Identification of the yeast TOP3 gene product as a single strand-specific DNA topoisomerase.

The TOP3 gene of the yeast Saccharomyces cerevisiae was postulated to encode a DNA topoisomerase, based on its sequence homology to Escherichia coli DNA topoisomerase I and the suppression of the poor growth phenotype of top3 mutants by the expression of the E. coli enzyme (Wallis, J.W., Chrebet, G., Brodsky, G., Golfe, M., and Rothstein, R. (1989) Cell 58, 409-419). We have purified the yeast TOP3 gene product to near homogeneity as a 74-kDA protein from yeast cells lacking DNA topoisomerase I and overexpressing a plasmid-borne TOP3 gene linked to a phosphate-regulated yeast PHO5 gene promoter. The purified protein possesses a distinct DNA topoisomerase activity: similar to E. coli DNA topoisomerases I and III, it partially relaxes negatively but not positively supercoiled DNA. Several experiments, including the use of a negatively supercoiled heteroduplex DNA containing a 29-nucleotide single-stranded loop, indicate that the activity has a strong preference for single-stranded DNA. A protein-DNA covalent complex in which the 74-kDa protein is linked to a 5' DNA phosphoryl group has been identified, and the nucleotide sequences of 30 sites of DNA-protein covalent complex formation have been determined. These sequences differ from those recognized by E. coli DNA topoisomerase I but resemble those recognized by E. coli DNA topoisomerase III. Based on these results, the yeast TOP3 gene product can formally be termed S. cerevisiae DNA topoisomerase III. Analysis of supercoiling of intracellular yeast plasmids in various DNA topoisomerase mutants indicates that yeast DNA topoisomerase III has at most a weak activity in relaxing negatively supercoiled double-stranded DNA in vivo, in accordance with the characteristics of the purified enzyme.

Identification of the Yeast TOP3 Gene Product as a Single Strand-specific DNA Topoisomerase* (Received for publication, April 3, 1992) Raymond A. Kim and James C. Wang From the Department of Biochemistry and Molecular Biology, Harvard University, Cambridge, Massachusetts 02138-2092 The TOP3 gene of the yeast Saccharomyces cerevisiae was postulated to encode a DNA topoisomerase, based on its sequence homology to Escherichia coli DNA topoisomerase I and the suppression of the poor growth phenotype of top3 mutants by the expression of the E. coli enzyme (Wallis, J. W., Chrebet, G., Brodsky, G., Golfe, M., and Rothstein, R. (1989) Cell 58,[409][410][411][412][413][414][415][416][417][418][419]. We have purified the yeast TOP3 gene product to near homogeneity as a 74-kDa protein from yeast cells lacking DNA topoisomerase I and overexpressing a plasmid-borne TOP3 gene linked to a phosphate-regulated yeast p H 0 5 gene promoter. The purified protein possesses a distinct DNA topoisomerase activity: similar to E. coli DNA topoisomerases I and 111, it partially relaxes negatively but not positively supercoiled DNA. Several experiments, including the use of a negatively supercoiled heteroduplex DNA containing a 29-nucleotide single-stranded loop, indicate that the activity has a strong preference for singlestranded DNA. A protein-DNA covalent complex in which the 74-kDa protein is linked to a 5' DNA phosphoryl group has been identified, and the nucleotide sequences of 30 sites of DNA-protein covalent complex formation have been determined. These sequences differ from those recognized by E. coli DNA topoisomerase I but resemble those recognized by E. coli DNA topoisomerase 111. Based on these results, the yeast TOP3 gene product can formally be termed S. cerevisiae DNA topoisomerase 111. Analysis of supercoiling of intracellular yeast plasmids in various DNA topoisomerase mutants indicates that yeast DNA topoisomerase I11 has at most a weak activity in relaxing negatively supercoiled double-stranded DNA in vivo, in accordance with the characteristics of the purified enzyme.
The TOP3 gene of the yeast Saccharomyces cerevisiae was originally termed EDR1, mutations in which were found to increase the frequency of RAD52-dependent loss of a suppressor tRNA marker flanked by the 6 sequences, short repetitive sequences that are normally present in the terminal repeats of the yeast transposon Ty (Rothstein et al., 1987;Wallis et al., 1989). Cloning and sequencing of the gene showed that it encoded a protein homologous to Escherichia coli DNA topoisomerase I, with 21% identical plus 18% conserved amino acids at corresponding positions (Wallis et al., 1989). Because of this homology, the gene and its product were termed TOP3 and DNA topoisomerase 111, respectively.
That the limited homology between the bacterial and yeast gene might be functionally significant was supported by the finding that expression of the bacterial enzyme in yeast top3 mutants appeared to suppress the slow-growth phenotype of the mutants (Wallis et al., 1989). The same study found, however, that the presence of the bacterial enzyme failed to suppress the higher recombination frequency between 6 sequences in top3 strains. The lack of direct evidence that the TOP3 gene encodes a DNA topoisomerase prompted us to purify and characterize its product. In this paper, we show that a 74-kDa protein can be purified to near homogeneity from yeast cells overexpressing a plasmid-borne TOP3 gene. Biochemical characterization of the purified protein shows that it possesses a distinct DNA topoisomerase activity: it partially relaxes negatively but not positively supercoiled DNA; negatively supercoiled heteroduplex DNA containing a short single-stranded loop is readily relaxed by it. A protein-DNA covalent complex in which the 74-kDa protein is linked to the 5' end of a DNA strand has been identified, and the nucleotide sequences at many sites of covalent adduct formation have been determined. The biochemical characteristics of the yeast enzyme suggest that it resembles more E. coli DNA topoisomerase I11 than E. coli DNA topoisomerase I. In vitro, yeast DNA topoisomerase I11 is much less efficient than either yeast DNA topoisomerase I or I1 in the relaxation of supercoiled DNA substrates. In vivo studies also indicate that yeast DNA topoisomerase I11 has at most a weak activity in relaxing negatively supercoiled intracellular DNA. The plausible biological roles of this enzyme are discussed in light of these findings.

