Site-specific Interaction of Vaccinia Virus Topoisomerase I with Base and Sugar Moieties in Duplex DNA*

Vaccinia DNA topoisomerase specifically binds and forms a covalent adduct at DNA sites containing a conserved sequence element 6’(C/T)CCTTJ in the scissile strand. The molecular interactions that contribute to recognition of the CCCTT motif in a synthetic DNA substrate have been examined using modification interference, modification protection, and analog substi- tution techniques. We report that topoisomerase makes contact with guanine nucleotide bases of the pentamer motif complementary strand (3’GGGAA) within the major groove of the DNA helix and that alteration of the binding surface by chemical modification is deleterious to the interaction. Additional contacts are made with guanine residues located outside the pentamer element. The enzyme is unable to form a covalent adduct with synthetic RNA substrates. Analysis of the cleavage of DNA duplexes containing 2’OMe sugars suggests that the inability of the vaccinia topoisomerase to cleave either an RNA duplex or an RNA:DNA hybrid can be accounted for by the interfering effects of a 2’ sugar substituent at two or more sites within the pentamer. Interaction with the sugar at the +2T nucleotide appears to be the most critical, as judged by the effects of single sugar substitutions.

A requirement for covalent complex formation between vaccinia topoisomerase and DNA is that the CCCTT sequence be in duplex form (4). Single nucleotide substitutions within the CCCTT motif profoundly affect cleavage efficiency (5). A hierarchy of mutational effects is observed depending on the nature of the base alteration and the position within the CCCTT sequence. The +2 T:A base pair appears to be the most critical position, insofar as any base change (either on the scissile or the nonscissile strand) virtually abrogates the cleavage reaction. At other positions, pyrimidine-to-pyrimidine or purine-to-purine substitutions can be tolerated to varying degrees, whereas pyrimidine-to-purine alterations (and vice versa) are generally inimical to the cleavage reaction The vaccinia topoisomerase, upon binding to a duplex DNA containing a CCCTT motif, specifically protects the region around the site of covalent adduct formation from DNase I digestion (5). The DNase footprint spans both sides of the cleavage site, from +13 to -13 on the scissile strand (+1 being the site of cleavage) and from +13 to -9 on the noncleaved strand. The size of the footprint is greater than the 11 bp' of duplex DNA that constitute the "minimal" substrate for covalent adduct formation (5). Also, the margins of the nuclease footprint extend beyond the minimal essential positions for strand cleavage (from +6 to -2) defined by DNA deletion and site phasing experiments (5).
Although the DNA mutational analyses and nuclease footprinting experiments confirm the sequence specificity of the topoisomerase-DNA interaction, they do not illuminate the pertinent principles of site recognition. This issue is addressed in the present study using a combination of modification interference, modification protection, and analog substitution techniques. The results demonstrate that vaccinia topoisomerase binds to the major groove of the DNA helix, making critical contacts with its recognition element at G residues on the nonscissile strand. Additional contacts are made with purine residues outside the pentamer element. Pentose analog substitution effects suggest that the enzyme also recognizes the sugar moiety of the polynucleotide chain and provide an explanation for the strict specificity of the vaccinia enzyme in its cleavage of DNA, but not RNA chains.

EXPERIMENTAL PROCEDURES
Enzyme Purijicatwn-Vaccinia DNA topoisomerase was expressed in Escherichia coli and purified as described (6). The heparin-agarose enzyme fraction used in the present study was the same preparation described previously (6).
Oligonucleotide Substrates-Synthesis of DNA and RNA oligonucleotides via Dimethoxytrityl-cyanoethyl phosphoramidite chemistry was performed by the Sloan-Kettering Microchemistry Laboratory using an Applied Biosystems model 380B or model 394 automated DNA synthesizer according to protocols specified by the manufac-The abbreviation used is: bp, base pair(s). turer. Standard deoxynucleoside and ribonucleoside phosphoramidites, as well as deoxyinosine phosphoramidite, were purchased from Applied Biosystems Inc. Other modified phosphoramidites (deoxyuracil and 2'0Me ribonucleosides) were purchased from Glen Research. Synthetic DNA oligonucleotides were labeled at the 5' end via enzymatic phosphorylation in the presence of [-y3'P]ATP and T4 polynucleotide kinase. The labeled oligonucleotide was gel purified and hybridized to complementary strand (present at 2-3-fold molar excess) as described (4, 5). RNA end labeling and hybridization procedures were identical to those used for DNA oligomers.
