Covalent Modification of p73a by SUMO-1 TWO-HYBRID SCREENING WITH p73 IDENTIFIES NOVEL SUMO-1-INTERACTING PROTEINS AND A SUMO-1 INTERACTION MOTIF*

Two-hybrid screening in yeast with p73a isolated SUMO-1 (small ubiquitin-like modifier 1), the enzyme responsible for its conjugation, Ubc-9, and a number of novel SUMO-1-interacting proteins, including thymine DNA glycosylase, PM-Scl75, PIASx, PKY, and CHD3/ZFH. A subset of these proteins contain a common motif, hhXSXS/Taaa, where h is a hydrophobic amino acid and a is an acidic amino acid, that is shown to interact with SUMO-1 in the two-hybrid system. We show here that p73a, but not p73b, can be covalently modified by SUMO-1. The major SUMO-1-modified residue in p73a is the C-terminal lysine (Lys). The sequence surrounding this lysine conforms to a consensus SUMO-1 modification site b(X)XXhKXE, where b is a basic amino acid. SUMO-1-modified p73 is more rapidly degraded by the proteasome than unmodified p73, although SUMO-1 modification is not required for p73 degradation. SUMO-1 modification does not affect the transcriptional activity of p73a on an RGC-luciferase reporter gene in SK-N-AS cells. Instead, SUMO-1 modification may alter the subcellular localization of p73, because SUMO-1modified p73 is preferentially found in detergent-insoluble fractions. Alternatively, it may modulate the interaction of p73 with other proteins that are substrates for SUMO-1 modification or which interact with SUMO-1, such as those identified here.

Covalent modification of proteins is widely used as a way of modifying their stability, activity, or localization. Examples of this are phosphorylation, acetylation, lipid modification, or glycosylation. Modification by covalent linkage to a second "tagging" protein was first observed with ubiquitin, a 76-amino acid polypeptide that is covalently linked to lysine residues in an acceptor protein by an enzymatic system involving two to three ubiquitin-activating and -conjugating enzymes (E1, E2, and E3). 1 Subsequent poly-ubiquitination usually signals the modified protein for degradation by the proteasome (1). Alternate outcomes for ubiquitinated proteins are activation or transport via an intracellular membrane vesicular system (2).
It has now become apparent that several other "ubiquitinlike" tagging molecules exist that are conjugated using enzy-matic systems similar but nonidentical to those used by ubiquitin (3,4). Two groups initially determined the nature of a modification of the Ran GTPase-activating protein (RanGAP1) involved in the interaction of this protein with RanBP2/ Nup358 at the nuclear pore complex (5,6). They called the modifying molecule GMP1 (GAP-modifying protein 1) or SUMO-1 (small ubiquitin-like modifier 1). SUMO-1 has since been identified several times as an interacting partner in the yeast two-hybrid system and given different names: with the promyelocytic leukemia gene product (PML) (PIC-1) (7), with the death domain of the FAS antigen or TNF-receptor (sentrin, DAP-1) (8,9), and with the RAD51 and RAD52 proteins involved in DNA recombination and repair (UBL-1) (10).
A budding yeast homologue of SUMO-1 (Smt3p) was identified by Meluh and Koshland (11) as a suppressor of mutations in Mif2 (mitotic instability factor 2), a protein thought to be the yeast equivalent of the CENP-C mammalian centromere protein. One putative role for SUMO-1/Smt3p is thus in assembly or maintenance of the centromere/kinetochore structure involved in chromosome segregation. Similarly, the fission yeast (Schizosaccharomyces pombe) equivalent of Smt3p, Pmt3p, has recently been shown to be involved in chromosomal segregation and the control of telomere length (12).
In yeast and mammalian cells, there is very little free Smt3p/ SUMO-1. Most (Ͼ90%) of the SUMO-1 detected on Western analysis of extracts from mammalian cells is conjugated to the RanGAP1 nuclear pore protein (18). Other proteins subject to modification by SUMO-1 are the PML and Sp100 proteins, which form part of the nuclear structures known as PODs (PML oncogenic domains) or nd10s (19). For PML, it has been shown that SUMO-1 modification is essential for its localization in PODs, with free PML being found in the nucleoplasm. Another substrate for SUMO-1 is I␤␣ (20), an inhibitor of NF␤, which is modified by SUMO-1 on the lysine residue also modified by ubiquitin. SUMO-1 modification prevents ubiquitination and thus results in stabilization of I␤␣ and consequently in inhibition of NF-␤ (20).
