Purification and Characterization of DNA Topoisomerase IV in Escherichia coli*

The subunits of topoisomerase IV (topo IV), the ParC and ParE proteins in Escherichia coli, were purified to near homogeneity from the respective overproducers. They revealed type I1 topoisomerase activity only when they were combined with each other. In the presence of Mg2+ and ATP, top0 IV was capable of relaxing a negatively or positively supercoiled plasmid DNA or converting the knotted P4 phage DNA, whether nicked or ligated, to a simple ring. However, supercoiling activity was not detected. The topoisom- erase activity was not detectable when the purified ParC and ParE proteins were combined with the purified GyrB and GyrA proteins, respectively. This is consistent with the result that neither a parC nor a parE mutation was compensated by transformation with a plasmid carrying either the gyrA or the gyrB gene. Simultaneous introduction of both the gyrA and gyrB plasmids corrected the phenotypic defect of parC and parE mutants. The results suggest that DNA gyrase can substitute for top0 IV at least in some part of the function for chromosome partitioning. Antisera were prepared against the purified ParC, ParE, GyrA, and GyrB proteins and used to investigate cellular localization of these gene products. ParC protein was found to be specifically associated with inner membranes only in the presence of DNA. This result suggests

The subunits of topoisomerase IV (topo IV), the ParC and ParE proteins in Escherichia coli, were purified to near homogeneity from the respective overproducers. They revealed type I1 topoisomerase activity only when they were combined with each other. In the presence of Mg2+ and ATP, top0 IV was capable of relaxing a negatively or positively supercoiled plasmid DNA or converting the knotted P4 phage DNA, whether nicked or ligated, to a simple ring. However, supercoiling activity was not detected. The topoisomerase activity was not detectable when the purified ParC and ParE proteins were combined with the purified GyrB and GyrA proteins, respectively. This is consistent with the result that neither a parC nor a parE mutation was compensated by transformation with a plasmid carrying either the g y r A or the g y r B gene. Simultaneous introduction of both the g y r A and g y r B plasmids corrected the phenotypic defect of parC and parE mutants. The results suggest that DNA gyrase can substitute for top0 IV at least in some part of the function for chromosome partitioning. Antisera were prepared against the purified ParC, ParE, GyrA, and GyrB proteins and used to investigate cellular localization of these gene products. ParC protein was found to be specifically associated with inner membranes only in the presence of DNA. This result suggests that one of the functions of top0 IV might be to anchor chromosomes on membranes as previously proposed for eukaryotic topoisomerase 11.
The mechanism of chromosome partitioning in Escherichia coli has been one of the most intriguing themes in the study of the bacterial cell cycle and cell division. In E. coli, the process of cell division proceeds apparently in the absence of any mitotic apparatus, such as the spindles and centrosomes found in eukaryotes, and yet replicated chromosomes are stably segregated into daughter cells. The mechanism of coupling chromosome replication with cell division probably involves chromosome-membrane interactions. Involvement of the cell surface structure in chromosome segregation was proposed by Jacob and his colleagues (1963) in their "replicon model." Since then, the attachment of chromosomes to the cell surface has been believed to be the mechanism for chro-* This work was supported by Grant-in-Aid 03261103 for scientific research on priority areas from the Ministry of Education, Science and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom reprint requests should be addressed. Tel mosome partitioning in prokaryotic cells but without a conclusive experimental result.
Investigations into the mechanism of chromosome partition have been undertaken by utilizing thermosensitive par mutants of E. coli. The par mutants are characterized by formation of large aggregated nucleoids at the restrictive temperature, and five classes of par mutations have been identified parA, parB, parC, parD, and parE (Hirota et al., 1968(Hirota et al., , 1971Hussain et al., 1987;Kat0 et al., 1988Kat0 et al., ,1990. It has been found that the Par phenotype described as parA and parD is due to mutations in the gyrB and gyrA genes, respectively (Kato et al., 1989;Hussain et al., 1987), andparB is most probably an allele of dnaG (for DNA primase) (Norris et al., 1986). On the other hand, parC and parE, both located around 65 min on the E. coli genetic map, represent new genes essential for chromosome partition. The gene products of parC and parE have considerable homology to the A and B subunits of DNA gyrase in primary structure. When crude cell lysates were prepared from ParC and ParE overproducers and combined, they showed enhanced topoisomerase activity, and the properties of ParC and ParE have suggested that ParC and ParE are components of a new topoisomerase, designated topo' IV, which possibly belongs to the class of type I1 topoisomerases (Kato et al., 1990).
