A proteome-wide approach identifies sumoylated substrate proteins in yeast.

The ubiquitin-related protein SUMO-1 is covalently attached to proteins by SUMO-1 ligases. We have performed a proteome-wide analysis of sumoylated substrate proteins in yeast. Employing the powerful affinity purification of Protein A-Smt3 (Smt3 is the yeast homologue of SUMO-1) from yeast lysates in combination with tandem liquid chromatography mass spectrometry, we have isolated potential Smt3-carrying substrate proteins involved in DNA replication and repair, chromatin remodeling, transcription activation, Pol-I, Pol-II, and Pol-III transcription, 5' pre-mRNA capping, 3' pre-mRNA processing, proteasome function, and tubulin folding. Employing tandem affinity purifications or a rapid biochemical assay referred to as "SUMO fingerprint," we showed that several subunits of RNA polymerases I, II, and III, members of the transcription repression and chromatin remodeling machineries previously not known to be sumoylated, are modified by SUMO-1. Thus, the identification of a broad range of SUMO-1 substrate proteins is expected to lead to further insight into the regulatory aspects of sumoylation.

The ubiquitin-like protein SUMO-1 (Smt3 in yeast) is a 100residue protein that is conjugated to substrate proteins by sequential thioester transfer reactions via specific E1 1 activating enzymes (Uba2/Aos1) and E2 (Ubc9) conjugating enzyme (1,2). SUMO is conjugated to specific lysine residues on substrate proteins typically exhibiting consensus sites hKxE, where h is a hydrophobic amino acid (3,4). Unlike ubiquitylation, sumoylation of target proteins does not lead to proteasomal degradation but can affect diverse functions of the protein, such as subcellular localization, protein/DNA interaction, or enzymatic activity (5,6).
Smt3 is highly conserved and essential in yeast. In addition, the conjugating and deconjugating machinery are conserved from yeast to humans and perform essential functions in yeast. The only yeast proteins known to be modified by Smt3 are the nonessential septins involved in cytokinesis and the essential Pol30, Top2, and Pds5 (7)(8)(9). A growing number of proteins that are modified by SUMO-1 in mammalian cells have been reported (6). Recent proteomic approaches in mammalian cells have identified new proteins that are subject to SUMO-1 and SUMO-2 modification (10 -12). The confirmation of sumoylation of target proteins in these approaches was limited by the need for protein specific antibodies.
Previously, we have shown that the deconjugating enzyme Ulp1 is tethered to the nuclear pore channel via nuclear import receptors (13). The biological significance of such a tethering mechanism is poorly understood because of the lack of knowledge about its substrate proteins. Yeast has served as a powerful, rapid, genetic and in vivo biochemical model system to gain mechanistic insights into protein function in eukaryotes. Hence, we applied a proteomic approach in yeast to unravel the SUMO proteome. We have enriched sumoylated proteins from yeast cell lysates using the high affinity tag of Protein A (ProtA). In this study, we provide a list of proteins that are potentially modified by Smt3. Using rapid biochemical assays, we confirmed sumoylation for some of the isolated proteins.

EXPERIMENTAL PROCEDURES
Yeast and Microbiological Methods-Yeast strains and plasmids used in this study are listed in the supplemental information. Microbiological techniques, plasmid transformation and recovery, mating, sporulation of diploids, and tetrad analysis were done as described by Santos-Rosa et al. (17). Genomic integration of TAP (TRP1 marker) in a SMT3 shuffle strain was performed according to Rigaut et al. (18).
Affinity Purifications and "SUMO Fingerprint"-The ProtA-Smt3 containing the yeast strain was grown to an A 600 nm of 3.5 in 2 liters of YPD medium (yeast extract/peptone/dextrose) at 30°C. The cells were washed and resuspended in 10 mM Tris-HCl, pH 9.4, and 10 mM dithiothreitol. After 15 min of incubation, the cells were pelleted and resuspended in 1.2 M sorbitol, 50 mM potassium phosphate, pH 7.4. The cells were spheroplasted with 10 mg of zymolyase (20T) at 30°C for 30 min. The pelleted spheroplasts were treated with 50 mg of iodoacetamide for 15-30 min on ice and then lysed by vortexing with glass beads in LB buffer (150 mM KCl, 5 mM MgCl 2 , 1% Triton X-100, 20 mM Tris-HCl). After centrifugation, the supernatant was incubated with IgG-Sepharose beads at 4°C for 1 h. The beads were washed with LB buffer and increasing concentrations of MgCl 2 (0.1-2 M) followed by a wash with LB buffer and final elution with 0.5 M acetic acid (pH 3.5). The eluate was lyophilized, resuspended in SDS-loading buffer, and analyzed on a SDS 4 -12% gradient polyacrylamide gel. Western blot analysis was performed using anti-ProtA antibodies (14).
