Polymeric Chains of SUMO-2 and SUMO-3 Are Conjugated to Protein Substrates by SAE1/SAE2 and Ubc9*

Conjugation of the small ubiquitin-like modifier SUMO-1/SMT3C/Sentrin-1 to proteins in vitro is dependent on a heterodimeric E1 (SAE1/SAE2) and an E2 (Ubc9). Although SUMO-2/SMT3A/Sentrin-3 and SUMO-3/SMT3B/Sentrin-2 share 50% sequence identity with SUMO-1, they are functionally distinct. Inspection of the SUMO-2 and SUMO-3 sequences indicates that they both contain the sequence ψKXE, which represents the consensus SUMO modification site. As a consequence SAE1/SAE2 and Ubc9 catalyze the formation of polymeric chains of SUMO-2 and SUMO-3 on protein substrates in vitro, and SUMO-2 chains are detectedin vivo. The ability to form polymeric chains is not shared by SUMO-1, and although all SUMO species use the same conjugation machinery, modification by SUMO-1 and SUMO-2/-3 may have distinct functional consequences.

Conjugation of the small ubiquitin-like modifier SUMO-1/SMT3C/Sentrin-1 to proteins in vitro is dependent on a heterodimeric E1 (SAE1/SAE2) and an E2 (Ubc9). Although SUMO-2/SMT3A/Sentrin-3 and SUMO-3/SMT3B/Sentrin-2 share 50% sequence identity with SUMO-1, they are functionally distinct. Inspection of the SUMO-2 and SUMO-3 sequences indicates that they both contain the sequence KXE, which represents the consensus SUMO modification site. As a consequence SAE1/ SAE2 and Ubc9 catalyze the formation of polymeric chains of SUMO-2 and SUMO-3 on protein substrates in vitro, and SUMO-2 chains are detected in vivo. The ability to form polymeric chains is not shared by SUMO-1, and although all SUMO species use the same conjugation machinery, modification by SUMO-1 and SUMO-2/-3 may have distinct functional consequences.
The small ubiquitin-like modifier SUMO-1 1 (also known as SMT3C, Sentrin, GMP1, UBL1, and PIC1) is a member of the ubiquitin-like protein family (1). SUMO-1 is known to be covalently conjugated to a variety of cellular substrates via a three-step enzymatic pathway analogous to that of ubiquitin conjugation. The E1-like enzymes for both SUMO-1 and the yeast homologue Smt3p exist as heterodimers known as SAE1/ SAE2 and Uba2p/Aos1p, respectively (2)(3)(4)(5). In the first step the SAE1/SAE2 heterodimer utilizes ATP to adenylate the C-terminal glycine of SUMO-1. Formation of a thioester bond between the C-terminal glycine of SUMO-1 and a cysteine residue in SAE2 is accompanied by the release of AMP. The second step is a transesterification reaction, which transfers SUMO-1 from the E1 to a cysteine residue within the SUMO-specific E2conjugating enzyme (Cys 93 in Ubc9). In the third step, Ubc9 catalyzes the formation of an isopeptide bond between the C terminus of SUMO-1 and the ⑀-amino group of lysine in the target protein. In contrast to the ubiquitin conjugation pathway no activity equivalent to an E3 ligase is required for SUMO-1 conjugation in vitro (2,4), suggesting that the specificity for target proteins is conferred by Ubc9 itself or the Ubc9⅐SUMO-1 thioester complex. This is supported by the observations that almost all SUMO-1-conjugated proteins bind Ubc9 in two-hybrid assays, and the acceptor lysine residues on target proteins appear to exist within the consensus motif KXE (where represents a large hydrophobic amino acid, and X represents any amino acid) (6 -8). Furthermore, SUMO-1 is thought not to form SUMO-1-SUMO-1 polymers, which are characteristic of ubiquitination.
Unlike the majority of ubiquitinated proteins, acceptors of SUMO-1 modifications are not targeted for degradation. In fact, in the case of the transcriptional inhibitor IB␣ the target lysine for SUMO-1 modification is the same as that of ubiquitin conjugation, thus blocking ubiquitination at that residue and stabilizing the protein (8). Transcriptional activity of specific proteins appears to be affected by SUMO-1 modification. For example, conjugation at a single site in the C terminus of p53 activates its transcriptional response (9,10). Furthermore, SUMO-1 modification of certain substrates is also known to have implications upon subcellular localization. The interaction of Ran GTPase-activating protein 1 (RanGAP1) with the Ran-GTP-binding protein 2 at the cytoplasmic face of the nuclear pore complex is dependent on SUMO-1 conjugation of RanGAP1 (11,12). Modification of the promyelocytic leukemia protein (PML) targets it to distinct nuclear bodies (13)(14)(15) and is required for Daxx recruitment to these structures (16,17).
