Ebp1, an ErbB3 receptor-binding protein, inhibits cell proliferation and acts as a putative tumor suppressor. Ebp1 translocates into the nucleus and functions as a transcription co-repressor for E2F-1. Here, we show that Ebp1 p42 isoform can be sumoylated on both K93 and K298 residues, which mediate its nuclear translocation and are required for its anti-proliferative activity. We find that translocation in liposarcoma (TLS)/FUS, an RNA-binding nuclear protein that is involved in pre-mRNA processing and nucleocytoplasmic shuttling, has Sumo1 E3 ligase activity for Ebp1 p42. Ebp1 directly binds TLS/FUS, which is regulated by genotoxic stress-triggered phosphorylation on Ebp1. Ebp1 sumoylation facilitates its nucleolar distribution and protein stability. Overexpression of TLS enhances Ebp1 sumoylation, whereas depletion of TLS abolishes Ebp1 sumoylation. Moreover, unsumoylated Ebp1 mutants fail to suppress E2F-1-regulated transcription, resulting in loss of its anti-proliferation activity. Hence, TLS-mediated sumoylation is required for Ebp1 transcriptional repressive activity.
Ebp1, a ubiquitously expressed protein, localizes in both the nucleus and the cytoplasm and binds ErbB3 receptor in human breast cancer cells (Yoo et al., 2000). Ebp1 is the human homolog of a previously identified cell–cycle-regulated mouse protein p38-2G4 (Radomski and Jost, 1995). The treatment of serum-starved AU565 breast cancer cells by the ErbB3/4 ligand heregulin (HRG) results in dissociation of Ebp1 from ErbB3 and translocation from the cytoplasm into the nucleus (Yoo et al., 2000). EBP1 overexpression inhibits proliferation of human fibroblasts, and this effect is linked to its nucleolar localization (Squatrito et al., 2004). Northern blotting analysis shows two transcripts with1.7 kb being the major one and 2.2 kb the minor one (Nakagawa et al., 1997). These observations are consistent with the report that two Ebp1 mRNAs occur in several normal human organs (Yoo et al., 2000). Recently, we report that the PA2G4 gene encodes two Ebp1 isoforms, p48 and p42, which differentially regulate PC12 cell survival and differentiation (Ahn et al., 2006; Liu et al., 2006). P48 is 54 amino acids longer than p42 at its N-terminus. The longer form p48 localizes in both the cytoplasm and the nucleolus and suppresses apoptosis, whereas the shorter form p42 predominantly resides in the cytoplasm and promotes cell differentiation (Liu et al., 2006). Ebp1 is phosphorylated on serine 360 (S360) by protein kinase Cδ, which strongly stimulates Ebp1 p48 to bind Akt and suppresses apoptosis (Ahn et al., 2006). Ebp1 p42 specifically associates with nucleophosmin (NPM)/B23 on epidermal growth factor stimulation, whereas p48 constantly binds NPM/B23. NPM/B23 regulates cell proliferation and survival through p42 and p48, respectively (Okada et al., 2007). Ebp1 also binds tumor suppressor retinoblastoma protein (Rb), leading to inhibition of the E2F1-regulated transcription (Xia et al., 2001; Zhang et al., 2003). Ebp1 contains an autonomous C-terminal transcriptional repression domain that binds histone deacetylase 2 (Zhang et al., 2005). It strongly suppresses both androgen receptor-mediated gene transcription and tumorigenesis of prostate cancer cells and salivary adenoid carcinoma cell metastasis in mice (Zhang et al., 2005; Yu et al., 2007). Thus, Ebp1 p42 acts as a tumor suppressor by repressing E2F-1 and androgen receptor-mediated transcription.
TLS (translocation in liposarcoma) was initially identified as the N-terminus of TLS-CHOP (CCAAT/enhancer binding protein (C/EBP) homologous protein), a fusion oncoprotein that is expressed as a consequence of the t(12;16) translocation, which is implicated in human myxoid and round cell liposarcomas (Crozat et al., 1993; Rabbitts et al., 1993). TLS or the related EWS (Ewing's sarcoma) gene is also fused to a panel of unrelated transcription factors through chromosomal translocation in some of the human sarcomas and leukemias. The merged oncoproteins contain an N-terminal domain from TLS or EWS (Sanchez-Garcia and Rabbitts, 1994). The C-terminal domain of TLS is replaced by the DNA-binding domain from the corresponding transcription factor. The N-terminal domain in the fusion oncoproteins has a critical role in transformation (Zinszner et al., 1994; Kuroda et al., 1997). TLS possesses an SYGQ-rich region in the N-terminus, an RNA recognition motif, a C2/C2 zinc-finger motif and at least one RGG repeat region in the C-terminus (Morohoshi et al., 1998). Therefore, TLS binds RNA in vitro and in vivo (Crozat et al., 1993; Prasad et al., 1994; Zinszner et al., 1997b); it is expressed at high levels in hematopoietic and non-hematopoietic tissues (Aman et al., 1996), and localized primarily in the nucleus, where it may be involved in pre-mRNA processing and nucleocytoplasmic shuttling, as well as in the regulation of basal transcription (Ron, 1997; Zinszner et al., 1997a).
