Mapping the DNA Topoisomerase III Binding Domain of the Sgs1 DNA Helicase*

Several members of the RecQ family of DNA helicases are known to interact with DNA topoisomerase III (Top3). Here we show that the Saccharomyces cerevisiae Sgs1 and Top3 proteins physically interact in cell extracts and bind directlyin vitro. Sgs1 and Top3 proteins coimmunoprecipitate from cell extracts under stringent conditions, indicating that Sgs1 and Top3 are present in a stable complex. The domain of Sgs1 which interacts with Top3 was identified by expressing Sgs1 truncations in yeast. The results indicate that the NH2-terminal 158 amino acids of Sgs1 are sufficient for the high affinity interaction between Sgs1 and Top3. In vitro assays using purified Top3 and NH2-terminal Sgs1 fragments demonstrate that at least part of the interaction is through direct protein-protein interactions with these 158 amino acids. Consistent with these physical data, we find that mutant phenotypes caused by a point mutation or small deletions in the Sgs1 NH2 terminus can be suppressed by Top3 overexpression. We conclude that Sgs1 and Top3 form a tight complexin vivo and that the first 158 amino acids of Sgs1 are necessary and sufficient for this interaction. Thus, a primary role of the Sgs1 amino terminus is to mediate the Top3 interaction.

Several members of the RecQ family of DNA helicases are known to interact with DNA topoisomerase III (Top3). Here we show that the Saccharomyces cerevisiae Sgs1 and Top3 proteins physically interact in cell extracts and bind directly in vitro. Sgs1 and Top3 proteins coimmunoprecipitate from cell extracts under stringent conditions, indicating that Sgs1 and Top3 are present in a stable complex. The domain of Sgs1 which interacts with Top3 was identified by expressing Sgs1 truncations in yeast. The results indicate that the NH 2 -terminal 158 amino acids of Sgs1 are sufficient for the high affinity interaction between Sgs1 and Top3. In vitro assays using purified Top3 and NH 2 -terminal Sgs1 fragments demonstrate that at least part of the interaction is through direct protein-protein interactions with these 158 amino acids. Consistent with these physical data, we find that mutant phenotypes caused by a point mutation or small deletions in the Sgs1 NH 2 terminus can be suppressed by Top3 overexpression. We conclude that Sgs1 and Top3 form a tight complex in vivo and that the first 158 amino acids of Sgs1 are necessary and sufficient for this interaction. Thus, a primary role of the Sgs1 amino terminus is to mediate the Top3 interaction.
The Saccharomyces cerevisiae SGS1 gene encodes a member of the RecQ family of DNA helicases. In addition to the RecQ protein of Escherichia coli, this family includes the human BLM, WRN, RECQL4, and RECQ5 proteins as well as Rqh1 from Schizosaccharomyces pombe (1)(2)(3)(4)(5)(6)(7). These proteins play an important role in DNA metabolism as mutations in the human genes give rise to diseases characterized by genome instability and a predisposition to cancer. Werner's syndrome cells, which result from mutations in WRN (2), display a genomic instability termed variegated translocation mosaicism (8). Bloom's syndrome cells, which result from mutations in BLM (1), are characterized by increased rates of sister chromatid exchange and sensitivity to DNA-damaging agents (9). Mutations in RECQL4 are found in a subset of Rothmund-Thomson syndrome cases. These cells are characterized by elevated rates of chromosomal breaks and rearrangements (5,10). All members of this family contain a COOH-terminal domain with homology to RecQ, and all those that have been tested exhibit a 3Ј-to 5Ј-DNA helicase activity (11)(12)(13)(14)(15). In addition to the helicase domain, the eukary-otic proteins contain a large NH 2 -terminal domain of about 650 amino acids whose sequence is poorly conserved between members. The NH 2 -terminal domain is important for activity in yeast (16), but with the exception of the 3Ј-to 5Ј-exonuclease domain of WRN (17,18) the biochemical function of the NH 2terminal domain is unknown.
A subset of the eukaryotic RecQ family members has been shown to interact with DNA topoisomerase III (Top3) 1 (19 -22). Eukaryotic Top3 was first identified as a hyperrecombination mutant in yeast that also displayed a slow growth phenotype (23). Top3 has since been identified in several organisms including S. pombe (21,24), Caenorhabditis elegans (25), and humans (26,27). Like the bacterial enzyme, eukaryotic Top3 is a type I 5Ј-DNA topoisomerase with weak superhelical relaxing activity and a strict requirement for substrates containing single-stranded DNA or strand-passing activity (28,29). The biological function of Top3 is unclear, but in addition to its relaxing activity E. coli topoisomerase III is notable for its ability to decatenate gapped single-stranded DNA circles (29). The recent demonstration that eukaryotic Top3 and E. coli RecQ helicase functionally interact to catenate fully duplex DNA circles (30) suggested a role for these enzymes at the termination of DNA replication to decatenate daughter chromosomes (31,32). Although it has been suggested that RecQ helicases might function to restart stalled replication forks (7, 33-35) a role for Top3 in this process is unclear.
The SGS1 gene of yeast was identified as a mutation that suppressed the slow growth phenotype of top3 mutants (22). Thus, in contrast to top3 strains, top3 sgs1 double mutants exhibit a near wild type growth rate as well as suppression of other top3 phenotypes (22,36). Compared with wild type cells the sgs1 single mutant displays increased rates of mitotic recombination, both at the ribosomal DNA locus and throughout the genome (22,37), as well as increased rates of chromosome loss and missegregation (38). Like mutations in BLM, SGS1 mutations result in a hypersensitivity to methyl methanesulfonate (MMS) (16) and hydroxyurea (HU) (39).
