Differential Ubiquitination Defines the Functional Status of the Tumor Suppressor Smad4*

Smad4 is an essential signal transducer of all transforming growth factor-β (TGF-β) superfamily pathways that regulate cell growth and differentiation, and it becomes inactivated in human cancers. Receptor-activated (R-) Smads can be poly-ubiquitinated in the cytoplasm or the nucleus, and this regulates their steady state levels or shutdown of the signaling pathway. Oncogenic mutations in Smad4 and other Smads have been linked to protein destabilization and proteasomal degradation. We analyzed a panel of missense mutants derived from human cancers that map in the N-terminal Mad homology (MH) 1 domain of Smad4 and result in protein instability. We demonstrate that all mutants exhibit enhanced poly-ubiquitination and proteasomal degradation. In contrast, wild type Smad4 is a relatively stable protein that undergoes mono- or oligo-ubiquitination, a modification not linked to protein degradation. Analysis of Smad4 deletion mutants indicated efficient mono- or oligo-ubiquitination of the C-terminal MH2 domain. Mass spectrometric analysis of mono-ubiquitinated Smad4 MH2 domain identified lysine 507 as a major target for ubiquitination. Lysine 507 resides in the conserved L3 loop of Smad4 and participates in R-Smad C-terminal phosphoserine recognition. Mono- or oligo-ubiquitinated Smad4 exhibited enhanced ability to oligomerize with R-Smads, whereas mutagenesis of lysine 507 led to inefficient Smad4/R-Smad hetero-oligomerization and defective transcriptional activity. Finally, overexpression of a mutant ubiquitin that only leads to mono-ubiquitination of Smad4 enhanced Smad transcriptional activity. These data suggest that oligo-ubiquitination positively regulates Smad4 function, whereas poly-ubiquitination primarily occurs in unstable cancer mutants and leads to protein degradation.

Smads are the primary intracellular signal transducers for TGF-␤ 1 superfamily proteins (1,2). Receptor-regulated Smads (R-Smads) are phosphorylated by the type I receptor serine/ threonine kinase of the heteromeric (type I/type II) cell surface receptor complex. Activated R-Smads oligomerize with a common effector Smad4 (also called Co-Smad), and translocate to the nucleus. After nuclear accumulation, Smads engage in transcriptional complexes by binding to DNA and by interacting with various transcription factors to regulate gene expression (1,3,4). Another class of Smad proteins, the inhibitory (I) Smads, participate in negative feedback loops whereby their genes are direct targets of the Smad pathway and their induced expression leads to inhibition of R-Smad phosphorylation by the type I receptors (5). In addition, multiple intracellular effectors that mediate signals from alternative signaling cascades can regulate the Smad pathway (1,6).
Proteasomal degradation of ubiquitinated R-Smads has been reported to be important for the maintenance of their steady state levels and for the shutdown of the activated Smad pathway after execution of its transcriptional roles (reviewed in Ref. 2). Furthermore, I-Smads act as adaptor proteins that mediate ubiquitination and degradation of the type I receptors (7,8). Much of the evidence on the role of ubiquitin-mediated degradation of Smad proteins depends on the identification of two related Hect domain E3 ligases, named Smurf1 and Smurf2, that interact specifically with R-Smads and I-Smads but not with Smad4 and catalyze their specific poly-ubiquitination (7)(8)(9)(10)(11). However, the precise physiological role of such ubiquitination events remains unclear and controversial to a certain extent (2). Whether ubiquitination of R-Smads is a necessary event to regulate the signaling capacity of the pathway or whether it could serve some alternative functions remains an open question.
Ubiquitination of the Co-Smad, Smad4, has been reported to occur only in cancer cells and is thought to be associated with oncogenic mutations that lead to Smad4 inactivation (1). Thus, point mutants isolated from human cancer cells and mapping in the N-terminal MH1 domain were shown to be unstable compared with wild type Smad4 (12,13). One such Smad4 mutant was analyzed in more detail, as was the homologous mutant of the R-Smad Smad2; both were found to be proteasomally degraded after poly-ubiquitination (13). Furthermore, a C-terminally truncated mutant of Smad4 also exhibits protein instability and ubiquitin-mediated proteasomal degradation (14). This mutant was demonstrated to be able to affect the stability of Smad2 with which Smad4 normally forms functional signaling complexes. These studies suggested that degradation of Smad4 is a prevalent mechanism of inactivation of TGF-␤ pathways in cancer cells. Finally, two other factors, oncogenic Ras overexpression and one subunit of the COP9 signalosome, CSN5/Jab1, could both mediate Smad4 ubiquitination and proteasomal degradation (15,16). Thus, although ubiquitin-mediated degradation of Smad4 has been established as a mechanism of inactivation of this tumor suppressor in human cancers, the importance of Smad4 ubiquitination under non-tumorigenic conditions has not yet been addressed. Many of the mechanistic details of Smad ubiquitination in general also remain largely unexplored.
We therefore set up to investigate some of the aspects of Smad4 ubiquitination by focusing on a comparative analysis of wild type Smad4 versus a panel of missense MH1 domain mutants, the functional properties of which we reported previously (12). We also attempted to map lysine residues in Smad4 that are prominent sites of ubiquitination. We show that all MH1 domain mutants are poly-ubiquitinated and degraded by proteasomes, thus extending the paradigm reviewed above to a larger group of cancer mutations. Furthermore, we demonstrate that wild type Smad4 is preferentially mono-or oligoubiquitinated, which is in agreement with its apparent stability. This difference is also clear when a panel of dominant negative E2 ubiquitin-conjugating enzymes was studied for their effects on wild type and mutant Smad4 stability. Mono-or oligo-ubiquitination of wild type Smad4 maps to and/or requires the MH2 domain. We have identified lysine 507 as one prevalent site of ubiquitination in the MH2 domain. This residue affects R-Smad/Co-Smad oligomerization and transcriptional activity, and based on these findings we postulate that ubiquitination of Smad4 may serve two alternative purposes, i.e. positive regulation of Smad oligomerization and signaling strength, and protein degradation of mutants associated with human carcinogenesis.