DNA gyrase, a type I1 topoisomerase in E. coli, has been shown to take part in chromosome partitioning (Steck and Drlica, 1984). It is conceivable that circular replicons such as E. coli chromosomes might remain linked topologically in catenanes as observed in small plasmids (Sakakibara et al., 1976). Even eukaryotic cells with linear replicons require topoisomerases for resolving intertwined replicons. Mutants defective in a type I1 topoisomerase of Saccharomyces cereuisiae and Schizosaccharomyces pombe show abnormalities in chromosome segregation; catenated plasmids as well as unsegregated chromosomes have been observed in an S. cereuisiae top0 I1 mutant, and topoisomerases can resolve the intertwined replicons in vitro (DiNardo et al., 1984;Holm et al., 1985;Uemura andYanagida, 1984,1986). The major function of the topoisomerase in chromosome segregation may be topological resolution of intertwined replicons. Recent researches, however, have illuminated other functions of topoisomerases (Wang, 1991;Wang et al., 1990). Topoisomerases may act as anchors in a nuclear scaffold (Gasser et al., 1986;Adachi et al., 1989;Sperry et al., 1989) and contribute to chromosome condensation (Uemura et al., 1987;Adachi et al., 1991). The lack of a topoisomerase function enhances some kinds of recombination in yeast, suggesting that topoisomerases contribute to the maintenance of genome stability (Christman et al., 1988;Kim and Wang, 1989;Wallis et al., The abbreviations used are: topo, topoisomerase; m-AMSA and o-AMSA, 4'-(9-acridinylamino)methanesulfon-m-anisidide and -0-anisidide, respectively; BSA, bovine serum albumin. 25676 1989; Aguilera and Klein, 1990). The function of a topoisomerase in prokaryotes has been suggested in the interaction of DNA gyrase with the par region of plasmid pSC101, and at this par region, the plasmid is associated with a cell membrane (Gustaffson et Wahle and Kornberg, 1988). The par region is required for stable partition (topographical segregation) of replicated plasmids into daughter cells but not for decatenation (topological resolution). DNA gyrase may function to maintain superhelicity of replicons, since plasmids lacking the par region are more relaxed and unstable (Miller et al., 1990).
In order to have an insight into the functions of top0 IV, we examined the topoisomerase activity and cellular localization of this protein. Although top0 IV is homologous to DNA gyrase in amino acid sequence, top0 IV showed a relaxation activity but not the supercoiling activity of DNA gyrase. Subunits were not interchangeable between DNA gyrase and top0 IV, either in vivo or in uitro. Nevertheless, an increase in gene dose of both gyrA+ and gyrB+ compensated the phenotypic defect of parC and parE, suggesting that DNA gyrase can take over the function of top0 IV even though at lower efficiency. Both top0 IV and DNA gyrase were found in association with the inner membrane. The ParC protein revealed a unique property of DNA-dependent association with inner membranes. Prokaryotic type I1 topoisomerases, top0 IV in particular, may be involved in anchoring chromosomal DNA and forming chromosome loops as suggested in regard t o eukaryotic topoisomerase 11. EXPERIMENTAL PROCEDURES Bacterial Strains, Plasmids, and Culture Media-The E. coli strains used were: C600 (F-, thi thr leuB lacy tonA supE), DH1 (F-, recA1, endA2, gyrA96 thi-I hsdRl7 supE44) (Sambrook et al., 1989), EJ812 (parC2215, a derivative of C600) (Kato et d., 1988), W3110parEI0 (Kato et al., 1990), YN2942 (TAPlO6) (A(int-cII1)BAM N::Kan cI857 A(cro-bioA)) (Inada et al., 1989), and RW1053 (Mizuuchi et al., 1984).
Bacteria were grown routinely in LB broth and antibiotic medium 3 (Difco) broth, or on antibiotic medium 3 plates. Where relevant, antibiotics were added to 50 pg/ml (ampicillin and kanamycin) and 15 pg/ml (chloramphenicol).
Plasmid Construction-Plasmid pJK831 was constructed by inserting the Hpd parC fragment of pJK800 (Kato et d., 1988) into the EcoRV site of pACYC184. To obtain pJK2032, first the HindIII fragment containinggyrA was cloned into the HindIII site of pUC118; the resultant plasmid was digested with Sal1 and ligated with the Sal1 Km' fragment of pUC4K, and the BamHI fragment containing both gyrA and Km' was excised and inserted into the BamHI site of pACYC184.
Preparation and Manipulation of DNA-The techniques for preparation of plasmid DNA, manipulation of DNA, and transformation were as described (Sambrook et d., 1989).
Purification of ParC and ParE Proteins-All operations were performed at 4 "C. The ParC and ParE proteins were purified by detecting protein bands in SDS electrophoretic gels.