Methylation Protection Assay-Purified topoisomerase was incubated with a 60-mer duplex DNA substrate (5' end labeled uniquely on the scissile or nonscissile DNA strand) in 50 mM Tris-HC1 (pH 7.5) for 5 min at 37 "C. The mixtures were made 0.04% in dimethyl sulfate and incubated for 5 min at 30 "C, after which glycerol was added to 5%. To separate topoisomerase-DNA complexes from unbound 3zP-labeled DNA, the reaction mixtures were electrophoresed through a nondenaturing 6% polyacrylamide gel containing 0.25 X TBE (22.5 mM Tris borate, 0.6 mM EDTA). The gel was autoradiographed wet. A discrete topois~merase-[~~P]DNA complex was resolved from free labeled DNA; the bound and free DNA fractions were excised separately from the gel and eluted by soaking the gel slice overnight at 4 "C in TE (10 mM Tris-HC1 (pH 8.0), 1 mM EDTA). The labeled DNA was recovered by ethanol precipitation and resuspended in water. Samples were mixed with an equal volume of 2 hi piperidine and heated at 90 "C for 30 min. DNA was precipitated serially with butanol and ethanol and then denatured in formamide and electrophoresed through a 12% polyacrylamide sequencing gel containing 7 M urea in TBE (90 mM Tris borate, 2.5 mM EDTA).
Methylation Interference Assay-The 60-mer DNA substrate (5' end labeled uniquely on the scissile or nonscissile DNA strand) was premethylated by treatment with 0.05% dimethyl sulfate for 5 min at 30 "C. Modification was halted by the addition of p-mercaptoethanol, and the DNA was recovered by ethanol precipitation. Purified topoisomerase was incubated with the methylated 60-mer DNA substrate in 50 mM Tris-HC1 (pH 7.5) for 5 min at 37 "C. Protein-bound and free DNA fractions were recovered after native gel electrophoresis, cleaved with piperidine, and analyzed by denaturing gel electrophoresis as described above for the methylation protection assay.
Assay of Covalent Complex Formation--Reaction mixtures (20 pl) containing 50 mM Tris-HC1 (pH 7.51, 5' "P-labeled DNA, and vaccinia topoisomerase were incubated at 37 "C for 5 min. Covalent complexes were trapped by the addition of 1% SDS to denature the bound topoisomerase. Samples were heated for 2 min at 95 "C and then electrophoresed through a 10% polyacrylamide gel containing 0.1% SDS. Free DNA migrated with the bromphenol blue dye front. Covalent complex formation was revealed by transfer of radiolabeled DNA to the topoisomerase polypeptide as detected by autoradiographic exposure of the dried gel. The extent of adduct formation was quantitated by scintillation counting of an excised gel slice containing the labeled protein and was expressed as the percent of the input 5' '*P-labeled oligonucleotide that was covalently transferred to protein.

Methylation
Interference-The interaction of vaccinia topoisomerase with duplex DNA is presumed to entail specific contacts between amino acid residues in the topoisomerase polypeptide and the paired bases of the conserved pentamer binding motif. The contributions of individual nucleotide bases to binding/cleavage specificity can be inferred from the effects of base modifications on protein binding. A 60-bp duplex DNA containing a single CCCTT motif was employed for this analysis. The structure of the 60-mer is shown below (Structure 1). ylated on purine residues by treatment with dimethyl sulfate. Methylated DNA was mixed with purified topoisomerase under standard binding reaction conditions described previously (4, 5, 7); the reactions were constituted such that about half of the input DNA was bound by the enzyme. Topoisomerase-DNA complexes were then separated from unbound 32Plabeled DNA by native gel electrophoresis (7). Bound and unbound DNA species were recovered from the gel, and the DNA was cleaved at methylated G residues via treatment with piperidine. Cleavage products were then analyzed by denaturing gel electrophoresis (Fig. 1). N-7 methylation of any of the guanine bases within the 3'GGGAA pentamer on the nonscissile strand strongly interfered with protein-DNA complex formation; this was evinced by the nearly complete exclusion, from the bound DNA fraction, of DNAs with piperidine-cleavable sites corresponding to G positions +3, +4, and +5 (Fig. 1, left panel, lane B ) .
