Roles of topoisomerases in maintaining steady-state DNA supercoiling in Escherichia coli.

DNA supercoiling is essential for bacterial cell survival. We demonstrated that DNA topoisomerase IV, acting in concert with topoisomerase I and gyrase, makes an important contribution to the steady-state level of supercoiling in Escherichia coli. Following inhibition of gyrase, topoisomerase IV alone relaxed plasmid DNA to a final supercoiling density (sigma) of -0.015 at an initial rate of 0.8 links min(-1). Topoisomerase I relaxed DNA at a faster rate, 5 links min(-1), but only to a sigma of -0.05. Inhibition of topoisomerase IV in wild-type cells increased supercoiling to approximately the same level as in a mutant lacking topoisomerase I activity (to sigma = -0.08). The role of topoisomerase IV was revealed by two functional assays. Removal of both topoisomerase I and topoisomerase IV caused the DNA to become hyper-negatively supercoiled (sigma = -0.09), greatly stimulating transcription from the supercoiling sensitive leu-500 promoter and increasing the number of supercoils trapped by lambda integrase site-specific recombination.

DNA supercoiling is essential for viability and is tightly regulated in the cell. Many enzymes not only are affected by the level of supercoiling in their DNA substrate but can also alter DNA supercoiling. For example, DNA replication initiation requires a negatively supercoiled template (1,2). DNA strand separation and unwinding by the progression of the replication machinery cause (ϩ) supercoil accumulation, which must be removed by topoisomerases for replication to be completed (reviewed in Refs. 3 and 4). Initiation of transcription of a number of genes also requires (Ϫ) supercoiling (5)(6)(7). In addition, tracking by the transcription machinery transiently separates DNA into supercoiled domains with (ϩ) supercoils in front of the RNA polymerase and (Ϫ) supercoils behind it (8,9). Many site-specific recombination systems require or are influenced by (Ϫ) supercoiling (10,11) and can transmogrify DNA supercoils into DNA catenanes or knots (12,13). DNA supercoiling not only affects strand separation and protein binding but also condenses DNA and promotes the interaction of distant sequences (reviewed in Ref. 14). Without this condensation, chromosome segregation may fail (15)(16)(17). Finally, the enzymes that control and maintain DNA supercoiling, the DNA topoisomerases, are themselves strongly regulated by DNA supercoiling (18 -21).
In eubacteria, there are four topoisomerases and each could have a role in maintaining DNA supercoiling (reviewed in Refs. 4 and 22). We focus on the properties of the topoisomerases of Escherichia coli. Topoisomerase I, encoded by the topA gene, and topoisomerase III, encoded by the topB gene, are type 1 topoisomerases, those that make single-stranded breaks in DNA and catalyze linking number (Lk) 1 changes in steps of one (reviewed in Ref. 4). topA deletions in E. coli cause an increase in (Ϫ) supercoiling and are tolerated only in the presence of compensatory mutations elsewhere that reduce the level of supercoiling (23,24). Topoisomerase III is dispensable for cell viability and has no known role in DNA supercoiling, but its deletion causes a mutator phenotype (25,26). The similarity of topoisomerase III to the important eukaryotic homologues encourages a search for key functions (27)(28)(29). The type-2 topoisomerases in bacteria, DNA gyrase, and topoisomerase IV are essential for cell viability (reviewed in Ref. 30). The type-2 enzymes make double-stranded DNA breaks and alter the Lk of DNA in steps of two. The genes encoding the topoisomerase IV subunits are parC, which is homologous to gyrA of gyrase, and parE, which is homologous to gyrB of gyrase. The importance of these two enzymes is underscored by the fact that they are the cellular targets for the widely prescribed quinolone antibiotics (reviewed in Refs. 30 and 31). The type 1 enzymes are not affected by quinolones in vitro (32).
Despite nearly 40% amino acid sequence identity between gyrase and topoisomerase IV (33), no overlap in specific function has been found. Although both enzymes are required for DNA replication, they have different roles. Gyrase introduces (Ϫ) supercoils, which are required for the initiation of replication (34), and removes (ϩ) supercoils to allow fork progression (see Ref. 3). The critical role for topoisomerase IV in replication is to unlink the catenated intermediates (15,35,36). Topoisomerase IV also unlinks catenanes generated by site-specific recombination (16). The decatenation of replication or recombination intermediates by topoisomerase IV is facilitated by the (Ϫ) supercoiling activity of gyrase (15,16). Prior to the discovery of topoisomerase IV, it was thought that gyrase carried out all of the functions now ascribed to topoisomerase IV (37)(38)(39).
The wild-type, steady-state supercoiling level, , of around Ϫ0.06 to Ϫ0.075 had long been thought to be regulated only by the opposing activities of topoisomerase I relaxing and gyrase introducing (Ϫ) supercoils. The following evidence supported this model. (i) The purified enzymes efficiently carry out these functions in vitro (40 -42). (ii) DNA isolated from topoisomerase I mutant strains generally is more (Ϫ) supercoiled than that from wild-type strains (43)(44)(45). (iii) For topoisomerase I deletion mutants to be viable, there must be compensatory mutations elsewhere, many of which mapped to the gyrase genes (23,24,46,47). (iv) Inhibition of DNA gyrase in cells leads to DNA relaxation and, in those DNA molecules being transcribed, (ϩ) supercoil accumulation. The (ϩ) supercoils were thought to arise because topoisomerase I removes the (Ϫ) supercoils behind RNA polymerase during transcription, but not the (ϩ) supercoils ahead (9,43).
