In vitro evolution of preferred topoisomerase II DNA cleavage sites.

Topoisomerase II is an essential enzyme that is the target for several clinically important anticancer drugs. Although this enzyme must create transient double-stranded breaks in the genetic material in order to carry out its indispensable DNA strand passage reaction, the factors that underlie its nucleotide cleavage specificity remain an enigma. Therefore, to address the critical issue of enzyme specificity, a modified systematic evolution of ligands by exponential enrichment (SELEX) protocol was employed to select/evolve DNA sequences that were preferentially cleaved by Drosophila melanogaster topoisomerase II. Levels of DNA scission rose substantially (from 3 to 20%) over 20 rounds of SELEX. In vitro selection/evolution converged on an alternating purine/pyrmidine sequence that was highly AT-rich (TATATATACATATATATA). The preference for this sequence was more pronounced for Drosophila topoisomerase II over other species and was increased in the presence of DNA cleavage-enhancing anticancer drugs. Enhanced cleavage appeared to be based on higher rates of DNA scission rather than increased binding affinity or decreased religation rates. The preferred sequence for topoisomerase II-mediated DNA cleavage is dramatically overrepresented ( approximately 10,000-fold) in the euchromatic genome of D. melanogaster, implying that it may be a site for the physiological action of this enzyme.

Topoisomerase II is an essential enzyme that is the target for several clinically important anticancer drugs. Although this enzyme must create transient doublestranded breaks in the genetic material in order to carry out its indispensable DNA strand passage reaction, the factors that underlie its nucleotide cleavage specificity remain an enigma. Therefore, to address the critical issue of enzyme specificity, a modified systematic evolution of ligands by exponential enrichment (SELEX) protocol was employed to select/evolve DNA sequences that were preferentially cleaved by Drosophila melanogaster topoisomerase II. Levels of DNA scission rose substantially (from 3 to 20%) over 20 rounds of SELEX. In vitro selection/evolution converged on an alternating purine/pyrmidine sequence that was highly AT-rich (TATATATACATATATATA). The preference for this sequence was more pronounced for Drosophila topoisomerase II over other species and was increased in the presence of DNA cleavage-enhancing anticancer drugs. Enhanced cleavage appeared to be based on higher rates of DNA scission rather than increased binding affinity or decreased religation rates. The preferred sequence for topoisomerase II-mediated DNA cleavage is dramatically overrepresented (ϳ10,000-fold) in the euchromatic genome of D. melanogaster, implying that it may be a site for the physiological action of this enzyme.
DNA topoisomerase II is an enzyme that is necessary for the survival of all proliferating cells (1,2). In addition to its normal functions in replication and mitosis, it is the target for some of the most widely prescribed drugs used in the treatment of human cancers (3)(4)(5)(6)(7)(8). The essential nature of topoisomerase II, as well as its role as a target for anticancer chemotherapy, extend from its unique status in the cell; it is the only enzyme known to create transient double-stranded breaks in the genetic material (9 -12). This ability to cleave and religate DNA in a concerted fashion allows topoisomerase II to disentangle topologically linked DNA molecules or alter the supercoiled state of nucleic acids without compromising the integrity of the genome (2). Conversely, when the cleavage/religation cycle of the enzyme is perturbed by anticancer drugs that enhance cleavage or inhibit religation, topoisomerase II is converted to a lethal enzyme that generates high levels of breaks in the DNA of treated cells (3,6,(12)(13)(14).
The reversibility of topoisomerase II-mediated DNA scission results from the fact that the enzyme forms a proteinaceous bridge that spans the double-stranded break and never releases its cleaved nucleic acid intermediate (15,16). Throughout its scission reaction, topoisomerase II remains covalently linked to the newly generated 5Ј termini of the cleaved DNA through the active-site tyrosyl residue of each of its two identical subunits (17,18). Although the covalent topoisomerase II-DNA "cleavage complex" is a fleeting intermediate in the catalytic cycle of the enzyme (9 -12), it ensures resealing of the double-stranded DNA break and prevents illegitimate recombination that would result from ligation of DNA termini to different nucleic acid molecules.
Sites of covalent topoisomerase II-DNA cleavage complex formation on any given nucleic acid substrate are reproducible and nonrandom, but the basis of this DNA sequence specificity remains obscure. Consensus DNA cleavage sequences have been determined for topoisomerase II from several eukaryotic species, ranging from Drosophila melanogaster to humans; however, these consensus sequences are generally weak and vary significantly from one another (9, 19 -22). In addition, DNA sequences that contain strong sites of enzyme-mediated scission have been identified that bear little relation to the published consensus for topoisomerase II from that species (16,19,23). Thus, the predictive value of consensus DNA cleavage sequences for the eukaryotic type II enzyme appears to be limited.
