INTRODUCTION

SMT3 is an essential Saccharomyces cerevisiae gene encoding a member of a family of ubiquitin-like proteins, including the mammalian protein SUMO-1 (1) (also called GMP1, PIC1, UBL1, or sentrin (2-5)). SUMO-1 was isolated as a protein covalently linked to the Ran-GTPase-activating protein (RanGAP1),¹ and it is also conjugated to a number of other, primarily nuclear, proteins (1-3, 6). Smt3p, which is 48% identical to SUMO-1 and 17% identical to Ub, also becomes attached to several proteins posttranslationally, and at least one of its essential functions is mediated by its attachment to another protein (7).

Ubiquitin (Ub) conjugation is carried out by a multistep pathway culminating in formation of an isopeptide bond between the C-terminal carboxyl group of Ub and the -amino group of a lysine side chain in an acceptor protein (8, 9). In the initial step, Ub-activating enzyme (E1) utilizes ATP to adenylate the Ub C terminus, which is then transferred to a conserved Cys residue in the E1, yielding an E1~Ub thioester, AMP, and pyrophosphate. Ub is transferred from the E1 to a Cys residue in a Ub-conjugating enzyme (E2). Cells contain multiple E2s (13 in yeast by sequence similarity) which are involved in ubiquitinating different proteins. In some cases Ub can be transferred directly from the E2 to the acceptor protein, but more often Ub-isopeptide bond formation is facilitated by a third heterogeneous class of proteins termed Ub-protein ligases or recognins (E3).

Several features of the Ub pathway are conserved in the early steps of the Smt3p conjugation pathway (7). Like Ub, the SMT3 translation product is proteolytically processed to expose its mature C terminus. Smt3p undergoes ATP-dependent activation by a heterodimeric activating enzyme consisting of Uba2p, a 71-kDa protein with extensive sequence similarity to the C-terminal region of E1s, including the active site Cys residue participating in the thioester (10), and Aos1p, a 40-kDa protein similar to the N termini of E1s (7). While the Smt3p- and Ub-activating enzymes are related, they do not interact with each other's substrates, suggesting that the two pathways are distinct.

One candidate to be the Smt3p-conjugating enzyme is Ubc9p, a member of the E2 sequence family whose Xenopus leavis homolog co-immunoprecipitates with a complex including SUMO-1-conjugated RanGAP1 (11) and whose human homolog interacts in a two-hybrid screen with SUMO-1 (4). UBC9 is an essential gene. Conditional ubc9 mutants arrest in the cell cycle at G₂/M and are impaired in proteolysis of both B-type cyclins (12) and G₁ cyclins (13). However, Ubc9p does not seem to be the E2 involved in cyclin B ubiquitination (14, 15). While it has been suggested that Ubc9p functions as an E2 in the Ub pathway, there is no clear biochemical data to support this hypothesis.

We have found that Ubc9p is a Smt3p-conjugating enzyme and that it is likely to constitute the only Smt3p-conjugating activity in yeast. Furthermore, Ubc9p does not seem to be a Ub-conjugating enzyme, suggesting that the ubc9 cell cycle defect actually results from impairment of Smt3p conjugation.

EXPERIMENTAL PROCEDURES

Standard techniques were used (16). S. cerevisiae strains used were DF5 (MAT trp1-1 ura3-52 his3-200 leu2-3,112 lys2-801) (17) and the DF5-derived strain YWO102 (MAT ubc9::TRP1 leu2::ubc9Pro-Ser::LEU2) (12), which was a generous gift of S. Jentsch (Heidelberg, Germany).

pET21b or pET11a (Novagen)-based Escherichia coli expression plasmids expressing N-terminally His₆- (18) and FLAG-tagged (19) Smt3p (HF-Smt3p) and His₆-tagged Aos1p and Uba2p have been described (7). pET21b-based plasmids for expressing Ubc9p-, Ubc2p-, and Pex4p-tagged C-terminally with His₆ were produced by polymerase chain reaction. The Ubc2p construct expresses Ubc2p truncated after Met¹⁵³, deleting most of the acidic C-terminal domain. A pET21b-based plasmid expressing N-terminally His₆- and FLAG-tagged ubiquitin bearing Lys⁴⁸  Arg, Lys⁶³  Arg and Gly⁷⁶  Ala mutations (HF-Ub(A76)) was also produced by polymerase chain reaction. The A76 mutation alters the kinetics of thioester formation (20). Construction details are available on request. Proteins were expressed in E. coli and purified by Ni-NTA chromatography as described (7). The H-Uba2p·Aos1p complex was further purified by Smt3p affinity and gel filtration chromatography (7). Purified recombinant Ubc4p was a generous gift of V. Chau (Wayne State University School of Medicine, Detroit, MI). His₆-tagged Uba1p was purified from yeast lysate of strain JD77-1A (MATa uba1::HIS3) (10) bearing plasmid pJD325, which expresses H-Uba1p from P_CUP1, both gifts of J. Dohmen (Heinrich-Heine Universität, Düsseldorf, Germany). Uba1p was purified by Ni-NTA chromatography as described for Uba2p (7) followed by Ub affinity chromatography (21).