EXPERIMENTAL PROCEDURES
Cell Culture, Ligands, Proteins, Antibodies, and Plasmids-Human embryonic kidney (HEK) 293T, mammary carcinoma MCF7 and MDA-MB-468, hepatoma HepG2, and colorectal carcinoma SW480 cells were obtained and cultured according to protocols from American Type Culture Collection. The human keratinocyte cell line HaCaT was a gift from N. Fusenig (German Cancer Research Center, Heidelberg, Germany).
Mouse monoclonal anti-FLAG (M2 and M5) antibodies were purchased from Sigma, mouse monoclonal anti-HA was from Roche Molecular Biochemicals, mouse monoclonal anti-Smad4 was from BD Transduction Laboratories, and rabbit polyclonal anti-Smad4 (DPC4) and mouse monoclonal anti-myc (9E10) antibodies were produced in house.
The promoter-reporter constructs 9xCAGA-luc 12xCAGA-luc (17) were gifts from J.-M. Gauthier (Laboratoire Glaxo Wellcome, Paris, France). The mammalian expression vectors pcDNA3 encoding constitutively active ALK5, 6myc-tagged Smad3 and Smad2, FLAG-tagged Smad4, Smad4⌬MH1, and Smad4⌬MH2, have been previously described (18 -20). The FLAG-tagged Smad4 linker (amino acids 142-321) construct was produced by PCR amplification from the full-length human Smad4 cDNA and subcloning to pcDNA3. The point mutants K507R and K507A of human Smad4, epitope-tagged with FLAG at their N termini, were constructed using site-directed mutagenesis with the QuikChange kit (Stratagene) according to the protocol from the manufacturer. The FLAG-tagged, human dominant-negative E2 ubiquitinconjugating enzymes, UbcH(CS)s, were constructed by replacement of the active site cysteine with serine by a polymerase chain reaction (PCR)-based approach. The K48R mutant ubiquitin was also constructed by a PCR-based approach. All mutations and sequences of the mutants were confirmed by DNA sequencing. The missense Smad4 mutants have been described previously (12). HA-tagged ubiquitin was a gift from D. Bohmann (European Molecular Biology Laboratory, Heidelberg, Germany).
Cell Transfections and Gene Reporter Assays-Transient transfections of cells using FuGENE6 (Roche Molecular Biochemicals) or cal-cium phosphate coprecipitation were performed according to protocol from the manufacturer and as previously described (12). Luciferase reporter assays were performed using the enhanced luciferase assay kit from BD Pharmingen Inc., according to the protocol from the manufacturer. All promoter-reporter experiments included triplicate determinations from independently transfected plates within each experiment and were repeated at least twice. The data presented in figures represent average values and standard deviations from representative single experiments with triplicate transfections.
Analysis of Smad4 Protein Expression and Stability-Transiently transfected or non-transfected cell monolayers were lysed in 0.5% Triton X-100, 0.5% deoxycholate, 20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10 mM EDTA, 1% aprotinin, 1 mM phenylmethylsulfonyl fluoride. After Bradford analysis of total protein levels, the lysates were adjusted prior to subsequent immunoblotting or immunoprecipitation. Endogenous Smad4 was either directly immunoblotted using the mouse anti-Smad4 antibody (Fig. 1A) or immunoprecipitated with rabbit DPC4 antibody before immunoblotting with mouse anti-Smad4 antibody (Fig. 1B). Transfected FLAG-tagged Smad4 was detected by direct immunoblotting of cell lysates with anti-FLAG M2 antibody. Immunoblotted proteins were detected by ECL, scanning on a CCD camera (Fuji LAS-1000), and densitometry was performed using the AIDA program of the scanner (Fuji).
For Smad4 stability measurements, transfected HEK-293T cells were treated with cycloheximide (50 g/ml, Sigma) for the last 2-12 h of the transfection and prior to cell lysis and anti-FLAG immunoblotting. Alternatively, pulse-chase experiments were performed in transfected HEK-293T cells by starving cells in methionine/cysteine-free MCDB 104 medium for 4 h, labeling cells with 50 Ci/ml [ 35 S]methionine/cysteine (Amersham Biosciences) for 2 h, and chasing in MCDB 104 supplemented with cold 50 g/ml methionine and cysteine. Cells were lysed and immunoprecipitations with FLAG M2 antibody were followed, and after washing the immunocomplexes with lysis buffer four times, the proteins were resolved by SDS-PAGE, the gels were dried and autoradiographed on a phosphorimager (Fujix BAS 2000, Fuji). Relative protein expression levels were quantified using the scanning densitometric software (AIDA) of the phosphorimager.
For Smad4 analysis after proteasomal inhibition, transfected HEK-293T cells were treated with 20 M MG-132 inhibitor (Calbiochem) for 16 h. A second pulse of inhibitor was added for additional 2 h, and then cells were lysed. All control cell treatments included equal volumes of dimethyl sulfoxide, the solvent of MG-132.