Two ubiquitin-like proteins, known as SUMO-2 (SMT3A, Sentrin-3) and SUMO-3 (SMT3B, Sentrin-2), have been identified that are related to SUMO-1 but are apparently functionally distinct (18 -20). SUMO-2 and SUMO-3 are very similar (95% sequence identity) but are relatively different from SUMO-1 (50% sequence identity). In vivo studies have indicated that PML is modified by SUMO-1 and SUMO-2/-3, although the functional significance of SUMO-2/-3 conjugation has not been revealed (21). Whether or not SUMO-2/-3 conjugates to RanGAP1 in vivo may depend on the expression levels of the SUMO proteins (19,20). Recent evidence indicates that SUMO-2/-3 is more abundant than SUMO-1 in COS-7 cells and that pools of free SUMO-2/-3 decrease when these cells are exposed to heat, ethanol, or hydrogen peroxide (19). Thus SUMO-2/-3 may be involved in the cellular response to envi-* This work was supported by the Medical Research Council. 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. ‡ Initially supported by an Engineering and Physical Sciences Research Council studentship. § Initially supported by the Biotechnology and Biological Sciences Research Council.
cDNA Cloning-The wild type (wt) full-length (FL) 2 cDNAs of SUMO-2 (309 nucleotides) and SUMO-3 (285 nucleotides) were cloned by reverse transcriptase-polymerase chain reaction (RT-PCR) using HeLa poly(A) ϩ RNA as template in the Titan TM one-tube RT-PCR system (Roche Molecular Biochemicals) according to the manufacturer's instructions. wt-SUMO-2 was cloned using primers 5Ј-TCCCCGCGC-CGCTCGGAATCCATGTCCGAG-3Ј and 5Ј-CCCGAATTCGGGACGGG-CCCTCTAGAAACT-3Ј, and wt-SUMO-3 was cloned using primers 5Ј-GAGGAGACTCCGGCGGGATCCATGGCCGACGAA-3Ј and 5Ј-GTAG-AATTCCAGGTTCCCTTTTCAGTAGAC-3Ј. Shorter DNA constructs that code for proteins terminated at the diglycine (276 nucleotides for wt-SUMO-2-GG and 279 nucleotides for wt-SUMO-3-GG) were amplified by PCR using a single downstream primer (5Ј-CCCGAATT-CCTAACCTCCCTGCTGCTGTTGGAACAC-3Ј) for both with the respective upstream primer as described above. The restriction sites introduced into the primers allowed the cleavage of the cDNAs by the enzymes BamHI and EcoRI and subsequent ligation into both pCDNA3-HA (22) and pGEX-2T. Note that our sequences for SUMO-2 and SUMO-3 cloned from poly(A) ϩ HeLa RNA were consistent with those cloned recently from a human B-lymphocyte library (23) that only differ from one another by 3 residues close to the N terminus (see Fig. 2A).
The mutant SUMO-2 and SUMO-3 K11R DNAs were amplified by PCR using the same single downstream primer described above for the wild type GG constructs along with mutant upstream primers 5Ј-ATC-GATGGATCCATGTCCGAGGAGAAGCCCAAGGAGGGTGTGAGGAC-AGAGAAT-3Ј and 5Ј-ATCGATGGATCCATGGCCGACGAAAAGCCCA-AGGAAGGAGTCAGGACTGAGAAC-3Ј, respectively. Cleavage with BamHI and EcoRI allowed the cloning of PCR products into pGEX-2T for GST fusion protein expression and pCDNA3-HA for transient cell transfection.
DNA encoding the C52A-SUMO-1 mutant, which does not form disulfide dimers, was PCR-amplified from the wild type protein using the internal primers 5Ј-GAATCATACGCTCAAAGACAG-3Ј and 5Ј-CT-GTCTTTGAGCGTATGATTC-3Ј and the external primers described previously for the cloning of the wt-SUMO-1-GG protein (22).
To generate a recombinant substrate containing a single SUMO modification site, GST was fused to an 11-amino acid sequence (PRK-VIKMESEE) representing amino acids 485-495 of PML (7). GST and the 11-amino acid modification site were separated by a recognition motif (EPVYFQG) for the tobacco etch virus (TEV) protease (24). An upstream primer complementary to pGEX-2T in the region of the BstBI restriction site (5Ј-GCTGAAAATGTTCGAAGATCGTTTATGTCA-3Ј) and a downstream primer containing both a BamHI restriction site and a sequence coding for the TEV protease recognition motif (5Ј-CAGGG-ATCCTTGGAAATAGACTGGTTCATCCGATTTTGGAGGATGGTC-3Ј) were used in PCR reactions using pGEX-2T as template. BstBI-and BamHI-cleaved PCR products were subsequently ligated into pGEX-2T to give the pGEX-2T-TEV plasmid. Into this expression plasmid DNA coding the 11-amino acid fragment of PML was inserted as described previously (7).
