SUMO modification of the Ets-related transcription factor ERM inhibits its transcriptional activity.

A variety of transcription factors are post-translationally modified by SUMO, a 97-residue ubiquitin-like protein bound covalently to the targeted lysine. Here we describe SUMO modification of the Ets family member ERM at positions 89, 263, 293, and 350. To investigate how SUMO modification affects the function of ERM, Ets-responsive intercellular adhesion molecule 1 (ICAM-1) and E74 reporter plasmids were employed to demonstrate that SUMO modification causes inhibition of ERM-dependent transcription without affecting the subcellular localization, stability, or DNA-binding capacity of the protein. When the adenoviral protein Gam1 or the SUMO protease SENP1 was used to inhibit the SUMO modification pathway, ERM-dependent transcription was de-repressed. These results demonstrate that ERM is subject to SUMO modification and that this post-translational modification causes inhibition of transcription-enhancing activity.

Covalent modification of proteins with ubiquitin-like proteins creates new proteins with unique protein surfaces that can mediate a range of protein-protein interactions (reviewed in Ref. 1). Modification of proteins by the small ubiquitinrelated modifier (SUMO) is increasingly recognized as an important regulatory mechanism that can affect the stability, subnuclear localization, and transcription-activating capacity of the protein (reviewed in Refs. 1 and 2). Several transcription factors are reported to undergo SUMO modification, including the androgen receptor (3,4), c-Myb (5), AP-2 (6), and c-Jun and p53 (7). In addition, other transcriptional regulators, such as GRIP (8) and histone deacetylases 1 (9) and 4 (10), are also SUMO targets (the currently established targets of SUMO are reviewed in Ref. 1). Three different types of enzyme constitute the SUMO pathway: ubiquitin-activating enzyme (E1), 1 ubiquitin-conjugating enzyme or carrier protein (E2), and ubiquitin-protein isopeptide ligase (E3). Although Ubc9 is the only E2 (conjugating enzyme) found in humans, multiple E3 ligases have recently been identified. SUMO modification is a reversible process, and several SUMO proteases have been identified in mammalian cells (reviewed in Ref. 1).
In mammals, the ets genes encode a large family of transcription factors, characterized by their ETS DNA-binding domain. On the basis of conservation of this and other domains, these factors have been subclassified into 13 groups (for a review, see Ref. 11). Some Ets transcription factors have been identified as targets of SUMO. For instance, Elk-1 is modified by SUMO, a modification that is reversed by signaling via the prototype extracellular signal-regulated kinase (ERK) MAPK pathway, which induces a switch from the repressed to the transcriptionally active state (12,13). SUMO modification of Elk-1 results in the recruitment of histone deacetylase activity to promoters; this indicates the existence of an important integration point for two protein-modifying pathways in the cell, the SUMO and deacetylation pathways, which combine to promote inhibition of transcription-enhancing activity (14). SUMO has also been shown to enhance the recruitment of the Ets transcription factor Tel into repressive domains such as the PML bodies (15)(16)(17).
The three PEA3 group members (PEA3/E1AF, ER81/ETV1, and ERM/ETV5) show high conservation of their ETS domain and of the two transcriptional activation domains (reviewed in Ref. 18). These factors are involved in a number of developmental processes. For example, they play a role in the organization of the germ layers showing high proliferation and migration rates (19) and in the development of motor and sensory neurons (20,21). These factors have also been found to be deregulated in cancer and are over-expressed in metastatic human breast cancer cells (22) and Neu-induced mouse mammary tumors (23,24).
Post-translational modifications such as phosphorylation regulate the function of the PEA3 group members. In particular, components of the MAPK pathway have been found to increase the transactivation capacity of these factors. This suggests that these factors may contribute to the nuclear response to cell stimulation and also to Ras-induced transformation (25)(26)(27)(28)(29)(30). The c-Jun NH 2 -terminal kinase/ stress-activated protein kinase (JNK/SAPK) and cAMP-dependent protein kinase (PKA) pathways are also involved in regulating the transcription-enhancing activity of the PEA3 group members (30 -32).
In the present study we have investigated whether ERM undergoes SUMO modification. We show that of the five putative SUMO modification sites of this protein, the first four can be conjugated to SUMO. We present the results of functional studies showing that this post-translational modification can inhibit ERM transcription-enhancing activity.

