Modulation of DNA Supercoiling Activity of Escherichia coli DNA Gyrase by F Plasmid Proteins

The letA (ccdA) and letD (ccdB) genes of F plasmid contribute to stable maintenance of the plasmid in Escherichia coli cells; a product of the latter has a lethal effect on the host cell and that of the former neutralizes functions of the letD. In cells that overpro- duce the LetD (CcdB) protein, the plasmid DNA is extensively relaxed. Correspondingly, DNA supercoiling activity in a cell-free extract of the overproducing strain decreases to a level of less than 1% of that seen in normal cells. However, the extract does not inhibit DNA gyrase reconstituted from purified subunits, thereby indicating that the intrinsic DNA gyrase is inactivated in the overproducing strain. Upon addition of purified LetA (CcdA) protein to the extract of LetD overproducing cells, the DNA supercoiling activity was fully restored. Using this rejuvenation as an assay, we purified the “inactivated gyrase” and obtained evi- dence that the LetD protein formed an isolable complex with the A subunit of DNA gyrase. Thus, the LetD and the LetA proteins constitute an opposing pair in modulating the DNA supercoiling activity of gyrase, prob- ably by direct interaction. The ccd operon of F plasmid, located in the primary replicon region (repFIA), consists of letA (ccdA), letD (ccdB), and resD and structure-function relation-ships

segregant model) (5, 6). However, the molecular basis for the inhibition of host cell growth by LetD protein and the suppression by LetA protein is poorly understood.
The LetD protein induces SOS functions although the killing activity of the protein is exhibited in both recA+ and recA-bacteria (4, 7-9). From studies of this LetD-mediated SOS induction, Bailone et al. (10) speculated that the SOS signal resides on the host chromosome rather than on the F DNA and further postulated that the LetD protein might inhibit chromosome decatenation by DNA gyrase, leading to generation of single-stranded DNA by the action of RecBCD helicase (10,ll). Phenotypic similarities between some gyrase mutants and the letD-expressing cells support this notion. E.
coli par mutants (12) show basically the same nucleoid morphology and filamentous cell growth as noted above, under restrictive conditions. The Par phenotype of parA and parD mutants have been ascribed to gyrB(,,) and gyrA(,,, mutations, respectively (13, 14). The mutants are considered to have lost control over the spatial location of septa, probably as a result of a primary defect at separation of the intertwined chromosomes after DNA replication (15). It may be that the LetD protein hampers cellular segregative machinery in which the par genes are involved. In attempts to obtain clues for target protein(s) of the LetD protein, mutants that are tolerant to the inhibitory effect of LetD protein (tld mutants, tolerance to letD product growth inhibition) have been isolated (16). We found one such mutant (tldC15) in gyrA, a structural gene for the A subunit of DNA gyrase (17). The mutant was able to form colonies under LetA-LetD+ conditions at 28 "C and the mutation itself rendered the cell temperature sensitive. The mutant GyrA protein was transdominant over wild type GyrA protein for LetD tolerance. In addition, increased dosage of the wild type gyrA gene overcame the growth inhibition of the letD gene product. These observations suggested that action of the LetD protein is on DNA gyrase.
We report here that the target of LetD protein is DNA gyrase, particularly the A subunit. In the LetD overproducing cells, DNA gyrase (AzB2), as well as a free form of the A subunit, which was shown to be present in 20-fold excess over the A2Bz tetramer, existed in an inactivated form. This inactivation is likely due to binding of the LetD protein to the A subunit. The LetA protein is able to rejuvenate the inactivated gyrase or the A subunit.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-The E. coli strain KP3998 (lacP) has been described elsewhere (18). Plasmid pMJR1560, a source of a lacP fragment, was kindly provided by M. Imai, Kyoto University. pKP1498 (18) contains a fragment carrying letA gene inserted into BamHI and HincII sites of a tac promoter-mediated expression vector pKP1500 (18). pKP1052, a pBR322 (19) derivative carrying the EcoRI fragment G (Spc') of NRI (20), was constructed in our laboratory. pKP1444, a pUC9 (21) derivative constructed in our laboratory,' contains a fragment carrying the Spc' gene derived from pKP1052 and a BamHI-PstI fragment carrying letA-ktD genes of F13-1 (42.84-43.6F) (4). A new EcoRI site was introduced just upstream of the initiation codon of the letD gene, using oligonucleotide-directed sitespecific mutagenesis. pKPlllO is a pUC9 derivative carrying the EcoRI fragment G of NRI.