In the present communication, we describe SUMO-1 modification of the p53-related p73␣ protein. p53 is the most widely * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Tel.: 33-5-61-00-41-67; Fax: 33-5-61-00-40-01; E-mail: adrian.minty@sanofi-synthelabo. com. 1 The abbreviations used are: E1, ubiquitin-activating enzyme; E2, ubiquitin; E3, ubiquitin-protein isopeptide ligase; PML, promyelocytic leukemia gene product; POD, PML oncogenic domain; HDAC, histone deacetylase complex. studied tumor suppressor and is mutated in over 50% of human tumors (21). It plays a key role in both the regulation of cell cycle checkpoints and the initiation of apoptotic cell death in response to DNA damage. The activity of p53 has been shown to be finely tuned by a variety of post-translational modifications (phosphorylation, acetylation, and glycosylation) and to be highly sensitive to conformational changes (22). The p53 molecule contains a number of well defined domains including an N-terminal transcriptional activation domain and a central core, corresponding to the DNA-binding domain, which is highly conserved in evolution and which contains the majority of the mutation hot spots in cancer cells. The rest of the molecule contains a linker region including the major nuclearization signal, an oligomerization domain, and a regulatory C-terminal region containing multiple phosphorylation and acetylation sites (21,22). Recently, this region has been shown to contain a lysine residue (386) that can be covalently modified by SUMO-1 (23,24).
We previously reported the existence of a p53-related gene, p73, mapping to a chromosomal locus (1p36.3) often deleted in neuroectodermal human cancers such as neuroblastomas (24). Subsequent work from several laboratories has described the existence of another p53 family member, more closely related to p73 than to p53, variously described as p63, KET, and p51 (25). p73 shows structural similarities with p53, including the presence of transcription-activating, DNA-binding, and oligomerization domains. It exists as multiple isoforms, resulting from differential splicing of C-terminal exons, of which the two major forms are the ␣ and ␤ isoforms containing 636 and 499 amino acids (25,26). p73 differs from p53 in that its levels of expression are not elevated in response to environmental stresses such as UV irradiation and actinomycin D treatment (25). However, in interaction with the protein kinase c-Abl it mediates an apoptotic response to ionizing radiation and to genotoxic agents such as cisplatin (27).
In contrast to p53, no functionally significant p73 mutations have so far been reported in cancer cells, but a monoallelic pattern of expression is sometimes observed (26). Analysis of p73 knock-out mice does not show an increased susceptibility to spontaneous tumorigenesis (28). However, these studies reveal that p73 plays key roles in a number of developmental processes that are nonoverlapping with the roles played by other p53 family members (28). The activities of p73 may be exerted at multiple levels. Firstly, p73 is a transcriptional activator eliciting a response different from that obtained with p53 (29). Secondly, the major form of p73 is often an N-terminally truncated form that would be incapable of transcriptional activation (28), suggesting the possibility of other nontranscriptional roles for p73.
During yeast two-hybrid screens using p73 as a bait, we isolated the cDNA for SUMO-1. Here we show that p73 can be covalently modified by SUMO-1, with the major modification occurring on the terminal lysine residue. A number of other SUMO-1-interacting proteins were isolated in the p73 twohybrid screening, and we have been able to deduce and confirm a novel SUMO-1 interaction motif. The nature of the proteins identified here suggests that one role for SUMO-1 may be in transcriptional regulation, perhaps co-ordinating this with other key cellular processes such as cell cycle checkpoints, chromosome segregation, DNA recombination and repair, and the induction of apoptosis.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture-The SK-N-AS neuroblastoma cell line (30) and the 293 embryonic kidney cell line (American Type Culture Collection CRL 1573) were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 1 mM sodium pyruvate and 10% fetal calf serum. The U937 monocytic cell line (American Type Culture Collection CRL 1593) was grown in RPMI (Life Technologies, Inc.) containing 10% fetal calf serum.
RNA Preparation and cDNA Library Construction-Total cellular RNA was extracted from SK-N-AS and U937 cells by the guaninidium thiocyanate/acid phenol method (31). Poly(A) ϩ RNAs were isolated using oligo(dT) magnetic beads (Dynal). 1 g of each poly(A) ϩ RNA was transcribed into cDNA using reverse transcriptase (Superscript, Life Technologies, Inc.