MATERIALS AND METHODS
Yeast Two-hybrid Screen-The yeast two-hybrid screening was performed with the MATCHMAKER two-hybrid system as recommended by the manufacturer (Clontech). A DNA fragment encoding amino acids 72-370 of ERM was generated by PCR amplification and subcloned in-frame with the LexA DNA-binding domain contained in the yeast expression vector pLexA (Clontech) at the EcoRI restriction site. The resulting plasmid was transformed using a standard polyethylene glycol/LiAc-mediated transformation procedure into the EGY48 strain of yeast. An oligo(dT)-primed mammary gland cDNA library (Clontech) constructed in the yeast galactose-inducible expression plasmid pB42AD was introduced into the EGY48 yeast strain harboring the pLexA-ERM construct. The resulting co-transformants were grown at 30°C in the absence of histidine, tryptophan, and uracil (selective minimal media). A total of 3 ϫ 10 6 clones were screened for the capacity to grow on selective minimal media. Colonies growing in the absence of leucine and scoring positive for ␤-galactosidase activity by the filter assay method were directly recovered from the plate and grown in selective media lacking leucine. cDNA inserts from the positive yeast colonies were electrotransformed into KC8 Escherichia coli tryptophan auxotrophic bacterial cells. Transformants were grown on M9 (ϪTrp, ϪAmp) minimal medium to select for plasmids containing the pB42AD vector. The cDNA inserts were then sequenced by automated sequence analysis (ABI sequence analyzer).
Plasmid Constructs-The full-length and amino-terminal truncated pSG5-ERM expression vectors and the pSV-HA-ERM vector have been described elsewhere (32). The 5Ј FLAG-tagged ERM vector (pSV-FLAG-ERM) was constructed using the ERM cDNA and the pSV plasmid used previously (33) (details are available upon request). The various ERM mutants (Lys to Arg and Glu to Ala) were generated with a QuikChange site-directed mutagenesis kit (Stratagene) and verified by DNA sequencing. His 6 -SUMO-1, -SUMO-2, and -SUMO-3 expression plasmids have been described elsewhere (34). As luciferase reporter plasmids, we used the 3xE74-tk-Luc vector (25) and the ICAM WT -Luc vector corresponding to the wild-type human ICAM-1 promoter fragment (bp Ϫ44 to Ϫ178) (Ets binding sites at positions Ϫ158 and Ϫ138) (35). The pSG5-␤Gal vector was used to normalize transfections (32). SENP1 and SENP1 mut expression vectors were kindly provided by Dr E. T. H. Yeh (University of Texas M. D. Anderson Cancer Center, Houston) (36) and the Gam1 and Gam1 mut expression vectors were a gift from Dr S. Chiocca (European Institute of Oncology, Milan, Italy).
Cell Culture and Transfections-Rabbit kidney (RK13), monkey kidney (COS-7), and human cervix (HeLa) carcinoma cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) heat-inactivated fetal calf serum (Invitrogen) and cultured at 37°C in water-saturated 5% CO 2 atmosphere. 1.5 ϫ 10 5 RK13 or HeLa cells/well were plated in 12-well plates, and the next day transient transfections were performed using the polyethyleneimine Exgen 500 procedure (Euromedex) using 150 ng of DNA/well including 25 ng of reporter plasmid, 10 ng of pSV expression vector, and 10 ng of ␤-galactosidase expression vector. When indicated, 20 ng of the plasmids expressing the wild-type or mutated SUMO-interfering molecule were used. Luciferase activity was determined 24 h after transfection and normalized with respect to the ␤-galactosidase activity (used as a measure of the transfection efficiency) as described previously (35). The data presented are means Ϯ S.E. of at least three independent experiments. To detect SUMO-modified forms of ERM, COS-7 or RK13 cells were transfected with 300 ng of ERM expression plasmid and, when indicated, with 700 ng of His 6 -tagged SUMO-1, -2, or -3 plasmids. After 24 h, cells were rinsed with phosphate-buffered saline and lysed in immunoprecipitation, immunoblot, or Ni 2ϩ puri-fication buffers. For the determination of protein half-life, COS-7 cells were transfected with ERM and SUMO-2 plasmids for 24 h and then treated with 80 g/ml cycloheximide (Sigma). After the indicated time, cells were processed for Western blotting.