Because the nucleic acid sites at which topoisomerase II acts probably govern (to at least some extent) the ability of the enzyme to carry out its physiological functions (1, 2, 24 -27), it is critical to understand the basis by which topoisomerase II selects its specific sites of action on DNA. Therefore, a systematic evolution of ligands by exponential enrichment (SELEX) 1 approach (28) was utilized to address this fundamental issue for topoisomerase II from Drosophila. This approach differs from previous studies that mapped and compared enzymemediated DNA cleavage sites in that it identifies "preferred" (i.e. highly selected) rather than consensus (i.e. average) sites of topoisomerase II scission. Following 20 rounds of selection/ evolution based on enzyme-mediated DNA scission, a predominant 18-mer sequence emerged. This sequence is dramatically overrepresented in the Drosophila euchromatic genome, suggesting that it may represent a site of physiological action of topoisomerase II. cerevisiae wild-type topoisomerase II and the mutant ytop2Y783F enzyme (in which the active-site tyrosine was replaced by a phenylalanine) were overexpressed and purified from yeast cells as described by Elsea et al. (30) except that the initial phosphocellulose chromatography was replaced by hydroxylapatite (29). The construct utilized for the overexpression of ytop2Y783F was the generous gift of Dr. J. E. Lindsley (University of Utah) and has been described previously (31). The ␣ and ␤ isoforms of human topoisomerase II were overexpressed in S. cerevisiae and purified as described previously (32). Etoposide and amsacrine were purchased from Sigma, and the quinolone CP-115,953 was the generous gift of Drs. T. Gootz and P. McGuirk (Pfizer Central Research, Groton, CT). All drugs were stored at Ϫ20°C as 10 mM stocks in Me 2 SO. Tris and urea were purchased from Sigma; SDS was from Merck; proteinase K was from U. S. Biochemical Corp.; restriction endonucleases, calf intestine alkaline phosphatase, and polynucleotide kinase were from New England BioLabs; Klenow and Taq DNA polymerases were from Promega; [␣-32 P]ATP (6000 Ci/mmol) and Sequenase DNA polymerase were from Amersham Pharmacia Biotech; and Bluescript SKϩ phagemid was from Stratagene. All chemicals were analytical reagent grade.
Selection Protocol-Cleavage complexes were established at 30°C by incubating 200 ng of oligonucleotide substrate (100 nM) with 2 g of Drosophila topoisomerase II (100 nM) for 10 min in 50 l of cleavage buffer (10 mM Tris-Cl (pH 7.9), 50 mM NaCl, 50 mM KCl, 0.1 mM EDTA, and 2.5% glycerol) containing 5 mM CaCl 2 . CaCl 2 , rather than MgCl 2 , was employed because topoisomerase II generates significantly higher levels of single-stranded DNA breaks in the presence of Ca 2ϩ (33). This is important because the subsequent amplification of the cleavage products requires an intact template DNA strand. Cleavage complexes were trapped by the addition of 50 l of 4% SDS, followed by 50 l of 4 mM EDTA, and were precipitated by the addition of 50 l of 400 mM Tris-Cl, pH 7.9, 250 mM KCl, and 5 g/ml tRNA. After incubation on ice for 10 min, precipitates were collected by centrifugation at 14,000 ϫ g for 15 min at 4°C. Samples were resuspended in 200 l of 10 mM Tris, pH 8.0, 100 mM KCl, 1 mM EDTA, and 5 g/ml tRNA for 10 min at 45°C, reprecipitated by shifting the temperature from 45°C to ice for 10 min, and collected by centrifugation as above. Precipitates were resuspended at 45°C for 10 min in 200 l of water containing 5 g/ml tRNA, ethanol-precipitated twice, washed with 95% ethanol, and dried under partial vacuum at room temperature. Selection reactions were carried out in quadruplicate and pooled prior to the second ethanol precipitation.
Amplification of Selected DNA-Selected DNA molecules (i.e. molecules isolated from cleavage complexes) were amplified under mutagenic conditions (34) as follows. The dried pellets from the selection procedure were resuspended in 198 l of 10 mM Tris-Cl, pH 8.3, 50 mM KCl, 7 mM MgCl 2 , 0.5 mM MnCl 2 , 1 mM dCTP, 1 mM TTP, 0.2 mM dATP, 0.2 mM dGTP, and 0.01% gelatin containing 2 ng/l each of primers 1 and 2. Amplification was initiated by the addition of 2 l (10 units) of Taq polymerase. Samples were overlaid with 100 l of light mineral oil and cycled in an Ericomp TwinBlock thermal cycler, using the following program: 5 min at 94°C; followed by 20 cycles each consisting of 1 min at 94°C, followed by 1 min at 50°C and 1 min at 72°C. After the last cycle, 2 g of each primer was added, reactions were incubated at 94°C for 5 min, and one additional cycle was run as above. Since the template DNA molecules for this procedure were heterogeneous, this last annealing and extension in the presence of excess primer was important to ensure that both strands of the DNA substrate used for the next round of selection were properly base-paired and exactly complementary for any individual DNA molecule.
Experiments with a control oligonucleotide of known sequence indicated that the rate of misincorporation was ϳ0.4 bp/oligonucleotide molecule/round (which included 20 cycles of amplification) of SELEX. Therefore, the products of the final cycle of amplification for any given round of SELEX should have contained no more than a single base pair mismatch for every 50 oligonucleotide molecules. Since there is no mechanism by which mismatched base pairs can be specifically maintained from round to round of the SELEX protocol, it is unlikely that the potential existence of a low level of mismatches in the DNA affected the overall selection process.