Now it is known that in many of these early experiments in which gyrase was inhibited by quinolone or coumarin drug addition, topoisomerase IV was also inhibited (16,48,49). Therefore, a potential role for topoisomerase IV in DNA relaxation would have been masked. Indeed, several results have implied such a role for topoisomerase IV. (i) Topoisomerase IV relaxes DNA in vitro (50). (ii) Overexpression of topoisomerase IV in E. coli reduces the expression of a gene that is sensitive to DNA supercoiling, tyrT (73). (iii) Overexpression of topoisomerase IV in E. coli as well as in Shigella flexneri compensates for the loss of topoisomerase I (33,51). (iv) In E. coli, when gyrase activity is blocked, integrase (Int) recombination is reduced approximately 7-fold when topoisomerase IV remains functional compared with when topoisomerase IV is blocked, showing that topoisomerase IV is able to relax DNA (16).
We have determined systematically the relative contributions of topoisomerase I, topoisomerase IV, and gyrase in regulating DNA supercoiling in vivo. We show that topoisomerase I and topoisomerase IV counter gyrase to maintain DNA supercoiling and that each topoisomerase has distinct roles.

EXPERIMENTAL PROCEDURES
Chemicals and Reagents-Radioactive [␣-32 P]dCTP (250 mCi/mmol; 1 Ci/ml) was obtained from NEN Life Science Products. The supercoiled DNA ladder, 1-kilobase ladder, and calf thymus topoisomerase I were obtained from Life Technologies, Inc. Proteinase K was obtained from Roche Molecular Biochemicals. Norfloxacin, fusaric acid, chlorotetracycline, chloroquine, RNaseA, and DNase I were obtained from Sigma. Multiprime DNA labeling kit was obtained from Amersham Pharmacia Biotech and Nytran (MSI) transfer membrane from Fisher.
Bacterial Strains and Plasmids-The bacterial strains used in this study are listed in Table I. Four isogenic sets of strains containing all combinations of drug-resistant and wild-type alleles of gyrA and parC were constructed. Two sets were based upon the parental E. coli strain C600 that we used in earlier work (15,48). The strains used for the Int experiments were derived from the parental lysogen strain W3101 (38) and were designated LZ33-38. The topoisomerase I mutant strains were LZ53 and LZ54. In the text, for simplicity, we refer to these strains by their topoisomerase allelic states. Plasmids pJB3.5d, 3500 bp (38); pBR322, 4363 bp (52); pLEU500Tc, 3825 bp (7); and pGP509, 4900 bp (53) have been described previously. Strain Construction-The construction of the Tn10-linked gyrA ϩ and gyrA L83 (15) and kan-linked parC ϩ , parC L80 , and parC K84 (48) strains was described previously. For the Int experiments, the Tn10-linked gyrA or kan-linked parC alleles were transduced with P1 vira (54) into W3101. The strains containing mutant topoisomerase I were made by transferring the kan marker from CAG12183 into strains N51 and KL16 (55,56). Then the gyrA L83 or gyrA ϩ gene with its linked kan marker was transduced into the lysogen RS2 (38). The presence of the lysogen was verified by temperature sensitivity at 42°C. The presence of the topA10 mutation was verified by decreased plasmid supercoiling.
Inhibition of Topoisomerases with Norfloxacin-We used norfloxacin (30 M) to inhibit wild-type gyrase and topoisomerase IV in vivo as described (16,48). Drug-resistant mutants in either or both enzymes allowed us to inhibit only gyrase, only topoisomerase IV, both enzymes, or neither enzyme.
Assays for Plasmid DNA Supercoiling-Cells were grown in LB with ampicillin (50 g/ml) with shaking at 30°C (lysogen and temperaturesensitive strains) or 37°C to a density of 70 Klett units (mid-log phase). At this point (time 0), norfloxacin (30 M) was added to inhibit gyrase and/or topoisomerase IV. Cells continued shaking at the same temperature through the experiment. Aliquots (1.9 ml) were removed at various times during the incubation, immersed in liquid nitrogen, stored at Ϫ80°C, and thawed on ice (15). We believe that this rapid freeze stops all cellular activities and allows for more accurate measurements of and may explain why others, in general, have obtained somewhat more relaxed supercoiling density values. Plasmid DNA was isolated by the alkaline lysis procedure (57) and treated with Rnase A.
DNA Relaxation with Calf Thymus Topoisomerase I-5 g of plasmid DNA (pJB3.5d or pBR322) was relaxed to completion with 10 units of calf thymus topoisomerase I in 50 l of buffer that contained 10 mM Tris-Cl, pH 7.9, 50 mM KCl, 0.1 mM EDTA, and 5 mM MgCl 2 . Incubation was for 1 h at 37°C.
Int Recombination-For all experiments, we verified that topoisomerase IV had been inhibited by measuring accumulation of catenated products of Int recombination exactly as in Ref. 16.
Western Analysis-Cells were grown to an A 600 of 0.6 at 30°C and shifted to 42°C for 40 min. For comparative Western analysis, equal amounts of cells were collected, resuspended in B50 buffer (50 mM NaCl, 0.5% Triton X-100, 10 mM Tris-Cl, pH 8) and lysed by vortexing with glassbeads at 4°C for 20 min in the presence of complete proteinase inhibitor (Roche Molecular Biochemicals). The amount of protein was measured using a BCA Protein Assay (Pierce). 20 g of protein was loaded in each lane and displayed using SDS-polyacrylamide gel electrophoresis. Western blotting was carried out as described (58) except that electrophoresis was performed in a 10% polyacrylamide gel containing SDS and that 10% methanol was included in the buffers during electrotransfer of proteins onto polyvinylidene difluoride membranes (Immobilon-P, Millipore). GyrA and GyrB were detected using GyrAand GyrB-specific (mouse) antibody (Lucent). Horseradish peroxidaseconjugated mouse (Promega) IgG-specific secondary antibodies were used for detection with an ECLϩPlus kit (Amersham Pharmacia Biotech). The autoradiographs were scanned by densitometry and quantified by NIH-IMAGE software. The experiments were repeated four times. The results were highly reproducible.
