Transcriptional activity of CCAAT/enhancer-binding proteins is controlled by a conserved inhibitory domain that is a target for sumoylation.

CCAAT/enhancer-binding proteins (C/EBPs) are basic region/leucine zipper transcription factors that function as regulators of cell growth and differentiation in numerous cell types. We previously localized transcriptional activation and inhibitory regions in one family member, C/EBP epsilon. Here we describe the further characterization of a C/EBP epsilon inhibitory domain termed regulatory domain I. We show that functionally related domains are present in C/EBP alpha, C/EBP beta, and C/EBP delta. These domains contain an evolutionarily conserved five-amino acid motif (the regulatory domain motif (RDM)) that conforms to the consensus sequence (I/V/L)KXEP. Mutagenesis studies revealed that the residues at positions 1, 2, and 4 of the RDM are critical for inhibitory domain function. Data base searches identified RDM-like sequences in a number of nuclear proteins. We found that small regions from c-Jun, JunB, and JunD containing this sequence also function as transcriptional inhibitory domains. Importantly, the RDM is similar to the recognition sequence for attachment of the ubiquitin-like protein, small ubiquitin-like modifier-1 (SUMO-1), and the conserved lysine residue of each C/EBP RDM served as an attachment site for SUMO-1. SUMO-1 attachment decreased the inhibitory effect of the C/EBP epsilon regulatory domain, suggesting that sumoylation may play an important role in modulating C/EBP epsilon activity as well as that of the other C/EBP family members.

The CCAAT/enhancer binding proteins (C/EBP) 1 form a subgroup of the basic region/leucine zipper superfamily of transcription factors (1,2). Six members of this family have been identified in mammalian cells, and four of these (C/EBP␣, C/EBP␤, C/EBP␦, and C/EBP⑀) possess similar arrangements of functional domains (3). Each protein possesses a bipartite carboxyl-terminal DNA binding domain that mediates homoand heterodimerization and sequence-specific binding to a common DNA recognition motif (4,5). The C/EBP family members have been implicated in the regulation of numerous cell processes, including differentiation, proliferation, and apoptosis. For example, mice bearing a targeted mutation in the C/EBP⑀ gene exhibit immune defects that result from defective neutrophil function (6). The phenotype of C/EBP⑀-deficient mice resembles that seen in humans with neutrophil-specific granule deficiency, and mutations in the C/EBP⑀ gene have been detected in patients with neutrophil-specific granule deficiency (7). Likewise, studies of mice carrying mutations in other C/EBP genes have shown that members of this family are critical for the development or function of numerous cell types, including hepatocytes, adipocytes, and macrophages (1,2).
Although the C/EBPs possess similar DNA binding specificities and dimerization properties, each protein exhibits unique functional properties in vivo. The appearance of specific phenotypes in each C/EBP-deficient mouse supports the contention that these proteins do not have fully redundant functions. Although this observation could be partially explained by nonoverlapping expression patterns of the different proteins, direct evidence for specific functions has come from experiments where the coding sequence for one C/EBP has been replaced with that encoding a different family member. For example, the insertion of the C/EBP␤ coding sequence into the C/EBP␣ locus rescued hepatic-specific defects in mice but could not rescue defects in adipose tissue (8).
Several studies have focused on the identification of functional domains within each C/EBP protein. As mentioned above, four family members (C/EBP␣, C/EBP␤, C/EBP␦, and C/EBP⑀) possess very similar structures. A potent multimodule transcriptional activation domain (AD) is located at the amino terminus of each protein (3,9,10), and domains that negatively modulate transcriptional activity have been identified in C/EBP␣, C/EBP␤, and C/EBP⑀ (3,(11)(12)(13). In C/EBP⑀, we defined two inhibitory domains (RDI and RDII) and identified a short amino acid motif that was critical for RDI function (14). Interestingly, negative regulatory domains have been identified in a number of other nuclear transcription factors, including c-Fos (15), c-Myb (16,17), Sp1/3 (18), and members of the nuclear hormone receptor superfamily (19,20). The inhibitory functions of these various domains within their cognate proteins have been variously attributed to inhibition of DNA binding, transcriptional activation, or synergy with factors bound to adjacent promoter elements (3,19). However, the relationship between these domains is currently unknown. A common feature of these domains is that many have been suggested to be binding sites for currently uncharacterized regulatory proteins (14,19).
