Broad-host-range plasmid RK2 encodes a post-segregational killing system, parDE, which contributes to the stable maintenance of this plasmid in Escherichia coli and many distantly related bacteria. The ParE protein is a toxin that inhibits cell growth, causes cell filamentation and eventually cell death. The ParD protein is a specific ParE antitoxin. In this work, the in vitro activities of these two proteins were examined. The ParE protein was found to inhibit DNA synthesis using an E. coli oriC supercoiled template and a replication-proficient E. coli extract. Moreover, ParE inhibited the early stages of both chromosomal and plasmid DNA replication, as measured by the DnaB helicase- and gyrase-dependent formation of FI*, a highly unwound form of supercoiled DNA. The presence of ParD prevented these inhibitory activities of ParE. We also observed that the addition of ParE to supercoiled DNA plus gyrase alone resulted in the formation of a cleavable gyrase–DNA complex that was converted to a linear DNA form upon addition of sodium dodecyl sulphate (SDS). Adding ParD before or after the addition of ParE prevented the formation of this cleavable complex. These results demonstrate that the target of ParE toxin activity in vitro is E. coli gyrase.
A key feature of bacterial plasmids is their remarkably stable maintenance in growing populations of bacteria. This stability is the result of plasmid-encoded processes that ensure that there is a sufficient number of copies of a plasmid in a bacterial cell and that each daughter cell at cell division receives at least one copy of the plasmid element. Plasmids have been shown to encode a variety of regulatory mechanisms that sense plasmid copy number and that respond to upward or downward fluctuations in their copy number. It is also known that low- and moderate-copy-number plasmids are not randomly distributed to daughter cells on cell division but, instead, carry partitioning genes that ensure the distribution of plasmid copies to each daughter cell. The presence of post-segregational killing (PSK) systems on a plasmid also contributes to plasmid maintenance by virtue of the ability of the system to kill or inhibit the growth of a daughter cell that does not receive a copy of the plasmid during cell division.
Active PSK systems have now been described for a number of plasmids in a wide range of bacteria (Gerdes et al., 2000). One major class of PSK systems, designated proteic systems, consists of two proteins, one of which has toxin activity and the other has antitoxin activity by virtue of its ability to form a tight complex with the toxin protein. The best understood proteic PSK system is encoded by the F plasmid and consists of two proteins, CcdB, a toxin, and its antitoxin, CcdA (Karoui et al., 1983; Ogura and Hiraga, 1983; Miki et al., 1984; Jafféet al., 1985). E. coli cells that have lost the F plasmid are killed because the CcdB toxin is relatively stable, but the CcdA antitoxin is readily degraded by the Lon protease (Van Melderen et al., 1994). CcdB has been shown to kill E. coli by binding to the A subunit of DNA gyrase and inhibiting its activity (Bernard and Couturier, 1992; Bahassi et al., 1995). Plasmid R1 and the closely related plasmid R100 also carry a proteic PSK system that has been studied with respect to the mode of action of the toxin (Bravo et al., 1988; Tsuchimoto et al., 1988). The PSK system of these plasmids is identical and, in the case of plasmid R1, the toxin, Kid, has been shown to inhibit DnaB-dependent initiation of DNA replication (Ruiz-Echevarria et al., 1995).
Plasmid RK2 is a 60 kb broad-host-range plasmid that, despite the relatively low copy number of five to eight copies per chromosome, is stably maintained in a wide range of Gram-negative bacteria (Thomas and Helinski, 1989). One region of RK2 that is important in the stable maintenance of this plasmid is the 3.2 kb par region, which consists of five genes on two divergent operons, parCBA and parDE (Saurugger et al., 1986; Gerlitz et al., 1990; Roberts et al., 1990; Roberts and Helinski, 1992). Both operons have been shown to contribute to the stable maintenance of intact RK2 and mini-RK2 replicons in E. coli and in several Gram-negative bacteria distantly related to E. coli (Sia et al., 1995; Sobecky et al., 1996; Easter et al., 1997; 1998). ParA has been shown to be a resolvase, acting at an in cis site located between the two operons, which converts multimeric forms of the plasmid to monomers (Eberl et al., 1994). The functions of the ParB and ParC proteins are not known (Johnson et al., 1999). It is clear, however, that the 9 kDa ParD protein and the 12 kDa ParE protein function as a proteic PSK system, in which the antitoxin ParD protein protects the bacterial cell by forming a complex with the ParE toxin protein (Johnson et al., 1996). ParD exists as a dimer in solution and binds to the dimer form of the ParE toxin to form a tetrameric complex (Johnson et al., 1996; Oberer et al., 1999). Stabilization of mini-replicons of RK2 by parDE is accompanied by growth inhibition, filamentation of plasmid-free segregants and cell death (Roberts et al., 1994). In a comparison of the activities of the proteic PSK systems of plasmids F (ccd), R1 (parD) and RK2 (parDE), it was found that the parDE system was markedly more effective than the other two in stabilizing mini-R1 plasmids in E. coli (Jensen et al., 1995). In the present study, we demonstrate that purified ParE inhibits DNA replication in vitro and that this inhibition results from inactivation of E. coli DNA gyrase by the toxin. Furthermore, the inactivation of gyrase by ParE is prevented by the presence of the specific antitoxin, ParD and can be reversed by the addition of the antitoxin to ParE-inhibited gyrase.
