RING finger-dependent ubiquitination by PRAJA is ... - Nature

Oncogene (2006) 25, 693–705

& 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc

ORIGINAL ARTICLE

RING finger-dependent ubiquitination by PRAJA is dependent on TGF-b and potentially defines the functional status of the tumor suppressor ELF T Saha1, D Vardhini1, Y Tang1, V Katuri1, W Jogunoori1, EA Volpe1, D Haines2, A Sidawy1,5, X Zhou3, I Gallicano3, R Schlegel4, B Mishra1 and L Mishra1,5 1 Departments of Surgical Sciences, Medicine, Lombardi Comprehensive Cancer Center, Georgetown University, Washington, DC, USA; 2Fels Institute for Cancer Research and Molecular Biology, Temple University, Philadelphia, PA, USA; 3Department of Cell Biology, Georgetown University, Washington, DC, USA; 4Department of Pathology, Georgetown University, Washington, DC, USA and 5Department of Veterans Affairs Medical Center, Washington, DC, USA

In gastrointestinal cells, biological signals for transforming growth factor-beta (TGF-b) are transduced through transmembrane serine/threonine kinase receptors that signal to Smad proteins. Smad4, a tumor suppressor, is often mutated in human gastrointestinal cancers. The mechanism of Smad4 inactivation, however, remains uncertain and could be through E3-mediated ubiquitination of Smad4/adaptor protein complexes. Disruption of ELF (embryonic liver fodrin), a Smad4 adaptor protein, modulates TGF-b signaling. We have found that PRAJA, a RING-H2 protein, interacts with ELF in a TGF-bdependent manner, with a fivefold increase of PRAJA expression and a subsequent decrease in ELF and Smad4 expression, in gastrointestinal cancer cell lines (Po0.05). Strikingly, PRAJA manifests substantial E3-dependent ubiquitination of ELF and Smad3, but not Smad4. D-PRAJA, which has a deleted RING finger domain at the C terminus, abolishes ubiquitination of ELF. A stable cell line that overexpresses PRAJA exhibits low levels of ELF in comparison to a D-PRAJA stable cell line, where ELF expression is high compared to normal controls. The alteration of ELF and/or Smad4 expression and/or function in the TGF-b signaling pathway may be induced by enhancement of ELF degradation, which is mediated by a high-level expression of PRAJA in gastrointestinal cancers. In hepatocytes, half-life (t1/2) and rate constant for degradation (kD) of ELF is 1.91 h and 21.72 min1 when coupled with ectopic expression of PRAJA in cells stimulated by TGF-b, compared to PRAJA-transfected unstimulated cells (t1/2 ¼ 4.33 h and kD ¼ 9.6 min1). These studies reveal a mechanism for tumorigenesis whereby defects in adaptor proteins for Smads, such as ELF, can undergo degradation by PRAJA, through the ubiquitin-mediated pathway. Oncogene (2006) 25, 693–705. doi:10.1038/sj.onc.1209123; published online 10 October 2005

Correspondence: Dr B Mishra and L Mishra, Laboratory of Developmental Molecular Biology, Georgetown University, Medical–Dental Building, Rooms NW210–213, 3900 Reservoir Road, NW, Washington, DC 20007, USA. E-mail: [email protected] and [email protected] Received 28 June 2005; revised 11 August 2005; accepted 11 August 2005; published online 10 October 2005

Keywords: ELF; PRAJA; TGF-b; ubiquitination; halflife

Introduction The transforming growth factor-beta (TGF-b) signaling pathway is essential for cell polarity and lineage, and is a tumor suppressor pathway in multiple cell types (Souchelnytskyi et al., 2002; Tang et al., 2002; Siegel and Massague, 2003). Smads are the intracellular mediators of TGF-b signals, and their regulation of Smad function is dependent on modulation by adaptor proteins such as SARA (Smad anchor for receptor activation), filamin, microtubules, and ELF (embryonic liver fodrin) (Roberts and Sporn, 1990; Heldin et al., 1997; Derynck and Zhang, 2003; Itoh et al., 2003). Recently, disruption of the anchor function of ELF has been demonstrated to lead to tumorigenesis. Elf/deficient mice have disrupted TGF-b signaling because of an alteration in ELF interactions with Smad3 and Smad4 (Tang et al., 2003b). Smad4 is required for gut endoderm lineage, and Smad4 þ / mice develop gastric tumors after 12 months (Xu et al., 2000). Smad2 plays a key role in gastrulation, while Smad3 is important for the establishment of the gastrointestinal mucosal immune response to TGF-b signals (Tang et al., 2003b). Smad3-deficient mice frequently develop gut abscesses and die between 1 and 10 months from impaired mucosal immunity (Yang et al., 1999). Defective liver development with loss of gastrointestinal epithelial cell shape and polarity is also seen in elf/ homozygotes, Smad2 þ / and Smad3 þ / double heterozygous mice. Strikingly, hepatocellular carcinoma (HCC) is seen in elf þ / heterozygotes (Weinstein et al., 2001). ELF therefore is one of the few adaptor proteins that acts as a tumor suppressor. The regulation of ELF abundance therefore is likely to be important in tumorigenesis. Although gene dosage is critically important in this family (as haploinsufficiency phenotypes result in aberrant gut, brain, liver hypoplasia, and gastrointestinal carcinoma) (Weinstein et al., 2000; Shi and

TGF-b signaling is modulated by ubiquitination of ELF by PRAJA T Saha et al

694

Massague, 2003; Siegel and Massague, 2003; Tang et al., 2003b), other mechanisms, such as targeted protein proteolysis, also play an important role in the regulation of protein levels (Xu and Attisano, 2000). Biological functions of many proteins are altered by their covalent attachment to polypeptide modifiers (Schwartz and Hochstrasser, 2003). The best-known example of this type of modification is ubiquitination (Li et al., 2003). Ubiquitin-dependent protein degradation is involved in the regulation of various cellular processes, including cell cycle progression, signal transduction, transcription, DNA repair, and protein quality control (Koepp et al., 1999; Laney and Hochstrasser, 1999; Hicke and Dunn, 2003). Recent studies have revealed the ability of Smads to interact with multiple components of the 26S proteasome system before and after Smad activation (Izzi and Attisano, 2004). Ubiquitin is an abundant 76-amino-acid polypeptide that can be covalently conjugated to specific proteins by the formation of an isopeptide bond between its carboxyl-terminus and the amino group of a lysine residue of the target protein (Wang, 2003; Janse et al., 2004). The selectivity of ubiquitination is largely mediated by the recognition of substrates by E3; posttranslational modifications, such as phosphorylation of the substrate, can regulate this interaction. To date, a number of RING E3 ligases have been characterized on a molecular level (Chen et al., 2005; Dupont et al., 2005; Zhong et al., 2005). E3s with known amino-acid sequences include the N-end rule E3s of yeast and mammals and members of the HECT (homologous to E6-AP C-terminus) family (Dinudom et al., 1998). Mammalian HECT E3s include E6-AP, which targets p53 for ubiquitination in the presence of human papillomavirus E6 (Honda and Yasuda, 1999), and Nedd4, which ubiquitinates epithelial sodium channel subunits (Zhu et al., 1999; Liu et al., 2000). Other E3s include Mdm2, which catalyses both its own ubiquitination and that of p53 (Koepp et al., 1999; Bonni et al., 2001); the anaphase-promoting complex (APC); and other F box and cullin-containing complexes whose substrates include Sic1p, G1 cyclins, inhibitor of nuclear factor kB (IkB), and b-catenin (Laney and Hochstrasser, 1999). Interactions involving ubiquitination are an integral part of the signaling functions of Smads and other proteins and include several ubiquitin pathways. For instance Smurf2 an E3 ubiquitin ligase, regulates both cytoplasmic receptor-regulated Smads (R-Smad; such as Smad1) and nuclear R-Smad (such as Smad2 and Smad3) ubiquitination and proteasomal degradation (Kavsak et al., 2000; Stroschein et al., 2001). Secondly, nuclear R-Smads (e.g. Smad3) show a novel ability to regulate the ubiquitination of several key regulators, such as the oncoprotein SnoN and the multidomain docking protein HEF1 (human enhancer of filamentation 1), through a physical interaction with different ubiquitin E3 ligases (HECT family, Skp1/cullin/F-box protein (SCF) family, and APC complex (Zhang et al., 2001)). Nuclear R-Smads are ubiquitinated by the action of the SCFFbw1a/Roc1 RING–H2 E3 ligase Oncogene

