| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J. Biol. Chem., Vol. 280, Issue 42, 35477-35489, October 21, 2005
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
¶||1
From the
Molecular Oncology Group, Departments of Biochemistry, ¶Medicine, and ||Oncology, McGill University, Montréal, Québec H3A 1A1, Canada
Received for publication, April 25, 2005 , and in revised form, July 25, 2005.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
-responsive reporter genes and comparable activation on vitamin D3 receptor transcription. Coexpression of SIP1 with Pc2 can partially relieve E-cadherin repression by SIP1. We further show that SIP1 sumoylation disrupts the recruitment of CtBP. Thus SIP1 sumoylation regulates its transcriptional activity in a promoter context-dependent manner and may represent an important intervention target to modulate EMT in tumorigenesis. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
(TGF-), matrix-metalloproteinases, and extracellular matrix components, have been implicated in the regulation of EMT and tumor progression, but the cross-talk among these pathways and their downstream targets remains largely unknown (6, 7).
One hallmark of EMT, loss of E-cadherin, is induced by various epigenetic mechanisms, including promoter hypermethylation as well as active transcriptional repression by several transcription factors (6), including Snail (8, 9) and its homologue Slug (10), Smad-interacting protein 1 (SIP1, also called ZEB2, for zinc-finger E-box-binding protein 2) (11) and its homologue 
EF1 (also called ZEB1) (12), twist (13), and the basic helix-loop-helix factor E12/E47 (14) (reviewed in Ref. 3).
SIP1 was originally identified from a yeast two-hybrid screen through its binding to Smad (15). SIP1 plays a crucial role in normal embryonic development of neural structures and the neural crest (16). SIP1 is a member of the zfh-1 family of two-handed zinc-finger transcription factors, which also includes EF1 (17, 18). They share the unique structure of two zinc-finger clusters separated by a homeodomain, where each of the zinc-finger clusters binds to an E-box element in the promoter region of target genes (17, 18). In addition to Smad-binding domains, zfh-1 family members also contain consensus binding motifs for the corepressor CtBP (19, 20). SIP1 and EF1 were both shown to be present in a CtBP corepressor core complex, which binds to the E-cadherin promoter and represses its transcription (21). However, it seems that SIP1 might also be able to repress E-cadherin independent of CtBP binding (22). Besides E-cadherin, SIP1 and EF1 have been shown to repress the transcription of many other genes depending on the cell types (23). This includes suppression of interleukin 2, immunoglobulin µ heavy chain, CD4, GATA-3, and 
4 integrin in hematopoietic cells; inhibition of p73 gene expression in mesenchymal cells, and repression of type I and type II collagen expression in osteoblasts. EF1 can also activate the vitamin D3 receptor (VD3R) gene (24) and ovalbumin (25), although the detailed mechanism is unknown. Recently, SIP1 and EF1 were shown to exert opposing effects on TGF-/Smad signaling through the recruitment of coactivators p300 and/or P/CAF and displacement of CtBP by EF1 (23, 26).
Smad is the key signal transducer of TGF-, which plays a pivotal role in multiple cellular processes, including proliferation, differentiation, adhesion, apoptosis, and importantly tumorigenesis (27). Phosphorylation of Smad2 and Smad3 by the ligand-activated transmembrane serine/threonine kinase receptors results in heteromeric complex formation with Smad4, and translocation into the nucleus, where transcription of target genes is regulated through direct DNA binding and/or the recruitment of transcriptional coactivators or corepressors (28–30).
Sumoylation is a covalent modification that adds small ubiquitin-like modifier (SUMO) to lysine residues. It occurs at a consensus motif 
KX(D/E), where is a hydrophobic amino acid (31, 32). This modification involves the coordinated action of multiple enzymes, in a manner very similar to ubiquitination. These enzymes include E1 SUMO-activating enzyme (heterodimer of SAE1/SAE2), E2-conjugating enzyme Ubc9, and an E3 ligase, which promotes the transfer of SUMO from Ubc9 to specific proteins (31, 32). The protein inhibitor of activated STAT (PIAS) family proteins (33–35), nucleoporin protein RanBP2 (36), and polycomb protein Pc2 (37) have been identified as SUMO E3 ligases. Sumoylation is negatively regulated by multiple proteases, which remove SUMO from its substrates (38). Unlike ubiquitination, which signals proteins for degradation by the proteasome or enhances trafficking of transmembrane proteins for degradation in the lysosome, protein sumoylation has been shown to regulate a variety of activities, including protein stability, subcellular localization, transcriptional regulation, DNA repair, as well as genome integrity (31, 32, 39, 40).
In this report, we show that SIP1 and its homologue EF1 are sumoylated. We have mapped the sumoylation sites to two conserved lysine residues, Lys391 and Lys866, in SIP1, and show that polycomb protein Pc2, acts as a SUMO E3 ligase for SIP1. Importantly, a sumoylation negative mutant of SIP1 shows more potent repression toward E-cadherin. We further show that Pc2-mediated sumoylation can partially relieve the E-cadherin repression by SIP1. Our results provide an example of how sumoylation regulates transcription in a promoter context-dependent manner.
