Proc. NatL Acad. Sci. USA Vol. 78, No. 2, pp. 843-847, February 1981 Biochemistry

Catenation and knotting of duplex DNA by type 1 : A mechanistic parallel with type 2 topoisomerases (nickdng-closing /w protein/linking number/DNA condensation/circular DNA) PATRICK 0. BROWN* AND NICHOLAs R. COZZARELLI*t *Departnment of Biochemistry and tDepartment of Biophysics and Theoretical Biology, The University of Chicago, Chicago, Illinois 60637 Communicated by Bernard Roizman, November 13, 1980

ABSTRACT a protein, a type 1 topoisomer- DNA molecules is nicked. The requirement for a nick was dem- ase, can catenate and knot duplex DNA circles. Previously, these onstrated by Y.-C. Tse and J. C. Wang, who independently activities were thought to be limited to type 2 topoisomerases. Catenation by a, requires a nick in one of the participating mol- discovered catenation by w (personal communication). Exten- ecules, but it is not necessary that both be nicked. Catenation does sive sequence homology between reacting molecules is not re- not depend on sequence homology and is greatly stimulated by quired. We therefore propose an alternative model for type 1 DNA-condensing agents such as spermidine. A eukaryotic type 1 reactions, analogous to our sign inversion model topoisomerase can also interlock duplex DNA circles. These ac- for gyrase (6), in that its characteristic feature is the deliberate tivities cannot easily be explained by the widely held topoisomer- passage of one segment of DNA through an enzyme-bridged ase model in which a reversible nick in DNA allows free rotation about the unbroken strand. We suggest instead passage of a DNA break in another. Our model differs from the swivel mechanism segment through a transient enzyme-bridged break in a single in that the enzyme holds fast to both sides of the nick to prevent DNA strand. This is analogous to the sign inversion mechanism free axial rotation, the change in Lk is intrinsically limited to of the type 2 topoisomerases, and thus expresses an essential one in each cycle of DNA breakage and reunion, enzyme-cat- mechanistic unity among topoisomerases. As expected for relax- alyzed movement of the DNA is not limited to axial rotation, ation by a single-strand passage, a, changes the linking number and a base-paired region is not required. of DNA in steps of 1. MATERIALS AND METHODS that alter topological invariants of DNA molecules are called topoisomerases. Until recently, the detection of topo- Enzymes. w purified from E. coli by chromatography on isomerases has depended on their ability to relax negatively DEAE-Sephacel, Sephacryl S-200, and phosphocellulose was supercoiled DNA or, in the case of DNA gyrases, to introduce used for most of the experiments. Results were checked by negative supercoils (1, 2). These reactions change a topological using additional purified preparations, kindly provided by R. invariant called the linking number (Lk), the net number of E. Depew (Northeastern Ohio Universities College of Medi- times one strand of a closed circular DNA molecule is wound cine) and by J. C. Wang (Harvard University). Purified DNA- around the other. It is now clear that other topological invariants untwisting enzyme from rat liver nuclei was a gift from J. Chain- can be altered by topoisomerases. Phage T4 topoisomerase (3) poux (University of Washington). All four topoisomerase prep- and bacterial DNA gyrases (4, 5) can reversibly catenate and arations were >80% pure. knot DNA rings, and catenating enzymes have been isolated DNA. Native monomers of ColE1 and pAO3 and from several eukaryotic sources (2). These topological isomer- phage 4X174 (OX) replicative form (RF)I DNA were prepared izations are all readily explained by a mechanism we have called as described (4, 11). Nicked DNA was prepared by treatment sign inversion (6), in which a duplex DNA segment is passed with DNase I (Worthington) until 40% of the starting material through a transient break across both strands of another such remained intact, and then purified by sedimentation (4). segment (6, 7). The sign inversion mechanism always changes Reactions. Catenation reactions were at 37C and the mix- Lk in steps of 2, and topoisomerases with this property are des- tures contained, in addition to enzyme and DNA, 20 mM ignated type 2 topoisomerases (3). Tris'HCl (pH 7.6), 20 mM KCl, 6 mM MgCl2, 5 