Gel Electrophoresis and Quantification-Purified supercoiled plasmids were analyzed by electrophoresis through 1.2% agarose gels with 0, 2, 4, 6, or 8 g/ml chloroquine and compared with plasmids of known supercoil density. The mean values reported were quantified by the band counting method (59). A of zero was taken as the midpoint of the topoisomer distribution after plasmid DNA relaxation in vitro with calf thymus topoisomerase I (see above). Plasmid DNA in the gels was transferred to Nytran membranes and probed with labeled plasmid. The Southern blots were exposed to Kodak XAR film or PhosphorImager (Molecular Dynamics) cassettes. The amount of radioactivity in each band was determined by PhosphorImager analysis. The rates for topoisomerase IV-or topoisomerase I-mediated DNA relaxation were calculated from the slopes of the least squares best fit lines. was calculated according to the formula, ϭ ⌬Lk/Lk 0 , where Lk 0 for pJB3.5d ϭ 3500 bp/10.45 bp/turn ϭ 335.
Extraction and Analysis of Cellular RNA-RNA was prepared from E. coli cultures in mid-exponential growth phase as described previously (60). In brief, 200 l of cultures were lysed by a 1-min incubation in a boiling bath with an equal volume of 20 mM sodium acetate (pH 5.2), 2% SDS, 0.3 M sucrose. The sample was then extracted twice with phenol and chloroform, and the nucleic acids were precipitated with ethanol. After the addition of 0.2 pmol of 5Ј-32 P-labeled primer, the sample was heated to 90°C in 4.5 l of 50 mM Tris-Cl, pH 8.0, 50 mM KCl and rapidly cooled. 25 units of RNasin were added prior to a 20-min incubation at 42°C. 12 l of 70 mM Tris-Cl, pH 8.0, 70 mM KCl, 15 mM MgCl 2 , 15 mM dithiothreitol, 1.3 mM deoxyribonucleoside triphosphates containing 50 units of moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) were added, and the solution was incubated for 2 h at 42°C. cDNA made from the transcripts was analyzed by electrophoresis in 6% polyacrylamide gels in 90 mM Tris borate (pH 8.3), 10 mM EDTA buffer containing 7 M urea. Gels were dried, and radioactive bands were quantified by exposure to storage phosphor screens and analysis using a Fuji BAS-1500 phosphoimager.

RESULTS
Experimental Approach-To measure the contributions of the topoisomerases in maintaining DNA supercoiling in E. coli, we inhibited topoisomerase I, topoisomerase IV, and DNA gyrase individually or in combination and assayed the resulting changes in plasmid supercoiling. Because gyrase and topoisomerase IV are essential enzymes, we tested the kinetics of the topoisomerases before the cells died. We inhibited gyrase with 30 M of the quinolone, norfloxacin, which completely blocks negative supercoiling by gyrase in vivo (16,48). This concentration of norfloxacin also blocks approximately 90% of wild-type topoisomerase IV activity in vivo, as assayed by the accumulation of catenated intermediates of DNA replication (48) or recombination (16). In some experiments we used the drug-resistant allele of topoisomerase IV (parC K84 ), which is resistant to this concentration of norfloxacin (16,48). Therefore, norfloxacin added to a topA ϩ parC K84 gyrA ϩ strain allows DNA relaxation by both topoisomerase I and topoisomerase IV (gyrase is inhibited), whereas the drug added to the topA ϩ parC ϩ gyrA ϩ strain reveals relaxation by topoisomerase I alone (topoisomerase IV and gyrase are blocked). To measure the contribution of only topoisomerase IV to DNA relaxation, we used a mutant in topoisomerase I, topA10, that has about one-hundredth the activity of wild-type topoisomerase I (23,46). Because of the low topoisomerase I activity, plasmid DNA isolated from this mutant strain is more (Ϫ) supercoiled than that from a wild-type strain (46). In a topA10 parC K84 gyrA ϩ strain, the addition of norfloxacin reveals essentially the DNA relaxation rate of topoisomerase IV alone (topoisomerase I and gyrase are inactive). The drug added to the topA10 parC ϩ gyrA ϩ tests whether any other enzyme contributes significantly to DNA relaxation (topoisomerase I, topoisomerase IV, and gyrase are inactive). The strains used in this study are listed in Table I.
For most of these experiments we used the plasmids pJB3.5d, which lacks the tetracycline resistance gene (tetA), and pBR322, which contains tetA. Expression of tetA anchors the transcription complex to the cell membrane, because the gene is coordinately transcribed and translated, and the polypeptide product inserts into the membrane (7,9,43,45,60,61). Instead, pJB3.5d contains bla, which confers ampicillin resistance. Export of ␤-lactamase is not coordinated with transcription of bla and the complex is not anchored to the cell membrane. Transcription-induced supercoiling in plasmids containing the bla gene is therefore much less efficient than in plasmids with the tetA gene (45).