|
EXPERIMENTAL PROCEDURES |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
EF1 cDNA was assembled through PCR from two overlapping expressed sequence tag clones (National Institutes of Health Image clones 4245215 and 6180372, Open Biosystems). Myc (six copies), HA, FLAG, GFP, GAL4-tagged SIP1, and EF1 were constructed by subcloning the cDNAs into pCS3–6Myc, pcDNA3.1-HA, pCMV5-FLAG, pEGFP-C2 (BD Biosciences), and pSG424, respectively. YFP-SIP1 was constructed by first replacing SUMO1 coding region in YFP-SUMO1 (41) and then transferring the YFP-SIP1 fusion region into pCMV5. Myc-tagged SIP1 deletion mutants were constructed by restriction enzymes digestion. SIP1 domains were subcloned into pGEX-4T-1 (Amersham Biosciences) to generate GST fusion proteins for use in GST pull-down assays. One or more different SIP1 and EF1 sumoylation sites mutants, as well as the CtBP-binding mutant of SIP1 (where four consensus motifs were simultaneously mutated as previously (22)), were made by using the QuikChange System (Stratagene) or PCR-based mutagenesis. All constructs and mutants were confirmed by sequencing. Myc-Smad4 and Myc-Smad4 (K113/159R), FLAG-Smad2 and FLAG-Smad3, FLAG-SUMO1 and 2FLAG-SUMO1 (34), Myc-SUMO1, and FLAG-PIAS3 and FLAG-PIASy (42) were as previously described (43, 44). FLAG-Pc2 and CFP-Pc2 (37), and FLAG-HDAC4 (45), were as described. FLAG-RanBP2 was constructed by PCR from BP2 
FG (36). FLAG-CtBP was kindly provided by J. White (McGill University). Antibodies, Cell Lines, Transfection, and Immunoprecipitation— HDAC4 antisera (46) were from X-J. Yang. Other antibodies used are: anti-Myc (BD Biosciences), anti-FLAG and anti-His (Sigma), anti-HA (16B12) (BAbCo), anti-GFP (Molecular Probes), anti-SUMO1 (GMP-1) (Zymed Laboratories Inc.), anti-GAL4 (RK5C1), anti-actin (I-19), anti-Mel-18 (C-20), anti-CtBP (H-440) (Santa Cruz Biotechnology).
HeLa, HEK293T, COS-1, Mv1Lu/L17 (mink-lung epithelial cells), and Madin-Darby canine kidney (MDCK) cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Invitrogen). Transient transfections were performed using Lipofectamine (Invitrogen). Immunoprecipitations were usually carried out as previously described (44) from cells lysed in TNTE buffer (10 mM Tris HCl, pH 7.8, 150 mM NaCl, 1 mM EDTA, 1.0% Nonidet P-40) plus protease inhibitors and phosphatase inhibitors. FLAG-SIP1-bound proteins were purified from FLAG-SIP1-transfected 293T cells by anti-FLAG-M2 affinity gel (Sigma) and eluted with 150 µg/ml 3xFLAG peptide (Sigma).
In Vivo Sumoylation Assay—In vivo sumoylation assay of SIP1 was carried out the same as previously described (43). Briefly, COS-1 cells were transiently transfected and lysed in modified radioimmune precipitation assay/SDS buffer (43, 47). The lysis buffer is a 3:1 mixture of radioimmune precipitation assay buffer (25 mM Tris-HCl, pH 8.2, 50 mM NaCl, 0.5% sodium deoxycholate, 0.5% Nonidet P-40, 0.1% SDS, 0.1% sodium azide) and SDS sample buffer (5% SDS, 0.15 M Tris-HCl, pH 6.8, 30% glycerol), plus 10 mM N-ethylmaleimide, protease inhibitors, and phosphatase inhibitors. Cells were then sonicated briefly and boiled for 10 min. The clear lysates from centrifugation were separated on SDS-PAGE and immunoblotted against appropriate primary antibody (anti-Myc or anti-HA). To confirm the in vivo sumoylation, lysates were diluted 10-fold into phosphate-buffered saline/0.5% Nonidet P-40 and immunoprecipitated with anti-Myc antibody followed by immunoblot with anti-FLAG antibody.
In Vitro Sumoylation Assay—The 35S-labeled, Myc-tagged wild-type, and mutant SIP1 were synthesized by in vitro translation using the SP6 TNT Quick Coupled Transcription/Translation System (Promega). Purification of GST-SUMO1 (48), Ubc9, E1, and in vitro sumoylation reactions were carried out as previously described (43). Alternatively, 35S-labeled Myc-SIP1 was incubated with an in vitro sumoylation control kit (LAE Biotech, Rockville, MD), in the absence or presence of increasing amount of bacterially purified His-Pc2 protein (AmProx, Carlsbad, CA). Reaction products were separated on 4–10% gradient gel and analyzed by autoradiography.
GST Pull-down Assay—N-terminal and C-terminal zinc-finger clusters (N-terminal, 90–383; C-terminal, 957–1156), and the middle region (384–956) were expressed as GST fusion protein in Escherichia coli together with empty vector. GST proteins were purified from glutathione-Sepharose 4B (Amersham Biosciences). Beads with 1 µg of GST proteins were mixed with 1 µg of His-Pc2 protein (AmProx) in phosphate-buffered saline/0.5% Nonidet P-40 buffer and subjected to the pull-down assay (44). SIP1-bound Pc2 were detected by anti-His antibody.
Luciferase Reporter Gene Assay—293T, Mv1Lu/L17, and COS-1 cells in 12 well plates were transfected by Lipofectamine. Human E-cadherin-luc (-1008/+49) (49) was used to measure transcriptional repression of E-cadherin by SIP1 constructs in 293T cells. TGF--responsive 3TP-lux and SBE4-luc were used to detect repression of TGF- signaling by SIP1 in Mv1Lu and L17 cells, where cells were treated with or without 250 pM TGF- for 18–24 h as previously described (44). L8G5-luc and LexA-VP16 (50) were used to measure the transcriptional repression activity of GAL4-SIP1 in 293T cells as described before (50). pGL3-VDR-luc (24) was used to compare Myc-SIP1 and its sumoylation mutant together with Myc-EF1, for their transactivation ability on VD3R in COS-1 cells. Luciferase activities were normalized with cotransfected -galactosidase control driven by pSV--gal (Promega). Data presented are means ± S.D. from at least three independent experiments.
