiss MicroImaging Inc.) with a digital camera (RETIGA EXi) using a Plan Apochromat ×63/1.4 oil DIC objective lens (Carl Zeiss MicroImaging Inc.). Photographs were taken at room temperature. Primary images were acquired using the ZEN 2011 program. Image memory content was reduced, and brightness contrast was adjusted using Photoshop 6.0 (Adobe). Pictures shown are representative images from three different experiments.

In vivo ubiquitylation assays

PC-3U and MEF cells were washed once in PBS, scraped in 1 ml of PBS, and centrifuged at 400g for 5 min. Noncovalent protein interactions were dissociated in 1% SDS, and the solution was boiled for 10 min. Samples were diluted 1:10 in PBS, containing 0.5% NP-40, 1 mM aprotinin, and 1 mM Pefabloc. The samples were cleared by centrifugation at 12,000g for 10 min and subjected to immunoprecipitation, followed by immunoblotting.

In vitro ubiquitylation assays

Recombinant GST-TRAF6 (E3) and His-p85α (SignalChem; about 0.1 μg at maximum concentration) were incubated in 20 mM tris (pH 7.4), 50 mM NaCl, 10 mM MgCl2, 10 mM dithiothreitol, 10 mM ATP, ubiquitin (0.5 μg/μl) (Sigma), 2 μM ubiquitin aldehyde (Biomol), 100 μM MG132 (Sigma), 0.1 μg of E1 (human recombinant from Biomol), and 0.2 μg of E2 Ubc13-Uev1A (Biomol) at 30°C for 1 hour and then subjected to SDS-PAGE.

Scratch wound healing assay

PC-3U and MEF cells were cultured in six-well plates. Forty-eight hours later, the cells were treated or not with different inhibitors. TGF-β was added to the cells after 1 hour, and “wounds” were made using a 200-μl pipette tip. Pictures of the wounds were taken immediately and 24 hours later using a Zeiss Axiovert 40CFL microscope. Primary images were acquired using the AxioVision program and analyzed by the TScratch program. The percentage of the open wound was calculated by dividing the area of the gap after 24 hours by that at 0 hour. Then, the percentage of the wound closure was calculated by subtracting the open wound percentage from 100. Pictures shown are representative images from three different experiments.

Transwell migration assay

Migration assays were performed in Transwell cell culture chambers with polycarbonate filters (24-well, 8-μm pore, Corning Costar). The filters were coated with fibrinogen (1 mg/ml) for 6 hours and then dried. Cells pretreated or not with the inhibitors for 1 hour were washed and resuspended at a density of 200,000 cells in 200 μl of the starvation medium without or with only TGF-β or TGF-β and the inhibitor and loaded to the upper chamber. Starvation medium (500 μl) without or with only TGF-β or TGF-β and the inhibitor was poured to the lower reservoir. After 6 hours of migration, cells on the upper membrane were scraped off, and those that migrated to the lower membrane were fixed by 4% formaldehyde, stained by the Giemsa staining method, and photographed using the Zeiss Axiovert 40CFL microscope. Primary images were acquired using the Zen program. Quantification of the number of the migratory cells was done by measuring the number of pixels using Photoshop 6.0 (Adobe). Pictures shown are representative images from three different experiments.

Plasmids

Expression vectors for wild-type 3×HA-Ub, Lys48-only 3×HA-Ub, and Lys63-only 3×HA-Ub were gifts from V. M. Dixit (Genentech). FLAG-tagged TRAF6-C70A was a gift from Z. J. Chen (University of Texas Southwestern Medical Center, Dallas, TX). Flag-AKT1 was a gift from W. C. Sessa (Yale University School of Medicine, New Haven, CT). HA-tagged AKT1 was purchased from Addgene. HA–TβRI TD (also called ALK5 TD) was a gift from P. ten Dijke (University of Leiden, Netherlands). The HA–TβRI TD–E161A plasmid with HA fused to the C terminus of the receptor was described previously (14). Flag-tagged p85α constructs were gifts from J. Y. Ahn (School of Medicine, Sungkyunkwan University, Suwon, Korea).

siRNA transfection

SMARTpool siRNA for TRAF6 and p85α and Individual siGENOME PIK3R1 siRNA that targets 3′-UTR were synthesized by Dharmacon Research. The SMARTpool TRAF6 and p85α siRNA were transfected with DharmaFECT (Dharmacon Research) according to the manufacturer’s protocol. The Individual siGENOME PIK3R1 siRNA was transfected with Lipofectamine 3000 (Invitrogen).

Site-directed mutagenesis

Site-directed mutagenesis was done using a QuikChange Lightning Site-Directed Mutagenesis kit (Agilent Technologies). The following primers were synthesized by Sigma: K567R, ggataaggtctggtctaatgctgttcatacgtttgtcaatttc and gaaattgacaaacgtatgaacagcattagaccagaccttatcc; K575R, ggtctctcgtccttctcagctggataaggtctgg and ccagaccttatccagctgagaaggacgagagacc; K513R, gccttcacgtctaaacttttctatgtattctttgctgtaccg and cggtacagcaaagaatacatagaaaagtttagacgtgaaggc; and K519R, gcataatcctttgtatttctctctcattgccttcacgtttaaac and gtttaaacgtgaaggcaatgagagagaaatacaaaggattatgc.

PI3K activity detected by ELISA

An ELISA assay (96-well) for detection of PIP3 was purchased from Echelon Biosciences Inc. MEFs or transfected PC-3U cells were lysed, and PI3K reactions were set up according to the manufacturer’s procedure. The PIP3 product of the kinase reactions was detected by ELISA.