Here, we have extended our previous studies on the RDI inhibitory domain of C/EBP⑀ (14). We show that a conserved five-amino acid motif within the inhibitory domains of C/EBP⑀, C/EBP␣, C/EBP␤, and C/EBP␦ is critical for regulatory domain function. Further, we demonstrate that a lysine residue within this motif is a site for covalent attachment of the small ubiquitin-like modifier-1 (SUMO-1) protein (21) and that attachment of SUMO-1 to this residue decreases the transcriptional inhibitory function of the regulatory domain. We propose that sumoylation of C/EBP proteins is a conserved modification that modulates the activity of each family member.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Expression plasmids for each full-length C/EBP were constructed using the pMEX expression vector (5). The coding sequence for the FLAG epitope tag (Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys) was inserted immediately following the translation initiation codon, and the coding sequence of each C/EBP was inserted as NcoI/ HindIII fragments. In these and all subsequent plasmids, mutations were introduced using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA) using conditions recommended by the manufacturer. A restriction enzyme recognition sequence was introduced where possible to aid in the detection of plasmids carrying the desired mutation. All constructs were sequenced at the Texas Tech University Biotechnology Core Laboratory to ensure that unintended changes were not introduced during the mutagenesis procedure. Gal4 fusion genes were constructed in the Gal0 plasmid (14). The generation of each of the wildtype Gal4 fusion plasmids bearing N-terminal regions of C/EBP␣, C/EBP␤, and C/EBP⑀ has been described previously (3,14). Gal4 fusions bearing amino acids 1-92 and 1-142 of C/EBP␦ were generated using EagI and BssHII restriction enzyme sites present in the mouse cDNA sequence. The coding regions for individual domains were generated as BamHI/BglII fragments and assembled into Gal4 fusion constructs by standard subcloning procedures. The VP16 AD segment inserted into Gal4 fusions consisted of amino acids 429 -456 (14). The FLAG-SUMO-1 expression vector (22) was a kind gift of Dr. Giuseppina Nucifora (University of Illinois at Chicago). A vector expressing a nonprocessable version of SUMO-1 was generated by site-directed mutagenesis, changing glycine 97 to alanine. The Tet-VP16 and TRE-UAS-TATA plasmids (23) were obtained from Dr. Noel Buckley (University of Leeds). The Gal4-responsive reporter plasmid (G5E1bLuc), which consists of five copies of the Gal4 binding site upstream of the adenovirus E1b minimal promoter, was described previously (3).
Cell Culture and Transfections-The monkey kidney COS-1 cell line was cultured in Dulbecco's modified Eagle's medium (BioWhittaker, Walkersville, MD) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT). All luciferase-based transfections were performed in 3.5-cm dishes using 2 ϫ 10 5 cells/well, and transfections for preparing extracts for Western blotting were performed in 10-cm dishes using 1.5 ϫ 10 6 cells/dish. Transfections were performed with the Effectene transfection reagent using conditions recommended by the manufacturer (Qiagen, Valencia, CA). The optimal concentration of each plasmid was determined empirically for each type of transfection experiment. Transfections that examined the activity of Gal4 fusion proteins contained 150 ng of the Gal4 expression plasmid and 150 ng of the G5E1bLUC reporter. Transfections that examined the ability of the C/EBP⑀ RD to inhibit the activity of the Tet-VP16 activator in trans contained 250 ng of the TRE-UAS-TATA luciferase reporter, 200 ng of the Tet-VP16 expression plasmid, and 250 ng of the Gal4-C/EBP⑀ expression plasmid. Transfections that tested the effect of co-expression of wild type or mutant SUMO-1 on RD function contained 150 ng of G5E1bLUC, 250 ng of the Gal4-C/EBP⑀ expression plasmid, and 20 or 40 ng of the SUMO-1 expression plasmid. The total DNA concentration was normalized in this last experiment using empty pCMV expression plasmid. Transfections for Western extract preparation contained 3 g of the appropriate expression plasmid. A control plasmid containing the Renilla luciferase gene under the control of the thymidine kinase promoter was included in all experiments for normalization of transfection efficiencies. All transfections were performed in triplicate at least three times and in some cases up to seven times. Statistical analyses were carried out using a two-tailed Student's t test. Luciferase assays were performed using the Dual-Luciferase Reporter Assay kit (Promega, Madison, WI).
Nuclear Extract Preparation and Western Blotting-Nuclear extracts were prepared from transfected COS-1 cells as described previously (24) except when isopeptidase inhibitors were used. In these samples, the cells were washed twice in phosphate-buffered saline containing 1 g/ml E-64 (Roche Molecular Biochemicals) and lysed in Dignam Buffer A (25) supplemented with 10 g/ml E-64. Samples were separated on 12 or 15% denaturing polyacrylamide gels. Proteins were transferred to nitrocellulose membranes, and immune detection was performed using the Supersignal chemiluminescence detection kit (Pierce) as described previously (26). The anti-Gal4 DNA binding domain (DBD) monoclonal antiserum (RK5C1) was obtained from Santa Cruz Biotechnology, Inc.