Confocal Immunofluorescence Microscopy—COS-1, HeLa, 293T, or MDCK cells grown on coverslips in 24-well plates were transfected with plasmids to be tested by Lipofectamine. 40 h post transfection, cells were fixed in 2% paraformaldehyde and analyzed by direct or indirect immunofluorescence microscopy in a LSM510 confocal system (Zeiss).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
EF1 sequences contain multiple KX(D/E) consensus motifs, so we asked whether SIP1 and EF1 can be modified by sumoylation. To address this, we first generated mammalian expression constructs for SIP1 and EF1. As previously observed for EF1 (51), both SIP1 and EF1 had apparent molecular masses higher than calculated (170 kDa for Myc-SIP1 and 190 kDa for Myc-EF1, Fig. 1A, lanes 1 and 3). To test for in vivo sumoylation, COS-1 cells were transfected with Myc-SIP1 and Myc-EF1, together with FLAG-SUMO1 (Fig. 1A). As a control for sumoylation, we included Smad4, a key mediator of TGF- signaling, which is modified by sumoylation at Lys-159 and Lys-113 (Fig. 1A, lanes 6 and 7) (43, 52–54). Cotransfection of FLAG-SUMO1 with Myc-SIP1 or Myc-EF1 led to slower migrating bands of 230 and 220 kDa, respectively (Fig. 1A, lanes 1–2 and 4–5). The slower migrating band of SIP1 was further "shifted" when a 2FLAG-SUMO1 construct was transfected (lane 3), confirming that the 230-kDa band of SIP1 is a SUMO conjugate. These results suggest that SIP1, as well as its homologue EF1, is subjected to sumoylation, the size difference being consistent with the addition of two or three SUMO molecules. SIP1 Sumoylation Sites Localize to the Repression Domain—To systematically map the sumoylation sites in SIP1, we generated several deletion mutants and tested their in vivo sumoylation to delineate domains of SIP1 required for sumoylation. SIP1 is a member of the two-handed zinc-finger transcription factor family, and has N- and C-terminal zinc-finger clusters, N-ZF and C-ZF, separated by a central homeodomain (HD), a Smad-binding domain, and CtBP-interacting domains (CIDs) (see Fig. 1B for details). Deletion of the N-terminal 303 amino acids failed to affect sumoylation (Fig. 1B, dm1, lane 3), whereas a protein with an N-terminal deletion of 413 residues showed a significant decrease in sumoylation (dm2, lane 4), suggesting that the region between N-ZF and the Smad-binding domain (303–413) regulates or contains the major sumoylation sites.
Interestingly, we found that the N-terminal 266 amino acids are necessary but not sufficient for efficient sumoylation of deletion mutants. This domain can promote the sumoylation of dm3, which alone showed very little sumoylation potential (lanes 6 versus 5). However the N-terminal 266 amino acids failed to rescue the modification of the C-ZF mutant (975–1214, dm7, lane 8). We speculate that the N-terminal 266 residues may serve as an adaptor to recruit components of the sumoylation machinery (E2-conjugating enzyme Ubc9 and/or E3 ligase) to dm3, which harbors sumoylation site(s). The different sumoylation potential of dm6 and dm7 argues that the region between 810 and 974 is responsible for the sumoylation of dm6. Thus, we have mapped the minimal sumoylation domain of SIP1 to 304–975, which is almost identical to the previously identified "repression domain" (335–998) of SIP1 (18). This result is consistent with the observation that sumoylation of most transcription factors lies within their repression domains (31, 40, 55).
Identification of SIP1 Sumoylation Sites—The minimal sumoylation domain (304–975) contains three consensus sumoylation motifs: Ile-Lys391-Thr-Glu, Ile-Lys555-Lys-Glu, and Ile-Lys866-Lys-Glu (Fig. 2A, top panel). Single mutants of each of these sites where lysine was substituted with an arginine residue were generated from Myc-SIP1, and their in vivo sumoylation was tested in COS-1 cells. Whereas the sumoylation of Myc-SIP1 (K391R) is significantly decreased when compared with wild-type, sumoylation of K555R or K866R is not affected (Fig. 2A, bottom panel, lanes 8-10). These results indicate that Lys391 is a major sumoylation site for SIP1, and other site(s) contribute to the residual sumoylation detected in the K391R mutant (lane 8). A double mutant of K391R/K555R was still modified by sumoylation, albeit inefficiently (data not shown), suggesting that Lys555 of SIP1 is not a sumoylation site.
Although most of the reported sumoylation target proteins are modified at a consensus KX(D/E) motif, there are several exceptions (32). To identify additional sumoylation site(s) in SIP1, we looked at deletion mutant dm5, which does not contain Lys391 but is sumoylated (Fig. 1B, lane 6). In addition to Lys866, which is a consensus site, we tested two nonconsensus sumoylation motifs: Val-Lys136-Asn-Ala, which is similar to the sumoylation motif Ala-Lys1182-Val-Asn in the homeodomain-interacting protein kinase 2 (56), and Val-Lys734-Pro-Met, which has an aligned sequence in EF1 (Ala-Lys657-Asn-Asn), also similar to Ala-Lys1182-Val-Asn in homeodomain-interacting protein kinase 2. Lysine to arginine single substitution was introduced into each possible sumoylation site, and their sumoylation was tested. Only the mutant K866R but not K136R or K734R, abolished the sumoylation of dm5 (Fig. 2A, bottom panel, lanes 2–5). The introduction of the K866R substitution into another deletion mutant dm6 also abolished its sumoylation (data not shown). This identifies Lys866 as a second sumoylation site for SIP1, required for the observed sumoylation of dm6. This is consistent with the absence of sumoylation of dm7, which lacks this site (Fig. 1B, lanes 7 and 8).
A SIP1 double mutant, where the two sumoylation sites Lys391 and Lys866 were simultaneously substituted with arginine, was constructed, and its in vivo sumoylation was tested together with each single mutant (Fig. 2B, top panel). Indeed, when compared with wild-type SIP1 or its single mutant, Myc-SIP1 (K391R/K866R) did not show any detectable slow migrating band (lane 6), consistent with Lys391 and Lys866 being sumoylated in wild-type Myc-SIP1. To confirm that SIP1 was sumoylated at these two residues, proteins were immunoprecipitated with anti-Myc antibody and then immunoblotted with anti-FLAG antibody (Fig. 2B, bottom panel). As expected, the 230-kDa slow migrating band of Myc-SIP1 can only be detected when FLAG-SUMO1 was cotransfected, confirming that the band is indeed a FLAG-SUMO1-modified form of Myc-SIP1 (lane 3 versus lane 2). Although a single mutation did not prevent the production of the slow migrating SUMO species, the K391R/K866R double mutant completely abolished any detectable FLAG-SUMO1 conjugates. These data strongly support that in mammalian cells SIP1 can be modified by sumoylation at Lys391 and Lys866.