Following amplification, samples were ethanol-precipitated, resuspended in 40 l of water and 4 l of loading buffer (10 mM Tris, pH 7.9, 60% sucrose), and subjected to electrophoresis in an 8% nondenaturing polyacrylamide gel at 10 watts for ϳ2 h. Amplification products were located by shadowing with ultraviolet light, and the DNA band was excised from the gel and eluted overnight in 400 l of 0.5 M ammonium acetate, 10 mM magnesium acetate, 1 mM EDTA. The eluted DNA was ethanol-precipitated and resuspended in 50 l of water, and its concentration was determined by measuring its absorbance at 260 nm. This DNA was then used as the substrate for a new round of selection.
To control for the possibility of contamination during the SELEX protocol, samples with no added DNA were run alongside the normal samples for every step of the selection and amplification procedures. When the DNA from the amplification reactions was gel-purified, a corresponding gel slice from the contamination control reaction lane was also excised and incubated overnight in the elution buffer. After ethanol precipitation and resuspension, the same volume of the control sample was used as the "DNA" for the contamination control reaction of the next round of selection and amplification. None of the contamination controls produced a band upon UV shadowing of the gel.
K ϩ /SDS DNA Cleavage and Religation Assays-Levels of cleavage complex formation were monitored by the K ϩ /SDS precipitation assay (35,36). Oligonucleotide pools (from each round that was assayed) were digested with EcoRI and BamHI restriction endonucleases, and the ends were filled in using Klenow fragment and [␣-32 P]dATP in the presence of nonradioactive nucleotides (37). Labeled oligonucleotides were gel-purified as described above. DNA cleavage reactions contained 100 nM Drosophila topoisomerase II and 5 nM oligonucleotide substrate and were carried out for 10 min at 30°C in 50 l of cleavage buffer containing either 5 mM CaCl 2 or MgCl 2 . For reactions that utilized yeast topoisomerase II, no KCl was used, and the NaCl concentration was 100 mM. For reactions that utilized human topoisomerase II␣ or ␤, the cleavage buffer was the same except that no NaCl was used and the KCl concentration was 100 mM. When DNA cleavage reactions were carried out in the presence of drugs, 100 M drug was included such that the final Me 2 SO concentration was 1%. Unless stated otherwise, DNA cleavage reactions were always carried out in the presence of 5 mM CaCl 2 .
As a prelude to determining rates of DNA religation, cleavage complexes were established in cleavage buffer containing CaCl 2 and trapped by the addition of EDTA (5 mM final concentration). Following incubation at 30°C for 2 min, NaCl was added (250 mM final concentration of additional salt), and samples were equilibrated at 30°C for 1 min. DNA religation was initiated by the addition of MgCl 2 (10 M final concentration) and terminated by the addition of SDS (2% final concentration) at various time points. Levels of DNA cleavage complex remaining were determined by the K ϩ /SDS precipitation protocol described above.
DNA Binding Assays-Binding of the mutant yeast topoisomerase II, ytop2Y783F, or the wild-type Drosophila enzyme to oligonucleotide pools was determined using a nitrocellulose filter assay (38). Reactions were carried out in 10 l of the appropriate cleavage buffer in the presence (yeast) or absence (Drosophila) of 5 mM MgCl 2 . Concentrations of oligonucleotide (radioactively labeled) and enzyme were as described above for the K ϩ /SDS assays. After incubation at 30°C for 10 min, reactions were spotted onto nitrocellulose filters (Millipore; presoaked in assay buffer), and the filters were washed three times with 1 ml of cold cleavage buffer. Filters were dried, and the radioactivity retained was determined by scintillation counting.
Cloning and Sequencing of in Vitro Selection Evolution Products-The oligonucleotide pool from round 20 of the SELEX procedure was reselected as above and amplified under nonmutagenic conditions: 30 cycles as above, but with 20 fmol of round 20 DNA as the template in 10 mM Tris-Cl (pH 9.0); 50 mM KCl; 0.1% Triton X-100; 1 mM each TTP, dATP, dGTP, and dCTP; and 1 g each of primers 1 and 2. Oligonucleotide products were digested with BamHI and EcoRI restriction endonucleases and ligated into linearized Bluescript SKϩ phagemid. Ligation was performed using the Boehringer Mannheim Rapid DNA Ligation Kit. Ligation products were used to transform Escherichia coli, and colonies were selected that contained the phagemid plus insert as described by the manufacturer. Phagemid from the individual clones was purified using the Boehringer Mannheim High Pure Plasmid Isolation kit, and the sequences of the inserts were determined using the Amersham Pharmacia Biotech Sequenase kit and a sequencing primer that was complementary to phagemid sequence near the insert (5Ј-AAAGCTGGAGCTCCACCGCG-3Ј).
Mapping of Cleavage Sites-Sites of DNA cleavage were determined as described previously (39). The composition of cleavage reactions was as above for the K ϩ /SDS assays except that the concentration of oligonucleotide was 80 nM.
Double-stranded DNA Cleavage-In order to verify that scission of the selected sequences was double-stranded, DNA cleavage reactions were carried out as above for the K ϩ /SDS assays (in Mg 2ϩ -containing buffer), terminated with SDS, digested with proteinase K (0.8 mg/ml) for 30 min at 37°C, and subjected to electrophoresis in a 14% nondenaturing polyacrylamide gel (cooled to 10°C) at 10 watts for ϳ2 h. Radioactive DNA cleavage products were visualized by PhosphorImager (Molecular Dynamics) analysis.