Ni-NTA-purified recombinant Aos1p and Uba2p were applied to an HF-Smt3p-Affi-Gel 15 column and washed as described (7). A few milligrams of each protein bound. Yeast nuclear extract, consisting of the soluble fractions (primarily the load) below the nuclear envelopes in a Nycodenz/sucrose flotation gradient, prepared as described (22), was a generous gift of R. Beckmann and D. Peter. Extract containing ~20 mg of protein was dialyzed against 50 mM BisTris (pH 6.5), 50 mM NaCl, 1 mM MgCl₂, and 1 mM -mercaptoethanol (-ME), brought to final concentrations 50 mM BisTris (pH 6.5), 75 mM NaCl, 5 mM MgCl₂, 2 mM ATP, and 0.5 mM -ME and applied to the Aos1p/Uba2p-preloaded HF-Smt3p column. The column was washed with 10 column volumes of the same buffer except containing 1 M NaCl and eluted as described (7). The eluate was exchanged into 50 mM Tris (pH 8.0), 300 mM NaCl, 1 mM MgCl₂, and 1 mM -ME using a Biomax-10 ultrafiltration unit (Millipore), bound to 0.5 ml of Ni-NTA-agarose, and washed with 5 column volumes of 50 mM sodium phosphate (pH 6.0), 300 mM NaCl, and 10% glycerol. The unbound and wash fractions were pooled and fractionated by SDS-PAGE on a 4-20% gradient gel (Novex) followed by Coomassie Blue staining. Bands were excised from the gel and identified by direct analysis of a Lys-C endoproteinase digest by MALDI-TOF mass spectrometry at the Rockefeller University Protein/DNA Technology Center as described (23). The six major peaks, with measured masses of 856.68, 1080.54, 1702.28, 1829.39, 2892.92, and 3463.25 Da, are within 1.6 mass units of the peptide masses predicted by a theoretical Lys-C digest of Ubc9p (using the ProFound and MS-Fit programs: http://chait-sgi.rockefeller.edu/cgi-bin/ProFoundandhttp://prospector.ucsf.edu/htmlucsf/msfit.htm), assuming the N terminus is acetylated and cysteine residues are modified by acrylamide.

Thioester formation reactions contained 50 mM BisTris (pH 6.5), 100 mM NaCl, 10 mM MgCl₂, 0.1 mM DTT and some of the following: 5 mM ATP, 30 µg/ml HF-Smt3p or HF-Ub(A98), 15 µg/ml Aos1p/Uba2p heterodimer or Uba1p, and ~15 µg/ml Ubc9p, Ubc4p, Pex4p, or Rad6p (Ubc4p and Pex4p contained noticeable amounts of contaminants). HF-Smt3p-containing reactions were incubated 30 min and HF-Ub(A76)-containing reactions 90 min at 25 °C and stopped by addition of SDS-containing loading buffer either lacking reducing agent or containing 100 mM DTT, followed by a 10-min incubation at 37 °C, SDS-PAGE (6-15% acrylamide gradient) and either Coomassie Blue staining or immunoblotting using the M2 anti-FLAG antibody (IBI/Kodak).

Whole yeast cell lysate from DF5 or YWO102 (both after 90-min incubation at 37 °C) were prepared as described (7) and exchanged into 50 mM Tris (pH 7.5), 150 mM NaCl, 1 mM MgCl₂, and 0.1 mM DTT using a Sephadex G-25 column. Reactions were incubated for 90 min at 25 °C and contained 25 mM Tris (pH 7.5), 75 mM NaCl, 10 mM MgCl₂, 50 µM DTT, 10 mg/ml whole yeast lysate, 30 µg/ml HF-Smt3p and, where indicated, 5 mM ATP and/or 5 µg/ml Ubc9p.