Reagents and Proteins-Sources were: IPTG? ATP, creatine phosphokinase (rabbit muscle, type I), phosphocreatine (di-Tris salt), and E. coli tRNA, Sigma; spermidine, Wako; bovine serum albumin (Fraction V), Boehringer Mannheim; protein standards for SDS-PAGE (high and low molecular weight range packages), and topoisomerase I (calf thymus), Bethesda Research Laboratories; proteinase K, E. Merck; agarose, T4 DNA ligase, and restriction enzymes, EcoRI, HindIII, PstI, and BamHI, Takara; gel filtration calibration kit, Pharmacia LKB Biotechnology Inc.; Vectastain Elite ABC kit (avidin DH, biotinylated horseradish peroxidase H, biotinylated anti-rabbit IgG) and substrate kit (3,3'-diaminobenzidine, 4-chloro-l-naphthol), Vector Laboratories; E. coli GyrA, GyrB, and topoisomerase I proteins, gifts from K. Sekimizu, University of Tokyo; rabbit antiserum against GyrA protein, a gift from J. Kato, University of Tokyo; rabbit antiserum against LetA-LetD complex, prepared by Takara custom services; LetA and LetD proteins, purified in our laboratory3; LetA-LetD complex, purified in our laboratory: The LetA, LetD, and LetA-LetD complex proteins were purified to apparent homogeneity from their overproducing strains by following the proteins by SDS-PAGE (detailed procedures will be described elsewhere). N-terminal amino acid sequences of the purified proteins were determined and they agreed with those deduced from the DNA sequences of the corresponding genes.
Construction of LetD Overproducing Plasmid-Three steps were taken to construct a LetD overproducing plasmid (i) construction of a LetD-LetA overproducing plasmid (ii) introduction of the lacP gene into the plasmid; and (iii) deletion of the letA gene. First, the letD gene flanked by EcoRI and BamHI sites in pKP1444 was cloned into multicloning sites of pKP1498 to yield pKP1656, so that the letD and letA genes were located in this order just downstream of the tac promoter and the Shine-Dalgarno sequence of the lac2 gene. Second, EcoRI fragment G of NRI (Spc*) in pKP1052 was inserted into an EcoRI site of pMJR1560 in order to provide a selective marker for the lacF fragment. From the resulting plasmid (pKP1874), a HindIII fragment carrying both lac19 and Spc' genes was cut out and inserted into a HindIII site in pKP1656 located just downstream of the letA gene, and pKP1876 was obtained. Third, a PstI fragment carrying the letA gene was deleted from the plasmid (pKP1876) that has two PstI sites, one in the multicloning site at the junction of the letD and letA genes and the other in the downstream of the ktA gene. The final construct, pKP1878, was not successfully introduced into E. coli strain KP3998 (lacP). Instead, the plasmid was introduced into KP3998 harboring F'lac (KP5621) and the strain (KP5621/pKP1878) was used to overproduce the LetD protein.
A control plasmid lacking the letD gene was constructed by inserting a HindIII fragment (lac19 and Spc') of pKP1874 into a HindIII site of pKP1500 in the same orientation as in pKP1878. The resulting plasmid pKP1879 was introduced into KP5621 to yield a control strain (KP5621/pKP1879).
Cell Growth and Preparation of Cell Extracts-Since the growth of E. coli cells was affected by a very low level of production of the LetD protein even in the presence of F'lac, plasmid pKP1878 was easily lost from the cells during growth, resulting in a poor production of LetD protein. To overcome this problem, E. coli KP5621/pKP1878 was first grown on nutrient agar plates containing ampicillin (100 pg/ml) and spectinomycin (50 pg/ml) at 28 "C. Cells forming colonies were scraped off with a sterile spatula and inoculated into 450 ml of PYGNM medium (1% polypeptone, 0.5% yeast extract, 0.5% glycerol, 0.5% NaC1, 0.01% MgSO,) in a 2-liter flask so as to adjust an initial AM of the cell suspension to be 0.2. The cells were grown at 28 "C to A695 of 0.6, at which point the temperature was shifted to 40 "C. After ' T. Orita and T. Miki 1 h, IPTG was added to the culture to a final concentration of 2 mM, and the cells were left to grow a t 40 "C for another 0.5, 1.0, and 1.5 h, depending on the experiment. The cells were harvested by centrifugation, suspended to A5g5 of 400 in 50 mM Tris-HCl (pH 7.5) containing 10% sucrose, and frozen immediately in liquid nitrogen. KP5621/pKP1879 was grown under the same conditions, except that the cells were inoculated in much smaller size (initial A695 of 0.05) and shifted to 40 'C a t A595 of 0.3 in order to adjust the final A595 to that of KP5621/pKP1878. Cell lysates were prepared using lysozyme and heat treatment (37 "C) in the presence of spermidine as described (22). Topological Analysis of Isolated Plasmid-After KP5621/pKP1878 and KP5621/pKP1879 were grown under the conditions specified, they were chilled rapidly by mixing with an equal volume of frozen medium, and plasmid DNAs (pKP1878 and pKP1879) were isolated by the alkaline lysis method with proteinase K treatment (23). The DNAs were electrophoresed on a 0.8% agarose gel (14 X 12.5 X 0.69 cm) in Tris borate/EDTA buffer (23) at 1.6 V/cm for 14 h. The gel was stained with 1 pg/ml ethidium bromide and photographed under ultraviolet light on Polaroid type 55 film.
