The type I TGF- receptor TRI is sumoylated

To examine the sumoylation of TRI or TRII, we expressed Flag-tagged rat TRI or TRII with Myc-tagged SUMO-1. Cell lysate immunoprecipitations with anti-Flag antibodies, followed by western blotting, detected Myc-tagged, sumoylated TGF- receptors. As shown in Fig. 1a, SUMO was conjugated to TRI, but not TRII, resulting in a shift of more than 20 kDa, similar to that of other sumoylated proteins, indicating that TRI is post-translationally sumoylated in vivo. TRI sumoylation was increased when the E2-conjugating enzyme Ubc9 was co-expressed with SUMO-1, suggesting that Ubc9 is involved in the sumoylation of TRI (Fig. 1b). Under conditions of Ubc9 overexpression and proportionally insufficient E3 SUMO ligase expression, up to three sumoylated TRI forms were observed. Because only one SUMO-1 can be linked to a Lys residue, we assume that, under these conditions, the initial, site-specific sumoylation can confer additional TRI sumoylation at other sites.

Figure 1: The type I TGF- receptor TRI is sumoylated.

(a) TRI, but not TRII, is sumoylated. Lysates of COS cells, expressing Flag-tagged (F-) TRI or TRII and Myc-tagged (M-) SUMO-1, were subjected to immunoprecipitation (IP) with anti-Flag antibody, followed by western blotting (WB) with anti-Myc to detect sumoylated TGF- receptors. (b) Increasing expression of Ubc9 enhances TRI sumoylation. COS cells, expressing Flag-tagged TRI, Myc-tagged SUMO-1 and increasing levels of Ubc9, were lysed and subjected to immunoprecipitation, followed by western blotting with the indicated antibodies. (c) In vitro sumoylation of TRI. Immunopurified Flag-tagged TRI was incubated with or without recombinant SUMO-1, the E1 enzyme Aos1/Uba2 and the E2-conjugating enzyme Ubc9. The reaction mixture was analysed by western blotting with anti-Flag antibody. (d) TGF- induces sumoylation of endogenous TRI. Lysates of Mv1Lu or MDA-231 cells, treated with or without TGF-, were immunoprecipitated with anti-TRI and immunoblotted with antibody against SUMO-1. (e) TRI, but not other type I receptors, is sumoylated. 293T cells, ectopically expressing the indicated type I receptor, Myc-tagged SUMO-1 and Ubc9, were lysed, and sumoylation was analysed by western blotting.

Full size image (65 KB)

We next evaluated whether TRI can be sumoylated in vitro. Immunopurified TRI was incubated with SUMO-1, the E1 activating SUMO enzyme, Aos1/Uba2, and Ubc9 in the presence of ATP. Western blotting detected a band with a size that was compatible with the attachment of a SUMO-1 protein to TRI and corresponded to the sumoylated TRI in vivo. When the E1 or E2 enzyme, or SUMO-1, was absent, this band was not detected (Fig. 1c). This result suggests that TRI was sumoylated in vitro.

Immunoprecipitation of TRI from Mv1Lu or MDA-231 cells, treated with or without TGF-, and immunoblotting with antibodies against SUMO-1, revealed that endogenous TRI was sumoylated and that TGF- induced TRI sumoylation (Fig. 1d). These data indicate that receptor activation by TGF- may induce sumoylation of TRI.

We assessed whether other type I TGF- family receptors could be sumoylated. Each type I receptor was expressed in the presence or absence of SUMO-1 and Ubc9, and sumoylation was analysed by immunoblotting. Whereas TRI was sumoylated, other type I receptors were not (Fig. 1e).

TRI kinase activity and phosphorylation are required for sumoylation of the TRI receptor

To further characterize whether activation of TRI affects its sumoylation, as apparent by the TGF--induced TRI sumoylation (Fig. 1d), we compared the in vitro sumoylation efficacy of immunopurified wild-type TRI and activated TRI (caTRI) with a Thr 202 to Asp mutation (Thr 202 in rat TRI corresponds to Thr 204 in human TRI), resulting in elevated kinase activity⁸. As shown in Fig. 2a, caTRI was sumoylated much more efficiently than wild-type TRI, suggesting that TRI activation, which normally occurs by TRII-mediated phosphorylation in response to TGF-, facilitates sumoylation of the receptor.

Figure 2: The kinase activities of TRI and TRII are required for TRI sumoylation.