Immunoprecipitation and Western Blot Analyses-Cells transfected with FLAG-tagged ERM were treated with 100 mM H 2 O 2 for 20 min and lysed in phosphate-buffered saline containing 1% SDS, 10 mM iodoacetamide, and protease inhibitors. After the addition of 9 volumes of phosphate-buffered saline containing 1% Triton X-100 and 10 mM iodoacetamide, the cellular extract was centrifuged at 100,000 ϫ g for 30 min at 4°C. The supernatant was then immunoprecipitated with anti-FLAG M2 affinity gel (Sigma) overnight at 4°C. After washing, the immunopurified proteins were eluted with Laemmli sample buffer and analyzed by SDS-PAGE. Prestained broad-range molecular weight markers (New England BioLabs) were used as standards in each SDS-PAGE. Proteins were then electrophoretically transferred to nitrocellulose membranes. Immunoblot analyzes were performed with the rabbit anti-ERM antibodies (22) or with the purified rabbit polyclonal antibodies against SUMO-1 or SUMO-2/3 (Abgent) followed by treatment with horseradish peroxidase-conjugated secondary antibody. Immune complexes were visualized by enhanced chemiluminescence according to manufacturer's instructions (Santa Cruz Biotechnology).
Identification of His-SUMO-ERM Conjugates-Transfected cells were lysed in denaturing buffer containing 6 M guanidine hydrochloride. His 6 -SUMO conjugates were purified by metal-chelate affinity chromatography as described previously (37). Briefly, cells plated in 6-well plates were washed twice in ice-cold phosphate-buffered saline and lysed in 400 l of lysis buffer (6 M guanidine hydrochloride, 0.1 M Na 2 HPO 4 /NaH 2 PO 4 , pH 8.0, 0.01 M Tris-HCl, pH 8.0)/35-mm well. The lysates were centrifuged at 100,000 ϫ g for 90 min at 4°C. Each sample was mixed with 20 l of packed Ni 2ϩ -nitrilotriacetic acid beads (Qiagen) and rotated 3 h at room temperature. The beads were then washed twice with lysis buffer, three times with 8 M urea, 0.1 M Na 2 HPO 4 / NaH 2 PO 4 , pH 6.4, and once with phosphate-buffered saline before being resuspended in Laemmli gel loading buffer and analyzed by SDS-PAGE. Proteins were then electrophoretically transferred to nitrocellulose membranes, and His-SUMO-ERM conjugates were detected by Western blotting using a rabbit polyclonal anti-ERM antibody as described above.
In Vitro SUMO Modification Reaction-[ 35 S]Methionine-labeled substrates for SUMO modification reactions were generated by in vitro transcription/translation in wheat germ extract according to the manufacturer's instructions (Promega). In vitro modification was carried out with purified recombinant products as described (34). Reaction products were fractionated by SDS-PAGE and detected by phosphorimaging.
Electrophoretic Mobility Shift Assay (EMSA)-ERM was produced and sumoylated in vitro as described above. A portion of the reaction mixture was mixed with 1 ng of the 32 P-labeled E74 probe (sense strand, 5Ј-GAGCTGAATAACCGGAAGTAACTCAT-3Ј) in the presence of 25 mM Hepes, pH 7.9, 25 mM KCl, 2 mM MgCl 2 , 1 mM EGTA, 0.05% Nonidet P-40, 10% glycerol, 1 mM dithiothreitol, and 1 g of poly(dI⅐dC). The mixture was incubated for 1 h at room temperature and loaded onto a 6% polyacrylamide gel. The gel was run at 4°C in 0.5ϫ Tris borate-EDTA buffer at 180 V (32).
Immunofluorescence Studies-RK13 cells plated on coverslip slides were transfected with the different FLAG-tagged ERM plasmids. Twenty-four h later the cells were fixed in 4% paraformaldehyde for 10 min at room temperature and then incubated for 2 h with 1:1000 diluted anti-FLAG antibody. This was followed by incubation for 1 h with 1:1000 diluted goat anti-rabbit Alexa 488 antibody.

SUMO Modification of ERM-To identify ERM-interacting
proteins, we screened a human mammary gland cDNA library in the yeast two-hybrid system using a nontransactivating portion of the ERM molecule (amino acids 72-370) as bait. Twelve isolated clones were found to encode the human Ubc9. GST pull-down assays were used to obtain direct evidence for ERM binding to Ubc9 (data not shown), thus confirming Ubc9 as a novel ERM-interacting protein.