Human wild type histone deacetylase 4 (wt-HDAC4) in pCDNA3.1 (a gift from T. Kouzarides, University of Cambridge (25)), which encodes HDAC4 with an N-terminal Myc epitope tag linked to a 6-histidine (His 6 ) peptide, was used as template with the internal primers 5Ј-GGCGTGCAGGTGAGGCAGGAGCCCATT-3Ј and 5Ј-AATGGGCTC-CTGCCTCACCTGCACGCC-3Ј with the Transformer TM site-directed mutagenesis kit (CLONTECH) as directed by the manufacturer. This generated the K559R-HDAC4 mutant DNA construct in pCDNA3.1. All DNA constructs were verified by automated DNA sequencing on an ABI PRISM TM 377 DNA Sequencer (St. Andrews University DNA Sequencing Unit).
Cell Culture and Transfections-293 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. For analysis of SUMO-1/-2-modified His 6 -HDAC4, 75-cm 2 flasks of subconfluent 293 cells were transfected with 12 g of total plasmid DNA as indicated in the figures. Lysates, protein purification, and Western blots were prepared as indicated.
Isolation of His 6 Proteins-His 6 -HDAC4 proteins were purified from transfected cells by guanidine cell lysis and nickel-nitrilotriacetic acid-Sepharose (Qiagen) purification as described previously (7). Proteins eluted from the nickel-nitrilotriacetic acid-Sepharose were fractionated by electrophoresis in 8% polyacrylamide gels containing SDS before anti-HA Western blotting as described above.
Expression and Purification of Recombinant Proteins-GST-SAE2/ SAE1 fusion protein was expressed in Escherichia coli B834 and purified by affinity chromatography using glutathione-Sepharose as described previously (26). The GST fusion protein was cleaved with thrombin and dialyzed to remove glutathione, and the SAE1/2 was purified by covalent affinity chromatography on a column of SUMO-1 linked to agarose as described previously (2). Expression of both proteins was confirmed by Western blotting, and the activity of recombinant protein was tested by thioester assay using 125 I-SUMO-1 and Ubc9 (data not shown).
Recombinant Ubc9 was expressed and purified as detailed previously (22). All SUMO proteins were expressed in E. coli strain B834, extracted, and purified as described previously (26). GST-SUMO proteins were bound to glutathione-Sepharose beads and either eluted with buffer containing 10 mM glutathione or cleaved by 17 units⅐ml Ϫ1 thrombin while bound. Eluted GST-SUMO proteins were dialyzed against 50 mM Tris (pH 7.5), 1 mM dithiothreitol (DTT) and concentrated using 30-kDa molecular mass cut-off microconcentrators to 10 mg⅐ml Ϫ1 before storage at Ϫ70°C. Thrombin-cleaved SUMO proteins were dialyzed against 20 mM ammonium bicarbonate (pH 8.2), 1 mM DTT. Protein samples were lyophilized and taken up in 50 mM Tris/HCl (pH 7.5), 1 mM DTT at a concentration of 10 mg⅐ml Ϫ1 . All cleaved SUMO protein masses were verified by MALDI-TOF mass spectrometry on a Micromass TofSpec 2E mass spectrometer (Micromass, Manchester, UK; University of St. Andrews Mass Spectrometry Service) before use in the assays.
The GST-TEV-PML protein was expressed and extracted as described for the GST fusion SUMO proteins. The resultant GST-TEV-PML protein (referred to herein as GST-PML) solution was dialyzed against 50 mM Tris/HCl (pH 7.5), 1 mM DTT and concentrated to 4 mg⅐ml Ϫ1 using 10-kDa molecular mass cut-off microconcentrators. All recombinant protein concentrations were determined using both the Bradford method (27) or calculated ⑀ 280 extinction coefficients for absorbance measurements at 280 nm.
Bacterial Expression of [ 35 S]SUMO Proteins-wt-SUMO-GG proteins were labeled by incorporation of [ 35 S]methionine/cysteine during isopropyl-1-thio-␤-D-galactopyranoside induction of bacterial cultures as described above in the presence of 35.75 mCi/liter [ 35 S]methionine/ cysteine (Amersham Pharmacia Biotech). Expressed proteins were then purified as outlined above for the unlabeled proteins.
In Vitro Expression of Proteins-In vitro transcription/translation of proteins was performed using 1 g of plasmid DNA and a wheat germ-coupled transcription/translation system according to the instructions provided by the manufacturer (Promega). [ 35 S]Methionine (Amersham Pharmacia Biotech) was used in the reactions to generate radiolabeled proteins.