DNAs methylated a t these three sites were represented prominently in the population of unbound DNAs (lane F ) . Thus, integrity of each of the Gs within the pentamer motif was a key determinant of DNA binding. In contrast, methylation at the nearby +7G residue (situated outside the conserved motif) appeared not to interfere with complex formation (this was more obvious on longer autoradiographic exposure of the gel in Fig. 1). A partial effect of methylation was noted at the -4G position of the nonscissile strand, as this cleavage site was underrepresented in the bound DNA population (Fig. 1, left panel, and other data not shown).
Methylation of two G residues of the scissile strand (+6G and +9G, located 5' of the pentamer element) interfered partially with protein-DNA complex formation (Fig. 1, right panel, and other experiments not shown). Note that a cleavage product of the radiolabeled scissile strand was detected in the protein-bound DNA fraction, but not in the free DNA population, which did not correspond to a G position in the scissile strand sequence (Fig. 1, right panel, denoted by the filled circle). Rather, this product was generated by strand scission at the topoisomerase +1 cleavage site. We suspect that the phosphotyrosyl DNA-protein intermediate was partially susceptible to hydrolysis during treatment with piperidine at 90 "C.
Methylation Protection-Intimate contacts between vaccinia topoisomerase and guanine bases within the 60-bp DNA were identified on the basis of protection from base-specific chemical modification. Preformed protein-DNA complexes were reacted in solution with dimethyl sulfate, and the bound DNAs were resolved preparatively from unbound species by native gel electrophoresis. Control experiments established that dimethyl sulfate treatment did not affect the stability of the protein-DNA complex (data not shown). Comparison of piperidine cleavage products from the bound and free DNA populations indicated that the all three G residues (+3, +4, and +5) within the pentamer element on the nonscissile strand were protected strongly from chemical modification when bound to the enzyme (Fig. 2, kit panel). Partial protection was afforded at -4 and -5 G of the noncleaved strand. The DNA was 5' radiolabeled on either the scissile or non-the gel shown in Fig. 2, left panel; other experiments (not scissile strand. Control experiments established that topoi-shown) indicated more clearly that only weak protection was somerase-mediated cleavage was confined to the CCCTT-conferred at the -5 position.) On the scissile strand, the +9G containing strand (data not shown). The 60-mer was meth-and +6G residues were protected by bound protein (Fig. 2,  The G bases shown to be protected from chemical modification when bound to the topoisomerase were the very same residues whose prior methylation interfered with DNA binding. The strongest and most critical interactions were with the three G bases of the pentamer element of the nonscissile strand. The N-7 position of guanosine is situated within the major groove of the DNA double helix; thus, the vaccinia topoisomerase recognized and bound to its pentamer element (in part, at least) through base-specific contacts in the major groove. Additional, albeit weaker, interactions were evident at G residues outside the pentamer, extending from +9 on the scissile strand to -5 on the nonscissile strand. The dimensions of the protein-DNA interface defined by G modification were in keeping with the DNase I protection experiments, which had shown protection by bound topoisomerase of the DNA Non-scissile Strrnd Scissile Strand on both sides of the site of strand scission. Inosine Substitution Effects-The role of individual purine bases of the pentamer element in DNA cleavage was examined by substitution of inosine at positions +5 to +1 on the nonscissile DNA strand. Synthetic nonscissile strand oligonucleotides (12-mer) containing a single inosine substitution were annealed to a 5' radiolabeled 12-mer scissile strand (5'TCGCCCTTATTC). Covalent complex formation was assayed by transfer of 5' end-labeled scissile strand from the duplex substrate to the enzyme to yield an SDS-resistant adduct detectable by SDS-polyacrylamide gel electrophoresis. The wild-type 12-mer substrate readily formed the covalent adduct (Table I). Complex formation was unaffected by I substitution for G at either position +5 or +4. Thus, the loss of the C-2 exocyclic amino group of the purine residue at +5 or +4 (situated within the minor groove) had no significant impact on cleavage efficiency. A 3-fold reduction in cleavage

AGCGGGAZTAAG
was noted when the +3 C:G base pair was changed to C:I (Table I); it was shown earlier that heteroduplex substitution of A for G at +3 on the nonscissile strand had a similar mild effect (2-fold) on scission of the 12-mer (5). It appears then that the purine 2-amino group at +3 of the consensus motif (although not essential per se) may contribute to the topoisomerase-DNA interaction. Inosine substitution for A at +2 abrogated DNA cleavage; this was consistent with the stringent requirement for a T A base pair at this position (5). A sharp reduction in cleavage was also observed when inosine was substituted for A at position +1 (Table I). These substitution effects could not be attributed to a failure of the radiolabeled scissile strand to hybridize to its I-substituted complement. Analysis of the annealed strands by native polyacrylamide gel electrophoresis revealed that the mobility of the hybridized scissile strand had been altered (i.e. retarded) relative to that of DNA that had not been subjected to annealing (data not shown). As in earlier studies (5), we interpreted this mobility shift as indicative of effective hybridization.