In this study, we show that p42 but not p48 is sumoylated, which is regulated by protein kinase Cδ-mediated phosphorylation. P42 sumoylation provokes its nucleolar translocation. TLS/Fus acts as a Sumo E3 ligase for Ebp1 p42. TLS directly binds to both Ubc9 and Ebp1 and promotes Ebp1 sumoylation. Disruption of Ebp1 sumoylation abolishes its anti-proliferative activity.
Ebp1 p42 but not p48 isoform can be sumoylated
Tumor suppressors including NPM/B23 and p53 are usually regulated by sumoylation. To investigate whether both Ebp1 isoforms, p42 and p48, can be sumoylated, we used stably transfected F293 cells, which express tetracycline (Tet)-inducible His-Sumo1 or His-Sumo3. F293 cells were transfected with green fluorescent protein (GFP) alone, GFP-p42 and p48, and the Tet-induced-His-Sumo modified proteins were pulled down with the Nickel affinity column. Immunoblotting analysis showed that Ebp1 p42 but not p48 was robustly sumoylated by both Sumo1 and Sumo3 (Figure 1a, top panel). Co-transfection with FLAG-Sumo1 also verified that p42 but not p48 was selectively sumoylated (Figure 1b). To examine whether Ebp1 can directly associate with SUMO1, we co-transfected GFP-Sumo1 into HEK293 cells with various glutathione S-transferase (GST)-tagged Ebp1 fragments. GST-pull-down assays showed that p48 did not bind to Sumo1, and deletion of its N-terminal 22 amino acids (a.a.; 23–394) elicited evident association (Figure 1c, lower left panel, lanes 2 and 3), whereas the fragments (a.a. 292–360) and (a.a. 23–136) robustly interacted with GFP-Sumo1 (lanes 1 and 4). Thus, the N-terminal 22 amino acids in p48 inhibit its association with Sumo1. Presumably, it explains why p48 cannot be sumoylated.
Genotoxic stress frequently elicits protein sumoylation (Watts, 2006; Huen and Chen, 2008). To assess whether DNA damage reagents can provoke Ebp1 sumoylation, we used a few DNA damage agents (actinomycin D, adriamycin and VP16), which have been shown to stabilize covalent DNA intermediates of the topoisomerase I (for example, VP16) or to inhibit RNA polymerases (for example, actinomycin D). Compared with DMSO control, 4 h treatment with actinomycin D, adriamycin and VP16 markedly elevated the sumoylation of Ebp1 p42 (Figure 1d, top panel). DNA damage induces phosphorylation of p53 at Ser15 and Ser20 and leads to a reduced interaction between p53 and its negative regulator, the oncoprotein MDM2 (Shieh et al., 1997; Meek, 1998). Immunoblotting analysis showed that these three agents provoked robust p53 phosphorylation (Figure 1d, bottom panel).
Ebp1 binds ErbB3 in breast cancer cells, and translocates into the nucleus on HRG stimulation (Yoo et al., 2000). To examine whether endogenous Ebp1 is also sumoylated in breast cancer cells, we treated MCF7 and AU565 cells with HRG and immunoprecipitated Ebp1 with Ebp1 antibody, which recognizes both p42 and p48. Ebp1, which was upregulated on HRG treatment, was sumoylated in AU565 but not MCF7 cells (Figure 1e, left top panel). As control, sumoylated Ebp1 was selectively immunoprecipitated by Ebp1 antibody but not control IgG (Figure 1e, right panels). To explore whether DNA damage agents also provoke a similar effect, we treated AU565 cells with actinomycin D for various time points. Reverse immunoprecipitation with anti-Sumo1 showed evident endogenous Ebp1 sumoylation by actinomycin but not DMSO when compared with control IgG (Figure 1e, bottom panels). Together, these observations support that Ebp1 p42 is sumoylated on growth factor stimulation or genotoxic stress.