) and the primer GATCCGGGCCCATTTTCTAC[AC-GT][ACGT][ACGT][ACGT][ACGT][ACGT]
. cDNAs were fractionated on Sephacryl S400 (Amersham Pharmacia Biotech), and fractions containing cDNA of approximately 500 -1500 nucleotides were selected for cloning in the pJGC cloning vector, derived from the pJG4-5 activation domain vector (32) by insertion of a polylinker containing ApaI and BamHI cloning sites between the EcoRI and HindIII sites. cDNA libraries were constructed by the primer adapter method (33).
Bait Plasmid Construction and Two-hybrid Screening of cDNA Libraries-Sequences corresponding to p73␣ (amino acids 85-636), p73␤ (amino acids 85-499), and p53 (amino acids 73-393) were inserted in the pEG202 vector between either the EcoRI or BamHI sites and the XhoI site (32) creating fusion proteins with the LexA DNA-binding domain. The pEG202.p73␣ bait was introduced using lithium acetate/ polyethylene glycol transformation with sheared single-stranded DNA carrier (34) into the EGY48 strain of Saccharomyces cerevisiae (containing the LEU2 gene under the control of six LexA operators) along with a reporter plasmid pSH18-34 containing the LacZ gene under the control of eight LexA operators. cDNA libraries were then similarly introduced, and yeast colonies were selected on Yeast Nitrogen Base (Difco) medium containing 2% glucose and leucine (but lacking tryptophan, histidine, and uracil). Approximately 10 6 transformed yeast were obtained with the U937 library and 2 ϫ 10 6 transformed yeast with the SK-N-AS library. After 3-4 days, colonies were replicated to nitrocelulose filters (Protran BA85; Schleicher & Schull), replated on dishes containing 2% galactose, 1% raffinose, 1 mg/ml 5-bromo-4-chloro-3indolyl ␤-D-galactopyranoside (X-gal) (Life Technologies, Inc.) lacking leucine, and grown for 4 -5 days. 20 yeast colonies from each cDNA library transformation showing a blue coloration were selected for further study.
Plasmid Identification-Plasmid DNA was extracted using a glass bead disruption method (35), and cDNA inserts in the pJGC plasmid were amplified by polymerase chain reaction using oligonucleotides flanking the cDNA insert and sequenced. The pJGC plasmid was isolated by selection in the KC8 bacterium in minimal A medium containing vitamin B1 and supplemented with uracil, histidine, and leucine but lacking tryptophan. It was then tested for interaction with the p73␣, p73␤, and p53 pEG202 bait plasmids by measurements of the ␤-galactosidase levels in transformed EGY48 24 h after galactose induction of the GAL1 promoter in pJGC, as described by Kippert (36).
Site-directed Mutagenesis-Amino acid substitutions were performed by limited polymerase chain reaction amplification of plasmid DNA using Pfu DNA polymerase and oligonucleotides containing the mutated codons, followed by digestion of remaining input plasmid DNA using the methylation sensitive enzyme Dpn1 (QuikChange; Stratagene).
Transient Transfection of Animal Cells-The p73␣, p73␤, p53, and SUMO-1 cDNAs were introduced into the pcDNA3 vector (Invitrogen) or an epitope-tagged vector derived from pcDNA3 by insertion of an optimalized ATG codon (CCACCATGGCG) and a c-Myc 9E10 epitope (EQKLISEEDL) between the HindIII and EcoRI sites. Plasmid DNA preparations were performed using the QIAfilter Plasmid Midi Kit (Qiagen). Approximately 10 6 cells were transfected in six-well dishes using 1-2 g of plasmid DNA and LipofectAMINE Plus reagents (Life Technologies, Inc.) as described by the manufacturer. Cells were scraped from the dish 20 -30 h after transfection, resuspended in denaturing SDS gel buffer (Bio-Rad) with 0.7 M ␤Ϫmercaptoethanol and analyzed on SDS-polyacrylamide gels.
Dual Luciferase Assays-SK-N-AS cells were transfected in six-well dishes as described above using the RGC firefly luciferase reporter gene (200 ng) and a pRL-CMV Renilla luciferase control (100 ng) (Promega). 20 h after transfection, cells were scraped in 250 l of Passive Lysis Buffer (Promega) and subjected to two cycles of freeze-thawing. 20 l of cell extract were analyzed with the Dual Luciferase Reporter Assay system, using a Lumistar luminometer. Luciferase activities of the RGC-luciferase vector were normalized based on the luciferase activities of the co-transfected pRL-CMV, to correct for variations in cell number and/or transfection efficiency between wells.
Indeed, when tested directly with a SUMO-1 bait, all of these proteins gave a positive result (Fig. 1B).