complex, which is exported to the cytoplasm and finally degraded there (Fukuchi et al., 2001). Thirdly, TGF-b induces Smad7-Smurf1 association that export to the cytoplasm, and target the receptor kinases, which then undergo degradation (Ebisawa et al., 2001; Asano et al., 2004). The Smurfs can also regulate ubiquitination and degradation of other target proteins, including the TGFb receptor complex and the transcriptional corepressor SnoN (Ebisawa et al., 2001; Fukuchi et al., 2001; Miyazono et al., 2003; Asano et al., 2004). These recently described physical links between Smads and E3 ligases of the 26S proteasome–ubiquitin system allow intracellular events triggered by the TGF-b family ligands to connect with those induced by many other extracellular regulators to form an extremely complex signaling network that regulates a wide range of biological activities (Izzi and Attisano, 2004). Proteasomal degradation of Smad4 occurs in tumor cells, which either harbor deleterious mutations in MADH4 or express activated oncoproteins such as Ras (Xu and Attisano, 2000). The ultimate degradation of nuclear Smads after prolonged ligand stimulation has been firmly established as a mechanism that shuts off the signaling pathway, yet the specific ubiquitination mechanism remains elusive. More recently, PRAJA has been identified as a similar RING finger protein with E3 ligase activity in bone morphogenetic protein (BMP) pathway ubiquitination (Sasaki et al., 2002). Praja open-reading frame (ORF) comprises 1188 bases encoding a 395-amino-acid protein of B60 kDa, and is involved in cell proliferation, apoptosis, juxtaposition, and architecture (also in Sanskrit: birth, development, and offspring) (Mishra et al., 1997). PRAJA ORF exhibits extensive homology to RING finger proteins at the C-terminus end between residues 347 and 373 (Mishra et al., 1997). Computer analysis of this domain demonstrates that it contains a perfect C3H2C3 Zn-binding motif or RING-H2 finger found in many regulatory proteins, as well as several human proto-oncogene products such as promyelocytic leukemia protein (PML), ret finger protein (RFP), BRCA1, and inhibitors of apoptosis (IAP) (Schwede et al., 2003). In this study, we show that expression of PRAJA is inversely related to that of ELF. PRAJA ubiquitinates ELF and Smad3, but not Smad4, in a TGF-b dependent manner. Here, we demonstrate that PRAJA E3 ligase activity regulates TGF-b signaling by controlling ELF abundance through ubiquitin-mediated degradation. Results ELF expression in liver tissue and regeneration In order to demonstrate ELF expression during liver regeneration, we performed immunohistochemical staining on sections from livers after 70% partial hepatectomy (PH) at time points 0, 5, and 15 min, and 1, 3, 6, 12, 48, and 72 h. Maximal staining for ELF was seen throughout the lobule at 0–5 min, and 6 h after PH, but diminished to almost undetectable levels at 1 and 48 h


695

after PH (Figure 1a), remaining only in centrilobular hepatocytes (Supplementary Figure 1b). For these studies, we utilized a ELF-specific polyclonal antibody (VA-1) (Mishra et al., 1998, 1999; Tang et al., 2002) that recognizes N-terminal sequence unique for ELF spectrin. Confocal experiments have shown ELF expression at the apical, canalicular surface of hepatocytes (Weinstein et al., 2001). TGF-b levels during PH have been described before. Normal quiescent livers contain low level of TGF-b expression. During liver regeneration, at 4 h after PH, TGF-b expression rises threefold above normal levels, peaking to an eightfold level at the end of the 72 h post-PH (Braun et al., 1988; Russell et al., 1988). A representative diagram has been presented in Figure 1b. PRAJA is expressed at distinct time points in liver regeneration We further investigated expression of PRAJA during liver regeneration after PH and determined for correla-

tion with ELF expression. We performed immunohistochemical staining, using polyclonal antibody specific for PRAJA (Mishra et al., 1997), on sections from livers after 70% PH at time points 0, 5, and 15 min, and 1, 3, 6, 12, 48, and 72 h. PRAJA labeling was detectable in adult centrilobular hepatocytes, but not in Kupffer cells or hepatic stellate cells. During liver regeneration, PRAJA expression was induced in a distinctly time-dependent manner. Maximum labeling for PRAJA was observed throughout the lobule around 15 min after PH, before diminishing to almost undetectable levels at 6 h and then rising again after 48–72 h (Figure 1a), remaining only in centrilobular hepatocytes. Thus, expression of PRAJA and ELF were inversely proportional to each other. We demonstrate here that during 4–6 h of the PH, peak expression of PRAJA occurs while TGF-b expression is already high (Figure 1a) compared to expression in normal quiescent liver. Low ELF expression during that time frame reflects ELF ubiquitination and destruction by high expression of PRAJA in the presence of high levels of TGF-b, supported by the subsequent experiments.

Figure 1 Expression of ELF is inversely proportional with the expression of PRAJA in liver tissue regeneration, Gastric cancer cells and in stable cell lines expressing PRAJA and D-PRAJA. (a) Graphical representation of immunohistochemical labeling of liver samples with antibodies specific to ELF and PRAJA, separately, that was collected at various time points after partial hepatectomy. All experimental values were normalized by dividing with the control value and represented as IDV/Ctrl where IDV stands for ‘Integrated density value’ and Ctrl stands for ‘Control’. (b) Fold increase in the expression of TGF-b during partial hepatectomy when compared with the normal liver. (c) Human gastric cancer cell lines (lane 1, NCI-N87 and lane 2, SNU-1) exhibit a loss of ELF expression with an overexpression of PRAJA (lanes 1 and 2) when compared to the wild-type gastric cell line (lane 3). Loss of expression of Smad4 is also observed in one gastric cancer cell line but not in the other cell line when compared to the wild-type cells (compare lanes 1 and 2). In all the cases, lysates were prepared and immunoblotted (IB) with antibodies specific to ELF, PRAJA, Smad4, or Actin. (d) Schematic representation of the wild-type (PRAJA) and the dominant-negative PRAJA (D-PRAJA). RING finger domain in PRAJA is indicated. (e) Cell lines were created in HepG2 cells that stably express PRAJA protein and its mutant form DPRAJA, as shown in the top panel. This blot was stained with antibody against V5, a short signal peptide to show the overexpression of the indicated proteins and serves as a control for the bottom panel. In the bottom panel, lysates were prepared from the stable cells and separated in a SDS–PAGE gel and immunoblotted with ELF antibody. Cells expressing PRAJA has low levels of ELF expression, but cells expressing D-PRAJA have elevated levels of ELF compared to HepG2 alone. Oncogene