SGS1 was cloned in a two-hybrid screen with TOP3, suggesting that Top3 interacted with the first 550 amino acids of Sgs1 (22). Because two-hybrid results do not provide evidence for direct binding, we set out to confirm this result biochemically, refine the domain of interaction, and determine whether binding was through direct protein-protein interaction. We identified an Top3⅐Sgs1 complex by coimmunoprecipitating and co-fractionating these proteins from yeast extracts. The results indicate that Sgs1 and Top3 are present in a stable complex and that the NH 2 -terminal 158 amino acids of Sgs1 are sufficient for complex formation. The proteins do not appear to form a simple heterodimer, however, because the full-length proteins cofractionate at a large native molecular weight. We determined that only the NH 2 -terminal 158 amino acids of Sgs1 were required to bind Top3 based on an enzyme-linked immunosorbent assay (ELISA) using purified proteins. These biochemical results are consistent with our observation that phenotypes caused by mutations in the first 158 amino acids of Sgs1 can be suppressed by overexpressing Top3, whereas larger deletions cannot.

EXPERIMENTAL PROCEDURES
Yeast Strains and Plasmids-Strain construction, growth, and transformation followed standard protocols (40). S. cerevisiae strain NJY620 expresses epitope-tagged versions of Sgs1 and Top3. This strain was constructed by modifying the chromosomal SGS1 gene of wild type strain CHY125 (41) by integrating BglII-linearized plasmid pJM1526, which places three consecutive HA epitopes (YPYDVPDYA) at the COOH terminus of Sgs1. This gene and protein are henceforth called SGS1-HA and Sgs1-HA, respectively. The chromosomal TOP3 gene was modified by integrating SphI-linearized pJM2565, which places a single V5 epitope (GKPIPNPLLGLDSTRTG, Invitrogen) followed by six histidines at the COOH terminus of Top3. This gene is henceforth referred to as TOP3-V5 and its encoded protein as Top3-V5. Strain WFY822 was created by integrating pJM2565 into strain NJY531 (sgs1::loxP) (16). Strain NJY560 was constructed by deleting the SGS1 and SLX4 genes of CHY125 (41) with loxP-KAN-loxP cassettes (42) and maintaining the strain with plasmid pJM500 (SGS1/URA3). SGS1 and sgs1-34 were integrated at the LEU2 locus of NJY560 to create strains BSY1228 and BSY1229, respectively. SGS1 mutant phenotypes were assayed as described (16).
Plasmid pJM1526, which expresses the epitope-tagged truncation Sgs1 645-1447 -HA, contains the insert from pSM105-HA (16) in the vector pRS405 (43). Plasmid pJM2565 contains a fragment of the TOP3 gene encoding a COOH-terminal in-frame fusion to the V5-His6 epitope (Invitrogen) in pRS404. To overexpress Top3 in yeast, TOP3 was subcloned downstream of the GAL1 promoter in pRS424 to make pJM2566. Plasmids expressing Sgs1-HA truncations were described (16), except for pKR1554 and pKR1555, which express epitope-tagged proteins Sgs1 1-158 -HA and Sgs1 1-322 -HA, respectively. To create these plasmids the first 474 and 966 base pairs of SGS1 were amplified by polymerase chain reaction so as to place an NdeI site in the context of the initiating ATG and an NotI site at the end of the coding region. These fragments were subcloned into NdeI/NotI-digested pSM100-HA (16). For expression of recombinant yeast proteins in E. coli, TOP3-V5 was subcloned into the T7-inducible vector pET11a (44), yielding plasmid pSAS402. Glutathione S-transferase (GST) fusion proteins were expressed by subcloning NdeI/BamHI fragments from pKR1554 and pKR1555 into pET11GTK-WF to create pKR1564 and pKR1565. Plasmid pET11GTK-WF was created by destroying the NdeI site of pET11GTK (45) and placing an in-frame NdeI downstream of the GST target coding region by polymerase chain reaction.
Yeast Extracts, Immunoprecipitations, and Immunoblotting-Extract preparation and chromatography were performed at 4°C. To prepare large scale extracts, yeast cells were grown in 12 liters of yeast extract-peptone-dextrose (YPD) at 30°C to A 600 ϭ 1.5; the medium was supplemented with an additional 2% dextrose and growth continued to A 600 ϭ 2.8, which yielded 90 g of cells, wet weight. Cells were washed once with H 2 O and resuspended in Buffer A (25 mM Tris-HCl (pH 7.5), 1 mM EDTA, 0.01% (v/v) Nonidet P-40, 10% (v/v) glycerol, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM DTT) plus 200 mM NaCl and the following protease inhibitors: 10 g/ml pepstatin, 5 g/ml leupeptin, 10 mM benzamidine, 100 g/ml bacitracin. The cells were broken in a bead-beater (Biospec Products) with 50% volume of glass beads in 30-s bursts (separated by 90-s pauses) for a total of 5 min of breakage. The lysate was centrifuged at 16,000 ϫ g for 10 min and the resulting supernatant cleared at 235,000 ϫ g for 90 min in a Beckman Ti45 rotor. This centrifugation was observed to pellet a significant portion of the chromatin as reported (46). The cleared lysate was precipitated by stirring 350 mg of (NH 4 ) 2 SO 4 /ml of lysate for 60 min followed by centrifugation at 188,000 ϫ g for 15 min. The pellet was resuspended in 84 ml of Buffer A and dialyzed to a conductivity of Buffer A plus 250 mM NaCl. Small scale extracts for immunoprecipitations were prepared as described (16). Protein concentrations were determined by the Bio-Rad protein assay using bovine serum albumin as a standard. Superose 6 chromatography was performed in Buffer B (25 mM Hepes-HCl (pH 7.5), 1 mM EDTA, 0.01% (v/v) Nonidet P-40, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM DTT) containing 150 mM NaCl at 0.4 ml/min. Fractions were collected, precipitated with trichloroacetic acid, and resolved by 10% SDS-PAGE.
