In normal epithelial tissues, the multifunctional cytokine transforming growth factor-β (TGF-β) acts as a tumor suppressor through growth inhibition and induction of differentiation whereas in advanced cancers, TGF-β promotes tumor progression through induction of tumor invasion, neoangiogenesis, and immunosuppression. The molecular mechanisms through which TGF-β shifts from a tumor suppressor to a tumor enhancer are poorly understood. We now show a role for the tumor suppressor phosphatase and tensin homologue deleted on chromosome 10 (PTEN) in repressing the protumorigenic effects of TGF-β. The TGF-β effector SMAD3 inducibly interacts with PTEN on TGF-β treatment under endogenous conditions. RNA interference (RNAi) suppression of PTEN expression enhances SMAD3 transcriptional activity and TGF-β–mediated induction of SMAD3 target genes whereas reconstitution of PTEN in a null cancer cell line represses the expression of TGF-β–regulated target genes. Targeting PTEN expression through RNAi in a PTEN wild-type cell line increases TGF-β–mediated invasion but does not affect TGF-β–mediated growth inhibition. Reconstitution of PTEN expression in a PTEN-null cell line blocks TGF-β–induced invasion but does not modulate TGF-β–mediated growth regulation. These effects are distinct from Akt and Forkhead family members that also interact with SMAD3 to regulate apoptosis or proliferation, respectively. Pharmacologic inhibitors targeting TGF-β receptors and phosphatidylinositol 3-kinase signaling downstream from PTEN cooperate to block TGF-β–mediated invasion. Thus, the loss of PTEN expression in human cancers may contribute to a role for TGF-β as a tumor enhancer with specific effects on cellular motility and invasion. (Cancer Res 2005; 65(24): 11276-81)

Transforming growth factor-β (TGF-β) regulates a diverse set of biological activities, including proliferation, apoptosis, differentiation, motility, extracellular matrix deposition, and angiogenesis (1). The breadth of cellular responses is derived from the hundreds of genes transcriptionally regulated by TGF-β. TGF-β initiates cellular effects through the activation of serine/threonine kinase receptors that phosphorylate and activate the intracellular mediators SMAD2 and SMAD3 (24) to promote or inhibit the transcription of target genes (46). The ability of TGF-β to modulate gene expression partially depends on interactions with other signaling pathways. Recent reports highlight the importance of interactions between SMAD3 and components of phosphatidylinositol 3-kinase (PI3K) signaling, a prosurvival pathway activated in many cancers. Akt directly binds SMAD3, sequestering SMAD3 in the cytoplasm and inhibiting TGF-β-induced apoptosis (7, 8), whereas FOXO family members interact with SMAD3 to activate transcription of p21WAF1/CIP1, but not plasminogen activator inhibitor-1 (PAI-1; ref. 9). Phosphatase and tensin homologue (PTEN) is a tumor suppressor deleted or mutated in many solid tumors, including glioblastomas (10). As PTEN antagonizes the effects of PI3K on phosphoinositide phosphorylation, the loss of PTEN permits the constitutive activation of signaling downstream from PI3K (10) to increase cellular resistance to apoptosis and tumor cell invasion. We report a novel mechanism through which PTEN could prevent invasion during cancer progression: PTEN can functionally interact with SMAD3 to negatively regulate TGF-β signaling and prevent TGF-β–mediated invasion.

Retroviral expression. Indicated sense sequences for PTEN RNA interference (RNAi) were cloned into pSuperRetropuro vector (OligoEngine, Seattle, WA). Short hairpin RNA 1 (shRNA1) human PTEN bp 700 to 722: 5′-CGGGAAGACAAGTTCATGTACTT-3′. shRNA2 human PTEN bp 319 to 339: 5′-GATCTTGACCAATGGCTAAGT-3′. pBabe-bleo vector containing PTEN, C124G, or G129E phosphatase-dead mutants was provided by Frank Furnari (Ludwig Institute for Cancer Research, San Diego, CA).

Immunoprecipitation. Endogenous lysates, collected from HaCaT cells treated with 100 pmol/L TGF-β for 15 minutes, were exposed to α-PTEN or α-matrix metalloproteinase 7 (Santa Cruz Biotechnology, Santa Cruz, CA) and protein A/G beads (Amersham, Piscataway, NJ). In overexpression studies, hemaglutinin (HA)- or Flag-tagged constructs were expressed in 293T cells using FuGENE 6 (Roche, Indianapolis, IN) and immunoprecipitated using α-HA (Santa Cruz Biotechnology).

Western blotting. Cell lysates were prepared as previously described (11) using α-SMAD2, α-SMAD4 (Zymed, Carlsbad, CA), α-SMAD3, α-PTEN, α-HA, α-PAI-1, α-p21, α-c-myc (Santa Cruz Biotechnology), α-Flag (Stratagene, La Jolla, CA), α-pAkt, α-Akt (Cell Signaling, Beverly, MA), and α-tubulin (Sigma, St. Louis, MO).

Luciferase reporter assay. Luciferase assays were done as previously described (11) with the addition of 100 pmol/L TGF-β or buffer 24 hours after transfection.

Thymidine incorporation assay. Thymidine incorporation assays were done as previously described (11) with addition of TGF-β and/or buffer 20 hours before thymidine addition.

Migration and invasion assay. For HaCaT, pretreated cells were wounded and treated with the indicated drug, DMSO, 100 pmol/L TGF-β, or buffer overnight. For U87MG, 25,000 pretreated cells were plated in serum-free media in the upper chamber of Matrigel-coated wells or uncoated control chambers (BD Biosciences, Bedford, MA). The lower chamber contained 1% serum with or without TGF-β. The ratio of cells invading through Matrigel to cells migrating through the control chamber determined the percentage of invading cells.