Because Ubc9 is the only-known SUMO-conjugating enzyme (38) and because ERM displays five minimal consensus sumoylation motifs (KXE; , large hydrophobic residue (39)), we sought to determine whether ERM is a sumoylation target using in vitro and in vivo approaches. To show directly that ERM is a SUMO modification substrate, 35 S-labeled ERM was generated by in vitro transcription/translation and incubated with purified components required for SUMO modification (34). As illustrated in Fig. 1A, analysis of the products indicates that a large proportion of the ERM was converted by SUMO-1, SUMO-2, or SUMO-3 to more slowly migrating forms in a reaction requiring the SUMO-activating enzyme E1 and the E2 enzyme Ubc9. To establish that ERM is modified by SUMO in vivo, we co-transfected RK13 cells with expression plasmids for ERM and His 6 -tagged SUMO-1, -2, or -3. Cells were lysed under conditions preserving SUMO modification, and ERM proteins were identified by Western blotting with an anti-ERM antibody. In the absence of SUMO, we observed a major 70-kDa protein species, corresponding to unmodified ERM. Higher molecular weight species were detected only when SUMO-1, SUMO-2, or SUMO-3 was present (Fig. 1B). Under these conditions, the presence of multiple high molecular weight species suggests that multiple SUMO modifications take place on ERM. Similar in vivo experiments were performed on other cell lines, such as COS-7. Again, the upper bands also appeared in the presence of each SUMO and were much pronounced in the presence of SUMO-2 (Fig. 1B). To prove that these upper bands corresponded to SUMO-modified forms of ERM, COS-7 cells transfected with ERM and His-SUMO-2 were lysed under highly denaturing conditions, and SUMO-modified proteins were isolated on nickel-agarose (Ni 2ϩ -nitrilotriacetic acid). Nickel-nitrilotriacetic acid-bound proteins were then eluted and analyzed by Western blot with anti-ERM antibody. As expected, unmodified ERM was absent, and the higher molecular weight ERM species were observed in the SUMO-conjugated cellular fraction (Fig. 1C). It thus appears that ERM is modified by SUMO at multiple sites both in vitro and in vivo.
SUMO Modification Sites on ERM-ERM contains five copies of the KXE sequence required for modification by SUMO ( Fig. 2A). To locate sites of SUMO modification on the ERM protein, we generated amino-terminal deletion mutants of ERM and tested their ability to be SUMO-modified in vivo by co-transfection with His-SUMO-2 (as in Fig. 1C). Although wild-type ERM was modified by SUMO at multiple sites, ERM-(276 -510) was modified at only one site and ERM-(354 -510) did not appear to be SUMO-modified (Fig. 2A). These data suggest that residues Lys 89 and Lys 263 are likely targets of SUMO modification, along with either Lys 263 or Lys 350 . It is unlikely that Lys 468 , although in a sequence matching perfectly the consensus sumoylation motif, is a site of SUMO modification.
To determine precisely the sites of SUMO modification, we therefore replaced each potential acceptor lysine, separately or in combination, with an arginine unable to act as a SUMO acceptor. COS-7 cells were then co-transfected with a plasmid coding for wild-type or mutated ERM together with the SUMO-2 expression vector. ERM was revealed by Western blotting. Fig. 2B shows that the single mutation of Lys 468 (KR 5 ) has no significant effect on ERM SUMO modification. A single mutation of Lys 89 (Fig. 2B, KR 1 ), Lys 263 (KR 2 ), or Lys 350 (KR 4 ) reduced the intensity of at least some SUMO-modified ERM species, and a more pronounced effect was observed for the Lys 293 mutation (KR 3 ). When Lys 263 and Lys 293 (KR 23 ) were mutated in combination, the more slowly migrating SUMOconjugated species of ERM were no longer observed, but the major 110 -120 kDa was still clearly detected. Further mutation of Lys 350 (Fig. 2B, KR 234 ) had little effect on the SUMO modification pattern, whereas mutation of Lys 89 in combination with KR 23 (KR 123 ) caused the amount of the major SUMOmodified species to decrease markedly. This modified form was practically undetectable when Lys 89 , Lys 263 , Lys 293 , and Lys 350 were mutated simultaneously (Fig. 2B, KR 1234 ), and no further change was observed with ERM mutated at all five consensus sumoylation sites (KR 12345 ). Simultaneous mutation at all of the five consensus sites completely prevented SUMO modification of ERM in vitro (Fig. 2C). These findings suggest that SUMO modification occurs only at consensus sites and that, of the five consensus sites identified in ERM, the motif containing Lys 468 in the carboxyl-terminal part of ERM is the only one that is not SUMO-modified in vivo.