Ubiquitination Assays-HEK-293T cells were transfected with FLAG-Smad4 and HA-ubiquitin or HA-ubiquitin K48R; 48 h after transfection cells were lysed by boiling in 200 l of phosphate-buffered saline, pH 7.4, containing 1% SDS until the pellet was dissolved, after which 1.8 ml of phosphate-buffered saline with 0.5% Nonidet P-40 and protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1% aprotinin) were added. Smad4 was immunoprecipitated with anti-FLAG M2; after four washes with lysis buffer, the immunocomplexes were resolved by SDS-PAGE and immunoblotted with anti-HA antibody for ubiquitin.
Mass Spectrometric Analysis of Mono-ubiquitinated Smad4 -HEK-293T cells were transfected, and ubiquitination assays were performed as described above. After immunoprecipitation, the beads were boiled in 100 l of 1% SDS for 5 min, 1 ml of lysis buffer was added, and a second anti-FLAG immunoprecipitation was performed. The proteins obtained after the second immunoprecipitation were resolved by SDS-PAGE and silver-stained (21). The non-ubiquitinated and mono-ubiquitinated bands (Fig. 6A) were cut out and digested in the gel (22). 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 (23). The peptides were eluted with acetonitrile-containing matrix (4-hydroxy-␣-cyanocinnamic acid) directly onto the target, and spectra were taken using a Bruker Biflex III MALDI-TOF-MS from Bruker Daltonics (Bremen, Germany).
The search for ubiquitinated peptides was accomplished by GPMAW (Lighthouse Data, Odense, Denmark), by scanning for the expected addition of either of the two most C-terminal tryptic peptides of human ubiquitin (GG or LRGG, of 114.05 or 383.22 Da, respectively) onto a Lys residue in the tryptic digest of human Smad4.
Smad4/R-Smad Complex Formation Assays-Association of transfected FLAG-or 6myc-tagged Smad4 with 6myc-or FLAG-tagged Smad2 or Smad3 was monitored by co-immunoprecipitation assays. Transfected HEK-293T cells were lysed and immunoprecipitated with anti-FLAG or anti-myc antibodies as described above, prior to immu-noblotting with the reciprocal antibody and ECL detection. To evaluate Smad4/R-Smad complexes more stringently, lysates of transfected HEK-293T cells were first immunoprecipitated using anti-FLAG agarose (Sigma), the immunocomplexes were washed four times with lysis buffer, and the bound Smad proteins were eluted from the beads by incubating for 2 h with FLAG peptide (10 g/ml) in lysis buffer. The eluted proteins were immunoprecipitated using rabbit anti-Smad4 antibody, and the final immunocomplexes were resolved on SDS-PAGE and immunoblotted.

Smad4 Is a Widely Expressed Protein with Slow Turnover in
Human Epithelial Cells-We screened a large panel of human epithelial cell types for the expression pattern of Smad4 using direct immunoblotting or immunoprecipitation followed by immunoblotting. Fig. 1A shows representative data from five such cell lines, one of which, the mammary carcinoma MDA-MB-468, was used as negative control because it lacks both Smad4 alleles (24). In all cell lines tested, we identified rather comparable levels of endogenous Smad4 expression, whereas MDA-MB-468 cells (and SW480 cells 2 ; Ref. 25) completely lacked Smad4 expression. The endogenous levels of Smad4 expression in human epithelial cells were roughly 20% of the levels of epitope-tagged Smad4 overexpressed in HEK-293T cells (Fig.  1C), as assessed using the same anti-Smad4 antibody. 2 We also examined the influence of several growth factor stimuli on the steady state levels of endogenous Smad4 (Fig. 1, A and B). In HepG2 and HEK-293T cells, stimulation with high doses of TGF-␤1 or BMP-7 for up to 24 h did not lead to any measurable effects on Smad4 levels. Stimulation of HEK-293T cells by a constitutively active epidermal growth factor receptor over the course of a 2-3-day transfection, or stimulation of porcine aortic endothelial cells stably transfected with plateletderived growth factor receptor ␤ by platelet-derived growth factor receptor BB, 2 neither had any measurable effects on endogenous Smad4 levels. Finally, high and low serum levels in the absence or presence of recombinant TGF-␤1 showed no appreciable effects on Smad4 protein expression in immortalized normal human keratinocytes HaCaT. 2 The expression of transfected Smad4 in HEK-293T cells after inhibition of protein synthesis by cycloheximide treatment remained rather robust and roughly equal to the non-treated control over a 12-h time period (Fig. 1, C and E). Similarly, metabolic labeling of transfected Smad4 in HEK-293T cells for 2 h followed by a cold amino acid chase for up to 12 h did not reveal any appreciable change in the Smad4 levels ( Fig. 1, D and E). From all these experiments, we conclude that human Smad4 exhibits rather stable expression patterns and that mitogenic or anti-proliferative growth factors do not appreciably change this pattern.
Missense in the MH1 domain of Smad4 that were originally described in human cancers, including pancreatic and colon cancers (12). One common feature of the four mutants analyzed in this previous study was their relative low levels of expression compared with wild type Smad4. The Smad4(L43S) mutation maps in a region close to the nuclear localization signal of Smad4 and exhibits the strongest phenotype in terms of protein instability ( Fig. 2D and Ref. 12). We analyzed the stability of this mutant after transfection in HEK-293T cells and blockade of protein synthesis by cycloheximide ( Fig. 2A), and after metabolic labeling and chasing with cold amino acids (Fig. 2B). We observed a dramatic and rapid down-regulation of the Smad4(L43S) protein levels in both experiments (Fig. 2, A-C) that strongly contrasted with the behavior of the wild type protein (Fig. 1). The half-life of this mutant was estimated at 4.5 h in transfected HEK-293T cells. The Smad4(L43S) mutant represents one of the most unstable Smad4 protein mutants described so far. We also analyzed the other three MH1 mutants, G65V, R100T, and P130S, using similar assays and measured their half-life (Fig. 2D). Smad4(P130S) exhibited intermediate phe-notype and a half-life of 11 h, Smad4(G65V) showed biphasic kinetics with a rapid down-regulation phase during the first 2 h followed by steady levels up to 12 h, and R100T showed very similar kinetics to L43S and a half-life of 5 h.