ParE inhibits E. coli oriC DNA replication in vitro
It is well established that the ParE protein is a toxin that causes extensive cell filamentation and that ParD is its specific antidote (Roberts et al., 1994; Johnson et al., 1996), but the mechanism of ParE action is not known. DAPI staining of ParE-induced cell filaments revealed a large nucleoid mass in the centre of the filaments (data not shown), suggesting that either DNA replication or segregation was inhibited. As interference with a number of processes that are important in maintaining chromosome structure and in chromosome partitioning, including DNA replication, is known to result in cell filamentation (Kornberg and Baker, 1992), we tested the effect of ParE on DNA replication in vitro. The assay used consisted of the Fraction II soluble protein extract of E. coli and an E. coli oriC plasmid DNA template. As shown in Fig. 1, ParE effectively inhibited DNA replication of oriC, and this inhibition, over a wide range of concentrations of the toxin, was prevented by the addition of the antitoxin, ParD.
ParE inhibits F1* formation from a chromosomal and a plasmid replication origin
In an attempt to determine which stage of oriC DNA replication is inhibited by ParE, the effect of the toxin was examined on the formation of an extensively unwound form of covalently closed circular oriC DNA (FI*) (Baker et al., 1986). The formation of this electrophoretically distinct form of DNA is the net result of E. coli DnaA protein interaction at the origin of replication leading to the formation of an open complex followed by recruitment of the DnaB–DnaC complex to the origin region and DnaB helicase activation at this open region. The combined action of the DnaB helicase and E. coli gyrase in the presence of ATP generates this extensively unwound form. Nicking and closing of DNA by gyrase removes positive supercoils introduced by progression of the helicase complex on the DNA template (Gellert et al., 1976). As shown in Fig. 2A, ParE addition to the reaction mixture results in a signifi-cant reduction in the amount of FI* as well as a reduction in the levels of both supercoiled (FI) and open circular DNA (FIII) forms. Concomitantly, there is an increase in the amount of DNA that is at the electrophoretic position of the linear form (FII) of oriC DNA. The addition of ParD prevents ParE conversion of the various forms of oriC DNA to the linear form (Fig. 2A, lane 6). This effect of ParE in the FI* reaction is not specific for oriC DNA. Formation of the FI* form of the replication origin (oriV) of RK2 requires the RK2 replication initiation protein, TrfA, in addition to E. coli DnaA, DnaB and DnaC proteins and gyrase. A similar effect of ParE addition is seen with the supercoiled form of plasmid RK2 oriV DNA as the template (Fig. 2B). As with oriC DNA, ParE addition reduces the levels of the FI*, FI and FIII forms of oriV DNA and increases the level of the linear (FII) form. ParD once again neutralizes the ParE effect (Fig. 2B, lane 6). The observed inhibition of FI* formation could be a result of inhibition of helicase or gyrase activity or the inhibition of an earlier step in the formation of the prepriming complex at the replication origin.
A target of ParE inhibitory activity is E. coli gyrase
The conversion of oriC and oriV supercoiled DNA to the linear form by the addition of ParE to the FI* reaction mixture is reminiscent of the action of quinolone anti-biotics or the F plasmid CcdB toxin protein on gyrase (Maxwell, 1997; Couturier et al., 1998). In both cases, the covalently linked enzyme–DNA intermediate in the catalysis of negative supercoiling of DNA by gyrase is stabilized by these topoisomerase poisons and, upon addition of sodium dodecyl sulphate (SDS), usually added to stop the reaction and prepare the reaction mixture for agarose gel electrophoretic analysis, double-strand breakage of the DNA occurs. On the basis of these observations and the fact that gyrase is present in the FI* reaction mixture, ParE protein was tested for its effect on the activity of the E. coli gyrase enzyme. With the exception of gyrase, all the proteins that are required for FI* formation were omitted from the reaction mixture. As shown in Fig. 3A, ParE induced conversion of the supercoiled DNA form of the plasmid RK2 mini-replicon to the linear form (FII). Increasing the amount of gyrase results in increasing levels of the FII form of the DNA. At the highest concentration of gyrase, there is apparent fragmentation of the DNA to lower molecular weight forms. The presence of the antitoxin protein, ParD, in the reaction mixture prevents ParE-induced DNA breakage (Fig. 3B). The formation of the FII form of the DNA is dependent on the presence of ATP (Fig. 4A). In the absence of ATP and the presence of gyrase, there is relaxation of the supercoiled DNA, resulting in more slowly migrating forms (Fig. 4B). If ethidium bromide is present during the agarose electrophoretic separation of the various DNA forms, the various relaxed forms of plasmid DNA generated in the absence of ATP run as a single band (Fig. 4A). The observed conversion of the supercoiled DNA to a largely full-length linear form of DNA by ParE action on gyrase mimics the action of quinolones on this topoisomerase (Maxwell, 1997). As shown in Fig. 5, the quinolone, oxolinic acid, produces a single double-strand break in supercoiled DNA, and this product of oxolinic acid action is found at a position similar to that of ParE inhibition of gyrase. However, unlike what is observed with ParE, in which ParD prevents ParE-induced DNA breakage (Fig. 5, lane 3), ParD fails to neutralize the effect of oxolinic acid (Fig. 5, lane 5). This is consistent with the specific interaction between the ParD and ParE proteins.