(a) Activated TRI is more sumoylated than wild-type TRI. In vitro sumoylation of immunopurified Flag-tagged wild-type and activated (ca) TRI in the presence or absence of recombinant SUMO-1, Aos1/Uba2 (E1), and Ubc9. The reaction mixture was analysed by western blotting for TRI. (b) Effects of the TRI kinase inhibitor and TRI dephosphorylation on TRI sumoylation. In vitro sumoylation was performed as in a with wild-type or activated (ca) TRI, as indicated, in the presence or absence of the TRI kinase inhibitor SB431542. The phosphates were removed from TRI with lambda phosphatase before in vitro sumoylation. (c) The kinase activities of TRII and TRI are required for efficient TRI sumoylation. 293T cells co-expressed a cytoplasmic receptor chimaera TRII-RI, in which the TRI cytoplasmic domain follows the TRII cytoplasmic domain, or chimaeras in which the TRII and/or TRI kinase activities were inactivated by point mutation (KR), with Myc-tagged (M-) SUMO-1 and Ubc9. Sumoylation of the chimaera was analysed by western blotting. (d) In vitro sumoylation of immunopurified cytoplasmic receptor chimaeras or each of the kinase-defective receptor chimaeras, used in c. (e) Diagram showing TGF--induced sumoylation of TRI in the receptor complex.

Full size image (70 KB)

Because the activated TRI has elevated kinase activity and increased phosphorylation⁸, we examined whether increased TRI sumoylation resulted from increased TRI kinase activity or phosphorylation. In vitro sumoylation of wild-type TRI or caTRI was decreased in the presence of SB431542, a specific TRI kinase inhibitor, although this was more easily detected with caTRI (Fig. 2b). These data suggested that the TRI kinase regulates TRI sumoylation. Because TRI did not phosphorylate SUMO-1, the E1 enzyme (Aos/Uba2) and Ubc9 (Supplementary Information, Fig. S1a), these data suggested that TRI autophosphorylation has a role in its sumoylation. To determine whether TRI phosphorylation regulates TRI sumoylation, we removed the Ser/Thr phosphorylation from TRI by using lambda phosphatase before in vitro sumoylation. The absence or decrease of TRI phosphorylation decreased the sumoylation of wild-type or activated TRI (Fig. 2b). This again was more easily detected with activated TRI than with wild-type TRI, as a result of the difference in sumoylation level. These results suggest that increased kinase activity together with increased phosphorylation contribute remarkably to the efficiency of TRI sumoylation.

TGF- binding to TRII results in stable complex formation of two TRII and two TRI receptors, in which TRII phosphorylates the TRI cytoplasmic domain and thereby activates the TRI kinase¹. The activated receptor complex permits autophosphorylation of the TRII and TRI dimers. To determine the roles of the TRII and TRI kinases in TRI sumoylation, we used a cytoplasmic chimaera that fused the TRI cytoplasmic domain to the TRII cytoplasmic domain⁹. In this complex, the TRII kinase activates the TRI kinase without the need to add TGF-. The receptor chimaera, expressed in the presence of SUMO and Ubc9, was sumoylated. Because TRII is not sumoylated (Fig. 1a), the sumoylation site is within the TRI cytoplasmic domain. Inactivation of the TRI kinase by Lys 230 to Arg mutation decreased the chimaera sumoylation (Fig. 2c), which was consistent with the decreased TRI sumoylation in the presence of SB431542 (Fig. 2b). Similar inactivation of the TRII kinase by Lys 277 to Arg mutation also decreased sumoylation of the chimaera in comparison with the wild-type, kinase-active version (Fig. 2c). Because the TRII cytoplasmic domain is not targeted for sumoylation, this result indicates that phosphorylation of the TRI cytoplasmic domain by the TRII kinase is important in the sumoylation of TRI. Mutation of both kinase ATP-binding sites in the chimaera blocked sumoylation.

The requirement for both receptor kinase activities for sumoylation was also studied in vitro. The efficacies of in vitro sumoylation of wild-type, TRI kinase-defective, TRII kinase-defective and TRI/II kinase-defective chimaeras, immunopurified from transfected cells, were compared (Fig. 2d). Inactivation of the kinase functions of TRII or TRI strongly decreased the sumoylation in vitro, whereas inactivation of both kinases abolished sumoylation of the chimaera (Fig. 2d).

These observations indicate that the kinase activities of both TRI and TRII, and the consequent phosphorylation of TRI, are required for efficient TGF--induced sumoylation of TRI in the receptor complex. This is consistent with the TGF--induced phosphorylation and consequent activation of TRI by TRII (Fig. 2e).

The TRI receptor is sumoylated on Lys 389

Sumoylation often occurs on a lysine residue (K) within a consensus sequence Kx(D/E), in which represents a large hydrophobic residue¹⁰. Because this consensus sequence is absent from the amino-acid sequence of TRI, each of the 20 lysine residues in the cytoplasmic domain was singly replaced by arginine, and the effect of each mutation on TRI sumoylation was tested. Lysates of cells co-expressing each Flag-tagged lysine mutant of TRI with SUMO-1 and Ubc9 were subjected to immunoprecipitation with anti-Flag antibody, followed by western blotting. TRI was not sumoylated when Lys 389 was replaced by arginine, whereas replacements of other lysine residues by arginine did not affect the sumoylation of TRI (Fig. 3a, and data not shown), indicating that Lys 389 is a major site for TRI sumoylation. The Lys 389 mutation also affected in vitro sumoylation, because no sumoylated TRI was detected when Lys 389 was replaced by arginine (Fig. 3b). These results indicate that Lys 389 is the only residue targeted for sumoylation.