mM spermi- Type 1 topoisomerases instead introduce transient single- dine'HCl, 2 mM dithiothreitol, 30% (vol/vol) glycerol, and strand breaks and are shown in this report to change Lk in steps bovine serum albumin at 50 ug/ml. Reaction products were of 1. According to the swivel model for these topoisomerases, treated with proteinase K at 0.1 gg/ml for 30 min at 370C. the enzyme would nick one DNA strand, allow the two ends Crosslinking. DNA was crosslinked by irradiation with long- to rotate around the unbroken strand, then reseal the nick. wavelength ultraviolet light (UV Products model B-1OA) in a Supercoil relaxation and. even knotting (8) and renaturation of 10-,ul volume containing 0.4 ug of DNA, 25 ng of 4,5',8-tri- single-stranded circles (9, 10) by type 1 enzymes have been methylpsoralen (trioxsalen, Calbiochem), 10 mM Tris-HCl (pH thought to occur by this swivel mechanism. It is hard to conceive 7.8), and 0.5 mM EDTA. To monitor crosslinking, products of of a plausible scheme whereby a swivel could catenate duplex a parallel reaction using Cfo I restriction endonuclease frag- DNA circles. But we report here our surprising discovery that ments of OX RFI DNA were denatured and then renatured in the two archetypal type 1 topoisomerases-Escherichia coli w the presence of a 20-fold molar excess of 4X174 viral single- protein and rat liver nicking-closing enzyme-efficiently cate- stranded DNA. Uncrosslinked fragments failed to self-renature nate duplex rings, provided that at least one of the reacting and were resolved- from crosslinked fragments by polyacryla- mide gel electrophoresis and quantitated by scanning the ethid- The publication costs ofthis article were defrayed in part by page charge ium bromide-stained gel. payment. This article must therefore be hereby marked "advertise- ment" in accordance with 18 U. S. C. §1734 solely to indicate this fact. Abbreviations: OX, phage 4X174; RF, replicative form. 843 Downloaded by guest on September 25, 2021 844 Biochemistry: Brown and Cozzarelli Proc. Nad Acad. Sci. USA 78 (1981) RESULTS ca Catenates Duplex DNA Circles. Interlocking or fusion of DNA rings can readily be assayed by the reduction of their elec- trophoretic mobility in agarose gels. A characteristic ladder of slower-moving forms is produced or, if the products are suf- ficiently large, the DNA remains at the origin (4). By using this itA assay, catenation by gyrase and topoisomerase II' (2) were easily detected in crude extracts of E. coli. Two additional activities were discovered when the extract was fractionated by DEAE-Sephacel chromatography. Surpris- ingly, one of these coincided with a supercoil-relaxation activity FIG. 2. Catenation of nonhomologous DNA. A 200-MAl catenation that had properties characteristic of w. The protein, purified reaction mixture containing 1 ,ug each of nicked pA03 and nicked OX on the basis of the relaxation activity, was a 10,000-dalton poly- RF DNA and 2 ,±g of co was incubated for 3.hr and spread for electron peptide that comigrated on NaDodSO4polyacrylamide gels microscopy. The catenane shown contains four pAO3 and two OX RF with authentic A. Fig. 1A shows catenation of nicked #X174 molecules. (x51,000.) RFI DNA by the purified protein at molar ratios to DNA circles ranging from 0.3 (lane g) to 150 (lane b). The proof that the tions of cations for catenation were nearly identical to their crit- multimers are catenanes is presented below. ical concentrations for induction of DNA aggregation under The ratio of catenation to relaxation activity was constant similar ionic conditions (12), suggesting that the driving force throughout purification. Two other preparations purified by for catenation was provided primarily by polycation-promoted using different procedures by R. E. Depew and by J. C. Wang DNA aggregation. had similar specific activities. For all three preparations, 10-20 Under standard conditions for DNA aggregation, supercoil times more enzyme was needed to catenate than to relax half relaxation was inhibited so that the ratio of relaxation events of the DNA substrate, under the respective optima. As ex- (unit changes in Lk) to net catenation events was about 10. pected for w (1), catenation was inhibited by single-stranded Catenation Does Not Depend on Extensive Sequence Ho- DNA and required Mg(II) (Fig. 1B, lanes band c). We conclude mology. co