(Santa Cruz, CA), and the anti-FLAG polyclonal antiserum was obtained from Sigma. The C/EBP⑀-specific rabbit polyclonal antiserum was raised against a specific epitope in the C-terminal portion of rat C/EBP⑀.

Identification of an Evolutionarily Conserved
Sequence within the C/EBP⑀ RDI-In a previous study, we performed a structure/function analysis of the C/EBP⑀ polypeptide and identified several autonomous domains, including two activation and two inhibitory domains (14). One inhibitory domain, termed RDII, contained several potential sites for phosphorylation by members of the MAP kinase superfamily; thus, RDII activity is probably dependent on phosphorylation of C/EBP⑀ in response to upstream signaling pathways. However, a second inhibitory domain, termed RDI, did not contain obvious signatures for post-translational modifications, and thus the mechanism of action of this domain remained unclear. Initial comparisons of the sequence of RDI with inhibitory domains from other transcription factors revealed weak sequence similarities, suggesting the existence of conserved inhibitory domains within multiple transcriptional regulatory proteins (14). We therefore performed extensive sequence comparisons of the RDI sequence with protein sequences in public databases. These searches revealed the existence of a five-amino acid motif that is conserved in three other members of the C/EBP family, C/EBP␣, C/EBP␤, and C/EBP␦ (see Table I), which represents the only conserved sequence within the experimentally defined minimal RDI domain (see sequence comparison in Ref. 2). This motif is defined by the consensus sequence (I/V/ L)KXEP and will be referred to hereafter as the regulatory domain motif (RDM). The availability of peptide sequences for each C/EBP protein from multiple species permitted exploration of the evolutionary conservation of this sequence. In all vertebrate species in which orthologous relationships to the mammalian C/EBP proteins could be established, the RDM sequence was conserved, in some cases exactly (Table I). For example, in C/EBP␣, the RDM sequence IKQEP was perfectly conserved from humans to fish. Significantly, C/EBP proteins from nonvertebrates (two Drosophila species (27) and two Aplysia species (28,29)) also contain sequences that conform to the RDM consensus. These comparisons suggest that the RDM might be an important component of an evolutionarily conserved transcriptional inhibitory domain in C/EBP proteins.
Functional Analysis of C/EBP RDMs-To test the contribution of the RDM to RDI function in C/EBP⑀, we performed mutagenesis studies using Gal4-C/EBP⑀ fusion proteins that either contained or lacked RDI. Our earlier studies mapped the minimal RDI element to the region between amino acids 97 and 128 of the mouse C/EBP⑀ protein, with the RDM located between amino acids 120 and 124 (14). Therefore, we performed mutagenesis on a Gal4 construct containing amino acids 1-128 of C/EBP⑀ and compared the activity of the wild type and mutant proteins to a fully active Gal4-C/EBP⑀ fusion protein that only contained activation domain sequences located between amino acids 1 and 64 (Fig. 1A). Each construct was co-transfected into COS-1 cells along with a Gal4-responsive luciferase reporter plasmid. In agreement with previous data, the inclusion of sequences encompassing amino acids 65-128 decreased the activity of the fusion protein to less that 10% of Gal4⑀-(1-64). Each residue within the RDM was first changed to alanine to test their respective contribution to this inhibitory effect. Changes introduced at positions 1, 2, and 4 essentially abolished the inhibitory function, whereas changes introduced at positions 3 and 5 were silent. Thus, residues that were critical for RD function were those that exhibited the highest degree of sequence conservation within the RDMs of each C/EBP. The identity of the residue at position 3 of the RDM appears not to be fixed, whereas the proline at position 5 may provide important structural information but may not contribute to RDM function per se. We also changed the three central charged amino acids within the C/EBP⑀ RDM to arginine. As before, changes at positions 2 and 4 diminished inhibitory domain function, whereas substitution of arginine for the glutamic acid residue at position 3 had little effect. Importantly, these data indicate that the presence of a lysine residue at position 2 is critical for RDM function.
We next extended these studies to include the other mammalian C/EBPs listed in Table I. Previous studies had mapped transcriptional inhibitory domains in C/EBP␣ and C/EBP␤ that overlapped the region encompassing their RDMs (3,12,13); however, no data were available for C/EBP␦. Therefore, we constructed pairs of Gal4 fusion constructs containing either minimal N-terminal ADs from each protein or extended regions that include the RDM from each protein ( Fig. 2A). These constructs were introduced into COS-1 cells, and the relative activities of the expressed proteins were examined as described above. In each case, the inclusion of RDM-containing regions resulted in significant inhibition of AD function ( Fig. 2A) to levels between 0.4 and 3.7% of the uninhibited control protein.