Genome Searches-Searches of the European Drosophila Genome Project data bases were conducted using the blastn search program (40) with the threshold set at 1000.

SELEX Scheme and Rationale-Topoisomerase II interacts
with ϳ28 bp on its DNA cleavage helix (41,42) and requires a minimum of 16 bp for efficient DNA scission (43). In an effort to define the mechanism by which the enzyme recognizes its site of action on nucleic acid substrates, previous studies have mapped sequences at which topoisomerase II cleaves DNA. This approach has resulted in a series of weak consensus cleavage sequences for several eukaryotic type II enzymes including Drosophila, chicken, mouse, and the ␣ and ␤ isoforms of human (19 -22). However, these consensus sites exhibit little sequence agreement among themselves and even disagree regarding the position of preferred bases relative to the point of cleavage. The variability between reported sequences notwithstanding, topoisomerase II displayed at least some level of specificity for 6 -10 sequence positions in each of these studies (9, 19 -22). Assuming an average of eight base-specific points of contact between topoisomerase II and its DNA cleavage site, it would be necessary to determine every site of action for the enzyme in ϳ65,000 bp of random DNA in order to generate all possible sequences (i.e. 4 8 combinations) at these positions. Since the size of DNA substrates utilized for the generation of consensus sequences generally ranged between 1000 and 10,000 bp (19 -22), it is clear that (on average) only a small fraction of the necessary cleavage sites have been sampled in order to confidently define the intrinsic specificity of topoisomerase II.
In light of the above, we have utilized an alternative approach to address this fundamental issue of topoisomerase II specificity. Rather than attempting to generate a consensus sequence for enzyme action based on mapping sites in a larger fragment of DNA, a SELEX protocol (28) was employed to select/ evolve preferred sites of DNA cleavage mediated by Drosophila topoisomerase II from a pool of ϳ10 12 potential sequences.
The scheme utilized for the present study is shown in Fig. 1. The initial DNA substrate employed for the SELEX protocol was a 60-mer oligonucleotide that incorporated two critical features: 1) it included 20 bp of defined flanking sequences (derived from pBR322) at each end that were devoid of topoisomerase II cleavage sites (16,44) and contained the indicated restriction endonuclease recognition sites for eventual cloning; and 2) it included a 20-bp core of random DNA sequence that was synthesized using an equimolar ratio of all four bases. This random portion of the substrate allowed the type II enzyme to select/evolve preferred sites of DNA cleavage from among as many as 4 20 sequences, affording topoisomerase II the opportunity to sample pools of potential cleavage sites many orders of magnitude larger than those used in previous studies.
Drosophila topoisomerase II was incubated with DNA substrate, and cleavage complexes were established in the presence of CaCl 2 (rather than the physiological divalent cation MgCl 2 (45)). Although the DNA cleavage site specificity of topoisomerase II appears to be the same in the presence of either divalent cation (23,33), the use of CaCl 2 provided two important advantages. First, levels of DNA cleavage generated in the presence of Ca 2ϩ are significantly higher than those generated in the presence of Mg 2ϩ (33). Second, topoisomerase II generates considerably more single-stranded DNA breaks in the presence of Ca 2ϩ than it does in Mg 2ϩ -containing reactions (18,33). The stimulation of single-stranded DNA scission by Ca 2ϩ is especially important, because it leaves one intact strand of DNA in the cleavage complex, which can then act as a template for the geometric amplification of the selected nucleic acid molecules.
Covalent topoisomerase II-DNA cleavage complexes formed in the presence of Ca 2ϩ were trapped by the addition of SDS and isolated by precipitation of the enzyme in the presence of KCl (35,36). DNA molecules that were not covalently attached to topoisomerase II (through at least one of the two strands) remained in solution. Following a series of washes, oligonucleotides that co-precipitated with the denatured enzyme were amplified to provide the substrate DNA for the next round of SELEX. As demonstrated below (see Fig. 2), the DNA pool was enriched for sites of DNA cleavage by topoisomerase II with each successive round.
Although the random portion of the initial oligonucleotide FIG. 1. General scheme for the SELEX protocol. Preferred topoisomerase II DNA cleavage sites were selected/evolved using the following procedure: 1) covalent topoisomerase II-DNA cleavage complexes were formed in the presence of Ca 2ϩ ; 2) cleavage complexes were trapped by the addition of SDS, precipitated with KCl, washed extensively, and redissolved; 3) primers were annealed to the oligonucleotides obtained from redissolved cleavage complexes; and 4) the enriched pool of DNA cleavage sequences was amplified under mutagenic conditions to provide substrate DNA for the next round of SELEX. The initial oligonucleotide cleavage substrate is shown at the top. It included a 20-bp core of random DNA flanked by constant sequences that were devoid of topoisomerase II cleavage sites (16,44) and contained sites for restriction endonucleases. Topoisomerase II (Topo II) was modeled after the crystal structure reported by Berger et al. (58).
substrate contained (in theory) all possible 20-mer sequences, amplification was performed under mutagenic conditions (34) to generate a slight "drift" in the sequences selected for each round. This was done to offset any bias in the initial pool or the loss of any important sequences early in the selection process, since mutagenic amplification has the potential to regenerate such lost sequences. As discussed under "Experimental Procedures," the DNA amplification protocol employed generated ϳ0.4 mutations/oligonucleotide/round of SELEX.