TGF-β induces SMAD3 and PTEN interaction. Based on prior reports showing physical and functional interactions between members of the TGF-β and PI3K signaling pathways, we hypothesized a physical or functional role for the PI3K pathway inhibitor PTEN in the regulation of TGF-β signaling. To determine if PTEN could physically interact with TGF-β effector SMADs, we did coimmunoprecipitation experiments in HaCaT cells, a human keratinocyte cell line with intact TGF-β responses and wild-type PTEN expression. Endogenous PTEN interacted with endogenous SMAD3 and, less significantly, with SMAD2 in a TGF-β-responsive manner (Fig. 1A). The structural basis of the interaction between SMAD3 and PTEN was further determined in overexpression studies. PTEN inducibly coimmunoprecipitated with full-length or NH2-terminal SMAD3 (Fig. 1B), suggesting that PTEN binds to the Mad homology 1 (MH1) domain of SMAD3 that contains the DNA-binding domain. Although PTEN did not interact with the SMAD3 COOH terminus or linker region (Fig. 1B), TGF-β-induced phosphorylation of the COOH-terminal SSxS sequence is likely important for conformational changes in SMAD3 to permit PTEN binding as the nonphosphorylated COOH-terminal MH2 domain interacts with the NH2-terminal MH1 domain (2). Indeed, a SMAD3 mutant in which COOH-terminal serines were mutated to alanine, mimicking nonphosphorylated SMAD3, did not inducibly interact with PTEN (Fig. 1C). These data suggest that TGF-β–mediated phosphorylation of COOH-terminal serines induces a conformational change in SMAD3 to permit NH2-terminal binding to PTEN.

Figure 1.

TGF-β treatment induces a physical interaction of SMAD2 and SMAD3 with PTEN. A, endogenous SMAD2 and SMAD3 interact with PTEN. PTEN was immunoprecipitated from whole-cell lysates prepared from HaCaT cells in the presence and absence of 100 pmol/L TGF-β for 15 minutes, resolved by SDS-PAGE, and immunoblotted for SMAD2, SMAD3, SMAD4, and PTEN. An immunoglobulin G (IgG)–matched control immunoprecipitation was done in parallel. B, PTEN interacts with the NH2 terminus of SMAD3. 293T cells were transfected with HA-tagged PTEN and Flag-tagged full-length SMAD3 (S3) or SMAD3 deletion mutants containing the DNA binding MH1 region and linker region (NL) or the linker region and MH2 ligand binding domain (LC). Cell lysates were prepared, PTEN and SMAD were immunoprecipitated, resolved by SDS-PAGE, and immunoblotted for the presence of PTEN and SMAD3 using the expression tags. C, COOH-terminal phosphorylation of SMAD3 is required for TGF-β-inducible binding to PTEN. 293T cells were transfected with HA-tagged PTEN and Flag-tagged wild-type SMAD3 (S3) or mutants of SMAD3 in which COOH-terminal serines have been mutated to alanine (S→A) or aspartic acid residues (S→D). Cell lysates were prepared, PTEN and SMAD were immunoprecipitated, resolved by SDS-PAGE, and immunoblotted for the presence of PTEN and SMAD3 using the expression tags. D, SMAD3 interacts with the NH2 terminus of PTEN. 293T cells were transfected with HA-tagged full-length PTEN or deletion mutants of PTEN containing amino acids 1 to 274 in the absence (1-274) or presence (▵274-346) of the extreme carboxy terminus and Flag-tagged SMAD3.

Figure 1.

TGF-β treatment induces a physical interaction of SMAD2 and SMAD3 with PTEN. A, endogenous SMAD2 and SMAD3 interact with PTEN. PTEN was immunoprecipitated from whole-cell lysates prepared from HaCaT cells in the presence and absence of 100 pmol/L TGF-β for 15 minutes, resolved by SDS-PAGE, and immunoblotted for SMAD2, SMAD3, SMAD4, and PTEN. An immunoglobulin G (IgG)–matched control immunoprecipitation was done in parallel. B, PTEN interacts with the NH2 terminus of SMAD3. 293T cells were transfected with HA-tagged PTEN and Flag-tagged full-length SMAD3 (S3) or SMAD3 deletion mutants containing the DNA binding MH1 region and linker region (NL) or the linker region and MH2 ligand binding domain (LC). Cell lysates were prepared, PTEN and SMAD were immunoprecipitated, resolved by SDS-PAGE, and immunoblotted for the presence of PTEN and SMAD3 using the expression tags. C, COOH-terminal phosphorylation of SMAD3 is required for TGF-β-inducible binding to PTEN. 293T cells were transfected with HA-tagged PTEN and Flag-tagged wild-type SMAD3 (S3) or mutants of SMAD3 in which COOH-terminal serines have been mutated to alanine (S→A) or aspartic acid residues (S→D). Cell lysates were prepared, PTEN and SMAD were immunoprecipitated, resolved by SDS-PAGE, and immunoblotted for the presence of PTEN and SMAD3 using the expression tags. D, SMAD3 interacts with the NH2 terminus of PTEN. 293T cells were transfected with HA-tagged full-length PTEN or deletion mutants of PTEN containing amino acids 1 to 274 in the absence (1-274) or presence (▵274-346) of the extreme carboxy terminus and Flag-tagged SMAD3.