A SUMO-1 Interaction Motif-When the cDNA sequences of the SUMO-1-interacting proteins were examined for the presence of common sequences, an 11-amino acid motif was detected in a subset of the proteins. This contained a central serine doublet separated by one amino acid (SXS), within which one serine was replaced by threonine in the human PKY cDNA ( Fig. 2A). This SXS triplet is flanked on the N-terminal side by predominantly hydrophobic amino acids and on the C-terminal side by acidic amino acids (D/E) ( Fig. 2A). This motif is evolutionarily conserved in the PKY and PIAS gene families ( Fig. 2A). In view of the subsequent mutagenesis experiments (see below), it is unclear whether the motif in SAE2 that served to derive this consensus would in fact be sufficient on its own to interact with SUMO-1.
The motif from the PM-Scl75 protein was used to construct a LexA-SXS motif fusion protein, which showed a very strong interaction with SUMO-1 in the two-hybrid system (equivalent to that of the dimerization of p53/p53 and stronger than the initial PM-Scl75/SUMO-1 interaction). Critical residues were identified by alanine replacement. Such an analysis shows that both serine residues are necessary (Fig. 2B, sequences c and d), as is the one amino acid spacing between these residues; either no or two amino acid spacing destroys SUMO-1 interaction (Fig. 2B, sequences e and f). The acidic C-terminal residues are

FIG. 2. Identification and testing of a SUMO-1 interaction motif.
A, the cDNAs described in Fig. 1 and two other cDNAs also identified in the p73 two-hybrid screen PML-3 (81) (M79464) and the SUMO-1-activating enzyme SAE2 (13) (AF090384) were compared in order to identify potential common motifs. One such motif was identified and found to be evolutionarily conserved. Sequences are from PKY (38), PKM (61), homeodomain-interacting protein kinases (62), PIAS (37), SAE2 (13), PML-3 (81), and PM-Scl75 (41). B, this motif from the PM-Scl75 protein was inserted in the pEG202 vector and tested with SUMO-1 in the pJGC vector in the two-hybrid system. Different amino acids in this motif were substituted with alanine, and the interaction with SUMO-1 was again tested by measuring ␤-galactosidase levels. These were normalized to that for the initial interaction of the PM-Scl75 motif. A similar motif from the protein furin (43) was also tested (sequence b). also crucial. Mutation of E8 or E10 completely destroys SUMO-1 interaction (Fig. 2B, sequences g and i), and mutation of E9 drastically reduces interaction. (Fig. 2B, sequence h). Although expression levels of the different mutants were not tested, it seems unlikely that differential expression of the mutant proteins (corresponding to single amino acid substitutions, additions or deletions) can explain the almost complete loss of the capacity to interact with SUMO-1. The SXS motif resembles one previously identified in the C terminus of the endopeptidase furin, which is implicated in its translocation into the trans Golgi network (43). However, the furin motif, which also has an acidic N terminus, does not interact with SUMO-1 (Fig. 2B, sequence b), indicating the potential importance of the hydrophobic residues (1)(2)(3)(4) in the SXS motif.
p73␣ Can Be Covalently Modified by SUMO-1-We tested whether p73 might be covalently modified with SUMO-1 by co-transfection of SK-N-AS neuroblastoma cells with p73␣ and SUMO-1. In the presence of p73 and SUMO-1, a prominent novel protein species was formed, showing a molecular mass approximately 20 kDa more than p73␣, both for the full-length and N-terminally truncated forms of p73␣ (Fig. 3, lanes b and  d). This corresponds to the apparent molecular size difference for SUMO-1-modified forms of RanGAP1 and PML seen on SDS-polyacrylamide gel electrophoresis (5,6,18). A similar high molecular mass endogenous p73 species was observed in cell and tissue extracts, such as those from primary cultures of epithelial cells, isolated from human nasal polyps (a cell extract provided by Dr. F. Tournier, University of Paris VII) (44) (Fig.  3, lane f).
In addition to the major modified species (Fig. 3, lane h, indicated with **), several minor higher molecular mass species were seen in the p73 transfected cells (Fig. 3, lane h, indicated with *), which were of variable intensity from one experiment to another (Fig. 3, lanes b and h, and Fig. 4, lane a). The fact that the lysine in ubiquitin used for polyubiquitination is absent from SUMO-1 suggests the possibility of multiple sites for SUMO-1 modification.