EXPERIMENTAL PROCEDURES
Strains-S. cerevisiae strains CH335 a his4-539 lys2-801 urd-52 (Holm et al., 1985) and CH1105 CY ade2-101 Aleu2 lys2-801 Atrpl urd-52 were obtained from Dr. Connie Holm, Harvard University. Strain TG205 a his4439 lys2-801 Atop1::URAJ top2-4 and its u r d derivative JCW2 have been described previously Giaever and Wang, 1988). Strain JCW173 a ade2-I01 Aleu2 lys2-801 Atop1::URAJ topJ-A3l::TRPl was derived from CH1105 by first switching the mating-type from CY to a, using the method of Herskowitz and Jensen (1991); two cycles of gene transplacement (Rothstein, 1983) were then carried out to introduce the mutations in the topoisomerase genes. The plasmid used in the deletion of TOPI was the one reported previously . For targeted inactivation of TOP3, a 31-bp' segment within TOP3 in pRK480 (see below), bounded by an NaeI and a HpaI site, was replaced by an 850bp segment containing the TRPl marker. An EcoRI to Sal1 restriction fragment containing the TRPl insert and its TOP3 flanking sequences was isolated and used for gene transplacement.
Plasmids for the Expression of Yeast TOP3-A partial EcoRI digest of S. cerevisiae strain CH335 (Holm et al., 1985) was used to prepare The abbreviations used are: bp, base pair(s); kb, kilobase(s); kbp, kilobase pair(s); EGTA, [ethylenebis(oxyethylenenitrilo)]tetraacetic acid; SDS, sodium dodecyl sulfate; HEPES, 4-(2-hydroxyethyl)-lpiperazineethanesulfonic acid. a genomic yeast DNA library in phage M13mp18 (Yanisch-Peron et al., 1985). To identify members containing TOP3, a pair of primers 5' CAGCTACGTAAAGGTG3' and 5'CCGTACTCTGTCCATCC3' were synthesized based on the published TOP3 nucleotide sequence of Wallis et al. (1989), and the polymerase chain reaction method (Mullis and Faloona, 1987) was used to obtain a 2.36-kb TOP3 fragment bounded by these primer sequences. Following purification of the fragment by agarose gel electrophoresis, it was cloned into the SmaI site of pUC18. Gel-purified restriction fragments from the TOP3 region of the clone were used to prepare 32P-labeled TOP3specific probes, using a commercial random-priming kit (Stratagene).
A clone in a yeast genomic library in phage M13 was identified by the use of the probes described above, and it was found to contain a 5.8-kb EcoRI fragment of yeast DNA. A 3.3-kb EcoRI-XbaI fragment containing the entire TOP3 gene was excised from this clone, subcloned first in between the same sites in pUC18, and then moved into pBluescript-SK (Strategene) as an EcoRI-SalI fragment. To facilitate the construction of the yeast TOP3 overexpression plasmids, a 7cycle amplification of a 450-bp segment in the pBluescript-TOP3 plasmid was carried out by the polymerase chain reaction, using a pair of primers 5'GGGATCCTCATGAAAGTGCTATGTGTC3' and 5'GGTATACTTGATCATTTTGmTGAG3'. The underlined ATG in the first primer corresponds to the initiation codon of the TOP3 open-reading frame, and the primer was designed to make this ATG a part of a BspHI site preceded by a BamHI site; the sequence of the other primer contains a BclI site 5'TGATCA3', a unique site in TOP3 about 430 bp downstream of the initiation codon. The gel-purified polymerase chain reaction product was used to replace the SrnaI-BclI segment of the TOP3 clone in pBluescript, yielding pRK480.
In the TOP3 overexpression clone pRK485, which was derived from pRK480 and a commercial vector pMAL-c (New England Biolabs), the following sequences were tandemly inserted in between the StuI and XbaI sites in the polylinker region of the vector: ( a ) GAATTCGCGGATCCTC3' and ( b ) a segment of yeast TOP3 from the initiation ATG codon to an XbaI site in the 3"untranslated region of the gene. The sequence in parenthesis in ( a ) encodes a decapeptide in the human c-myc protein, an epitope specifically recognized by a mouse monoclonal antibody MCY1-9E10.2 (ATCC/ NIH Repository Number CRL 1729). The bulk of the (a) sequence ending at the underlined GGATCC BamHI site was derived from a clone pMCT6 (Munro and Pelham, 19861, and routine recombinant DNA methodology was used in splicing these segments together to yield pRK485. Nucleotide sequencing of a region in the final clone, from the end of the malE gene of the vector to the beginning of the yeast TOP3 coding sequences, was carried out to ensure that the final product contained the desired junction region. Two clones pRK490 and pRK500 for the overexpression of yeast TOP3 in yeast were derived from pNKY2070, which was kindly provided to us by Drs. Wendy Raymond and Nancy Kleckner, Harvard University. The LEU2+ plasmid pNKY2070 was originally constructed for the overexpression of RAD50 from a phosphate-regulated PH05 gene promoter (Arima et ai., 1983). To obtain pRK490, a segment of pRK485, from an NcoI site near the end of the malE gene to a SalI site immediately downstream of the XbaI end of yeast TOP3, was used to replace the RAD50 gene between NcoI and SalI in pNKY2070. In this swap, the initiation codon of RAD50 in pNKY2070 was replaced by an internal methionine codon 50 amino acids from the end of the malE protein. In pRK500, the NcoI to BspHI segment in pRK490 was deleted; the joining of the matching NcoI and BspHI ends places the entire open reading frame of yeast TOP3 under the control of the pH05 promoter, with its initiation codon at the same position as that of the RAD50 gene in pNKY2070 (Raymond, 1990).