TLS/FUS is a Sumo E3 ligase for Ebp1
To search for Sumo E3 ligases that regulate p42 sumoylation, we transfected mammalian-expressing GST and GST-p42 constructs into His-Sumo1 and His-Sumo3 cells. The transfected proteins were pulled down with glutathione beads, and the co-precipitated proteins were eluted and resolved on SDS–PAGE (polyacrylamide gel electrophoresis), followed by silver staining. Similar protein binding profiles were observed for both His-Sumo1 and His-Sumo3 samples, and prominent protein bands were subjected to proteomic analysis (Figure 2a, left panel). As expected, we identified NPM/B23, which confirmed our previous finding that Ebp1 associates with B23 in the nucleolus (Okada et al., 2007). In addition, we found numerous ribonucleoproteins, verifying a previous report (Squatrito et al., 2004). Among these binding partners, we also identified a zinc-finger containing protein TLS, which is a nucleocytoplasmic shuttling protein that binds RNA and mediates pre-mRNA processing (Ron, 1997; Zinszner et al., 1997a). Nucleolar proteins, including hBRE1 and TLS, were confirmed by immunoblotting to associate with Ebp1 (Figure 2a, right panel). To investigate whether TLS regulates p42 sumoylation in intact cells, we co-transfected FLAG-Sumo1 and TLS and Ubc9 into HEK293 cells with GFP-p42. P42 was prominently sumoylated when either E2 (Ubc9) or E3 (TLS) enzyme was overexpressed. The maximal effect occurred when both E2 and E3 enzymes were highly expressed (Figure 2b), suggesting that TLS can facilitate p42 sumoylation in intact cells. An in vitro sumoylation assay further supported this observation. Both the N-terminus of TLS (a.a. 1–290) and full length of TLS but not its C-terminus (a.a. 357–525) robustly stimulated Ebp1 sumoylation (Figure 2c). To ascertain that TLS is required for p42 sumoylation, we knocked down endogenous TLS with its siRNA. Knocking down of TLS significantly abolished p42 sumoylation (Figure 2d). Together, these results show that TLS acts as an E3 sumo ligase for p42 and stimulates its sumoylation.
TLS/FUS binds p42 Ebp1
To explore whether the substrate Ebp1 p42 interacts with TLS, we co-transfected GST-p42 and GST-p48 into HEK293 cells with FLAG-TLS. GST pull-down assays showed that both isoforms associated with TLS (Figure 3a, left panels). By contrast, GST control failed to bind TLS (Figure 3a, right panels). To assess whether the interaction occurred between the endogenous proteins, we treated K562 cells with DNA damage agent adriamycin (3 μM) and actinomycin D (10 nM) for different time points and immunoprecipitated Ebp1. Immunoblotting showed that TLS bound to Ebp1 in a time-dependent manner. The binding affinity correlated with Ebp1 sumoylation and S360 phosphorylation status (Figure 3b), indicating that DNA damage regulates the association between Ebp1 and TLS, and subsequently mediates Ebp1 sumoylation, for which Ebp1 S360 phosphorylation might be implicated. Moreover, Ebp1 antibody but not control IgG selectively precipitated TLS from K562 cells in the absence of stimulation, underscoring that endogenous TLS and Ebp1 form a tight complex (Figure 3c). We made similar observations in AU565 cells (data not shown). To investigate the role of Ebp1 phosphorylation in its interaction with TLS, we conducted a co-immunoprecipitation assay with p42 S360D, a phosphorylation mimetic mutant, and S360A, an unphosphorylable mutant. As expected, GST-p42 S360D exhibited a stronger binding affinity to TLS than did wild-type p42, whereas p42 S360A failed to associate with TLS (Figure 3d), supporting that p42 S360 phosphorylation is essential for the TLS/Ebp1 complex formation. Interestingly, p42 S360D showed much stronger sumoylation activity than did wild-type p42 in His-sumo1 stably transfected F293 cells. Co-transfection of TLS further enhanced the effect. In contrast, p42 S360A was barely sumoylated regardless of TLS expression (Figure 3e). Thus, these observations support that p42 phosphorylation promotes its association with TLS and augments its sumoylation.