I n Vitro DNA Supercoiling Assay-The assay measures conversion of the relaxed closed-circular form of pKPlllO DNA to the supercoiled form, as demonstrated by agarose gel electrophoresis. The substrate DNA was prepared by treating the supercoiled form of the DNA with calf thymus topoisomerase I according to the manufacturer's instruction. The activities of gyrase (AzB2), GyrA, and GyrB in cell-free extracts were selectively measured by addition of none, excess purified GyrB, and excess purified GyrA, respectively. The purified GyrA and GyrB proteins were free of contaminating GyrB and GyrA activities, respectively. The standard reaction mixture (19.5 pl) contained 35 mM Tris-HC1 (pH 7.5), 6.5% (w/v) glycerol, 0.14 mM EDTA, 4 mM MgC12, 24 mM KCl, 1.8 mM spermidine-HCl, 1.4 mM ATP, 0.1 mg/ml creatine kinase, 8 mM creatine phosphate, 90 pg/ml E. coli tRNA, 0.36 mg/ml bovine serum albumin, 5 mM dithiothreitol, 0.24 pg (43 fmol of DNA circle) of the relaxed closedcircular DNA, and enzyme fractions to be assayed. The components above, except for the enzyme fractions, were combined in a total volume of 13.5 pl (premix) and prewarmed at 25 "C. The enzyme mixture (6 pl) containing cell lysates, GyrA, or GyrB, was preincubated at 25 "C for 5 min. Reaction was initiated by addition of the premix into the enzyme mixture, left to proceed at 25 'C for 60 min, and terminated by adding EDTA (10 mM) and SDS (0.5%). The product DNA was electrophoresed and the gel was photographed as above. One unit of DNA supercoiling activity is defined as converting 50% of the DNA in the standard assay to supercoiled species which migrate at the fully supercoiled position, as measured by a densitometric scan (2202 UltroScan laser densitometer, LKB) of negatives of the photographs (24). When the activities were too low to be able to generate the supercoiled species within the assayable range, units were determined by comparing the patterns of DNA on agarose gels with those produced by activity of known units. Enzymes were diluted when necessary into 50 mM Tris-HC1 (pH 7.5), 50% (w/v) glycerol, 1 mM EDTA, 50 mM KCl, 3.6 mg/ml bovine serum albumin, and 5 mM dithiothreitol.
When an ATP-regenerating system consisting of creatine kinase and creatine phosphate was not included in the assay, activity that inhibited DNA gyrase was detected in E. coli cell extracts. We purified this activity to apparent homogeneity and it proved to be glycerol kinase (EC 2.7.1.30) encoded by the glpK gene located at 88 min on the E. coli genome (25). The enzyme catalyzes phosphorylation of glycerol which in turn generates ADP from ATP. We found this ADP production to be inhibitory to the supercoiling activity of DNA gyrase. Levels of glycerol kinase activity were the same in the LetD overproducing strain and the control strain.
I n Vitro Restoration of the DNA Supercoiling Activity by LetA Protein-The assay was carried out essentially as the DNA supercoiling reaction, except that the LetA protein was included in the enzyme mixture. In order to determine units of the restored supercoiling activity, the LetA protein was titrated into reactions containing a fixed amount of a cell extract of the LetD overproducing strain.
A series of such titrations was carried out with several different amounts of the cell extract and the reactions showing the maximal restoration for each series were employed for calculation of the units, as described above.