696

At 72 h post–PH, PRAJA expression is low, thus ELF remains accumulated in the cells, although TGF-b expression is high. Cancer cell lysates show overexpression of PRAJA Loss of ELF is crucial for tumorigenesis, as is observed in our studies in hepatocellular, gastric, and colon cancers (Tang et al., 2005). We demonstrate in this manuscript that ubiquitination of ELF by RING-H2 finger, PRAJA plays a critical role in its degradation via the proteasomal pathway. To test for overexpression of PRAJA in cancer cells, we chose two gastric cancer cell lines (NCI-N87) and (SNU-1) that show loss of ELF. Both cell lines showed a fivefold increase in PRAJA expression when compared to the normal gastric cells (Figure 1c). The loss of ELF paired with the overexpression of PRAJA provides further support for an important role for PRAJA in ELF ubiquitination. Stable cell lines expressing full-length PRAJA demonstrate loss of ELF To further demonstrate the reciprocal behavior of both ELF and PRAJA, we constructed two expression plasmids, one containing full-length PRAJA tagged with a small peptide V5 at C-terminus, and the other containing a deletion mutation of PRAJA, termed DPRAJA in the same expression plasmid (in which the RING-H2 finger domain at the C-terminus has been deleted). A schematic representation of the constructs is shown in Figure 1d. Using these recombinant plasmids, we have generated two stable cell lines in HepG2 cells that constitutively express both V5-PRAJA and V5-DPRAJA (Figure 1e). Deletion of RING-H2 domain results in the loss of ubiquitination, which is reflected on the rise in ELF expression levels. ELF is downregulated in cell lines that overexpress PRAJA, suggesting an interaction between the two. Conversely, ELF is upregulated and accumulates in the D-PRAJA-overexpressing stable cell line when compared to ELF levels in HepG2 cells. These studies indicate that interactions between ELF and PRAJA occur via RING finger domain, supporting a functional role in the differential expression of PRAJA and ELF, as was initially observed during PH. PRAJA overexpression modulates TGF-b-induced cell proliferation To examine PRAJA function in relation to TGF-b response, we performed thymidine incorporation assays

on HepG2 cell lines treated with TGF-b and other cytokines. Cells were transfected with expression vectors encoding PRAJA or D-PRAJA, or with expression vector alone. We observed a differential expression under the influence of TGF-b (Supplementary Figure 2). We also observed an increase in cell proliferation in cells transfected with PRAJA in the presence of TGF-b and not in presence of platelet-derived growth factor (Supplementary Figure 2), suggesting that overexpression of PRAJA modulates the TGF-b pathway. In vitro and in vivo association of PRAJA with ELF in the presence or absence of TGF-b To determine for ELF interactions with PRAJA, first, an in vitro binding assay was performed by exogenously producing and labeling PRAJA and ELF, using [35S]methionine and a rabbit reticulocyte-coupled transcription and translational system (Chatterjee-Kishore et al., 2000; Yin et al., 2002). Recombinant plasmids containing ELF, PRAJA and D-PRAJA were utilized in this assay to generate the corresponding proteins. In vitro translated proteins were immunoprecipitated with the respective antibodies and used in this assay. Both PRAJA and D-PRAJA are found to interact weakly with ELF in the absence of TGF-b (Figure 2a). Increasing the concentration of ELF does not improve this interaction. This demonstrates that the association of ELF and PRAJA in a cell-free system can be saturated. To determine whether PRAJA associates endogenously to ELF in the presence of TGF-b, we performed a protein–protein interaction assay using HepG2 cell lysates. TGF-b treatment was performed before harvesting the cells at indicated time points, with subsequent immunoprecipitation (IP) of PRAJA. As shown in Figure 2b, PRAJA co-immunoprecipitates with ELF at 30–60 min after TGF-b treatment, suggesting a role for PRAJA in ELF localization. The same association was observed when ELF antibody was used in IP and blotted with PRAJA antibody (Figure 2b). Subcellular distributions of ELF and PRAJA were examined upon their expression in HepG2 cells by confocal microscopy (Figure 2c). We observed that both ELF and PRAJA were distributed diffusely across the cytoplasm and the cell membrane of the cells (Figure 2c; I and IV subsequently), with weak colocalization signals (as shown in Figure 2c; VII). Upon TGF-b stimulation, PRAJA label appears distinctly at the cell membrane at 30 min and 1 h later, especially at cell–cell contact sites

Figure 2 In vitro and In vivo interaction between PRAJA and ELF in the presence and absence of TGF-b stimulation. (a) TNT-T7coupled transcription/translation system from Promega was used to produce labeled exogenous proteins from the recombinant plasmids with one copy of ELF, PRAJA, and D-PRAJA using [35S]methionine. ELF shows a weak association with PRAJA and DPRAJA in the absence of TGF-b, as indicated on the top. (b) HepG2 cells were collected after incubation with TGF-b at different time points. Cells were then immunoprecipitated using antibody specific to PRAJA followed by immunoblotting (IB) with antibody specific to ELF, to see the interaction between ELF and PRAJA. Prominent protein–protein interactions were observed at 30–60 min after TGF-b treatment as shown by the accumulation of the 200 kDa band of ELF. (c) Confocal microscopy demonstrates the subcellular distribution of ELF and PRAJA in a TGF-b-dependent manner. HepG2 cells were incubated with TGF-b for 30 min and 1 h followed by immunofluorescence detection of ELF or PRAJA by confocal microscopy with two different fluorescence probes. ELF is labeled red and PRAJA is labeled green. Top panel demonstrates membrane and punctate vesicle labeling of ELF. Middle panel shows PRAJA labeling cell membrane and Co-Loc (bottom panel) demonstrates colocalization of ELF with PRAJA at cell membrane at both time points, appearing as yellow spots (arrows). The final row represents the transmission picture (DIC) of the above cells. Oncogene


697

indicated by the arrows in Figure 2c, VIII and IX. This phenomenon was observed in 30% nonsynchronized cells in culture. The bottom panel represents the phase pictures of the respective cells (Figure 2c; X, XI, and XII). This again further supports in vivo ELF interaction with PRAJA.

PRAJA mediates ubiquitination of ELF and itself in the presence of TGF-b The finding that ELF associates with the RING finger protein PRAJA prompted us to hypothesize that ELF may be a candidate substrate for PRAJA. We examined PRAJA E3 ubiquitin ligase activity for ELF in vivo.

Oncogene


698

We also tested the TGF-b components Smad3 and Smad4 to determine whether they were also ubiquitinated by PRAJA. HepG2 cells were transfected with expression vectors encoding PRAJA, ELF, and HA-tagged ubiquitin. Cells were harvested following 30 mins treatment of TGF-b and the lysates were collected after 18 h. Ubiquitinated polypeptides were immunoprecipitated with ELF antibody and analysed by SDS–polyacrylamide gel electrophoresis (SDS-PAGE). Western blot using an antibody to HA was performed to detect HA– ubiquitin-conjugated proteins. Ubiquitinated products were recognized by higher-molecular-mass products with a smear above the 200 kDa band representing ELF (Figure 3a, lane 1). In similar assays with Smad3 and Smad4, PRAJA did not generate ubiquitinated Smad4 products (Figure 3a, lane 5), although ubiquitinated Smad3 conjugates were observed (Figure 3a, lane 4). As RING finger subunits are essential components with E3 ubiquitin ligase activity and because these RING proteins recruit and activate E2s, we tested whether PRAJA might be an E3 ligase for itself and act as both E2 and E3. The products of substrateindependent ubiquitination reactions with PRAJA were analysed as above. Ubiquitinated conjugates of PRAJA were observed as a smear on top of wild-type PRAJA (Figure 3a, lane 3). RING finger mutations abolish E3 activity of PRAJA RING finger domains are conserved cysteine-rich amino-acid sequences that form two interleaved Znbinding sites. There are a total of eight cysteines and histidines that constitute the sites of metal coordination. Computer analysis of this domain shows that it contains a perfect C3H2C3 Zn-binding motif or RING finger found in many regulatory proteins, such as EL5 DNAbinding protein, as well as several human protooncogenes, such as Cbl, PML, RFP, and BRCAI and IAP (Figure 5b). RING-H2 fingers are classified based on the presence of a histidine in the fifth coordination site. The RING finger motif of PRAJA is suggested to be the recognition module for E2 and therefore necessary for E3 activity. Thus, we examined the effect of the mutant form of PRAJA, without the RING consensus, to determine for the abrogation of PRAJA induced ubiquitination of ELF in vivo and in the presence of TGF-b. A mutant form of PRAJA (DPRAJA) was generated with deletion of the RING finger motif (Figure 1d), resulting in the loss of E3 activity. Cotransfection in HepG2 cells with expression vectors tagged with D-PRAJA (mutant), ELF, and HA– ubiquitin do not ubiquitinate ELF in a dominantnegative mode. These results indicate that ELF is ubiquitinated in mammalian cells and that PRAJA is involved in its ubiquitination process via its RING finger (Figure 3a, lane 2). Dose-dependent ubiquitination of ELF by PRAJA To test the possibility that ubiquitination of ELF by PRAJA is influenced by the enzyme–substrate ratio, we performed similar in vivo ubiquitination assays as Oncogene