We noticed that the sumoylated form of K391R mutant runs faster than that of K866R (Fig. 2B, top panel, lanes 4 and 5). Similar migration aberrance was previously reported for Smad4 (43, 53) and BKLF (57), a basic Krüppel-like transcription repressor. It was shown that Lys197, one of the sumoylation sites for BKLF, lies in a critical domain in the protein where sumoylation or mutation significantly affects the mobility of the protein on SDS-PAGE (57). It would be interesting to see whether Lys866 of SIP1 also lies in a similar critical microenvironment.
We also tested whether SIP1 can be sumoylated at Lys391 and Lys866 in a purified in vitro sumoylation system (43). The 35S-labeled, Myc-tagged SIP1 and its sumoylation variants were synthesized by in vitro translation. Addition of GST-SUMO1, E1-activating enzyme, and Ubc9 led to the sumoylation of Myc-SIP1 (Fig. 2C). Multiple bands of sumoylated SIP1 were detected (lane 2), which could represent SIP1 conjugated with a multi-SUMO1 chain. Although conjugation of a multi-SUMO chain is generally believed to happen on SUMO2 or SUMO3 rather than SUMO1 (48), similar polymerization of SUMO1 has been reported in vitro for yeast septins cdc3, cdc11 (33), CtBP (37), and in vivo for Smad4 (52). Under these experimental conditions, Myc-SIP1 (K866R) displayed a similar sumoylation pattern to the wild-type protein, whereas K391R only showed weak sumoylation (lanes 6 and 4). Simultaneous substitution of both lysines to K391R/K866R, completely abolished any slow migrating SIP1 species (lane 8). Thus, in agreement with the in vivo data, SIP1 is also sumoylated at Lys391 and Lys866 in vitro.
|
|
EF1 comprise the zfh-1 protein family. Alignment of the two sumoylation sites of human SIP1 with the same region from human EF1 and their counterparts from other organisms revealed that the positions of the two sumoylated lysines are highly conserved (Fig. 2D, top panel). Interestingly, the second sumoylation site (Lys866 in SIP1) is adjacent to a consensus CtBP-binding motif. From the alignment, two sumoylation consensus motifs, Ile-Lys347-Thr-Glu, Ala-Lys774-Lys-Glu were found in EF1. To test if these sites are sumoylated in EF1, a double mutant HA-EF1 (K347R/K774R) was constructed and tested for sumoylation following transient transfection in COS-1 cells (Fig. 2D, bottom panel). The slow migrating EF1 bands observed for the wild-type protein were completely abolished in the double mutant (lanes 3 and 4), suggesting that EF1 is sumoylated at Lys347 and Lys774. Taken together, our results indicate that SIP1 is sumoylated at two conserved sites, Lys391 and Lys866. Pc2, but Not PIAS Family Proteins, Acts as a SUMO E3 Ligase for SIP1—Although in vitro sumoylation of proteins can take place when only E1 and E2 proteins are included, SUMO E3 ligase can enhance the modification in vitro and in vivo (31, 32). The PIAS family proteins (33–35), nucleoporin RanBP2 (36), and polycomb protein Pc2 (37), have been identified as SUMO E3 ligases. Histone deacetylase 4 (HDAC4) was recently shown to stimulate the sumoylation of MEF2 transcription factors (45).
To identify which of the reported SUMO E3 ligases could enhance SIP1 sumoylation, HA-SIP1 was tested for sumoylation in the absence or presence of different FLAG-tagged SUMO E3 ligases (Fig. 3A). In the absence of cotransfected E3, HA-SIP1 is weakly sumoylated (lane 2). Cotransfection with FLAG-PIAS3, FLAG-PIASy, and FLAG-HDAC4 did not enhance SIP1 sumoylation (lanes 6–8), even when expressed at high levels (data not shown). Nucleoporin protein RanBP2 also failed to promote SIP1 sumoylation (data not shown). In contrast, polycomb protein Pc2 enhances SIP1 sumoylation in a dose-dependent manner (lanes 3–5), suggesting that Pc2 can act as a SUMO E3 ligase for SIP1.
To establish if Pc2 can act as an E3 ligase for SIP1, we examined the ability of Pc2 to promote SIP1 sumoylation in vitro. For this, bacterially purified His-Pc2 was included in the Myc-SIP1 in vitro sumoylation reaction (Fig. 3B). Under these conditions, multiple sumoylated bands of SIP1 were easily detected (lanes 1 and 2). Polycomb protein Pc2 promotes a modest enhancement of the majority of SIP1 sumoylated species in a dose-dependent manner, with the exception of a high molecular weight band (lanes 2–6). Thus, purified Pc2 shows E3 ligase activity toward SIP1 sumoylation in vitro.
A SUMO E3 ligase must recognize and bind the target protein to provide substrate selectivity. SIP1 and Pc2 were recently shown to be present in a CtBP corepressor complex by tandem affinity purification tag purification (21), but so far there is no evidence for their direct interaction. We first tested their possible interaction following transient overexpression. FLAG-Pc2 was cotransfected with a series of Myc-SIP1 deletion mutants (Fig. 1B, left panel) into 293T cells. Proteins from cell lysate were immunoprecipitated with anti-Myc antibody followed by immunoblot using anti-FLAG antibody (Fig. 3C). Specific binding of Pc2 to SIP1 was readily detected when FLAG-Pc2 was coexpressed with Myc-SIP1 (lane 1). All deletion mutants except dm9 (lane 2), which contains only the C-ZF domain, were able to interact with Pc2. Notably, both the N- and C-terminal regions of SIP1 were important in the binding to Pc2, because two deletion mutants, dm8-(1–303) and dm4-(810–1214), were able to bind Pc2 (lanes 8 and 5). The interaction between Pc2 and SIP1 deletion mutants, with the exception of dm9, were also successfully detected in a reciprocal coimmunoprecipitation assay (data not shown).