Figure 3: The TRI receptor is sumoylated on Lys 389.

(a) Mutation of Lys 389 abolishes TRI sumoylation. 293T cells expressed Flag-tagged wild-type (WT) or mutant TRI with the indicated lysine-to-arginine mutation, with Myc-tagged SUMO-1 and Ubc9. Cell lysates were subjected to immunoprecipitation with anti-Flag antibody, followed by western blotting for SUMO-1. (b) In vitro sumoylation of wild-type TRI or K389R TRI. Immunopurified TRI was subjected to in vitro sumoylation followed by immunoblotting to detect sumoylation. (c) Proposed structure of the TRI cytoplasmic domain. The Lys 389, L45 loop, GS region and ATP-binding site are indicated. N and C indicate the N and C termini. (d) Sequence alignment of the TRI sequence containing Lys 389 and corresponding regions of other type I receptors.

Full size image (85 KB)

The proposed structure of the TRI cytoplasmic domain (Fig. 3c) predicts that Lys 389 is located in the hinge between the EF helix and the F helix, and is exposed at the surface of the C lobe of the kinase domain. The sumoylation site faces the same orientation as the GS region, which is phosphorylated by TRII on the binding of TGF-, and the L45 loop, which specifies the interaction with Smad, although in a separate protein domain. The GS region and L45 loop, both located in the N lobe, interact with the Smad for phosphorylation by TRI. The exposure of Lys 389 at the protein surface predicts that conjugation with SUMO strongly affects the cytosolic surface of TRI and may regulate the binding of Smad to the L45 loop and GS domain of TRI, and interactions of additional proteins with the receptor complex.

Sequence comparisons (Fig. 3d) show that Lys 389 is not conserved in other TGF--family type I receptors, with the exception of the activin receptor ActRIB/ALK-4. This is consistent with the absence of sumoylation of these type I receptors in vivo (Fig. 1e). The lack of ActRIB/ALK-4 sumoylation suggests that other determinants besides the target lysine are needed for sumoylation of TRI. It is unlikely that this is due to the serine versus threonine difference, four residues preceding the sumoylated lysine residue in TRI compared with ActRIB/ALK-4 (Fig. 3d), because S385T replacement did not affect the in vitro sumoylation of TRI (Supplementary Information, Fig. S1b).

TRI sumoylation regulates Smad interaction and activation

To evaluate whether sumoylation of the exposed Lys 389 affects Smad activation, we examined the interaction of Smad3 with caTRI. Because this interaction is hard to detect by immunoprecipitation, probably as a result of its low affinity and transient nature, we examined the interaction of the Smad3D407E mutant with caTRI. The D407E mutation in the MH2 domain was identified in Smad2 in colorectal carcinoma, and affects the interaction of Smad with TRI and heteromerization with Smad4 (ref. 11). Increased caTRI sumoylation by co-expressing Ubc9 and SUMO-1 enhanced the interaction of TRI with Smad3D407E. In contrast, coexpression of SUMO and Ubc9 did not enhance the interaction of Smad3D407E with caTRI carrying the sumoylation-resistant K389R mutation (Fig. 4a). We also incubated immobilized glutathione S-transferase (GST)–Smad3D407E with a mixture of sumoylated and unsumoylated TRI. Western blotting of purified GST–Smad3–TRI complexes showed preferential binding of Smad3 to sumoylated TRI, in comparison with unsumoylated TRI, even though the latter was in large excess (Fig. 4b). This result, together with the data in Fig. 4a, indicates that sumoylation of TRI enhances Smad3 recruitment and suggests that TRI sumoylation enhances Smad activation.

Figure 4: TRI sumoylation regulates Smad activation and TGF- responses.