Partial Purification of LetA . GyrB-dependent Supercoiling Actiuity-The DNA supercoiling activity that appeared with addition of LetA and GyrB proteins was purified, using the rejuvenation assay. The cell extract (18.5 ml, 233 mg of protein) of KP5621/pKP1878, grown under full induction conditions, was treated with ammonium sulfate (0.35 mg/ml of cell extract) and the precipitate was collected by centrifugation (31,000 X g, 0 "C, 60 min). The suspension of the pellet (4 ml. 64 mg of protein) in buffer A (50 mM Tris-HCI (pH 7.5), 15% (v/v) glycerol, 0.1 mM EDTA, 5 mM dithiothreitol) was dialyzed against the same buffer and loaded onto a 3-ml DEAE-trisacryl (type M, IBF Riotechnics) column equilihrated with huffer A + 50 mM KCI. The column was washed with 5-column volumes of the equilihration buffer and the activity was step eluted by 5-column volumes of huffer A + 300 mM KCI. A portion of the active peak fraction (100 PI, 0.8 mg of protein) was filtered through a FPLC-Superose 12 column (25 ml) equilihrated with huffer R (50 mM Tris-HCI (pH 7.5). 10% (v/v) glycerol, 0.1 mM EDTA, 5 mM dithiothreitol, 100 mM KCI) at a flow rate of 0.1 ml/min; 200-pl fractions were collected. All operations were carried out a t 0-4 "C.
Immunological Detection of C-vrA and htl) Proteins-The immunohlot. method of Burnette (26) was carried out, hut with modifications. Proteins resolved hy SDS-PAGE were transferred electrophoretically onto memhranes (Immohilon PVDF (Millipore) for detection of GyrA protein and Zeta-Prohe (Rio-Rad) for detection of LetD protein by a semi-dry blotter (Trans-Blot SD, Rio-Rad) following the manufacturer's instructions. The hlots were hlocked in a solution of 5% skim milk (Difco) in Tris-saline (20 mM Tris-HCI, pH 7.5, 150 mM NaCI), and then incuhated in the primary antihody diluted in an antibody buffer (1% skim milk in Tris-saline) for a t least 2 h followed I)y t.hree washes in Tris-saline containing 0.05?4 Tween 20. The anti-GyrA and the anti-LetD sera were diluted 500-and 50-fold, respect ively. The hlots were incuhated next with a hiotinylated anti-rahhit IgG (Vector Laboratories), a t a dilution of 1 in 2000, in antihody huffer and then with an avidin-hiotinylated horseradish peroxidase complex (Vector Lahorat,ories) in antihody buffer with washes in Trissaline/Tween 20 between the incuhations. After a thorough washing twice with Tris-saline/Tween 20 and once with Tris-saline, the hlots were suhjected to visualization reactions. 4-Chloro-1-naphthol (for detection of GyrA) and 3J"diaminohenzidine (for detection of LetD) were used as suhstrates for horseradish peroxidase, according to the manufacturer's instructions (Vector Laboratories).
All operations were carried out a t ambient temperature.
Other Methods-Isolation of plasmid DNA, plasmid construction, and transformation were as descrihed (23). SDS-PAGE and gel staining with Coomassie Brilliant Rlue were essentially as descrihed (27). 15% SDS-polyacrylamide gels were used for analyses of LetD and LetA proteins and 12.5% gel for GyrA protein. Protein concentration was determined by the method of Bradford (28) with hovine serum alhumin as a standard. Enzymes and cell lysates stored in small aliquots a t -80 "C showed no changes in activity.

RESULTS
Amplification of the letD Gene Product-The letD gene of F plasmid was cloned on an expression vector derived from plasmid pUC9 (18). In the resulting plasmid (pKP1878), expression of the letD gene depends on the tac promoter and the lac2 Shine-Dalgarno sequence. T o ensure a tight repression of the transcription of the letD gene, the lacl" gene was inserted into the plasmid. However, co-existence of a F'lac plasmid that carried a single copy of the letA-letD operon was necessary to establish the strain transformed with the LetD overproducing plasmid, presumably because a deleterious effect of leaky expression of the letD gene had to be suppressed by the LetA protein.
Amplification of the LetD protein was achieved by two means; one was heat treatment of the cells, which increased copy number of the plasmid, and the other was addition of IPTG to the medium, which induced transcription from the tac promoter. Upon induction, a polypeptide of 11 kDa was overproduced in the cells finally up to 20% of the total protein ( Fig. lA). During the first 1 h after the temperature shift from 28 to 40 "C, optical density of the cells continued to increase whereas the number of viable cells began to decline, thereby suggesting filamentation of the cells. This was confirmed by light microscopic observation. More than 99% of the plasmid containing cells were killed immediately after the addition of  IPTG, presumably because of a high level of the LetD protein (Fig. 1H).