Close modal

The phosphatase activity of PTEN is located on the NH2 terminus (12, 13) whereas phosphorylation of the COOH terminus of PTEN regulates stability and binding to other proteins (14, 15). In overexpression experiments, TGF-β induced an interaction between SMAD3 and the NH2 terminus of PTEN (amino acids 1-274) despite reduced PTEN protein stability on deletion of the COOH terminus (Fig. 1D). Although SMAD3 bound to PTEN in or near its phosphatase domain, the interaction did not require the phosphatase activity of PTEN as SMAD3 inducibly interacted with PTEN mutants that lack total phosphatase (C124G) or lipid phosphatase (G129E) activities (data not shown; ref. 16). The addition of the extreme PTEN COOH terminus to the NH2 terminus of PTEN (the Δ274-346 mutant) prevented the induction of SMAD3 binding by TGF-β (Fig. 1D), indicating that the COOH terminus of PTEN regulates SMAD3 binding. Furthermore, a PTEN mutant lacking COOH-terminal phosphorylation sites, A5, bound strongly to SMAD3 (data not shown), showing that phosphorylation of the COOH tail of PTEN negatively regulates the interaction with SMAD3. Thus, SMAD3 and PTEN physically associate in both NH2 termini with regulation by COOH termini of each protein.

RNAi to PTEN increases transcription of TGF-β-responsive promoters. To determine the functional significance of PTEN expression on TGF-β/SMAD3–mediated transcription, we targeted PTEN expression with stable expression of shRNA directed against PTEN with two different shRNA sequences that decreased PTEN expression by ∼75% and 99%, respectively (Fig. 2A). In HaCaT, shRNA targeting of PTEN expression augmented TGF-β transcriptional activation of multiple TGF-β-responsive promoters (Fig. 3A and B and Supplementary Fig. S1A and B). Transcriptional activity of the TGF-β-responsive 3TP-luc reporter was induced by TGF-β in an inverse relationship to PTEN expression levels (Fig. 3A). Targeting PTEN expression additionally enhanced TGF-β–mediated transcription of a reporter containing concatamerized SMAD3 DNA binding elements (SBE; Supplementary Fig. S1A; ref. 3). Expression of SMAD3 in cells expressing an SBE reporter induced transcription in a dose-dependent manner whereas decreased PTEN expression enhanced this effect (Fig. 3B and Supplementary Fig. S1B). Together, theses results show that PTEN represses TGF-β- and SMAD3-mediated transcriptional activation.

Figure 2.

PTEN does not prevent SMAD3 nuclear accumulation but can modulate TGF-β-mediated protein expression. A, RNAi targeting of PTEN expression increases TGF-β-mediated PAI-1 protein expression. HaCaT cells stably infected with retroviruses encoding shRNA sequences directed against PTEN (shRNA1 or shRNA2) were treated with increasing concentrations of TGF-β for 24 hours. Total cell lysates were resolved by SDS-PAGE and immunoblotted for PAI-1, p21, c-myc, and PTEN, with tubulin as a loading control. B, reconstitution of PTEN expression in PTEN-null cells prevents TGF-β-mediated induction of PAI-1. PTEN-null U87MG glioma cells stably infected with wild-type PTEN or phosphatase-dead mutants of PTEN (C124G, G129E) were treated with increasing concentrations of TGF-β for 24 hours. Total cell lysates were analyzed for PAI-1, p21, c-myc, pAkt, Akt, HA (for PTEN or mutant PTEN), and tubulin as a loading control. C, RNAi targeting of PTEN expression does not enhance TGF-β-mediated SMAD3 nuclear translocation. HaCaT cells stably infected with retroviruses encoding two shRNA sequences directed against PTEN (shRNA1 or shRNA2) were treated with increasing concentrations of TGF-β for 15 minutes. Isolated nuclear extracts were analyzed for total expression of SMAD3 or β-catenin as a control. D, RNAi targeting of PTEN expression does not enhance TGF-β-mediated SMAD3 COOH-terminal serine phosphorylation. HaCaT cells stably infected with retroviruses encoding shRNA sequences directed against PTEN (shRNA1 or shRNA2) were treated with increasing concentrations of TGF-β for 15 minutes. Isolated nuclear and cytoplasmic extracts were resolved by SDS-PAGE and immunoblotted for expression of phosphorylated SMAD2/3 with antibodies recognizing the COOH-terminal SSxS motif. Whole-cell lysates were also analyzed for total SMAD2 and SMAD3.

Figure 2.

PTEN does not prevent SMAD3 nuclear accumulation but can modulate TGF-β-mediated protein expression. A, RNAi targeting of PTEN expression increases TGF-β-mediated PAI-1 protein expression. HaCaT cells stably infected with retroviruses encoding shRNA sequences directed against PTEN (shRNA1 or shRNA2) were treated with increasing concentrations of TGF-β for 24 hours. Total cell lysates were resolved by SDS-PAGE and immunoblotted for PAI-1, p21, c-myc, and PTEN, with tubulin as a loading control. B, reconstitution of PTEN expression in PTEN-null cells prevents TGF-β-mediated induction of PAI-1. PTEN-null U87MG glioma cells stably infected with wild-type PTEN or phosphatase-dead mutants of PTEN (C124G, G129E) were treated with increasing concentrations of TGF-β for 24 hours. Total cell lysates were analyzed for PAI-1, p21, c-myc, pAkt, Akt, HA (for PTEN or mutant PTEN), and tubulin as a loading control. C, RNAi targeting of PTEN expression does not enhance TGF-β-mediated SMAD3 nuclear translocation. HaCaT cells stably infected with retroviruses encoding two shRNA sequences directed against PTEN (shRNA1 or shRNA2) were treated with increasing concentrations of TGF-β for 15 minutes. Isolated nuclear extracts were analyzed for total expression of SMAD3 or β-catenin as a control. D, RNAi targeting of PTEN expression does not enhance TGF-β-mediated SMAD3 COOH-terminal serine phosphorylation. HaCaT cells stably infected with retroviruses encoding shRNA sequences directed against PTEN (shRNA1 or shRNA2) were treated with increasing concentrations of TGF-β for 15 minutes. Isolated nuclear and cytoplasmic extracts were resolved by SDS-PAGE and immunoblotted for expression of phosphorylated SMAD2/3 with antibodies recognizing the COOH-terminal SSxS motif. Whole-cell lysates were also analyzed for total SMAD2 and SMAD3.