To investigate whether the observed decreased levels of Smad4(L43S) and the other three MH1 domain mutants (G65V, R100T, and P130S) are the result of proteasomal degradation, we measured their levels of expression after transfection in HEK-293T cells in the absence and presence of the proteasomal inhibitor MG-132 (Fig. 2E). Whereas inhibition of proteasomal activities had small effects on wild type Smad4 levels, it dramatically enhanced the levels of all missense MH1 domain mutants. Thus, the apparent instability of the missense mutants is likely to be the result of proteasomal degradation.
To further confirm that ubiquitin-mediated proteasomal degradation was the mechanism responsible for the apparent instability of the missense cancer mutants of Smad4, we coexpressed the various mutants together with enzymatically inactive mutants of human E2 ubiquitin-conjugating enzymes (Ubc) that are known to act as dominant negative inhibitors of endogenous ubiquitination pathways (26). The effects of such dominant negative E2 Ubcs on wild type Smad4 protein levels were measurable but modest (Fig. 3). However, UbcH3(CS), and to a lesser extent UbcH6(CS) and UbcH7(CS), rescued the weak or undetectable expression of the missense Smad4 mutants. We therefore conclude that the missense MH1 mutants are indeed heavily degraded via an ubiquitin-mediated proteasomal pathway, and furthermore, that UbcH3 appears to play a selective role in this mechanism. In contrast, the low degree of ubiquitination and proteasomal degradation of wild type Smad4 may be catalyzed by several E2 Ubcs, as we did not observe any specificity in the weak effects of dominant negative Ubcs.
Wild Type Smad4 Is Mono-or Oligo-ubiquitinated, Whereas Missense Smad4 Mutants Are Additionally Poly-ubiquitinated-We then examined whether Smad4 and its mutants could undergo ubiquitination in intact cells (Fig. 4). We coexpressed in HEK-293T cells wild type or missense mutant Smad4, tagged with a FLAG epitope, alone or together with ubiquitin tagged with an HA epitope, and performed immunoprecipitations of denatured proteins from the cell extract with the anti-FLAG antibody and immunoblotting of the resolved proteins with the anti-HA antibody. To make sure that we were not identifying coprecipitating ubiquitinated proteins, the primary immunoprecipitates were boiled in the presence of SDS and reprecipitated with the anti-FLAG antibody prior to immunoblotting (see Fig. 6). Such analyses showed that wild type Smad4 exhibited ubiquitinated products of low molecular weight that could correspond to mono-, di-, and possibly triubiquitinated Smad4 (Fig. 4A). Very low levels of high molec-ular weight poly-ubiquitinated chains were detected. In contrast, all missense mutants, and especially Smad4(L43S), exhibited significant levels of poly-ubiquitination and a different pattern of oligo-ubiquitinated protein bands. The presence of the oligo-ubiquitinated Smad4 species was detectable even by plain immunoblotting of total cell lysates (Fig. 4A, bottom  panel).
To obtain more direct evidence about the degree of Smad4 oligo-ubiquitination, we repeated the ubiquitination assays for wild type and L43S mutant Smad4 in the presence of wild type ubiquitin or mutant K48R ubiquitin (Fig. 4B). Lysine 48 of ubiquitin is one of the most commonly used residues for the creation of branched polymers of ubiquitin that are added on protein substrates that will eventually become proteasomally degraded (26). Thus, this mutant ubiquitin should not affect wild type Smad4 mono-ubiquitination, but it should affect polyubiquitination of missense mutant Smad4. Indeed, we could readily detect mono-ubiquitinated wild type Smad4 using ubiquitin K48R. In contrast, this K48R ubiquitin allowed formation only of minor species of oligo-ubiquitinated Smad4(L43S) and reduced the degree of poly-ubiquitination of Smad4(L43S) significantly (Fig. 4B, right panel). These data suggest that indeed Smad4 can be ubiquitinated and the patterns of ubiquitination differ between wild type and missense MH1 mutant Smad4 proteins. The former is primarily mono-or oligo-ubiquitinated, whereas the latter is mono-and poly-ubiquitinated by ubiquitin chains that are linked via either lysine 48 of ubiquitin or alternative lysines.