ERM Is SUMO-modified in Response to Stress in Vivo-Because cell stresses such as heat shock or oxidative stress induce SUMO conjugation (40), we investigated the ability of endogenous SUMO to modify ERM in vivo upon exposure of cells to hydrogen peroxide (H 2 O 2 ). For this purpose, COS-7 cells were transfected with plasmids coding for wild-type or KR 12345 FLAG-ERM, and transfected cells were treated or not with H 2 O 2 . The cell extracts were then subjected to immunoprecipitation with anti-FLAG antibody, and the immunoprecipitates were analyzed by Western blot with antibodies FIG. 1. ERM is modified by SUMO in  vitro and in vivo. A, SUMO modification of ERM in vitro. ERM was labeled with [ 35 S]Met by in vitro translation in a wheat germ extract and incubated in the presence of recombinant SUMO-1, SUMO-2, or SUMO-3 plus E1 (SAE1/2) and E2 (Ubc9). A control reaction lacking these molecules was also carried out. The products were analyzed by SDS-PAGE and phosphorimaging. The in vitro translated protein gives a two-band pattern because of internal methionine initiation. B, in vivo modification of ERM by SUMO. RK13 and COS-7 cells were transfected for 24 h with an ERM-expressing plasmid alone or with this plasmid plus a plasmid expressing His-tagged SUMO-1, -2, or -3. ERM was detected by Western blotting with an anti-ERM antibody. The upper bands correspond to SUMO-modified ERM forms. C, COS-7 cells were transfected as described in B with plasmids expressing His-tagged SUMO-2 and ERM. Crude extract (Crude) and SUMO-conjugated proteins isolated by nickel affinity chromatography (Ni 2ϩ ) were analyzed by SDS-PAGE, and ERM was then detected by Western blotting with an anti-ERM antibody.
against ERM or SUMO. With the anti-ERM antibody, unmodified ERM was detected independently on H 2 O 2 treatment. In contrast, more slowly migrating ERM species were observed specifically in the H 2 O 2 -treated cells. These H 2 O 2 -induced more slowly migrating ERM species were not observed with the FLAG-ERM KR 12345 mutant (Fig. 3); their appearance was thus dependent on the presence of the sumoylation sites. To determine which form of SUMO was conjugated to ERM, extracts from H 2 O 2 -treated cells immunoprecipitated with the anti-FLAG antibody were analyzed by Western blotting using either an anti-SUMO-1 or an anti-SUMO-2/3 antibody. The more slowly migrating forms of ERM were detected with the anti-SUMO-2/3 antibody but not with the anti-SUMO-1 antibody. These forms were absent when ERM KR 12345 was used. This indicates that endogenous SUMO-2 or/and -3 are responsible for H 2 O 2 -induced ERM sumoylation in vivo.
SUMO Sites in ERM Correlate with Inhibition of Transcription-enhancing Activity-To investigate the effect of SUMO modification on ERM transcription-enhancing activity, we analyzed the ability of wild-type and mutant forms of the protein to activate ERM-dependent reporter genes. In experiments with the human Ets-responsive ICAM-1 minimal promoter (35), the wild-type ERM caused an ϳ20-fold increase in transactivation (Fig. 4A, WT). ERM singly mutated at any one of the five consensus sumoylation sites showed the same or nearly the same transactivation power as wild-type ERM (Fig. 4A, left  panel). Triply mutated ERM proteins KR 123 and KR 234 , however, showed an ϳ3-fold higher transactivation capacity. No further increase was observed with the KR 1234 and KR 12345 ERM mutants (Fig. 4A). The second luciferase reporter vector used contains three E74 Ets-binding sites cloned upstream from the minimal thymidine kinase promoter. Although ERM specifically interacts with the E74 binding site, it cannot activate the E74 reporter construct without stimulation by posttranslational modification such as phosphorylation (32). As shown in Fig. 4A (right panel), RK13 cells co-transfected with the E74 reporter and a plasmid expressing either the wild-type or a singly mutated ERM showed similar, near basal level luciferase activity. When a multiply mutated ERM was used, transactivation increased about 4-fold (KR 234 ) to 8-fold (KR 123 , KR 1234 and KR 12345 ) (Fig. 4A). As assessed by immunofluorescence, the sumoylable (Fig. 4B, WT) and the nonsumoylable (KR 12345 ) ERM forms similarly localized in the nucleus in a diffuse manner. Localization of ERM was not affected by overexpression of SUMO-2. Altogether, these results clearly show that the nonsumoylable ERM mutant has an enhanced transcriptional capacity and that it is not due to change in intracellular localization.