For the SUMO fingerprint, lysates from 10 units A 600 nm of cells were prepared using glass bead lysis in 20% cold trichloroacetic acid. The protein pellets were washed with cold ethanol, air-dried, and resus-* 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.
pended in 100 l of SDS sample buffer, and 5 l were analyzed on 4 -12% and 3-8% gradient gels. Western analysis was performed using anti-ProtA antibodies as described by Santos-Rosa et al. (17).
Site-directed Mutagenesis-Site-directed mutagenesis of Rpc128 was performed using the QuikChange site-directed Mutagenesis kit (Stratagene). Rpc82 was mutagenized using nested PCR-based techniques. Mutations were verified by DNA sequencing.
Protein Identification by Liquid Chromatography-Mass Spectrometry/Mass Spectrometry-The gel lane containing purified ProtA-Smt3 conjugates was cut in 48 slices across the entire separation range of the gel. Each gel slice was treated with trypsin as described by Shevchenko et al. (19). The generated peptide pools were subjected to partial peptide sequencing using nanocapillary reverse-phase chromatography (75-m inner diameter column) coupled directly to an ion trap mass spectrometer (LCQ DecaXP, Thermo Finnigan). The proteins were unambiguously identified by searching the peptide sequencing spectra against a translated version of the yeast genome using the software Mascot (Matrix Science). In cases where the identification of a protein was based on only a single sequenced peptide, the identification was independently verified by searching the mass spectrometry data using the sequence tag algorithm (20).

A Proteomic Approach to Identifying Sumoylated Substrate
Proteins in Yeast-To enrich for sumoylated proteins by affinity purification, a yeast strain expressing ProtA-tagged Smt3 instead of endogenous Smt3 was used. This modified Smt3 version was found to be functional in the temperature range from 23 to 37°C in a yeast strain lacking chromosomal SMT3 (Fig. 1A). Under conditions that preserve sumoylation (see "Experimental Procedures"), a cell lysate from the strain expressing ProtA-Smt3 was passed over an IgG-Sepharose column to enrich for ProtA-Smt3-modified proteins. Because sumoylated proteins could be part of multimeric complexes, we treated ProtA-Smt3 conjugates immobilized on IgG-Sepharose with increasing concentrations of MgCl 2 . This treatment is known to disrupt protein-protein interactions within protein complexes but does not dissociate the ProtA tag from IgG-Sepharose (14). Indeed, this harsh MgCl 2 incubation removed many nonsumoylated proteins, but ProtA-Smt3 conjugates remained bound to the beads (Fig. 1B). Finally, the ProtA-Smt3 conjugates were eluted from IgG-Sepharose by acetic acid and analyzed by SDS-PAGE. After staining with Coomassie Brilliant Blue, numerous bands ranging from 40 to 200 kDa became visible (Fig. 1B, lane 8; (Coomassie)). The corresponding Western blot using antibodies against the ProtA tag revealed that many of these bands carried ProtA-Smt3 (Fig. 1B, lane 8; (Western)). To elucidate the identity of these proteins, the gel lane was sliced into 48 pieces across the entire separation range to sample modified proteins without bias as to size or abundance. Each gel slice was trypsinized and subjected to mass spectrometric analysis (see "Experimental Procedures"). The data obtained are summarized in Table I. Although all of the proteins listed in Table I were unambiguously identified, we were unable to directly sequence the Smt3-modified peptides of the identified proteins. Therefore, it is possible that a fraction of the identified proteins are contaminants (e.g. abundant enzymes of metabolic/anabolic pathways or "sticky" proteins) or high affinity interactors to genuinely sumoylated proteins that were not removed by MgCl 2 treatment. However, for a representative number of identified proteins, we show that they are indeed sumoylated (see below). We have grouped the proteins isolated via ProtA-Smt3 into classes according to available functional annotations (Table I). Evidently, our proteomic approach has identified factors that are known to be sumoylated in yeast. These are the yeast septins Cdc3, Cdc11, and Sep7 (4) and Pol30 (PCNA), which is involved in DNA synthesis and repair (7). The identifi-  Table I and "Experimental Procedures"). cation of known sumoylated proteins confirms the validity of the applied method and demonstrates how our proteomic approach can be used to identify SUMO-targeted proteins in yeast. However, our list exhibits a vast number of proteins, which were previously not known to be sumoylated in yeast. Notably, these SUMO-carrying substrate proteins point to diverse functions in many cellular pathways.