Again, mutations were introduced into the RDM sequences of each protein to test whether they contributed to the inhibitory function, although these experiments only focused on the perfectly conserved lysine and glutamic acids residues at positions 2 and 4. In each case, proteins bearing the indicated mutations were significantly more active than the wild type protein, particularly in the case of C/EBP␣ and C/EBP␦. Although the introduction of mutations into the RDM of C/EBP␤ resulted in significant increases in the activity of the resultant proteins, these proteins were still significantly less active than the protein containing the AD alone (Gal4-␤-(1-83)). Further analysis of the inhibitory domain of C/EBP␤ indicated the existence of additional inhibitory sequences in this protein, 2 and further studies are under way to define these sequences. Nevertheless, these studies clearly demonstrated the critical importance of the RDM for inhibitory domain function in each C/EBP.
The C/EBP⑀ RDI Inhibits AD Function Both in Cis and in Trans-We previously provided evidence that the RDI element of C/EBP⑀ functioned as a site for protein/protein interactions (14). To further investigate the mechanism of action of this domain, we next tested the ability of this domain to inhibit the action of a heterologous activation domain in both cis and trans arrangements. Single or multiple copies of the C/EBP⑀ RDI domain (amino acids 65-128) were attached to a fusion protein containing the Gal4 DBD and the VP16 AD, and the activity of the resulting proteins was tested as before (Fig. 3A). A Gal4-VP16 protein carrying a single RDI element possessed less than 10% of the activity of the parental protein, and attachment of additional RDI elements yielded further inhibition that was proportional to the number of RDI elements. We next tested whether the RDI element functioned only when located adjacent to an AD by inserting a mutant form of the RDI element between the VP16AD and one or two copies of the wild type RDI element. Attachment of a mutant RDI element resulted in a modest decrease in the activity of the Gal4-VP16 fusion protein; however, the further attachment of one or two The peptide sequences of four C/EBP family members (C/EBP␣, C/EBP␤, C/EBP␦, and C/EBP) from a variety of species were compared to identify residues conserved with the region encompassing the minimal negative regulatory domain of C/EBP (amino acids 97-128 of the mouse protein). A five-amino acid motif was identified that was highly conserved in each C/EBP in a variety of vertebrate species and also in nonvertebrate C/EBPs. This sequence has been named the regulatory domain motif (RDM), and the derived consensus is shown at the bottom right. The species from which each sequence was derived, the 15 functional RDI elements in this more distal arrangement led to dramatic transcriptional repression (Fig. 3A).
To examine RDI function in trans, Gal4 fusion proteins carrying single or triple copies of wild type or mutant RDI elements, without any AD sequences, were generated. These proteins were tested for their ability to inhibit the activity of an activator protein consisting of the DNA binding domain from the Tet repressor fused to the VP16 AD (14). The effect of each RDI-containing fusion protein was compared with the Gal4 DBD alone on a reporter construct containing seven copies of the Tet response element (TRE) in a distal position and five Gal4 binding sites located in a proximal position (TRE-UAS-TATA; Fig. 3B). Coexpression of Gal4 fusion proteins containing single or triple RDI elements decreased the measured activity of the Tet-VP16 activator protein to ϳ10 and 1% of the control value, respectively. This effect was dependent on the integrity of the RDM, since fusion proteins containing mutant RDI elements had no inhibitory effects in this assay. These data suggest that the RDM can recruit proteins to a promoter that inhibit transcription in a position-independent and dosedependent manner.
Identification of Functional RDM-containing Transcriptional Inhibitory Domains in Jun Family Members-Data base searches revealed the existence of several hundred proteins that contained sequences similar to the RDM sequences found in the C/EBPs. 3 The majority of these proteins were nuclear proteins with known or predicted functions in transcription, replication, or DNA repair. Therefore, we were interested in determining whether the proteins identified in these searches contained transcriptional inhibitory domains similar to the domains described above in the C/EBPs. We selected the three members of the Jun family of bZIP transcription factors for these studies, since each contains an RDM-like sequence located adjacent to the N-terminal end of their DNA binding domains (Fig. 4A). Short segments of each Jun protein were attached to a Gal4-VP16 fusion protein as described above. The effect of single or, in some cases, multiple copies of each Jun module on the transcriptional activity of the resulting fusion proteins was assessed in transfection assays in COS-1 cells. Single copies of the c-Jun and JunD modules efficiently inhibited the activity of the Gal4-VP16 protein, whereas the JunB module had less of an effect (Fig. 4B). Inhibition was dependent 3 S. C. Williams, unpublished observations.