DNA Relaxation by Topoisomerase I and Topoisomerase IV Following Gyrase Inhibition-The change in the Lk of pJB3.5d following gyrase inhibition with norfloxacin was analyzed by gel electrophoresis as illustrated in Fig. 1A-C; the mean values from two experiments are plotted in Fig. 1, D and E. Fig.  1A shows the relaxation of DNA by topoisomerase I alone after the inhibition of gyrase (left), and the data are plotted in Fig. 1, D and E (Ⅺ). Relaxation by both topoisomerase I and topoisomerase IV is shown in Fig. 1A (right) and summarized in Fig.  1, D and E (E). In the control, in the absence of norfloxacin, the DNA supercoil density was the same for both strains ( ϭ Ϫ0.077). This suggests that drug-resistant and wild-type topoi-somerase IV are functioning equally well in the absence of the drug. By 5 min after the addition of norfloxacin (Fig. 1A, lane  4), topoisomerase I alone relaxes plasmid DNA to a of Ϫ0.057 Cells were grown at 30°C and at time 0 norfloxacin (Nor) was added to a final concentration of 30 M. Samples were removed 0.5, 2, 5, 10, 20, 30, 60, and 120 min later, and plasmid pJB3.5d DNA from these cells was analyzed by electrophoresis through 1.2% agarose gels containing 2 g/ml (A and B) or 1 g/ml (C) chloroquine. Chloroquine allows separation of supercoiled topoisomers. On the left of the figure, the DNA is still (Ϫ) supercoiled. On the right, as the DNA is relaxed by topoisomerase IV, it migrates on these chloroquine gels as (ϩ) supercoiled topoisomers, which interdigitate between the (Ϫ) supercoiled bands. Autoradiographs of the Southern blots are shown. The mid-point topoisomer of plasmid relaxed to completion with calf thymus topoisomerase I migrated at the position indicated and was assigned a ϭ 0. The other values were extrapolated from this point by topoisomer band counting (59). The grid below the gels indicates enzyme activity; ϩ indicates functioning, Ϫ indicates not functioning, and Ϯ indicates partially functioning. A, following norfloxacin addition to the LZ36 (topA ϩ parC ϩ gyrA ϩ ) strain (left), relaxation is by topoisomerase I alone. Relaxation by both topoisomerase I and topoisomerase IV is seen with the LZ38 (topA ϩ parC K84 gyrA ϩ ) strain (right). B, wild-type supercoiling levels are maintained when all three topoisomerases are inhibited in the topA10 parC ϩ gyrA ϩ (LZ53, left) strain. Relaxation by topoisomerase IV alone follows drug addition to the topA10 parC K84 gyrA ϩ (LZ41) strain (right). C, wild-type supercoiling levels are also maintained when all three topoisomerases are inhibited in the topA20 parC ϩ gyrA ϩ strain (2158, left). Topoisomerase IV alone relaxes DNA to near completion in the topA20 parC K84 gyrA ϩ strain (2159, right). In this experiment 6, 30, or 60 M norfloxacin was added for 15 min, and the plasmid was pBR322. The position of negatively supercoiled DNA, for these conditions, is shown (sc). D, the graphic representation of relaxation by topoisomerase IV and topoisomerase I is shown. Following the addition of norfloxacin, DNA relaxation is shown for the LZ53 (topA10 parC ϩ gyrA ϩ ) (q), LZ38 (topA ϩ parC K84 gyrA ϩ ) (E), LZ41 (topA10 parC K84 gyrA ϩ ) (f), and LZ36 (topA ϩ parC ϩ gyrA ϩ ) (Ⅺ). The x axis is time following inhibition of gyrase; the y axis is the mean value of plasmid DNA. The data shown in Fig. 1, A and B, plus an additional experiment were plotted. Error bars represent the standard deviation of two experiments (the 2, 5, 10, 20, 30, and 60 min time points only). When error bars are not shown, the deviation was smaller than the symbols. E, the early time points from D are replotted to show more clearly the initial kinetics. and, by 120 min, further relaxes the DNA to a Ϸ Ϫ0.048 (lane 9). The relaxation rates were calculated from the slopes of the best-fit curves shown in Fig. 1, D and E. The initial rate for topoisomerase I relaxation was ϳ5 Lk min Ϫ1 . When topoisomerase IV is active along with topoisomerase I (Fig. 1A,  right), the initial rate was the same. In contrast to supercoil relaxation by topoisomerase I alone, topoisomerase IV with topoisomerase I relaxes supercoils after 5 min (lane 13), but at the considerably slower rate of 0.12 Lk min Ϫ1 . Also, the distribution of the DNA topoisomers widens. After 20 min (lane 15), the maximal relaxation ( Ϸ Ϫ0.015) was reached (Fig. 1D), and the 60 and 120 min time points were identical (not shown).
The DNA relaxation by topoisomerase IV alone is shown in Fig. 1B, right, and Fig. 1, D and E (f). The topA10 mutation causes the supercoiling level, before the addition of norfloxacin (time 0), to be more negative (lane 9) than in the topA ϩ strains in Fig. 1A. The initial rate of topoisomerase IV-mediated DNA relaxation, calculated from Fig. 1D to be ϳ0.8 Lk min Ϫ1 , is much less than the initial rate by topoisomerase I. After 15 min (Fig. 1B, lane 13), the residual rate of relaxation by topoisomerase IV was the same (0.13 Lk min Ϫ1 ) as relaxation by topoisomerase IV plus topoisomerase I (0.12 Lk min Ϫ1 ). These data show that topoisomerase IV is slower than topoisomerase I, but removes more (Ϫ) supercoils. Topoisomerase IV by itself reaches the same end point as topoisomerase IV in conjunction with topoisomerase I. Relaxation by either topoisomerase I or topoisomerase IV appears to be distributive in vivo because supercoiling changes for the entire population. Topoisomerase IV relaxation is distributive in the test tube under optimal conditions as well (36,62).