Extensive Relaxation of Plamid DNA in the I A D Overproducing Cells-To examine the effects of the LetD protein on DNA gyrase, we first measured the superhelical density of plasmid DNA in the LetD overproducing cells. The IRtD overproducing plasmid (pKP1878) was extracted from cells collected a t different stages of induction, and the superhelical density was assessed by agarose gel electrophoresis ( Fig. 2A ). We also analyzed a control plasmid (pK1'1879) isolated from the same host strain carrying the plasmid grown under the same condition (Fig. 2R). Upon the addition of IPTG, the plasmid DNA in the LetD overproducing cells became extensively relaxed. It should be noted that the relaxation was readily recognized in an agarose gel, without chloroquine. In contrast, the control plasmid remained supercoiled throughout the time course. Therefore, relaxation of the plasmid was attributed to large amplification of the LetD protein. Copy number of the control plasmid increased, as expected after heat treatment, but there was no such increase with the overproducingplasmid. This may have occurred because DNA replication, in which an action of DNA gyrase is required, was halted after adding IPTG. However, since the level of amplification of the LetD protein exceeded that explained solely by derepression of the tac promoter, it is equally probable that a n unusually tangled plasmid DNA due to loss of decatenation activity of DNA gyrase could not be isolated using alkaline lysis.
Complete Disappearance of DNA Gyrase Activity in the LetD Overproducing Cells-Cell-free extracts were prepared from the LetD overproducing strain (KP5621/pKP1878) and its control strain (KP5621/pKP1879), and DNA supercoiling activity in the extracts was assayed. The extract from the control strain catalyzed conversion of relaxed closed-circular DNA to faster migrating species and finally to a form whose electrophoretic mobility indistinguishable from that of naturally supercoiled DNA. On the other hand, no topological change was ohserved when comparable amounts of the extract from the overproducing strain were titrated in the reaction (Fig. 3 A ) . As shown in Fig. 3R, DNA supercoiling activity in the extract of LetD overproducing strain was retained before the induction and was dramatically lowered to a nondetectahle level in a manner dependent on the expression of the 1etD gene. Thus, it is evident that the loss of supercoiling activity was caused by the overproduction of LetD protein.
One possible explanation for the disappearance of DNA supercoiling activity is enhancement of the DNA relaxing activity such as that of topoisomerase I of E. coli. However, no significant DNA relaxing activity was observed either in the extract of the overproducing strain or in that of the control strain, when assayed under the same condition as for the DNA supercoiling reaction (data not shown). Alternatively, the disappearance of supercoiling activity could be due to activity inhibiting DNA gyrase. When DNA gyrase reconsti- tuted from its purified subunits was incubated with the extract of LetD overproducing strain in the DNA supercoiling assay, no such inhibition was detected (data not shown, see "Experimental Procedures"). These observations strongly suggest that the DNA gyrase activity had disappeared from cells upon induction of the LetD protein.
DNA gyrase is a tetramer of two each of A (GyrA) and R (GyrR) subunits (29). T o determine which of the subunits was affected, GyrA and GyrR activities in the cell extracts were assayed separately (Table I). In the extract of the control strain, the total GyrR activity (assayed in the presence of purified GyrA in excess) was equal to that of the A& tetramer (assayed in neither presence of GyrA nor GyrR). Therefore, almost all the active GyrR protein was present in the A& complex. The total GyrA activity (assayed in the presence of excess GyrB) was 20-fold higher than that of the AZR2 tetramer, thereby indicating that 20 times more GyrA protein existed in a free form than that in the A2R2 complex. Compared with these activities, the activities of both CyrA (including the free form) and GyrR almost completely disappeared in the extract of LetD overproducing cells.