Immunoprecipitations (IPs) were performed at 4°C essentially as described (16). Unless otherwise indicated, all IPs were performed by incubating extract with 1 l of anti-HA (Roche Molecular Biochemicals, 5 g/l) or anti-V5 (Invitrogen, 1 g/l) antibodies for 1 h in RIPA buffer (150 mM NaCl, 50 mM Tris-HCl (pH 7.5), 1% (v/v) Nonidet P-40, 0.5% (w/v) deoxycholate, 0.1% (w/v) SDS) (47). 20 l of protein-A Sepharose beads (Amersham Pharmacia Biotech) was added to each sample, followed by rocking for 1 h. The immune complexes were then washed three times with 1 ml of RIPA buffer. Following SDS-PAGE the gels were transferred to nitrocellulose membranes (48) and treated with either anti-V5-horseradish peroxidase or anti-V5 as the primary antibody (1:10,000). Blots were treated with anti-mouse horseradish peroxidase conjugate secondary antibody as required (1:10,000; Life Technologies, Inc.) and developed with chemiluminescence reagents (Life Technologies, Inc.) to detect Top3-V5. Blots were reprobed with anti-HA (1:10,000) as the primary antibody and treated as above to detect Sgs1-HA. For phosphate labeling experiments, yeast cells were grown and labeled with [ 32 P]PO 4 as described (49). Extract preparation and immunoprecipitations were then performed as described above.
Purification of Recombinant Yeast Proteins-Plasmids pET11GTK (expressing GST alone), pKR1564 (GST-Sgs1 1-158 -HA), pKR1565 (GST-Sgs1 1-322 -HA), and pSAS402 (Top3-V5) were transformed into E. coli BL21-RIL cells (Life Technologies, Inc.). Cells were grown by shaking in LB medium containing 0.1 mg/ml ampicillin at 37°C to an A 600 of 0.4. To induce the expression of the recombinant protein, cultures were treated with isopropyl-1-thio-D-galactopyranoside at a final concentration of 0.1 mM for 2 h at 37°C, except for cells expressing Top3-V5, which were induced for 6 h at 20°C. Induced cells were pelleted and resuspended in Buffer A plus protease inhibitors (above) containing 250 mM NaCl for GST and GST fusions, and 150 mM KCl for Top3. Extractions and chromatography were performed at 4°C, except where noted. Cell suspensions were incubated with 0.1 mg/ml lysozyme for 30 min and then sonicated three times for 1 min using a Branson sonifier 450 microtip at setting 4, 60% duty cycle. Lysed cells were clarified by centrifugation at 32,500 ϫ g and the supernatant collected as extract.
GST and GST-Sgs1 1-158 -HA proteins were purified by batch binding the extract from 1 liter of cells to 1 ml of glutathione-Sepharose 4B resin (Amersham Pharmacia Biotech) for 2 h. The resin was washed with 3 column volumes of Buffer A plus 250 mM NaCl, then half-column volume fractions were eluted at room temperature with Buffer A (pH 8.0) plus 150 mM NaCl and 10 mM glutathione. The peak fraction was determined by Bradford assay and SDS-PAGE, then 200 l was fractionated on a Superdex 75 (Amersham Pharmacia Biotech) gel filtration column in Buffer B plus 150 mM NaCl to achieve greater purity. GST-Sgs1 1-322 -HA extract from 2 liters of cells was diluted in Buffer A to a conductivity of Buffer A plus 50 mM NaCl and bound to an SP-Sepharose (Amersham Pharmacia Biotech) column at 20 mg of extract/ml of resin. SP-Sepharose was washed with 3 column volumes of Buffer A plus 200 mM NaCl, then GST-Sgs1 1-322 -HA was eluted in Buffer A plus 500 mM NaCl. The resulting SP 500 mM pool was diluted in half with Buffer A and affinity purified by glutathione-Sepharose 4B and Superdex 75 chromatography as above.
Top3-V5 containing extract from 3 liters of cells was bound to P-11 phosphocellulose (Whatman) at a ratio of 10 mg of extract/ml of resin in Buffer A plus 150 mM KCl. The column was washed with 3 column volumes of Buffer A plus 400 mM KCl, then Top3-V5-containing fractions were eluted from the column in Buffer A plus 600 mM KCl. Top3-V5-containing fractions were precipitated with 400 mg/ml (NH 4 ) 2 SO 4 for 1 h and then pelleted at 32,500 ϫ g. The resulting Top3-V5-containing pellet was resuspended in Buffer N (25 mM Tris-HCl (pH 8.0), 0.01% (v/v) Nonidet P-40, 10% (v/v) glycerol, 0.1 mM phenylmethylsulfonyl fluoride, 250 mM NaCl) plus 20 mM imidazole and batch bound to 1.5 ml of Probond nickel resin (Invitrogen) for 4 h. Resin was poured into a column and washed with 3 column volumes of Buffer N plus 20 mM imidazole and 10 column volumes of Buffer N plus 50 mM imidazole. Top3-V5 protein was then eluted in 6 half-column volume fractions of Buffer N plus 250 mM imidazole.