described above, with increasing doses of PRAJA in the presence of a constant quantity of the substrate, ELF and TGF-b (100 pM final concentration). Low levels of ubiquitination of ELF were observed with low levels of PRAJA: ELF ratios (3 :1), whereas slower migrating, higher levels of ubiquitinated forms of ELF were observed with higher levels of PRAJA:ELF ratios (6 :1 and 8 :1) (Figure 3b). This finding reveals that an efficient ubiquitination reaction can occur between ELF and PRAJA in cells with high levels of PRAJA. Treatment with proteasomal inhibitors MG132 and MG115 leads to ELF accumulation To test whether ubiquitinated ELF is targeted by the 26S proteasome, we conducted a time-point study in which HepG2 cells treated with MG132 and MG115 (data not shown) were harvested at 0, 2, 4, and 8 h. Western blot analysis with an antibody to ELF revealed an accumulation of ELF at 6–8 h (Figure 3c). Actin was used as a loading control. These results suggest that ELF is ubiquitinated by PRAJA and is marked for degradation by the proteasomal pathway. ELF half-life analysis in TGF-b-stimulated HepG2 cells To determine whether the stability of ELF protein may be differentially regulated in proliferating cells, the half-life of ELF was measured in both untreated and TGF-b-treated HepG2 cells. TGF-b treatment was also performed in cells transfected with PRAJA and D-PRAJA. ELF protein half-life was measured by determining the level of protein at various time points (1–4 h) after treatment with the protein synthesis inhibitor cycloheximide (Pan and Haines, 1999; Alarcon-Vargas et al., 2002). Interestingly, Figure 4a (top panel) shows that ELF protein degraded much more rapidly in cycloheximide-treated, TGF-b-stimulated cells that were transfected with PRAJA, than in unstimulated PRAJA-transfected cells. The half-life (t1/2) of ELF protein in TGF-b-stimulated PRAJAtransfected cells is 1.91 h, whereas in PRAJA-transfected cells, without stimulation, ELF half-life (t1/2) is 4.33 h (Figure 4b, left panel and Table 1). In each case, the broken line corresponds to the TGF-b-treated cells and the solid line corresponds to TGF-b-untreated cells. When the conditions are kept constant, protein degradation follows first-order kinetics and the rate of degradation reaction is described by a single apparent rate degradation constant (kD) and can be calculated by the following equation: t1/2 ¼ ln 2/kD ¼ 0.693/kD. The values of the rate constant of the degradation (kD) in different experimental samples have been given in Table 1. Twofold higher kD value is obtained in the TGF-b-stimulated PRAJA transfected cells when compared to unstimulated PRAJA-transfected cells (Table 1). Similar results were obtained with ELF degradation when HepG2 cells are transfected with D-PRAJA or transfected with vector. All have higher t1/2 values and lower kD values for ELF (Table 1) compared with PRAJA transfection both in the presence


699

Figure 3 PRAJA ubiquitinates ELF and Smad3 but not Smad4 when stimulated with TGF-b: (a) In vivo ubiquitination assay was completed to assess the role of PRAJA in the TGF-b signaling pathway components. Expression vectors containing PRAJA, D-PRAJA, ELF, Smad3, Smad4, or Actin, and HA tagged ubiquitin were transfected in HepG2 cells. Cells were harvested after 30 min of TGF-b treatment. Co-immunoprecipitation (Moser et al., 2001) was performed by the protein of interest followed by immunoblotting (IB) with antibody to ‘HA’ for each transfection. ELF showed prominent ubiquitination by PRAJA, represented by ubiquitinated adducts at the top (lane 1) that demonstrated the E3 ligase activity of PRAJA. The D-PRAJA that has deleted RING-H2 finger motif at the C-terminus did not ubiquitinate ELF (lane 2), which supports the fact that RING H2 motif is important for such activity. PRAJA performed as both E2-conjugating protein and E3 ligase when in the absence of the substrate and ubiquitinated itself (lane 3). PRAJA also ubiquitinated Smad3 (lane 4) but did not ubiquitinate Smad4 (lane 5). Positive controls for ELF, PRAJA, Smad3, and Smad4 are in lanes lanes 6, 7, 8, and 9, respectively. (b) Dose-dependent ubiquitination assay: PRAJA induces low and high ubiquitinated adducts in a dosage-dependent manner in vivo when stimulated with TGF-b for 30 min before harvesting. Immunoblotting (IB) was performed with HA-specific monoclonal antibody after co-immunoprecipitation (Moser et al., 2001) with antibody specific to ELF (VA-1). Low ubiquitinated conjugates with lower doses (lane 1) and high ubiquitinated conjugates with higher doses (lane 3) of PRAJA were observed. Lane 2 corresponds to the intermediate ubiquitinated products. An increase in the expression of PRAJA (bottom panel) was shown. (c) Accumulation of ELF was observed in the presence of the 26S proteasomal inhibitor MG132 in the presence of TGF-b. HepG2 cells were harvested at 0, 2, 4, and 8 h after treatment with (left panel)/without (right panel) proteasomal inhibitor MG132 (10 mm). MG132-treated cells (left panel) showed a marked accumulation of ELF at 8 h, compared to 2 and 4 h time point, indicating that ubiquitinated ELF is targeted by the 26S proteasome for degradation.

Oncogene


700

Figure 4 Half-life determination of ELF in HepG2 cells: (a) HepG2 cells (2 105) were plated in a six-well tissue culture plate and transfected with PRAJA (top panel) and D-PRAJA (middle panel) 24 h later. Bottom panel represents wild-type HepG2 cells. In each case, protein synthesis inhibitor, cycloheximide, was added to the culture and maintained for indicated time points. TGF-b stimulation was completed 30 min before harvesting, where indicated. Cells harvested in SDS cell lysis buffer, before being fractionated in 4–12% NuPage gels and immunoblotted with ELF antibody. Ectopic and endogenous levels of PRAJA are shown as indicated. Actin was also shown as a loading control. (b) The decay curve of ELF was generated from the band intensities that was shown in (a) and the equation was generated using CRICKET graphics software in Macintosh. Solid line represents untreated and the broken line represents treatment with TGF-b.