Close modal
Figure 3.

PTEN represses TGF-β– and SMAD3-mediated transcription. A, RNAi targeting of PTEN expression with shRNA increases TGF-β–mediated transcription of the 3TP-luc reporter. HaCaT cells stably infected with shRNA directed against PTEN (shRNA1 or shRNA2) were transfected with 3TP-luc, a reporter containing TGF-β–responsive elements of the PAI-1 promoter, and β-galactosidase and then treated with 100 pmol/L TGF-β for 24 hours. *, P < 0.05, ANOVA comparison of control versus TGF-β–treated pSuper vector–infected cells. **, P < 0.01, ANOVA comparison of TGF-β–treated pSuper vector–infected versus shRNA1-infected cells. ***, P < 0.001, ANOVA comparison of TGF-β–treated pSuper vector–infected versus shRNA2-infected cells. B, SMAD3-mediated transcription of the 12xCAGA-luc reporter is enhanced with the targeted disruption of PTEN expression with shRNA. HaCaT cells stably infected with shRNA directed against PTEN (shRNA1 or shRNA2) were transfected with pGL2 basic control or 12xCAGA-luc, a reporter containing concatamerized SMAD3 DNA binding elements, and β-galactosidase in the presence of increasing concentrations of SMAD3. *, P < 0.01, ANOVA comparison of control versus 0.2 μg SMAD3–transfected pSuper vector–infected cells. **, P < 0.001, ANOVA comparison of 0.2 μg SMAD3–transfected pSuper vector–infected versus PTEN shRNA1– or PTEN shRNA2–infected cells. ***, P < 0.001, ANOVA comparison of 0.5 μg SMAD3–transfected pSuper vector versus PTEN shRNA1– or PTEN shRNA2–infected cells. C, targeted disruption of PTEN expression with shRNA increases basal transcription of the TIE reporter but does not prevent TGF-β–mediated repression. HaCaT cells stably infected with shRNA directed against PTEN (shRNA1 or shRNA2) were transfected with TIE-luc, a reporter containing the TGF-β inhibitory elements from the c-myc promoter, or MIE-luc, a control reporter containing mutated TGF-β inhibitory elements. Cells were also transfected with β-galactosidase and then treated with 100 pmol/L TGF-β for 24 hours. *, P < 0.001, ANOVA comparison of untreated pSuper retro–infected cells versus TGF-β–treated pSuper retro–infected cells or untreated shRNA2-infected cells as indicated. **, P < 0.001, ANOVA comparison of TGF-β–treated pSuper retro–infected cells versus TGF-β–treated shRNA2-infected cells. D, restoration of PTEN blocks SMAD3-mediated transcription of the 4xCAGA-luc reporter. PTEN-null U87MG cells were transfected with pGL2 basic control vector or 4xCAGA-luc containing concatamerized SMAD3 DNA binding elements in the presence of SMAD3 and wild-type PTEN or phosphatase-dead PTEN mutants (C124G and G129E). *P < 0.001, ANOVA comparison of control versus SMAD3-transfected vector control cells. **, P < 0.001, ANOVA comparison of SMAD3- and PTEN-transfected cells versus SMAD3- and vector control–transfected or SMAD3- and C124G mutant PTEN–transfected cells. ***, P < 0.05, ANOVA comparison of SMAD3- and PTEN-transfected cells versus SMAD3- and G129E mutant PTEN–transfected cells.

Figure 3.

PTEN represses TGF-β– and SMAD3-mediated transcription. A, RNAi targeting of PTEN expression with shRNA increases TGF-β–mediated transcription of the 3TP-luc reporter. HaCaT cells stably infected with shRNA directed against PTEN (shRNA1 or shRNA2) were transfected with 3TP-luc, a reporter containing TGF-β–responsive elements of the PAI-1 promoter, and β-galactosidase and then treated with 100 pmol/L TGF-β for 24 hours. *, P < 0.05, ANOVA comparison of control versus TGF-β–treated pSuper vector–infected cells. **, P < 0.01, ANOVA comparison of TGF-β–treated pSuper vector–infected versus shRNA1-infected cells. ***, P < 0.001, ANOVA comparison of TGF-β–treated pSuper vector–infected versus shRNA2-infected cells. B, SMAD3-mediated transcription of the 12xCAGA-luc reporter is enhanced with the targeted disruption of PTEN expression with shRNA. HaCaT cells stably infected with shRNA directed against PTEN (shRNA1 or shRNA2) were transfected with pGL2 basic control or 12xCAGA-luc, a reporter containing concatamerized SMAD3 DNA binding elements, and β-galactosidase in the presence of increasing concentrations of SMAD3. *, P < 0.01, ANOVA comparison of control versus 0.2 μg SMAD3–transfected pSuper vector–infected cells. **, P < 0.001, ANOVA comparison of 0.2 μg SMAD3–transfected pSuper vector–infected versus PTEN shRNA1– or PTEN shRNA2–infected cells. ***, P < 0.001, ANOVA comparison of 0.5 μg SMAD3–transfected pSuper vector versus PTEN shRNA1– or PTEN shRNA2–infected cells. C, targeted disruption of PTEN expression with shRNA increases basal transcription of the TIE reporter but does not prevent TGF-β–mediated repression. HaCaT cells stably infected with shRNA directed against PTEN (shRNA1 or shRNA2) were transfected with TIE-luc, a reporter containing the TGF-β inhibitory elements from the c-myc promoter, or MIE-luc, a control reporter containing mutated TGF-β inhibitory elements. Cells were also transfected with β-galactosidase and then treated with 100 pmol/L TGF-β for 24 hours. *, P < 0.001, ANOVA comparison of untreated pSuper retro–infected cells versus TGF-β–treated pSuper retro–infected cells or untreated shRNA2-infected cells as indicated. **, P < 0.001, ANOVA comparison of TGF-β–treated pSuper retro–infected cells versus TGF-β–treated shRNA2-infected cells. D, restoration of PTEN blocks SMAD3-mediated transcription of the 4xCAGA-luc reporter. PTEN-null U87MG cells were transfected with pGL2 basic control vector or 4xCAGA-luc containing concatamerized SMAD3 DNA binding elements in the presence of SMAD3 and wild-type PTEN or phosphatase-dead PTEN mutants (C124G and G129E). *P < 0.001, ANOVA comparison of control versus SMAD3-transfected vector control cells. **, P < 0.001, ANOVA comparison of SMAD3- and PTEN-transfected cells versus SMAD3- and vector control–transfected or SMAD3- and C124G mutant PTEN–transfected cells. ***, P < 0.05, ANOVA comparison of SMAD3- and PTEN-transfected cells versus SMAD3- and G129E mutant PTEN–transfected cells.