The crude lysate of the ParC overproducing strain was prepared from an 800-ml culture as described (Kato et QL, 1990) except that the pellet of ammonium sulfate fractionation was suspended in 4 ml of KP1 buffer (5 mM potassium phosphate, pH 6.8, 1 mM 2-mercaptoethanol) and dialyzed against KP1 buffer. The dialysate was centrifuged a t 15,000 rpm for 15 min, and the pellet was saved. Almost all the ParC protein was precipitated. The pellet was washed with 2 ml of KP1 buffer, suspended in 2 ml of KP2 buffer (5 mM potassium phosphate, pH 6.8, 1 mM 2-mercaptoethanol, 0.3 M NaCl, 10% glycerol), centrifuged at 15,000 rpm for 15 min, and the supernatant, in which most of the ParC protein was contained, was saved. It was supplemented with 0.2 ml of 75% glycerol and loaded onto a hydroxylapatite column (Seikagaku Kogyo Corp., Tokyo, Japan) (1.0 cm2 X 10 cm) equilibrated with KP2 buffer. The column was washed with 5 column volumes of KP3 buffer (50 mM potassium phosphate, pH 6.8, 1 mM 2-mercaptoethanol, 0.3 M NaCl, 10% glycerol), and proteins were eluted with a 200-ml gradient of potassium phosphate (50-200 mM) in KP3 buffer. The fractions containing ParC protein (50 ml) were pooled, diluted by adding a 310-ml solution of 48 mM Tris-HC1, pH 8.0, 1.2 mM EDTA, and 11.6% glycerol, and loaded onto a Q-Sepharose fast flow column (Pharmacia LKB Biotechnology Inc.) (0.5 cm2 X 5 cm) equilibrated with TE, buffer (10 mM Tris-HC1, pH 8.0, 1 mM EDTA, 1 mM 2-mercaptoethanol, 10% glycerol) containing 0.15 M NaC1. The column was washed with 10 column volumes of TE, buffer containing 0.15 M NaC1, and protein was eluted with a 200-ml gradient of NaCl (0. 15-0.25 M) in TE, buffer. The fractionated ParC protein was diluted with 1 volume of TE, buffer and concentrated by loading onto a Q-Sepharose fast flow column (Pharmacia; 0.5 cm2 X 2 cm) equilibrated with TE, buffer containing 0.15 M NaCl and eluting with TE, buffer containing 0.3 M NaC1. After addition of 0.1 volume of 75% glycerol, the concentrated fraction was loaded onto a Sephacryl S400 column (Pharmacia) 2.4 cm2 X 85 cm) equilibrated with TE, buffer containing 0.3 M NaC1, and protein was eluted with the same buffer. The ParC fractions were pooled, concentrated using the Q-Sepharose fast flow column as above, and frozen quickly in small aliquots with liquid nitrogen.
The crude lysate of the ParE overproducing strain was also prepared as described (Kato et al., 1990) except that cells from a 1.6liter culture in a suspension of 28 ml of 10 mM Tris-HCI, pH 8.0, 5 mM EDTA, 10% sucrose, and 0.6 M NaCl were lysed by addition of 1.4 ml of 10 mg/ml lysozyme. The supernatant was treated with 0.3% Polymin P, and solid ammonium sulfate was added to 40% saturation. After removing the pellet, ammonium sulfate was further added to 60% saturation, and the protein precipitate was saved. The pellet of ammonium sulfate fractionation was suspended in 4 ml of KP1 buffer and dialyzed against the same buffer. The dialysate was centrifuged at 15,000 rpm for 15 min, and the pellet was saved. ParE protein was precipitated as found in preparing ParC. The pellet was washed and extracted as in the ParC preparation. Then the fraction containing ParE protein was mixed with 0.3 ml of 75% glycerol and loaded onto a hydroxylapatite column (Seikagaku Kogyo Corp.) (1.0 cm2 X 10 cm) equilibrated with KP2 buffer. The column was washed with 5 column volumes of KP2 buffer, and proteins were eluted with a 200ml gradient of potassium phosphate (5-100 mM) in KP2 buffer. The fractions containing ParE protein (50 ml) were pooled, diluted by addition of 170 ml of 47 mM Tris-HC1, pH 8.0, 1.3 mM EDTA, and 12.9% glycerol, and loaded onto a Q-Sepharose fast flow column (Pharmacia) (0.5 cm2 X 5 cm) equilibrated with TE, buffer containing 0.15 M NaC1. The column was washed with 10 column volumes of TE, buffer containing 0.15 M NaCl, and protein was eluted with a 200-ml gradient of NaCl (0. 15-0.25 M ) in TE, buffer. The fractionated ParE protein was concentrated with a Q-Sepharose fast flow column as described for the ParC preparation. After addition of 0.1 volume of 75% glycerol, the concentrated ParE was loaded onto a Sephacryl S200 column (Pharmacia) (2.4 cm2 X 85 cm) equilibrated with TE, buffer containing 0.3 M NaCl and eluted with the same buffer. The ParE fractions were pooled, concentrated using a Q-Sepharose fast flow column as above, and frozen quickly in small aliquots with liquid nitrogen.
Purification of GyrA and GyrB Proteins-DNA gyrase A and B subunits were purified from overproducing strains as described (Mizuuchi et d., 1984) with several modifications as follows.
Assay of Topoisomerase Actiuity-The standard assay mixture (20 p l ) for topoisomerase IV contained 50 mM Tris-HCI, pH 7.5, 100 pg/ ml BSA, 10 mM MgCI,, 1 mM dithiothreitol, 0.5 mM ATP, 70 mM NaCI, and 0.55 pg of supercoiled pBR322 DNA or 0.2 pg of P4 phage DNA. The DNA gyrase activity was assayed as described above for top0 IV except that 5 mM spermidine was added, and pBR322 DNA relaxed with calf thymus top0 I (Takara Shuzo Corp., Kyoto, Japan) was used as substrate. Reaction mixtures were incubated for 1 h a t 30 "C. The reaction was stopped with 2 p1 of a solution containing 16.7% SDS and 0.017% bromphenol blue. 5 pl of the samples were electrophoresed through 0.7% agarose gels using TBE (89 mM Trisborate, 89 mM boric acid, 2 mM EDTA) buffer. The gel was stained with ethidium bromide and photographed under UV illumination.
Positively supercoiled DNA was prepared with the reverse gyrase of Sulfobbus (Kikuchi and Asai, 1984). Reverse gyrase and knotted DNA from defective heads of P4 phage were gifts from Dr. A. Kikuchi (Mitsubishi-Kasei Institute of Life Sciences, Tokyo, Japan).