TLS E3 ligase directly binds to Ubc9 and is sumoylated
SUMO E3 ligases, similar to RING domain ubiquitin E3s, do not possess intrinsic enzymatic activity, but act as adapters, which bring together the E2 (Ubc9) and the substrate (Hochstrasser, 2001; Jackson, 2001). To explore whether the TLS E3 ligase indeed interacts with Ubc9, we conducted a GST pull-down assay. The N-terminus of TLS but not its C-terminus interacted with Ubc9 (Figure 4a), agreeing with the in vitro sumoylation results (Figure 4a). TLS also specifically interacted with Ubc9 in the co-transfected HEK293 cells (Figure 4b). To test whether endogenous Ubc9 interacts with TLS, we treated various cells with genotoxic agents and immunoprecipitated Ubc9. Actinomycin D treatment increased the association between Ubc9 and TLS (Figure 4c). We made a similar observation with adriamycin (data not shown). Combined with the finding that Ebp1 also co-precipitated with TLS (Figure 3), these data show that these three proteins (Ebp1/TLS/Ubc9) might form a complex on genotoxic stress. Co-expression of SUMO-1 with FLAG-TLS resulted in the presence of a slower migrating FLAG-reactive band, indicating that TLS may be sumoylated (Figure 4d, left lower panel). Immunoprecipitation with anti-TLS displayed an evidently slower migrating band, suggesting the conjugation of SUMO to TLS (Figure 4d, upper left panel). The co-immunoprecipitation study showed that endogenous SUMO1 also associated with TLS in intact cells (Figure 4d, right panel). Hence, TLS directly binds to the E2 conjugation enzyme Ubc9 and SUMO1, and itself is also sumoylated as well.
Ebp1 K93 and K298 residues are sumoylation residues
SUMOylation takes place on lysine residues that are usually embedded within the core consensus motif Ψ-Lys-X-Glu (where Ψ is a hydrophobic amino acid, and X is any amino acid) (Rodriguez et al., 2001). In exploring the sequence of Ebp1, we noticed that amino acids 64–67, FKKE; 92–95, LKSD; 106–109, VKID; and 297–300, AKHE correspond to a motif that is identified as a consensus sumoylation element present in numerous SUMO substrates. To assess the potential sumoylation sites, we co-transfected GFP-Sumo1 into HEK293 cells with various Myc-Ebp1 constructs with K mutated into R. The co-immunoprecipitation assay showed that p42 but not p48 was potently sumoylated, and that the K93R or K298R mutation disrupted Ebp1 sumoylation (Figure 5a). Hence, K93 and K298 residues are major sumoylation sites on Ebp1 p42. To confirm that these two residues are indeed the major sumoylation sites, we co-transfected FLAG-TLS into F293 cells with various p42 constructs. As expected, wild-type p42 was readily sumoylated, and TLS overexpression markedly increased p42 sumoylation. By contrast, none of the K93R, K298R and (K93,298R) mutated proteins were sumoylated even in the presence of TLS (Figure 5b). These results support that K93 and K298 are the major sumoylation sites on Ebp1, and abolishing one residue impairs the other site sumoylation.
p42 predominantly occurs in the cytoplasm, whereas p48 resides in both the cytoplasm and the nucleolus (Liu et al., 2006). Sumoylation frequently triggers protein nuclear translocation. To explore whether p42 sumoylation is required for its nucleolar translocation, we co-transfected GFP-Sumo1 into HEK293 cells with Myc-p42 wild-type and unsumoylated mutant (K93,298R). Immunofluorescent staining showed that Myc-p42 wild-type and unsumoylated mutant alone mainly resided in the cytoplasm, and that co-transfection with Sumo1 provoked wild-type p42 evident nucleolar translocation, whereas unsumoylated mutants predominantly localized in the cytoplasm (Figures 5c and e, left panels). Co-transfection of GFP-Sumo1 elicits p42 nucleolar residency only in a portion of co-transfected HEK293 cells. Sumo fusion protein system has been used for effective production of native proteins (Lee et al., 2008). It has also been used to study the biological functions of sumoylation substrates. For instance, fusion of Sumo1 to the C-terminus but not N-terminus of c-Fos leads to the dramatic stabilization of c-Fos and reduction of its transcriptional activity (Bossis et al., 2005). We found that GFP-Sumo1-p42 recombinant protein exclusively localized in the nuclei of all transfected cells and displayed different subnuclear patterns. Interestingly, some of the transfected GFP-Sumo1-Ebp1 accumulated in the nucleolus, partially colocalizing with Arf (Figures 5d and f), supporting that sumoylation of Ebp1 triggers its nucleolar translocation. These data obtained with artificial fusion protein must be interpreted with caution, because GFP-Sumo1-p42 is localized into some nuclear aggregates that might be just artifacts, because of sumo interaction with other proteins. The nucleolar residency of Ebp1 p42 wild-type but not unsumoylated mutants was further confirmed by immunostaining with nucleolar markers NPM/B23 and Arf in Myc-p42 and GFP-p42 transfected HEK293 cells (Supplementary Figure 1). Unsumoylated mutants were unable to reside in the nucleolus, underscoring that sumoylation is required for p42 nucleolar localization. Moreover, VP16 also provoked wild-type GFP-p42 nucleolar translocation; by contrast, unphosphorylated mutant GFP-p42 S360A remained in the cytoplasm (Figure 5g), indicating that VP16-provoked phosphorylation in p42 is required for its nucleolar translocation. Thus, sumoylation is sufficient and necessary for p42 nucleolar translocation.