Close modal

RNAi to PTEN enhances transcription of a promoter repressed by TGF-β. To determine if the effects of PTEN on TGF-β–mediated transcription were limited to activation, we evaluated the activity of a promoter repressed by TGF-β (4). Loss of PTEN expression enhanced the basal activity of a reporter with concatamerized TGF-β inhibitory elements (TIE; Fig. 3C). Whereas targeting PTEN expression did not abolish the transcriptional repression of TIE sequences by TGF-β, TGF-β did not repress transcription to basal levels in the absence of PTEN (Fig. 3C).

Reconstitution of PTEN, but not of phosphatase-dead mutant PTEN, blocks TGF-β-responsive promoters. In experiments reciprocal to those in HaCaT cells, we reconstituted the expression of wild-type PTEN or phosphatase-dead PTEN mutants in the PTEN-null human U87MG glioma cell line and evaluated the ability of SMAD3 to induce transcription of an SBE reporter. Wild-type PTEN inhibited SMAD3-dependent transcription whereas the expression of C124G and G129E mutant PTEN proteins did not (Fig. 3D). Reconstitution of wild-type PTEN expression in U87MG cells also inhibited 3TP-luc in a dose-dependent manner (data not shown). Together, these data show that PTEN inhibits SMAD3-dependent transcription through a mechanism requiring PTEN phosphatase activity although the phosphatase activity itself is not required for the PTEN-SMAD3 interaction.

PTEN expression does not prevent TGF-β-induced SMAD3 nuclear accumulation. As SMAD3 and PTEN interact on both functional and physical levels, we sought to elucidate the mechanism by which PTEN inhibits SMAD3-dependent transcription. SMAD3-mediated transcription is regulated by its COOH-terminal phosphorylation and translocation to the nucleus; thus, we examined the effect of PTEN expression on this pathway. RNAi suppression of PTEN expression in HaCaT cells did not enhance TGF-β-induced SMAD3 nuclear accumulation (Fig. 2C) or COOH-terminal phosphorylation of SMAD2 or SMAD3 (Fig. 2D). Restoration of wild-type PTEN expression in U87MG cells also did not prevent TGF-β–mediated SMAD3 nuclear translocation or COOH-terminal phosphorylation (data not shown). Although other PI3K pathway members (specifically Akt) regulate SMAD3 transcription by cytoplasmic sequestration (7, 8), our data show that PTEN regulates SMAD3 through a distinct mechanism.

RNAi to PTEN alters TGF-β-regulated c-myc and PAI-1 expression. As TGF-β modulates cellular behaviors through transcriptional regulation of target genes, we examined the effects of PTEN on the ability of TGF-β to regulate expression of several proteins that critically regulate cell behavior. TGF-β potently induces the expression of several cyclin-dependent kinase inhibitors, including p21WAF1/CIP1 (5), while repressing c-myc expression (4) to decrease cell cycle progression. TGF-β induced p21WAF1/CIP1 and repressed c-myc expression in HaCaT cells (Fig. 2A) that are potently growth inhibited by TGF-β (Fig. 4A). In contrast to FOXO family member effects on TGF-β signaling, the loss of PTEN expression did not alter the ability of TGF-β to induce p21WAF1/CIP1 expression (Fig. 2A). shRNA targeting of PTEN expression in HaCaT elevated basal c-myc expression but did not completely block TGF-β–mediated repression of c-myc expression (Fig. 2B), similar to our results in transcriptional assays (Fig. 3C).

Figure 4.