We then investigated whether change in the transcriptional activity of the sumoylation defective ERM mutant is due to variation of protein stability as previously reported for other proteins (41,42). We thus examined the steady-state levels of wild-type and nonsumoylable ERM following inhibition of protein synthesis. COS-7 cells transfected with ERM plasmid were treated from 20 to 80 min with the protein synthesis inhibitor cycloheximide and analyzed by Western blotting. In contrast to actin, which is not influenced by cycloheximide treatment, wild-type ERM is relatively unstable with a half-life of about 40 min; this relatively weak stability is also found in the nonsumoylable KR 12345 ERM mutant (Fig. 5A). Similar experiments were also performed in cells transfected with wild-type ERM and SUMO-2 to examine the stability of the SUMOmodified forms of ERM. As illustrated in Fig. 5A, the SUMOmodified forms showed similar stability as compared with un- modified wild-type ERM or the KR 12345 ERM mutant. These results indicate that neither mutation of the sumoylation sites nor sumoylation of ERM significantly affects ERM stability.
We also tested whether the DNA-binding capacity of the nonsumoylable ERM form varies from that of wild-type protein. Using EMSA with nuclear extracts of COS-7 cells overexpressing ERM or its nonsumoylable mutant KR 12345 , we showed that the specific DNA binding was not affected by the mutation of the sumoylation sites (data not shown). We had shown previously, however, that in vivo only a small proportion of the overexpressed ERM protein is SUMO-modified, and the influence of sumoylation on ERM DNA-binding capacity was thus difficult to evaluate in these conditions (Fig. 2B). To circumvent that, we subjected ERM produced by in vitro translation to in vitro SUMO modification and then tested its DNAbinding ability. Under these conditions, ERM was almost fully SUMO-modified (Fig. 5B, SDS-PAGE), and gel shift analysis with an E74 probe specifically recognized by ERM (32) revealed a lower mobility of SUMO-modified ERM-DNA complex (Fig.  5B, EMSA). However, similar band intensities were observed for the SUMO-modified ERM-DNA and the unmodified ERM-DNA complexes (Fig. 5B, EMSA; compare lanes 2 and 3). This indicates that difference in transcriptional activity between wild-type and sumoylation-defective ERM is independent of DNA-binding ability.
Sumoylation Has a Negative Effect on Transcription Regulation by ERM-We thus have shown that SUMO modification of ERM did not affect the subcellular localization, the protein stability, and the DNA-binding activity of ERM, whereas mutation of the lysine within the SUMO sites induced transcriptional activation. However, in addition to sumoylation, the lysine residue can be modified by a number of post-translational modifications including methylation, ubiquitination, and acetylation. To ensure that enhanced transcriptional activity of ERM mutated at all of the lysine acceptor sites (KR 12345 ) is specifically due to defective SUMO modification, we tested the activity of the mutant protein obtained by changing the glutamic acid of each consensus sumoylation site to alanine, without mutating the lysine residue of the site. Such mutations are reported to disrupt SUMO transfer by affecting the interaction of the substrate with the SUMO-conjugating enzyme Ubc9 (43). In RK13 cells, the effect on transcription of ERM mutated at all five glutamic acid residues (EA 12345 ) was similar to that of the KR 12345 ERM mutant in both the ICAM-1 and E74 reporter systems (Fig. 6A). Similar results were also obtained in HeLa cells. As expected, the EA 12345 ERM mutant, like the KR 12345 ERM mutant, could not be modified by SUMO, in contrast to wild-type ERM (Fig. 6B), thus confirming that the loss of SUMO modification correlates with enhanced transcriptional activation of ERM.