Domain Mapping of Ubiquitinated Smad4 -To identify the domains of ubiquitination in Smad4, we performed in vivo ubiquitination experiments of various deletion mutants of wild type Smad4 in intact cells (Fig. 5). Smad4 contains 12 lysines in its MH1 domain, 1 lysine in the linker domain, and 7 lysines in the MH2 domain, most of which are proposed to be exposed on the surface of the protein based on crystallographic studies (27). In wild type Smad4, we detected oligo-ubiquitinated protein bands linked to the linker-MH2 domain of the protein (Fig. 5, lanes 9 and 18). No detectable ubiquitinated linker domain alone (Fig. 5, lanes 7 and 16), or MH1-linker domain (Fig. 5, lanes 5 and 14) of Smad4 was observed. In contrast, the degree of poly-ubiquitination of the MH1-linker domain of Smad4(L43S) 2 was significantly higher, as was the case for the full-length mutant protein (Fig. 4). We therefore conclude that the conserved MH1 and MH2 domains of Smad4 have different ubiquitination capacities. The MH2 domain is more efficiently ubiquitinated but accumulates mainly mono-or oligo-ubiquitin chains, whereas the wild type MH1 domain is weakly ubiquitinated but becomes more heavily ubiquitinated in cancer missense mutants, and accumulates primarily poly-ubiquitin chains. It must be noted that this level of analysis does not distinguish between poly-ubiquitination of a single site on the substrate Smad4 and mono-ubiquitination of multiple sites of the substrate.
Lysine 507 Is a Major Site of Ubiquitination in Smad4 -To identify specific lysine residues within each one of the MH1linker and linker-MH2 domains of Smad4, we used large scale preparations of ubiquitinated Smad4 MH1-linker 2 and linker-MH2 domains in intact cells, after double immunoprecipitation (Fig. 6A). Such preparations were resolved by SDS-PAGE and stained with silver nitrate; the mono-ubiquitinated protein bands were localized and digested with trypsin within the gel pieces. Peptide fragments were then eluted and subjected to mass spectrometry to identify Smad4 and ubiquitin peptide fragments, but more importantly to identify branched peptides that corresponded to the covalent bond of the C-terminal ubiquitin glycine to the ⑀-amino group of a lysine in Smad4 (Fig. 6B, inset). This approach worked successfully only for the linker-MH2 domain that reproducibly gave high yield mono-ubiquitinated protein bands (Fig. 6A), but failed for the MH1-linker domain of wild type or L43S mutant Smad4. Inspection of the detected peptide masses indeed identified a branched peptide centered on lysine 507 of the MH2 domain (Fig. 6B).
Lysine 507 is part of the L3 loop of Smads (Fig. 6C) that is known to participate in recognition of the L45 loop on TGF-␤ type I receptors and of phosphoserines (28). Lysine 507 is fully conserved among all Smads in the family and among all members of the family in various species (Fig. 6D). In Smad4, lysine 507 has been proposed to recognize, together with other critical residues of the MH2 L3 loop and a basic patch, the phosphorylated diserine motif of an R-Smad (29,30). In R-Smads, the equivalent lysine residues are thought to form the phosphoserine binding motif that recognizes the phosphorylated glycine/serine-rich (GS) box in conjunction with the L45 loop of the activated type I receptors of the superfamily (30,31). In I-Smads, the equivalent lysine may be involved in recognition of the phosphorylated GS box in the type I receptor, although this is not fully confirmed yet (32). In conclusion, we identified one conserved lysine residue in the MH2 domain of Smad4, which is ubiquitinated.
Mutagenesis of Lysine 507 Leads to Decreased Smad4 Ubiquitination and Defective TGF-␤ Signaling-To functionally analyze the role of lysine 507 of Smad4, we mutated this residue to arginine or alanine. We first performed ubiquitination assays in intact cells and compared the efficiency and pattern of ubiquitination between wild type and lysine 507 mutant Smad4 (Fig. 7A). In all experiments we observed a reduced efficiency of Smad4 ubiquitination when lysine 507 was mutated to alanine (Fig. 7A) or arginine. 2 This analysis suggested that the absence of lysine 507 reveals ubiquitination that either targets additional lysines in Smad4 or alternatively targets a new lysine not previously engaged in the context of the wild type protein; similar observations have also been made for many other ubiquitinated proteins (26).
Because lysine 507 is known to form part of the phosphoserine recognition patch of the MH2 domain that contributes to Smad oligomerization, we tested the effects of lysine 507 mutations on R-Smad/Smad4 complex formation (Fig. 7B). As predicted, both K507A and K507R mutants exhibited weak interactions with Smad2 (Fig. 7B, left panel) and Smad3 (Fig.  7B, right panel), after overexpression of the constitutively active type I receptor for TGF-␤ in HEK-293T cells. These data derived from intact cells are consistent with the in vitro data published earlier using purified recombinant Smad MH2 domains (28).
We therefore predicted that mutations in lysine 507 should also have effects on the transcriptional activity of Smad4. To this end, we performed transient promoter-reporter assays (Fig. 7C) in a cell line that lacks endogenous Smad4 protein expression, the colon carcinoma SW480, after reconstitution of wild type or lysine 507 mutant Smad4 together with a synthetic gene reporter that contains 12 copies of the Smad binding element of the plasminogen activator inhibitor I promoter fused upstream of the luciferase cDNA (17). In this cell line the transcriptional activity of the Smad-dependent promoter is very weak because of the absence of endogenous Smad4 (Fig.  7C and Ref. 25). After reconstitution of wild type Smad4, we could measure strong ligand-dependent promoter activity. In contrast, the mutant Smad4 constructs exhibited significantly weaker ability to transactivate the promoter, K507A being less active than K507R. Thus, the ubiquitinated lysine 507 in the Smad4 MH2 domain plays important functional roles in TGF-␤ signal transduction, raising the possibility that ubiquitination of lysine 507 might regulate additional processes than its proteasomal degradation. This was also suggested by the fact that wild type Smad4 exhibits poor poly-ubiquitination but efficient monoand oligo-ubiquitination.