ELISAs-To detect a direct interaction between Top3 protein and the NH 2 terminus of Sgs1, 15 pmol of purified GST and GST-Sgs1-HA fragments were first immobilized in DYNEX Imulon 2 HB 0.4-ml wells. Immobilization of GST and GST-Sgs1-HA fragments was carried out in 75 l of PBS (10.1 mM Na 2 HPO 4 , 2.4 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl), pH 7.2, containing 0.1% Tween 20 (PBST) and 1 mM DTT by shaking at 60 rpm for 1 h at room temperature. After immobilization, wells were washed once with 0.4 ml of PBST plus 1 mM DTT, then blocked with 0.4 ml of 5% dried milk (w/v) in PBST plus 1 mM DTT for 1 h at room temperature. After blocking, cells were washed three times with 0.4 ml of PBST plus 1 mM DTT. A titration of 0 -40 pmol of Top3 protein was added to each set of coated wells in PBST plus 1 mM DTT in a volume of 75 l and incubated for 30 min at room temperature. After incubation with Top3-V5 protein the wells were washed three times with 0.4 ml of PBST. To detect the Top3-V5 protein, 100 l of anti-V5 antibody (diluted 1:5,000 in PBST plus 0.5% dried milk) was added to each well for 1 h at room temperature. Wells were then washed three times with PBST, and 100 l of anti-mouse horseradish peroxidase conjugate secondary antibody (diluted 1:5,000 in PBST plus 0.75% (w/v) dried milk) was added to each well and incubated for 1 h at room temperature. After the secondary antibody incubation, wells were washed three times with 0.4 ml of PBST, and then 200 l of 3,3Ј,5,5Јtetramethylbenzidine liquid substrate system for ELISA (Sigma) was added to each well and incubated for 30 min at room temperature. After incubation, 100 l of 0.5 N H 2 SO 4 was added to each well and the A 450 of each solution read to determine the amount of Top3-V5 protein present.

RESULTS
Functional Complementation of Epitope-tagged SGS1 and TOP3-To characterize the interaction between Top3 and Sgs1, we constructed yeast strains whose chromosomal copies of the SGS1 and TOP3 genes were modified to express the COOH-terminally tagged proteins Sgs1-HA and Top3-V5 (see "Experimental Procedures"). These strains allowed us to immunoprecipitate and immunoblot the products of stable singlecopy genes expressed under their native promoters. To verify that the epitope-tagged alleles behaved like wild type, we tested their ability to complement various sgs1 and top3 phenotypes. Two very sensitive measures of SGS1 and TOP3 activity are resistance to the DNA-damaging agent MMS and resistance to the DNA synthesis inhibitor HU (16). The strains expressing the tagged proteins were serially diluted and replica plated to medium containing MMS or HU. As shown in Fig. 1, the epitope-tagged strains grew as well as wild type on YPD plates and did not show the HU or MMS hypersensitivity characteristic of sgs1 or top3 strains. For example, top3 mutants grow very slowly on YPD; SGS1 TOP3-V5 cells do not display the slow growth of SGS1 top3 cells and in fact grow at the wild type rate (data not shown). Similarly, sgs1 strains grow somewhat slower than wild type, and SGS1-HA TOP3 cells grow noticeably faster than sgs1 cells. Whereas sgs1 and top3 single mutants are hypersensitive to MMS and HU (Fig.  1), the SGS1-HA and TOP3-V5 strains do not display either of these sensitivities; these strains grow like wild type in the presence of these drugs as does the SGS1-HA TOP3-V5 doubletagged strain. Based on these growth phenotypes we conclude that the epitope-tagged alleles SGS1-HA and TOP3-V5 function exactly like wild type.
Coimmunoprecipitation and Cofractionation of Sgs1 and Top3-To identify an interaction between Sgs1 and Top3, ex-tracts were prepared from a wild type strain and from strain NJY620 expressing Sgs1-HA and Top3-V5. Following incubation of the extracts with anti-HA or anti-V5 antibodies, the immune complexes were precipitated with protein A beads and analyzed by immunoblot. Using extracts from cells expressing the tagged proteins, we observed that anti-V5 precipitated Top3-V5, as expected, and coprecipitated Sgs1-HA (Fig. 2A,  lane 6). Similarly, anti-HA precipitated Sgs1-HA, as expected, and coprecipitated Top3-V5 ( Fig. 2A, lane 4). These signals are specific to the epitope-tagged proteins as extract from the untagged wild type strain showed no bands of corresponding size. We note that under optimal conditions Top3-V5 coprecipitated Sgs1-HA more efficiently than Sgs1-HA coprecipitated Top3-V5 ( Fig. 2A, compare lanes 2 and 4 with 6 and 8). The simplest explanation for this effect is that there is an excess of Top3 over Sgs1 protein in the extract. Such a result is consistent with the genetics of this system; lowering the Top3:Sgs1 ratio either by mutating TOP3 (22) or by overexpressing SGS1 (16) results in a profound growth defect.