(a) Interaction of Smad3 with TRI. Activated (ca) TRI or its K389R mutant were coexpressed with Myc-tagged (M-) SUMO-1 and Ubc9, and/or Smad3D407E in 293T cells. The lysates were subjected to immunoprecipitation, and analysed by western blotting. (b) In vitro interaction of Smad3 with TRI. Immobilized GST or GST–Smad3(D407E) were incubated with in vitro sumoylated and non-sumoylated Flag-tagged TRI. Adsorbed proteins were subjected to western blotting for TRI. The lower panel shows Coomassie blue staining of GST and GST–Smad3(D407E) used for the adsorption. (c) The indicated fibroblasts were subjected to western blotting to assess the expression of TRI. (d) Biotin-labelled cell-surface proteins from the indicated MEFs were subjected to avidin precipitation, followed by western blotting to assess the cell-surface expression levels of TRI. (e) Wild-type and K389R TRI have similar kinase activities. TRI, immunopurified from transfected 293T cells, was subjected to kinase reactions without or with TRI kinase inhibitor. (f, g) Lack of TRI sumoylation confers a lower level of Smad3 (f) or Smad2 (g) activation. Tgfbr1^{- /-} fibroblasts stably expressing wild-type or K389R TRI were treated without or with TGF- for the indicated time. Cell lysates were analysed by western blotting. (h) Lack of TRI sumoylation decreases Smad3-mediated transcription. Tgfbr1^{- /-} fibroblasts stably expressing wild-type TRI or K389R mutant TRI were transfected with the Smad3-responsive (CAGA)₁₂-luciferase reporter. Luciferase activities were measured. The error bars represent s.d. (n = 2) (i) Lack of TRI sumoylation decreases TGF--induced endogenous gene expression. Tgfbr1^{- /-} fibroblasts stably expressing wild-type TRI or K389R mutant TRI were treated without or with TGF-. Smad7 mRNA was quantified with real-time PCR. The error bars represent s.d. (n = 3) (j) Lack of TRI sumoylation decreases TGF--induced growth inhibition. Tgfbr1^{- /-} fibroblasts stably expressing wild-type or K389R TRI, or with an empty vector, were cultured without or with the indicated dose of TGF- for 3 days. The cells were then counted. The error bars represent s.d. (n = 3). Uncropped images of blots in a and f are shown in Supplementary Information, Fig. S4.

Full size image (102 KB)

Because TRI phosphorylates Smad2 and Smad3 after the binding of TGF-, we investigated whether sumoylation of activated TRI affects Smad3 phosphorylation. TRI-defective mouse embryonic fibroblasts (MEFs) derived from Tgfbr1^{- /-} mice¹² were retrovirally infected to express wild-type TRI or sumoylation-resistant K389R TRI. Stably selected cell populations, expressing either TRI form at equal levels (Fig. 4c), showed equivalent cell-surface levels of wild-type or mutant TRI (Fig. 4d), suggesting that the K389R mutation did not affect cell-surface transport or stability of TRI. K389R TRI also showed a similar phosphorylation level to that of wild-type TRI, resulting primarily from the TRI kinase activity, because SB431542 abolished this phosphorylation (Fig. 4e). Fractionation of cell lysates did not reveal differences in subcellular compartmentalization of wild-type versus mutant TRI (Supplementary Information, Fig. S1c). We then compared wild-type and K389R TRI for their ability to phosphorylate Smad3 in response to TGF-. In fibroblasts expressing wild-type TRI, TGF- induced Smad3 phosphorylation within 15 min, whereas, in cells expressing K389R TRI, the Smad3 phosphorylation kinetics in response to TGF- was slower, with first detection at 30 min. Furthermore, the overall level of Smad3 activation was lower in cells expressing K389R TRI than in cells expressing wild-type TRI (Fig. 4f). Similar results were seen with TGF--induced activation of Smad2 (Fig. 4g). Replacement of Lys 393 in ActRIB/ALK-4, which is not sumoylated and corresponds to Lys 389 in TRI, with arginine did not affect Smad3 activation (Supplementary Information, Fig. S1d).

The differences in level and kinetics of Smad2 and Smad3 phosphorylation by wild-type versus K389R TRI, together with the results of the interaction of Smad3 with TRI, suggest that TRI sumoylation enhances the Smad interaction with TRI, allowing more efficient phosphorylation and activation of Smad2/3 in response to TGF-.

TRI sumoylation regulates functional responses to TGF-

Using Tgfbr1^{- /-} fibroblasts ectopically expressing wild-type or K389R TRI, we characterized the effect of TRI sumoylation on Smad-mediated transcription; that is, the functional consequence of Smad activation. We used a reporter in which tandem Smad-binding sites control luciferase transcription. Cells expressing K389R TRI showed decreased transcription from the Smad3-responsive promoter compared with cells expressing wild-type TRI (Fig. 4h). We also compared the endogenous expression of the TGF--responsive Smad7 gene by reverse transcriptase polymerase chain reaction (RT–PCR). Cells expressing K389R TRI showed decreased expression of Smad7 messenger RNA in response to TGF-, compared with cells expressing wild-type TRI (Fig. 4i). Similar results were obtained with two additional populations of Tgfbr1^{- /-} fibroblasts ectopically expressing TRI or K389R TRI at similar levels (Supplementary Information, Fig. S2). These results suggest that TRI sumoylation defines the TGF--induced transcriptional regulation.