In Vitro SUMO Conjugation Assays-Noncompetitive SUMO conjugation assays were performed in 10-l volumes containing between 1.4 and 10 g (as indicated) of either unlabeled or 35 S/ 125 I-labeled SUMO proteins, an ATP-regenerating system, and buffer (50 mM Tris (pH 7.5), 5 mM MgCl 2 , 2 mM ATP, 10 mM creatine phosphate, 3.5 units⅐ml Ϫ1 creatine kinase) and 0.6 units⅐ml Ϫ1 inorganic pyrophosphatase in either the absence or presence of 1 l of [ 35 S]methionine-labeled substrate (HDAC4 or promyelocytic leukemia protein) or varying concentrations of GST, GST-PML, GST-SUMO-1-FL, GST-SUMO-2-FL, or GST-SUMO-3-FL as indicated in the figure legends. Unless otherwise stated, assays contained 120 ng (1.1 pmol) of purified recombinant SAE1/SAE2 and 650 ng (35.9 pmol) of Ubc9. Reactions were incubated at 37°C for between 45 min and 4 h as described. After termination with SDS sample buffer containing ␤-mercaptoethanol, reaction products were fractionated by electrophoresis in polyacrylamide gels (8 -10%) containing SDS, stained, destained, and dried before analysis by phosphorimaging. Reactions were incubated for 4 h at 37°C before termination with SDS sample buffer containing ␤-mercaptoethanol followed by electrophoresis in a 10% polyacrylamide gel containing SDS. Gels were stained and destained before drying and exposure to a phosphorimaging screen for 10 min. The mutant proteins C52A-SUMO-1 and K11R-SUMO-2/-3 were used to avoid experimental interference from SUMO-1-SUMO-1 disulfide linked dimers and the formation of poly(SUMO-2) and poly(SUMO-3).
Tryptic Digests and MALDI-TOF Mass Spectrometric Analysis of SUMO-2 Conjugation Assays-After a 3-h incubation a portion (10 l) of the SUMO-2 conjugation assays carried out either in the absence of substrate or in the presence of GST-PML was dialyzed against 50 mM ammonium bicarbonate on a 0.025-m VS membrane disc (Millipore, Bedford, MA). This procedure not only exchanged the buffer but also removed ATP thus stopping the reaction. 0.5 l of trypsin (bovine, sequencing grade, Roche Diagnostics, 1 g⅐l Ϫ1 ) was added to the dialyzed sample, and the digestion was allowed to proceed overnight at 37°C. The tryptic digest was then analyzed by MALDI-TOF mass spectrometry. 0.5 l of tryptic digest was applied to the target along with 0.5 l of 0.1% trifluoroacetic acid to acidify the sample and 0.5 l of ␣-cyano-4-hydroxycinnamic acid matrix solution (10 mg⅐ml Ϫ1 in 75% acetonitrile, 2.5% formic acid) and allowed to dry. The sample was analyzed using a TofSpec 2E mass spectrometer (Micromass) in reflectron mode.

A Completely Characterized System for SUMO-1 Modification in Vitro-
To facilitate the biochemical analysis of SUMO-1 conjugation an in vitro system has been developed that contains recombinant, bacterially produced components. The assay contains SAE1/SAE2, Ubc9, SUMO-1, and, as substrate, GST fused to an 11-amino acid sequence containing the SUMO modification site located between amino acids 485-495 of PML (7). Each of the components was analyzed in a polyacrylamide gel containing SDS and was highly purified as judged by Coomassie Blue staining (Fig. 1A). SUMO-1 was labeled with 35 S by the addition of [ 35 S]methionine/cysteine to the bacterial growth medium. In comparison with previous assay methods that used cell extract sources of SAE1/2 and in vitro translated protein substrates, the GST-PML fusion was efficiently utilized as a substrate for modification with 35 S-labeled SUMO-1 in this system (Fig. 1B). The specificity of the reaction was assessed by comparing modification of GST with that of the GST-PML fusion and quantitation of the 35 S-labeled products by phosphorimaging. Although GST contains 21 lysine residues, none of them conform to the SUMO modification consensus KXE, and modification of GST was less than 1% of that observed for the GST-PML fusion (Fig. 1C). The extent of modification was such that after fractionation by polyacrylamide gel electrophoresis the reaction products could be easily monitored by Coomassie Blue staining. Under these conditions the product of the reaction was GST-PML bearing a single SUMO-1 modification with no evidence for the formation of SUMO-1 multimers (Fig. 1, D  and E).