The previous experiment indicates that Sgs1 and Top3 interact in cell extracts but does not address the strength of the interaction or whether these proteins require DNA to interact. We addressed these questions by varying the conditions of the immunoprecipitation from nonstringent (Buffer A plus 150 mM NaCl) to very stringent (RIPA buffer plus 50 g/ml ethidium bromide). As shown in Fig. 2B, the intensity of the Sgs1-HA signal that coprecipitated with Top3-V5 was unaffected by changing these conditions. Likewise, the efficiency with which Top3-V5 was coprecipitated with Sgs1-HA was unaffected by changing these conditions (Fig. 2B, lower panel). Both proteins were found to coprecipitate even under the harshest conditions. We conclude that Sgs1 and Top3 are stably bound and that their interaction is not mediated by DNA.
If Sgs1 and Top3 are present in a complex then they would be expected to cofractionate over a gel filtration column. An extract from NJY620 cells was fractionated over a Superose 6 gel filtration column, and the fractions were immunoblotted to determine the elution volumes of Sgs1-HA and Top3-V5 (Fig.  3). A portion of the Sgs1-HA and Top3-V5 proteins were found to elute with a similar profile, the peak of which corresponds to a native molecular mass of ϳ1.3 MDa (Fig. 3, top and middle  panels). Additional Top3-V5 signal was detected in a second peak close to the void volume, although this signal was not associated with Sgs1-HA (Fig. 3, middle blot). When WFY822 (sgs1 TOP3-V5) extract was fractionated on a Superose 6 column, only the Top3-V5 signal eluting near the void volume was detected (Fig. 3, bottom panel). We conclude that the 1.3-MDa peak of Top3 is Sgs1-dependent, and the Top3-V5 signal near the void is likely to represent aggregated Top3-V5 protein that is in excess of Sgs1-HA. If a single polypeptide of Sgs1 were to interact with a single polypeptide of Top3, the expected size would be 240 kDa. The larger size of 1.3 MDa suggests that these proteins have a different stoichiometry or are complexed with additional proteins.
TOP3 and SGS1 Interact through the NH 2 Terminus of SGS1-To determine the domain(s) of Sgs1 responsible for interaction with Top3, strain WFY822 (TOP3-V5 sgs1::loxP) was transformed with a series of plasmids expressing Sgs1-HA truncations under the control of the native SGS1 promoter (16). Extracts were prepared and IPs performed under RIPA conditions. A fragment of Sgs1 consisting of amino acids 1-652 (Sgs1 1-652 -HA) coprecipitated with Top3-V5 (Fig. 4A), consistent with a Top3 interaction domain in the NH 2 -terminal 550 amino acids as determined by the two-hybrid assay (22). In contrast, no interaction was detected between Top3-V5 and the DNA helicase domain of Sgs1 (Sgs1 645-1447 -HA) (Fig. 4B). These results indicate that the interaction between Sgs1 and Top3 is mediated through the NH 2 terminus of Sgs1. To map the interaction domain more accurately, fragments of Sgs1-HA containing all but the NH 2 -terminal 158 or 322 amino acids were expressed in the presence of Top3-V5. Neither Sgs1 159 -1447-HA nor Sgs1 323-1447 -HA was successfully coprecipitated with Top3-V5 (Fig. 4, C and D). This result indicates that the first 158 amino acids are necessary to detect an interaction with Top3-V5 under these conditions. We tested whether the first 158 amino acids were sufficient for this interaction and observed that Sgs1 1-158 -HA was indeed coprecipitated with Top3-V5 (Fig. 4E). When Sgs1 1-158 -HA was immunoprecipi-tated with anti-HA it migrated in between a doublet of small IgG chains on SDS-PAGE. Comparing this signal with that of a control IP from an untagged strain confirms the identity of this band as Sgs1 1-158 -HA (Fig. 4F). We were unable to test whether Sgs1 1-322 -HA could be coprecipitated with Top3-V5 because this protein was insoluble when expressed in yeast (data not shown). We conclude that the first 158 amino acids of Sgs1 are necessary and sufficient to interact with Top3-V5 in vivo.
After determining that the first 158 amino acids of Sgs1 are sufficient for interacting with Top3 in yeast extracts, we wanted to find out if this was due to a direct protein-protein interaction. We initially tried to express Top3-V5 and Sgs1-HA fragments in rabbit reticulocyte lysates and immunoprecipitate them under mild conditions (Buffer A plus 150 mM NaCl), but these assays revealed no interaction (data not shown). We then turned to the more sensitive ELISA to identify a direct interaction. GST fusions of Sgs1 1-158 -HA and Sgs1 1-322 -HA were expressed in bacteria and purified on glutathione beads. As shown in Fig. 5A, unfused GST protein and GST-Sgs1 1-322 -HA were highly purified, whereas GST-Sgs1 1-158 -HA contained several smaller bands that are likely to be breakdown products because their abundance varied between preparations. Recombinant full-length Top3-V5 was highly purified using Ni-affinity chromatography (Fig. 5A). ELISA wells were coated with 15 pmol of purified GST, GST-Sgs1 1-158 -HA, or GST-Sgs1 1-322 -HA and nonspecific sites blocked with 5% dried milk. Increasing amounts of purified Top3-V5 protein were then incubated in a series of wells prior to washing and detecting bound Top3-V5 with anti-V5 antibody and a chromogenic substrate. This assay revealed weak background binding of Top3-V5 to unfused GST protein that saturated at 30 pmol of input Top3-V5 (Fig. 5B). In contrast, both GST-Sgs1 1-158 -HA and GST-Sgs1 1-322 -HA bound increasing amounts of Top3-V5 protein. At the highest input level of Top3-V5, these Sgs1 domains bound three times more Top3-V5 than GST alone. This result demonstrates a direct protein-protein interaction between the amino terminus of Sgs1 and Top3. Little difference between GST-Sgs1 1-158 -HA and GST-Sgs1 1-322 -HA was detected, confirming that the first 158 amino acids contains a significant portion of the interacting domain.