We also examined the contribution of TRI sumoylation to the antiproliferative response to TGF-. We seeded the fibroblasts expressing wild-type or K389R TRI in parallel with the parental Tgfbr1^{- /-} cells as control cells, and determined the proliferative response after adding TGF-. Cells lacking TRI were not affected in their proliferation by TGF-, whereas those expressing wild-type TRI responded with decreased proliferation, as assessed by cell number (Fig. 4j). In contrast with wild-type TRI, cells expressing K389R TRI showed a decreased growth inhibitory response to TGF-. This result suggests that sumoylation regulates the TRI-mediated antiproliferative response to TGF- and renders the cells more responsive to TGF-.

TRI sumoylation enhances invasion and metastasis of Ras-transformed cells

Because autocrine TGF- signalling regulates cancer progression^{13, 14} we postulated that resistance to sumoylation, while suppressing TGF- growth inhibitory activities, affects tumour progression. To address this issue, the Tgfbr1^{- /-} fibroblasts, carrying a control empty vector or ectopically expressing wild-type or K389R TRI at similar levels, were transduced with a control vector or a vector expressing activated Ha-Ras (Leu 61) to generate tumorigenic cell populations. Mutant Ras was expressed and activated extracellular signal-regulated kinase (ERK) MAP kinase to similar extents in all three Ras-transformed cell populations (Fig. 5a). Cells expressing activated Ras had a transformed phenotype, which was apparent from the altered cell morphology and loss of contact inhibition (data not shown).

Figure 5: Lack of TRI sumoylation decreases TGF--regulated invasion and metastasis.

(a) Ras-transformed Tgfbr1^{- /-} fibroblasts stably expressing wild-type or K389R TRI, or with an empty vector, were subjected to western blotting for TRI, Ras or phospho-ERK1/2 (pERK1/2) as a marker of Ras activation. (b, c) TRI-mediated TGF- responsiveness of Ras-transformed cells promotes invasion, which is decreased by lack of TRI sumoylation. Cells were seeded on a Matrigel-coated Transwell filter and incubated for 24 h to allow invasion towards 10% serum. Cells that migrated through the filter were stained with crystal violet. The white stipples indicate the pores in the filter. A representative picture and quantification of invaded cells are shown in b and c, respectively. Error bars represent s.d. (n = 4) (d, e) Ras-transformed Tgfbr1^{- /-} fibroblasts expressing wild-type or K389R TRI, or with an empty vector, were injected into the tail vein of nude mice. The lung tumour nodules were counted after 3 weeks. (d) Representative pictures of lungs from mice with Ras-transformed MEFs are shown in the upper panels, and corresponding haematoxylin/eosin-stained sections of tumour nodules at the same magnification (10 objective) are shown in the lower panels. (e) Quantification of tumour nodules in the lungs. Error bars represent s.e.m. (n = 6). The single and double asterisks indicate P < 0.05 and P < 0.01, respectively, compared with wild-type TRI.

Full size image (88 KB)

We examined the invasion of the Ras-transformed cells by using a modified Boyden chamber assay in which cells migrate through Matrigel towards serum. Ras-transformed Tgfbr1^{- /-} cells ectopically expressing wild-type TRI showed a higher invasion activity than Ras-transformed Tgfbr1^{- /-} cells lacking TRI (Fig. 5b), indicating that the invasive capacity of Ras-transformed MEFs depends on TRI signalling. Ras-transformed Tgfbr1^{- /-} MEFs expressing K389R TRI were less invasive than cells expressing wild-type TRI, indicating that, in this system, lack of TRI sumoylation impairs the TRI-dependent invasion of transformed cells (Fig. 5b, c). These observations are consistent with the role of sumoylation of TRI in TGF--induced gene expression and growth inhibition (Fig. 4h–j).

Metastasis is a complex process, requiring cell growth, migration, invasion, intravasation and extravasation, and cell survival in the circulatory system and at the metastatic site. Using a mouse tail-vein injection model, autocrine TGF- signalling was shown to enhance the ability of tumour cells to establish metastatic nodules within the lung^{15, 16}. To determine the roles of TRI and sumoylation of TRI in the formation of metastatic nodules in this model, we compared the ability of the Tgfbr1^{- /-} MEF derivatives to colonize the lung. Colonization of the lungs by MEFs was fully dependent on expression of activated Ha-Ras (Fig. 5d, e; data not shown). Ras-transformed Tgfbr1^{- /-} cells gave rise to only few very small lung tumour nodules (Fig. 5d, e). Tumour cells were proliferative with a high incidence of apoptosis (Supplementary Information, Fig. S3h). The tumours were morphologically heterogeneous, and large cells with massive nuclei were indicative of chromosomal instability (Supplementary Information, Fig. S3e). Remarkably, Ras-transformed Tgfbr1^{- /-} cells ectopically expressing wild-type or K389R TRI developed numerous large metastatic nodules (Fig. 5d, e). MEFs expressing sumoylation-defective TRI gave rise to fewer tumour nodules than cells expressing the wild-type receptor (Fig. 5e). These nodules were generally smaller than those from wild-type MEFs expressing TRI (Fig. 5d, and data not shown), although their histological appearance (Fig. 5d) and proliferative rates (Supplementary Information, Fig. S3a) were similar. These results suggest that, in this model of TGF--mediated metastasis, TRI sumoylation contributes to tumour progression by enhancing tumour cell extravasation, survival and/or growth at the metastatic site.