and other type 1 topoisomerases have always been that the catenation activity is intrinsic to ct. assumed to act at base-paired regions (1, 8, 13). To determine The optimal conditions for catenation included low ionic if extensive sequence homology was required for catenation, strength and the presence of a polycationic DNA-condensing we used unrelated DNA molecules that could be distinguished agent, such as spermidine (Fig. 1B, lane a), spermine (lane e), by electron microscopy. With a mixture of pAO3 (1683 base or cobaltic hexammine (lane f) (12). The threshold concentra- pairs) (11) and #X DNA (5386 base pairs), heterologous products predominated (Fig. 2). (Although most of the products of the spermidine-promoted reactions were networks containing >100 A B circles, electron micrographs of more easily interpreted small a bc d e f g a b c d e f oligomers are presented.) Homologous circles are not prefer- Top of gel-- e-Top of gel entially linked. In a reaction with supercoiled ColEl and an equal mixture, by weight, of nicked ColE1 and nicked OX RFI DNA, the ratio of nicked ColEl-closed ColEl to nicked OX-closed ColEl dimers was very close to 1. Thus catenation by w shows no dependence on sequence homology. Role of Nicks. We were surprised that nicked DNA was far superior to supercoiled DNA as a substrate for catenation, be- cause cl can bind preferentially to the latter (1). The basis for this result was uncovered by the work of Y.-C. Tse and J. C. f S relaxed - and .: nicked A B a b c d e f g h top

6 v- supercoi led v£ %o0l 2S nicked -* a2. n-cked 6 O5X < relaxed

FIG. 1. Catenation by co. (A) The 20-ul catenation reaction mix- 0 2o 40 60 G6 b tures containing 0.2 Mkg of nicked OX RF DNA and: lane a, no cv; lane % ncked b, 1000 ng of a; lane c, 200 ng of c; lane d, 100 ng of a; lane e, 20 ng native of a; lane f, 4 ng of co; or lane g, 2 ng of w were incubated for 90 min then electrophoresed through a 0.7% agarose gel. The ethidium bro- FIG. 3. Role of nicks in catenation. (A) The 20-MAI reaction mixtures mide-stained gel is shown. Because the band at the origin is difficult contained 0.2 Ag of the indicated mixture of native and nicked OaX RF to quantitate, catenation is best estimated from depletion of mono- DNA and 40 ng of a. After 1 hr, the products were resolved by elec- meric starting material. (B) Reaction mixtures contained 0.1 Mig each trophoresis and quantitated by scanning a photographic negative of of nicked and native OX RFI DNA and 40 ng of a, with the following the gel. (B) The 20-MuI reaction mixtures contained 0.1 pg each of: lanes modifications: lane a, none; lane b, minus MgCl2; lane c, plus 0.2 MUg a-c, supercoiled and relaxed ColEl; lanes d-f, supercoiled and nicked of viral #X DNA; lane d, minus spermidine; lane e, same as lane d, plus ColEl; lanes g-i, relaxed and nicked ColEl; and no w (lanes a, d, and 0.2 mM spermine; lane f, same as lane d, plus 1 mM cobaltic g), 10 ng of c (lanes b, e, and h), or 20 ng of et (lanes c, f, and i). The hexammine. products of the 1-hr reactions were resolved by electrophoresis. Downloaded by guest on September 25, 2021 Biochemistry: Brown and Cozzarelli Proc. Natl. Acad. Sci. USA 78 (1981) 845 Wang (personal communication) showing that a nicked DNA A B molecule appears to be required for catenation. As illustrated in Fig. 3A, the extent of catenation increased monotonically a bc with the ratio of nicked to supercoiled substrate. A break in the DNA and not the absence of supertwists is required, because relaxed closed DNA circles could not replace nicked DNA (Fig. 3B, lanes a-f). The effect of nicks is enzyme-specific, because catenation by gyrase (4) shows no enhancement by nicks. Only one of each pair of DNA circles needed to be nicked for catenation by w, and remarkably there was no preference catenated -> for either a nicked or supercoiled circle as the second member cdimer of a linked pair. Thus, the proportion of catenated dimers com- posed of two nicked monomers and the proportion consisting of one closed and one nicked paralleled the proportions of un- reacted monomers that were nicked or closed, respectively. Dimers composed of two closed circles were never detected. Although closed relaxed DNA could not substitute for nicked DNA (Fig. 3B, lanes a-c), in the presence of nicked circles, relaxed and supercoiled DNA molecules were