Ebp1 sumoylation is required for its anti-proliferative activity
Previous studies have shown that sumoylation might enhance substrate stability. For instance, sumoylation increases HIF-1α (hypoxia-inducible factor-1α) stability and its transcriptional activity (Bae et al., 2004); sumoylation of Oct4 enhances its stability, DNA binding and transactivation (Wei et al., 2007). To explore whether p42 sumoylation also affects its stability, we co-transfected FLAG-Sumo1 into HEK293 cells with GST-p42 wild-type or unsumoylated mutant (K93,298R). The transfected cells were treated with protein translation inhibitor cycloheximide for various time points. In the absence of exogenous SUMO1, wild-type p42 half-life was <6 h, and co-transfection of SUMO1 substantially increased its stability. By contrast, (K93,298R) was almost completely degraded at 6 h regardless of SUMO1 (Figure 6a, left panels), supporting that sumoylation significantly stabilizes p42. NPM/B23 strongly binds Ebp1 (Okada et al., 2007). Depletion of NPM/B23 significantly decreased its half-life even in the presence of Sumo1 (Figure 6a, right panels), supporting that NPM/B23 might stabilize Ebp1 p42. To test whether sumoylation accelerates growth-suppressive activity of p42, we transfected human cancer cells with various constructs and conducted a 5-bromo-2-deoxyuridine (BrdU) incorporation assay. GFP-p42 overexpression resulted in significant cell proliferation decrease, whereas the cell proliferation rate remained similar for both GFP control and GFP-Sumo1. Co-transfection of both Sumo1 and p42 further suppressed cell proliferation, which was comparable with the repressive effect by the recombinant GFP-Sumo1-p42 protein (Figure 6b), supporting that sumoylation of p42 stimulates its growth-suppressive activity.
Ebp1 overexpression inhibits cancer cell proliferation (Zhang et al., 2005; Yu et al., 2007). Ebp1 has been shown to repress various cell-cycle-related gene transcriptions, including cyclin E1 and E2F-1. To examine the effect of sumoylation on transcriptional suppressive activity of Ebp1, we co-transfected E2F-1 luciferase plasmid into HEK293 cells with Ebp1 p42 and (K93,298R) mutant constructs. Compared with the control vector, ectopic expression of p42 decreased luciferase activity by ∼40%, agreeing with previous observations that overexpression of p42 suppresses cell proliferation. Co-transfection of p42 with either TLS or SUMO1 further enhanced the repressive effect of p42, and the maximal inhibitory effect occurred in TLS, SUMO1 and p42 co-transfected cells. By contrast, transfection of SUMO1 alone or co-transfection of TLS and SUMO1 slightly increased E2F-1 transcription activity (Figure 6c, left panel). Nevertheless, unsumoylated p42 (K93,298R) was unable to suppress the luciferase activity. TLS and SUMO1 lost their stimulatory effects, suggesting that Ebp1 sumoylation is required for its transcription-suppressive activity (Figure 6c, right panel). To further evaluate the effect of sumoylation on the cell growth inhibitory activity of p42, we conducted a colony formation assay with AU565 cells that were stably transfected with wild-type p42 and various unsumoylated mutants. Compared with control, p42 suppressed cancer cell growth. However, K298R mutant failed to inhibit colony formation. Strikingly, p42 K93R and (K93,298R) mutants markedly enhanced cell proliferation (Figure 6d, left panel). Consequently, knocking down of endogenous TLS but not CtBP2, a nuclear transcription co-repressor, selectively increased BrdU incorporation and colony formation (Figure 6d, right panel). These data indicate that abolishing p42 sumoylation somehow facilitates cell proliferation. Together, these results support the notion that sumoylation is required for the anti-proliferative activity of Ebp1 p42.
In this study, we identify that TLS/FUS acts as a SUMO E3 ligase for Ebp1 p42. TLS promotes p42 sumoylation at both K93 and K298 residues. Moreover, p42 sumoylation provokes its nucleolar translocation and is required for its stability and repressive activity on E2F-1 transcription factor. Abolishing p42 sumoylation cripples its anti-proliferative activity.