SELEX Generates Oligonucleotides That Contain Preferred Sequences for Topoisomerase II-mediated DNA Cleavage-The SELEX procedure described above was employed for 20 cycles, with progress being monitored by K ϩ /SDS precipitation of cleavage complexes containing radiolabeled DNA (Fig. 2). While only 3% of the initial random oligonucleotide pool (round 0) was cleaved by topoisomerase II, the emergence of preferred cleavage sequences was evident as early as round 4. By round 20, nearly 20% of the selected oligonucleotide pool was cleaved by the Drosophila type II enzyme. In light of the fact that the DNA cleavage/religation equilibrium of topoisomerase II normally lies far toward religation (9 -13), 20% represents an unusually high level of enzyme-mediated DNA scission. Four more rounds of SELEX were carried out, but the levels of DNA cleavage appeared to plateau by round 20 (data not shown).
Since the SELEX protocol employed Ca 2ϩ as the divalent cation for DNA cleavage reactions (see above), a control experiment was performed that monitored cleavage complex formation of the SELEX pools in the presence of Mg 2ϩ (Fig. 2, inset). As expected (18,33), levels of topoisomerase II-mediated DNA scission were lower in Mg 2ϩ than those generated in the presence of Ca 2ϩ . This difference notwithstanding, the emergence of preferred DNA sequences mirrored the trend observed in the Ca 2ϩ -containing reactions. This finding supports the previous observation that the DNA cleavage site specificity of topoisomerase II is not dictated by the nature of its divalent cation cofactor (18,23,33).
Topoisomerase II-mediated Cleavage of Oligonucleotides Generated by SELEX Is Stimulated by Anticancer Drugs-Beyond its critical physiological functions, topoisomerase II is the target for a number of anticancer drugs that are in wide clinical use (5, 7, 8). These agents act by increasing levels of topoisomerase II-mediated DNA cleavage (3,6,(12)(13)(14). Consensus sequences reported for drug-stimulated DNA scission generally differ from those reported for drug-free reactions and show specificity at fewer positions (4,8). Consequently, it has been questioned whether drug-stimulated scission takes place primarily at a subpopulation of "intrinsic" topoisomerase II DNA cleavage sites or rather is induced at a novel population of drug-specific sites.
In order to address this issue, the effects of three structurally diverse topoisomerase II-targeted drugs on enzyme-mediated DNA cleavage of the round 0 and 20 SELEX pools were determined (Fig. 3). These experiments were performed in the presence of Mg 2ϩ and 100 M etoposide, amsacrine, or CP-115,953 (or 1% Me 2 SO as a solvent control). While cleavage of the round 0 (random) substrate displayed little sensitivity (Ͻ2-fold stimulation) to the three drugs examined, cleavage of the round 20 oligonucleotide pool was stimulated between 4-and 6-fold. This result indicates that oligonucleotide substrates that are enriched for "intrinsic" topoisomerase II cleavage sites are also enriched for drug-inducible sites. It further supports the hypothesis that drug-induced DNA cleavage complexes are formed primarily at sites intrinsic to the enzyme rather than at a novel population of drug-specific sequences (12,46).

Oligonucleotide Cleavage Substrates Selected/Evolved by Drosophila Topoisomerase II Are Not Universally Preferred Substrates for Type II Enzymes from Other Species-Previous
studies indicate that type II topoisomerases, even from diverse eukaryotic organisms, will often cleave a given DNA substrate at a similar array of sites (20,22,23,47,48). However, since consensus DNA cleavage sequences differ considerably for enzymes from different species (9, 19 -22), it is obvious that generalizations from enzyme to enzyme may not be appropriate. To determine whether DNA cleavage substrates selected by Drosophila topoisomerase II are also preferred substrates for enzymes from other species, the ability of yeast (S. cerevisiae) topoisomerase II as well as the ␣ and ␤ isoforms of the human enzyme to cleave the round 0 and 20 SELEX pools was determined. As seen in Fig. 4, a significant enhancement of cleavage (ϳ3-fold) for the round 20 pool over the round 0 substrate was observed for yeast topoisomerase II. However,  neither of the human isoforms displayed any appreciable specificity for the round 20 SELEX pool. These data provide further evidence that the specificity of type II topoisomerases from different species is not necessarily conserved for any given preferred sequence.
Mechanism of DNA Cleavage Enhancement-Higher levels of topoisomerase II-DNA cleavage complex formation can result from increased binding between the enzyme and its substrate DNA, from increased cleavage within the noncovalent topoisomerase II-DNA complex, or from both (3,9,12). As a first step toward determining the mechanistic basis for the evolution of preferred cleavage sequences, the binding of topoisomerase II to the round 0 and 20 DNA pools was characterized. In this experiment, enzyme-DNA binding was monitored by a nitrocellulose filter protocol (38). A mutant yeast topoisomerase II in which the active-site tyrosine was replaced with a phenylalanine (ytop2Y783F) was utilized for this study (31). The use of this mutant enzyme allowed binding to be monitored in the absence of DNA cleavage, even when Mg 2ϩ was present in assay mixtures. As seen in Fig. 5, ytop2Y783F displayed a similar binding affinity for the round 0 and 20 SELEX pools.