and absence of TGF-b (compare Figure 4a and b, and Table 1). PRAJA panels in Figure 4a demonstrate PRAJA protein levels in PRAJA-transfected and -untransfected cells where PRAJA levels remain unchanged. Degradation kinetic analysis of ELF is consistent with the third-order polynomial function (Figure 4b, left panel and equation on it) in cells transfected and stimulated by PRAJA and TGF-b. In all other cases, ELF degradation by curve fit follows a second-order polynomial function (Figure 4b, middle and right panel with the equation). Degradation of ELF is highly regulated by PRAJA in a TGF-b-dependent manner. This again strengthens our conclusion that ELF and PRAJA interact with each other and that PRAJA acts as an E3 ubiquitin ligase and ubiquitinates ELF, initially modifies ELF and subsequently destroys it. Discussion Inhibition of the antiproliferative effect of TGF-b is often associated with progression of tumorigenesis in human cancers (Markowitz and Roberts, 1996). Here, we demonstrate strong evidence that the RING finger (RING-H2) protein PRAJA ubiquitinates ELF, a tumor Oncogene

suppressor and a Smad adaptor protein that is involved in TGF-b signaling, and that PRAJA may play an important regulatory role in liver regeneration. A model representing PRAJA as a potential regulatory protein in TGF-b signaling pathway is shown in Figure 5a. In brief, we propose that, upon TGF-b induction, TGF-b serine/threonine kinase receptor II activates TGF-b receptor I. This is followed by ELF phosphorylation and subsequently a complex is formed between ELF and Smad3 that is in turn phosphorylated by the type I receptor. Thereafter, PRAJA emerges as an E3 ubiquitin ligase that attaches ubiquitin molecules to ELF and localizes the complex towards the plasma membrane. At the same time, ubiquitin chains elongate and ELF is displaced from the signaling pathway and then becoming a target for 26S proteasome and is destroyed. Under normal circumstances, Smad4, Smad3, and phosphorylated ELF form a complex that maintains the signaling pathway for the normal growth of the cells. Indeed, disruption of TGF-b signaling is seen in almost all gastrointestinal cancers. A ribbon and stick model of the best-fit superposition of the RING-H2 domain of PRAJA, simulated by the Swiss–Prott model protein data bank (PDB) repository, is shown in Figure 5b. The hypothetical energy


701

minimum tertiary structure is designed from the structurally similar RING-H2 finger protein EL5 (PDB code: 1LYM). The backbone of this motif is shown in green and contains one a chain and two b-pleated sheets nearby (dashed square box). The protein is folded in such a way that it can accommodate two Zn molecules in the clefts shown by yellow arrows and this Zn and RING finger motif are both responsible for its function.

Figure 5 Theoretical and computer-generated model for ELF– PRAJA interaction: (a) A hypothetical schematic diagram depicting PRAJA’s possible role in the determination of ELF’s fate. Ubiquitinated ELF is targeted by the 26S proteasome; ubiquitinated ELF may confer crucial regulatory functions for ELF and Smad3. TGF-b stimulates ELF/Smad3 association in a phosphorylation-dependent manner, followed by association with PRAJA. I, TGF-b type I receptor; II, TGF-b type II receptor; and S3, Smad3. The red filled circles indicate ubiquitin molecules. The dark blue ovals indicate PRAJA. The orange filled circles indicate activated phosphate groups. The turquoise ovals indicate an ankyrin-based complex that associates with ELF in the absence of TGF-b. (b) A ribbon model of the best-fit superposition of the RING finger (RING-H2) (C3H2C3) domain of PRAJA, simulated by the Swiss– Prott model PDB repository. The hypothetical tertiary structure was designed from the structurally similar RING finger protein EL5 (PDB code: 1LYM). This structure contains one a helix and two b-pleated sheets that remain in the white box shown in the figure. It also accommodates two zinc molecules in the clefts that were indicated by yellow arrows.

Studies using PH in mouse indicate that, in liver regeneration, PRAJA peaks at 15 min and 48–72 h by immunohistochemical labeling after PH. ELF induction occurs later, during liver regeneration, showing a peak at 6 h after PH with a concomitant decrease of PRAJA (6 h after PH). Similarly, at 48 and 72 h, PRAJA levels increase, with a concomitant decrease in ELF expression. Our findings suggest that PRAJA could be a key mediator of ELF degradation and that endogenous levels of ELF are dynamically regulated through a possible ELF-PRAJA feedback loop via the ubiquitinmediated proteasomal pathway. The differential activity of PRAJA, both as an essential regulatory protein in hepatocyte development and as an ubiquitinator of ELF, an important tumor suppressor adaptor protein, inhibits apoptosis, and is possibly dictated by its own intracellular concentration. Our hypothesis is that low levels of PRAJA maintained in normal gastrointestinal cells are likely to play important regulatory roles for normal development. However, when PRAJA activities are high, PRAJA-mediated ubiquitination induces ELF degradation, abrogating the ELF-Smad TGF-b signaling pathway, thereby leading to tumorigenesis (Figure 1). The results in this study demonstrate that the RING finger protein PRAJA is regulated by TGF-b; that it facilitates ubiquitination of ELF, a tumor suppressor adaptor protein involved in TGF-b signaling; and that it may play an important regulatory role in liver development. Results of the assays we utilized suggest that PRAJA may have differential effects in determining ELF fate. In gastrointestinal cell differentiation, biological signals for TGF-b are transduced through transmembrane serine/threonine kinase receptors that signal to Smad proteins (Roberts and Sporn, 1990; Heldin et al., 1997; Souchelnytskyi et al., 2002; Derynck and Zhang, 2003; Itoh et al., 2003; Siegel and Massague, 2003; Tang et al., 2003b). Interactions involving ubiquitination are an integral part of the signaling functions of Smads, involving at least six ubiquitin pathways (Kretzschmar and Massague, 1998). The common mediator, Smad4, a tumor suppressor, is often inactivated in human gastrointestinal cancers (Zhang et al., 1997). However, the mechanism by which Smad4 mutations interfere with Smad activity remains uncertain. Our previous studies have shown that disruption of ELF, a Smad4 adaptor protein, inactivates TGF-b signaling, with elf þ / mutant mice that develop HCCs. Conforming to these observations, the present study indicates that PRAJA, a RING finger E3 ubiquitin ligase, interacts with ELF and ubiquitinates it. ELF associates with PRAJA in a TGF-b-dependent manner.

Table 1 Determination of t1/2 (half-life of ELF) and kD (rate constant of the degradation of ELF) in our experimental samples Samples t1/2 (h) kD (min1)

ELF+PRAJA (+TGF-b)

ELF+PRAJA (TGF-b)

ELF+D-PRAJA (+TGF-b)

ELF+D-PRAJA (TGF-b)

ELF (+TGF-b)

ELF (TGF-b)

1.91 21.72

4.33 9.6

3.83 10.85

4.25 9.78

3.66 11.34

4.25 9.78 Oncogene


702

After 30 min of TGF-b induction, a strong interaction is observed (Figure 2b and c), which is approximately when the ELF-Smad heteromeric complex translocates into the nucleus. These studies raise possibilities that PRAJA interaction and degradation of ELF could result in the disruption of Smad4 signaling, with a potential role in the development of HCCs. Indeed, we have found a fivefold increase in PRAJA expression with a concomitant decrease in ELF and Smad4 expression in gastrointestinal cancers (Po0.05) (Figure 1c). It is interesting to note that escape from TGF-b growth inhibition occurs in the presence of PRAJA (60% more growth), as well as, to a lesser extent, with D-PRAJA, suggesting a dominant-negative role for D-PRAJA. These results are similar to those for Cbl, where substrate recognition and association are independent of activation (Joazeiro et al., 1999). Treatment with potent proteasomal inhibitors MG132 result in the accumulation of ELF at the end of 8 h, indicating that PRAJA ubiquitinates ELF and that its degradation is mediated via the proteasomal pathway (Figure 3). The functional role of ubiquitinated ELF and the possible role of RING-mediated ubiquitination await further study. One possibility is that RING finger protein PRAJA ubiquitinates ELF (Figure 3) and therefore provides it with an important cellular function, which is further supported by the ELF– PRAJA interactions modulated by TGF-b (Figure 2). These observations are analogous to studies on other oncogenic E3 ligases, such as Mdm2 and its substrate, p53 (Li et al., 2003). RING E3s, which mediate ubiquitination, represent the largest E3 family known to date and include the c-Cbl proto-oncogene product, the multisubunit SCF and APC cell cycle-regulatory complexes, the Mdm2 protooncogene product that regulates the p53 tumor suppressor, BRCA1, and members of the IAP family of antiapoptotic proteins (Sun, 2003). In addition, it is possible that, in vivo, changes in conformation, alterations induced by phosphorylation, and changes in intermolecular associations, such as dimerization, may influence the capacity of RING finger proteins to function with E2s in ubiquitination. Adhering to these attributes, the compact RING module of PRAJA could confer an additional regulatory role on a wide array of TGF-b signaling proteins that associate with it. Recent studies also show that PRAJA ubiquitinates Dlxin-1 and Msx2 proteins and thus regulates the transcriptional function of homeodomain protein DLX5 through Dlxin1 via an ubiquitin-dependent degradation pathway (Sasaki et al., 2002). Importantly, this observation confirms the role of PRAJA in the TGF-b/BMP pathway. The importance of RING finger domain in ubiquitination is further supported by the fact that in vivo assays with the RING finger mutant PRAJA (D-PRAJA) abolish ubiquitination of ELF (Figure 3). RING mutations of BRCA1 are associated with familial carcinomas. Strikingly, PRAJA manifests substantial E3-dependent ubiquitination of ELF, but not Smad4, while still exhibiting ubiquitinated Smad3 conjugates. This finding suggests a role for PRAJA in the membrane Oncogene