The deleterious effects of inosine substitution for A at +2 and +1 might have been caused by the single base mismatch per se. Thus, a second series of substrates was prepared in which the I-substituted noncleaved strand was annealed with a labeled scissile oligonucleotide containing a compensatory T 4 C mutation that would to restore base pairing. Heteroduplex mutant substrates containing only the T + C mutations in the scissile strand were included as controls. As shown in Table 11, the suppressive effects of the +2 C:A mispair on covalent adduct formation could not be overcome by replacement with the paired C:I moiety. Similarly, the cleavage of the +1 C:A heteroduplex substrate (which was reduced 20fold relative to the standard substrate) was enhanced only slightly (2.5-fold) when the compensatory +1 C:I pair was introduced. It has been pointed out that changing T:A to C:I base pairs alters the surface of the major groove while preserving the potential binding surface of the minor groove (8). This has prompted a strategy for assessment of the relative contributions of major versus minor groove base substituents to DNA protein interactions (8), i.e. if major groove contacts Topoisomerase I  should be detrimental to the interaction, whereas if the critical contacts are exclusively in the minor groove, then the I:C substitution should be well tolerated. That I:C substitution so strongly depresses the cleavage reaction (by 90-fold at +2 and by 10-fold at +1) suggests that the relevant contacts at positions +2 and +1 are in the major groove.
Uracil Substitution Effects-Individual thymines within the CCCTT motif of the scissile strand were replaced with deoxyuracil. U substitution for T at +2 or +1 had no influence on strand scission (Table 111). Thus, the 5-methyl group of thymine was not essential for interaction of the enzyme with its recognition sequence.
Vaccinia Topoisomerase Does Not Cleave RNA-The presumption that topoisomerases act exclusively on DNA substrates has been called into question by the recent findings of DiGate and Marians (9) that at least one DNA topoisomerase, E. coli topoisomerase I11 (a type I enzyme), is capable of cleaving RNA strands. RNA scission by topoisomerase 111 entailed covalent adduct formation between the enzyme and the 5' phosphate group of the incised bond. Furthermore, the nucleotide sequences of topoisomerase I11 RNA and DNA cleavage sites were found to be identical (9). A key question is whether other DNA topoisomerases have similar abilities to break and rejoin RNA strands. The sequence specificity of the vaccinia topoisomerase allows us to approach this issue in a straightforward fashion using synthetic RNA oligonucleotides identical in base sequence to the scissile and nonscissile strands of standard DNA cleavage substrates.