The Ser385Tyr mutation of TRI, implicated in metastatic cancer, confers sumoylation resistance

Mutations in TGF- signalling mediators, including TGFBR1, have been associated with human cancers¹³. Among these, a missense mutation, S387Y, in TGFBR1 was enriched in breast and head-and-neck cancer metastases, in comparison with corresponding primary tumours^{17, 18}. This mutation confers diminished TGF- signalling and is the only mutation in TGFBR1 or TGFBR2 known to associate specifically with tumour metastases¹⁷. Because the corresponding residue in rat TRI, namely Ser 385, localizes close to the Lys 389 sumoylation site of TRI (Fig. 6a), we examined whether rat S385Y TRI was sumoylated. In contrast with wild-type TRI, this mutant was not sumoylated in cells overexpressing SUMO-1 and Ubc9 (Fig. 6b). In vitro, replacement of Ser 385 with Tyr (Fig. 6c) or Ala (Fig. 6d) also strongly decreased sumoylation. In contrast, replacement of Ser 385 with Thr, which may, similarly to Ser, be targeted for phosphorylation by Ser/Thr kinases, did not affect sumoylation (Supplementary Information, Fig. S1b). These results indicate that this single amino-acid substitution prevents TRI sumoylation.

Figure 6: The Ser385Tyr mutation impairs TRI sumoylation and function.

(a, b) Ser385Tyr TRI is not sumoylated. (a) The rat TRI sequence is shown with Lys 389 as sumoylation site four amino-acid residues away from Ser 385, which is equivalent to Ser 387 in human TRI. (b) 293T cells co-expressing wild-type, K389R or S385Y TRI, with Myc-tagged (M-) SUMO-1 and Ubc9, were lysed and analysed by western blotting to detect sumoylation. (c, d) In vitro sumoylation of S385Y mutant TRI (c) or S385A TRI (d) in comparison with wild-type TRI. Immunopurified TRI was subjected to in vitro sumoylation followed by immunoblotting to detect sumoylation. (e) Tgfbr1^{- /-} fibroblasts stably expressing wild-type, K389R or S385Y TRI were transfected with the Smad3-responsive (CAGA)₁₂-luciferase reporter. Luciferase activities without or in response to added TGF- were measured. Error bars represent s.d. (n = 3). (f) Representative pictures of lungs from mice with Ras-transformed MEFs are shown in upper panels, and corresponding haematoxylin/eosin-stained sections of tumour nodules at the same magnification (10 objective) are shown in the lower panels. (g) Quantification of tumour nodules in the lungs. Error bars represent s.e.m. (n = 6). Double asterisk indicates P < 0.01 compared with wild-type TRI.

Full size image (82 KB)

We examined whether the S385Y mutation decreased the responsiveness to TGF-, as observed with the sumoylation-defective K389R TRI (Fig. 4). Using the Smad3-responsive reporter of Fig. 4h, cells expressing S385Y TRI responded to TGF- with a lower level of transcription than cells expressing wild-type TRI, but a higher level than cells expressing K389R TRI (Fig. 6e), indicating that the decreased TGF- responsiveness associated with S385Y TRI correlates with impaired TRI sumoylation. We also evaluated the S385Y TRI mutation in the lung colonization tumour model. Ras-transformed MEFs expressing S385Y TRI formed fewer and smaller metastatic nodules than cells expressing wild-type TRI (Fig. 6f, g). Although this difference in tumour nodule formation is qualitatively comparable to that of K389R TRI cells, the efficiency of metastatic nodule formation was more impaired in S385Y TRI cells than in K389R TRI cells. This quantitative difference between cells expressing K389R TRI or S385Y TRI, when compared with their differential activities in transcription assays, raises the possibility that the impaired TRI sumoylation resulting from the S385Y mutation may be complemented with an additional defect of relevance to cancer progression.

Plasmids.