The Principal SUMO-1 Modification Site in p73 Is the Cterminal Lysine-When p73␣ and p73␤ were transfected into SK-N-AS cells, SUMO-1-modified forms were detected for p73␣ but not for p73␤ (Fig. 3, lanes h and i). Similarly, after Cterminal deletion up to amino acid 450, p73 no longer showed SUMO-1 modification (Fig. 3, lane g). After deletion of the last 18 amino acids of p73␣, no major SUMO-1-modified form was seen, but minor forms were still apparent (Fig. 3, lane j).
The major modification site in the last 18 amino acids of p73␣ was identified by site-directed mutagenesis as being the final lysine residue, number 627 (Fig. 4A). Similar experiments on p53 identified the C-terminal lysine residue (Lys 386 ) as being the major SUMO-1 modification site (23,24). In agreement with all but one of the previously identified SUMO-1 modification sites, the p53 lysine 386 and the p73␣ lysine 627 have a glutamic acid at ϩ2 and a hydrophobic amino acid at position Ϫ1 (Fig. 4B). In addition, as suggested by Sternsdorf et al. (45), a basic residue (arginine, lysine, or histidine) is found at position Ϫ4 or Ϫ5 (Fig. 4B). A similar consensus sequence is found for the C-terminal lysine of p63␣ (Fig. 4B). (46) reported that p73␣ and  and d). Lane f, a similar high molecular mass form of p73 is seen in total cell extracts from primary cultures of human nasal epithelial cells analyzed with an anti-p73 antibody. Lanes g-j, p73 sequences lacking the N-terminal activation domain (amino acids 1-84) and with differing lengths of C-terminal deletions and an N-terminal c-Myc tag were co-transfected with c-Myc-tagged SUMO-1 into SK-N-AS cells. SUMO-1-modified and unmodified p73 forms were detected using an anti-c-Myc antibody. The major SUMO-1-modified form of p73 is indicated by the double asterisk and minor forms are indicated by a single asterisk.

FIG. 4. SUMO-1 modification of the C-terminal lysine residue of p73␣.
A, the C-terminal lysine of p73 (Lys 627 ) was modified to arginine by site-directed mutagenesis, and this mutant was tested for SUMO-1 modification by co-transfection in 293 cells with SUMO-1, as described in Fig. 3. The major SUMO-1-modified form of p73 is indicated by a double asterisk. B, comparison of the sequences surrounding the modified lysines in p53 and p73␣ with those identified in other SUMO-1-modified proteins (45) reveals a consensus for glutamic acid at position ϩ2, a hydrophobic amino acid at position Ϫ1, and a basic amino acid (His, Arg, or Lys) at position Ϫ4 or Ϫ5. A predicted SUMO-1 modification site in p63␣ (49) is also indicated. p73␤ isoforms undergo differential degradation by the proteasome, with p73␣ being sensitive and p73␤ insensitive to this degradation. We have investigated whether the SUMO-1 modification on lysine 627 plays a role in this instability by transfecting p73 expression plasmids into SK-N-AS cells and following the accumulation of the different p73 protein forms in the presence and absence of the proteasome inhibitor MG-132. p73␣ containing a mutation of the major SUMO-1 modification site (lysine 627) to arginine shows a similar small but reproducible (1.5-3-fold) increase in accumulation in the presence of MG132 as the wild-type p73␣ (Fig. 5A). Both SUMO-1-modified and unmodified forms of p73 are increased by MG132 treatment (Fig. 5, A and B), whereas the levels of SUMO-modified RanGAP1 and of PCNA are unchanged (Fig. 5, C and D).
Quantification of the modified and unmodified p73 forms in several experiments showed that the amount of SUMO-1-modified p73 is increased to a greater extent in the presence of MG132 (5-10-fold) than is that of the unmodified p73 (1.5-3fold) (Fig. 5, A and B). Although the interconvertibility of the two p73 forms complicates the analysis, this result would suggest that SUMO-1 modification potentiates proteasomal degradation of p73.
SUMO-1 Modification May Alter p73 Localization-When performing detergent extractions using RIPA (1% Nonidet P-40, 0.5% sodium deoxycholate) buffer to prepare cell extracts, we often found that the SUMO-1-modified form of p73 was preferentially recovered in the detergent insoluble pellet fraction (Fig. 5B), whereas the nonmodified p73 was found in both soluble and pellet fractions. Treatment of cells with MG132 leads primarily to an increase in p73 accumulation in the pellet fraction (Fig. 5B). In contrast, the SUMO-modified form of RanGAP1 (Fig. 5C), which accumulates in nuclear pore complexes, and other nuclear proteins such as PCNA (Fig. 5D) are very efficiently extracted in the detergent-soluble fraction.