As shown in Fig. 3, there was no detectable label transfer from 5' end-labeled RNA to the topoisomerase when the enzyme was incubated with an 18-bp RNA hybrid labeled either on the CCCUU-containing strand or on the complementary strand containing 3'GGGAA. Nor did covalent adduct formation occur when topoisomerase was incubated with either of the two 18-mer oligonucleotides in single-stranded form. An RNADNA hybrid composed of radiolabeled CCCTT-containing RNA strand annealed to a complementary DNA nonscissile strand was inert in covalent complex formation. The same nonscissile strand when annealed to a labeled scissile DNA strand constituted a duplex that was readily cleaved. Although a trace level of label transfer was detected when topoisomerase was incubated with radioactive scissile DNA strand hybridized to a complementary RNA oligonucleotide, we observed a similar trace level of cleavage activity with scissile DNA strand in single-stranded form. Cleavage of the CCCTT-containing RNA strand (either as RNARNA hybrid or RNADNA hybrid) could not be induced by inclusion of magnesium over a concentration range of 0.5-10 mM (Fig. 4). Similarly, cleavage of the (CCCTT)DNA (GGGAA)RNA substrate was not enhanced by inclusion of either magnesium or manganese over the same concentration range (data not shown). These results indicate that vaccinia topoisomerase is unable to catalyze site-specific cleavage of RNA strands.
Effects of 2' Sugar Substitutions-Failure to cleave RNA, or to even to cleave a DNARNA hybrid, may reflect sensitivity to the global conformation of the helix (e.g. A-form uersus B-form). Alternatively, the presence of a 2'OH group on Enzyme -+ + + -+ + + + + Top Strand @ @ @@ --R @@a individual sugars may interfere with specific protein-nucleic acid backbone interactions essential for covalent adduct formation. The latter possibility was tested using synthetic substrates containing 2'OMe substituents at selected positions within the pentamer element. Initially, we examined the cleavage of substrates in which consecutive sugar residues were substituted in groups of two or three (e.g. CmCmCm or UmUm) on either the scissile or nonscissile strand. These sugar-substituted species were hybridized to unmodified complementary strands or to other sugar-substituted derivatives (Fig. 5 ) . With respect to the nonscissile strand, 0-methylation of the three consecutive G nucleotides virtually abrogated strand scission (lune 4 ) , whereas methylation of the neighboring two A nucleotides had no effect whatsoever (lune 6 ) .
On the scissile strand, 2' 0-methylation of the three C nucleotides had only a mild effect (lane 8 ) , whereas sugar substitution of the two U nucleotides was sufficient to abolish the reaction completely (lane 14). Not surprisingly, the methylated strands that were ineffective when annealed to wild-type oligonucleotides remained so when hybridized with sugar- substituted strands (lanes IO, 16, and 18). It was noteworthy that the effect of combining of the CmCmCm scissile strand with the mAmA nonscissile strand was to eliminate cleavage (lane 12), even though neither substitution alone had such an effect. These data suggested that sugar-specific interactions within the pentamer motif play a significant role in site recognition and that the relative importance of individual sugars might vary with the context of the sugar-phosphate backbone on the opposite DNA strand. This theme is underscored by the experiments presented in Tables IV and V, in which 18-mer oligonucleotides containing single 2'OMe sugars were tested for their ability to support covalent complex formation. The presence of a methylated sugar at any one of the five nucleotides of the 3'GGGAA element of the nonscissile strand had no appreciable affect on cleavage efficiency (Table IV, experiment 1). Within the CCCTT sequence of the scissile strand, single site methylations were also benign, with one exception. The 2'OMe group at position +2U was sufficient to reduce cleavage 9-fold (Table  IV, experiment 2). That such an effect was limited to the +2 residue reinforces the findings of the base substitution analyses that the integrity of the +2 position is the most critical single determinant of topoisomerase specificity (5). Because the presence of an OMe group at the +lU nucleotide (ie. vicinal to the scissile bond) had little effect on strand scission, we can infer that the failure to cleave an RNA chain was not because of an inherent inability of the topoisomerase to transesterify across a ribonucleotide linkage. Certain sugar substitutions of one strand that by themselves had no effect on covalent adduct formation were found to be strongly inhibitory when combined with a singly substituted complementary strand. A hierarchy of synergistic effects was apparent ( Table V). The relative importance of individual sugars within the pentamer motif could be inferred from the capacity of a given 0-methyl substitution to elicit interference with cleavage in a "cross-strand" combination. By this criterion, the sugars at nucleotides +3G and +4G on the nonscissile strand (and, to a limited degree, +1A) were relevant to the topoisomerase-DNA interaction, whereas sugars a t +5G and +2A appeared unimportant. On the scissile strand, the hierarchy of sugar substitution effects was +2U >> +lA > +4C > +3C. Modification of the +5C sugar of the scissile strand had little impact on the cleavage reaction in any of the cross-strand combinations tested. The effects of 2'OMe ribose substitution at positions outside the conserved pentamer motif have not been evaluated.