RESULTS

E1s and E2s can be readily purified by "covalent affinity chromatography" (21) by adding ATP to a protein mixture containing these enzymes and incubating with an affinity column to which Ub has been coupled. E1s and E2s form thioester bonds with the column-bound Ub and then can be gently eluted in buffer containing a thiol reducing agent, such as DTT, which breaks the thioester bonds.

We attempted an analogous approach to purifying the Smt3p-conjugating enzyme(s) using a column linked to His₆ (18)- and FLAG-tagged (19) Smt3p (HF-Smt3p) (mature processed Smt3p having C-terminal Gly⁹⁸ will be referred to as Smt3p). Because the Smt3p-conjugating enzyme would be predicted to bind the column by displacing Uba2p/Aos1p, which is necessary to activate the column-bound Smt3p, the Smt3p column was prebound with recombinant His₆-tagged Uba2p and Aos1p to improve the efficiency of conjugating enzyme binding. Next, ATP-supplemented yeast nuclear extract was applied to the column, which was then washed and eluted with DTT. Nuclear extract was used because Uba2p has been reported to be nuclear (10) and because we had previously found a yeast cytosolic fraction to be inactive in Smt3p conjugation (data not shown). The eluate contained predominantly recombinant Uba2p and Aos1p (data not shown), but as these bore His₆ tags, they were selectively removed by Ni-NTA chromatography. The resulting fraction contained four major bands of 90, 50, 40, and 19 kDa and a number of minor bands (Fig. 1). The 90- and 40-kDa bands are likely to represent the endogenous Uba2p and Aos1p from the yeast lysate. The other two bands were analyzed by direct mass spectrometric analysis of a protease digestion mixture (see "Experimental Procedures"). The 50-kDa band contained the elongation factor EF1, and the 19-kDa band contained Ubc9p.

[View Larger Version of this Image (31K GIF file)]

The isolation of EF1 in this fractionation was intriguing, as EF1 has been connected to Ub-like systems before, as an isopeptidase involved in Ub-dependent proteolysis of N--acetylated substrates (24). However, it is also an extremely abundant protein and a frequent contaminant in affinity purifications.² We attempted to purify EF1 and add it to some of the reactions described below, but did not obtain any conclusive results.

Smt3p-thioester formation assays were performed using purified, His₆-tagged recombinant proteins expressed in E. coli. Incubation of the Uba2p/Aos1p heterodimer with ATP and HF-Smt3p promoted formation of the ~105-kDa HF-Smt3p~Uba2p thioester (7) (Fig. 2, lane 6). When purified Ubc9p was also added to this reaction, a different HF-Smt3p-containing band formed at ~38 kDa (Fig. 2, lanes 2 and 7). In addition to Ubc9p, formation of this product required ATP, Uba2p, and Aos1p. Furthermore, the vast majority of this product could be destroyed by incubation with DTT (Fig. 2, lanes 1-8). The data are consistent with the 38-kDa product being the HF-Smt3p~Ubc9p thioester. The DTT-sensitive bond cannot be a disulfide bond, as HF-Smt3p does not contain any cysteine residues. Coomassie Blue staining of the reaction products provided additional evidence that Ubc9p was the other component of the 38-kDa product, as the Ubc9p bands were significantly depleted upon ATP incubation and formation of the 38-kDa band and reappeared upon addition of DTT (Fig. 2, lanes 1-3; Ubc9p ran as a ~20-kDa doublet under nonreducing conditions). Also, formation of the 38-kDa product was entirely dependent on the presence of Uba2p/Aos1p, strengthening the analogy between this reaction and that of Ub, E1s, and E2s. This reaction also formed small amounts of a series of DTT-insensitive HF-Smt3p-containing bands, one of which ran at the same position as the HF-Smt3p~Ubc9p thioester. These bands could represent a polymer of HF-Smt3p or isopeptide-containing conjugates of Smt3p to Ubc9p.