The preferential recovery of SUMO-1-modified p73 in the insoluble fraction may thus result from targeting of p73 modified by SUMO-1 to particular subcellular structures, although SUMO-1 modification is not required for the presence of p73 in this fraction because the mutant p73K627R distributes in a similar fashion to unmodified wild-type p73 (not shown). Alternatively, it may represent differential SUMO-1 modification of p73 in different cellular compartments or differential SUMO-1 cleavage during extraction. We have attempted to reduce isopeptidase cleavage during extraction by the addition of protease inhibitor mixtures. In addition, in certain experiments N-ethylmaleimide was included at concentrations from 10 nM to 1 M in both washing and extraction buffers to inhibit the thiol protease activities reported for SUMO-1 hydrolases (47). This addition did not affect the preferential recovery of SUMO-1-modified p73 in the pellet fraction (Fig. 5E).
SUMO-1 Modification Does Not Affect the Transcriptional Activity of p73-To test the potential modulation of the transcriptional activities of p73␣, we measured the activation of an RGC-luciferase reporter gene in SK-N-AS cells by different p73 forms in the presence or absence of SUMO-1. As can be seen in Fig. 6, the level of activation of the reporter gene by p73␣ is not affected by co-transfection with a large excess of a SUMO-1 expressing plasmid in these experiments. A similar result was found for activation by p53 (not shown). The transcriptional activities of p73␣ and of the mutant p73␣ K627R in these experiments are equivalent and are lower than that of p73␤ (Fig. 6).

DISCUSSION
p73␣ Is Modified by SUMO-1, but p73␤ Is Not-We show here that p73␣ is a novel substrate for SUMO-1 modification. It is of interest that of the two p73 isoforms (which differ only in their C termini), p73␣ is a good substrate for SUMO-1 modification, whereas p73␤ is not. Accordingly, deletion experiments show that the lysines involved in SUMO-1 modification are contained in the C-terminal region of p73␣ not present in p73␤. It remains to be determined whether p73␣ is also a substrate for modification by SUMO-2/3 (48).
p73␣ shows one major SUMO-1-modified form and several minor modified forms. Because the lysine in ubiquitin used for polyubiquitination (lysine 48) is absent from SUMO-1, this suggests that there is one major site and several minor sites for SUMO-1 modification. Site-directed mutagenesis shows that the major modification site of p73␣ is the extreme C-terminal lysine. Similarly, the C-terminal lysine residue of p53 (lysine 386) has recently been identified as the major SUMO-1 modi- FIG. 5. Effect of the proteasome inhibitor MG132 on the accumulation of SUMO-1-modified and unmodified forms of p73, and detergent extraction of these forms. A, wild-type p73␣ and p73␣K627R were transfected into SK-N-AS cells, and 20 h after transfection the cells were cultured for a further 8 h in the presence and absence of 5 M MG132 (Calbiochem). p73 accumulation was detected, using an anti-p73␣ antibody, in total cell extracts prepared by direct lysis in denaturing SDS gel buffer (Bio-Rad). One experiment is shown for p73␣K627R (lanes a and b) and two experiments are shown for p73␣ (lanes c and d and lanes e and f). B-D, p73␣ and SUMO-1 were transfected into SK-N-AS cells, and 20 h after transfection the cells were cultured for a further 8 h in the presence or absence of 5 M MG132 (Calbiochem). Cell extracts were prepared by direct lysis in denaturing SDS gel buffer (Bio-Rad) for the total extract (T) or lysis in RIPA buffer (50 mM Tris pH8, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 1 mM dithiothreitol) containing complete protease inhibitor mixture (Roche Molecular Biochemicals) for 15 min at 4°C and centrifugation at 15,000 rpm to separate insoluble pellet (P) and soluble (S) fractions. Following solubilization of the pellet fraction by boiling in SDS gel buffer, samples were analyzed on denaturing SDS-acrylamide gels. The presence in the different fractions of p73 was detected using an anti-p73␣ antibody (B), that of SUMO-1-modified RanGAP1 and p73 was detected with an anti-SUMO-1 antibody (C), and that of PCNA was detected with an anti-PCNA antibody (D). E, SK-N-AS cells were transfected as in B and cultured without MG132. Cell extracts were prepared as in B in the presence and absence of the thiol protease inhibitor N-ethylmaleimide (NEM) (1 M). The presence in the different fractions of p73 was detected using an anti-p73␣ antibody. fication site (23,24). When the amino acid sequences surrounding lysine 386 of p53 and lysine 627 of p73␣ are compared with those of SUMO-1-modified lysines in other proteins (45), they show a consensus for glutamic acid at the position ϩ2, a hydrophobic amino acid at position Ϫ1, and a basic amino acid (lysine, arginine, or histidine) at position Ϫ4 or Ϫ5 (Fig. 4B). A similar consensus sequence is found for the C-terminal lysine of p63 (49) (Fig. 4B).