Protein Purifications-Cells of a S. cereukine strain JW173 Atopl top3 leu2 were first transformed to LEU2' with either pRK490 or pRK5OO. A colony of transformed cells was picked and grown overwith 2% glucose. Cells from 0.5 liter of culture were twice-pelleted night at 30 "C in minimal medium lacking leucine and supplemented and washed with sterile distilled water, and then resuspended in four 4-liter flasks, each containing 2 liters of YEP low phosphate medium supplemented with 2% glucose (for compositions of various media, see Sherman, 1991). The flasks were placed on a platform orbital shaker (New Brunswick) in a 30 "C chamber for 16 h. In the low phosphate medium, induction of the pH05 promoter-linked TOP3 fusion protein (in cells harboring pRK490) or TOP3 protein (in cells harboring pRK500) occurs, and the expression of a high level of either protein is apparently detrimental to cell growth; thus, during the 16-

5'GGGGGC(ATGGAGCAAAAGCTCATTTCTGAAGAGGAC)-TT
h period, the cells underwent fewer than 4 doublings. Cells were harvested by centrifugation for 10 min at 5 "C and 5,000 rpm; approximately 4 g of wet-packed cells were obtained from each liter of culture. All subsequent steps were carried out at 5 "C, and various pelleting steps were done by centrifugation at 12,000 rpm for 30 min. Cell pellets from 8 liters of culture were combined, washed once with ice-cold water, pelleted again, and resuspended in 100 ml of buffer A (20 mM Tris. HCl, pH 7.5, 1 mM EDTA, 1 mM EGTA, 1 mM NaHS03, 1 mM leupeptin, 10% glycerol, and 1 mM final concentration of freshly added phenylmethylsulfonyl fluoride). The suspension was mixed with an approximately equal volume of acid-washed glass beads (Sigma 150-212 pm) in a "Bead-beater" (Biospec Products), and cells were disrupted by 5 bursts of blending, each for a duration of about 1 min. The bulk of cell lysate was poured off the beads, and the beads were washed twice, each time with about the same volume of buffer A as that of the wet-packed beads. The lysate and washes were combined and clarified by centrifugation (Fraction

I, approximately 200 ml).
A freshly made 10% solution of Polymin P (Bethesda Research Laboratories) was added dropwise to Fraction I while it was being stirred gently. The mixture was stirred for an additional hour, and the pellet was removed by centrifugation. Saturated ammonium sulfate solution was then added slowly to the Polymin P supernatant (Fraction II), to a final concentration of 60% saturation. The precipitate was collected by centrifugation and then resuspended in sufficient volume of buffer A to reduce the electric conductivity to that of buffer A plus 300 mM KC1 (Fraction 111). Fraction I11 was chromatographed on a DE52 column (5 X 5 cm) pre-equilibrated with buffer A + 300 mM KCI. The flow-through (Fraction IV) was diluted with equal volume of buffer A and loaded on a 2.5 X 7-cm P11 phosphocellulose column pre-equilibrated with buffer A plus 150 mM KCI. After washing the P11 column with 10 bed volumes of buffer A plus 150 mM KCl, the column was eluted with 5 bed volumes of buffer A plus 400 mM KCl. The eluate (Fraction V) was applied directly to a 2.5 X 5-em single-stranded DNA-agarose column (Bethesda Research Laboratories) pre-equilibrated with buffer A plus 400 mM KCI. After washing the column with several bed volumes of the same buffer, a 100-ml linear gradient of buffer A plus 400 rn"1.5 M KC1 was applied. Fractions of 2 ml each were collected, and peak fractions containing the malE-TOP3 fusion protein or the TOP3 protein itself were found in the 0.7-1.5 M KC1 region of the gradient. Aliquots of fractions were examined by SDS-polyacrylamide gel electrophoresis, and peak fractions based on the patterns of Coomassie Blue-stained protein bands were pooled, diluted with buffer A to an equivalent KC1 content of 150 mM in conductivity, and loaded on a 2-ml P11 column. Elution with buffer A plus 400 mM KC1 yielded about 3 ml of Fraction VI. The final step described above served the purpose of concentrating the protein and reducing the KC1 concentration to 400 mM.

RESULTS
Purification of S. cerevisiae TOP3 Gene Produt-Initial attempts to detect the yeast TOP3 protein in crude or fractionated cell extracts as an activity capable of relaxing negatively supercoiled DNAs were unsuccessful: no ATP-independent relaxation activity was detectable in extracts of Atopl cells devoid of DNA topoisomerase I or in Atopl top2-4 cells lacking DNA topoisomerase I and containing a temperaturesensitive DNA topoisomerase 11; fractionation of cell extracts by several procedures commonly employed in the purification of enzymes involved in DNA metabolism, including ammonium sulfate precipitation and step-elution from a phosphocellulose column, also failed to yield an active fraction. These failures led us to adopt a strategy of overexpressing first an immunologically marked TOP3 protein and purifying it without an enzymatic assay; the characteristics of the tagged protein could then serve as a guide in the purification of the untagged protein.
Three plasmids were successively constructed for the expression of the TOP3 gene itself or TOP3 gene fused to coding sequences of other polypeptides. The first one, pRK485, was derived from a vector pMAL-c, a commercially available plasmid for the overexpression of fused genes in E. coli (see "Experimental Procedures" for plasmid construc-tions). In this overexpression clone, the yeast TOP3 gene is fused to the E. coli malE gene (Duplay et al., 1984). Codons for two peptide motifs, those for a tetrapeptide IEGR and those for a decapeptide EQKLISEEDL, are present in between the malE and TOP3 coding sequences. The decapeptide is specifically recognized by a monoclonal antibody MYC1-9E10.2, and its presence provides an immunotag in the fusion protein; the tetrapeptide motif is the recognition sequence of the factor X. endoproteinase and was carried over from the original pMAL-c cloning vector.