Rejuvenation of Inactivated DNA @rase bv I A A Protein-
If the disappearance of the DNA gyrase activity in the LetD overproducing cells was due to inactivation of the protein hut not to its degradation or repression of the gyr genes. the inactivated gyrase may be rejuvenated. Since LetA protein neutralizes the lethal effect of LetD protein in vioo ( 1 , 4 ) , we tested whether the DNA supercoiling activity in the extract of LetD overproducing cells could be restored by adding purified LetA protein to the extract. As shown in Fig. 4, we observed the LetA-dependent conversion of relaxed closedcircular DNA to forms with increased mohility on an agarose gel. The LetA protein ihelf exhibited neither DNA supercoiling activity (Fig. 4 ) nor stimulation of supercoiling by the purified DNA gyrase (data not shown). A maximal restoration of the supercoiling activity was obtained a t roughly a 1:l ratio of LetA to LetD, assuming that 20% of the total protein in the extract is the LetD protein. The restoration capacity of LetA protein was further confirmed by carrying out the assav with fractions of DEAE-5PW HPLC column chromatogra-

TABLE I DNA suprcoilinp acticitirs in crll rxtrarts of I,&D
ovrrproducinp and control strains DNA supercoiling assay WAS performed using cell extracts of KP5621lpKP187R and KP5621/pKI'IR79 described in the legend tn  phy, the final step of purification of the LetA protein. As shown in Fig. 5, the restoration capacity coincided with a peak of the LetA protein. Using E. coli topoisomerase I whose unique characteristic is to relax negatively but not positively supercoiled DNA (30), it was proven that the restored DNA supercoiling activity introduced negative turns (data not shown). In addition, the restored DNA supercoiling activity was sensitive to oxolinic acid, an inhibitor of t h e A subunit of DNA gyrase (data not shown). These results strongly suggest that an inactivated form of DNA gyrase was present in the LetD overproducing cells and that the LetA protein had a capacity to rejuvenate it. Further titration of the LetA protein in the rejuvenation assay showed an inhibitory effect (Fig. 4). Whether this inhibition is on the rejuvenation process or on the action of DNA supercoiling remains to be examined. The LetA protein itself did not inhibit supercoiling activity of the purified DNA gyrase even when an excess amount of the protein was added (data not shown).
The extent of the restoration by LetA protein is summarized in Table I. When the LetA protein was added to the extract from the control cells, there was no stimulation of the DNA supercoiling activity. In contrast, the addition of the LetA protein to the extract of LetD overproducing cells restored the DNA gyrase activity t o a level comparable with that seen in the control extract; the activity of A,H, complex (assayed in the presence of none or GvrA in excess) was restored to 40-50% of that in the control extract. Moreover, the activity of free GyrA protein (assayed in the presence of GyrB in excess) was restored to a level a s high as 90'; of that in the control extract. It seems, therefore, that the primary target of the LetD protein is the A subunit of DNA gyrase. When the A subunit in the A& complex is inactivated, the R subunit probably remains in the complex so that its activity is not detectable even in the presence of active GvrA protein.
The LetA protein has the capacity to rejuvenate the inactivated form of the GyrA protein whether it is in a free form or in the A& complex.
Association of LetD Protein with thP Inactivatpd A Subunit of DNA Gyrase-The restoration of supercoiling activity by the LetA protein made WAY for an assay to purify the insctivated gyrase in the LetD overproducing cells. DNA supercoiling activity that appeared in the presence of both Let.A and GyrR proteins (LetA .GyrR-dependent DNA supercoiling activity, see Fig. 6) was followed since it was abundant and stable ( Table I). The extract of LetD overproducing cells was fractionated first hy ammonium sulfate precipitation. The LetA. GyrR-dependent supercoiling activity was recovered in the precipitate along with a large amount of the LetD protein. A significant portion of the supercoiling activity that appeared in the presence of only LetA protein (LetA-dependent DNA supercoiling activity, see Fig. 6) remained in the soluble fraction, however, the yield was low. The precipitate fraction was then subjected to a DEAR-trisacryl column chromatography. Almost all the LetD protein was recovered in a flowthrough fraction of the Chromatography, whereas the Let A . GyrB-dependent supercoiling activity was step eluted with a buffer containing 300 mM KC1 in a single peak with a recovery of 84%. This activity was then size fractionated by Superose 12 gel filtration. In the same experiment but under separate runs, the purified LetD and GyrA proteins as well as the standard protein markers were analyzed (Fig. 7 ) . The LetA. GyrB-dependent supercoiling activity was recovered in a single peak with a 38% recovery. Stokes' radii of the LetA. GyrBdependent supercoiling actiyity, GyrA protein, and the LetD protein were 68,64, and 22 A, respectively (Fig. 7 A ) . It should be emphasized that the LetA. GyrB-dependent supercoiling activity was eluted at a position close to but corresponding to a somewhat larger molecule than that of the GyrA protein.
Among the many bands still present in peak fractions of the LetA e GyrR-dependent supercoiling activity visualized by silver staining of an SDS gel, two bands corresponding to the sizes of GyrA and LetD coincided with the LetA. GyrRdependent supercoiling activity (data not shown). Location of GyrA and LetD proteins was therefore examined directly by immunoblotting. Fractions containing the LetA. GyrR-dependent supercoiling activity were subjected to SDS-PAGE, transferred to a membrane, and monitored with antibodies directed against GyrA and LetD proteins (Fig. 7, R and C).