incorporated nicked Ll |i equally into catenated networks (Fig. 3B, lanes d-i). Proof that the DNA Multimers Are Catenanes. The data presented thus far do not distinguish conclusively among four linear - possible linkages in the DNA multimers produced by w. These are: (i) covalently fused rings (figure 8s), (ii) one strand of one circle passing between the two strands of another, (iii) both strands of one circle passing between the two strands of an- other, and (iv) catenation. The first two possibilities were ruled out by a crosslinking experiment (Fig. 4). An equal mixture of FIG. 5. Knotting by w. (A) For knotting, spermidine was omitted supercoiled and nicked ColE 1 monomers was the substrate for from the catenation reaction mixture. Reaction mixtures contained 0.3 W, and product multimers were separated from unreacted mon- pg of nicked ColEl (containing 5% linear contaminants) and: lane a, omers by sedimentation (4). The multimers were then cross- no w; lane b, 0.5 jg of co; or lane c, 2 jZg of co. Electron micrographs linked with trioxsalen on average once per 500 base pairs and of knotted nicked OX RFI are shown in B. (x62,000.) treated with EcoRI endonuclease, which cuts ColEl. If the omitted from a catenation reaction mixture containing nicked junctions were of forms 1 or 2, most of the resulting ColE1 lin- circular DNA and high levels of w (5- to 100-fold molar excess ears could not separate, because in the former case they would over DNA), a regular series of bands with increased electro- be covalently linked and in the latter topologically linked by the phoretic mobilities appeared (Fig. 5A). This pattern is typical crosslinks. In fact, >95% of the DNA was released as linear of nicked, knotted DNA circles (3). When the products of a monomers. reaction that converted 20% of the nicked circles into higher A structure of the third kind would be unstable if the "in- mobility forms were examined by electron microscopy, 26% (29/ vaded" double-helix were nicked, resolving into a catenane or 110) of the circular monomers appeared to be knotted (Fig. 5B). free monomers after migration of the junction to and through Less than 1% of control DNA had structures that could be in- a nick. Thus, we need to rule out only structures with invading terpreted as knots. nicked circles. Heterologous multimers formed between closed Even though spermidine was omitted for the knotting re- circular ColE 1 and nicked OX RFI were crosslinked and treated action, a low yield (12%) ofcatenated oligomers, chiefly dimers, with EcoRI endonuclease. If the junctions were of form 3, the was observed at these high enzyme levels. Furthermore, the crosslinked ColEl linears would have remained topologically yield of knotted forms (up to 35%) was over an order of mag- linked to the circular OX DNA; but, in fact, only OX-OX link- nitude higher than expected at equilibrium in the absence of ages were spared. We conclude that the multimers produced deforming forces (14). These results suggest that binding of w by w are catenanes. can deform the DNA to favor knotting and promote aggregation. Knotting of Duplex DNA. Knotting is an intramolecular an- Relaxation by w Changes Linking Number in Steps of One. alog of catenation. When the DNA-aggregating agent was Although it is commonly thought that type 1 topoisomerases such as w have been shown to change Lk in steps of one, the experiments published to date show only that this is true if the linear catenated step has a unique value. In the swivel model for topoisomerases, Col El multi mers vu axial rotation of the nicked DNA intermediate and therefore ALk is not intrinsically constrained to one turn. Because catena- tion and knotting cannot be simply explained by the swivel model, we reinvestigated the linking number change by W. FIG. 4. Proof that the DNA multimers are catenanes. A 500-gLl Reaction conditions were chosen to favor distributive relaxation reaction mixture contained 10 jig each of nicked and native ColEl and by w (high ionic strength, low supercoil density), because it 10 Mg of w. The reaction was stopped after 3 hr by addition of EDTA would be difficult to distinguish multiple processive relaxation to 10 mM and heating to 65°C for 5 min. Multimers were purified by events in steps of 1 from a ALk of >1 per round. The time sedimentation onto a 1.75-g/ml CsCl shelf, thendialyzed against5 mM course of relaxation of two different topoisomers of pAO3 DNA Tris HCl (pH 7.8)/0.1 mM EDTA. Next, 0.4 Hg of the multimers was crosslinked (1/500 base pairs) with trioxsalen and half was then di- with roughly half the native superhelical density (Fig. 6) shows gested with EcoRI endonuclease. The undigested (lane a) and EcoRI- that the first reaction product indeed had a ALk of 1. About 15% digested sample (lane b) were electrophoresed through an agarose gel. of the -4.5 topoisomer (Fig. 6A) had gained a single link before Downloaded by guest on September 25, 2021 846 Biochemistry: Brown and Cozzarelli Proc. Nad Acad. Sci. USA 78 (1981)