Ubiquitination Enhances the Formation of Functional Smad3-Smad4 Oligomers and Their Transcriptional Activity-We evaluated the above hypothesis to a first degree of approximation, by measuring the influence of ubiquitin overexpression, which as we have shown leads to Smad4 ubiquitination (Fig. 4), on the efficiency of oligomerization between Smad4 and Smad3, one of the two R-Smads of the TGF-␤ pathway (Fig.  8). We used a coimmunoprecipitation assay for the two Smads in the absence and presence of pathway activation using the constitutively active type I receptor for TGF-␤. We monitored efficient and receptor activation-dependent complex formation between the two Smads. Unexpectedly, overexpression of ubiquitin resulted in dramatic enhancement of the observed ligandinduced complex between Smad3 and Smad4 (Fig. 8A, compare  lanes 5 and 6). Furthermore, ubiquitin overexpression positively affected the complex formation between Smad3 and Smad4 even in the absence of pathway activation by the receptor (Fig. 8A, compare lanes 3 and 4). Ubiquitin overexpression did not appreciably affect the steady state levels of wild type Smad4 or Smad3 but induced the presence of mono-or oligoubiquitinated Smad4 species as observed before. Wild type ubiquitin had the same effects on Smad3-Smad4 oligomerization as ubiquitin(K48R) (Fig. 8A, compare lanes 4 and 6 and  lanes 7 and 8), a fact that enhances the evidence that monoubiquitination is sufficient for the observed enhanced oligomer formation. Finally, the mono-ubiquitinated Smad4 species were also detectable in the complexes with Smad3 but only after pathway activation by the constitutive receptor (Fig. 8A,  lanes 6 and 8, upper panel, solid arrowheads). To corroborate the above results more rigorously, we repeated the coprecipitation assays in the presence of ubiquitin using a double immunoprecipitation protocol (Fig. 8B). Accordingly, Smad3-Smad4 complexes were immunoprecipitated (IP1) by antibody targeting Smad3, and then the complexes were washed and eluted with excess antigenic peptide (FLAG) and the eluted material was re-precipitated using antibody against Smad4 this time. Finally the secondary precipitates were analyzed by immunoblotting. This protocol gave very clean background, and so Smad3-Smad4 oligomers were observed only after receptor activation (Fig. 8B, lane 2). Under such conditions, ubiquitin also led to robust Smad3-Smad4 oligomers even in the absence of receptor stimulation (Fig. 8B, lane 3), despite the equal levels of Smad protein expression.
To test the contribution of lysine 507 in the observed effect of ubiquitin on Smad3-Smad4 oligomer formation, we re- peated the coprecipitation experiments using the mutant Smad4(K507A) and ubiquitin(K48R) (Fig. 9). Compared with wild type Smad4, Smad4(K507A) did not form robust complexes in response to constitutive receptor activation (Fig. 9,  lanes 8 and 9), reproducing the data of Fig. 7B. Overexpressed ubiquitin(K48R) was able to elicit a weak but detectable positive effect on Smad3-Smad4(K507A) oligomerization, and this effect could not be further enhanced by receptor activation (Fig.  9, lanes 10 and 11). These results suggest that, in addition to lysine 507, other lysines in Smad4 may contribute to the observed enhancement of Smad oligomerization observed by ubiquitin. However, lysine 507 appears to play a major role in this mechanism and in addition it contributes strongly to the ligand-dependent nature of the ubiquitin effect. Thus, we conclude that mono-or oligo-ubiquitination of wild type Smad4 may be a novel mechanism to regulate the efficiency of oligomerization and, thus, the strength of TGF-␤ signaling.
Finally, we tested whether ubiquitin could affect a downstream Smad signaling response using a well established promoter-reporter system, the CAGA 9 -luciferase promoter (17). This promoter is highly specific, as it requires the combined effects of Smad3 and Smad4 to be induced by TGF-␤. Because overexpression of ubiquitin could affect multiple endogenous proteins, including transcription factors, co-activators, and members of the RNA polymerase II machinery, we chose this synthetic promoter to be able to measure specific effects that would affect the Smads only. To increase the specificity of the assays, we also employed the SW480 colon carcinoma cells that lack endogenous Smad4 expression, and reconstituted Smad4 via transfection (Fig. 10). Reconstitution of Smad4 resulted in a clear TGF-␤-dependent activation of the promoter. Co-expression of wild type ubiquitin did not lead to statistically significant modulation of the promoter response (Fig. 10A). However, co-expression of ubiquitin(K48R), which would only lead to mono-ubiquitination of Smad4, resulted in statistically significant (based on a two-tailed t test, at 95% confidence interval) enhancement of the promoter activity. Repeating the experiment with the lysine 507 mutant of Smad4 (Fig. 10B) unfortunately did not lead to conclusive results, as this mutant always resulted in very weak and highly variable activation of the promoter. Co-expressing wild type or K48R 2 ubiquitin together with the Smad4(K507A) mutant never resulted in enhanced Brackets with asterisks mark a characteristic mass peak cluster of human Smad4 linker-MH2 domain that remains unaltered between unmodified and mono-ubiquitinated proteins. A bracket with arrow indicates the mass peak cluster corresponding to the Smad4-ubiquitin peptide conjugate. The first mass peak within each cluster represents the mono-isotopic ion of the peptide. The corresponding peptide sequences are shown on top of the spectra, and arrows indicate sites of trypsin cleavage. C, structural model of the Smad4 MH2 domain. The Smad4 MH2 structure obtained from the protein data base (PDB number 1YGS) was visualized using RasMol software. The side chain of lysine 507 and the L3 loop are indicated in the ribbon diagram of the MH2 structure. D, sequence of the site of Smad4 mono-ubiquitination. The amino acid sequence of all Smad proteins corresponding to the L3 loop and flanking regions is shown. An arrow indicates the conserved lysine that corresponds to lysine 507 in Smad4, and all conserved amino acids among all eight Smad members are shown with gray background.
promoter activity above background levels. We conclude that the enhanced Smad oligomerization induced by mono-ubiquitination of Smad4 can lead to increased transcriptional effects. However, as the approach taken here, by overexpressing ubiquitin in mammalian cells, can affect multiple other proteins, including R-Smads or the TGF-␤ receptors, it is not possible to conclusively answer that Smad4 mono-ubiquitination has a direct role on the transcriptional activity of Smad proteins. Further studies need to address the exact molecular mechanism of such ubiquitin-mediated regulation in the Smad pathway.