Two-dimensional agarose gel electrophoresis was carried out as described (Kikuchi and Asai, 1984).
Preparation of Antisera-The ParC, ParE, GyrA, and GyrB proteins were purified as described above and further separated by preparative SDS gel electrophoresis. Gel blocks containing the purified proteins were cut off, crushed, and directly used as antigens. A rabbit was immunized by injecting 250 pg of each antigen subcutaneously, in Freund's incomplete adjuvant, and given two booster injections 2 and 4 weeks later with 125 pg of each antigen in Freund's incomplete adjuvant. The animals were bled 6 weeks after the first immunization.
Fractionation of Cell Lysates and Crude Membrane Fractions-To fractionate the cell lysate into the soluble and crude membrane fractions, E. coli C600 cells harvested in late log phase were disrupted by two passages a t 20,000 p.s.i. through a prechilled French pressure cell in T E buffer (50 mM Tris-HCI, pH 7.5, 10 mM EDTA, 1 mM 2mercaptoethanol) or T M buffer (50 mM Tris-HCI, pH 7.5, 10 mM MgC12, 1 mM 2-mercaptoethanol). DNase I, when included, was added t o a final concentration of 20 pg/ml before disruption. Unbroken cells were removed by centrifuging a t 3,000 rpm for 20 min, and the supernatant was centrifuged a t 39,000 rpm for 60 min to sediment crude membrane fractions.
To fractionate the crude membrane fraction into inner and outer membrane fractions in the presence of Mg"', the flotation gradient method was used with some modification (Ishidate et al., 1986). T M buffer was used during fractionation. The concentration of DH1 cell suspension was adjusted to 30 Am nm units, and the cells were disrupted as mentioned above. After removing unbroken cells, the lysate was layered on a two-step gradient consisting of 0.8 ml of 60% sucrose and 2.5 ml of 25% sucrose. After centrifugation in a Beckman S W 50.1 rotor a t 40,000 rpm for 3.5 h, the crude membrane fraction at the 25-60% sucrose interface was pooled. 0.6 ml of the crude membrane fraction was mixed with 1.4 ml of 67% sucrose and layered on 0.5 ml of 67% sucrose. Then 3 ml of 50%, 3 ml of 45%, 2 ml of 40%, 1 ml of 35%, and 0.5 ml of 30% sucrose were layered successively, and the sample was centrifuged a t 36,000 rpm for 72 h in a Beckman S W 41Ti rotor. Samples were collected from the tops of the tubes, and aliquots were analyzed for D-hCtate dehydrogenase activity to detect fractions containing inner membrane (Futai, 1973), for protein profiles by SDS gel electrophoresis to detect outer membrane proteins, and for ParC, GyrA, and GyrB proteins by Western blot.
The effect of DNase I treatment after fractionation was investigated by incubating the inner membrane fraction containing ParC proteins, which was isolated by flotation gradient as described above, with DNase I. The inner membrane fraction (2.9 ml) was prepared using four tubes of a Beckman SW 41Ti rotor and divided into two portions. One portion (1.4 ml) received DNase I a t a final concentration of 25 pg/ml, and the other (1.4 ml) did not. Both portions were incubated a t 4 "C for 15 h after addition of 1.1 g of sucrose and subjected to fractionation by the flotation gradient method and to sample analyses as described above.

Purification of ParC and ParE Proteins-Overproducers for
ParC and ParE proteins have been constructed, and the enhanced relaxation activity of their lysates in vitro has suggested the overproduction of the active proteins (Kato et al., 1990). The ParC and ParE proteins were separately purified to near homogeneity from each overproducer by detecting the ParC and ParE protein bands in SDS electrophoretic gels ( Fig. 1, Table I). Topoisomerase activity was detected only in the presence of both proteins (Fig. 2 A ) . A t the final purification stage, the ParC and ParE fractions eluted from the chromatographic column were assayed for the topoisomerase activity in the presence of the other purified subunit, and the amount of the ParC and ParE protein was shown to correlate with the degree of relaxation activity (data not shown). These results also confirm that top0 IV consists of the ParC and ParE proteins. When the overproducing strains were lysed with a French press or sonication, ParC protein was stable, but ParE protein was degraded, especially in the presence of Mg2+ (data not shown). The instability of the ParE protein depends on the method of lysis. ParE protein was more stable when cells were lysed by freezing and thawing (Fig. lB, lane 1). The protease which might degrade the ParE protein has not been identified. OmpT protease is not responsible because the instability was little influenced by an o m p r mutation (data not shown). Unless the ParC and ParE proteins were purified in the presence of NaCl a t a concentration higher than 0.15 M, they aggregated. Even in the buffer containing 0.3 M NaC1, these proteins may not exist as monomers, since they were eluted near the void volume in gel filtration with  Table I   Amounts of ParC and ParE proteins were estimated by quantifying and comparing the intensity of the ParC and ParE bands on Western blot membranes with that of the purified ParC and ParE proteins as standards. The total amount of protein in the cell extracts was taken as 100%. One unit of topoisomerase activity was defined as the amount of enzyme required to fully relax 0.55 pg of supercoiled pBR322 plasmid DNA under standard assay conditions. The specific activities of the final samples were 1 X lo4 units/mg for each of the ParC and ParE ParE proteins.