Developing a SUMO Fingerprint-To confirm that ProtA-Smt3-purified proteins are sumoylated in vivo, we developed a simple verification assay, wherein a suspect target protein was genomically TAP-tagged at the C terminus to allow sensitive detection using anti-ProtA antibodies. In a next step, this TAPtagged putative target was co-expressed with either of the three different Smt3 constructs that differ in size: Smt3 (ϳ15 kDa), myc 3 -Smt3 (ϳ35 kDa), and GFP-Smt3 (ϳ45 kDa). The mobility of the target protein-SUMO conjugates generated in the three different yeast strains should decrease depending on the size of the different SUMO constructs. Thus, if a target is conjugated to Smt3, myc 3 -Smt3, or GFP-Smt3, one would expect a stepwise decrease in mobility on SDS-PAGE, thus generating a SUMO fingerprint for that substrate protein. To validate the assay, we tested a known SUMO substrate, Cdc3, which exhibits four sumoylation sites (4) (see also Table I). Cell lysates were prepared from Cdc3-TAP strains, which expressed either Smt3, myc 3 -Smt3, or GFP-Smt3 (see "Experimental Procedures"). After SDS-PAGE and Western blotting using anti-ProtA antibodies, both the unmodified and the various SUMOmodified Cdc3 forms could be discerned, which differ in size according to the Smt3 construct used (Fig. 2). Thus, the Cdc3 SUMO fingerprint confirmed the sumoylation of Cdc3 (4). Next, we applied this assay to a number of proteins present in our "SUMO catalogue," for which sumoylation has not been yet demonstrated. Clearly, Ubc9 (SUMO E2 carrier protein), Tup1, Ssn6, and Pdr1 (transcription factors), Rsc8 (chromatin remodeling), Bop3 (unknown function), and Rpc82 (Pol-III subunit) all revealed a typical SUMO fingerprint (Fig. 2). Altogether, we concluded from these studies that affinity purification of ProtAtagged Smt3 yielded novel sumoylated substrate proteins.
RNA Polymerases I, II, and III Are Sumoylated-In our SUMO proteome, we have obtained several subunits of RNA Pol-I, -II, and -III. To show sumoylation for these subunits in a direct way, we affinity-purified RNA Pol-I, -II, and -III complexes using several bait proteins. To probe for sumoylation, we affinity-purified Pol-III (via TAP-tagged Rpc82 and Rpc128) from cells that expressed myc-tagged Smt3 to allow Western analysis. In agreement with the SUMO fingerprint, sumoyl-  This study ated Rpc82 is detected in the purified Pol-III complex via the myc epitope (Fig. 3A, lanes 1 and 3). The second prominent SUMO-modified band in Pol-III could correspond to Rpc128 (Fig. 3A, lanes 1 and 3). To assign the Western bands to sumoylated versions of Rpc82 and Rpc128, we mutated the several consensus sumoylation sites in these proteins. Notably, the K17R mutation in Rpc128 and the K406R mutation in Rpc82 caused the disappearance of the attributed SUMO bands (Fig.  3A, compare lane 1 with 2 and lane 3 with 4). Mutating the other consensus sites (K927R in Rpc128, K406R in Rpc82) did not change the sumoylation pattern (data not shown). Similar studies were performed to demonstrate that subunits of RNA polymerases I and II are SUMO-modified as well. As baits for TAP purifications, we used Rpa190 (Pol-I), Rpb2 (Pol-II), Rpb8 (common subunit of Pol-I/II/III), and Rpac40 (common subunit of Pol-I/III). As shown in Fig. 3B, purified RNA polymerases I and II exhibited sumoylated bands mainly in the high molecular weight range of the gel (M r ϭ Ͼ100,000). By comparing the different sumoylated bands in various Pol-I/ II/III preparations, we concluded that Rpa190 and Rpa135 (the first and second subunit of Pol-I), Rpb1 (first subunit of Pol-II), and Rpc128 and Rpc82 (second and third subunit of Pol-III) are sumoylated. As expected, a mixed pattern of sumoylation was observed for purified Rpac40 and Rpb8, which were present in different RNA polymerases. Taken together, our data show that all three yeast RNA polymerases are targets of sumoylation. Notably, mainly the larger subunits of Pol-I/II/III become modified by SUMO.