localization of the ELF and the Smad signaling complex. Furthermore, it is observed that in the absence of a substrate, PRAJA has the capacity to ubiquitinate itself. This ability is consistent with recent observations that multiple, otherwise unrelated RING finger proteins, including Mdm2, AO7, Siah1, kf-1, have the inherent capacity to ubiquitinate themselves (Lorick et al., 1999). RING mutations of BRCA1 are associated with familial carcinomas (Brzovic et al., 1998). A potential model for RING-mediated E3 activity is one in which the RING and surrounding regions not only associate with E2-ubiquitin but also provide a favorable environment for the transfer of ubiquitin from E2 to an available lysine. Such a mechanism is analogous to the mechanism in models for the function of N-end rule E3s (Ohta et al., 1999). RING proteins, defined by eight cysteines and histidines that coordinate two zinc ions, vary substantially in length and composition. RINGs have cysteines in the first three and last three coordination sites and a His in the fourth site. Additionally, proteins bearing this motif have either a Cys [C3HC4 RING (RING-HC)] or a His [C3H2C3 RING (RINGH2)] in the fifth position. The first, second, fifth, and sixth cysteines/histidines coordinate one cation, and the third, fourth, seventh, and eighth coordinate the second (Saurin et al., 1996). As determined with consensus sequences, PRAJA falls into the RING-H2 category, with the amino acids indicated predicted to be coordination sites. Here, we establish such a mechanism by showing that pathway and growth of cells that are dependent on an adaptor protein ELF is inactivated by PRAJA, an E3 ligase. Multiple other cancers derived from mesoendodermally derived epithelium are associated with TGF-b/ BMP pathway inactivation, where it may regulate progenitor cell fate (Souchelnytskyi et al., 2002; Siegel and Massague, 2003; Tang et al., 2003a). Indeed, the functions of TGF-b are complex and extend beyond their role in the inhibition of cell growth. TGF-b induces the growth of mesenchymal cells, alters synthesis of extracellular matrix components, and metalloproteases, which is involved in cell invasion (Roberts and Sporn, 1990; Heldin et al., 1997; Souchelnytskyi et al., 2002; Derynck and Zhang, 2003; Itoh et al., 2003; Tang et al., 2003b). TGF-b signals also modulate the immune response to tumors and are thought to play a role in tumor angiogenesis (Wieser, 2001). The development of gastrointestinal tumors in elf þ /and elf þ //Smad4 þ / mutants indicates a crucial role for ELF, a b-spectrin, which acts as an essential adaptor protein for the proper transmission of signals generated by the TGF-b pathway. These studies demonstrate that loss of expression of ELF through PRAJA could play an important role in the development of gastrointestinal tumors, which are among the most lethal forms of cancers.

Materials and methods Animals and PH PH was carried out, as described before (Bhanumathy et al., 2002), on Harlan mice. Briefly, a median incision line was


703 made, exposing the large median lobe and the left lateral lobe. The hepatic duct was ligated and both lobes were excised. Thus, hepatic parenchyma constituting 65–75% was removed, leaving the right lateral lobe and the small caudate within the peritoneum. Following PH, the animals were killed at different time points of 0, 5, and 15 min, and 1, 3, 6, 12, 48, and 72 h. Subsequently, the liver residual mass was excised at different time points of 0, 5, and 15 min, and 1, 3, 6, 12, 48, and 72 h, and fixed in 4% paraformaldehyde for paraffin section. The slides from the paraffin section were stained with the respective antibodies as indicated and quantitated by using the DP70 software (Olympus) in a PC computer directly attached to the microscope (Olympus). Construction of the plasmids and stable cell lines cDNA sequence of PRAJA was amplified by gene-specific primer and inserted into either pcDNA3.1/V5-His-TOPO (V5PRAJA) or pCMV mammalian expression vectors (pCMVPRAJA) (Invitrogen). cDNA sequences corresponding to the RING-H2 domain (347–391 amino acids) of wild-type PRAJA were deleted by site-specific polymerase chain reaction to construct V5-D-PRAJA or pCMV-D-PRAJA, and then subsequently subcloned into the same mammalian expression vector. These recombinant plasmids were used in transfection studies. For the generation of the stable cell lines, PRAJA and D-PRAJA mammalian expression plasmids were transfected into HepG2 cells and the stable integrants were selected using 600 mg/ml Geneticin (Invitrogen). Cell culture and treatments The human HCC cell line, HepG2 (American Type Culture Collection (ATCC) #HB-8065), along with stable cell variants expressing V5-PRAJA and V5-D-PRAJA, were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 2 mM L-glutamine, 100 U/ ml penicillin, and 100 mg/ml streptomycin in 5% CO2. Human gastric cancer cell lines NCI-N87 (ATCC #CRL-5822) and SNU-1 (ATCC #CRL-5971) were maintained in the RPMI 1640 medium supplemented with 10% FBS, 2 mM L-glutamine, and antibiotics in 5% CO2 (all media and reagents were purchased from GIBCO, Invitrogen). TGF-b (Sigma) was added to a final concentration of 100 pM and cultures were incubated for 15, 30, and 60 min. Transfection and thymidine incorporation assays Cells were seeded at a density of 2 105 cells/well in six-well dishes and were then transfected using either ELF or PRAJA expression constructs, or the vector alone (3 mg of DNA/well), using Fugene 6 as a transfection reagent (Roche). Following transfections, cells were washed with 2 DMEM after 12–18 h and then treated with 5 mg/ml of TGF-b before being incubated for an additional 24 h. The cells were pulsed with [3H]thymidine for the final 2 h of the treatment period. Cells were then collected by trypsinization and put on a filter paper for the quantification of [3H]thymidine incorporation, using an LS 6500 Multi-purpose Scintillation Counter (Beckman). All experiments were repeated at least three times, and similar results were obtained each time. Immunohistochemistry An indirect immunoperoxidase procedure was used for immunohistochemical localization of ELF and PRAJA protein in regenerating mouse liver. An antibody to a peptide corresponding to amino acids 145–159 (CLRRKYRSREQPQS) specific for PRAJA was produced in rabbit as described previously and used to detect PRAJA by immuno-