The expression plasmids for N-terminally Flag-tagged or Myc-tagged SUMO-1, or Ubc9 or C-terminally Flag-tagged TRI have been described^{33, 34}. The expression plasmid for the Flag-tagged TRII–TRI cytoplasmic chimaera has also been described⁹. Site-directed mutagenesis with the QuikChange kit (Stratagene, La Jolla, CA) generated plasmids encoding C-terminally Flag-tagged TRI or TRII–TRI cytoplasmic chimaeras with amino-acid substitutions, and mutations were confirmed by DNA sequencing. To generate retroviral vectors expressing Flag-tagged TRI, the coding region was inserted into the HpaI site of LNCX³⁵. The retroviral vector pBABE-H-Ras(Leu 61)-IRES-Puro³⁶ was provided by R. Davis (University of Massachusetts Medical School, Worcester, MA). The expression plamid for Myc-tagged Smad3(D407E)¹¹ was provided by K. Miyazono (University of Tokyo, Tokyo, Japan). Expression plasmids for haemagglutinin (HA)-tagged ALK-1, ALK-3 or ALK-6 were generated by subcloning the coding regions from pcDNA3-HA-ALK1, pcDNA3-HA-ALK3 and pcDNA3-HA-ALK6 (refs 37, 38), provided by P. ten Dijke (University of Leiden, Leiden, The Netherlands), into the EcoRI/XhoI site of pRK5. A plasmid encoding Flag-tagged ALK-2 was made by inserting the coding region from pRK5-Myc-ALK2 into the EcoRI/SalI site of pXFIF³⁹. To generate a plasmid encoding Flag-tagged ALK-4, the coding region from pcDNA1-hALK4 (ref. 37), provided by C. H. Heldin (Ludwig Institute for Cancer Research, Uppsala, Sweden) was amplified by PCR and inserted into the EcoRI/SalI site of pRK5 (ref. 40). The reporter plasmid (CAGA)₁₂-luciferase⁴¹ was also provided by P. ten Dijke.

Cell culture and transfections.

Tgfbr1^{- /-} MEFs¹² were provided by S. Karlsson (Lund University Hospital, Lund, Sweden), immortalized and cultured in DMEM medium with 10% fetal bovine serum (FBS). Cells were plated at 2 10⁵ cells per well in six-well plates and transfected with reporter and -galactosidase plasmids using Lipofectamine Plus (Invitrogen). One day after transfection, cells were transferred to medium containing 0.2% FBS with or without TGF- (1–5 ng ml^{- 1}) for 16 h. Cell extracts were prepared and assayed for luciferase activity as described⁴². Luciferase activities were normalized to -galactosidase activity from a co-transfected -galactosidase plasmid.

Immunoprecipitations and immunoblotting.

COS or 293T cells were harvested 48 h after transfection and lysed by brief sonication in lysis buffer (25 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% TritonX-100, 10% glycerol, and protease inhibitor cocktail). Lysates were subjected to immunoprecipitation with anti-Flag M2 agarose (Sigma-Aldrich). Immune complexes were washed three times with lysis buffer and subjected to western blotting with anti-Flag or anti-Myc antibodies. To detect sumoylation of endogenous TRI, Mv1Lu or MDA-231 cells were cultured for 3 h in DMEM with 0.2% FBS with or without TGF- (10 ng ml^{- 1}). Cells were washed twice with PBS, harvested and lysed by sonication in lysis buffer containing 10 mM N-ethylmaleimide. The lysates were precleared with mouse IgG and incubated with rabbit anti-TRI antibody (Santa Cruz Biotech) or IgG. Immune complexes were precipitated with protein G beads, and immunoblotted with mouse anti-SUMO-1 (Cell Signaling Technologies) or rabbit anti-TRI antibodies (Santa Cruz Biotech).

In vitro sumoylation.

293T cells were transfected with plasmids expressing Flag-tagged proteins. Lysates were subjected to immunoprecipitation with anti-Flag M2 agarose, and immune complexes were eluted with Flag peptide (Sigma-Aldrich). The immunopurified proteins were incubated with 2 g of recombinant SUMO-1 (BIOMOL International, Plymouth Meeting, PA), 0.5 g of Aos1/Uba2 (BIOMOL International), and 0.1 g of Ubc9 (BIOMOL International) for 2 h in 20 l of 50 mM Tris-HCl pH 7.5, 5 mM MgCl2, 2 mM ATP. The reaction mixture was analysed by western blotting with anti-Flag antibody (Sigma-Aldrich). The TRI kinase was inhibited by adding 5 M SB431542 (Sigma-Aldrich). TRI was dephosphorylated by treatment with lambda protein phosphatase (New England Biolabs) for 30 min.

GST adsorption assays.

Immunopurified Flag-tagged TRI was subjected to in vitro sumoylation. The mixture of sumoylated and unsumoylated TRI was incubated with immobilized GST–Smad3(D407E) in 50 mM Tris-HCl pH 7.5, 120 mM NaCl, 0.1% Nonidet P40, 10% glycerol and protease inhibitor cocktail. After pull-down, precipitates were subjected to SDS-PAGE followed by western blotting with anti-Flag antibody to detect TRI.