Sephacryl s400 and S200 (exclusion limit of 8,000 and 250 kDa, respectively). Even though the purified ParC and ParE proteins may be aggregated or polymerized, they can function as a topoisomerase as described in the next section.
Topoisomerase Activity of Top0 IV-As described in the previous work, the combined crude cell lysates of the ParC and ParE overproducers showed an enhancement in relaxation activity (Kato et al., 1990). Using the purified ParC and ParE proteins, we were able to detect the relaxation activity when both proteins were mixed in the presence of ATP and M e (Fig. 2 A ) , but supercoiling activity was not detected (data not shown). The optimal NaCl concentration for the relaxation activity was 70-80 m M (Fig. 2B). Mn2+, although less effective, was able to substitute for M$+ (Fig. 2 A ) . As shown in Fig. 2 A , especially in the absence of ATP, accumulation of possible cleavable complexes was observed even in the absence of inhibitors. The accumulation of possible cleavable complexes was also observed when lower amounts of ParC or ParE proteins were present in the reaction mixture (Fig. 2C). The result of a titration experiment with ParC and ParE proteins suggests that the molar ratio of the ParC to the ParE protein in top0 IV might be 1:l (Fig. 2C).
In order to know whether top0 IV can relax positively supercoiled DNA, top0 IV was incubated with the positively supercoiled DNA prepared with reverse gyrase, and the product was analyzed by two-dimensional gel electrophoresis. Positively supercoiled DNA can be clearly distinguished from negatively supercoiled DNA by two-dimensional gel electrophoresis; in the second dimension, positively supercoiled DNA has a higher mobility due to the increase in the superhelical density by ethidium bromide, while the migration of negatively supercoiled DNA was retarded because of relaxation by ethidium bromide. (In the presence of a rather low concentration of ethidium bromide, since highly negative supercoils run as fast as highly positive supercoils (Fig. 2 0 , (ii), lanes 1 and 41, only moderately negative supercoils can be distinguished from positive supercoils). The result clearly showed that top0 IV could also relax positively supercoiled DNA (Fig. 20). In E. coli, top0 IV is the first example of an enzyme that can relax positively supercoiled DNA, since bacterial top0 I and intact DNA gyrase cannot (Gellert, 1981;Wang, 1985). DNA gyrase has no activity for relaxing positively supercoiled DNA in the absence of ATP or P,y-imido-ATP but can convert positively supercoiled DNA to negatively supercoiled DNA by supercoiling activity in the presence of ATP (Gellert, 1981).
Unknotting activity was investigated using knotted P4 phage DNA as substrate. DNA extracted from a P4 mutant phage is knotted and nicked in both strands, and both type I and I1 topoisomerases can convert it to an unknotted circle, although P4 phage DNA ligated in vitro can be unknotted only by type I1 topoisomerases. As shown in Fig. 2E, top0 IV can untangle the knotted DNA, either nicked or ligated; in this figure the knotted P4 DNA appears as a smear, while the unknotted phage DNA is detected as a sharp band in agarose gel electrophoresis. To confirm the structure of the substrate DNA, the nicked and the ligated DNA were incubated at 75 "C for 5 min. After the heat treatment, the ligated phage DNA was not changed in electrophoretic appearance, while the nicked DNA was converted to a simple ring to form a single distinct band. The unknotting activity, as well as the relaxation activity, of top0 IV required both subunits, ATP and M$+ (data not shown). Fig. 2 F shows the effects of inhibitors. The inhibitors of DNA gyrase, oxolinic acid and novobiocin, inhibited the relaxation activity of top0 IV a t almost the same concentration as for DNA gyrase. (The detailed analysis of inhibition of top0 IV activity by inhibitors will be described elsewhere.) This result is consistent with the fact that the homology in amino acid sequence is higher between top0 IV and DNA gyrase than between top0 IV and eukaryotic type I1 topoisomerases (Kato et al., 1990).' It is also consistent with the result that the inhibitors of eukaryotic top0 11, m-AMSA and VP16, did not inhibit the relaxation activity of top0 IV (Fig.  2F, (ii)). However, interestingly, o-AMSA, the isomer of m-AMSA, which inhibits eukaryotic top0 11 only slightly, inhibited the top0 IV (Fig. 2F, (ii)).
Compensation of the parC and parE Mutations by Increasing Gene Dosage of Both gyrA and gyrB-The amino acid sequence homology between top0 IV and DNA gyrase suggested the possibility that the subunits of top0 IV might be substituted by the subunits of DNA gyrase. To test this possibility, topoisomerase activities were investigated in uitro for relaxation, supercoiling, and unknotting in combinations of ParC and GyrB or ParE and GyrA. No activity was detected, at least under the reaction conditions for DNA gyrase or top0 IV (data not shown), suggesting that the subunits of top0 IV and DNA gyrase are not interchangeable.