TLS interacts directly with Ubc9, Sumo1 and its substrate Ebp1 p42, and promotes SUMO1 modification of p42, which fulfill all the criteria applicable for an Ebp1 p42 SUMO E3 ligase. TLS might form a multiple protein complex containing sumoylation machinery. Proteomic analysis identified 498 proteins in the purified nucleolar compartment, and TLS is one of them (Andersen et al., 2005). TLS is localized primarily in the nucleus, where it may be involved in pre-mRNA processing and nucleocytoplasmic shuttling, as well as in the regulation of basal transcription (Ron, 1997; Zinszner et al., 1997a). TLS has been proposed to act as a proto-oncogene through chromosomal translocation. N-terminal part of TLS fuses with CHOP in myxoid liposarcoma carrying the t(12;16) translocation (Crozat et al., 1993; Rabbitts et al., 1993), and in different types of human myeloid leukemia in which the C-terminus of TLS is replaced by the DNA-binding domain of ERG (Shimizu et al., 1993; Panagopoulos et al., 1995). BCR/ABL induces increased expression of TLS by preventing its proteasome-dependent degradation (Perrotti et al., 2000). TLS deficiency in mice results in defective B-lymphocyte development and activation, high levels of chromosomal instability and perinatal death (Hicks et al., 2000). It has been reported that TLS (−/−) mice display male sterility and enhanced radiation sensitivity (Kuroda et al., 2000). All these characteristics resemble another nucleolar protein NPM/B23, which is a ubiquitously expressed nucleolar phosphoprotein that shuttles continuously between the nucleus and the cytoplasm. NPM is translocated or mutated in various lymphomas and leukemias, forming fusion proteins (NPM-ALK, NPM-RARalpha and NPM-MLF1) (Falini et al., 2007). Nonetheless, the role of NPM/B23 in oncogenesis is controversial as it has been attributed to both oncogenic and tumor-suppressive functions. NPM/B23 is essential for embryonic development and the maintenance of genomic stability (Grisendi et al., 2005). Our previous study shows that NPM/B23 binds Ebp1 p42 in the nucleolus, which is regulated by EGF. P42 S360D, a phosphorylation mimetic mutant, strongly binds NPM/B23 (Okada et al., 2007). Interestingly, p42 also strongly binds TLS, which is regulated by DNA damage stress, and Ebp1 phosphorylation increases the interaction (Figure 3). Consistently, immunofluorescent staining with p-Ebp1 S360 antibody shows stronger colocalization by Ebp1 and TLS in the nucleus (Supplementary Figure 2). NPM/B23 is sumoylated on K263 residue, which is essential for NPM/B23 nucleolar residency (Liu et al., 2007). Interestingly, TLS itself is also sumoylated (Figure 4). Conceivably, TLS sumoylation might mediate its nucleolar residency. TLS promotes p42 sumoylation and augments its anti-proliferative activity (Figure 6), indicating that TLS might possess tumor-suppressive activity, although it has been proposed as a proto-oncogene owing to its N-terminal-fused hybrid proteins that possess robust oncogenic effects. Nevertheless, whether wild-type TLS itself possesses any oncogenic activity remains unknown. However, it is worth noting that the knocking down of TLS induces BrdU incorporation (Figure 6d), and this effect might not be exclusively attributed to Ebp1 p42. Nonetheless, p42 suppresses cell proliferation and TLS-promoted sumoylation further enhances its repressive effect. Moreover, the knocking down of TLS diminishes p42 sumoylation. Together, these data indicate that TLS siRNA-induced BrdU incorporation might be, at least in part, regulated by Ebp1 p42.
In this study, we have provided compelling evidence supporting that TLS acts as a physiological E3 sumo ligase for p42. First, TLS binds sumoylation machinery including both SUMO1 and Ubc9 and the substrate p42. Second, overexpression of TLS substantially enhances p42 sumoylation in intact cells and depletion of TLS by siRNA abrogates p42 sumoylation. Third, the in vitro sumoylation assay shows that TLS strongly promotes Ebp1 p42 sumoylation, for which its N-terminus, which binds Ubc9, has an essential role. Hence, our finding establishes that TLS is a new member of the growing list of SUMO E3 ligases.
SUMOylation regulates protein–protein interactions, subcellular localization and stability (Hay, 2001; Muller et al., 2001). Here, we show that p42 Ebp1 sumoylation mediates its association with physiological binding partners. We found that p42 K93R and K298R single mutants decreased their binding affinity to TLS, and that the double mutant K93,298R failed to bind TLS (Supplementary Figure 3). Hence, p42 sumoylation is critical for its association with TLS. Moreover, we also show that sumoylation is essential for its stability. Unsumoylated mutants displayed a shorter half-life than did wild-type proteins (Figure 6a). Many proteins that are important for regulating gene expression, including promoter-specific transcription factors, cofactors and chromatin-modifying enzymes, have been found to be reversibly modified by SUMO. In most cases, SUMOylation of transcriptional regulators correlates with inhibition of transcription (Gill, 2005). Our data support that sumoylation of p42 augments its repressive effect on E2F-1 transcriptional activity, elevating its anti-proliferative activity. On the other hand, abrogating p42 sumoylation cripples its anti-proliferative activity, leading to the upregulation of cancer cell growth and colony formation (Figure 6).