Mass spectrometric analysis of polyubiquitylated p85α

HEK293T cells in five 10-cm dishes were transfected, and ubiquitylation assays were performed, as described above. p85 was immunoprecipitated and then subjected to SDS-PAGE and silver staining (57). Bands corresponding to nonubiquitylated and polyubiquitylated p85α were cut out and trypsin-digested in the gel (58). In brief, the protein bands were destained, washed, and treated with ammonium bicarbonate. After complete drying under nitrogen, trypsin (porcine; modified sequence grade from Promega) was allowed to absorb into the gel piece. After overnight digestion, the peptides were extracted and then desalted and concentrated on a handmade microcolumn, POROS R2 20, packed in a gel loader tip (59). The peptides were eluted with acetonitrile-containing matrix (4-hydroxy-α-cyanocinnamic acid) directly onto the target, and spectra were taken using a Bruker Biflex III matrix-assisted laser desorption/ionization–time-of-flight mass spectrometer from Bruker Daltonics. The search for ubiquitylated peptides was accomplished by GPMAW (Lighthouse Data) by scanning for the expected addition of the most C-terminal tryptic peptide of human ubiquitin [GG (114.05 Da)] onto a lysine residue in the tryptic digest of human p85α. Spectra and associated data are provided in fig. S12.

In situ PLA

The prostate cancer tissue sections were pretreated with deparaffinization, antigen retrieval, and permeabilization. PLA was performed with a Duolink Detection kit for Brightfield (Sigma) according to the manufacturer’s instructions. Images were acquired with Pannoramic 250 Flash. Signals were quantified using Duolink ImageTool. p85α antibody was from Novus Biologicals. pSMAD2 and pAKT Ser473 antibodies were from Cell Signaling.

Malignant prostate cancer tissues were collected from men undergoing prostatectomy. The tumors had Gleason scores ranging from 6 (3 + 3) to 9 (4 + 5) and were staged as pT2 or pT3. All experiments were performed in agreement with the relevant guidelines and regulations for handling human tissues. The use of patient tissues was approved by the Regional Ethical Review Board (permit no. 03-482).

Immunohistochemistry

Immunohistochemistry was performed as previously described (33). In brief, the tissue slides were deparaffinized in xylene, rehydrated through graded alcohols, incubated for antigen retrieval in the Retriever 2100 (ProteoGenix). Endogenous peroxidase activity was blocked with 3% hydrogen peroxidase in methanol. After incubation with the primary antibody at 4°C overnight, the sections were incubated with REAL EnVision detection system (DAKO). The reaction was visualized by REAL DAB+ chromogen (DAKO) under microscopic control. After stopping the reaction with tap water, the sections were treated with hematoxylin, dehydrated, and mounted. Negative controls are shown in fig. S11.

Statistical analysis

Statistical analyses between all groups were performed using one-way analysis of variance (ANOVA). F test was used to derive P values. Quantification of immunoblots was done by measuring the density of the bands using the AIDA 2D densitometry program. Data are means ± SEM of three independent experiments (n = 3), except for figs, S6, S8, and S10 (B and C), which were repeated twice (n = 2). P values of <0.05 were considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001 within one set of samples; §P < 0.05, §§P < 0.01, §§§P < 0.001 between different sets of samples).

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/10/486/eaal4186/DC1

Fig. S1. AKT interacts with TRAF6 in a TGF-β–dependent manner.

Fig. S2. TGF-β induces Lys63-linked polyubiquitylation and activation of AKT.

Fig. S3. TGF-β–induced polyubiquitylation and activation of AKT require the E3 ligase activity of TRAF6.

Fig. S4. TGF-β–induced activation of AKT depends on the PI3K pathway and is not dependent on the kinase activities of TβRI or TβRII.

Fig. S5. TGF-β–induced activation of AKT is not dependent on the kinase activity of TAK1 or the PDGFR.

Fig. S6. SMAD7 is required for Lys63-linked polyubiquitylation of p85α and AKT.

Fig. S7. TRAF6, pAKT Ser473, TβRI, and p85α colocalize at the cell membrane.

Fig. S8. Mutation of Lys513 and Lys519 to arginine does not affect binding of p85α to p110α.

Fig. S9. PI3K activity is increased by TGF-β stimulation and requires the E3 ligase activity of TRAF6 but not the kinase activity of TβRI.

Fig. S10. TGF-β–induced polyubiquitylation of p85α on Lys513 and Lys519 is important for AKT activation and migration of p85−/− MEFs.

Fig. S11. Negative control for immunohistochemistry and in situ PLA.

Fig. S12. Mass spectrometry analysis.

REFERENCES AND NOTES

Acknowledgments: We thank A. Moustakas and our colleagues at the Ludwig Institute for Cancer Research, Uppsala Branch and the Department of Medical Biosciences, Umeå University, Sweden for valuable discussions. We also thank S. K. Gudey for technical assistance and U. Engström and Å. Engström for mass spectrometry analysis. Funding: This work was partly supported by grants from the Knut and Alice Wallenberg Foundation (KAW 2012.0090), Swedish Cancer Society (CAN 2014/674), and Swedish Medical Research Council (K2013-66X-15284-04-4 and 2015-02757). Author contributions: A.H., N.T., A.M., and J.S. performed the experiments. A.B. provided the human samples. S.I. generated the Smad7−/− MEFs. C.-H.H. and M.L. planned the project. A.H., A.B., C.-H.H., and M.L. prepared the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article