DISCUSSION
Vaccinia DNA topoisomerase binds to and incises duplex DNA at specific sites. Although prior studies had shown that site selection is dictated by a pentapyrimidine sequence in the scissile strand (5'CCCTT, or close congeners thereof), the 58.2 GCACAGCGGGAATAAGGG molecular interactions contributing to recognition of this motif had not been explored. The experiments presented above reveal that topoisomerase makes direct contact with purine nucleotide bases of the pentamer motif complementary strand within the major groove of the DNA helix and that alteration of the binding surface by chemical modification is deleterious to the interaction. Additional contacts are made with guanine residues located upstream of the pentamer motif and with residues 3' (downstream) of the site of strand scission.
The observed base-specific contacts outside the pentamer element occur at positions beyond the minimal region sufficient for site-specific cleavage of synthetic substrates (+6 to -2). No sequence conservation or strong nucleotide bias at positions outside the pentamer emerged from the mapping of topoisomerase cleavage sites within linear plasmid DNA (2). Yet, these sites could be classified as having higher or lower affinity for topoisomerase based on several criteria (2). (Higher affinity sites were cleaved at low enzyme concentration, were less sensitive to competition, and were most refractory to religation promoted by salt, divalent cations, and elevated temperature. Cleavage at lower affinity sites required higher enzyme concentration and was more sensitive to competition and induced religation.) This suggested that site affinity is influenced by the DNA sequence flanking the pentamer element. Studies of the effects of base changes within the pentamer element indicate that some alterations that abolish cleavage of a minimal substrate are actually tolerated by the vaccinia enzyme when they occur in the context of either a larger DNA or particular flanking sequences (5). Although flanking sequence effects on the vaccinia topoisomerase cleavage reaction have not been explored systematically, our findings that contact is made with nucleotide bases in the immediate flanking regions provide a rationale for the earlier observations and a framework for further studies. It is worth noting that the sequences of the synthetic DNA substrates used in the present study (and in earlier reports from this laboratory) are based closely on the sequence surrounding one of the high affinity sites mapped within plasmid DNA. Comparison of the protection and interference profiles of DNA substrates containing different flanking sequences may shed light on the issue.
Synthetic 18-bp RNA substrates containing a CCCUU motif are not cleaved by vaccinia topoisomerase under conditions permissive for the cleavage of DNA substrates of identical sequence. The observation that E. coli DNA topoisomerase I11 (a type I enzyme) can incise RNA strands (9) is therefore not applicable to this virus-encoded member of the eukaryotic family of type I enzymes. Control experiments using DNA substrates make clear that this is not caused by substitution of uracil for thymine within the pentamer element. Rather, our analysis of the cleavage of synthetic DNA duplexes containing 2'OMe sugars suggests that the inability of the vac-cinia topoisomerase to cleave either an RNA duplex or an RNADNA hybrid can be accounted for by the interfering effects of a 2' sugar substituent at two or more sites within the pentamer motif. A key implication is that DNA topoisomerase makes specific contacts with the sugar moiety that are relevant to site recognition and/or catalytic activity. Interaction with the sugar at the +2T nucleotide appears to be the most critical, as judged by the effects of single sugar substitutions. Although we have not undertaken to assess the effects of 2'OMe sugar substitutions outside the pentamer motif, preliminary footprinting experiments with the chemical nuclease copper/phenanthroline show that the vaccinia topoisomerase protects the sugar-phosphate backbone of the scissile strand over a region extending from position +10 to -10.' ' J. Turner and S. Shuman, unpublished data.
We anticipate that the application of additional chemical footprinting methods (i.e. with different base and backbone specificities) will help define at higher resolution the spectrum of protein-nucleic acid interactions that contribute to binding and cleavage. Efforts to crystallize the covalent enzyme-DNA intermediate are being made concurrently.