Generation of stable cell lines.

Tgfbr1^{- /-} cells were infected with the LNCX-based retroviral vector expressing Flag-tagged wild-type or K389R TRI, and stably infected cell populations were generated as described⁴³. The expression levels of TRI were assessed by western blotting with anti-Flag antibody.

To generate Ras-transformed cells, Tgfbr1^{- /-} cell populations with an empty vector or expressing Flag-tagged wild-type or K389R TRI were transduced with the retroviral vector pBABE-H-Ras(Leu 61)-IRES-Puro or control vector pBABE-Puro, and selected with 2 g ml^{- 1} puromycin. The expression levels of H-Ras(Leu 61) were examined by western blotting with anti-c-H-Ras antibody (Calbiochem, San Diego, CA), and phosphorylated ERK1/2 was detected by western blotting with anti-phospho-ERK1/2 antibody (Cell Signaling Technologies).

In vitro kinase assays.

Immunopurified receptors were incubated at 30 °C for 30 min in 10 mM HEPES-KOH pH 7.5, 5 mM MgCl2 and 5 mM CaCl2 with or without 5 M TRI kinase inhibitor SB431542. The reaction mixture was subjected to SDS-PAGE, followed by autoradiography.

Cell-surface TGF- receptor biotinylation and precipitation.

Tgfbr1^{- /-} cells expressing wild-type or K389R TRI, or transfected with an empty vector, were grown to confluence and labelled with sulpho-N-hydroxysuccinimido biotin (Pierce) at 4 °C for 2 h. Cells were washed with 100 mM glyine and lysed by brief sonication in 25 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton cX-100, 10% glycerol, and protease inhibitor cocktail. Biotinylated cell-surface proteins were precipitated with avidin-immobilized beads (Pierce) and subjected to western blotting with anti-Flag antibody to detect cell-surface TRI.

TGF-response analyses.

To evaluate Smad3 phosphorylation, Tgfbr1^{- /-} cells expressing wild-type or K389R TRI were treated without or with 2.5 ng ml^{- 1} TGF- for 10, 30, 60 or 120 min, and cell lysates were analysed by western blotting with anti-phospho-Smad3 (Cell Signaling Technologies) or anti-Smad3 antibody (Invitrogen).

To perform the cell growth inhibition assay, 5 10⁴Tgfbr1^{- /-} cells expressing wild-type or K389R TRI or having an empty vector were grown for 3 days in DMEM with 10% FBS without or with TGF-, and cell numbers were counted with a haemocytometer.

To quantify Smad7 mRNA expression, Tgfbr1^{- /-} cells expressing Flag-tagged wild-type or K389R TRI were treated with or without TGF- (2.5 ng ml^{- 1}) for 4 h. RNA was isolated with RNeasy kit (Qiagen) and used as a template for reverse transcriptase. The Smad7 mRNA was quantified by real-time PCR with cyber-green (Invitrogen), and normalized against RPL19 mRNA. The primer sequences were as follows: Smad7, 5'-TCTGGACAGTCTGCAGTTGG-3' (forward) and 5'-TCCTGCTGTGCAAAGTGTTC-3' (reverse); RPL19, 5'-GGAAGAGGAAGGGTACTGCC-3' (forward) and 5'-GGATTCCCGGTATCTCCTGAG-3' (reverse).

In vitro invasion assay.

In vitro invasion assays were performed with Biocoat Matrigel Invasion chambers (BD Biosciences). Cells (2.5 10⁴) were seeded into the upper insert of the chamber and incubated for 24 h, allowing invasion through Matrigel towards 10% serum. The invaded cells were fixed with 96% ethanol and stained with 0.05% crystal violet.

Tumour formation.

To perform lung tumour formation assays, 5 10⁵ cells were injected into the tail vein of eight-week-old nude mice³⁶. Three weeks after injection, the mice were labelled by intraperitoneal injection of 100 mg kg^{- 1} 5-bromo-2-deoxyuridine (BrdU). At 1 h after injection of BrdU, the mice were killed, lung tumour nodules were counted, and lungs were fixed in 4% paraformaldehyde and processed for histological analysis of paraffin-embedded tissue sections.

Immunostaining for BrdU was performed with biotin-conjugated mouse anti-BrdU antibody (Alexis Biochemicals); detection was with the Vectastain Elite ABC kit detection system (Vector Laboratories).

Note: Supplementary Information is available on the Nature Cell Biology website.

Author contributions

J.S.K., R.J.A. and R.D. conceived and designed the studies; J.S.K. and E.F.S. performed the experiments; J.S.K., E.F.S., R.J.A. and R.D. prepared the manuscript.