FIG. 1. Mapping of RDM residues critical for regulatory domain function in C/EBP⑀.
A, the transcriptional activity of Gal4 fusion protein proteins bearing either the first 64 or 128 amino acids of mouse C/EBP⑀ were examined by transient transfection with a Gal4responsive luciferase reporter plasmid into COS-1 cells. The locations of the tripartite N-terminal activation domain (ADI) and the RDI of C/EBP⑀ are labeled. A protein containing amino acids 65-128 (which includes the RDM as shown) possessed less than 10% of the activity of the Gal4-⑀-(1-64) fusion protein (the activity of Gal4-⑀-(1-64) was arbitrarily set to 100). Point mutations were introduced into the RDM sequence to assess the relative contribution of each residue to regulatory domain function. In the first series, each amino acid in the RDM was changed to alanine within the context of the Gal4-⑀-(1-128) fusion protein, whereas in the second series the central three amino acids were changed to arginine. The relative activity of each protein is shown with asterisks to denote those proteins that displayed significantly (p Ͻ 0.05) higher activity than the wild type Gal4-⑀-(1-128) protein (i.e. proteins whose activity was derepressed). B, a representative Western blot demonstrating that activity differences were not directly related to differences in steady-state protein levels.

FIG. 2. The RDM is a functional component of negative regulatory domains in C/EBP␣, C/EBP␤, and C/EBP␦.
A, the activities of Gal4 fusion proteins carrying either the N-terminal AD of C/EBP␣, C/EBP␤, or C/EBP␦ or extended sequences encompassing the RDM from each protein were tested as described above in COS-1 cells. In each case, the inclusion of RDM-containing sequences caused a significant decrease in activity, with the magnitude of the change varying between proteins. Mutations were introduced that changed either of the conserved residues at positions 2 and 4 of the RDM, and activities of the mutant proteins were compared back with the longer, repressed protein in each case. The relative activity of each protein is listed with asterisks to denote those proteins whose activity was significantly different from the wild type, RDM-containing protein (p Ͻ 0.05). B, representative Western blot showing steady-state levels of the indicated proteins. on the integrity of the RDM, since changing the lysine residue at position 2 of the c-Jun RDM to alanine nearly abolished its inhibitory effect (compare G4V-c-Jun with G4V-(c-Jun) m ). To further test whether a functional inhibitory domain exists in JunB, a protein bearing two copies of the RDM-containing element was tested. Strong inhibition of the activity of the chimeric protein was observed at a level similar to the effect of two copies of the c-Jun element. Thus, a functional RDMcontaining inhibitory domain is present in each Jun family member.
The C/EBP RDM Sequences Are Attachment Sites for SUMO-1 -Several recent studies have defined a consensus sequence for attachment of the SUMO proteins as (I/V/L)KXE (21), which is contained within the consensus RDM sequence described above. In addition, the lysine residue in the c-Jun RDM sequence is an attachment site for SUMO-1 (30). Therefore, we tested whether the C/EBP family members could be modified by covalent attachment of SUMO-1. Initially, expression vectors encoding two isoforms of C/EBP⑀ (p29 and p32) were cointroduced into COS-1 cells either with or without an expression vector for SUMO-1, and cell extracts were prepared and analyzed by Western blotting. The C/EBP⑀ proteins and SUMO-1 both carried FLAG epitope tags, and thus duplicate Western blots were probed with antisera directed against either FLAG or C/EBP⑀ to detect these proteins. When expressed alone, the C/EBP⑀ proteins migrated at the expected positions, with multiple closely spaced bands perhaps representing differentially phosphorylated isoforms (Fig. 5A, lanes 1 and 2). In the presence of SUMO-1 and isopeptidase inhibitors, slower migrating complexes were observed whose positions were dependent on the mass of the C/EBP⑀ isoform (lanes 4 and 5). These slower migrating complexes were detected using both anti-C/EBP⑀ and anti-FLAG antisera. However, the slower migrating complexes were not detected following transfection of an expression plasmid encoding a C/EBP⑀ protein bearing alanine substitutions in the KEE core of the RDM (lanes 3 and 6).
To confirm that the slower migrating complexes represented

FIG. 3. The C/EBP⑀ regulatory domain inhibits transcription in a copy-dependent manner both in cis and in trans.