Is SUMO-1 Modification Involved in Regulating the Stability or Localization of p73?-Ubiquitin modification is primarily, though not exclusively, involved in regulating protein stability, including that of p53 (1,2). We investigated whether SUMO-1 modification plays a role in modulating the stability of p73, because Lee and La Thangue (46) reported that p73␣ is sensitive to degradation by the proteasome, whereas p73␤ is not. Our results, using an inhibitor of proteasomal degradation, MG132, show that both SUMO-1-modified and nonmodified p73␣ are degraded via the proteasome. p73␣ mutated in the major SUMO-1 modification site and wild-type p73␣ show similar up-regulation by MG132 treatment. SUMO-1 modification would thus not seem to be a major factor influencing p73␣ degradation by the proteasome. However, SUMO-1-modified p73␣ was stabilized to a greater extent than unmodified p73 by treatment with MG132, suggesting that SUMO-1 modification potentiates proteasomal degradation. Although some proteins have been shown to be stabilized by SUMO-1 modification because of a resulting inhibition of ubiquitination on the modified lysine residue (20,50), other SUMO-1-conjugates have been shown to be degraded via the proteasome (51,52). SUMO-1 modification may induce conformational changes potentiating ubiquitination or may influence protein degradation via modulation of E3 ubiquitin ligases, as found for the ubiquitin-like modifier Rub1 in yeast (3).
As for PML (19), SUMO-1-modified p73 may have a particular subcellular localization because we have found that SUMO-1-modified p73 is preferentially isolated in the detergent insoluble fraction. We attempted to exclude the possibility that this result reflects preferential SUMO-1 cleavage in the soluble fraction during cell lysis by addition of protease inhibitor mixtures and by the detection of SUMO-1-modified Ran-GAP1 in the soluble fraction. However, RanGAP1 and PML have recently been reported to be differentially sensitive in vivo to the nuclear SUMO-1 hydrolase SENP1 (53). We cannot thus completely rule out a differential isopeptidase sensitivity of p73-SUMO-1 in the soluble and insoluble fractions as an explanation for our results, although the inclusion in cell washing and lysis buffers of N-ethylmaleimide, which has been shown to inhibit SUMO-1 hydrolases in vitro (47), did not increase the amount of SUMO-1-modified p73 in the soluble fraction.
We have so far been unable to identify p73 accumulation in PODs that contain a large amount of nuclear SUMO-1 after MG132 treatment. However, the low percentage of p73 that is modified by SUMO-1 makes identification of this fraction uncertain. For p53, nuclear aggregates induced by leptomycin B treatment (which prevents nuclear export), have recently been localized adjacent to PODs (54).
Is SUMO-1 Modification Involved in the Transcriptional Activity of p73␣?-Two recent reports (23,24) show that SUMO-1 modification of p53 on lysine 386 increases the transactivation activity of p53 on reporter genes. We did not find this result examining activation by p73␣, or by p53, of the RGC p53responsive element in SK-N-AS cells. This may reflect experimental differences in promoter constructs, in cell types, or in levels of SUMO-1 modification.
We find here, as previously reported (46,56), that the ␤ isoform, which is not subject to SUMO-1 modification, is more transcriptionally active than the ␣ isoform, which can be SUMO-1-modified. The C-terminal region of p73␣ has been shown to modulate its transcriptional and growth regulatory properties (55)(56)(57), acting both as a positive and negative regulator. Although our experiments do not provide evidence for a direct effect of SUMO-1 modification on the transcriptional activity of transfected p73, such a modification could indirectly modulate activation of the endogenous p73␣ protein by influencing interaction with other co-regulatory proteins such as the c-Abl tyrosine kinase (27) or the histone deactetylation complex (see below).
Novel SUMO-1-interacting Proteins and a SUMO-1-interacting Motif-Among the proteins we originally isolated in our p73 two-hybrid screen, the majority were subsequently found to interact with SUMO-1. These include PML, PM-Scl 75, thymine DNA glycosylase, PIASx, PKY, CHD3/ZFH, and one of the SUMO-1-activating enzymes, SEA2. Five of the protein sequences interacting in the two-hybrid system with SUMO-1 contained a motif with a central SXS (or SXT) triplet preceded by predominantly hydrophobic amino acids and followed by predominantly acidic amino acids ( Fig. 2A). We have confirmed that this motif can interact with SUMO-1 in the two-hybrid system. The serine/threonine and acidic residues essential for this interaction constitute a double CKII kinase site ((S/ T)XX(D/E)) (58), and the interaction may thus be regulated by phosphorylation.