Formation of Polymeric Chains of SUMO-2 and SUMO-3 in Vitro-Inspection of the SUMO-1, SUMO-2, and SUMO-3 sequences revealed that a consensus SUMO modification site (KXE) is present in the N-terminal regions of SUMO-2 and SUMO-3 but is absent in the sequence of SUMO-1 ( Fig. 2A). This raises the possibility that SUMO-2 and SUMO-3 could be used as substrates for SUMO modification and thus form poly-(SUMO) chains. SUMO-1, SUMO-2, and SUMO-3 were therefore expressed and purified as both the full-length pro-protein precursors and the shorter active forms (exposing the C-terminal diglycine motif). GST fusion and thrombin-cleaved versions of each protein were analyzed by SDS-polyacrylamide gel electrophoresis and Coomassie Blue staining (data not shown). The thrombin-cleaved proteins were essentially homogeneous, and their mass, which was determined by mass spectrometry, corresponded to that predicted from the sequence. To test the possibility that SUMO-2 and SUMO-3 could be conjugated to themselves the purified proteins were incubated in an assay containing purified SAE1/SAE2 and Ubc9. Analysis of the reaction products by polyacrylamide gel electrophoresis and Coo- massie Blue staining revealed that in the presence of all the assay components a large number of more slowly migrating species are observed that are consistent with the formation of poly(SUMO-2) (Fig. 2B) and poly(SUMO-3) (Fig. 2C) chains. In each case the formation of these products was dependent on the presence of SAE1/SAE2 and Ubc9. To compare the ability of SUMO-2 and SUMO-3 to act as acceptors for SUMO modification with that of SUMO-1 and GST-PML, GST fusions of the unprocessed forms (which cannot be conjugated to protein substrates) of SUMO-1-(1-101) (SUMO-1-FL), SUMO-2-(1-103) (SUMO-2-FL), and SUMO-3-(1-95) (SUMO-3-FL) and GST-PML were added in increasing amounts to a modification reaction containing 125 I-labeled SUMO-1 and limiting amounts of Ubc9. Incorporation of SUMO-1 into conjugated product was determined and plotted as a function of added substrate (Fig.  2D). GST-SUMO-2-FL was relatively efficiently utilized as a substrate (0.9 pmol of product formed in the presence of 100 pmol of substrate) albeit less efficiently than GST-PML (1.6 pmol of product formed in the presence of 100 pmol of substrate), whereas GST-SUMO-3-FL was not a good substrate for SUMO-1 modification (0.2 pmol of product formed in the presence of 100 pmol of substrate). GST-SUMO-1-FL modification was at background levels (less than 0.1 pmol of product formed in the presence of 100 pmol of substrate).
Identification of Sites of SUMO-2 Modification-Although the accumulation of slowly migrating species was consistent with the formation of poly(SUMO-2) chains, it was important to establish that this was indeed the case and to demonstrate that these chains could be linked to substrate proteins. In vitro reactions containing SAE1/SAE2 and Ubc9 were therefore set up in the additional presence of GST-PML, SUMO-2, or GST-PML and SUMO-2. Analysis of the reaction products by poly-acrylamide gel electrophoresis and Coomassie Blue staining revealed that in the presence of SUMO-2, dimers and trimers of SUMO-2 were present (Fig. 3A). In the presence of GST-PML, but with no added SUMO-2, no additional products were generated. When GST-PML and SUMO-2 were incubated in the conjugation assay a series of more slowly migrating products were generated that are consistent with the formation of poly-(SUMO-2) chains linked to GST-PML (Fig. 3A). To establish the identity of lysine residues involved in the formation of poly(SUMO-2) chains and attachment to the GST-PML substrate, samples from the reaction products analyzed in Fig. 3A were digested with trypsin and analyzed by MALDI-TOF mass spectrometry. Trypsin digestion of the assay mixture, containing SUMO-2 but no substrate, gave signals corresponding to the expected peptides produced by cleavage at the carboxyl side of lysine and arginine residues (Fig. 3B). In particular a signal was detected at 3873.5 Da that corresponds to aa 59 -92 of SUMO-2. This peptide contains Arg 60 , which under these conditions appeared to be resistant to trypsin cleavage with only a very small portion of the completely digested product (3570.7 Da) being detected. Additionally a pair of signals were detected at 5048.9 and 5352.1 Da. These signals correspond to digestion fragments in which aa 61-92 of SUMO-2 are covalently linked by an isopeptide bond to aa 8 -20 of SUMO-2 and aa 59 -92 are linked by an isopeptide bond to aa 8 -20. Hence, these signals confirmed the presence of SUMO-2 multimers with a linkage between the C-terminal Gly and Lys 11 .