The data obtained strongly suggested that sumoylation inhibits the transcriptional activity of ERM. We thus determined whether decreased sumoylation reverses the inhibition. In reporter gene assays, we used two proteins known to interfere with the SUMO conjugation pathway: the adenoviral protein Gam1, which inhibits the SUMO pathway by inducing the degradation of SUMO-activating (E1) and -conjugating (E2) enzymes (44); and SENP1, a SUMO-specific protease involved in desumoylation (36). Neither protein had any effect on basal transcription from the ICAM-1 and E74 reporter plasmids (Fig.  7). In the E74 reporter system, both Gam1 and SENP1 caused a major (ϳ6-fold) increase in the activity of wild-type ERM. This increase was dependent on the presence of the sumoylation sites, since neither Gam1 nor SENP1 altered the transcriptional activity of the KR 12345 ERM mutant (Fig. 7). In reporter system based on the ICAM-1, results were similar. Both interfering proteins increased transactivation by wildtype ERM, and KR 12345 -induced activity remained unaltered in the presence of these molecules. Moreover, the enhancement of ERM transactivation capacity was due to the ability of these two proteins to interfere with the SUMO pathway, as the transcriptional activity of ERM was only marginally changed by coexpression of inactive versions of Gam1 and SENP1 (Gam1 mut , Gam1 L258,265A ; SENP1 mut , SENP1 R630L,K631M ) (Fig.  7). These data therefore demonstrate that the transcriptional activity of ERM is inhibited by the SUMO pathway. DISCUSSION ERM transcription-enhancing activity is regulated by posttranslational modifications such as phosphorylation via the MAPK and cAMP-dependent protein kinase pathways (25,32). Here we show that ERM interacts with the SUMO-conjugating enzyme Ubc9 and is modified by SUMO. We further show that SUMO modification of this Ets transcription factor affects its ability to activate transcription.
We have demonstrated that four of the five lysines located in optimal sumoylation consensus motifs (KXE (39)) within the ERM amino acid sequence, Lys 89 , Lys 263 , Lys 293 , and Lys 350 , are modified by SUMO both in vitro and in vivo. When either the acceptor lysines or the glutamic acid residues of these sites are mutated, SUMO modification of the mutated ERM does not occur. These SUMO acceptor sites are conserved in human, mouse, chicken, and zebrafish ERM homologues. This suggests that SUMO modification of the ERM transcription factor plays an important role in vertebrates. These sites are also perfectly conserved in the two other PEA3 group members, ETV1 and PEA3. We have observed that both of these also undergo in vivo SUMO modification (data not shown). On the contrary, the fifth sumoylation consensus site (Lys 468 ), which is not modified in ERM, is not conserved in zebrafish ERM, human ETV1, or mouse PEA3. In these three molecules, the glutamic acid of this consensus site is replaced by an aspartic acid (data not shown).
ERM can thus be modified by SUMO at multiple sites, and multiple SUMO-modified forms of ERM are indeed detected in cells. Although SUMO-2 seems to conjugate preferentially with ERM, the SUMO modification profiles obtained for ERM in RK13 cells are similar for SUMO-1, -2, and 3. Unlike SUMO-2, SUMO-1 cannot form poly-SUMO chains (34). It is likely that the more slowly migrating SUMO-modified forms observed reflect monosumoylation of ERM at multiple sites. SUMO modification at a single acceptor site also occurs. The corresponding modified protein must be the major ERM species (about 110 -120 kDa), because this SUMO-modified ERM form is the only one remaining when three of the four SUMO conjugation sites are mutated. The observed molecular mass of this modified ERM form is, however, much higher as expected (about 85 kDa; ERM is about 70 kDa, and His-SUMO is about 15 kDa). Although surprising, this phenomenon is common for SUMOmodified molecules and is probably due to an altered conformation of the modified proteins.