A top cf gel- B

a .:/ *e V . . -. ricked CcE-l-

nicked 4OX-> '9 0 min native ColEI - FIG. 7. Linking of double-stranded DNA by a eukaryotic type 1 topoisomerase. (A) The 20-pi reaction mixtures contained: lanes a-c, 0.3 pugof native ColE1 (contaminated with 5% nicked molecules); lanes d-e, 0.3 ,g of nicked ColEl; plus no enzyme (lanes a and d), 50 ng of rat liver type 1 topoisomerase (lanes b and e), or 100 ng of topoiso- merase (lanes c and f). After 3 hr, samples were electrophoresed through a 0.7% agarose gel. (B) A 20-LI reaction mixture contained 0.05 ug of nicked pAO3, 0.05 pug of OX RFI, and 0.1 pg of rat liver enzyme. After 2 hr, a deproteinized sample was spread for electron microscopy. (a) Typical "key ring" form consisting of a central nicked pA03 molecule into which are catenated six 4X and one pAO3 molecule. (b) Tetramer containing two pAO3 and two X5X monomers. (c) Trimer ALk 0 1 2 0 1 2 3 containing two OX monomers linked to a single pAO3. (a-c, X 17,000.) that catenation by this enzyme preparation is Mg(II)-indepen- FIG. 6. Relaxation by w changes linking number in steps of one. The 10-lI. reaction mixtures contained 30 mM TrisHCl (pH 7.6), 4 mM dent (J. Champoux, personal communication). We conclude MgCl2, 16 mM KCl, 160 mM NaCl, 1 mM dithiothreitol, serum albu- that: (i) The eukaryotic type 1 topoisomerase can interlock du- min at 80 pg/ml, either 1 pug (in A) or 0.2 ,g (in B) of w, and 0.08 pg plex circular DNA; (ii) molecules to be linked need not have of a purified topoisomer of pAO3 underwound by about 4.5 (A) or 5.5 extensive sequence homology; and (iii) probably one (but not (B) turns relative to nicked DNA under reaction conditions. The ALk of the molecules to be linked must be relative to the starting material is shown at the bottom of the figure. both) nicked. After the times indicated at 20TC (A) or 15TC (B), the reaction products were resolved by electrophoresis through a 0.5% agarose/2% poly- DISCUSSION acrylamide composite gel. A tracing of a photographic negative of the stained gel is shown. Type 1 topoisomerases have previously been shown to catalyze supercoil relaxation (1), knotting of single-stranded circular any product with ALk of 2 appeared. The later-appearing prod- DNA (8), and renaturation of complementary single-stranded ucts with ALk of 2 and 3 were, we conclude, derived from the circular DNA (9, 10). In this report we have shown that both ALk of 1 and 2 precursors, respectively. The strand-passage E. coli w and a eukaryotic type 1 topoisomerase can interlock models considered in Discussion require that ALk = 1 per cat- duplex circular DNA, an activity that had been thought forbid- alytic cycle, as shown by the results in Fig. 6. It is not ruled out den for such enzymes. The co catenation reaction, which we that, under reaction conditions favoring rapid processive relax- have examined in greater detail, has four illuminating proper- ation, ALk exceeds one per cycle, but this seems unlikely. ties: (i) extensive sequence homology between reacting DNA A Eukaryotic Type 1 Topoisomerase Can Interlock Duplex molecules was not required; (ii) at least one of each pair of DNA. To explore the generality of catenation by type 1 en- molecules to be linked had to be nicked; but (iii) the other DNA zymes, we incubated nicked OX RF (Fig. 7A, lanes a-c) or a circle did not have to be nicked; there was no preference among mixture of supercoiled and nicked ColEl (Fig. 7A, lanes d-f) nicked, supercoiled, and relaxed circles as the second reactant; with the nicking-closing enzyme from rat liver nuclei. Agarose (iv) the net frequency of intermolecular DNA strand passage gel electrophoresis revealed, in both cases, a band at the top