To be sure that the small residual topoisomerase I activity in the topA10 mutant was not contributing significantly to DNA relaxation and therefore the relaxation rate we measured was only from topoisomerase IV, we repeated the last experiment in a strain where topoisomerase I is disrupted by a Tn10 insertion. In this topA20 parC K84 gyrA ϩ strain, the DNA is relaxed to the same extent as in the topA10 mutant when gyrase is inhibited with norfloxacin (Fig. 1C, right side). Therefore, to-FIG. 2. Effect on DNA supercoiling of inhibiting topoisomerase IV. We measured DNA supercoiling following the selective inhibition of topoisomerase IV with norfloxacin (Nor). The experimental protocol and labeling was as in Fig. 1 except that all strains contained a drugresistant gyrase allele (gyrA L83 ). The gel was 1.2% agarose and contained 4 g/ml chloroquine. The autoradiograph of the Southern blot of the gel is shown. The values were determined relative to ϭ 0 as detailed in the legend of Fig. 1.  To test whether topoisomerase IV and topoisomerase I are the only enzymes that significantly relax DNA supercoils in E. coli, we inhibited both of these enzymes by adding the drug to topA10 parC ϩ gyrA ϩ (Fig. 1B, left, and Fig. 1, D and E (q)) and topA20 parC ϩ gyrA ϩ strains (Fig. 1C, left). Very little relaxation was seen after norfloxacin addition in either of these strains. From this result, we draw two conclusions. First, topoisomerase I and topoisomerase IV are the only important DNA relaxation enzymes in the cell. If there were another relaxation activity, it would also have to be inhibited by norfloxacin. Because topoisomerase III is not inhibited by norfloxacin (32) and the E. coli sequence does not indicate another topoisomerase (63), this is unlikely. Second, the residual topoisomerase I activity in the topA10 mutant, whereas sufficient for viability, does not perceptively relax plasmid DNA following gyrase inhibition (Fig. 1B, lanes 1-8; Fig. 1, D and E (q)).
Previous work showed that residual decatenation by topoisomerase IV took place with 30 M norfloxacin (16), but here we find that topoisomerase IV relaxation is completely blocked by this amount of the drug. We measured the K i for topoisomerase IV relaxation in vivo to be 2.5 M (data not shown). This is significantly lower than the K i of 15 M for topoisomerase IV decatenation (16). This shows that the relaxation by topoisomerase IV is more sensitive to drug than decatenation by topoisomerase IV in vivo and may indicate that decatenation is a more potent activity of topoisomerase IV than relaxation.
All of the above experiments employed norfloxacin to block gyrase or topoisomerase IV. To be sure that the drugs affected only these topoisomerases, we inhibited gyrase by a different means. We used a temperature-sensitive mutant in gyrase. Following a shift to 42°C, the DNA was relaxed by topoisomerase IV and topoisomerase I to the same extent as when gyrase was inhibited with drug (data not shown).
Inhibition of Topoisomerase IV Affects DNA Supercoiling Level-We asked whether the removal of topoisomerase IV activity affects DNA supercoiling in the presence of gyrase activity. This would be indicated by a shift to more negatively supercoiled DNA, as seen with the removal of topoisomerase I activity. In our first experiment, we blocked topoisomerase IV with norfloxacin in cells containing a drug-resistant gyrase and wild-type topoisomerase I (topA ϩ parC ϩ gyrA L83 ). Plasmid DNA was analyzed before or 1, 20, and 60 min after the addition of the drug (Fig. 2). In the absence of norfloxacin, all three enzymes are functioning, and the level of supercoiling remains the same over the time course of the experiment (Fig. 2, lanes  1 and 2). By 60 min after inhibition of topoisomerase IV with drug, the Lk shifts by Ϫ2 to ϭ Ϫ0.081 (lane 5). This shift is the same as seen in the topA10 mutant (Fig. 2, lane 6) and occurs despite the presence of fully functional topoisomerase I.

FIG. 3. Effect on DNA supercoiling of inactivating topoisomerase IV by mutation.
We measured DNA supercoiling in temperature sensitive parE ts (ParE10, lanes 4 -6) and parC ts (ParC1215) (lanes 10 -12) mutants and in isogenic wild-type parE ϩ and parC ϩ strains (W3110; lanes 1-3, and C600, lanes 7-9). Cells were grown to mid-log phase at 30°C. Aliquots were taken (time ϭ 0), and cells were shifted to the restrictive temperature, 43°C, for up to 60 min. Plasmid DNA (pBR322) was subjected to electrophoresis through a 1.2% agarose gel containing 8 g/ml chloroquine to resolve the very negatively supercoiled DNA. The ethidium bromide-stained gel is shown. The amount of plasmid DNA increased with time for the wild-type strains, reflecting cell doubling. Nicked plasmid migrates as indicated. The at time ϭ 0 was a thermal distribution around Ϫ0.075. Faster migrating bands were more negatively supercoiled as indicated.  When topoisomerase IV and topoisomerase I activities were both absent, the DNA supercoiling level for the entire plasmid population shifted an additional two or three Lk to ϭ Ϫ0.09 (Fig. 2, lanes 9 and 10). These data are summarized in Table II.
Although the specific values were not directly measured for some of the plasmids shown in Table II, the change in Lk was. The change in was then calculated from the Lk values. We carried out two experiments to determine whether the Lk shift was caused by the specific inactivation of topoisomerase IV relaxation activity, as opposed to secondary effects during the incubation with drug. First, we added 30 M norfloxacin to a strain containing both a drug-resistant topoisomerase IV and a drug-resistant gyrase, topA ϩ parC K84 gyrA L83 , and found a shift of ϳϪ1 Lk (Table III, line LZ37). Thus, at least half of the effect of the drug results from the inhibition of topoisomerase IV. However, it is possible that 30 M norfloxacin partially overcomes the drug resistance of parC K84 . Second, we inhibited topoisomerase IV by a method independent of drugs. We measured the level of DNA supercoiling in plasmids isolated from conditional lethal mutants in parC or parE at the permissive and nonpermissive temperatures (Fig. 3). At 30°C (time ϭ 0), ϭ Ϫ0.075 for all strains (Fig. 3, lanes 1, 4, and 7) except parC ts , where ϭ Ϫ0.077 (lane 10). For this mutant, 30°C is only a semipermissive temperature (15). The topoisomer distribution for the parE ϩ parental strain shifted by Ϫ2 after 20 min at 43°C as a result of heat shock (64), but the 30°C value was restored by 60 min (Fig. 3, lanes 1-3). This heat effect was more persistent in the isogenic wild-type parent for parC ts , where the ⌬Lk remained at Ϫ2 (Fig. 3, lanes 7-9). A temperature shift of the culture to 43°C for either conditional lethal allele of topoisomerase IV caused a ⌬Lk of Ϫ4 compared with the Lk at 30°C (Fig. 3, compare lanes 4 to 6 and 10 to 12; see Table III). Therefore, the removal of topoisomerase IV relaxation activity by mutation or by drug results in more negatively supercoiled DNA even when the other topoisomerases are active.