The other two plasmids were constructed for the expression of TOP3 in yeast. In pRK490 (Fig. lA, top), an open reading frame derived from an NcoI-Sal1 restriction fragment of pRK485 was placed downstream of an inducible yeast p H 0 5 gene promoter. This open reading frame contains, in succession, codons for the following stretches of amino acids: (a) the last 50 amino acids of E. coli malE protein, which begins with an in-frame methionine codon coincident with the NcoI site, ( b ) some 30 amino acids, including the IEGR and the decapeptide immunotag described above and amino acids encoded by linker oligonucleotides that were introduced during various stages of cloning, and ( c ) the entire yeast TOP3 polypeptide (Wallis et al., 1989). In pRK500 (Fig. Ut, bottom), the open reading frame downstream from the p H 0 5 promoter is that of the TOP3 gene.
The first plasmid pRK485 appeared to express the desired TOP3 fusion protein in E. coli, but the protein was unstable and little intact protein could be isolated from cell extracts (results not shown). Induction of the p H 0 5 promoter in yeast cells harboring the second plasmid, pRK490, produced a protein readily detectable by immunostaining with antibody MYC1-9E10.2. The molecular mass of this protein, as estimated from its electrophoretic mobility in SDS-polyacrylamide gel, is 82 kDa, which is in agreement with that expected from the p H 0 5 promoter-linked open reading frame containing TOP3.
When cells of a yeast strain JCW173 Atop1 top3 leu2 were transformed to LEU+ with pRK490, the slow growth phenotype of the strain in rich media, which is attributable to the top3 mutation (Wallis et al., 1989), was no longer observed (results not shown). This suggests that the 82-kDa TOP3 fusion protein is functionally similar to the yeast TOP3 gene product itself. In low phosphate media, however, the overproduction of the 82-kDa fusion protein is apparently detrimental to cell growth, and pRK490 transformants of either JCW173 or its TOP3' parent were found to grow poorly.
The 82-kDa fusion protein was purified to near homogeneity from cells harboring pRK490, using SDS-polyacrylamide gel electrophoresis and immunostaining of the fusion protein to monitor its concentration in various fractions. The purified fractions were found to possess a DNA topoisomerase activity (see below). Together, the genetic and biochemical results provide strong evidence that the 82-kDa TOP3 fusion protein is functional in uivo and possesses the catalytic activity of the TOP3 gene product in vitro.
Similar to results obtained with pRK490, transformation of Atop1 Atop3 cells with pRK500 abolishes the poor growth phenotype of the parent cells in rich media. To purify the plasmid-borne TOP3 gene product, JCW173 Atop1 top3 cells harboring pRK500 were first grown in a rich medium, and cells pelleted centrifugally from the culture were resuspended in a low phosphate medium to induce the p H 0 5 promoterlinked TOP3 gene. Purification of the TOP3 product from the induced cells was carried out according to the purification scheme for the 82-kDa TOP3 fusion protein (see "Experimental Procedures"). cerevisiae TOP3 gene, and MaZE-myc denotes sequences derived from the amino-terminal region of E. coli mnZE gene and a decapeptide from human c-myc (see the text for details). B, N, and S denote, respectively, BspHI, NcoI, and SaZI restriction site; the joining of B to N, denoted by NIB in pRK500, destroyed both sites. B, purification of yeast DNA topoisomerase 111. Electrophoretic patterns of proteins in the various fractions are depicted. SDS-polyacrylamide gel electrophoresis was carried out as described in Laemmli, 1970, using 5% stacking and 8% running gels. Following electrophoresis, the gel was stained with Coomassie Blue.   . 1B shows the Coomassie Blue-stained protein bands in the various fractions following their resolution by SDS-polyacrylamide gel electrophoresis. The purification of a protein with an apparent molecular mass of 74 kDa, which is that expected for the yeast TOP3 gene product, parallels that of the 82-kDa malE-TOP3 fusion protein. The 74-kDa protein was purified to near homogeneity following chromatography on single-stranded DNA embedded in agarose (Fig. lB, rightmost three lanes). As will be shown below, fractions containing purified 74-kDa protein contain a DNA topoisomerase activity, and the levels of this activity in fractions from DNAagarose chromatography correlate well with the intensities of the 74-kDa band in these fractions.
The TOP3 Gene Product Zs a DNA Topoisomerase-Purified 74-kDa protein has a distinct albeit weak relaxation activity with negatively supercoiled DNA substrates. Fig. 2A depicts the electrophoretic patterns of a 7.6-kbp negatively supercoiled plasmid DNA upon treatment with the protein under various conditions. Electrophoresis was carried out in a 0.1 M Tris borate buffer containing 5 pg/ml chloroquine. The untreated supercoiled DNA ran as a cluster of positively supercoiled topoisomers ahead of the nicked DNA band in the presence of the intercalating chloroquine (lane 1 ). Incubation of the DNA with the 74-kDa protein a t 37 "C, in 40 mM HEPES buffer (pH 7.5) and 1 mM MgC12, increased the mobilities of the DNA topoisomers due to increases of their linking numbers (lane 2). When incubation was carried out a t 65 "C instead of 37 "C, further relaxation of the negatively supercoiled DNA was observed (lane 3); above 65 "C, the relaxation activity was inactivated (results not shown). At either 37 or 65 "C, relaxation of the negatively supercoiled DNA was incomplete. Under the gel electrophoresis conditions employed, completely relaxed substrate DNA in the A 0 FIG. 2. A , relaxation of negatively supercoiled DNA by purified yeast DNA topoisomerase 111. Each sample contained 200 ng of a negatively supercoiled 7.6-kbp plasmid pK1 and 10 ng of purified DNA topoisomerase 111 in a total volume of 20 pl. Unless stated otherwise, the reaction buffer contained 40 mM HEPES-KOH (pH 7.5 at 20 "C) and 1 mM MgC12, and incubation was for 15 min at 37 "C. The treated samples were phenol-extracted and analyzed by electrophoresis in an 0.8% agarose gel, using 0.1 M pH 8.3 Tris borate, 2.5 mM EDTA, and 5 pg/ml chloroquine as the electrophoresis buffer. reaction buffer ran as a highly positively supercoiled band (lane 4 ) with a mobility greater than those of the yeast DNA topoisomerase 111-treated topoisomer clusters in lanes 2 and 3. When positively supercoiled DNA was used as the substrate, no change in the linking numbers of the topoisomers was detectable (results not shown).