Because dm8-(1–303) and dm4-(810–1214) overlap the SIP1 zinc-finger clusters, N-ZF-(90–383) and C-ZF-(957–1156), we speculated that Pc2 might bind to SIP1 through the zinc-finger clusters. To test this hypothesis, we subcloned the SIP1 N-ZF, C-ZF, and middle repression domain and examined their direct binding to Pc2 in a GST pull-down assay. GST or GST-SIP1 domain proteins were incubated with bacterially purified His-Pc2 protein, and SIP1-bound Pc2 was detected by immunoblot with anti-His antibody (Fig. 3D, top panel). As expected, a GST fusion of SIP1, N-ZF, and C-ZF, but not the middle repression domain, can significantly bind to Pc2 (lanes 1–4). Similar amounts of GST and GST-SIP1 domain proteins were used (bottom panel). Thus, N- and C-terminal zinc-finger clusters of SIP1 are each sufficient for direct binding to Pc2.
Pc2 has been shown to recruit Ubc9 and CtBP into characteristic subnuclear domains, PcG bodies (37, 58, 59). To establish whether Pc2 and SIP1 colocalize, their subcellular localization was examined by immunofluorescence microscopy. First, YFP-SIP1 and CFP-Pc2 were transfected separately, and their subcellular localization was visualized by direct immunofluorescence. CFP-Pc2 localized to a nuclear body structure, characteristic of PcG bodies (37) (Fig. 3E, top panel). YFP-SIP1 was also found exclusively in the nucleus (Fig. 3C, top panel). Interestingly, YFP-SIP1 showed homogenous dispersed staining in about half of the transfected cells (58%) and punctate nuclear staining in the other half of cells (42%). The SIP1-positive punctate structure did not colocalize with Mel-18, another component of PcG bodies (middle panel). However, when YFP-SIP1 and CFP-Pc2 were cotransfected, a significant change in SIP1 nuclear staining was observed. All YFP-SIP1 colocalized with CFP-Pc2 (Fig. 3E, bottom panel), suggesting that, under these conditions, YFP-SIP1 is recruited to PcG bodies by Pc2. A similar colocalization of SIP1 and Pc2 into PcG bodies was observed in cells cotransfected with GFP-SIP1 and FLAG-Pc2, and in other cell lines tested, including COS-1, HeLa, MDCK, and 293T (data not shown). In contrast, SIP1 nuclear staining was not altered upon coexpression of PIAS1, PIAS3, PIASy, or RanBP2.3
From the findings that Pc2 can bind directly to SIP1 (Fig. 3D), as well as recruit the E2 enzyme Ubc9 (37, 59) and SIP1 (Fig. 3E) into PcG bodies, and promote the SIP1 sumoylation both in vivo and in vitro (Fig. 3, A and B), we conclude that Pc2 can act as a SUMO E3 ligase for SIP1. Hence, SIP1 is the second identified substrate for Pc2.
Sumoylation Does Not Affect Nuclear Localization of SIP1—Protein sumoylation has been shown to regulate a variety of events from protein stability, subcellular transport, transcriptional regulation, and DNA repair to maintenance of genome integrity (31, 32). To explore the functional significance of SIP1 sumoylation, we first compared the stability of SIP1 and sumoylation mutant by pulse-chase assay, but failed to observe any significant difference in protein half-life (data not shown).
To test the effects of SIP1 sumoylation on its subnuclear localization, COS-1 cells were transfected with YFP-SIP1 (K391R/K866R), and their nuclear staining was visualized by direct immunofluorescence (Fig. 4A). In a similar manner to wild-type (Fig. 3E), the SIP1 sumoylation mutant showed a distribution into two nuclear pools, homogenous nuclear staining (57%) and a punctate nuclear staining (43%). Hence, SIP1 sumoylation does not affect its nuclear distribution or nuclear structure.
We further examined the effect of Pc2 on localization of the SIP1 sumoylation mutant. As shown in Fig. 4B, cotransfection of CFP-Pc2 into COS-1 cells drastically changed the nuclear localization of YFP-SIP1 (K391R/K866R), where it colocalizes with Pc2 into the subnuclear domain of PcG bodies. Thus Pc2 recruitment of SIP1 is sumoylation-independent.
Sumoylation Selectively Regulates Transcriptional Activities of SIP1— SIP1 is a transcription factor that binds to the E-box of target genes to repress their transcription (11, 15, 17, 18). Tethering the repression domain of SIP1 to a LexA-DNA binding domain was shown to repress the transcription of multiple coactivators, including TFE3, MEF2C, and c-Myc (18). Importantly, SIP1 can repress transcription of E-cadherin and some TGF- reporter genes (11, 15, 23). Because previous experiments didn't show significant difference of SIP1 and its sumoylation mutant in subcellular localization and protein stability (Fig. 4 and data not shown), we studied the effects of SIP1 sumoylation on transcription regulation.
|
|
We then studied effects of SIP1 sumoylation on E-cadherin repression. YFP-SIP1 and its sumoylation mutant were transfected into 293T cells, and their ability to repress a human E-cadherin promoter (49) was measured (Fig. 5B, top panel). Under the two DNA concentrations tested, the SIP1 sumoylation mutant showed consistently more potent repression on E-cadherin transcription. For example, whereas 400 ng of YFP-SIP1 can repress E-cadherin transcription to 60.7%, the YFP-SIP1 mutant represses E-cadherin to 83.3%. A similar trend was observed using other tagged SIP1 constructs sets, including Myc-SIP1, HA-SIP1, and GFP-SIP1 (data not shown). These results are consistent with the conclusion that SIP1 sumoylation having a negative effect on its repression of E-cadherin transcription.