Before we could conclude, however, that topoisomerase IV contributed to the supercoiling level of DNA in cells, we had to consider a possible complication. The inhibition of topoisomerase IV by drug or mutation could secondarily induce gyrase expression, which would then result in more negatively supercoiled DNA. The transcription of gyrase genes is induced by a lowering of (Ϫ) supercoiling (18,65,66). To test this possibility, we measured the amount of gyrase using immunoblotting of whole cell extracts of the parC ts and parE ts strains and their wild-type parents. Lysates containing equal amounts of total protein were analyzed on polyacrylamide gels, transferred to nitrocellulose, and probed with antibodies to either GyrA (Fig.  4A) or GyrB (Fig. 4B) protein. Whereas the shift to 42°C caused an approximately 2-fold reduction in gyrase protein in all strains, there was no difference between the wild-type and topoisomerase IV mutant strains (Fig. 4, A and B, compare  lanes 2 to 4 and 6 to 8; Table IV). This finding rules out the possibility that the change in DNA supercoiling is caused by an increase in gyrase levels. We conclude, therefore, that topoi- FIG. 5. Topoisomer distribution of plasmid pLEU500Tc in strains with different topoisomerase activities. Norfloxacin (ϩNor) was added to inhibit the topoisomerases as indicated. Plasmid DNA was isolated from strains, and topoisomers were separated by electrophoresis through 1% agarose gels containing 2 g/ml chloroquine. Shown is the autoradiogram of the Southern blot. Nicked DNA migrated as shown. The plasmid extracted from the ⌬topA strain (DM800, lanes 1 and 6) exhibits a bimodal distribution of topoisomers. The remaining strains yielded unimodal distributions of topoisomers.  somerase IV has a direct role in setting the steady-state level of DNA supercoiling.
leu-500 Transcription-All of the above results were obtained by measuring the mean supercoil density of purified plasmid DNA. We explored next whether the supercoiling changes we measured had physiological consequences in vivo. Transcription from the leu-500 promoter is increased by (Ϫ) supercoiling in both Salmonella typhimurium and E. coli (7,67) and therefore provides a measure of the functional level of DNA supercoiling in vivo. This system has been used to analyze DNA supercoiling in ⌬topA compared with wild-type strains. Plasmids purified from ⌬topA cells have a wide distribution of topoisomers, with a fraction of hypernegatively supercoiled DNA (7,60). The level of leu-500 transcription is dependent upon the fraction of hypernegatively supercoiled DNA (7,60).
We asked how the supercoiling alterations we see after inhibiting the topoisomerases affect leu-500 transcription. To measure transcription initiating at the leu-500 promoter, we transformed our set of strains with the plasmid pLEU500Tc. This plasmid contains a complete tetA gene and leu-500 promoter (7). Strains were treated with norfloxacin (or not), and RNA was isolated. The activity of the leu-500 promoter was measured using run-off reverse transcription. A second transcript originating from the constitutive antitet promoter is detected by the same primer and provides an internal reference that is independent of supercoiling (7,67). leu-500 promoter activity is expressed as the fraction of leu-500 transcripts compared with the total transcripts detected by the primer. As previously observed, the leu-500 promoter is active in the ⌬topA strain, here two experiments averaged 38% of the total transcripts (Table V). This compares with an average background value of 9% with fully active topoisomerases. This value of ϳ10% held true for all strains with wild-type topoisomerase I. The leu-500 promoter activity was similar for the three topA10 strains that we used: topA10 parC ϩ gyrA ϩ (LZ39), topA10 parC K84 gyrA ϩ (LZ41), and topA10 parC ϩ gyrA L83 (LZ54) ( Table V). The mean ratio of leu-500 transcripts for the topA10 strains was 30%. This is a little lower than the corresponding expression in ⌬topA cells but significantly greater than that found in wild-type cells. Although hypernegatively supercoiled, the plasmids purified from the topA10 cells exhibited a topoisomer profile that is very different from that in ⌬topA cells. The topoisomer distribution of pLEU500Tc from topA10 strains is unimodal and never reaches the extreme hypernegative supercoiling level seen in the small fraction of plasmid DNA from ⌬topA cells (Fig. 5, compare lanes 1 and 2).
We next examined the effect of topoisomerase IV relaxation activity on leu-500 expression. Following gyrase inhibition, the activity of the leu-500 promoter was reduced in the norfloxacintreated topA10 parC K84 gyrA ϩ cells as the DNA was relaxed by topoisomerase IV. In four independent experiments, leu-500 promoter activity was reduced, on average, from 31 to 6% (LZ41, Table V). This shows that topoisomerase IV can relax DNA in vivo in a functionally measurable manner.
Transcription levels were already low (ϳ10%) when topoisomerase I was functional in topA ϩ cells (LZ34, Table V). The inhibition of gyrase activity by the drug had minimal additional effects, if any, in these cells, with leu-500 transcription reduced to 7% (LZ34, Table V). Under these conditions, the plasmid pLEU500Tc was relaxed by topoisomerase IV alone (Fig. 5, compare lanes 4 and 5; LZ41, Table III).