A, inhibition in cis. A series of Gal4 fusion plasmids containing the potent AD from the viral VP16 protein with or without single or multiple copies of the C/EBP⑀ regulatory domain (RDI; amino acids 64 -128) were constructed. A second series of plasmids was also constructed in which the regulatory domain closest to the AD contained an inactivating mutation in the RDM. Each protein was expressed in COS-1 cells, and its activity was assessed on a Gal4-responsive luciferase reporter plasmid. The relative activity of each protein is listed with asterisks to denote those proteins with significantly (p Ͻ 0.05) lower activity than the Gal4-VP16 parental protein, whose activity was set to 100. B, inhibition in trans. A series of Gal4 fusion plasmids bearing either one or three copies of the wild type or RDM mutant RDI (RDIm) domain were constructed. Each protein or the parental Gal4 DBD alone was co-transfected into COS-1 cells along with a plasmid encoding a Tet-VP16 activator protein and a reporter plasmid containing seven TREs and five Gal4 binding sites (UAS). The relative activity of the reporter plasmid in the presence of each pair of proteins is listed. *, p Ͻ 0.05.

FIG. 4. The RDM is a component of a conserved negative regulatory domain in three members of the Jun family.
A, sequence comparisons revealed the presence of RDM-like sequences located Nterminal to the bZIP DBD of c-Jun, JunB, and JunD. The amino acid sequence of each protein in the region is shown with the putative RDM highlighted. The proteins are relatively nonconserved in the surrounding region, except for the five amino acids immediately following the RDM. Position 1 in the JunB RDM is occupied by phenylalanine, a hydrophobic amino acid that does not occur in any C/EBP RDM. Short segments (26 -28 amino acids) of each mouse Jun cDNA was amplified by PCR, and the RDM-containing region was placed adjacent to the VP16 AD in Gal4 fusion proteins as diagrammed. B, the activity of a series of Gal4-VP16 fusion proteins carrying the single, double, or mutant copies of the Jun sequences was tested by co-transfection with a Gal4-responsive luciferase reporter plasmid in COS-1 cells. In the case of c-Jun, one construct carried a lysine to alanine substitution in the conserved lysine at position 2 of the RDM (indicated with an X in the c-Jun segment). The activity of each protein is listed as described before. *, p Ͻ 0.05. BR, basic region; LZ, leucine zipper. SUMO-1-modified C/EBP⑀, additional co-transfection experiments were performed using a C/EBP⑀ protein (K121A) bearing a specific mutation in the lysine presumed to be the SUMO-1 attachment site and a mutant SUMO-1 protein that cannot be processed for attachment (SUMOGA). As before, the slower migrating complex (which resolved into three distinct species in this gel) was observed using the wild type but not the K121A mutant protein (Fig. 5B, compare lanes 2 and 4). The appearance of the slower migrating complex(es) was dependent on the presence of a functional SUMO-1 protein, since complex formation was not observed when the SUMOGA mutant protein was coexpressed at levels equivalent to the wild type protein (lane 3).
We next tested whether the other C/EBP proteins could also be modified by SUMO-1 attachment within their RDM sequences. FLAG-tagged wild type C/EBP␣, C/EBP␤, and C/EBP␦ proteins and C/EBP␣ and C/EBP␤ proteins bearing substitutions at the conserved lysine residue in their RDMs were co-expressed with wild type and mutant SUMO-1 proteins and analyzed by Western blotting as before (Fig. 6, A and B). In each case, a specific slower migrating complex was observed that was dependent on both the integrity of the RDM sequence and on the presence of an attachment competent SUMO-1 protein. Therefore, sumoylation appears to be a modification common to each of the C/EBPs examined here and may represent an important regulatory modification of C/EBP transcriptional activity.
Sumoylation Decreases the Inhibitory Function of the C/EBP⑀ RDI Element-Having shown that the RDM is a target for FIG. 5. The C/EBP⑀ RDM is a target for attachment of SUMO-1. A, three FLAG-tagged C/EBP⑀ proteins were expressed in the absence (lanes 1-3) or presence (lanes 4 -6) of SUMO-1 in COS-1 cells. The p29 and p32 isoforms of C/EBP⑀ are synthesized by translational initiation at one of two in-frame AUGs in the C/EBP⑀ cDNA, and the p34m protein carries alanine substitutions in the central KEE motif of the RDM. Nuclear extracts were prepared in the presence of isopeptidase inhibitors and analyzed by Western blotting with antisera directed against either the FLAG epitope or a specific epitope in C/EBP⑀. The migration of bands corresponding to unmodified and sumoylated C/EBP⑀ proteins is indicated, and the position of molecular weight markers is shown on the right. B, sumoylation of C/EBP⑀ required the lysine within the RDM and a functional SUMO-1 protein. Wild type C/EBP⑀ or a mutant protein carrying a specific alanine substitution in lysine 121 was co-expressed with a wild type SUMO-1 protein (SUMOGG) or a nonprocessable SUMO-1 mutant protein (SUMOGA). Nuclear extracts were prepared and analyzed by Western blotting with an anti-FLAG antiserum as described above.