DISCUSSION
Although ubiquitin-mediated degradation of R-and I-Smads has been recently established as a mechanism of efficient down-regulation of TGF-␤ superfamily pathways (2), equivalent post-translational modification of the common Smad effector of all such pathways, Smad4, has not yet been associated with regulation of normal signal transduction. In this report we provide evidence for differential roles of mono-or oligoubiquitination of wild type Smad4 and poly-ubiquitination of mutant Smad4. Our data propose a positive mechanistic role for Smad4 mono-or oligo-ubiquitination on TGF-␤ signal transduction, possibly by enhancement of R-Smad/Co-Smad oligomerization.
Recent reports have illustrated the importance of ubiquitination of cancer-derived mutants of Smad4 (13,14). The analysis of three additional point mutants, L43S, G65V, and P130S in this study (Figs. 1-4), further confirms that genetic mutations that presumably lead to errors in Smad4 protein folding eventually contribute to a selective mechanism of ubiquitination and degradation of these proteins. We therefore conclude that all major cancer-derived missense mutations or deletions in Smad4, the most commonly mutated gene in the Smad family, lead to efficient loss of Smad4 function by mobilizing its proteasomal degradation. Thus, ubiquitin-mediated proteolysis is intimately linked to the loss of a major tumor suppressor protein, Smad4, in human cancer.
Additional conditions that can lead to Smad4 degradation have been recently reported and include oncogenic Ha-Ras or overexpression of the COP9 signalosome and proteasome subunit CSN5/Jab1 (15,16). Our extensive screen for mitogenic and anti-proliferative factors that might modulate the endogenous levels of Smad4 in several epithelial cell types that respond to these factors did not reveal any obvious effect (Fig.  1). We therefore conclude that oncogenic stimuli such as Ha-Ras must operate at very high levels of signaling to elicit detrimental effects on Smad4 protein fate.
We showed that preferential poly-ubiquitination of Smad4 missense mutants can occur in addition to the mono-or oligoubiquitination of the proteins (Figs. 4 and 5). In contrast, the apparently stable wild type Smad4 protein can only be monoor oligo-ubiquitinated, and such modification cannot be linked to enhanced turnover of this protein (Fig. 1). Thus, Smad4 can be differentially regulated by selective ubiquitin modification. Our effort to identify specific ubiquitin pathways leading to poly-ubiquitination and degradation of missense Smad4 mutants versus pathways that regulate oligo-ubiquitination of wild type Smad4 was based on an extensive analysis of dominant negative E2 Ubcs (Fig. 3). UbcH3 and, to a lesser extent, UbcH6 and UbcH7 exhibited higher sensitivity in regulation of missense Smad4 proteolysis. A similar analysis performed previously for mutant Smad4(R100T) suggested UbcH5 as a possible mediator of selective degradation of this mutant and the homologous Smad2 mutant (13). In the present study, domi- nant negative UbcH5 isoforms were rather inactive toward the four MH1 mutants analyzed; however, one might anticipate different contributions by these E2 Ubcs, as they are known to act rather promiscuously (26). An effort to analyze the ubiquitination pathways based on specific E3 ligases that recognize Smad4 will be described elsewhere.
The fact that all four MH1 domain mutants lead to preferential poly-ubiquitination of Smad4 suggests that the putative errors in protein folding are recognized by chaperones that possibly act in a domain-specific manner. In contrast, a different mechanism seems to be responsible for ubiquitination of the MH2 domain in Smad4, as the resulting pattern of modification is different (Fig. 4). It would be attractive to identify distinct ubiquitination machineries (including E2 Ubcs as suggested from Fig. 3) that mediate poly-ubiquitination versus mono-or oligo-ubiquitination of Smad4. Our effort to map the three major domains of Smad4 for their ability to support their mono-, oligo-or poly-ubiquitination (Fig. 5) must be taken into account with caution. It is formally possible that distinct domains of a protein may exhibit different patterns of ubiquitination in isolation, whereas in the context of the full-length protein they may actually contribute significantly either as direct substrates, providing lysine residues, or as protein interfaces for recognition by the ubiquitination machinery. Thus, our failure to detect efficient ubiquitination of the MH1 and linker domains in isolation does not preclude their involvement in Smad4 ubiquitin-mediated modifications. We also analyzed the ubiquitination pattern of Smad4(L43S) MH1-linker domain and detected robust poly-ubiquitination but not oligo-ubiquitination of that mutant domain. 2 It would be interesting to analyze the domain-specific pattern of ubiquitination of cancer mutants that map in the MH2 domain. Although one such mutant that leads to Smad4 instability via ubiquitin-mediated proteolysis has been analyzed (14), no detailed mapping of ubiquitination sites in this mutant has been performed yet. In the R-Smad, Smad3, ubiquitin-mediated proteolysis that is catalyzed by the E3 ligase complex Roc1/SCF has been proposed to operate by modification of residues in the MH2 domain (33). We can therefore conclude that, although the current understanding of specific lysine acceptor sites on Smad proteins that undergo ubiquitination is largely unexplored, the present report provides evidence about a possible structural division of Smad4. Its MH1 domain seems to be poly-ubiquitinated, whereas its MH2 domain mono-or oligo-ubiquitinated. The linker domain may facilitate these modifications (e.g. by serving as recognition surface by the Smurf family of E3 ligases).