The results in vitro are consistent with those obtained in J. Kato " + " + + + + + " -+ + + " + * " " " + + " -+ + + + + + + + -  1-7, respectively). D, relaxation of negatively and positively supercoiled pBR322 DNA. Negatively ((i), lanes 2-4 and (ii), lanes  1-3) and positively ((i), lanes 5 and 6 and (ii), lanes 4 and 5 ) supercoiled DNA were used as substrate for the relaxation assay of vivo. The temperature-sensitive growth of parC or parE mutants was not corrected when these mutants were transformed with either a gyrA or a gyrB plasmid; nor was the parA (gyrB) defect compensated by the introduction of either a parC or a parE plasmid into a parA (gyrE) mutant (data not shown). Interestingly, simultaneous introduction of gyrA and gyrB plasmids into the parC or parE mutants corrected the temperature-sensitive growth of the mutants, though compensation was not complete in comparison with the Complementation by parC or parE plasmids; the colony size of the parC or parE mutants carrying both gyrA and gyrB plasmids was somewhat smaller than that of the mutants carrying parC or ParE plasmids at the nonpermissive temperature (data not shown). The parA (gyrE) mutation was not compensated even in the presence of both parC and parE plasmids. The results suggested that at least a part of top0 IV function could be substituted by DNA gyrase but that some of the functions of DNA gyrase could not be replaced by those of top0 IV.
Interaction of ParC with Inner Membranes-In the previous work, we found that the ParC protein synthesized in minicells was detected in crude membrane fractions but only when minicells were fractionated in the presence of Mg2+ (Kato et al., 1988). Since in minicells, ParC protein was overproduced due to a multicopy parC plasmid, this result could have been attributable to the overproduction of ParC protein. To eliminate such a possibility, the localization of the subunits of top0 IV and DNA gyrase was investigated in a strain in which the cellular level of these proteins should be normal. E. coli C600 was disrupted by two passages through a French pressure cell, and the crude cell lysate was fractionated into soluble and crude membrane fractions by centrifugation in the presence of Mg2+ or EDTA. The subunits were detected by Western blotting using antisera prepared against the purified ParC, top0 IV. (200 ng each of the ParC and ParE proteins were contained in the reaction mixtures.) Positively supercoiled DNA was prepared with reverse gyrase (Nakasu and Kikuchi, 1985). The same samples were examined by one-dimensional gel electrophoresis, the samples were electrophoresed first in the vertical direction (from top to bottom) in the standard condition and then reelectrophoresed in the horizontal direction (from left to right) in the same buffer containing  1-4) and the knotted P4 DNA that was ligated in vitro (lanes 5-8) were used for the unknotting assay. (100 ng each of the ParC and ParE proteins were contained in the reaction mixtures.) Lanes I and 5, no enzyme; lanes 2 and 6, incubated with no enzyme a t 75 "C for 5 min; lanes 3 and 7, incubated with top0 IV a t 37 "C for 60 min; lanes 4 and 8, digested with EcoRI; lane 9, X-phage DNA digested with HindIII. The nicked and knotted DNA was converted to simple nicked rings by heat treatment (75 "C for 5 min), but the ligated and knotted DNA was not. F, effect of inhibitors on top0 IV-catalyzed relaxation of negatively supercoiled pBR322. Negatively ParE, GyrA, and GyrB proteins. When the amounts of the subunits were estimated by quantifying the intensity of the bands on Western blot membranes, in the presence of EDTA, almost all of the subunits of top0 IV and DNA gyrase were detected in the soluble fractions; in contrast, in the presence of M e , about half of both GyrA and GyrB and almost all of ParC were localized to the membrane fraction. As described in the previous section, ParE protein was degraded in the presence of M e so that it was difficult to study its cellular localization, although the localization of the degraded protein appeared almost the same as that of ParC protein (Fig. 3).
In order to know whether the ParC protein in the crude membrane fraction was really associated with the membrane or not, the crude membrane was further fractionated. Because in the presence of M e the crude membrane fraction contains a lot of ribosomes, we adopted the flotation gradient method t o fractionate the crude membrane (Ishidate et al., 1986). In the flotation gradient method, a crude membrane fraction is placed near the bottom under the sucrose gradient set up stepwise and centrifuged as described under "Experimental Procedures." By this procedure, only membrane fractions were floated, leaving ribosomal fractions at the starting position; as a result, the inner and outer membrane fractions were separated and detected without interference by the ribosomal fraction. The crude membrane fraction subjected to the flotation gradient was prepared by sedimenting through 25% sucrose cushioned with 60% sucrose on which crude membranes were sedimented in a layer. As described in Fig. 4, the crude membranes were fractionated into two major peaks of membranes by flotation through the sucrose gradient. Distribution of the activity of a marker enzyme for the inner membrane, D-laCtate dehydrogenase, and the electrophoretic profile of the outer membrane proteins suggested that the upper peak contained mainly inner membranes and that a major portion of the outer membrane was found in the lower peak. Using the antiserum against ParC protein, we studied the distribution of the ParC protein. T o our surprise, very different results were obtained when fractionation was carried out in the presence or absence of DNase I. The majority of ParC protein remained at the starting position, but not in the floated fractions, in the presence of DNase I (Fig. 4A), while 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 GyrA  I (lanes 9-12). The membrane fractions were suspended in 5 ml of TE buffer. Proteins (5 p1 of each fraction) were separated by SDS gel electrophoresis, GyrA, GyrB, ParC, and ParE proteins were detected by Western blotting with antisera against these proteins, and the amounts of the proteins were estimated by quantifying the intensity of the bands. Only the bands of these subunits are shown; there were almost no other clear bands, suggesting that there was apparently no cross-reaction. Lanes 1, 5,  and 9, cells before disruption; lanes 2, 6, and 10, crude cell lysates  after disruption; lanes 3, 7, and 11, soluble fractions; and lanes 4 , 8,  and 12, crude membrane fractions.