Previous studies indicate that topoisomerase II will bind DNA in the absence of a divalent cation, albeit with a decreased affinity (45,49). Therefore, to extend the above results to the Drosophila enzyme under conditions that did not allow DNA cleavage within the noncovalent complex, the binding of Drosophila topoisomerase II to DNA was determined in the absence of a divalent cation. Once again, there was no significant difference in binding to the two SELEX pools (Fig. 5, inset). Thus, the increased cleavage complex formation for the round 20 SELEX pool is not due to an increased binding affinity of topoisomerase II for the DNA.
Levels of cleavage within a topoisomerase II-DNA complex are dependent on the relative rates of DNA scission and religation by the enzyme (9,10,12,13). Unfortunately, direct measurement of the rate of cleavage is technically unfeasible, since it probably does not represent the rate-determining step of cleavage complex formation. It is possible, however, to directly measure the apparent first order rate of DNA religation within cleavage complexes (33,50). As seen in Fig. 6, rates of DNA religation for the round 20 SELEX pool were ϳ3-fold faster than those observed for the initial round 0 substrate. Two conclusions may be inferred from the above findings. First, the SELEX protocol did not select/evolve DNA sequences on the basis of decreased religation rates. Second, it is likely that the average rate of topoisomerase II-mediated scission is considerably faster for the selected DNA sequences in the round 20 pool than for the initial random oligonucleotide substrate.
Sequence of Oligonucleotides in the Round 20 SELEX Pool-The round 20 SELEX pool was digested with BamHI and EcoRI restriction endonucleases and ligated into Bluescript SK phagemid that had been previously digested with these enzymes. This construct was used to transform E. coli, and the sequences of 37 of the resultant clones were determined (Fig. 7). All of the clones sequenced were found to have random regions that were 18 rather than 20 bp in length, indicating that a 2-bp deletion occurred at some point in the selection process. As determined by sequence analysis of oligonucleotide pools, the deletion emerged between SELEX rounds 4 and 8 (data not shown).
It appears that the SELEX protocol converged on a single preferred sequence for topoisomerase II-mediated DNA cleavage (Fig. 7). The vast majority (36/37) of the clones examined contained the same sequence (typified by clone 1), alternating T and A residues with a centrally located CA dinucleotide at positions 9 and 10. The remaining clone (clone 17) was identical to the others, except that it contained a G (rather than an A) at position 12.
Sites of Topoisomerase II-mediated DNA Cleavage within Preferred Sequences-The two sequences selected by Drosophila topoisomerase II were examined to determine the specific sites of DNA scission by the enzyme. Oligonucleotides utilized for these studies consisted of the selected/evolved 18-mer sequences bordered by the original constant flanking sequences. Fig. 8 shows a representative cleavage site map of the predominant sequence (clone 1) that was identified by SELEX. The enzyme cleaved this oligonucleotide at four principal sites on each strand, with seven of the eight total sites occurring 3Ј to a T nucleotide (Figs. 8 and 9, top). Additional minor sites of cleavage were also observed. Sites of cleavage on the two strands occurred immediately 3Ј to pyrimidine residues and aligned with the expected four-base stagger that is characteristic of topoisomerase II-mediated DNA scission (35,16). In the one clone that differed in sequence (clone 17), the sites of cleavage observed were the same as for the predominant clone (Fig. 9, top).
It should be noted that the multiple sites of topoisomerase II-mediated DNA scission within the selected/evolved sequences do not result from the formation of multiple cleavage complexes on individual oligonucleotide molecules. Since the enzyme protects at least 28 bp of DNA, as determined by footprint analysis (41,42), only a single topoisomerase II homodimer is capable of forming a cleavage complex within the preferred sequence of any of these oligonucleotide molecules at any given time.
Levels of cleavage at specific sites in the predominant sequence (clone 1) were generally 2-4-fold higher in the presence of Ca 2ϩ than in Mg 2ϩ , site 3 being the notable exception (Fig. 9,   bottom). Similar levels of scission were observed for clone 17, except that cleavage at site 3 (which encompasses the A:T 3 G:C substitution) was decreased by ϳ50% relative to site 3 in the predominant sequence (data not shown).
Drug Stimulation of Topoisomerase II-mediated DNA Cleavage at Specific Sites-The stimulation of DNA cleavage observed for SELEX products in the presence of drugs (see Fig. 3) suggests that at least some of the sites within the selected/ evolved sequences are drug-inducible. Fig. 10 shows the effects of Me 2 SO (solvent), etoposide, and amsacrine on the relative levels of cleavage at the four principal sites of complex formation observed in the absence of drugs. While Me 2 SO had no significant effect on the levels of complex formed, both etoposide and amsacrine stimulated topoisomerase II-mediated DNA cleavage within the clone 1 sequence.