Top3 Overexpression Complements Mutations in the NH 2 Terminus of Sgs1-We previously used a synthetic lethal screen to identify several novel "SLX" mutants that require SGS1 for viability (41). Phenotypically, sgs1⌬ and slx4⌬ single mutants are viable, but the sgs1⌬ slx4⌬ double mutant is dead. Because Sgs1 activity is essential for viability in this background, slx4⌬ mutants provide a genetic system to identify functional domains of SGS1. Structure-function analysis previously revealed that small NH 2 -terminal deletions or mutations in the DNA helicase domain of Sgs1 were lethal (16).
To address the question of why NH 2 -terminal deletions of Sgs1 were defective in this assay, we tested whether overexpression of Top3 could rescue the synthetic lethal phenotype. The starting strain, NJY560 (slx4⌬ sgs1⌬ pJM500, SGS1/ URA3), is nonviable on medium containing the drug 5-FOA because it selects against the SGS1/URA3 plasmid, which is essential for viability in this background. NJY560 was first transformed with a plasmid expressing the TOP3 gene under control of the inducible GAL1 promoter (pJM2566, GAL1p-TOP3) and then with a series of SGS1 deletions in a LEU2 vector. In contrast to the LEU2 vector alone, wild type SGS1 allowed these cells to grow on 5-FOA (Fig. 6A). Complementation of the synthetic lethal phenotype by SGS1 is independent of Top3 overexpression because growth is observed under both repressed (glucose) and induced (galactose) conditions. As ex-

FIG. 2. Coimmunoprecipitation of Sgs1-HA and Top3-V5. Panel
A, extracts were prepared from strain NJY620 (SGS1-HA TOP3-V5) expressing Sgs1 and Top3 epitope-tagged proteins (even numbered lanes) and a wild type strain (CHY125) expressing no epitope-tagged proteins (odd numbered lanes). 1 mg of each extract was immunoprecipitated with anti-HA or anti-V5 antibodies under RIPA conditions and the products resolved by SDS-PAGE. After transfer to nitrocellulose the membrane was probed with anti-V5-horseradish peroxidase to detect Top3 (left) or anti-HA to detect Sgs1 (right). Panel B, IPs were performed as above except that the NJY620 extract was prepared, and the immune complexes were washed under the following conditions: Buffer A plus 150 mM NaCl, RIPA buffer, or RIPA buffer plus 50 g/ml ethidium bromide, as indicated. The upper blot was probed with anti-HA to detect Sgs1, and the bottom blot was probed with anti-V5horseradish peroxidase to detect Top3. pected, a helicase-defective allele of SGS1 (sgs1-hd) and all NH 2 -terminal truncations of Sgs1 were lethal when streaked onto 5-FOA plates containing glucose (Fig. 6A). In contrast, when these strains were streaked on 5-FOA galactose, the sgs1-⌬ N50 and sgs1-⌬ N158 alleles displayed complementing activity. Neither sgs1-hd nor sgs1-⌬ N322 complemented, even when Top3 was overexpressed. These results indicate that although DNA helicase activity is required under all conditions, the first 158 amino acids of Sgs1 are not required if Top3 is overexpressed. The suppression of the sgs1-⌬ N158 allele is not specific to the synthetic lethal phenotype because Top3 overexpression also suppressed the MMS hypersensitivity of this allele (data not shown). The simplest explanation for these results is that deletion of amino acids 1-158 significantly impairs the interaction between Top3 and Sgs1, and increasing the Top3 concentration restores this interaction. Based on this assay we conclude that the size of the interaction domain in vivo must be larger than amino acids 1-158 and smaller than 1-322.
The sgs1-34 mutation was isolated as a temperature-sensitive allele of SGS1 caused by the amino acid change Q31P. At the restrictive temperature (37°C) sgs1-34 behaves like sgs1-⌬N158, suggesting that it lacks Sgs1 NH 2 -terminal function. 2 As a result of this mutation, strain BSY1229 (sgs1-34 slx4) is viable at 25°C but not at 37°C. We tested whether Top3 overexpression could suppress this NH 2 -terminal point mutation. Strain BSY1229 was transformed with pJM2566 (GAL1p-TOP3) and the transformants streaked on selective plates containing glucose or galactose at 25 or 37°C (Fig. 6B). When Top3 2 V. Kaliraman and S. J. Brill, manuscript in preparation.

FIG. 4. Coimmunoprecipitation of Top3 and Sgs1 deletions.