histochemical labeling (Mishra et al., 1997). For ELF, the primary antibody was rabbit anti-mouse VA-1 (against Nterminal amino acids 2–14, ELQRTSSVSGPLS, specific for ELF) described previously (Mishra et al., 1998, 1999). Paraffin-embedded sections (8 mm) were deparaffinized in xylene and graded alcohol and rinsed in 1 phosphatebuffered saline (PBS). Endogenous peroxidase was quenched using 3% hydrogen peroxide (Sigma). Nonspecific binding sites were blocked in 1 PBS containing 5% goat serum and 1 mg/ml bovine serum albumin (BSA) for 1 h. The sections were incubated overnight at 41C in a humidified chamber with primary antibody diluted to 2.5–5 mg/ml in 1 PBS containing 1 mg/ml BSA and 0.05% Triton X-100. All further steps were performed at room temperature, including four, 5-min rinses with 1 PBS containing 1% goat serum after each successive step. Sections were incubated with peroxidase-conjugated goat anti-rabbit antibody (Jackson ImmunoResearch Laboratories) that was diluted in PBS with 1% serum, for 30 min. After four rinses, 200–500 ml of the insoluble peroxidase substrate diaminobenzidine (Pierce) was added to cover the entire tissue on the slide for color development, prior to an additional 2 min rinse in distilled water. Counterstaining was then performed with Harris’s hematoxylin solution (modified) (Sigma) for 1 min, followed by a rinse in distilled water for 5 min. Sections were dehydrated by passages through graded alcohol concentrations and finally with xylene. Coverslips were mounted using DPX (Fluka Labs) and monitored under a microscope. In vitro interaction assay TNT-T7-coupled transcription/translation system from Promega was used to produce exogenous proteins from the constructed plasmid (Chatterjee-Kishore et al., 2000; Yin et al., 2002). [35S]methionine was used in the reaction mixture to make the proteins label and thus can be detected by autoradiography. The produced proteins were then immunoprecipitated from the reaction mixture, as described below, by respective antibodies and used in our interaction assay. The binding assay was performed in 50 mM Tris, pH 7.4, 2 mM ATP, and 5 mM MgCl2, before being incubated at 371C for an hour. Following incubation, it was again immunoprecipitated and resuspended in SDS lysis buffer. The samples were separated in 4–12% NuPage Bis–Tris gel (Invitrogen). The gel was then dried and autoradiographed after 4 days. Co-IP and immunoblot To analyse ELF–PRAJA interactions under the influence of TGF-b, HepG2 cells were treated with 100 pmol/ml final concentration of TGF-b. Cells were harvested at different time points of 0, 15, 30, and 60 min, and were lysed in 150 mM NaCl, 1% Triton X-100, 0.5% deoxycholic acid, 50 mM Tris buffer pH 7.5, containing 1 mM phenylmethylsulfonyl fluoride, 10 mg/ml aprotinin, and 10 mg/ml leupeptin. For IP, lysed cells were incubated with respective antibodies as indicated in the figures for 2 h at 41C and then incubated with protein G plus agarose beads (Amersham) overnight at 41C. Individual host serum was used in the same way that serves the negative control in IP. Immunocomplexes were washed with detergent– PBS and were fractionated on SDS–PAGE (6% or 4–12% gel; Novex) and blotted to nitrocellulose membrane. Western blotting (WB) was performed as before (Mishra et al., 1998; Mishra et al., 1999). Enhanced chemiluminescence (ECL) kit (Perkin-Elmer Life Sciences) was used to develop the immunoblots. The loading control was performed under the same conditions using actin specific antibody (Santa Cruz Biotechnology Inc.). Oncogene


704 Confocal laser scanning immunofluorescence microscopy Colocalization studies were performed with anti-ELF (polyclonal, VA-1), anti-PRAJA (polyclonal) antibody on mouse embryonic fibroblasts. Primary antibodies were visualized with either tetramethyl rhodamine isothiocyanate-conjugated goat anti-rabbit immunoglobulin G; fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse immunoglobulin G, or cyanine (Cy5) (all secondary antibodies from Jackson ImmunoResearch Laboratories). The samples were analysed on a Bio-Rad MRC-600 confocal microscope (Bio-Rad), with an ILT model 5470 K laser (Ion Laser Technology) as the source for the crypton–argon ion laser beam. FITC-stained samples were imaged by excitation at 488 nm and Rhodamine-stained samples were imaged by excitation at 568 nm with a 598–621 bandpass emission filter with a 505–540 bandpass emission filter. Cy5-stained samples were imaged by excitation at 638 nm with a 647–670 bandpass emission filter using a 60 (numerical aperture 1.3) objective and 20 objective. Digital images were analysed using Metamorph software (Universal Imaging). Figures were prepared using Adobe Photoshop. In vivo ubiquitination assays To detect ubiquitinated forms of ELF, HepG2 cells were transfected/cotransfected with expression vectors encoding ELF, PRAJA, Smad, and HA-tagged ubiquitin individually as well as in combination. At 24 h post-transfection, cells were treated for 3 h with a potent 26S proteasome inhibitor MG132 (10 mM final concentration; Calbiochem) and 30 min with TGF-b (100 pM final concentration) before harvest, and were lysed in RIPA buffer (100 mM NaCl, 1.0% NP-40, 0.5% deoxycholate, and 50 mM Tris (pH 8.0)) for 30 min at 41C. Following incubation, the lysates were collected as a supernatant after centrifugation at maximum speed at 41C temperature. Samples were analysed by Western blot with a protein-specific antibody. ELF was immunoprecipitated using VA-1 (ELF specific antibody) and was subjected to SDS– PAGE. Ubiquitinated ELF conjugates were detected by WB using antibodies against hemagglutinin (HA) (12CA5; Boehringer Mannheim, Mannheim, Germany). Similar assays were performed to detect ubiquitinated conjugates of Smad3 and Smad4 with antibodies specific for Smad3 and Smad4 (Santa Cruz Biotechnology Inc.). Cycloheximide treatment of cells for determination of half-life of ELF HepG2 cells were seeded at 2 106 cells/ml in culture medium in six-well tissue culture dishes the day before the experiment. Transfection with PRAJA and D-PRAJA was performed as

above. Where indicated, 24 h following transfection, 75 mg/ml final concentration of cycloheximide solution in DMSO (Sigma) was added to the cells and maintained at indicated time points (1–4 h) (Pan and Haines, 1999; Alarcon-Vargas et al., 2002). Cells were treated with TGF-b for 30 min prior to harvesting at different time points after cycloheximide treatment and washed with PBS before lysates were prepared in SDS cell lysis buffer. The samples were fractionated in 4–12% NuPage Bis–Tris gel (Invitrogen). Decay curves were prepared using the band intensity of both ELF and actin at different time points and the equation was generated using the Cricket Graphics program in Macintosh.

Abbreviations RING, Really Interesting New Gene; TGF-b, transforming growth factor-beta; ELF, embryonic liver fodrin; SARA, Smad anchor for receptor activation; ORF, open-reading frame; HECT, homologous to E6-AP C-terminus; HEF1, human enhancer of filamentation 1; APC, anaphase-promoting complex; IkB, inhibitor of nuclear factor kB; PML, promyelocytic leukemia protein; RFP, ret finger protein; IAP, inhibitors of apoptosis; PH, partial hepatectomy; PBS, phosphate-buffered saline; BSA, bovine serum albumin; ATCC, American Type Culture Collection; HCC, hepatocellular carcinoma; DMEM, Dulbecco’s modified Eagle’s medium; FBS, fetal bovine serum; IP, immunoprecipitation; IB, immunoblotting; FITC, fluorescein isothiocyanate; SDS, sodium dodecyl sulfate; ECL, enhanced chemiluminescence; WB, Western blotting; SDS–PAGE, SDS–polyacrylamide gel electrophoresis; VA-1, ELF-specific polyclonal antibody; HA, hemagglutinin; R-Smad, receptor-regulated Smad; PDB, protein data bank. Acknowledgements We thank Drs M Zasloff and S Evans for their critical review of the manuscript, and R Redman and A Rashid for their invaluable assistance with histopathology specimens and extensive confirmatory review of all slides. Grant Support: NIHR01 DK56111 (LM), NIHR01 CA106614-01A2 (LM), NIHR01 DK58637 (BM), VA Merit Award (LM), and R Robert and Sally D Funderburg Research Scholar (LM).