We also tested effects of SIP1 sumoylation on TGF-/Smad repression in TGF--responsive cell lines with two widely used reporter genes, 3TP-lux and SBE4-luc (43, 44). L17 cells transfected with Myc-SIP1 or its sumoylation mutant, were compared for repression on 3TP-lux in the absence and presence of TGF- (Fig. 5C, top panel). As reported before (15, 23), Myc-SIP1 can repress the basal as well as TGF--induced transcription of 3TP-lux in a dose-dependent manner. In contrast to E-cadherin, Myc-SIP1 (K391R/K866R) showed almost identical repression potential to the wild-type SIP1 protein. A similar phenomenon was observed using the SBE4-luc/Mv1Lu system (Fig. 5D). Under both conditions, SIP1 expression levels were similar (Fig. 4C, bottom panel, and data not shown). Thus, although SIP1 sumoylation alleviates the transcriptional repression of E-cadherin, sumoylation does not affect its repression on other target genes, including at least two TGF--responsive genes.
EF1 has been reported to activate the transcription of certain genes, such as TGF--responsive reporter genes (23, 26), VD3R (24), and ovalbumin (25). Although EF1 and SIP1 showed some redundancy in their function, there are circumstances where these two members behave differently (18, 23, 26). So far it is not clear whether SIP1 could activate transcription. We therefore compared Myc-SIP1, Myc-SIP1 (K391R/K866R), and Myc-EF1 for their effects on VD3R reporter genes (24). As previously reported (24), EF1 mediates transactivation of the VD3R promoter (Fig. 5E). Interestingly, Myc-SIP1 can also activate VD3R transcription and has a higher transactivation potential than Myc-EF1, under conditions where their expression were comparable (Fig. 5D and data not shown). Furthermore, the SIP1 sumoylation mutant showed a similar ability to activate VD3R transcription. Thus, SIP1 sumoylation does not affect its ability to activate the transcription of VD3R.
Functional Involvement of Pc2-mediated SIP1 Sumoylation in E-cadherin Repression—To test more directly whether Pc2-mediated SIP1 sumoylation could attenuate its repression on the E-cadherin promoter, YFP-SIP1 or empty vector was cotransfected with sumoylation machinery components (SUMO1, Pc2, or SUMO1 plus Pc2) and their effect on E-cadherin transcription was measured (Fig. 6, left panel). Notably, Pc2 on its own can significantly potentiate E-cadherin transcription (lanes 3). When YFP-SIP1 was included, E-cadherin transcription was repressed (lane 5 versus lane 1), consistent with Fig. 5B. Moreover, coexpression of Pc2 with YFP-SIP1 led to the loss of the repression (Fig. 6, lane 7). Under these conditions, transfected genes were expressed at comparable levels (bottom panel), and SIP1 sumoylation was detectable and was promoted by Pc2 (Fig. 3A, and data not shown). To discriminate the effect of Pc2-mediated SIP1 sumoylation on E-cadherin transcription, the relative E-cadherin repression by YFP-SIP1 (in percentage) was plotted against the plasmid transfected (Fig. 6, right panel). This reveals that coexpression of Pc2, with YFP-SIP1, and presumably sumoylation of SIP1, could relieve E-cadherin repression, at least in part. Thus, Pc2-mediated sumoylation of SIP1 might contribute to the derepression of E-cadherin transcription.
SIP1 Sumoylation Disrupts the Recruitment of Corepressor CtBP—To determine the mechanism by which SIP1 sumoylation selectively attenuates the E-cadherin repression but not TGF-/Smad signaling (Fig. 5, B–D), we compared the binding of SIP1 and its sumoylation mutant to Smad3 and CtBP. As shown in Fig. 7A, the SIP1 sumoylation mutant showed similar binding ability to Smad3 (lanes 2 and 3). Binding to other receptor-regulated Smads, including Smad2 and Smad1, were also not affected by mutation of the sumoylation sites (data not shown). In contrast, SIP1 binding to CtBP is significantly elevated in the mutant protein (lanes 5 and 6), suggesting that SIP1 sumoylation may disrupt CtBP recruitment to attenuate its repression. As previously reported (22), simultaneous mutation of the CtBP-binding consensus motifs, abolished the ability of SIP1 to bind CtBP (lane 7).
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
EF1 and SIP1 were also shown to be phosphorylated at multiple serine/threonine sites (63).4 We show herein that SIP1 is covalently modified by sumoylation at two conserved lysine residues (Lys391 and Lys866) and that polycomb protein Pc2 can act as a SUMO E3 ligase for SIP1. Importantly, we found that sumoylation of SIP1 regulates its transcriptional activity in a promoter context-dependent manner.
We first mapped the SIP1 sumoylation sites to its repression domain. A similar observation has been reported for other transcription factors (31, 40, 55). Among the three consensus sumoylation motifs in the repression domain of SIP1, Ile-Lys391-Thr-Glu and Ile-Lys866-Lys-Glu were found to be sumoylated (Fig. 2, A and B). Accordingly, sumoylation was also observed in the homologous protein EF1 at the corresponding sites, Ile-Lys347-Thr-Glu and Ala-Lys774-Lys-Glu (Fig. 2D). Consistent with the observation that SIP1 Ile-Lys555-Lys-Glu is not sumoylated, its corresponding motif in EF1, Cys-Lys504-Ser-Glu, is not a consensus sumoylation motif. These results suggest that sumoylation of zfh-1 proteins might be evolutionally important for their functions. SIP1 Lys391 is a major sumoylation site, whereas the effect of Lys866 sumoylation was apparent only when Lys391 was absent, suggesting possible synergy between the two sites.
Polycomb Protein Pc2 Is a SUMO E3 Ligase for SIP1—The PIAS family of E3 ligases have been shown to promote the sumoylation of the majority of the reported target proteins, whereas RanBP2 and Pc2 only enhance the sumoylation of certain proteins, including SP100 (36), histone deacetylase 4 (HDAC4) (45, 48, 64), and corepressor CtBP (37, 59, 65). The gap between the increasing numbers of sumoylated proteins and known E3 ligases suggests that unknown SUMO E3 ligases wait to be identified. To support this notion, Nse2 was recently reported to act as SUMO E3 ligase in Schizosaccharomyces pombe (66), class IIa HDACs (HDAC4, HDAC5, HDAC7, and HDAC9) were found to stimulate the sumoylation of MEF2 transcription factors (45).