Strain WFY822 (sgs1⌬ TOP3-V5) was transformed with plasmids expressing the indicated Sgs1-HA protein fragments under the control of the natural SGS1 promoter. Cells were grown under selective conditions, whole cell extracts were prepared, and the indicated IPs were performed under RIPA conditions. The precipitated Sgs1-HA proteins were then detected by immunoblot using anti-HA antibody. The 50-kDa Sgs1 1-158 -HA protein (panel E) migrates between two small immunoglobulin proteins present in the anti-HA antibody. A control extract from the untransformed parent strain (panel F) was used to indicate the positions of the immunoglobulin bands that are detected by the secondary antibody. The following Sgs1-HA expression plasmids were used: -HA, or GST was immobilized in ELISA wells, blocked with 5% milk protein, and challenged with the increasing amounts of Top3-V5. The wells were washed, and the bound Top3-V5 was detected by treatment with anti-V5, horseradish peroxidase-conjugated secondary antibodies, and a chromogenic reagent. The reaction was stopped and the absorbance measured at 450 nm. expression was repressed by growth on glucose plates, the strain grew at 25°C but not at 37°C. However, when Top3 was overexpressed by growth in the presence of galactose the strain was able to grow at 37°C (Fig. 6B). The suppression of sgs1-34 by Top3 overexpression is allele-specific, as two other SGS1 temperature-sensitive alleles whose mutations map to the DNA helicase domain could not be suppressed (data not shown). As above, we conclude that the sgs1-34 mutation impairs the binding of Top3 to Sgs1 at the restrictive temperature and that increasing Top3 concentration restores the interaction.
Post-translational Modification of Sgs1-We suspected that the failure to identify a strong Sgs1-Top3 interaction by in vitro translation might be due to the requirement for an in vivo modification to Sgs1. We addressed this question by asking whether Sgs1 is phosphorylated in vivo. Yeast cells expressing Sgs1-HA or a control HA-tagged protein were grown in the presence of 32 P i . Extracts were prepared from these strains, and proteins were immunoprecipitated and analyzed by SDS-PAGE and autoradiography. As shown in Fig. 7, immunoprecipitation of Sgs1-HA results in a 32 P-labeled band migrating at 220 kDa, as expected for Sgs1-HA. The 220-kDa band is specific to Sgs1-HA as only the expected 100-kDa protein is precipitated with anti-HA from a control extract. An additional control revealed that only the expected 69-kDa RPA1 protein was immunoprecipitated from the Sgs1-HA extract with an antiserum to RPA1. Taken together, these data indicate that phosphorylated forms of Sgs1 and RPA1 do not interact under these conditions. We conclude that Sgs1 is phosphorylated during exponential growth in vivo.

DISCUSSION
Although a DNA fragment encoding the amino-terminal 550 amino acids of Sgs1 was isolated in a two-hybrid screen using Top3 as bait (22), there has been no biochemical confirmation of this interaction or any evidence of a direct interaction between these two proteins. To address the in vivo association of these proteins we created epitope-tagged alleles of SGS1 and TOP3 which were stably integrated at their chromosomal locations and expressed under their native promoters. These alleles were active in all of the biological assays we examined, indicating that the tagged proteins retain wild type function. Our data show that Sgs1 and Top3 can be coimmunoprecipitated under stringent buffer conditions including 0.1% SDS and ethidium bromide. These results are consistent with the idea that these proteins are present in a stable complex in vivo.
As summarized in Fig. 8, our deletion analysis indicates that a domain as small as the NH 2 -terminal 158 amino acids of Sgs1 is able to bind to Top3 in vivo. ELISAs using highly purified proteins provide biochemical evidence that Top3 binds Sgs1 1-158 through direct protein-protein interactions. As expected for such a physical interaction, Top3 overexpression suppressed phenotypes associated with mutations in the Sgs1 amino terminus. The fact that overexpressed Top3 suppressed Sgs1 with NH 2 -terminal deletions as large as 158 amino acids suggests that the Top3 interaction domain extends further than residues 1-158 in vivo (Fig. 8). Because overexpressed Top3 failed to suppress a deletion of 322 amino acids, we conclude that the Top3 interaction domain in vivo is larger than amino acids 1-158 and smaller than 1-322.
Recent evidence suggests that Top3 interacts with some, but not all, RecQ family members. Bacterial Top3 and RecQ were shown to interact functionally in vitro to catenate doublestranded DNA circles (30). In S. pombe, the top3 ϩ gene has been identified as an essential gene whose lethal phenotype is suppressed by mutations in rqh1 ϩ , the S. pombe RecQ homolog (21). In human cells, immunolocalization studies indicate that BLM is present in promyelocytic leukemia nuclear bodies together with Top3␣ (19,50). The human RecQ5␤ protein was also shown to colocalize and coimmunoprecipitate with Top3 (4). In contrast, there is as yet no evidence that WRN interacts with Top3, although it is associated with a large complex of replication proteins including topoisomerase I (51 6. Overexpression of Top3 suppresses NH 2 -terminal mutations in Sgs1. Panel A, strain NJY560 (sgs1⌬ slx4⌬ pJM500, SGS1/ URA3) was transformed with plasmid pNJ2566 (GAL1p-TOP3, TRP1) and the following LEU2 plasmids: pSM100 (SGS1), pJM531 (sgs1-⌬N50), pSM109 (sgs1-⌬N158), pSM110 (sgs1-⌬ N322), pSM100-hd (sgs1-hd), and pRS415 (vector). Transformants were streaked onto medium lacking tryptophan and leucine but containing 5-FOA and the indicated sugar to select stains growing in the absence of the URA3 plasmid pJM500. Panel B, strains BSY1228 (SGS1 slx4⌬) and BSY1229 (sgs1-34 slx4⌬) were transformed with plasmid pNJ2566 and streaked in duplicate onto selective plates lacking tryptophan but containing either glucose (glc) or galactose (gal) at 25 or 37°C. and BLM amino acid sequences, however, reveals no obvious similarities or motifs that might mediate the Top3 interaction. Our studies identified a single Top3 binding domain of Sgs1, whereas human Top3 was found to bind a second region of BLM (residues 1266 -1417) (20). This difference may be because that the far Western method included chemical cross-linking of the two proteins, which was not used in our studies. We conclude that this interaction is either not conserved in yeast or is not sufficiently stable to be detected by the methods used here.