Competing interest statement The authors have no competing financial interests.

References Alarcon-Vargas D, Tansey WP, Ronai Z. (2002). Oncogene 21: 4384–4391. Asano Y, Ihn H, Yamane K, Kubo M, Tamaki K. (2004). J Clin Invest 113: 253–264. Bhanumathy CD, Tang Y, Monga SP, Katuri V, Cox JA, Mishra B et al. (2002). Dev Dyn 223: 59–69. Bonni S, Wang HR, Causing CG, Kavsak P, Stroschein SL, Luo K et al. (2001). Nat Cell Biol 3: 587–595. Braun L, Mead JE, Panzica M, Mikumo R, Bell GI, Fausto N. (1988). Proc Natl Acad Sci USA 85: 1539–1543. Brzovic PS, Meza J, King MC, Klevit RE. (1998). J Biol Chem 273: 7795–7799. Chatterjee-Kishore M, van Den Akker F, Stark GR. (2000). J Biol Chem 275: 20406–20411. Oncogene

Chen D, Kon N, Li M, Zhang W, Qin J, Gu W. (2005). Cell 121: 1071–1083. Derynck R, Zhang YE. (2003). Nature 425: 577–584. Dinudom A, Harvey KF, Komwatana P, Young JA, Kumar S, Cook DI. (1998). Proc Natl Acad Sci USA 95: 7169–7173. Dupont S, Zacchigna L, Cordenonsi M, Soligo S, Adorno M, Rugge M et al. (2005). Cell 121: 87–99. Ebisawa T, Fukuchi M, Murakami G, Chiba T, Tanaka K, Imamura T et al. (2001). J Biol Chem 276: 12477–12480. Fukuchi M, Imamura T, Chiba T, Ebisawa T, Kawabata M, Tanaka K et al. (2001). Mol Biol Cell 12: 1431–1443. Heldin C-H, Miyazono K, ten Dijke P. (1997). Nature 390: 465–471. Hicke L, Dunn R. (2003). Annu Rev Cell Dev Biol 19: 141–172.


705 Honda R, Yasuda H. (1999). EMBO J 18: 22–27. Itoh S, Thorikay M, Kowanetz M, Moustakas A, Itoh F, Heldin C-H et al. (2003). J Biol Chem 278: 3751–3761. Izzi L, Attisano L. (2004). Oncogene 23: 2071–2078. Janse DM, Crosas B, Finley D, Church GM. (2004). J Biol Chem 279: 21415–21420. Joazeiro CA, Wing SS, Huang H, Leverson JD, Hunter T, Liu YC. (1999). Science 286: 309–312. Kavsak P, Rasmussen RK, Causing CG, Bonni S, Zhu H, Thomsen GH et al. (2000). Mol, Cell 6: 1365–1375. Koepp DM, Harper JW, Elledge SJ. (1999). Cell 97: 431–434. Kretzschmar M, Massague J. (1998). Curr Opin Genet Dev 8: 103–111. Laney JD, Hochstrasser M. (1999). Cell 97: 427–430. Li M, Brooks CL, Wu-Baer F, Chen D, Baer R, Gu W. (2003). Science 302: 1972–1975. Liu X, Elia AE, Law SF, Golemis EA, Farley J, Wang T. (2000). EMBO J 19: 6759–6769. Lorick KL, Jensen JP, Fang S, Ong AM, Hatakeyama S, Weissman AM. (1999). Proc Natl Acad Sci USA 96: 11364–11369. Markowitz SD, Roberts AB. (1996). Cytokine Growth Factor Rev 7: 93–102. Mishra L, Cai T, Levine A, Weng D, Mezey E, Mishra B et al. (1998). Int J Dev Biol 42: 221–224. Mishra L, Cai T, Yu P, Monga SP, Mishra B. (1999). Oncogene 18: 353–364. Mishra L, Tully RE, Monga SP, Yu P, Cai T, Makalowski W et al. (1997). Oncogene 15: 2361–2368. Miyazono K, Suzuki H, Imamura T. (2003). Cancer Sci 94: 230–244. Moser MJ, Gong Y, Zhang MN, Johnston J, Lipschitz J, Minuk GY. (2001). Dig Dis Sci 46: 907–914. Ohta T, Michel JJ, Schottelius AJ, Xiong Y. (1999). Mol Cell 3: 535–541. Pan Y, Haines DS. (1999). Cancer Res 59: 2064–2067. Roberts AB, Sporn MB. (1990). Peptide growth factors and their receptors. In: Handbook of Experimental Pathology, Sporn MB, Roberts AB (eds). Springer-Verlag: New York, pp. 419–472. Russell WE, Coffey Jr RJ, Ouellette AJ, Moses HL. (1988). Proc Natl Acad Sci USA 85: 5126–5130.

Sasaki A, Masuda Y, Iwai K, Ikeda K, Watanabe K. (2002). J Biol Chem 277: 22541–22546. Saurin AJ, Borden KL, Boddy MN, Freemont PS. (1996). Trends Biochem Sci 21: 208–214. Schwartz DC, Hochstrasser M. (2003). Trends Biochem Sci 28: 321–328. Schwede T, Kopp J, Guex N, Peitsch MC. (2003). Nucleic Acids Res 31: 3381–3385. Shi Y, Massague J. (2003). Cell 113: 685–700. Siegel PM, Massague J. (2003). Nat Rev Cancer 3: 807–821. Souchelnytskyi S, Moustakas A, Heldin C-H. (2002). Trends Cell Biol 12: 304–307. Stroschein SL, Bonni S, Wrana JL, Luo K. (2001). Genes Dev 15: 2822–2836. Sun Y. (2003). Cancer Biol Ther 2: 623–629. Tang B, Vu M, Booker T, Santner SJ, Miller FR, Anver MR et al. (2003a). J Clin Invest 112: 1116–1124. Tang Y, Katuri V, Dillner A, Mishra B, Deng CX, Mishra L. (2003b). Science 299: 574–577. Tang Y, Katuri V, Iqbal S, Narayan T, Wang Z, Lu RS et al. (2002). Oncogene 21: 5255–5267. Tang Y, Katuri V, Srinivasan R, Fogt F, Redman R, Anand G et al. (2005). Cancer Res 65: 9228–9237. Wang T. (2003). Front Biosci 8: d1109–d1127. Weinstein M, Monga SP, Liu Y, Brodie SG, Tang Y, Li C et al. (2001). Mol Cell Biol 21: 5122–5131. Weinstein M, Yang X, Deng C. (2000). Cytokine Growth Factor Rev 11: 49–58. Wieser R. (2001). Curr Opin Oncol 13: 70–77. Xu J, Attisano L. (2000). Proc Natl Acad Sci USA 97: 4820–4825. Xu X, Brodie SG, Yang X, Im YH, Parks WT, Chen L et al. (2000). Oncogene 19: 1868–1874. Yang X, Letterio JJ, Lechleider RJ, Chen L, Hayman R, Gu H et al. (1999). EMBO J 18: 1280–1291. Yin Y, Wang ZY, Mora-Garcia S, Li J, Yoshida S, Asami T et al. (2002). Cell 109: 181–191. Zhang Y, Chang C, Gehling DJ, Hemmati-Brivanlou A, Derynck R. (2001). Proc Natl Acad Sci USA 98: 974–979. Zhang Y, Musci T, Derynck R. (1997). Curr Biol 7: 270–276. Zhong Q, Gao W, Du F, Wang X. (2005). Cell 121: 1085–1095. Zhu H, Kavsak P, Abdollah S, Wrana JL, Thomsen GH. (1999). Nature 400: 687–693.

Supplementary Information accompanies the paper on Oncogene website (http://www.nature.com/onc)

Oncogene