the ParC protein was detected in the inner membrane fraction in the absence of DNase I (Fig. 4B). This result suggests that the ParC protein is associated with the inner membrane only in the absence of DNase I, i.e. in the presence of DNA. The distribution of the GyrA and GyrB proteins, about half of which were localized to the crude membrane fraction in the presence of M e (see above), was not affected by DNase, i.e. similar amounts of these proteins were found in the inner membrane fraction regardless of DNase I (data not shown). The result suggests that most of the gyrase subunits found in the crude membrane fraction, which account for about half of the total cellular subunits, are associated with the inner membrane. The existence of ParC in the inner membrane fraction only in the absence of DNase I suggests that the association of ParC with the inner membrane requires the participation of DNA. T o confirm the role of DNA in the interaction between ParC protein and the inner membrane, a further analysis was carried out. The fractionated inner membrane fraction containing ParC protein was divided into two portions, and DNase I was added to one of them. The two portions were then subjected to a second flotation gradient fractionation. As shown in Fig. 5, with the DNase I-treated sample, about half of the ParC was left at the startingposition, while almost all of the ParC of the other sample, to which DNase I was not added, was again recovered in the inner membrane fraction. Vesicle formation of the inner membrane may be the reason why only half of the ParC protein was dissociated from the inner membrane. These results suggest that DNA is contained in the ParC inner membrane fractions and that DNA is necessary for the association of ParC with the inner membrane. DISCUSSION In E. coli, two genes, parC and parE, have been found to code for a new type I1 topoisomerase, top0 IV. Their amino acid sequences are homologous to those of the GyrA and GyrB subunits of DNA gyrase, respectively, and an enhancement in relaxation activity was detected when crude cell lysates prepared from their overproducers were mixed (Kato et al., 1990). Therefore, there are two type I1 topoisomerases in E. coli, and both of these enzymes are essential for chromosome segregation, because the top0 IV mutants showed the Par phenotype at the nonpermissive temperature as observed in some DNA gyrase mutants. Purification and characterization of top0 IV were necessary to uncover the function of top0 IV and to answer the question why more than one type I1 topoisomerase is needed for chromosome segregation.
The subunits of top0 IV, ParC and ParE proteins, were purified from the overproducers for these subunits, and the purified proteins were shown to have topoisomerase activity. The coincidence of distribution of the topoisomerase activity with that of ParC or ParE protein in chromatographic fractionation confirmed that ParC and ParE proteins constitute top0 IV and that these subunits are the minimum components required to form a complex having topoisomerase activity. Both ParC and ParE proteins were found to become insoluble under low ionic strength conditions, and the purified proteins may aggregate or polymerize, even in the presence of a rather high concentration of NaCl, as suggested by the elution pattern on gel filtration. This feature might be an artifact due to overproduction, or it may be inherent in the nature in these proteins. In the presence of EDTA, most of the ParC protein of a cell that overproduced neither of them was found to be in a soluble fraction. However, when the soluble fraction was further fractionated by centrifugation through a sucrose gradient, ParC was detected a t a slightly lower position than most of the other soluble proteins (data not shown). DNA gyrase is thought to function as a tetramer containing two GyrA and two GyrB subunits, and further experiments would be necessary to clarify the composition of top0 IV in uiuo.
ParE protein was found to be degraded when cells were drastically disrupted by French press or sonication. The degradation might be due to a protease, because sonication of the purified ParE proteins did not cause degradation i n uitro (data not shown). In E. coli, an enzyme named top0 11', composed of GyrA and a small protein fragment derived from GyrB, has been purified. Both GyrB and ParE might have some common structures susceptible to a preferential attack by proteases. The physiological role of the instability, if any, remains unknown. Purified top0 IV was found to have relaxation activity but not supercoiling activity i n uitro, though the amino acid sequence was homologous to that of DNA gyrase. Top0 IV as well as top0 I might function in relaxation i n uiuo, in contrast to the action of DNA gyrase. A defect of top0 I was compensated by increasing the gene dose of both parC and parE (Kato et al., 1990). One of the suppressor mutations of topA (topo I), toc, which was reported to be associated with ampli- Procedures." A, fractionation in the presence of DNase I. DNase I was added before disruption with a French press (see "Experimental Procedures"). After centrifugation, samples were collected from top to bottom of the tube (corresponding to left to right in the panels).
(i), absorbance a t 280 nm; (ii), the activity of D-IaCtate dehydrogenase (a marker of inner membrane); (iii), protein profiles by SDS gel electrophoresis; (iu), the ParC proteins were detected using an anti-ParC serum by Western blot. R, fractionation in the absence of DNase I. Fractionation, sampling, measurement of absorbance a t 280 nm (i), and D-lactate dehydrogenase activity (ii), analysis of the protein profiles (iii), and detection of the ParC protein (iu) were carried out as in A except that DNase I was not added during fractionation.