FIG. 6. C/EBP␣, C/EBP␤, and C/EBP␦ are all modified by covalent attachment of SUMO-1. A, plasmids encoding FLAG-tagged wild type C/EBP␣ or C/EBP␤ or proteins in which the conserved lysine in the RDM was mutated were co-expressed with SUMOGG or SUMOGA in COS-1 cells. Nuclear extracts were prepared and analyzed by Western blotting using an anti-FLAG antiserum as described above. The band migrating slightly slower than the main C/EBP␤ species is likely to be a phosphorylated form of C/EBP␤. B, a similar experiment was performed using C/EBP␦, except that a mutant C/EBP␦ protein was not used.
SUMO-1 attachment, we next tested the consequence of this modification on the function of the RDI element of C/EBP⑀. Increasing amounts of the wild type and mutant SUMO proteins were co-expressed with Gal4-C/EBP⑀ fusion proteins bearing either wild type or mutant RDM sequences (Fig. 7). Co-expression of wild type SUMO-1 resulted in a consistent increase in the activity of the wild type Gal4-C/EBP⑀ fusion protein, with a maximal 7-fold increase observed with co-transfection of 40 ng of the SUMO-1 expression vector. By contrast, co-transfection of an expression vector encoding the SUMOGA mutant protein did not result in a significant change in Gal4-C/EBP⑀ activity. Furthermore, the activity of the RDM mutant Gal4-C/EBP⑀ protein was essentially unaffected by co-expression of either the wild type or mutant SUMO-1 protein. Therefore, attachment of SUMO-1 appears to relieve the inhibitory effect of the C/EBP⑀ RDM. DISCUSSION These studies have identified an evolutionarily conserved transcriptional regulatory domain within four members of the C/EBP family that is capable of inhibiting the activity of an adjacent transcriptional activation domain both in cis and in trans. These domains are characterized by the presence of a conserved five amino acid motif ((I/V/L)KXEP) that we named the RDM. Similar functional domains containing RDM-like sequences are also present in another family within the bZIP superfamily, the Jun proteins. The RDM is similar to the consensus sequence for covalent attachment of the SUMO-1 protein, and we demonstrated that C/EBP␣, C/EBP␤, C/EBP␦, and C/EBP⑀ can all be modified by attachment of SUMO-1 to the lysine at position 2 of the RDM. Attachment of SUMO-1 to the RDM of C/EBP⑀ decreased the inhibitory function of the regulatory domain of C/EBP⑀, suggesting that sumoylation may be a common mechanism for regulating the activity of C/EBP and perhaps other bZIP proteins.
The existence of sequences within C/EBP proteins with the capacity to inhibit transcription was first indicated by early structure/function studies on C/EBP␣ (12). These studies identified a domain, originally named the "attenuator," that was located between amino acids 108 and 170 of the rat C/EBP␣ protein (12). Subsequent studies identified inhibitory domains in mammalian (3) and avian (13) C/EBP␤ proteins and in mammalian C/EBP⑀ (11,14). Our data now show that the RDM is a functional component of the inhibitory domains of each of these proteins. Furthermore, the conservation of RDM-like sequences in each C/EBP throughout evolution supports the potential importance of this motif for C/EBP function. This is perhaps best exemplified by the existence of RDM-like sequences in nonvertebrate C/EBPs from Drosophila and Aplysia, although functional studies on these potential RDMs have not yet been performed.