Digestion of the assay mixture containing both SUMO-2 and GST-PML with trypsin gave the signals indicated above and a pair of new signals (Fig. 3C) Lysine 11 in SUMO-2 and SUMO-3 Is Required for the Formation of Poly(SUMO) Chains-To establish the importance of lysine 11 in the formation of poly(SUMO) chains these residues were changed to arginine, which preserves the charge on the molecule but cannot be utilized for the formation of an isopeptide bond. K11R-SUMO-2 and K11R-SUMO-3 were expressed and purified in an identical fashion to that of the corresponding wild type proteins. Wild type and mutant forms of SUMO were assayed for the generation of poly(SUMO) chains in the absence of substrate, and the reaction products were analyzed by electrophoresis in a polyacrylamide gel followed by Coomassie Blue staining. Whereas multimeric forms of SUMO-2 and SUMO-3 were evident with the wild type proteins, these forms were absent when the assays were performed with K11R-SUMO-2 and K11R-SUMO-3 (Fig. 4A). Inclusion of 125 I-labeled GST-SUMO-2-FL substrate in the reactions containing wild type SUMO-2 and SUMO-3 led to the formation of GST-SUMO-2-FL linked to SUMO and various poly(SUMO) chains (Fig.  4B). When the same reactions were performed with K11R-SUMO-2 and K11R-SUMO-3, the formation of GST-SUMO-2-FL linked to monomeric SUMO was unaffected, whereas formation of GST-SUMO-2-FL linked to poly(SUMO) chains was reduced (Fig. 4B). To analyze modification by the various SUMO forms in the context of a full-length protein substrate, 35 S-labeled in vitro translated PML was incubated with the wild type versions of SUMO-1, SUMO-2, and SUMO-3 in the presence of SAE1/SAE2 and Ubc9. As expected PML was modified at multiple sites by SUMO-1 (13), but a more extensive range of more slowly migrating modified products was observed The SUMO-SUMO multimers can be seen as a "ladder " of increasing molecular weight species in the complete assays. Samples of creatine kinase alone (CK) and SUMO-2/-3 alone are also shown. D, assays for SUMO-1 conjugation using as substrate varying quantities of GST-PML and the GST fusions of the full-length SUMO proteins (GST-SUMO-1-FL, GST-SUMO-2-FL, and GST-SUMO-3-FL) as indicated were set up as described under "Experimental Procedures". Using 0.55 pmol of SAE1/2, 0.55 pmol of Ubc9, and 125 pmol of 125 I-SUMO-1 per 10-l reaction, assays were incubated for 2 h at 37°C before termination with SDS sample buffer containing ␤-mercaptoethanol. Samples were fractionated by electrophoresis in a 10% polyacrylamide gel containing SDS before drying and analysis by phosphorimaging. Quantities of conjugated 125 I-SUMO-1 were measured by MacBas software analysis for each assay condition and substrate, and these data are represented graphically as indicated.
when the reactions were carried out with SUMO-2 and SUMO-3 (Fig. 4C). When the same reactions were carried out with K11R-SUMO-2 and K11R-SUMO-3 the more slowly migrating species were no longer detected, and the pattern reverted to that observed for SUMO-1 (Fig. 4C). Thus it is likely that poly(SUMO) chains of SUMO-2 and SUMO-3 are formed on protein substrates through the formation of isopeptide bonds between the C-terminal glycine residue of one SUMO molecule and lysine 11 of another molecule of SUMO-2 or SUMO-3.
Ubc9 Does Not Discriminate between SUMO-1, SUMO-2, and SUMO-3-Although the data presented in Figs. 1, 2, 3, and 4 indicate that SAE1/SAE2 and Ubc9 utilize SUMO-1, SUMO-2, and SUMO-3 for conjugation to protein substrates, it does not indicate if the enzymes preferentially utilize a particular form of SUMO. To address this question a competition analysis was carried out in which 125 I-labeled C52A-SUMO-1 was incubated with GST-PML in the presence of SAE1/SAE2 and limiting Ubc9 and a range of concentrations of unlabeled SUMO-1, SUMO-2, and SUMO-3. To simplify analysis the mutant C52A-SUMO-1 was used to stop the formation of disulfide-linked dimers of SUMO-1 and the K11R mutants of SUMO-2 and SUMO-3, which cannot form poly(SUMO) chains, were used. After incubation the reaction products were separated by electrophoresis in a polyacrylamide gel, and the 125 I-SUMO-1 conjugated to GST-PML was quantitated by phosphorimaging.
The addition of increasing amounts of unlabeled SUMO-1, SUMO-2, and SUMO-3 to the reaction resulted in a dosedependent decrease in the amount of 125 I-SUMO-1 conjugated to GST-PML. Analysis of the competition data indicated that each form of SUMO competes equally efficiently for formation of 125 I-SUMO-1-GST-PML (Fig. 5). Thus SAE1/SAE2 and Ubc9 utilize SUMO-1, SUMO-2, and SUMO-3 for conjugation to protein substrates with little evidence that these enzymes have the ability to discriminate between the various forms of SUMO.