Loss of PTEN enhances TGF-β–mediated invasion but not growth inhibition. A, loss of PTEN expression is not sufficient to block TGF-β–mediated growth inhibition. HaCaT cells, which are potently growth inhibited by TGF-β, were infected with retrovirus encoding shRNA directed against PTEN. Cells were treated with increasing concentrations of TGF-β for 24 hours and cell proliferation was measured by thymidine incorporation. *, P < 0.001, ANOVA comparison of control versus TGF-β treatment. B, TGF-β–mediated migration is enhanced by the loss of PTEN expression. HaCaT cells stably infected with shRNA directed against PTEN were treated with 100 pmol/L TGF-β for 24 hours. Cells were wounded and treated with 100 pmol/L TGF-β for an additional 18 hours. The average area of the wounds was measured and recorded as median area. *, P < 0.01, ANOVA comparison of pSuper retro vector control versus TGF-β treatment. **, P < 0.001, ANOVA comparison of PTEN shRNA control versus TGF-β treatment. Representative images were taken at ×200. C, TGF-β-mediated invasion is repressed by the restoration of PTEN to PTEN-null cells. PTEN-null U87MG glioma cells infected with vector control or PTEN were treated with 100 pmol/L TGF-β for 24 hours and allowed to migrate into Matrigel-coated chambers or control inserts for 18 hours. The number of migrating and invading cells was counted and recorded by a blinded observer. The ratio of invading cells (Matrigel) to migrating cells (control insert) is plotted. Representative images of U87 cells in the Matrigel invasion assay are shown at ×200. D, dual targeting of the TGF-β and PTEN pathways with small-molecule inhibitors prevents TGF-β-mediated migration. HaCaT cells were treated for 24 hours with 100 pmol/L TGF-β in the presence of SB-431542 (SB), a TGF-β type I receptor inhibitor, or rapamycin (Rap), where indicated. Cells were wounded and treated with TGF-β in the presence or absence of inhibitor for an additional 18 hours. The average area of the wounds was measured with ImageJ software and recorded as median area. *, P < 0.001, ANOVA comparison of control versus TGF-β treatment in the absence of drug. **, P < 0.001, ANOVA comparison of TGF-β treatments in the absence or presence of SB-431542. ***, P < 0.01, ANOVA comparison of TGF-β treatments in the absence or presence of rapamycin. ****, P < 0.05, ANOVA comparison of TGF-β treatments in the presence of rapamycin or in the presence of rapamycin plus SB-431542. Representative images are shown at ×200.

Figure 4.

Loss of PTEN enhances TGF-β–mediated invasion but not growth inhibition. A, loss of PTEN expression is not sufficient to block TGF-β–mediated growth inhibition. HaCaT cells, which are potently growth inhibited by TGF-β, were infected with retrovirus encoding shRNA directed against PTEN. Cells were treated with increasing concentrations of TGF-β for 24 hours and cell proliferation was measured by thymidine incorporation. *, P < 0.001, ANOVA comparison of control versus TGF-β treatment. B, TGF-β–mediated migration is enhanced by the loss of PTEN expression. HaCaT cells stably infected with shRNA directed against PTEN were treated with 100 pmol/L TGF-β for 24 hours. Cells were wounded and treated with 100 pmol/L TGF-β for an additional 18 hours. The average area of the wounds was measured and recorded as median area. *, P < 0.01, ANOVA comparison of pSuper retro vector control versus TGF-β treatment. **, P < 0.001, ANOVA comparison of PTEN shRNA control versus TGF-β treatment. Representative images were taken at ×200. C, TGF-β-mediated invasion is repressed by the restoration of PTEN to PTEN-null cells. PTEN-null U87MG glioma cells infected with vector control or PTEN were treated with 100 pmol/L TGF-β for 24 hours and allowed to migrate into Matrigel-coated chambers or control inserts for 18 hours. The number of migrating and invading cells was counted and recorded by a blinded observer. The ratio of invading cells (Matrigel) to migrating cells (control insert) is plotted. Representative images of U87 cells in the Matrigel invasion assay are shown at ×200. D, dual targeting of the TGF-β and PTEN pathways with small-molecule inhibitors prevents TGF-β-mediated migration. HaCaT cells were treated for 24 hours with 100 pmol/L TGF-β in the presence of SB-431542 (SB), a TGF-β type I receptor inhibitor, or rapamycin (Rap), where indicated. Cells were wounded and treated with TGF-β in the presence or absence of inhibitor for an additional 18 hours. The average area of the wounds was measured with ImageJ software and recorded as median area. *, P < 0.001, ANOVA comparison of control versus TGF-β treatment in the absence of drug. **, P < 0.001, ANOVA comparison of TGF-β treatments in the absence or presence of SB-431542. ***, P < 0.01, ANOVA comparison of TGF-β treatments in the absence or presence of rapamycin. ****, P < 0.05, ANOVA comparison of TGF-β treatments in the presence of rapamycin or in the presence of rapamycin plus SB-431542. Representative images are shown at ×200.

Close modal

In reciprocal experiments, reconstitution of wild-type or mutant PTEN expression in U87MG cells did not alter p21WAF1/CIP1 or c-myc expression (Fig. 2B). The failure of TGF-β to induce p21WAF1/CIP1 or repress c-myc expression in U87MG cells shows two molecular mechanisms through which this glioma cell line is resistant to TGF-β–mediated growth inhibition. Furthermore, the restoration of PTEN did not reestablish TGF-β–mediated control of these cell cycle–related target genes (Fig. 2B).

To evaluate TGF-β regulation of an extracellular matrix protein implicated in tumor cell migration and invasion, we determined the effects of targeting PTEN expression on the ability of TGF-β to regulate PAI-1 expression. Although PAI-1 inhibits proteases involved in invasion, PAI-1 effects are complex as PAI-1 may positively regulate invasion in a cell type/state–dependent manner (17). Indeed, PAI-1 expression is an independent marker of poor prognosis for glioblastomas, a highly invasive tumor (18). TGF-β induced PAI-1 protein expression in both HaCaT and U87MG (Fig. 2A and B), showing that TGF-β–mediated regulation of the extracellular matrix is maintained in these cell lines. RNAi to PTEN in HaCaT cells increased the efficacy with which TGF-β induced PAI-1 protein expression (Fig. 2A). In the reciprocal experiment, reconstititution of wild-type, but not phosphatase-dead PTEN, expression in U87MG cells blocked TGF-β–mediated PAI-1 expression (Fig. 2B). Thus, PTEN expression inversely correlates with the ability of TGF-β to induce PAI-1.