Ebp1 binds NPM/B23 (Okada et al., 2007). Here, we show that Ebp1 p42K298R mutant fails to bind NPM/B23, whereas p42 K93R strongly binds NPM/B23 (Supplementary Figure 4A), suggesting that K298 residue has a critical role in mediating the interaction between p42 and NPM/B23, for which p42 sumoylation might not be essential, because neither K93R nor K298R is sumoylated (Figure 5a). Ebp1 directly interacts with pRB. Overexpression of Ebp1 in MCF-7 and AU565 (Rb(+)) cells inhibits the activity of E2F-1 (Xia et al., 2001). Our previous study shows that wild-type NPM/B23 but not unsumoylated NPM/B23 (K263R) binds Rb. Sumoylated NPM/B23 exhibits robust affinity to pRb (Liu et al., 2007). We found that wild-type p42 upregulates NPM/B23 binding to pRb; in contrast, p42 K298R mutant blocks NPM/B23 binding to pRb (Supplementary Figure 4B), indicating that p42/B23 interaction is indispensable for NPM/B23 interaction with pRb. It has been reported before that DNA-damaging agents, actinomycin D and ultraviolet radiation, induce the dephosphorylation of pRB at Cdk phosphorylation sites and upregulate the binding of pRB to E2F-1, resulting in repression of E2F-1-mediated transcription (Inoue et al., 2007). It can be noted that, genotoxic stress strongly upregulates p42 sumoylation (Figure 1d). Conceivably, DNA damage activates protein kinase Cδ, which phosphorylates p42 and provokes its sumoylation through TLS. The sumoylated p42, in turn, associates with pRB, blocking E2F-1 transcription activity. Collectively, our finding that TLS promotes Ebp1 sumoylation provides a molecular mechanism for post-translational modification mediating Ebp1 transcriptional repressive activity.
Materials and methods
Cells and reagents
His-Sumo stably transfected F293 cells were grown in complete medium containing 1 × Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) containing 10 μg/ml of blasticidin and 100 μg/ml of hygromycin in a humidified incubator at 37 °C, 5% CO2. Expression of the His-SUMO proteins was induced by adding Tet to the culture medium at a final concentration of 1 μg/ml (a generous gift from Dr Van G Wilson, Texas A&M University, College Station, TX, USA). Anti-FLAG monoclonal antibody (M2), anti-hemagglutinin-horseradish peroxide (HA-HRP), anti-Myc and anti-GST-HRP monoclonal antibodies were purchased from Sigma (St Louis, MO, USA); anti-GFP monoclonal antibody was purchased from Roche (Basel, Switzerland). Anti-Ebp1 antibody was purchased from Millipore (Billerica, MA, USA). Anti-Ebp1 S360 has been previously described (Ahn et al., 2006). GFP-Sumo1-Ebp1 construct was prepared by three-step PCR amplification: (1) Sumo 1 was amplified using the following two primers: (a) HindIII Sumo1N 5′-ACTCAAGCTTCATG TCTGACCAGGAGGCAAAACC-3′ and (b) Sumo1Ebp1 (R) 5′-CTGTTTCTTCCATAATCATAACTGTTGAATGACCCCC-3′. (2) Ebp1 was amplified using the following two primers: (c) Sumo1Ebp1 (F) 5′-GGGGGTCATTCAACAGTTATGATTATGGAAGAAACAG-3′ and (d) Kpn1Ebp1 5′-ACTCGGTACCTCAGTCCCCAGCTTCATTTTC-3′. (3) The PCR product from the above two PCRs were mixed and amplified with primers a and d. The resulting fused PCR product was subcloned into p-EGFPC2 vector between HindIII/KpnI sites. All chemicals were purchased from Sigma.
Preparation and transfection of siRNA
The siRNA of TLS was purchased from Dharmacon (Lafayette, CO, USA) (catalog no. M180733-00); The siRNAs were transfected using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). The sequence for shRNA of NPM/B23 is 5′-GATCCCGAGGAAGATGCAGAGTCAGAAGATGAAGAGAAGCTTGTCTTCATCTTCTGACTCTGCATCTTCCTCTTTTTT-3′. The shRNA subcloned into pGE-1 between BamHI and XbaI, which was then cleaved by XhoI/XbaI and ligated into pAdTrack-CMV for preparing adenovirus.