Inhibition of topoisomerase IV with norfloxacin in topA ϩ parC ϩ gyrA L83 cells did not activate the leu-500 promoter (average with and without norfloxacin ϭ 8%) even though the supercoiling of the plasmid shifted by a ⌬Lk ϭ Ϫ2 (Fig. 5, lane  9; LZ35, Table III). However, when topoisomerase IV was in-hibited in a topA10 background (topA10 parC ϩ gyrA L83 ), the activation of the leu-500 promoter was higher than with topA10 alone (from 24 to 48%; LZ54, Table V) or even with ⌬topA (38%) (DM800 , Table V). Compared with the supercoiling level when all of the topoisomerases were working (Fig. 5, lane 8), the topA10 mutation caused pLEU500Tc to shift Ϫ5 Lk (Fig. 5,  lane 10). The superhelical density became even more negative when topoisomerase IV was inhibited in the topA10 cells, shifting an additional Ϫ4 Lk (Fig. 5, lane 11; LZ54, Table III). Although hyper (Ϫ) supercoiled, the level did not reach the extreme level observed in the small fraction of hypernegatively supercoiled DNA from the ⌬topA cells (Fig. 5, compare lane 1 or  6 with 11). The Lk changes are larger for pLEU500Tc than for pJB3.5d, because pLEU500Tc contains the tetA gene and is a FIG. 6. Measurements of effective supercoiling levels ( eff ) using Int site-specific recombination in vivo. A, schematic representation for random collision site synapsis by Int. The number of supercoils trapped between the two Int sites, attB and attP, dictates the complexity of the resultant recombinant catenane product. B, relationship between Int product nodes and substrate ⌬Lk for pJB3.5d (71). C, a schematic for the recombination experiments is shown. Cells grown to mid-log at 30°C were shifted to 43°C where Int is expressed but is inactive. Upon downshift to 30°C, Int recombines the substrate plasmid pJB3.5d to make multiply interlinked catenanes. 60 M norfloxacin was added to inhibit topoisomerase IV, which normally unlinks the catenanes to release free 0.6-and 2.9-kilobase circles. D, samples were removed 5 min after downshift. Plasmid DNA was isolated, nicked with DNase I, and displayed on 1% agarose high resolution gels. Densitometric scans of the autoradiograms are shown. Top, catenane complexity when all three topoisomerases are blocked in the topA10 parC ϩ gyrA ϩ (LZ35) strain and the of the wild-type plasmid is Ϫ0.075. The mean catenane node number was 4.43, which corresponds to a eff of Ϫ0.023. Bottom, the mean catenane node number was 5.58 for the topA10 parC ϩ gyrA L83 (LZ54) strain when the of the purified DNA was Ϫ0.09. eff in this case is Ϫ0.039. P is the unlinked product; S is substrate. The ordinate axis is arbitrary PhosphorImager units. The direction of gel electrophoresis is shown. little larger. These data indicate that both topoisomerase I and topoisomerase IV play roles in controlling supercoiling during transcription of leu-500 but that topoisomerase I makes the greater contribution.
Effective DNA Supercoiling Levels in Vivo and in Vitro-Much of what we know about the structure of DNA comes from analyses using pure DNA or computational studies (68,69). However, inside the cell DNA is compacted, surrounded by small molecules, and bound to proteins, RNA, and membranes, factors not taken into account in either the structural or computational studies. The values that we have reported were determined following the removal of the cellular components that might sequester or insulate DNA supercoiling. The purified plasmids were compared with DNA standards that were relaxed to completion with calf thymus topoisomerase I in vitro. Several experimental techniques have been devised for measuring the "effective" supercoiling level in the cell ( eff ) (reviewed in Ref. 70), and it has been found that the supercoiling density of purified plasmid DNA is approximately two times eff . The origin of this difference is not fully understood, but the above mentioned factors surely play a part.
One system that provides a measure of eff is site-specific recombination by Int, which converts plectonemic supercoils between the recombination sites into a readily measurable topological property, the number of catenane or knot interlinks (Fig. 6A). There is a linear relationship between the number of interlinks and in vitro (38,71). By inhibiting topoisomerase IV and topoisomerase I, we were able to set different unimodal levels of hyper-negatively supercoiled DNA, and to test the effect on the number of links in the catenated recombination products. We then calculated eff using the previously determined standard curve (71). The experimental protocol is shown schematically in Fig. 6C. When topoisomerase IV and gyrase were inhibited with norfloxacin in the topA10 parC ϩ gyrA ϩ strain, for the purified plasmid was Ϫ0.075, and the mean catenane node number was 4.3 (Fig. 6D, top), which corresponds to eff Ϸ Ϫ0.023. For the topA10 parC ϩ gyrA L80 strain, norfloxacin addition led to a mean catenane node number of 5.58 (Fig. 6D, bottom). Whereas the purified value was Ϸ Ϫ0.09, the number of catenane nodes corresponds to eff Ϸ Ϫ0.039. These data show that changes resulting from inhibiting topoisomerase IV and topoisomerase I that are measured with purified plasmids indeed reflect effective supercoiling changes in the cell and that the overall DNA structure does not dramatically change with more negative supercoiling.

Topoisomerase IV and Topoisomerase I Counter Gyrase to
Maintain DNA Supercoiling-We conclude that topoisomerase IV plays an important role in maintaining the steady-state level of supercoiling in E. coli. This conclusion is based upon the following findings. (i) Topoisomerase IV alone relaxed DNA supercoils in vivo nearly completely when gyrase activity was removed. (ii) The selective removal of topoisomerase IV, even when topoisomerase I was fully functional, shifted plasmid DNA to more negatively supercoiled values. (iii) Removal of topoisomerase IV in a topoisomerase I mutant caused hyper (Ϫ) supercoiling. (iv) The removal of topoisomerase IV in a topoisomerase I mutant greatly stimulated leu-500 promoter activity and increased the number of supercoils trapped by Int recombination.