DISCUSSION
Employing a powerful affinity purification of functional ProtA-tagged yeast SUMO from lysates in combination with mass spectrometry, we could isolate a large number of potential SUMO substrates and confirm sumoylation for a few of these proteins by different methods. Conspicuous targets of sumoylation are nuclear proteins, which are part of different multisubunit gene expression machineries. These include components of complexes involved in chromatin remodeling, rRNA and mRNA transcription, and mRNA processing (see Table I).
In agreement with these observations, both SUMO-conjugating and deconjugating enzymes are predominantly located in the nucleus and at the nuclear pore complexes (3). Moreover, our findings are consistent with data from the mammalian system, which have shown that many SUMO substrates are part of nuclear complexes with a wide role in transcription and genome organization (6). Moreover, the isolation of conjugates based on the ProtA methodology could be exploited for other ubiquitin-like proteins.
To show SUMO-1 modification, we have developed a rapid and diagnostic assay called the SUMO fingerprint. How sensi-  tive is our assay? Not all the proteins obtained in the proteomic approach exhibited the SUMO fingerprint, suggesting either the proteins represent contaminants or low abundance SUMO conjugates. Indeed, in the case of Rpc128, we were unable to obtain a SUMO fingerprint. However, upon enrichment by TAP-purification, we have been able to detect the sumoylation of Rpc128 via myc tagging of Smt3 and Western blot analysis. Moreover, Rna1 did not show the SUMO fingerprint or a myc-Smt3 Western signal when TAP-purified, 2 suggesting that it is a very low abundant conjugate or a contaminant in our preparation. The mammalian homologue of Rna1 (RanGAP) was shown to be modified by SUMO-1 (15). Nevertheless, both the methods could be useful to analyze SUMO modification in protein complexes and follow the changes of sumoylation patterns during formation and turnover of these large multimeric assemblies. Interestingly, the human homologues of the yeast RNA polymerase subunits Rpc160, Rpc128, Rpc82, Rpc37, Rpc25, Rpac40, and Rpabc27 are found to be "potentially" sumoylated in a proteomic approach using mammalian cells (10). Thus, sumoylation of RNA polymerases is conserved, suggesting that this post-translational modification is important for gene expression. Whether this modification affects the catalytic activity or assembly of RNA polymerases remains to be shown.
Did our large scale approach cover the entire SUMO proteome? We did not find Top2 (topoisomerase II) and Pds5 (sister chromatin cohesion), which are known to be sumoylated in yeast. It is possible that strong association with chromosomes (8,9) prevented their release into the soluble cell lysates. Moreover, sumoylated proteins that were not identified in our proteomic approach may be low abundant. Although we could not unravel the entire SUMO proteome, the potent affinity purification of Smt3-ProtA in combination with mass spectrometry was able to identify several unknown SUMO substrates. Indeed similar proteomic approaches in mammalian cells lead to the isolation of a distinct set of proteins (10,12). It is thus conceivable that only several different proteomic approaches will give a more comprehensive list of sumoylated proteins.
SUMO-1 has been implicated as a versatile regulator of protein function (16). However, two fundamental questions regarding its biological function remain to be answered. Why is SUMO-1 essential in most eukaryotic cells? What are these essential processes? The loss of Smt3 modification in Pol30, Top2, and Pds5 were shown not to be essential for the viability of yeast (7)(8)(9). Our mutational analysis, which inactivated the major sumoylation sites within Rpc128 or Rpc82, indicated that cell growth is not inhibited. 3 Thus, the influence of SUMO-1 on Pol-III function remains unknown. It is possible that inactivation of single sumoylation sites are not sufficient to affect Pol-III assembly/function. Clearly, more work is required to understand why Pol-III, as well as Pol-I and Pol-II, transcription machineries are sumoylated in vivo. Further, insights into these issues can be gained using the power of yeast genetics. Our catalogue of sumoylated proteins in yeast can trigger further studies to reveal the essential nature of sumoylation.