Formation of SUMO-2 Chains in Vivo-To provide evidence that formation of poly(SUMO) chains can occur in vivo our objective was to use a SUMO substrate that contained a single modification site and demonstrate that although SUMO-1 gave a single modified species, multiple modified species could be detected with SUMO-2. HDAC4 contains a single site of modification at lysine 559, and changing this residue to an arginine abolishes SUMO-1 modification in vitro (Fig. 6A). To examine the modification of HDAC4 with SUMO in vivo HA-SUMO-1 was cotransfected with His 6 -HDAC4, modified forms of HDAC4 were isolated on nickel-Sepharose, and SUMO-1 modification was detected by Western blotting with an antibody to the HA tag. A single SUMO-1-modified form of HDAC4 was detected when both HA-SUMO-1 and His 6 -HDAC4 were cotransfected. This form was not detected when the K559R mutant of His 6 -HDAC4 was cotransfected with HA-SUMO-1 (Fig. 6B). Cotransfection of His 6 -HDAC4 with HA-SUMO-2 and analysis of The GST-PML/SUMO-2 assay was set up as described under "Experimental Procedures." Assays lacking GST-PML and SUMO-2, containing no GST-PML but SUMO-2, containing GST-PML but no SUMO-2, and containing both components were incubated for 3 h. Half of the assay mixture was separated by SDSpolyacrylamide gel electrophoresis (A); the remainder was digested with trypsin, and the resultant peptides were analyzed by MALDI-TOF mass spectrometry. Portions of the spectra obtained for the assay containing SUMO-2 without GST-PML (B) and for the assay containing both SUMO-2 and GST-PML (C) are shown. In A the asterisk (*) marks an unidentified SUMO-dependent species, and molecular weight markers are also indicated. Peptides corresponding to peaks from the mass spectrometry analysis are shown in the material bound to nickel-Sepharose by Western blotting with an HA antibody revealed the presence of an additional more slowly migrating species of SUMO-2-modified His 6 -HDAC4 (Fig. 6B) that is consistent with the formation of two linked SUMO-2 molecules attached to a single site in HDAC4. This conclusion was supported by cotransfection of His 6 -K559R-HDAC4 and HA-SUMO-2, which resulted in loss of both SUMO-2-modified species (Fig. 6B), and by cotransfection of HA-K11R-SUMO-2 and His 6 -HDAC4, which eliminated the more slowly migrating SUMO-2-modified species. These data suggest that SUMO-2 chains can be conjugated to protein substrates in vivo. DISCUSSION Since their identification almost 5 years ago, the reason for the existence of three very similar ubiquitin-like proteins, SUMO-1, SUMO-2, and SUMO-3, has been unclear. Although a considerable amount of information has accumulated relating to SUMO-1, studies on SUMO-2 and SUMO-3 have been limited. Distinct roles for SUMO-1 and SUMO-2 have emerged from investigations using mammalian cells transfected with SUMO-1 and SUMO-2/-3 constructs in which differing profiles of conjugated proteins are generated when analyzed by Western blotting (19). 3 In fact the species generated by modification with SUMO-2/-3 appear to be of a higher molecular weight than those from SUMO-1 transfection experiments. Furthermore, it was demonstrated that in HeLa cells co-transfected with plasmids expressing SUMO-1 and SUMO-2 tagged with fluorescent proteins, the modifiers largely colocalize in PML nuclear bodies, although SUMO-1 alone is found associated with the nuclear membrane (19). 4 These data suggest that although SUMO-1 and SUMO-2/-3 may share many of the same substrates, the functional consequences of modification by SUMO-1 and SUMO-2/-3 may be quite different.