The demonstration that the RDM of C/EBPs is a site for covalent attachment of the SUMO-1 protein may have significant implications for understanding the mechanisms by which this domain might control the activity of this family of transcription factors. SUMO-1 is a 102-amino acid protein that displays ϳ18% sequence identity to ubiquitin (21,31). SUMO-1 is attached to the ⑀ amino group of lysine residues in target proteins using a mechanism that is functionally analogous to ubiquitination. SUMO-1 is synthesized as a precursor protein that is first cleaved by C-terminal hydrolases at the C-terminal side of a diglycine sequence. It is then attached through a thioester bond to a heterodimeric E1 activating enzyme (SAE1/ SAE2 in mammalian cells) and transferred to the Ubc9 E2 enzyme (21). Although sumoylation can occur in vitro in the absence of an E3 ligation activity, recent studies have identified a number of proteins that possess SUMO E3 ligase activity that may increase the efficiency and/or selectivity of the sumoylation reaction (32,33). Sumoylation is a reversible reaction, and a number of isopeptidases have been identified that possess desumoylation activities (21). The list of proteins that can be targets for sumoylation continues to grow and consists primarily of nuclear proteins in mammalian cells, consistent with our observation that most RDM-containing proteins are likely to function in the nucleus. Known targets include RanGAP1 (34), PML (35), IB␣ (36), p53 (30) and androgen receptor (37). Significantly, negative regulatory domains in other transcription factors, including c-Myb and c-Jun have also been identified as sumoylation targets, and the lysine residue within the c-Jun RDM sequence identified in our studies is the demonstrated site for covalent attachment of SUMO-1 (30,38). Therefore, sumoylation of negative regulatory domains in transcription factors appears to be a common mechanism for regulating their activity.
Although our data indicate that sumoylation may modulate the activity of C/EBP proteins, the mechanism underlying this effect is not yet clear. Sumoylation does not appear to target proteins for degradation, an important role for ubiquitination (39). Instead, sumoylation may modify the activity of target proteins via a number of different mechanisms (21,31). First, sumoylation may block alternative lysine-targeted modifications such as acetylation or ubiquitination. For example, sumoylation of IB␣ occurs on a lysine residue that is also targeted for ubiquitination, and thus attachment of SUMO stabilizes IB␣ by blocking ubiquitination and subsequent degradation (36). Stabilization of IB␣ results in prolonged inhibition of NF-B activity and thus may have strong significance for regulation of NF-B activation in inflammatory and other reactions. Although both Drosophila and Aplysia C/EBPs have been shown to be targets for ubiquitination (40,41), ubiquitination of mammalian C/EBPs has not yet been reported. In addition, we did not detect any significant differences in the steady state level of Gal4 fusion proteins carrying wild type or mutant RDM sequences, suggesting that sumoylation is unlikely to modulate the stability of C/EBP proteins. Likewise, although transcription factor acetylation is emerging as a powerful regulatory mechanism (42), there is no evidence as yet that C/EBPs are acetylated. Sumoylation may also affect the subcellular, or more specifically the subnuclear localization of target proteins as, for example, appears to be the case for PML (43). Both C/EBP␣ and C/EBP␤ are known to associate with the nuclear matrix (44) and to localize to specific dotlike structures within the nucleus (45,46), although whether these domains represent sites where C/EBP proteins are active or inactive remains uncertain. C/EBP⑀ is also localized to specific nuclear subdomains 4 ; however, this localization appears to be unaffected by the integrity of the RDM. Nevertheless, further definition of the specific subnuclear location of C/EBP proteins may provide important insights into the role of sumoylation in controlling C/EBP⑀ activity.
Finally, sumoylation modulates the transcriptional activity of a growing number of target proteins, including androgen receptor (37), p53 and c-Jun (30), c-Myb (38), and two members of the histone deacetylase family (47,48). Again, the exact mechanisms underlying SUMO-dependent regulation of transcriptional activity are poorly understood, but they may involve modulation of critical protein/protein interactions. Previous models to explain the action of the inhibitory domains of C/EBP␤ envisioned the formation of intermolecular interactions that prevented AD access to the transcriptional machinery (3,13). These interactions may be disrupted by sumoylation, thereby leading to activation of the modified protein.
Alternatively, the inhibitory domain may be a site for interactions with an as yet unknown inhibitory protein, and sumoylation results in dissociation of this protein. This latter model is more consistent with our current data, particularly the ability of the C/EBP⑀ RDM to inhibit the activity of an adjacent AD in trans. One attractive set of candidates for this function are the protein inhibitor of activated STAT (PIAS) proteins (49). PIAS proteins were first identified as inhibitors of the STAT family of transcription factors (50) and possess SUMO E3 ligase activity (32). There are six members of the PIAS family, and individual family members can act as both positive and negative co-regulators for a variety of transcription factors from different families (32). Experiments are currently under way to test whether PIAS proteins interact with the inhibitory domain of the C/EBP proteins and to determine the consequence of such interactions on RD function and sumoylation. Furthermore, there are two other SUMO proteins in mammalian cells, SUMO-2 and SUMO-3 (21), and we are interested in determining whether these two proteins can also be attached to C/EBP RDM sequences. Clearly, elucidation of the full complement of RDinteracting factors will be critical for our understanding of the contribution of inhibitory domains to C/EBP function in vivo.