Two bands which migrated to the positions of GyrA and LetD proteins cross-reacted with the corresponding antibodies. These protein peaks coincided perfectly with the LetA-GyrRdependent supercoiling activity. Stoichiometry of these proteins, calculated by comparing the immunoblots of known amounts of purified GyrA and LetD proteins serving as standards, was roughly 1 to 1. Taken together, these results strongly suggest that the A subunit of the DNA gyrase was inactivated in the LetD overproducing cells as a result of formation of a GyrA-LetD complex. Establishment of the components and stoichiometry of the complex awaits further purification.

DISCUSSION
We constructed a plasmid that allows a high level production of the LetD (CcdR) protein and we examined its effect on DNA supercoiling activity. Upon induction of the LetD protein, cells carrying the plasmid formed filaments and were killed. In these cells, DNA supercoiling activity was decreased to a nondetectable level, in a manner dependent on the production of LetD protein. Riochemical analyses of cell-free extracts of the LetD overproducing strain revealed that this was due to inactivation of intrinsic DNA gyrase. The extracts did not contain inhibitory capacity for an extrinsic DNA gyrase. A free form of A subunit of gyrase (GyrA), which was shown to exist in a 20-fold excess over the A& tetramer in normal cells, was also completely inactivated in the LetD overproducing cells. On the other hand, the LetA protein, genetically suggested to be a suppressor of the LetD function (1,4), possessed the ability to rejuvenate the inactivated form of both DNA gyrase and GyrA protein. Addition of the IRtA protein to the extract of LetD overproducing cells led to almost full rejuvenation of the inactivated proteins. Using this rejuvenation as an assay, the inactivated GvrA protein was partially purified. Gel filtration of the inactivated GvrA protein showed it to be somewhat larger than the active GvrA protein. Accordingly, a protein that cross-reacted with an antibody against the LetD protein was co-chromatographed with the inactivated GyrA protein, thereby implying association between the LetD protein and the GyrA protein. This complex formation might be the basis for the GyrA inactivation. Rased on the present observations, we conclude that the A subunit of DNA gyrase is a primary target of the LetD protein and that disappearance of the DNA gyrase activity from the cell accounts for the killing effect of the LetD protein. Furthermore, the results indicate that the IAtD and LetA proteins constitute an opposing pair in modulating the DNA supercoiling activity of the DNA gyrase; the LetD protein inactivates the GyrA protein probably by a direct interaction, whereas the LetA protein is capahle of reversing the inactivation process.
Attempts to reconstitute in vitro the inactivation of gyrase using purified LetD protein or extract of the LetD overproducing cells have not succeeded, possibly because cofactors or conditions might be lacking. One possible explanation is that t,he inactivation process requires a chaperone function (31). The notion derives from observations that mutations in RroES and groEL genes, whose products facilitate folding of newly synthesized polypeptides and their assembly into oligomeric structures (32), overcome the LetD-mediated killing I If?, 17). In contrast to the inactivation of DNA gyrase, the reverse reaction was demonstrated in vitro; purified LetA protein possessed biochemical activity to rejuvenate the inactivated GyrA protein. Since the LetD and LetA proteins form a tight complex of 69 (64)4 kDa (33), it is tempting to postulate that the LetA protein removes the LetD protein from a GyrA-.LetD complex by forming a LetA-LetD complex. The assumption that the LetA-LetD complex is inert as an inactivator is consistent with the in vivo observation that cells expressing both M A and k t D genes show no growth inhibition (1). However, it is unclear how the LetD protein in the LetA-LetD complex becomes free to he activated in cells that had lost the F plasmid. There is a report of instability of the LetA protein, and if this notion is valid, it would explain the activation of LetA-LetD complex (3.7). We observed no such instability with respect to the rejuvenation capacity of the purified LetA protein during normal handling, although the question of stability of the protein was not directly addressed. Complete reconstitution of both inactivation and reactivation of DNA gyrase from purified proteins is necessary to clarify all components involved in the process, and their molecular mode of action.