events leading to catenation was at least 10% of the frequency of the gel. Nicked DNA was the preferred substrate. The prod- of intramolecular strand passages leading to relaxation. ucts are indistinguishable by electron microscopy from the cate- The swivel model can account for the linking of two DNA nanes produced by w, but we have not yet unambiguously de- strands only if they are base-paired so that the intact strand can termined their structure. With native OX RFI (contaminated provide a pivot for the transient free ends. Thus it follows from with 10% nicked circles) and nicked pAO3 DNA as substrates, property i that the swivel mechanism does not operate in cate- electron microscopy showed that pAO3 was linked to both pAO3 nation. Moreover, it is not obvious why a swivel should be lim- and OX circles (Fig. 7B), but 45X-OX links were only rarely ited to 1 turn per catalytic cycle, as would be required to explain seen. When several molecules were linked to a central ring, the the steps of 1 in w-catalyzed relaxation. We propose that these central ring was usually the nicked pAO3 (Fig. 7B, panel a). type 1 topoisomerases, like the type 2 topoisomerases, are fun- Because neither multimer production nor supercoil relaxation damentally enzymes that pass one segment of DNA through a were stimulated by 1 mM ATP, they were not due to contam- transient break in another, while holding on to both the 3' and ination by the ATP-dependent type 2 topoisomerase (2). The 5' ends created by the break. The critical distinction is that eukaryotic 70,000-dalton type 1 enzymes are unique in not re- whereas type 2 enzymes pass one duplex segment through an- quiring Mg(II) for supercoil relaxation (13). It is thus significant other, changing Lk by 2, type 1 topoisomerases pass a DNA Downloaded by guest on September 25, 2021 Biochemistry: Brown and Cozzarelli Proc. NatL Acad. Sci. USA 78 (1981) 847

segment through a single strand and Lk changes in steps of 1. pairing in relaxed DNA. There is thus no evidence for the local The surprisingly high frequency of catenation suggests that base denaturation required by the first class of models. pairing between the two passing DNA segments may not in The second class of models does not have the disadvantage general be important for w reactions. Our data (Fig. 5) sug- of a high-energy locally denatured intermediate. In one ver- gesting that w promotes DNA aggregation and deforms DNA sion, two w molecules could act in concert as a type 2 enzyme, to favor knotting could be explained by a binding thatjuxtaposes but this model fails to explain the role of the essential nick. the sugar-phosphate backbones of two DNA segments. Because More reasonably, catenation by wt could occur by the single-step strand separation is favored by negative supercoiling, the de- passage of a double-helical segment of DNA through a transient pendence of w activities associated with relaxation on super- break opposite the nick. Because such a mechanism involves helical density (1) suggests that such a non-base-paired config- a transient double-strand break, it is in effect a type 2 mecha- uration may be needed for intramolecular strand passages. nism. Because co binds selectively to single-stranded DNA, and Mechanisms for catenation by w either invoke iterated pas- the DNA backbone opposite a nick would resemble single- sages of single strands through single strands or postulate a sin- stranded DNA in its flexibility, a selective binding at nicks gle-step passage of a double helix through one or two strands. would not be implausible. This mechanism could explain why We have discussed elsewhere (7) an example of the former class co typically generates structures in which several rings are ca- of models, in which catenation occurs by successive passages tenated into a central nicked ring. The DNA sequence speci- of the two strands of a duplex DNA circle through the intact ficity of co would create a particularly receptive substrate in strand of a nicked circle, with diffusion of each invading single which this sequence is opposite a nick. A disturbing feature of strand to and through the nick. Such models generate an in- this model is the requirement that the enzyme be able to pass, termediate in which a DNA strand is threaded between the two in spite of their obvious structural differences, a single- or dou- strands