Screening DNA sequence data bases for other proteins containing the SXS motif identified RanBP2/Nup358. Although the fit to our consensus sequence is not ideal (KKPEDSPS DDDVL), in that acidic amino acids are found at positions normally constituted by hydrophobic amino acids ( Fig. 2A), this sequence maps to the minimal domain determined for Ran-GAP1-SUMO-1 binding: 2550 -2837 (59). A number of other potential SUMO-1-interacting proteins have been identified. These include c-Myc, DNA repair proteins (XPG, XRCC1, and the Ku70 regulatory subunit of the DNA protein kinase), centromeric proteins (CENP-B), components of the origin of repli- cation (ORC1 and ORC2), and viral proteins such as the cytomegalovirus IE2 protein. In the latter case, this protein has been recently shown to be SUMO-1-conjugated (60).
The SXS motif has been functionally implicated in SUMO-1 interaction using the yeast two-hybrid system, where it may be interacting directly with SUMO-1 or indirectly via Ubc9 or one of the SUMO-1-activating enzymes. Recent experiments on the mouse homologue of PKY-related kinase PKM (61), the homeodomain-interacting protein kinase 2 (62) (Fig. 2A), show that the SXS motif is part of a sequence that specifies localization of mouse homeodomain-interacting protein kinase 2 in nuclear speckles and that interacts in vitro with Ubc9 (63). The SXS motif is also conserved in the PIAS transcription factor family ( Fig. 2A), including the androgen receptor-interacting protein ARIP (64) and the protein Miz-1, a N-terminally truncated form of PIASx␤ that interacts with the homeobox domain protein Msx2 (65). These and other (66) findings suggest that global modulation of SUMO-1 levels might co-ordinately regulate transcription of diverse genetic programs.
SUMO-1-interacting Proteins and Transcriptional Repression-Another of the SUMO-1-interacting proteins is the CHD3/ZFH zinc finger-containing helicase, whose expression is associated with cell growth (39,40). This protein does not contain an SXS motif but may be modified by SUMO-1 since it contains a potential SUMO-1 modification site (VKKE) within the 140 C-terminal amino acids identified here as the SUMO-1-interacting region. CHD3 has been shown to be present in histone deacetylase complexes (HDAC) (67), which have been implicated in transcriptional repression by p53 (68). This interaction between p53 and HDAC is indirect and is mediated at least in part by the co-repressor Sin3a. In the case of the lymphoid lineage-determining factors of the Ikaros gene family, interactions with HDAC have been shown to proceed both through Sin3 and through NURD/Mi-2 complexes (69,70), the latter containing CHD3. In view of the interactions detected here, p73-CHD3 and p53-CHD3 interactions via SUMO-1 may also participate in transcriptional repression mediated by these proteins.
Other proteins interacting with both SUMO-1 and HDAC complexes include the homeodomain-interacting protein kinase 2 (62,71) and unliganded nuclear receptors such as the androgen receptor and the glucocorticoid receptor (72)(73)(74). This may also be the case for retinoic acid receptors that were found to interact in two-hybrid studies with thymine DNA glycosylase (75), which we show here to interact with SUMO-1. The Drosophila transcriptional repressor Tramtrack 69 is also modified by SUMO-1 (76). SUMO-1 modification may play a key role in the balance between transcriptional activation and repression.
If this is true for p73 and its homologue p63, this may offer one explanation for the finding of N-terminally truncated forms of the ␣-isoforms of p63 and p73 (28,77), which would be unable to activate transcription of target genes. These N-terminally truncated forms are, however, still able to bind to DNA, and could thus mediate transcriptional repression via SUMO-1-modulated interaction with HDAC complexes. Transcriptional repression by p53 is involved in the apoptotic activity of this protein (68), and this could also be the case for the apoptotic activity of p73 (24,25), implicating both full-length and N-terminally truncated forms. Alternatively, N-terminally truncated forms can act as dominant negative inhibitors of p53or p73-induced transcription and apoptosis (28,48,78,79). The role of the different p73 forms in modulating apoptosis during development (80) and the contribution of SUMO-1 modification remain to be fully investigated.