The partial relaxation of negatively supercoiled DNA by the 74-kDa protein was detectable in a range of monovalent counterion concentrations below 150 mM; little difference was observed when NaCl, KC1, potassium acetate, or potassium glutamate was used. The omission of sulhydryl reagents or the presence of 10 mM dithiothreitol similarly showed no effect. Addition of exogenous Mg(I1) was not obligatory for the relaxation reaction; as shown in lane 5 of Fig. 2 A , incubation of negatively supercoiled DNA with the enzyme in 40 mM HEPES buffer (pH 7.5) resulted in partial relaxation of the DNA (compare the untreated and treated DNA run in lanes 1 and 5, respectively). The extent of relaxation of the DNA in the sample treated in the absence of added Mg(I1) was reduced relative to that incubated in the presence of 1 mM MgC12 (compare samples in lanes 5 and 2). However, when reaction was carried out in 40 mM HEPES plus 1 mM EDTA, no DNA relaxation activity was detectable (Fig. 2 A ,  lane 6). Addition of MgClz in molar excess (3 mM) relative to the concentration of EDTA restored activity (Fig. 2 A , lane 7). Activity was also readily detected when Mn(I1) instead of Mg(I1) was added to the HEPES-EDTA buffer, but Ca(I1) and Zn(I1) were found to be ineffective (data not shown). The optimal Mg(I1) concentration for the relaxation of negatively supercoiled DNA by the 74-kDa protein is around 1 mM; increasing the concentration to 5 or 10 mM resulted in less complete relaxation of the negatively supercoiled DNA (Fig.  2, lanes 8 and 9, respectively).
The results described above show that S. cerevisiae TOP3 gene encodes a 74-kDa protein, which, as predicted by its sequence homology to E. coli DNA topoisomerases I and I11 (Wallis et al., 1989;DiGate and Marians, 1989), possesses a DNA topoisomerase activity. This activity will be referred to as yeast DNA topoisomerase 111.
Yeast DNA Topoisomerase ZZZ Is Single Strand-specific-The partial relaxation of negatively but not positively supercoiled DNA by yeast DNA topoisomerase I11 is reminiscent of the reactions catalyzed by bacterial DNA topoisomerases I and can be interpreted in terms of a requirement for a short single-stranded segment in the DNA substrate (Wang, 1971). This single-stranded segment could either be present in the DNA before the binding of the enzyme, or it could be induced by the binding of the enzyme. In either case, the process is favored by negative supercoiling and disfavored by positive supercoiling of the DNA.
The requirement of a short single-stranded segment in the substrate of yeast DNA topoisomerase I11 is supported by several experiments described below. As shown in Fig. 2B , 1982). Double-stranded MP8 DNA was linearized at the SmBI site far away from the EcoRI-Hind111 region, denatured, and annealed with single-stranded circular MP8A29 viral DNA. The nicked circular heteroduplex DNA, which contains a 29-nucleotide-long single-stranded loop, was purified by agarose gel electrophoresis from the linear rena-tured MP8 homoduplex DNA and any remaining singlestranded species. As shown in lune 3 of Fig. 3, the gel-purified DNA migrated as the nicked circular species. Ligation of the nicked heteroduplex DNA in the presence of ethidium, followed by the removal of ethidium after ligation, converted a significant fraction of the nicked rings to the negatively supercoiled form, which migrated as an intense band near the bottom of Fig. 3, lune 7. Three minor bands are discernible in the lane 7 sample. In the order of increasing mobility, the first minor band is most likely the linear homoduplex DNA, as its mobility is the same as that of linear duplex mp8 DNA (lune 2 of Fig. 3). The next minor band is most likely the singlestranded circular form of MP8A29, the mobility of which is about the same as single-stranded circular MP8 DNA (lune 1 of Fig. 3). The fastest migrating minor band is probably a single-stranded circular DNA from a spontaneous deletion mutant of MP8, which is a contaminant in some of the preparations of MP8 phage or its derivatives.
Incubation of the heteroduplex DNA run in lane 7 of Fig. 3 with yeast DNA topoisomerase I11 resulted in the conversion of the negatively supercoiled band to a set of topoisomers extending all the way to the nicked DNA band (lune 6 of Fig.  3). This change is particularly striking in comparison to that observed for negatively supercoiled homoduplex MP8 DNA. The electrophoretic pattern of the homoduplex DNA under the electrophoresis conditions employed was unaffected by incubation with the yeast enzyme (compare the untreated and treated samples run, respectively, in lanes 4 and 5 of Fig. 3).