The polycomb protein Pc2 has been shown to act as a SUMO E3 ligase, by recruitment of the E2 enzyme Ubc9, and the substrate into PcG bodies, which might act as a center for sumoylation in a similar manner to promyelocytic leukemia protein bodies (37). So far the corepressor CtBP is the only identified substrate for Pc2 (37, 59), although CtBP can also be sumoylated by PIAS family proteins (59, 65). Structure-function analysis has revealed that separate domains in Pc2 are responsible for its ligase activity toward CtBP. Both the N-terminal region of Pc2, which shows minimal ligase activity in vitro, and the C terminus of Pc2, which acts as scaffold for the E2 enzyme Ubc9 and substrate CtBP, are required for Pc2 to promote the sumoylation of CtBP (59). In this study, we found that overexpression of Pc2, but not other SUMO E3 ligases, including PIAS3, PIASy, and RanBP2, can enhance the sumoylation of SIP1 in a dose-dependent manner (Fig. 3A, and data not shown). While overexpression of Pc2 could promote SIP1 sumoylation indirectly through binding and/or modification of other proteins such as CtBP, more direct evidence of Pc2 as an E3 for SIP1 comes from the in vitro sumoylation assay, where bacterially purified His-Pc2 exerts E3 ligase activity toward SIP1 sumoylation in vitro (Fig. 3B). Although the in vitro effect of Pc2 on SIP1 sumoylation is modest, a similar effect was reported on CtBP (37). Furthermore, two separate domains in SIP1, N- and C-terminal zinc-finger clusters, are each sufficient for binding to Pc2 to a similar extent (Fig. 3D), and Pc2 can recruit SIP1 into PcG bodies (Fig. 3E). These results indicate that Pc2 can act directly as an E3 ligase for SIP1 sumoylation, identifying SIP1 as the second substrate for Pc2. Notably, Pc2 alters the pattern of SIP1 sumoylation in vitro (Fig. 3B), suggesting a possible effect on choice of target lysines.
It is widely believed that E3 provides substrate specificity. Hence it would be important for future research to identify the Pc2 domain that mediates binding and ligase activity to SIP1, which might help explain how Pc2 provides substrate specificity toward SIP1.
Functional Significance of SIP1 Sumoylation—Protein sumoylation is an important post-translational modification, which can regulate the function of target proteins in multiple aspects. For SIP1, we found that sumoylation does not affect its subcellular localization (Fig. 4), nor does it affect the stability of SIP1 as assayed by pulse-chase (data not shown).
Immunofluorescence demonstrated that the SIP1 sumoylation null mutant can still be recruited into PcG bodies as the wild-type (Fig. 4B), suggesting that recruitment of SIP1 into PcG bodies is sumoylation-independent. Such sumoylation-independent recruitment of substrate by SUMO E3 into a subnuclear domain has previously been reported on LEF1 by PIASy (35) and on CtBP by Pc2 (59).
As a two-handed zinc-finger protein, SIP1 has been shown to bind to E-boxes at the consensus 5'-CACCT-3' sequence through its N- and C-terminal zinc-finger clusters (11, 17, 18). The sumoylation sites of SIP1 lie between the zinc-finger clusters in the central repression domain and is therefore not expected to affect its E-box-binding activity. In agreement with this, the SIP1 sumoylation null mutant regulates the VD3R promoter in a similar manner to the wild-type SIP1 (Fig. 5E).
SIP1 Sumoylation Regulates Transcription in a Promoter Context-dependent Manner—An increasing body of transcription factors has been shown to be modified by sumoylation. SUMO conjugation to transcription factors is generally associated with transcriptional repression (31, 32, 39, 40, 55). Two general models have been postulated to explain SUMO-dependent transcriptional repression, although they are not mutually exclusive (31). SUMO-modified transcription factors could recruit corepressor molecules to promoters and induce changes in chromatin structure associated with repression (31, 39). For example, corepressors HDAC1 (67), HDAC2 (68), HDAC6 (69), and Daxx (54) are recruited to the SUMO-modified histone H4 (67), the Elk-1 transcription factor (68), the coactivator p300 (69), and the Smad4 transcription factor (54), respectively, to mediate their sumoylation-dependent repression. An alternative model to explain the repressive effect of SUMO on transcription, involves the recruitment of SUMO-modified proteins into the repressive environment of particular subnuclear domains, such as promyelocytic leukemia protein bodies and PcG bodies (31). For example, sequestration of the PIASy-sumoylated transcription factor, LEF1, into nuclear bodies accounts for the repression of LEF1 activity (35). In a similar manner, the sequestration of sumoylated Sp3 into promyelocytic leukemia protein bodies correlates with the silencing of its transcriptional activity (60). Moreover, the corepressor CtBP can be recruited into the repressive PcG bodies by Pc2 (59).
Only a few examples of direct transactivation by protein sumoylation were reported, including heat shock transcription factors and nuclear factor of activated T cells (31, 32, 39, 40). In this study, we have shown that sumoylation of SIP1 has a negative effect on its repression activity. Specifically, SIP1 sumoylation potentiates transcription. First, in the GAL4 fusion assay, which has been used to test the effect of sumoylation of many transcription factors, including SP3 (60), Smad4 (43), MEF2D (45), Ikaros (61), the SIP1 sumoylation null mutant, K391R/K866R, has consistently lower transcriptional activity when compared with the wild-type (Fig. 5A), and this effect is more significant when the coactivator VP16 was coexpressed. Second, coexpression of SUMO1 with Pc2 could partially relieve repression on an E-cadherin promoter (Fig. 6). Together these data support the interpretation that SIP1 sumoylation attenuates its transcription repression of the E-cadherin promoter. Similarly, sumoylation of the corepressor Ikaros was recently shown to alleviate its transcriptional repression activity (61).