While this work was in progress additional evidence of an interaction between Sgs1 and Top3 was reported. Bennett et al. showed that Sgs1 fragments bind Top3 in yeast extracts and inhibit Top3 activity in vitro (52). Maximal inhibition was obtained with a fragment of Sgs1 spanning residues 1-283. In addition, recent two-hybrid mapping studies identified a weak Top3 interaction domain between residues 1 and 116 of Sgs1 and a stronger interaction domain between residues 1 and 282 (53). Consistent with our genetic results (Fig. 6), these investigators proposed that a specific interaction between Sgs1 and Top3 is required for certain Sgs1 functions. The Top3 interacting domains identified in these studies agree closely with the Top3 binding domain identified in our experiments, as well as the NH 2 -terminal Top3 binding domain of BLM (20). Taken together with earlier data, it now appears that Sgs1 is more closely related to BLM than to WRN. Amino acid sequence analysis had initially shown that Sgs1 and BLM lack the exonuclease domain found in the NH 2 terminus of WRN (17). Genetic complementation experiments have also shown that BLM is capable of complementing the HU hypersensitivity, top3 slow growth suppression, and premature aging phenotypes of sgs1 mutants that WRN could not (39,54). Given that an interaction between Sgs1 and Top3 appears to be essential for Sgs1 activity, it will be of interest to determine whether the ability of human BLM to complement yeast sgs1 phenotypes depends on its ability to bind yeast Top3.
The mapping of the Top3 interaction to the region of amino acids 1-158 is significant in that small NH 2 -terminal truncations of Sgs1, such as expressed by the sgs1-⌬N158 allele, produce "hypermorphic" phenotypes that are more extreme than the null phenotypes (16). Considering that the phenotypes of top3 mutants are more extreme than sgs1 mutants, one might hypothesize that the full NH 2 -terminal domain of Sgs1 is required for complete Top3 activity in vivo. Although this model is consistent with the ability of overexpressed Top3 to suppress the sgs1-⌬N158 phenotype, it is inconsistent with the fact that Sgs1 fragments inhibit Top3 activity in vitro (52) and that sgs1-⌬N158 produces growth defects even in a top3 back-ground (16). An alternative model to explain these results is that a third factor interacts with Top3 and Sgs1. Deletions of the Sgs1 NH 2 -terminal 158 amino acids might result in growth defects by reducing binding to Top3 as well as this yet to be identified third factor.
The Top3⅐Sgs1 complex isolated from yeast is resistant to RIPA buffer, suggesting that the interaction is very stable. This affinity is retained even in a complex between Top3 and the relatively small Sgs1 1-158 -HA fragment (Fig. 4). Given this apparent high affinity, it is surprising that multiple studies have required very sensitive methods to detect interactions between Top3 and RecQ helicases in vitro. We required a very sensitive ELISA, and it was reported that a Top3⅐Sgs1 complex formed in vitro was dissociated by the relatively gentle conditions of 140 -250 mM NaCl (52). As mentioned above, chemical cross-linking was used to detect an interaction between Top3 and BLM (20). This suggests that the Top3⅐helicase complexes formed in vitro are different from those formed in vivo. This difference in affinity may simply reflect suboptimal binding conditions in vitro; higher protein concentrations and/or cotranslation might be required to form a stable complex. Alternatively, the correct interaction between Top3 and Sgs1 might depend on other cellular factors or modifications. As a test of this we found that Sgs1 is phosphorylated. The biological function of the phosphorylation is unknown, but it might regulate Sgs1 DNA helicase activity or the interaction of Sgs1 with Top3 or other proteins.
The Top3⅐Sgs1 complex eluted from a Superose 6 column at an approximate size of 1.3 MDa, indicating that it exists in a multimeric complex. Although we cannot rule out the possibility that a small amount of DNA mediates this complex, we feel it is unlikely for the following reasons. First, high speed extracts were used to remove bulk chromatin. Second, Top3 and Sgs1 coimmunoprecipitated despite treatment with ethidium bromide, which has been found to disrupt protein-DNA interactions. Third, we have examined the elution of doublestranded DNA by Superose 6 chromatography and found that DNA larger than 4 kilobases elutes in the void, whereas 1-kilobase DNA fragments elute at ϳ2 MDa. Thus, if the complex at 1.3 MDa were mediated by DNA, the fragments would have to be very small and discreet in size.
If the 1.3-MDa complex is a multimer, it might consist of a hexamer of Top3 with a hexamer of Sgs1 which would be expected to run at that size. Other helicases have been shown to exist as hexamers, such as E. coli DnaB (55) and, more significantly, human BLM (56). Alternatively, other proteins might be present in the complex as reported recently for BLM. In addition to its presence in promyelocytic leukemia bodies, BLM is associated with a number of human DNA repair proteins in the BASC complex including some that are conserved in yeast (50,57,58). Based on the interactions of BLM in human cells, it is possible that Sgs1 and Top3 are associated with additional proteins that contribute to its large native molecular weight and stable association. Purification of the Top3⅐Sgs1 complex will be required to determine which of these models is correct.