Loss of PTEN expression alone is not sufficient to prevent TGF-β–mediated growth inhibition. Changes in TGF-β–mediated gene expression observed in the presence and absence of PTEN could modulate cellular behaviors, such as proliferation or invasion, which are important in cancer. In HaCaT cells, the elevation of c-myc expression detected with RNAi targeting of PTEN (Fig. 2A) did not increase basal cell proliferation or prevent TGF-β–mediated growth inhibition (Fig. 4A), likely due to the potent TGF-β–mediated induction of p21WAF1/CIP1 (Fig. 2A). Similarly, reconstitution of PTEN expression in U87MG cells did not enhance TGF-β–mediated growth inhibition (Supplementary Fig. S1C). Together, these data show that restoration of PTEN expression in PTEN-null cancer cells does not reestablish the potent growth inhibitory effects of TGF-β observed in epithelial cells and that loss of PTEN expression does not prevent TGF-β–mediated growth inhibition during cancer progression.

PTEN represses TGF-β–mediated cellular migration and invasion. Both TGF-β and PTEN regulate tumor cell invasion. We therefore examined the effects of PTEN expression on TGF-β mediated cellular motility/invasion. The effects of decreased PTEN expression on HaCaT cell migration were determined in a wound closure assay because these cells did not invade through the artificial extracellular matrix of the Matrigel assay (data not shown). RNAi to PTEN specifically altered cell migration in response to TGF-β but did not effect basal migration of HaCaT cells (Fig. 4B). Similarly, reconstitution of wild-type PTEN expression in U87MG cells blocked TGF-β–mediated invasion in the Matrigel invasion assay (Fig. 4C). Together, these results confirm a functional interaction between PTEN and SMAD3 on TGF-β transcriptional targets with a biological consequence for TGF-β–mediated invasion.

Targeting TGF-β and mammalian target of rapamycin signaling with pharmacologic inhibitors prevents TGF-β–mediated migration. As we showed that antagonism of PTEN through RNAi increased TGF-β–mediated migration, mutation or loss of PTEN during cancer progression may promote proinvasive TGF-β signals in addition to effects of PI3K activation. During cancer progression, the balance of TGF-β effects on tumors shifts towards tumor progression as growth inhibition/differentiation responses are lost. We previously showed that a novel type I TGF-β receptor inhibitor, SB-431542, decreased glioma cell migration (11). Based on this prior work and our new understanding of the interaction between PTEN and TGF-β signaling, we investigated the effects of a combination of TGF-β receptor and PI3K pathway inhibitors. In parallel to our results showing minimal effect of PTEN on TGF-β–mediated proliferation, SB-431542 and rapamycin (sirolimus), an inhibitor of the PI3K downstream target mammalian target of rapamycin, did not cooperate to reduce cell proliferation although both agents inhibited basal proliferation of U87MG cells (data not shown). In contrast, the combination of SB-431542 and rapamycin blocked TGF-β–mediated migration more effectively than either monotherapy in HaCaT cells (Fig. 4D). Together, our results suggest that there may be a therapeutic benefit for the specific combination of SB-431542 and rapamycin as well as a potential benefit for other small-molecule inhibitor combinations that target both TGF-β and PI3K signals.

The dichotomous effects of TGF-β in cancer are now well appreciated but poorly understood. TGF-β–mediated growth inhibition and regulation of differentiation are important for suppressing early cancers whereas TGF-β–mediated extracellular matrix deposition promotes invasion in late cancers (1). As the mutation and deletion of the PTEN tumor suppressor gene occurs in many advanced cancers (12) in which TGF-β functions to promote tumor progression, we hypothesize that loss of PTEN contributes to this shift in TGF-β activity. The loss of PTEN alone does not permit escape from TGF-β-mediated growth inhibition but may facilitate cell proliferation induced by mutations in TGF-β-independent pathways. Further, the loss of PTEN may increase expression of PAI-1 and other proteins that contribute to tumor invasion in TGF-β-rich microenvironments. However, this novel role for PTEN in the control of tumor cell migration and invasion has yet to be shown in vivo. We are working to address the limitations of cell culture models through the use of genetically engineered mouse models of cancer.

The precise mechanism through which loss of PTEN elevates proinvasive TGF-β/SMAD3 signals remains under investigation. Our data show that the phosphatase activity of PTEN is required to inhibit TGF-β-mediated transcription. PTEN may dephosphorylate SMAD3 at sites other than the COOH terminus to regulate its activity. Other SMAD3 phosphorylation sites are already known to be regulated by mitogen-activated protein kinases and cyclin-dependent kinases (19, 20). As PTEN does not alter SMAD3 nuclear translocation, the mechanism through which PTEN regulates SMAD3/TGF-β activity is distinct from that of other PI3K signal transduction members. Furthermore, the PTEN/SMAD3 interaction broadens our knowledge of PI3K pathway components in TGF-β signaling and introduces a new level of complexity for interactions between these two major signal transduction pathways in cancer.