fication of the tolC region near parC and parE genes, may be due to increase in gene dosage of parC and parE (Dorman et al., 1989). However, parC and parE mutations were compensated by increasing the gene dose of both gyrA and gyrB genes as described in the text. This result suggests that at least a part of top0 IV functions can be substituted by DNA gyrase. Purified top0 IV showed unknotting activity, as does DNA gyrase. Therefore, compensation of top0 IV mutation by DNA gyrase might reflect their roles in unknotting, decatenation, and resolution of intertwined chromosomes. These could be the main functions of top0 IV, since top0 IV shows rather high unknotting activity in uitro in comparison with the activity of DNA gyrase (data not shown). On the other hand, the parA (gyrB) mutation was not compensated by transformation with both parC and parE plasmids. Supercoiling activity is unique to DNA gyrase, while both top0 I and IV can only relax supercoiled DNA. The parA mutation may also cause a decrease in negative superhelicity, which cannot be compensated by top0 IV. If the function of top0 IV parallels that of DNA gyrase in decatenating the replicated chromosome, it is not understood why aparC or parE single mutation causes temperature-sensitive growth. Several explanations are possible. The decatenating function might be essentially the same between top0 IV and DNA gyrase. Then, the matter would simply depend on a quantitative problem of decatenase activity. If DNA gyrase were engaged in both resolution of daughter chromosomes and maintenance of the negative superhelicity, a cell would have to be provided with quite a high level of DNA gyrase, which might in turn disturb a balance of superhelical density. To solve such a problem, the cell seems to have developed the top0 IV system that takes over the decatenation function. Alternatively, the substrate specificity of DNA gyrase and top0 IV may be different. Top0 IV might have been so evolved as to recognize some unique topological states of DNA in folded chromosomes. DNA gyrase may be versatile in function but rather inefficient for resolving catenated chromosomes in uiuo, so that a higher level of DNA gyrase may be required for chromosome decatenation, which top0 IV would efficiently fulfill. Some functions of top0 IV might remain in the parC and parE mutants, and the functions could not be substituted by DNA gyrase, although other functions of top0 IV, which are defective in the mutants, could be compensated by DNA gyrase as shown in the text. DNA gyrase and top0 IV are probably differentiated in function in uiuo. DNA gyrase may be engaged mainly in supercoiling DNA, while top0 IV works in resolving chromosome catenanes. Nevertheless, DNA gyrase is indispensable for chromosome segregation, because the gyrA (parD) or the gyrB (parA) single mutation results in a Par phenotype as in top0 IV mutations. The Par phenotype caused by the defect of DNA gyrase might not be due to simple lack of decatenating and resolving activity but to a secondary effect of decrease in superhelicity.
Top0 IV was inhibited by inhibitors for DNA gyrase, novobiocin and oxolinic acid, at almost the same concentration as that effective for DNA gyrase (Fig. 2F, ( i ) ) . These results lead to the question of why the parC and parE mutants have not been isolated as resistant mutants against the DNA gyrase inhibitors so far. The reason is thought to be that the functions of top0 IV, relaxation and unknotting (decatenation), are shared with top0 I and DNA gyrase (and possibly top0 111) and that a partial defect of top0 IV might be compensated by other topoisomerases as shown in the text. However, supercoiling activity is unique to DNA gyrase. Nearly complete inhibition of top0 IV, as caused by the par@" and parE" mutations, might be necessary for cell killing, although partial inhibition of DNA gyrase might lead to cell death. Much higher concentration of the inhibitors might be needed for the lethal inhibition of top0 IV than for that of DNA gyrase. A more precise analysis of the effects of inhibitors is necessary and is in progress.
Immunological studies of the localization of top0 IV and DNA gyrase uncovered the interesting feature of interaction with the inner membrane. The ParC protein, especially, was suggested to interact with the membrane only in the presence of DNA. The ParC-inner membrane interaction may not be a nonspecific interaction, in which the ParC protein merely binds to nonspecific DNA and the DNA is associated with the membrane, because major DNA fraction was recovered from the soluble fraction but not from the membrane fraction after disruption of cells with a French press at rather high pressure used in this experiment. The ParC protein might anchor chromosomes to the inner membrane as eukaryotic top0 I1 is thought to anchor chromosomes to nuclear scaffolds. Since a mutant defective in the ParC-DNA-inner membrane interaction has not been isolated, the function involved in the interaction has not yet been clarified as is the case with eukaryotic top0 11. Among the parC mutants isolated so far, there may be mutants defective in the ParC-DNA-inner membrane interaction; we are now characterizing the mutants. In eukaryotes, chromosomes attach to the nuclear scaffold via specific DNA regions named SAR or MAR; by analogy, the ParC-inner membrane interaction might need specific DNA regions. The amount of ParC protein in a cell was estimated using antiserum and was almost the same as those of GyrA and GyrB according to the results of Gellert (1981) (about 500-1000 molecules/cell for each subunit; the result was referred to by Yang and Ames (1988)). There may be a lot of possible specific sites like SAR. One of the candidates is the family of REP sequences, which were reported to be specific binding sites of DNA gyrase (Yang and Ames, 1988). However, the specific interaction between DNA gyrase and repetitive extragenic palindromic sequences was not confirmed by the results of other groups (Higgins et al., 1988;Gilson et al., 1990). Other specific sequences or specific structures in chromosomes may be responsible for binding of top0 IV and DNA gyrase. Identification of the DNA regions and membrane component(s) that are necessary for interaction between type I1 topoisomerases, DNA gyrase and top0 IV, and the inner membrane is essential for understanding the function of type I1 topoisomerases in uiuo.