The observation, that SIP1 sumoylation only partially relieves transcriptional repression of E-cadherin, might be in part due to the fact that protein sumoylation is under dynamic and tight regulation and that only a small fraction of the substrate is modified at any given time (32). SIP1 sumoylation is also inefficient even when SIP1 was overexpressed (Figs. 1A, lanes 2 and 3, and 3A, lane 2). In addition, other transcription factors, including Snail (8, 9), Slug (10), EF1 (12), twist (13), and basic helix-loop-helix factor, E47 (14) (reviewed in Ref. 3), have been shown to repress E-cadherin transcription. These factors all bind to the E-box elements of E-cadherin promoter with different affinity, with Snail being the highest (3). Hence the modest effect of SIP1 sumoylation on E-cadherin transcription may reflect competition of a small pool of sumoylated SIP1 with unmodified SIP1 and other repressors that bind to E-box elements. The SIP1 homologue EF1 (ZEB1) was also shown to repress E-cadherin transcription modestly in a similar study (65).
In this study, SIP1 sumoylation was found to regulate its transcriptional activity in a promoter context-dependent manner. Although SIP1 sumoylation attenuates its transcription repression activity on an E-cadherin promoter (Fig. 5B), sumoylation does not affect its repression of some TGF- reporter genes or activation of the VD3R gene (Fig. 5, C–E). Similarly, sumoylation of Ikaros was shown to selectively interfere with its repressor activity, but not its role as a transcription potentiator (61). In addition, we have previously shown that sumoylation of Smad4 depends on promoter context to repress or activate transcription (43).
SIP1 Can Activate VD3R Transcription—During this study, we have unveiled a novel function for SIP1, as a transcriptional activator for VD3R (Fig. 5E). This is supported by the existence of an acidic amino acid-rich domain in the C terminus of SIP1, which is generally linked to transcription activation (70), and activation of VD3R by the homologue protein EF1 (24). So far, the mechanism of transactivation by EF1 and SIP on VD3R is unknown. For EF1, recruitment of transcription coactivator p300 and/or P/CAF might be responsible, because its unique N-terminal domain can bind directly to these coactivators (26). However, such a model for recruitment would not be expected for SIP1, because SIP1 in unable to bind to p300 or P/CAF (26).
SIP1 Sumoylation Disrupts the Recruitment of the Corepressor CtBP—SIP1 is an active transcription repressor for many transcription factors and target genes, and it is generally believed that, as shown for its homologue EF1 (19, 20), repression is dependent on the recruitment of corepressor CtBP (18, 26), through the four putative CtBP-binding motifs (PLXL(S/T)) in its CtBP-interacting domain (18, 22). Direct evidence was provided by chromatin immunoprecipitation demonstrating that a SIP1-containg CtBP repression complex can bind to the E-cadherin promoter and repress its transcription (21). Repression of TGF-/Smad signaling by SIP1 was also attributed to recruitment of CtBP (26).
To explore the mechanism how SIP1 sumoylation selectively regulates transcription, we studied the interaction between SIP1 and Smads, and found that the SIP1 sumoylation mutant showed similar binding ability to Smad3 and other receptor-regulated Smads as wild-type (Fig. 7A and data not shown), which supports our observation that SIP1 sumoylation does not affect repression of TGF- signaling (Fig. 5, C and D).
We also examined the binding between SIP1 and CtBP and found that the SIP1 sumoylation mutant binds more CtBP than the wild-type (Fig. 7A). Accordingly, coexpression of YFP-SUMO1, and presumably sumoylation of SIP1, decreases the endogenous CtBP binding to SIP1 (Fig. 7B). These results suggest that SIP1 sumoylation can disrupt corepressor CtBP recruitment. However, we cannot exclude that competition for SUMO-binding proteins, or sumoylation of other proteins, such as CtBP itself, may also contribute to the difference in CtBP binding that we observe. Because the SIP1 sumoylation motif Ile-Lys866-Lys-Glu, is adjacent to one of the CtBP-binding motifs, PLNLT (shown in Fig. 2A), it is tempting to speculate that SUMO conjugation at Lys866 could prevent binding of CtBP to that motif and, at least in part, may contribute to the disruption of CtBP recruitment. Similarly, sumoylation of the corepressor Ikaros was shown to disrupt corepressor recruitment (61). It was also shown that sumoylation and binding of CtBP to PDZ are mutually exclusive and adversely regulate CtBP activity due to the close localization of the sumoylation site on CtBP to the PDZ binding site (65).
In conclusion, the E-cadherin transcriptional repressor SIP1 is modified by sumoylation. Sumoylation can be mediated by polycomb protein Pc2, and it attenuates the transcriptional activity of SIP1 in a promoter context-dependent manner. Notably, SIP1 sumoylation can partially relieve E-cadherin repression. Thus SIP1 sumoylation by Pc2 can be an important intervention target to modulate EMT in tumorigenesis.
|
FOOTNOTES |
|---|
1 To whom correspondence should be addressed: McGill University Health Centre, 687 Pine Ave. West, Montréal, Québec H3A 1A1, Canada. Tel.: 514-934-1934 (ext. 35845); Fax: 514-843-1478; E-mail: morag.park{at}mcgill.ca.
2 The abbreviations used are: EMT, epithelial-mesenchymal transition; SIP1, Smad-interacting protein 1; SUMO, Small ubiquitin-like modifier; TGF-, transforming growth factor-; CtBP, C-terminal-binding protein; HDAC, histone deacetylase; PIAS, protein inhibitor of activated STAT; PIASy, protein inhibitor of activated STAT y; PcG, polycomb group; PML, promyelocytic leukemia protein; VD3R, vitamin D3 receptor; MDCK, Madin-Darby canine kidney cells; P/CAF, p300/CBP-associated factor; HA, hemagglutinin; GST, glutathione S-transferase; E1, ubiquitin-activating enzyme; E2, ubiquitin carrier protein; E3, ubiquitin-protein isopeptide ligase; STAT, signal transducers and activators of transcription; CMV, cytomegalovirus; N-ZF, N-terminal zink-finger; C-ZF, C-terminal zinc-finger; CID, CtBP-interacting domain; YFP, yellow fluorescent protein; HDAC4, histone deacetylase 4; HD, homeodomain.
3 J. Long and M. Park, unpublished work.
4 J. Long, D. Zuo, and M. Park, unpublished work.
|
ACKNOWLEDGMENTS |
|---|
EF1/ZEB1 constructs, and some of the TGF--responsive reporter assays, respectively, and members of the Park laboratory for comments on the manuscript. |
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