of a double helix, with concomitant disruption of base ble-stranded DNA segment through the transient break. Fig. pairing. Therefore, unless these unpaired regions were already 8 illustrates how this mechanism might operate to catenate (A) present, formation of such an intermediate would be impeded or relax (B) duplex DNA. The eukaryotic DNA-untwisting en- by a substantial energy barrier. Although the energetic cost of zyme shows no specific binding to single-stranded DNA (13). denaturation is substantially lower in an underwound DNA Thus, this enzyme might execute the sequence shown in Fig. molecule, there was no preference for negatively supercoiled 8B forwards to relax negative twists and backwards to relax pos- DNA over relaxed DNA for catenation with a nicked circle. itive twists, without locally unwinding DNA. Furthermore, experiments revealed no detectable difference Our discovery of catenation and knotting by type 1 topoiso- in Lk of closed circular pAO3 DNA prepared by ligating in the merases has invalidated a supposed key distinction between the absence or in the presence of w at molar ratios to DNA of from two types of topoisomerases, but our results actually sharpen 1 to 25, suggesting that w does not appreciably disrupt base and simplify the discrimination. Type 1 enzymes pass a DNA segment through a transient, enzyme-bridged, one-strand break, change Lk in steps of 1, can be induced to generate DNA interrupted in one strand, and can reversibly knot or catenate DNA rings with one intact strand. In contrast, type 2 enzymes pass a DNA segment through a transient, enzyme-bridged, two- strand break, change Lk in steps of 2, can be induced to gen- erate DNA interrupted in two strands, and can reversibly knot or catenate DNA rings with two intact strands. This work was supported by National Institutes of Health Grants GM- 21397 and CA-19265. P.O.B. was supported by National Institutes of Health Fellowship GM-07281. 1. Wang, J. C. & Liu, L. F., (1979) in MolecularGenetics, ed. Taylor, J. M. (Academic, New York), Part 3, pp. 65-88. 2. Cozzarelli, N. R. (1980) Science 207, 953-960. 3. Liu, L. F., Liu, C. C. & Alberts, B. M. (1980) Cell 19, 697-708. 4. Kreuzer, K. N. & Cozzarelli, N. R. (1980) Cell 20, 245-254. 5. Mizuuchi, K., Fischer, L. M., O'Dea, M. H. & Gellert, M. (1980) FIG. 8. Model for wreactions. wis depicted as a C-shaped structure. Proc. Natl Acad. Sci. USA 77, 1847-1851. In one arm is the active site for DNAbreakage and rejoiningthat bonds 6. Brown, P. 0. & Cozzarelli, N. R. (1979) Science 206, 1081-1083. with the 5' end of the transient break, and in the other arm is a tight 7. Morrison, A., Brown, P. O., Kreuzer, K. N., Otter, R., Gerrard, binding site for the 3' side of the break. At the end of either sequence, S. P. & Cozzarelli, N. R. (1980) in Mechanistic Studies of DNA at least one of the arms must relinquish its grip, to release the trav- Replication and , eds. Alberts, B. M. & ersing DNA segment from the hollow. (A) Catenation or knotting. (i) Fox, C. F. (Academic, New York), pp. 785-807. The enzyme arms bind opposite a nick, and another double-helical 8. Liu, L. F., Depew, R. E. & Wang, J. C. (1976)J. Mol. Biol. 106, DNA segment is bound at the base of the C. (ii) The strand opposite 439-452. the nick is broken and the second DNA segment then passes through 9. Champoux, J. (1977) Proc. NatL Acad. Sci. USA 74, 5328-5332. the gap into the hollow of the C, whereupon, (iii) the transiently broken 10. Kirkegaard, K. & Wang, J. C. (1978) Nucleic Acids Res. 5, is restored. (B) Relaxation. (i) On binding to neg- 3811-3820. atively supercoiled DNA, wlocally separates the DNA strands, holding 11. Oka, A., Nomura, N., Morita, M., Sugisaki, H., Sugimoto, K. one strand at the base of the C and the other with its arms. The tor- & Takanami, M. (1979) Mol. Gen. Genet. 172, 151-159. sional strain in the underwound DNA tends to induce a left-handed 12. Widom, J. & Baldwin, R. L. (1980)J. Mol. Biol. 144, 431-453. helical configuration in the melted region. (ii) The first strand is bro- 13. Champoux, J. (1978) Annu. Rev. Biochem. 47, 449-479. ken, the second strand passes through the break into the hollow of the 14. Frank-Kamenetskii, M. D., Lukashin, A. V. & Vologodskii, A. V. C, and (iii) the break is resealed. (1975) Nature (London) 258, 398-402. Downloaded by guest on September 25, 2021