In the presence of etoposide, site 3 exhibited the highest level of cleavage. This is consistent with the reported preference for a C residue immediately 5Ј to the site of etoposide-stimulated DNA cleavage (see Fig. 9, top) (51). Cleavage at site 2 in clone 17 increased 3-fold relative to that in clone 1 (data not shown), consistent with the T 3 C substitution 5Ј to the site of scission on the bottom strand. Site 3 in clone 17 was cleaved at ϳ50% of the level observed in the predominant clone (data not shown).
In the presence of amsacrine, site 2 exhibited the highest level of cleavage. Both halves of this site match the reported preference for an A residue immediately 3Ј to the site of amsacrine-stimulated DNA cleavage (51). Although site 4 also contains A residues 3Ј to both sites of cleavage, levels of amsacrineinduced scission were considerably lower than observed for site 2. The reason for this difference is not apparent, but it is notable that cleavage at site 4 was poor (relative to the other sites) under every condition tested.
Topoisomerase II-mediated Cleavage of the Preferred DNA Sequence Is Double-stranded-The SELEX protocol took advantage of the fact that topoisomerase II generates significant levels of single-stranded DNA breaks in the presence of Ca 2ϩ (33). This raises the question as to whether the enzyme cleaves the preferred sequence in a double-stranded fashion in the presence of its physiological divalent cation, Mg 2ϩ (45). To address this issue, cleavage complexes trapped with clone 1 in the presence of Mg 2ϩ were subjected to electrophoresis in a 14% nondenaturing polyacrylamide gel. As seen in Fig. 11, multiple DNA cleavage products (consistent with the multiple sites of cleavage within the oligonucleotide) were observed under nondenaturing conditions. Thus, topoisomerase II-mediated scission of the preferred sequence is (to at least some extent) double-stranded in nature. Fig. 11 also shows that cleavage complexes formed with clone 1 do not enter the gel if treatment with proteinase K is omitted (lane 3), demonstrating that the DNA in these complexes is protein-associated. Furthermore, the formation of cleavage complexes is reversed by the addition of EDTA prior to SDS (lane 4). Both of these characteristics are hallmarks of topoisomerase II-mediated DNA scission (9,10,12).
The Preferred Topoisomerase II DNA Cleavage Sequence Is Dramatically Overrepresented in the Euchromatic Genome of D. melanogaster-Statistically, an exact match for an 18-mer sequence should occur once in every 70 billion bp of random DNA (i.e. once in 4 18 bp). Consequently, it is doubtful that this sequence should appear by random chance even a single time in the 250-megabase pair genome of D. melanogaster. The odds of finding a match for such a sequence are reduced even further at the present time in light of the fact that Ͻ15 megabase pairs of the Drosophila euchromatic genome have been sequenced. To test this assumption, 10 individual 18-mer sequences were generated in a random fashion and used to search the European Drosophila Genome Project data base. No exact matches in the Drosophila genome were found for any of the 10 random sequences examined. Of the 10, eight did not even yield partial matches (using the default settings for the blastn search routine with the threshold set at 1000). The other two each yielded a single partial match at 15 of 18 or 17 of 18 positions (the latter with an insertion).
In marked contrast, a search of this same data base revealed Ͼ60 exact matches for the predominant Drosophila topoisomerase II DNA cleavage sequence in the euchromatic genome of this organism. Of these exact matches, 20 were found in known gene sequences, with 16 of them occurring near the 5Ј-or 3Ј-end of the genes (Fig. 12). Thus, it appears that the preferred topoisomerase II DNA cleavage sequence selected by SELEX is dramatically overrepresented in the Drosophila genome. Statistically (based on the total number of matches), it occurs approximately once every 250,000 bases. (A similar prevalence of this sequence was found in the 14-megabase pair S. cerevisiae genome, consistent with the finding that it also is a preferred substrate for yeast topoisomerase II.) Even ac-  11. Cleavage of the selected/evolved sequence is doublestranded in the presence of Mg 2؉ . Topoisomerase II-DNA cleavage complexes were established with clone 1 oligonucleotide in Mg 2ϩ -containing buffer, trapped by the addition of SDS, digested with proteinase K, and subjected to electrophoresis in a 14% nondenaturing polyacrylamide gel at 10°C. Samples shown are a control in the absence of enzyme (DNA), cleavage with topoisomerase II (Topo), cleavage with enzyme but omitting the proteinase K treatment (ϪPro K), and cleavage with enzyme that was reversed by the addition of EDTA prior to SDS (ϩEDTA). Locations of the origin, intact substrate DNA, and heterogenous double-stranded cleavage products are indicated.
counting for the increased A-T content (ϳ57%) of the Drosophila euchromatic genome, this sequence should appear less than once every 8 billion base pairs. Therefore, the sequence TATATATACATATATATA is present at a level that is at least 10,000 times higher than predicted by random chance. In addition, an exact match for this sequence was found within the mitochondrial genome of Drosophila, which is Ͻ20 kilobase pairs in length. Taken together, these findings suggest that the DNA cleavage sequence selected by topoisomerase II is of physiological importance.
It is notable that clone 17 (TATATATACATGTATATA), which differed from the predominant cleavage sequence by only an A 3 G substitution, yielded one exact match in the Drosophila euchromatic genome, at the 3Ј-end of the gene encoding the kinase suppressor of ras. The frequency with which this cleavage sequence occurs in the genome relative to that of clone 1 (1 out of 60) is comparable with the frequency with which it was found in the round 20 SELEX pool (1 out of 37). No exact matches for this sequence were found in the yeast genome data base.