[View Larger Version of this Image (37K GIF file)]

We also asked whether Ubc9p was capable of forming thioesters with Ub as well as with Smt3p. When the yeast E2s Ubc4p, Pex4p (Pas2p/Ubc10p), or Rad6p (Ubc2p), which are involved in bulk proteolysis (25), peroxisome biogenesis (26), and DNA repair (27), respectively, were mixed with Uba1p (the yeast E1) and His₆- and FLAG-tagged Ub(A76) (see "Experimental Procedures"), addition of ATP caused these E2s to shift almost quantitatively into higher molecular weight forms (Fig. 3, lanes 1-6). These products were DTT-sensitive and contained HF-Ub(A76) (data not shown), suggesting that they were the corresponding Ub-E2 thioesters. No thioester product between Ubc9p and Ub could be detected under these conditions either by Coomassie staining (Fig. 3, lane 8) or by immunoblotting with the antibody against the FLAG epitope (Fig. 3, lane 9). Conversely, extremely little or no thioester product was detected between Smt3p and either Ubc4p, Pex4p, or Rad6p (data not shown), demonstrating that the transthiolation reaction is very specific.

[View Larger Version of this Image (41K GIF file)]

When HF-Smt3p and ATP were incubated with whole yeast cell lysate from wild-type cells, a series of DTT-resistant high molecular weight HF-Smt3p-containing bands formed (7) (Fig. 4, lane 2). The pattern of Smt3p-containing bands was similar but not identical to that seen in vivo, with a greater proportion of HF-Smt3p found in very high molecular mass bands >200 kDa (7). The ubc9-1 mutant, which has a Ser residue in place of Pro69, is temperature-sensitive by virtue of the fact that Ubc9(P69S)p is rapidly degraded at 37 °C (28). When yeast lysate made from this mutant strain after incubation at 37 °C was used in the same reaction, no Smt3p conjugation was detected even on long exposure. (Fig. 4, lane 3). Addition of recombinant Ubc9p restored the Smt3p conjugation activity (Fig. 4, lane 4). These results suggest a direct requirement for Ubc9p in HF-Smt3p conjugation to other proteins in these lysates. Thus, Ubc9p is likely to be the only Smt3p-conjugating enzyme in yeast, at least that is expressed under normal growth conditions.

[View Larger Version of this Image (52K GIF file)]

DISCUSSION

Our results show that the Smt3p conjugation pathway is distinct from the Ub pathway, at least up to formation of the thioester with the conjugating enzymes. However, sequence comparisons of of the Smt3p-conjugating enzyme Ubc9p to the Ub-conjugating enzymes Rad6p (Ubc2p), Ubc4p, and Pex4p (Pas2p/Ubc10p) do not reveal any large continuous regions of dissimilarity, and the degree of sequence similarity between Ubc9p and any of these three E2s is comparable with that among the E2s. Ubc4p is 37% identical to Ubc2p, 35% identical to Pex4p, and 34% identical to Ubc9p. Yet the transthiolation reaction involving Aos1p/Uba2p and Smt3p is extremely selective for Ubc9p, and the reaction involving Uba1p and Ub selects strongly against Ubc9p. One possible factor in this discrimination may be that Ubc9p is much more basic than the other proteins, with a pI of 9.2 as compared with 6.3 for Ubc4p, 5.4 for Pex4p, or 4.0 for Rad6p. How this specificity is generated depends on whether conjugating enzymes interact directly with the activating enzyme, the Ub-like protein, or both.

Although our inability to detect Ub~Ubc9p thioesters does not prove that Ubc9p never participates in Ub conjugation in vivo, it is unlikely that the metabolic stabilization of cyclins observed in ubc9 mutants (12, 13) results directly from reduced ubiquitination of these proteins by Ubc9p, both because of our results and because the E2s that participate in cyclin B ubiquitination have been isolated and do not include Ubc9p (14, 15). It is also very unlikely that Smt3p conjugation substitutes for ubiquitination in targeting cyclins for proteolysis, as cyclins have been heavily studied and never found to be Smt3p-conjugated. Furthermore, Smt3p, which is only 17% identical to Ub, probably does not target its substrates for proteasome-dependent proteolysis. Smt3p conjugation could affect cyclin proteolysis by activating some component of the ubiquitination/proteolysis machinery, or its effect could be several steps removed from ubiquitination. It also has not been excluded that Ubc9p could be required for the conjugation of a different Ub-like protein, which could mediate the cell cycle effect. Smt3p, Aos1p, and Uba2p are all essential genes (7, 10), which is consistent with their being required for cells to transit mitosis, but the arrest phenotypes of conditional alleles of these genes have not been characterized. Ubc9p, Smt3p, and SUMO-1 interact genetically or in the two-hybrid system with a variety of DNA-binding proteins, including centromere-binding proteins (29, 30), proteins involved in recombination (4, 31) and transcription factors (4, 32-34), any of which might affect cell cycle progression. Identification of the targets of Smt3p conjugation should address these questions.