As our knowledge has developed, comparisons between ubiquitin and the SUMO proteins have revealed the extent of structural similarity but functional diversity. Like ubiquitin, it appears that SUMO conjugations can be induced by cell stress (9,19,28,29). Although SUMO and ubiquitin are conjugated and deconjugated by distinct pathways, the enzymes involved in their metabolism display significant degrees of sequence, structural, and functional similarity (for review, see Ref. 30). Indeed there is convincing evidence to support the theory that the E1 enzymes for both ubiquitin and the SUMO proteins share common prokaryotic ancestors (31). It has been shown here that in contrast with their close family member SUMO-1, SUMO-2 and SUMO-3 share the capacity to form polymeric chains with their relatively distant family member ubiquitin. Tryptic digests and MALDI-TOF mass spectrometry analysis of SUMO-2 conjugation assays both in the presence and absence of a defined substrate (GST-PML) show that this self-conjugation occurs via lysine 11, which exists within a SUMO modification motif (KXE, where is a large hydrophobic residue, K is the target lysine, and E is glutamic acid). Furthermore, mutation of lysine 11 in SUMO-2 and SUMO-3 blocks the assembly of polymeric chains in reactions either lacking or in the presence of substrate (GST-SUMO-2-FL or in vitro translated full-length PML protein). This is also recognized in vivo as the histone deacetylase protein HDAC4 that is modified by SUMO-1 at a single lysine residue (Lys 559 ) is modified by SUMO-2 chains. Although SUMO-2 and SUMO-3 are 96% identical, SUMO-2 appears to be a better substrate for chain formation than does SUMO-3 (Figs. 2D and 4). Thus the 3-amino acid difference in the N-terminal region appears to be partially inhibitory to conjugation of SUMO-3 in comparison with SUMO-2. Whether this inhibition is due to differing secondary structures of the two modifiers in this region or disruption of specific amino acid interactions with Ubc9 remains to be determined. The cellular and substrate-specific significance of these polymeric forms of SUMO-2 and SUMO-3 are not clear. One possibility is that in a manner analogous to the enhanced recognition of multiubiquitinated over monoubiquitinated proteins by 26 S protease complex proteins (for review, see Ref. 32), poly(SUMO)-modified proteins may represent a more distinct signal to proteins that bind them. Furthermore, the possibility that SUMO-1, SUMO-2, and SUMO-3 form heteromultimers is not unrealistic. In fact GST-SUMO-2-FL is conjugated by all three SUMO proteins in vitro (Figs. 2D and 4B). Thus it is possible that SUMO-1 may act as a SUMO chain terminator. This is supported by the observation that SUMO-1 and SUMO-2 appear to largely co-localize in mammalian cells, and although many of the identified SUMO-1 substrates are of relatively low molecular weight, the majority of SUMO-2 conjugated proteins found in crude cell lysates are of over 100 kDa in mass (19). 3 Interestingly sequence comparison between proteins of the SUMO family from different species reveals that although the KXE sequence required for polymeric chain formation is present in the N-terminal region of Saccharomyces cerevisiae Smt3p, it is not conserved in all species.
To aid investigations into the dynamics of SUMO-1, SUMO-2, and SUMO-3 conjugation we developed an in vitro assay that conjugates recombinant SUMO-1, SUMO-2, or SUMO-3 onto a GST fusion substrate presenting only the 11amino acid region surrounding one of the lysines targeted for SUMO-1 conjugation in PML ( 485 PRKVIKMESEE 495 ). The assay uses recombinant Ubc9 and SAE1/2 and either 5 mM ATP or an ATP-regenerating system with 0.6 units⅐ml Ϫ1 inorganic pyrophosphatase (see "Experimental Procedures" for details). Mass spectrometry and analysis of duplicate assays using GST alone as substrate revealed the specificity of the system for the 11-residue tag-GST protein and confirmed that the consensus lysine was, as suspected, the target for modification. By eliminating the requirement for crude cell extracts (used to provide SAE1/2) and in vitro translated protein substrates, such an assay is an important tool for the dissection and analysis of the enzymes of the SUMO conjugation pathway.
In comparison with assays using cell extracts and in vitro translated substrates, these conjugation reactions generate substantial quantities of SUMO-modified product. They can conjugate ϳ40 pmol of substrate/h (calculated from Fig. 1, C and D) in a 10-l reaction containing 250 pmol of GST-PML and 35.9 pmol of Ubc9, and thus reaction progress can be monitored by Coomassie-stained polyacrylamide gel electrophoresis. However, although this represents a significant improvement over existing methods, these reactions proceed relatively slowly with a Ubc9 turnover rate of ϳ1/h (or 3/h in Ubc9-dependent assays, Fig. 2D). Thus it is possible that although SUMO conjugation reactions can progress in vitro in the presence of only SAE1/2 and Ubc9, further components may be required to increase the reaction rate in vivo.
To determine whether Ubc9 could act as a discretionary factor in the SUMO system, an E2-dependent competition assay containing limiting amounts of Ubc9 was used to analyze the specificity of Ubc9 for SUMO-2/-3 in comparison with SUMO-1. The data showed that wt-SUMO-1, K11R-SUMO-2, and K11R-SUMO-3 compete with 125 I-C52A-SUMO-1 to ap-proximately the same extent for conjugation to GST-PML, supporting the idea that the contrasting cellular characteristics of SUMO-1 and SUMO-2/-3 are unlikely to rely solely upon Ubc9dependent conjugation in vivo. In addition to these findings, the recent identification of SUMO isopeptidases with apparently differing specificities for SUMO-1 and SUMO-2/-3 (33,34) suggest that the overall conjugation state of SUMO-1-and SUMO-2/-3-modified proteins may in fact be regulated at the level of removal rather than conjugation. While we still understand little about functional heterogeneity of SUMO-1 in comparison with SUMO-2/-3, the discovery of SUMO-2 and SUMO-3 multimers adds a new dimension of complexity to the SUMO conjugation system.