Yeast DNA Topoisomerase III Can Link Covalently to u DNA 5' Phosphoryl Group-The experiments described below show that similar to the cases with a number of other DNA topoisomerases (reviewed in Wang, 1985), a covalent complex between yeast DNA topoisomerase I11 and a DNA 5' phosphoryl group can be trapped by the addition of sodium dodecyl sulfate to the enzyme-DNA complex. The DNA samples were phenol-extracted and loaded in the sample wells of a 0.8% agarose slab gel. Following electrophoresis in TBE buffer (0.1 M Tris borate, 2.5 mM EDTA), the gel was blotted unto a nylon membrane (Genescreen, Du Pont) and probed with an M13derived 3ZP-labeled probe.
A pair of partially complementary oligonucleotides 5'-AATTTTGG)3' and 5'-TCGA(CGAAAATTCATAAATAG-CGAAAACCCGCG), which were originally synthesized for the conversion of a DNA with a 5'-TCGA overhang to one with a 12-nucleotide phage X left cohesive end (nucleotides in parentheses in the two sequences are complementary to each other), were used in these experiments. The oligonucleotides were 32P-labeled either a t their 5' ends by treatment with polynucleotide kinase, or, following annealing to allow pairing of the 31-nucleotide-long complementary sequences, at their 3' ends by repairing with the Klenow fragment of E. coli DNA polymerase I.
When yeast DNA topoisomerase I11 was first incubated with denatured 5'-end-labeled oligonucleotides, treatment of the complexes with SDS yielded no labeled protein-DNA covalent complex (Fig. 4, lune 4 ) . The same experiment with oligonucleotides with 32P-labeled nucleotides a t their 3' ends yielded, however, a cluster of radiolabeled bands following SDS-polyacrylamide gel electrophoresis (Fig. 4, lane 5). This GGGCGGCGACCT(CGCGGGTTTTCGCTATTTATGAA-1 2 3 4 5 6 -7. 5-""" FIG. 4. Transfer of "P from labeled DNA oligonucleotides to yeast DNA topoisomerase 111. The two partially complementary single-stranded oligodeoxyribonucleotides were annealed and labeled at their 5' or 3' ends as described in the text. The labeled oligonucleotides were purified by phenol extraction and exhaustive dialysis, first against 10 mM Tris-HCI, pH 8, 0.1 mM EDTA, and 1 M NaCI, and then against 10 mM Tris-HCI, p H 8,O.l mM EDTA. To form the protein-DNA covalent complex, labeled oligonucleotides were alkalidenatured by treatment with 0.1 M NaOH, chilled, neutralized with an equal volume of 0.1 M HCI and 0.1 M pH 7.5 Tris-HC1, and desalted by alcohol precipitation, washing, and resuspension in distilled water. Approximately 1.8 pmol of oligonucleotides and 160 ng of yeast DNA topoisomerase 111 were incubated as described in the legend to Fig. 2, and the reaction mixture was split into two equal portions. CaC12 was added to one to a final concentration of 2 mM, followed by the addition of 1 unit of staphylococcal nuclease (Worthington). Incubation was continued for 5 min for both samples, and an equal volume of a 2 X SDS-loading buffer, which contained 0.125 M Tris-HC1, pH 6.8, 4% SDS, 20% glycerol, 10% 2-mercaptoethanol, and 0.01% bromphenol blue, was added to each. Control experiments with the same radiolabeled oligonucleotides were also carried out with E. coli DNA topoisomerase I instead of yeast DNA topoisomerase 111, following the procedures of Tse et al. (1980). All samples were placed in a boiling water bath for 5 min, and the denatured protein in each was precipitated with 10% trichloroacetic acid. The pellet in each tube was washed twice with icy 10% trichloroacetic acid and resuspended in 50 pl of the 2X SDS-loading buffer. Several 1-pI portions of 1 M NaOH were gradually added to each sample to neutralize the excess acid in the acid-washed samples; neutrality was gauged by the return of the bromphenol-containing solution to a blue color. Electrophoresis in SDS-10% polyacrylamide gel was carried out as described (Laemmli, 1970) and was terminated when the blue tracking dye ran off the gel. The gel was stained with Coomassie Blue, destained, and dried between two sheets of gel-drying membrane (Promega). Autoradiography of the dried gel was done a t room temperature. Lane 1, 5'-end-labeled oligonucleotides treated with E. coli DNA topoisomerase I; lane 2.3"end-labeled oligonucleotides treated with E. coli DNA topoisomerase I; lane 3, same as in lane 2, except that the reaction product was digested with staphylococcal nuclease following incubation with E. coli DNA topoisomerase I; lanes 4-6, same as samples in lanes 1-3, respectively, except that yeast DNA topoisomerase 111 was used in place of E. coli DNA topoisomerase I. cluster of bands can be attributed to the attachment of oligonucleotides of varying lengths to the protein. In support of this notion, the slower migrating bands disappeared upon treatment of the products with staphylococcal nuclease (Fig.  4, lune 6). Staining of the same gel with Coomassie Blue prior to autoradiography also showed that the position of yeast DNA topoisomerase I11 coincided with the front of the radiolabeled cluster of bands (result not shown).
For an oligomer n nucleotides long with m of its 3"proximal nucleotides radiolabeled, cleavage within the first ( nm ) nucleotides from the 5' terminus would give a radiolabeled protein-oligonucleotide covalent complex only if the topoisomerase is linked to a 5'-phosphoryl end generated by the cleavage; cleavage within the last m nucleotides by such an enzyme would, however, give a radiolabeled protein-oligonucleotide covalent complex whether the topoisomerase is linked to a 5'-or 3'-phosphoryl end generated by the cleavage. Thus, based on the results shown in lunes 5 and 6 of Fig. 4 alone, an unequivocal conclusion could not be drawn on whether yeast DNA topoisomerase I11 is linked to a 5' or 3' phosphoryl group in the covalent complex. However, when these results are combined with the absence of a radiolabeled covalent complex with 5' end-labeled oligonucleotides (Fig. 4, lane 4 ) , it can be concluded that yeast DNA topoisomerase I11 is linked to a 5'-phosphoryl group in the formation of the covalent intermediate. Further confirmation of this conclusion was obtained through parallel experiments with E. coli DNA topoisomerase I. It is well-established that this enzyme is linked to a 5"phosphoryl group in the covalent intermediate (Depew et al., 1979;Tse et al., 1980); the labeling patterns of the E. coli enzyme upon incubation with the same oligomers and treatment with SDS are basically the same as tb.ose with yeast DNA topoisomerase I11 (Fig. 4, lanes 1-3).