Selective addition of SUMO-1, -2, and -3 (SUMO-2 and -3 FIG. 7. Effect of inhibitors of the SUMO pathway on ERM transcription-enhancing activity. RK13 cells were transiently transfected for 24 h as described for Fig. 4A in the presence of either the wild-type (Gam1) or mutated (Gam1mut) adenoviral SUMO-interfering protein or the wild-type (SENP1) or mutated (SENP1mut) SUMO protease. The transcriptional activity of ERM (WT) and ERM-KR 12345 (KR 12345 ) proteins was determined on ICAM and E74 promoters, and the results are as described for Fig. 4A.   FIG. 6. Effect of glutamic acid-to-alanine mutations on transcriptional activation by ERM. The glutamic acid residue present in each of the five SUMO modification sites (KXE) of ERM was replaced with an alanine (EA 12345 ). A, the transcription-activating capacity of the EA 12345 ERM mutant was compared with that of wild-type ERM (WT) and of the KR 12345 ERM mutant. Experiments were performed in RK13 and HeLa cells, and the results are presented as in Fig. 4A. B, in vivo SUMO modification of EA 12345 ERM was carried out in COS-7 cells, as described for Fig. 1B, and compared with wild-type ERM. being highly homologous) to protein targets has not been studied extensively. Studies have indicated that SUMO-1 can conjugate with proteins via mechanisms similar to the conjugation mechanism of SUMO-2 and SUMO-3, but it has been suggested that SUMO-1 on the one hand and SUMO-2/-3 on the other show different substrate specificities (40,45). When ectopically expressed, all three SUMO isoforms have the capacity to become conjugated to ERM, although SUMO-2 conjugates predominate in COS-7 cells. Interestingly, when ERM was expressed in COS-7 cells treated with H 2 O 2 , only the covalent attachment of endogenous SUMO-2 and/or -3 was visualized. This is in agreement with the observation that the incorporation of SUMO-2/-3 into conjugates increases in response to oxidative stress (40). Whether this actually reflects a preferential modification by SUMO-2 remains, however, to be determined. SUMO modification has been reported for several proteins, and most of them exert their main function within the nucleus as transcription factors or transcriptional co-activators or corepressors (for review, see Ref. 1). Here, we show that the Ets transcription factor, ERM, is submitted to SUMO modification and that this post-translational modification negatively affects its transcription-activating function, because SUMO modification-deficient mutant ERM proteins display a greater capacity to activate transcription. Moreover, two molecules known to down-regulate SUMO modification by distinct mechanisms (Gam1 and SENP1) enhanced ERM transcriptional activity, whereas they induced only a slight variation of the KR 12345 transcriptional activity. Altogether, these data argue for an important role of SUMO modification on ERM transcriptional activity. A correlation between SUMO modification and inhibition of transcription-regulating activity has previously been observed for other transcription factors such as Elk-1, Sp3, and c-Myb (for review, see Ref. 1). Most of these SUMO-modified proteins possess fewer SUMO acceptor sites than ERM, which has four of them. This relatively high number of SUMO sites may be responsible for fine regulation through sumoylation. It appears that SUMO modification of all sites is not necessarily required to block the full transcriptional activity of ERM. Although mutation of a single sumoylation site does not significantly increase the action of ERM on the promoters used, mutation of three of the four sites (KR 123 and KR 234 ) is sufficient to obtain the full transcriptional de-repression observed when ERM is mutated at all five consensus sites (KR 12345 ). This applies, however, only to the ICAM-1 promoter context, because on the E74 promoter, ERM KR 123 displays a weaker transcription-enhancing effect than ERM KR 234 . This suggests that according to the promoter context, the repressive effect of sumoylation may vary in function of the number and identity of SUMO-modified sites.
As reported for other SUMO-modified proteins, such as the Ets protein Elk-1 (13), only a small proportion of total ERM protein is SUMO-conjugated. However, mutation of the sumoylation sites on ERM or inhibition of the sumoylation process has a significant effect on the ERM transcription-enhancing activity. This is probably because SUMO modification is a highly dynamic process and this modification, although required to initiate transcriptional inhibition, is not necessary to maintain this inhibition. The molecular mechanism associated to SUMO on inhibition of transcription-enhancing activity is still unclear. Here, we show that this inhibition is not related to changes in the stability or DNA-binding activity of ERM. It might be that repression could be linked to SUMO itself, because SUMO possesses an intrinsic repressive ability when fused to the DNA binding of Gal4 (1). Such a repressive mechanism could involve histone deacetylase recruitment, as dem-onstrated previously for p300 (46) and Elk-1 (14). It is, however, possible that SUMO interferes with the recruitment of co-factors by ERM by decreasing the interaction with transcriptional elements required for transcriptional activity. In fact, SUMO-induced transcriptional activity of ERM could be regulated by CBP, which is a co-activator of the PEA3 group members (47) and a target of SUMO modification. Indeed, sumoylation could disfavor the recruitment of CBP by ERM, thus reducing the ERM transcriptional activity, and reduced ERM activity could be enhanced by the SUMO-dependent repression of CBP activity by recruitment of HDAC6 (46).
It has been shown that SUMO conjugation targets proteins to different cellular localizations. We observed no difference in nuclear localization between wild-type ERM and its SUMO conjugation-deficient mutant. Yet, we cannot exclude the possibility that SUMO modification of ERM might direct the protein, in a subtle manner not currently observed, into some particular nuclear domains or might affect the nuclear import/ export shuttling of this transcription factor, as shown previously for the Ets proteins Tel and Elk-1 (17,48). More experiments are needed to specify the mechanisms by which SUMO controls the activity of ERM.