Surprisingly, ubiquitination of the MH2 domain in Smad4 appears not to be linked to any appreciable level of protein degradation (Fig. 1). To the contrary, in several experiments we have observed either neutral or slightly positive effects on wild type Smad4 stability when ubiquitin was overexpressed in HEK-293T cells (Figs. 4 and 5). This led to mapping more carefully ubiquitination sites in the wild type Smad4 linker-MH2 domain (Fig. 6). We employed a novel approach based on mass spectrometry, and we were successful in identifying one such acceptor lysine (Lys-507) that revealed the diagnostic branched peptide between a Smad4 MH2 tryptic fragment and the C terminus of ubiquitin (Fig. 6B). Our effort to map additional sites in the MH1, linker, or MH2 domains using mass spectrometry as a diagnostic tool were not fruitful because of the low level of ubiquitination of such putative sites. Lysine 507 thus appears as a major ubiquitin acceptor site in Smad4 and cannot be directly linked to the mode of poly-ubiquitination observed in the cancer mutants. This was tested by introducing a mutation of lysine 507 in the context of the MH1 missense mutants, and this did not seem to alter the degree or pattern of ubiquitination of the cancer mutants. 2 We therefore reasoned that mono-or oligo-ubiquitination of lysine 507 in Smad4 may serve other functional roles.
The critical position of lysine 507 in the L3 loop of Smad4 was recently recognized from detailed crystallographic studies of the R-Smad/Co-Smad complex (28 -30). Lysine 507 is fully solvent-accessible, and its side chain protrudes from the L3 loop surface to the neighboring space (Fig. 6C). Thus, structurally and biochemically, Lys-507 represents a reasonable site for modification by ubiquitin. Such modification might affect the structural conformation of the MH2 domain and have functional impact on Smad4, but it is unlikely that it would disturb the proper folding of the domain causing detrimental side effects. Our data derived from intact cells (Fig. 7) fully support the structural hypothesis that lysine 507 would positively contribute to TGF-␤ signaling. Indeed, mutagenesis of this residue in Smad4 led to decreased ubiquitination, decreased oligomerization with R-Smads, and decreased transcriptional activity (Fig. 7). Ubiquitination of Smad4(K507A) or Smad4(K507R) was still detectable, suggesting the presence of additional lysines that serve as ubiquitin acceptor sites. However, one might argue that the observed phenotypes of the K507A or K507R mutants were not the result of a decrease of Smad4 ubiquitination but instead were the result of structural effects on the oligomer interface. The latter interpretation of our data is possible; however, the evidence of Figs. 8 and 10 strongly argues that mono-or oligo-ubiquitination of Smad4 is also important for signal transduction. We reproducibly observed enhanced R-Smad/Co-Smad complexes when ubiquitin was overexpressed in mammalian cells (Fig. 8A). This was corroborated by careful oligomerization assays using sequential immunoprecipitations (Fig. 8B). Furthermore, we were able to visualize the presence of the mono-ubiquitinated Smad4 in Smad3-Smad4 complexes, strengthening our hypothesis that ubiquitin addition at the MH2 domain does not necessarily perturb the fold or function of this domain. We therefore propose that oligo-ubiquitination of Smad4 may be a mechanism to convert "inactive" Smad4 to an "active" conformation, presumably by offering additional protein surfaces to the MH2 domain, thus facilitating its interaction with co-factors that may stabilize or mediate the function of the R-Smad/Co-Smad complex. Whether indeed ubiquitination at lysine 507 is essential for Smad-mediated signal transduction was difficult to address at this stage, because simple mutagenesis of the critical residue may lead to pleiotropic effects. Despite this, we attempted to first measure the contribution of lysine 507 to the enhanced oligomerization effect of ubiquitin (Fig. 9). We can conclude that lysine 507 indeed participates critically in this process; however, additional lysines are obviously involved, as mutation of lysine 507 did not fully block the effect of ubiquitin on oligomerization. Second, we performed promoter-reporter assays to measure the contribution of ubiquitin to overall signal transduction by TGF-␤ (Fig. 10). Although the specificity of such assays is weak as discussed under "Results," we did obtain positive evidence for an enhancement of Smad signaling, especially by the mutant ubiquitin(K48R), which leads primarily to mono-ubiquitination of target proteins. Further analysis using purified mono-or oligo-ubiquitinated Smad4 and a faithful cell-free system of signal transduction is required to elucidate such a mechanism that would resemble recent discoveries of alternative roles of ubiquitin or other ubiquitin-like proteins on the targeting, mobilization, and recognition of endosomal proteins or nucleocytoplasmic shuttling proteins, such as Smad4 (34).
Our model for a novel role of ubiquitin in the regulation of the activation state of the Smad pathway complements its already established role in shutting down the TGF-␤ pathway, and underscores the importance of a systematic search for additional ubiquitination sites in Smad4 and other Smads. In addition to conventional scanning mutagenesis, the mass spectrometric approach described here provides a good tool to measure biologically important post-translational protein modifications based on ubiquitin or its relatives.