Nucleotide Sequences of Yeast DNA Topoisomerase 1 1 1 Cleavage Sites Resemble Those of E. coli DNA Topoisomerase 111-Based on the finding described above that the yeast enzyme is linked to a 5'-phosphoryl group in the DNA-protein covalent intermediate, DNA strands uniquely radiolabeled at their 5' ends were used to map the cleavage sites: cleavage of such a strand would yield a protein-free labeled strand with a 3"hydroxy end, and the position of the 3' end could be mapped precisely by electrophoresis in a sequencing gel, using sequencing markers generated by priming the complementary strand with an oligonucleotide identical in sequence with the uniquely labeled end. Table I lists the nucleotide sequences of 30 yeast DNA topoisomerase I11 cleavage sites that have been mapped in two single-stranded DNA fragments. One was the viral strand of a restriction fragment of the replicative form of phage 4x174, which was radiolabeled at its 5' end generated by cutting with restriction endonuclease AvaII (nucleotide 5042). The other was a pBR322 restriction fragment from the Hind111 site (nucleotide 29) to the BamHI site (nucleotide 375), with its BamHI end radiolabeled. All sequences listed in Table I are in the 5' to 3' direction, and the internucleotide bond between positions -1 and +1 is the site of cleavage in each sequence.
The distributions of bases at various positions relative to the sites of cleavage are also tabulated at the bottom of Table  I. These distributions appear to be nonrandom. At position -3, for example, there appears to be a strong preference for the presence of an A. The very strong preference for a C at position -4 of the cleavage sites of E. coli DNA topoisomerase I (Tse et al., 1980;Dean et al., 1983) is not observed in the cleavage sites of yeast DNA topoisomerase 111. The particular +X174 DNA fragment was chosen because its sequence over- +3 4   (Dean, 1984). Significantly, with the exception of the second sequence from the top of the list, the other four sites were all found to be cleaved by both yeast and E. coli DNA topoisomerase 111. Relaxation of Supercoiled DNA by Yeast DNA Topoisomerase 1 1 1 in Vivo-In yeast cells expressing E. coli DNA topoisomerase I, it has been shown that plasmid DNA becomes positively supercoiled upon inactivation of both DNA topoisomerases I and I1 (Giaever and Wang, 1988). This phenomenon has been attributed to the preferential removal of negative supercoils by the bacterial enzyme from oppositely supercoiled domains generated by transcription or other processes involving the translocation of macromolecular assemblies along DNA (Liu and Wang, 1987). The very fact that positive supercoils accumulate in intracellular yeast DNA under these conditions shows that no DNA topoisomerase other than DNA topoisomerases I and I1 can efficiently remove positive supercoils in vivo. Furthermore, because the expression of the E. coli enzyme is needed for the accumulation of positive supercoils in the absence of yeast DNA topoisomerases I and 11, it appears that no other DNA topoisomerase in yeast can efficiently and preferentially relax negatively supercoiled intracellular DNA (Giaever and Wang, 1988). The present finding that purified yeast DNA topoisomerase I11 has a distinct albeit weak relaxation activity for negatively supercoiled DNA raised again the issue whether this enzyme might have a similar activity in uiuo. Examination of the endogenous 2-pm plasmid in yeast strain JCW2 Atop1 top2-4 cells harboring YEptopA-PGAL1, a plasmid expressing E. coli DNA topoisomerase I from the glucose-repressible and galactose-inducible yeast GAL1 gene promoter, confirmed the published results that 2-pm plasmid would become positively which the E. coli topA gene encoding E. coli DNA topoisomerase I is expressed from a galactose-inducible and dextrose-repressible GAL1 gene promoter (Giaever and Wang, 1988), was initially grown a t a permissive temperature of 26 "C in minimal media containing either 2% galactose or dextrose. The cultures were subsequently heated a t 35 "C for 2 h to heat-inactivate the top2-4 encoded mutant DNA topoisomerase 11. Following isolation of DNA from the cultures, twodimensional electrophoresis of the DNA samples in a 0.8% agarose slab was carried out as described previously (Giaever and Wang, 1988), using TBE buffer containing 0.6 and 3 pg/ml chloroquine in the first and second dimension analysis, respectively. The gel was blotted onto a nylon membrane and hybridized successively with radioactive probes derived from either yeast rDNA or 2 pm plasmid sequences. Only the autoradiogram obtained through the use of the radiolabeled rDNA probe is shown in the figure. The sample in lane I contained DNA from cells grown in the dextrose medium, in which the E. coli topA gene is repressed; the sample in lane 2 contained DNA from cells grown in the galactose medium, in which the E. coli gene is induced. In each case, the diagonal line is due mostly to sheared rDNA-containing chromosomal DNA and nicked circular and linear derivatives of extrachromosomal rDNA rings. supercoiled only when the GAL1 promoter was induced to express E. coli DNA topoisomerase I in the absence of active yeast DNA topoisomerase I or I1 (Giaever and Wang, 1988).