Most solid cancers develop from the dysregulation of multiple pathways. Therefore, the successful targeting of a single molecular pathway as with imatinib mesylate for chronic myelogenous leukemia is unlikely to induce widespread tumor responses: effective molecular targeted therapies will require targeting two or more molecular pathways in combination. In this study, we further validate the functional interactions between two important pathways involved in the pathogenesis of multiple cancers. Our results suggest that targeting both the TGF-β receptor and targets downstream of PI3K may benefit patients through the reduction of cancer cell proliferation and migration. Further, expression and activation of individual pathway components could partially determine cancer sensitivity to other pathway inhibitors. For example, disruption of Akt/PI3K activities may not completely reverse the effects of PTEN loss on TGF-β signaling but PTEN loss could be a marker for tumors with strong responses to anti-TGF-β therapies. Further study of the mechanisms of PTEN and other PI3K family member interactions with TGF-β signal transduction may both contribute to a greater understanding of their function in cancer and create new therapeutic opportunities.

Note: A.B. Hjelmeland is a Paul Brazen/American Brain Tumor Association Fellow. J.N. Rich is a Damon Runyon-Lilly Clinical Investigator and a Sidney Kimmel Cancer Foundation Translational Scholar.

Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

Grant support: Pediatric Brain Tumor Foundation of the United States (J.N. Rich), Accelerate Brain Cancer Cure (J.N. Rich), Southeastern Brain Tumor Foundation (A.B. Hjelmeland), American Heart Association (C.D. Kontos), and NIH grants NS047409, P50 CA108786 (J.N. Rich), and HL70165 (C.D. Kontos).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1
Siegel PM, Massague J. Cytostatic and apoptotic actions of TGF-β in hsomeostasis and cancer.
Nat Rev Cancer
2003
;
3
:
807
–21.
2
Shi Y, Massague J. Mechanisms of TGFβ signaling from the Cell Membrane to the Nucleus.
Cell
2003
;
113
:
685
–700.
3
Zawel L, Dai JL, Buckhaults P, et al. Human Smad3 and Smad4 are sequence-specific transcription activators.
Mol Cell
1998
;
1
:
611
–7.
4
Frederick JP, Liberati NT, Waddell DS, Shi Y, Wang XF. Transforming growth factor β-mediated transcriptional repression of c-myc is dependent on direct binding of Smad3 to a novel repressive Smad binding element.
Mol Cell Biol
2004
;
24
:
2546
–59.
5
Datto MB, Yu Y, Wang XF. Functional analysis of the transforming growth factor β responsive elements in the WAF1/Cip1/p21 promoter.
J Biol Chem
1995
;
270
:
28623
–8.
6
Westerhausen DR, Jr., Hopkins WE, Billadello JJ. Multiple transforming growth factor-β-inducible elements regulate expression of the plasminogen activator inhibitor type-1 gene in Hep G2 cells.
J Biol Chem
1991
;
266
:
1092
–100.
7
Conery AR, Cao Y, Thompson EA, et al. Akt interacts directly with SMAD3 to regulate the sensitivity to TGF-β induced apoptosis.
Nat Cell Biol
2004
;
6
:
366
–72.
8
Remy I, Montmarquette A, Michnick SW. PKB/Akt modulates TGF-β signaling through a direct interaction with Smad3.
Nat Cell Biol
2004
;
6
:
358
–65.
9
Seoane J, Le HV, Shen L, Anderson SA, Massague J. Integration of Smad and forkhead pathways in the control of neuroepithelial and glioblastoma cell proliferation.
Cell
2004
;
117
:
211
–23.
10
Haas-Kogan D, Shalev N, Wong M, et al. Protein kinase B (PKB/Akt) activity is elevated in glioblastoma cells due to mutation of the tumor suppressor PTEN/MMAC.
Curr Biol
1998
;
8
:
1195
–8.
11
Hjelmeland MD, Hjelmeland AB, Sathornsumetee S, et al. SB-431542, a small molecule transforming growth factor-β-receptor antagonist, inhibits human glioma cell line proliferation and motility.
Mol Cancer Ther
2004
;
3
:
737
–45.
12
Li J, Yen C, Liaw D, et al. PTEN, a putative protein tyrosine phosphatase gene mutated in human brain, breast, and prostate cancer.
Science
1997
;
275
:
1943
–7.
13
Lee JO, Yang H, Georgescu MM, et al. Crystal structure of the PTEN tumor suppressor: implications for its phosphoinositide phosphatase activity and membrane association.
Cell
1999
;
99
:
323
–34.
14
Georgescu MM, Kirsch KH, Akagi T, Shishido T, Hanafusa H. The tumor-suppressor activity of PTEN is regulated by its carboxyl-terminal region.
Proc Natl Acad Sci U S A
1999
;
96
:
10182
–7.
15
Vazquez F, Grossman SR, Takahashi Y, et al. Phosphorlation of the PTEN tail acts as an inhibitor switch by preventing its recruitment into a protein complex.
J Biol Chem
2001
;
276
:
48627
–30.
16
Myers MP, Pass I, Batty IH, et al. The lipid phosphatase activity of PTEN is critical for its tumor suppressor function.
Proc Natl Acad Sci U S A
1998
;
95
:
13513
–8.
17
Durand MK, Bodker JS, Christensen A, et al. Plasminogen activator inhibitor-I and tumour growth, invasion, and metastasis.
Thromb Haemost
2004
;
91
:
438
–49.
18
Muracciole X, Romain S, Dufour H, et al. PAI-1 and EGFR expression in adult glioma tumors: toward a molecular prognostic classification.
Int J Radiat Oncol Biol Phys
2002
;
52
:
592
–8.
19
Matsuura I, Denissova NG, Wang G, et al. Cyclin-dependent kinases regulate the antiproliferative function of Smads.
Nature
2004
;
430
:
226
–31.
20
Matsuura I, Wang G, He D, Liu F. Identification and characterization of ERK MAP kinase phosphorylation sites in Smad3.
Biochemistry
2005
;
44
:
12546
–53.

Supplementary data