DISCUSSION
Topoisomerase II is an essential enzyme that is involved in virtually every aspect of DNA metabolism (1,2). Fundamental to all aspects of its catalytic function, topoisomerase II must create transient double-stranded breaks in the backbone of the genetic material (9 -12). Although the enzyme displays a reproducible pattern of cleavage on any given DNA substrate, the factors that underlie its nucleotide specificity remain an enigma. In an attempt to define the DNA site specificity of topoisomerase II, previous studies have determined consensus sequences for enzyme action based on the nucleotide analysis of multiple (ranging from 16 to 93) cleavage sites (19 -22). In general, the consensus sequences reported from this approach have been weak and bear little relationship to one another. Consequently, they have not proven to be as useful a tool for elucidating the DNA site specificity of topoisomerase II as originally hoped.
It is not entirely surprising that consensus sequences reported for enzyme action have been weak in nature. First, since topoisomerase II displays at least some level of specificity for an average of eight nucleotide positions (based on consensus sequences), cleavage over a large number of base pairs (approaching 4 8 ) would have to be analyzed in order to define a consensus with a high degree of confidence. Clearly, this is a difficult criterion to meet and has not been approached in previous studies. Second, topoisomerase II probably carries out general functions (such as the control of superhelical density or DNA untangling) in a global, rather than a highly specific, manner (1,2). Therefore, to fulfill this aspect of its physiological role, the enzyme must be able to act at a wide variety of DNA sequences.
Beyond its global functions, however, topoisomerase II plays specific roles in chromosome organization, condensation/decondensation, and segregation (1,2,9,10,12,52). To fulfill these latter responsibilities, the enzyme appears to act at specific regions in the genetic material (matrix/scaffold attachment regions (i.e. MAR and SAR sequences) and centromeric sequences, for example) (27,48,53,54). Thus, against a background of low stringency sites, topoisomerase II may have highly preferred sites of action within the genome. The present study used a SELEX protocol to identify a candidate for such a site from Drosophila.
The cleavage sequence that was selected/evolved (TATATA-TACATATATATA) is rich in A:T base pairs and is made up of alternating purine/pyrimidine residues. Earlier studies that derived a consensus DNA cleavage sequence for Drosophila topoisomerase II (19) or characterized interactions between topoisomerase II and MAR/SAR sequences (25,48,53) suggested that A/T richness contributes to topoisomerase II-mediated DNA scission. Furthermore, a study that overlaid several consensus sequences for the enzyme demonstrated that topoisomerase II cleaves runs of alternating purines and pyrimidines (55). Thus, the preferred sequence identified in the present work supports these previous observations. Several aspects of the sequence identified by the SELEX protocol merit special note. First, the DNA site that was selected/evolved for cleavage by Drosophila topoisomerase II in the absence of DNA cleavage-enhancing drugs was also a preferred sequence for the action of anticancer agents targeted to the enzyme. This agrees with the previous hypothesis (12,46) that drugs stimulate DNA cleavage at sites for which topoisomerase II has some level of intrinsic specificity.
Second, the selected sequence contains multiple sites for topoisomerase II-mediated DNA scission. Clustering of strong DNA cleavage sites for the enzyme has been observed previously (46). The relationship between site proximity and strength is not known. Due to the small distance between the sites in this sequence, it is clear that only a single molecule of the enzyme can be accommodated on this sequence at a given time. Therefore, increased DNA scission cannot be due to multiple simultaneous cleavage events. More likely, the presence of multiple cleavage sites in close proximity may impede linear diffusion of topoisomerase II to distal sites on the DNA, thereby increasing the local concentration of the enzyme.
Third, this DNA sequence was a poor cleavage substrate for human type II enzymes. This latter result implies that the assumed "universal" nature of topoisomerase II site specificity (20,22,23,47,48) may not hold true for highly preferred sites of scission.
Finally, the selected/evolved topoisomerase II DNA cleavage sequence is dramatically overrepresented in the Drosophila genome and is often found at the 5Ј or 3Ј extremes of expressed genes. A previous study that mapped (at low resolution) topoisomerase II cleavage sites within 830 kilobase pairs of cloned Drosophila DNA found a correlation between fragments that contained SAR sequences and those rich in enzyme-mediated scission (27). Since MAR/SAR sequences generally display a high A/T content (56), and the frequency of the preferred topoisomerase II cleavage sequence in the Drosophila genome (every ϳ250 kilobase pairs) is on the order of the size of chromosomal loop domains (57), it is tempting to speculate that the sequence obtained in the present study represents an attachment region. However, further in vivo experimentation will be required to determine the functional significance of the sequence and its potential physiological interactions with topoisomerase II.
In summary, the sites of action of topoisomerase II on DNA profoundly influence its catalytic activity. To further our understanding of how the enzyme recognizes its DNA substrate, a SELEX protocol was used to identify a highly preferred sequence for DNA cleavage mediated by Drosophila topoisomerase II. Results of the present study afford a unique perspective toward defining the intrinsic DNA cleavage specificity of the type II enzyme and may ultimately reveal relationships that link the site specificity of topoisomerase II to its physiological functions.