Examination of the extrachromosomal rDNA rings in the same strain (Kim and Wang, 1989) showed, however, that a small fraction of the extrachromosomal rDNA became positively supercoiled upon inactivation of DNA topoisomerase 11, even when cells were grown in a glucose medium (Fig. 5,  left). The presence of positively supercoiled rDNA rings was not due to leaky topA expression from the repressed GAL1 promoter, as lysates obtained from JCW2 Atopl top2-4 or TG205 Atop1 top2-4 cells devoid of the topA plasmid also contained comparable amounts of positively supercoiled rDNA rings (data not shown). In JCW2 Atopl top2-4 cells harboring YEptopA-PGAL1, induction of the topA gene followed by inactivation of DNA topoisomerase I1 a t 35 "C greatly increased the proportion of rDNA rings in the positively supercoiled form (Fig. 5, right).
These results suggest that in addition to DNA topoisomerases I and 11, there might be a relaxation activity in yeast that specifically removes negative supercoils. The activity is weak relative to the relaxation activity of DNA topoisomerase I1 a t 26 "C, as no positively supercoiled species was detectable in JCW2 or TG205 cells at this permissive temperature for the top2-4 mutant enzyme. I t is plausible that DNA topoisomerase I11 is responsible for this weak relaxation activity that specifically removes negative supercoils in rDNA rings in uiuo. However, because positive supercoiling of intracellular DNA is undetectable in the presence of active DNA topoisomerase I or 11, a direct confirmation of this possibility would require the use of a top1 top2 top3 triple DNA topoisomerase mutant; attempts to construct such a mutant have so far been unsuccessful.

DISCUSSION
That the TOP3 gene of S. cereuisiue might encode a DNA topoisomerase was previously postulated based primarily on its sequence homology with E. coli DNA topoisomerase I (Wallis et al., 1989). The findings described under "Results" show clearly that the protein indeed possesses a DNA topoisomerase activity; based on these results, the yeast TOP3 gene product can be formally termed yeast DNA topoisomerase 111.
In comparison with yeast DNA topoisomerases I and 11, yeast topoisomerase I11 is rather ineffective in the relaxation of supercoiled DNAs. In vitro, positively supercoiled DNA is refractory to the enzyme, and negatively supercoiled DNA is only partially relaxed by it. In uiuo, the enzyme has at most a weak activity in the removal of negative supercoils. These results suggest that, in contrast to the earlier postulate based on the suppression of the poor growth phenotype of yeast top3 mutants by E. coli DNA topoisomerase I (Wallis et al., 1989), the biological function of yeast DNA topoisomerase I11 is unlikely to be related to its relaxation of supercoiled doublestranded DNA.
Yeast DNA topoisomerase I11 is more avid with DNA substrates containing single-stranded regions. Relaxation of negatively supercoiled DNA by the yeast enzyme is more effectively inhibited by single-stranded DNA than by doublestranded DNA or tRNA. More strikingly, the introduction of a 29-nucleotide-long single-stranded DNA loop in a negatively supercoiled DNA converts it into a much better substrate for yeast DNA topoisomerase I11 (Fig. 3).
This preference for single-stranded DNA is reminiscent of the properties of both bacterial DNA topoisomerases I and 111. It has been shown previously that bacterial DNA topoisomerase I preferentially relaxes negatively supercoiled DNA, that the relaxation reaction is strongly inhibited by exogenous single-stranded DNA fragments, and that positively supercoiled DNA is a very poor substrate unless the DNA contains a short single-stranded region (Wang, 1971;Kung and Wang, 1977;Kirkegaard and Wang, 1985). Several properties of E.
coli DNA topoisomerase I11 also indicate that its action requires the presence of single-stranded regions in its substrates (Dean et al., 1983;DiGate and Marians, 1988).
The common nucleotide sequences of the preferred sites of DNA cleavage by yeast DNA topoisomerase I11 indicate, however, that the yeast enzyme is more closely related to E. coli DNA topoisomerase I11 than to E. coli DNA topoisomerase I. Both yeast and E. coli DNA topoisomerase I11 also exhibit a stronger preference for single-stranded regions than E. coli DNA topoisomerase I, and both lack a carboxyl-terminal segment corresponding to the carboxyl-terminal quarter of E. coli DNA topoisomerase I. Significantly, mutations in the gene topB encoding E. coli DNA topoisomerase I11 have recently been observed to stimulate recombination between repetitive sequences.* The molecular mechanism underlying the hyper-recombination phenotype of yeast top3 or E. coli topB mutants is unclear. One plausible interpretation previously proposed is that such an enzyme might be a component of a cellular assembly involved in the reduction of pairing of DNA strands of complementary sequences (Wang et al., 1990;Wang, 1991).. The reduction of recombination by enzyme systems has been raised before (see for example, Radman, 1991), and there is substantial genetic and biochemical evidence that various DNA topoisomerases are involved in the maintenance of genome stability (Christman et al., 1988;Kim and Wang, 1989;Wang et al., 1990). In a purified system for the replication of pBR322, E. coli * J. H. Miller, personal communication.
DNA topoisomerase I11 has been shown to be a key component in the unlinking of parental strands at the terminal stage of replication (DiGate and Marians, 1988). It is plausible that yeast DNA topoisomerase I11 may function similarly in vitro and in uiuo, either by itself or as a component of a more complex assembly. Because of the multiplicity of bidirectional replication origins in eukaryotic cells, inefficient unlinking of intertwined parental DNA strands near the terminal stage of replication might affect more the growth of eukaryotic than prokaryotic cells. Finally, DNA topoisomerases have emerged as important targets of antimicrobial and anticancer agents (reviewed in Liu, 1989). Because most of the DNA topoisomerase-targeting agents interfere with the DNA rejoining step catalyzed by these enzymes and thus convert normal cellular entities into DNA damaging agents, the efficacy of a drug does not depend on it target being an essential enzyme. The identification of the yeast TOP3 gene product as a DNA topoisomerase thus adds a potential target in the search of new therapeutics of this type.