We propose that topoisomerase IV, topoisomerase I, and gyrase interplay to set the steady-state level of DNA supercoiling as schematized in Fig. 7. Variations of plasmid supercoil density from the wild-type level of ϳϪ0.075 following the inhibition of topoisomerase I, topoisomerase IV, and/or gyrase are shown. If gyrase activity is blocked, then either topoisomerase I or topoisomerase IV removes negative supercoils to a Ϸ Ϫ0.05. After that, only topoisomerase IV continues relaxing the DNA to a Ϸ Ϫ0.015. Removal of topoisomerase I alone or topoisomerase IV alone from cells with functional gyrase changes the mean supercoil density to ϭ Ϫ0.081. The supercoiling levels in plasmids that contained the tetA gene (pLEU500Tc and pBR322) followed the same general trend, but the removal of topoisomerase I alone altered the DNA to a more (Ϫ) supercoiled level than did the removal of topoisomerase IV alone. Removal of both enzymes had the same effect in all plasmids, resulting in a unimodal topoisomer distribution with Ϸ Ϫ0.09. Dissecting the roles of these three topoisomerases required the systematic inhibition of one, two, or all three enzymes.
Roles of Topoisomerase IV and Topoisomerase I in DNA Relaxation-Although both topoisomerase IV and topoisomerase I relax (Ϫ) supercoiled DNA, their unique roles in vivo are shown by the following findings. (i) Topoisomerase I relaxes DNA with an initial rate of 5 Lk min Ϫ1 , whereas topoisomerase IV relaxes DNA with a slower initial rate of 0.8 Lk min Ϫ1 . (ii) Topoisomerase I can relax only very negatively supercoiled FIG. 7. Steady-state levels of DNA supercoiling. The values of plasmid pJB3.5d or pGP509 DNA are shown as they deviated from the wild-type level of ϳϪ0.075 following the inhibition of topoisomerase I, topoisomerase IV, and/or gyrase. If gyrase activity is blocked, then either topoisomerase I or topoisomerase IV removes negative supercoils until Ϸ Ϫ0.05. After ϭ Ϫ0.05, topoisomerase I relaxation activity ceases, and topoisomerase IV continues relaxing the DNA to a Ϸ Ϫ0.015. The removal of either topoisomerase I or topoisomerase IV from a cell with functional gyrase shifts the DNA to a more (Ϫ) supercoiled level. In the topA10 mutant, was Ϫ0.081. Removal of topoisomerase IV activity in a wild-type topoisomerase I strain shifted the DNA supercoiling to the same level. When both topoisomerase I and topoisomerase IV are removed, shifted to Ϫ0.09. Removal of both enzymes shifted the DNA to a unimodal topoisomer distribution with Ϸ Ϫ0.09.
DNA ( Ͻ Ϫ0.05), whereas topoisomerase IV relaxes DNA nearly completely. Topoisomerase I does not relax DNA to completion in vitro (40) and topoisomerase IV does. In addition, perhaps protein, RNA, or membrane binding affects DNA accessibility, and topoisomerase IV, but not topoisomerase I, is able to dislodge these factors during relaxation. (iii) Topoisomerase I plays a more important role than topoisomerase IV in the relaxation of transcription-induced supercoiling, presumably because the supercoils are transient and topoisomerase I has a higher rate of relaxation. In the tetA-containing plasmid pLEU500Tc, the removal of topoisomerase I had a larger effect on than the removal of topoisomerase IV. A recent study has also shown that topoisomerase I is more important for controlling transcription-induced supercoiling than it is for maintaining steady-state supercoiling (72). The predominant role of topoisomerase I in controlling transcription-induced supercoiling was also shown functionally. Removal of topoisomerase I alone stimulated transcription from the leu-500 promoter, whereas the removal of topoisomerase IV alone did not (Table V). Previous work established the importance of twin supercoiled domain effects in the stimulation by (Ϫ) supercoiling of transcription from a leu-500 promoter (7,67). (iv) Topoisomerase IV, but not topoisomerase I, can relax the (ϩ) supercoils generated by replication or transcription (49). 2 DNA Supercoiling in Vivo-Taking advantage of the understanding of how the topoisomerases interplay to set DNA supercoiling levels in vivo, one can test better the effect of various levels of DNA supercoiling on physiological functions. Previously, the testable supercoiling range was only ϳ Ϫ0.08 to Ϫ0.05. The limitation was from the inadvertent inhibition by quinolones or coumarins of topoisomerase IV along with gyrase. A deletion of topA in the somewhat problematic DM800 strain caused extremely hypersupercoiled DNA ( Ͻ Ϫ0.09), but it was only for a fraction of the plasmid DNA. The rest of the plasmid DNA had a of only -0.06. One can now analyze DNA supercoiling ranging from values of Ϫ0.09 to Ϫ0.015 by utilizing a drug-resistant form of topoisomerase IV or a temperature-sensitive DNA gyrase. This supercoiling is unimodal in the range of ϭ Ϫ0.09 to ϳϪ0.05, with the topoisomer distribution broadening as approaches ϳϪ0.015. In addition, DNA relaxation is a simpler in vivo assay for topoisomerase IV activity than decatenation, the only previously known assay to test whether topoisomerase IV had, for example, become drugresistant (15,16,48).
We and others have argued previously that the essential function of topoisomerase IV in vivo is to unlink catenated intermediates of DNA replication (reviewed in Refs. 3,30). Although decatenation clearly remains a very important role, the discovery that topoisomerase IV is involved in maintaining the steadystate level of DNA supercoiling raises the question of the contribution of DNA relaxation to cell death when topoisomerase IV activity is removed. Past results on DNA supercoiling and topoisomerases should now be reexamined in light of the demonstrated expanded role for topoisomerase IV in the cell. This is particularly so for processes that were thought previously to be dependent upon topoisomerase I and gyrase alone.