While mechanisms by which the LetD protein inactivates gyrase remain to be determined, one differing from the inhibitory action of quinolone antibiotics (34), potent inhibitors of the A subunit of DNA gyrase, seems to be operative. Nalidixic acid and related compounds cause an immediate shut-off in DNA synthesis by trapping a covalently bound gyrase-DNA complex, which interferes with movement of the replication fork (35, 36). This biochemical view of the inhibition well explains a genetical observation that nalidixic acid-resistant gyrA mutation (Nal') is recessive against the wild type (Nal") allele (37). In contrast, under physiological circumstances in which the LetD protein functions to kill cells, the LetD protein shows no intriguing effect on DNA synthesis (5).5 Accordingly, a class of gyrA mutations (tu-) rendering cells resistant to the killing effect of LetD protein is dominant over the wild type (LetD") allele (17). With respect to the inhibitory mechanism, the LetD-mediated gyrase inhibition creates an SOS signal (11); merely the elimination of gyrase activity is not sufficient for RecA induction (38). One of the tld-mutations causes an amino acid substitution at the 214th residue in the N-terminal domain of the GyrA protein (17), in which the nicking and closing activities reside (39). The LetD protein might interfere with some aspects of the nicking and closing events in the DNA supercoiling reaction, although other possibilities would need to be considered. As in the case of antigyrase drugs, the LetD protein as a protein inhibitor of the GyrA protein might provide a means to elucidate the yetto-be elucidated mechanisms and structure-function relationships of DNA gyrase.
DNA gyrase plays a fundamental role in cellular processes, including DNA replication, transcription, recombination, as well as the resolution of daughter chromosomes (40, 41). These pleiotropic effects may be attributed to various enzymatic activities of the gyrase and also to the fact that most processes involved in the DNA transaction require a supercoiled state of DNA and a higher order structure of the chromosome in which the gyrase might be a structural component (42,43). Consequently, a number of gyrase mutants so far isolated and cells treated with various inhibitors show different phenotypes depending on the manner in which gyrase is altered or affected (41). When the LetD protein functions to kill cells, the cells form filaments with large unseparated nucleoids irregularly clustered in the center of cells (4, 5). Some normal size anucleate cells are also generated by the aberrant division of the filaments (5). Within the same time span, DNA synthesis in these cells continues, albeit at a slightly reduced rate (5): Therefore, DNA gyrase in these cells seems to be impaired in a specific manner so as not to be able to conduct a proper segregation of the replicated nucleoids without a primary defect in DNA synthesis. Certain conditional lethal mutants carrying an amber mutation in the gyrA gene or temperature-sensitive mutations in the gyrB gene show essentially the same phenotype (13-15,44). Under physiological circumstances, the LetD protein might preferentially inhibit the decatenation activity of DNA gyrase, which was shown to be required for the resolution of duplicated nucleoids (45). In addition, a perturbation of the topological state of DNA or the higher order structure of chro-T. Miki, unpublished results. mosome caused by the LetD protein might affect processes of chromosome segregation.
Although opposing actions of DNA gyrase and DNA topoisomerase I play a central role in determining the level of DNA supercoiling in bacterial cells (40, 46), accumulating evidence suggests that there are other proteins that influence DNA supercoiling. The hns (osmZ) gene product, an abundant DNA-binding protein (H-NS), is one such example (47, 48). The topological changes brought about by the H-NS protein may alter expression of a variety of genes dispersed throughout the E. coli genome (49, 50). It has also been suggested that HU (heterodimer of HU-1 and HU-2)) a histone-like protein, and products of genes in the minB locus which are implied in the proper positioning of septa, are involved in modulation of DNA supercoiling. A reduced level of DNA supercoiling was found in a strain lacking both HU-1 and HU-2 or one with an unbalanced expression of genes in the minB locus (51-53). Interestingly, defective chromosome segregation and cell filamentation, the phenotype reminiscent of the ktD-expressing cells, were observed with these strains (53-55). The gem gene of phage Mu encodes a protein that affects DNA supercoiling presumably through interaction with the B subunit of DNA gyrase (56, 57). The protein enables an efficient proliferation of the phage upon infection by altering expression of the host genes through effects on DNA supercoiling. In addition, a cyclical variation of gyrase activity or supercoiling state of DNA in the bacterial cell was suggested by the interesting features of Mu phage carrying a mutation in the gem gene (Mu gemts2) (58). The present study revealed that the LetD and LetA proteins are in this category. Among such proteins, the LetD and LetA are the only ones seen to have direct effects on DNA gyrase. In addition, the LetD and LetA proteins are unique in that they can reversibly modulate activity of the DNA gyrase. Although the killing effect of LetD protein has been considered to have biological importance (5)) the rejuvenation capacity of LetA protein found in the present study suggests a different view concerning the LetD-mediated inactivation of DNA gyrase. It is feasible that the LetD and LetA proteins form an opposing pair biologically significant for modulation of the DNA gyrase. In this light, the roles of the LetA and LetD proteins during the steady-state growth of bacteria carrying the F plasmid might need to be reconsidered.