Co-immunoprecipitation and Nickel column pull-down
Dishes (10 cm) of HEK293 cells were co-transfected with 5 μg of various Myc-p42 constructs and GFP-Sumo1 by Nova Factor. In 24 h, the transfected cells were washed once in phosphate-buffered saline (PBS), lysed in 1 ml lysis buffer (50 mM Tris, pH 7.4, 40 mM NaCl, 1 mM EDTA, 0.5% Triton X-100, 1.5 mM Na3VO4, 50 mM NaF, 10 mM sodium pyrophosphate, 10 mM sodium β-glycerophosphate, 1 mM phenylmethylsulfonyl fluoride, 5 μg/ml aprotinin, 1 μg/ml leupeptin and 1 μg/ml pepstatin A) and centrifuged for 10 min at 14 000 g at 4 °C. After normalizing the protein concentration, 2 μl anti-myc antibody and 40 μl of 50% slurry protein A/G agarose were added to the supernatant and incubated with rotation for 2 h at 4 °C. The agarose pellet was washed three times with 1000 μl lysis buffer each time. The agarose was then resuspended in 30 μl sample buffer separated by SDS–PAGE followed by immunoblotting using anti-GFP antibody. F293 cells were transfected with Myc-42 and induced with Tet for 24 h. His-tagged proteins were pulled down with nickel column. After extensive washing with the above lysis buffer without EDTA, the co-precipitated proteins were analysed by immunoblotting with anti-Myc antibody.
Recombinant protein purification and in vitro SUMOylation assay
Recombinant GST-p42 and His-SUMO1 proteins were expressed in Escherichia coli (BL21). GST-p42 proteins were purified with glutathione-sepharose beads (GE Healthcare, Piscataway, NJ, USA), and His-Sumo1 was purified with Ni2+ column and dialysed against PBS containing 10% glycerol overnight at 4 °C. AOS1 and Ubc9 were from BIOMOL (Plymouth Meeting, PA, USA). All the purified recombinant proteins (TLS and its fragments from mammalian HEK293 cells) were dialysed against PBS containing 10% glycerol. An in vitro SUMOylation analysis of GST-p42 was carried out with the SUMOylation kit (BIOMOL).
Cells were fixed with either methanol at −20 °C for 20 min or 10% formalin at room temperature for 20 min. The fixed cells were blocked by 10% normal goat serum in PBS for 1 h, and incubated with primary antibodies for 1 h. Cells were then incubated with secondary antibodies (Alexa Fluor 594-tagged goat anti-mouse IgG, Alexa Fluor 488-tagged goat anti-rabbit IgG or Alexa Fluor 594-tagged goat anti-mouse IgG antibodies, Invitrogen) for 1 h and counterstained for DNA with 4′,6-diamidino-2-phenylindole. After incubation with antibodies, cells were washed extensively in PBS. Cells were examined under a fluorescence microscope.
The E2F-1 promoter-luciferase reporter construct contains a luciferase reporter gene. Cells (1 × 105) in six-well plates were transfected with 0.25 μg of reporter plasmid with various Ebp1 constructs at the indicated concentrations. Renilla thymidine kinase was included as an internal control. After transfection, the cells were placed in complete medium for 48 h. Control or treated cells were harvested to analyse luciferase activity as previously described (Xia et al., 2001).
Colony inhibition assays
AU565 cells (1 × 104) were seeded into individual wells of six-well plates. Ebp1 plasmid DNA (2 μg) was stably transfected using Fugene 6 Roche. The number of colonies surviving after 3 weeks of G418 (500 μg/ml) selection was determined microscopically.
BrdU incorporation assay
Cells were seeded into six-well plates at 1 × 105 cells per well, cultured overnight and transfected with various siRNA. The transfected cells were pulse labeled with BrdU (10 μm; GE Healthcare). The incorporation was continued for 10 min. After washing with cold PBS, the cells were fixed with 4% paraformaldehyde for 15 min and treated with 2 M HCl at 37 °C for 30 min. The coverslips were blocked with 2% FBS/PBS and 0.4% Triton X-100 for 10 min at room temperature. For immunostaining, mouse monoclonal antibody against BrdU (1:200) and anti-GST-FITC-conjugated antibody were used.
Conflict of interest
The authors declare no conflict of interest.
This work is supported by grants from the National Institute of Health (RO1, CA127119) to K Ye. We thank Dr Yang Liu at Medical Research Service (VA Puget Sound Health Care System, Seattle, WA, USA) for TLS constructs.
About this article
Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)