Reversal of ParE inhibitory activity by ParD
To determine whether the effect of ParE on gyrase activity can be reversed by the antitoxin, ParD was added to the reaction mixture after the addition of the ParE protein. As shown in Fig. 6, the ParD protein is not only able to protect gyrase from the inhibitory action of ParE when both proteins are added before the addition of gyrase, but ParD is also able to reverse the formation in vitro of a ‘poisoned’ gyrase complex by ParE (Fig. 6). The addition of ParD for a 30 min period, after incubation of ParE with gyrase, ATP and supercoiled DNA for 30 min, completely reversed the formation of the cleavable gyrase–DNA complex (Fig. 6, lane 4). This result is similar to that found for the CcdA–CcdB post-segregational killing system of plasmid F, in which the action of the CcdB toxin protein in trapping gyrase in a cleavable complex can be reversed by the subsequent addition of CcdA antitoxin protein (Bahassi et al., 1999).
The ParE toxin protein inhibits E. coli gyrase and converts supercoiled plasmid DNA to a singly cleaved linear form. By analogy with the extensive studies carried out on the mechanism of action of quinolone antibiotics (Maxwell, 1997) and the F plasmid CcdB protein (Couturier et al., 1998), it is likely that ParE interacts with one of the subunits of gyrase and stabilizes a gyrase–DNA cleavable complex. The addition of SDS, to stop the gyrase reaction before gel electrophoresis, results in the release of the double-strand DNA breaks in the ParE–gyrase–DNA complex and the generation of a linear form of the DNA. In view of the essential role of gyrase in DNA replication and the formation of a prepriming complex (F1*) on supercoiled DNA, it is not surprising that ParE is a strong inhibitor of these reactions in vitro (Figs 1 and 2). The activity of the ParD antitoxin in preventing this inhibition by ParE in vitro is consistent with the established activity of this antitoxin in vivo in preventing ParE killing (Roberts et al., 1994).
Despite the lack of amino acid sequence homology between ParE and the extensively studied CcdB protein (Roberts and Helinski, 1992), there are a number of properties that these two toxins share. Both proteins are relatively low-molecular-weight dimers in solution (Steyaert et al., 1993; Johnson et al., 1996). A linear DNA product of gyrase inhibition is generated by both ParE (Fig. 3) and CcdB (Bernard and Couturier, 1992) after treatment with SDS, not unlike that observed for the inhibitory activity of the quinolone oxolinic acid (Fig. 5) (Liu, 1989). However, unlike oxolinic acid, the inhibitory activity of both ParE (Fig. 4A) and CcdB (Bernard et al., 1993) requires ATP. Not surprisingly, given the specificity of the ParD protein in neutralizing the activity of ParE, the presence of this antitoxin fails to prevent oxolinic acid inactivation of gyrase (Fig. 5).
Escherichia coli gyrase is a type II topoisomerase and consists of two GyrA and two GyrB subunits (Reece and Maxwell, 1991; Maxwell, 1997). The GyrA subunits contain the catalytic core of the enzyme, whereas the GyrB subunits play a key role in ATP binding and hydrolysis. CcdB has been shown to target the GyrA subunit both in vitro and in vivo (Bernard and Couturier, 1992). Although there is no information to date as to which gyrase subunit is the target for ParE activity, it is known that an E. coli strain expressing the GyrA462 mutant protein, which is resistant to CcdB inhibition, is suscep-tible to killing by ParE (Roberts et al., 1994). Although most mutations to quinolone resistance are found in GyrA, several mutations have been mapped to the gyrB gene (Maxwell, 1997). The CcdB and quinolone-resistant mutations are located in different domains of GyrA, and the mutant proteins do not cross-protect (Couturier et al., 1998).
It is known for both quinolones and CcdB, and it is likely for ParE, that the mechanism of toxic activity is to trap gyrase into a cleavable complex. However, the subsequent events that lead ultimately to cell killing are not fully understood. The critical role of gyrase in maintaining the appropriate level of negative superhelical turns in DNA in vivo is well established, and interfering with this enzymatic activity is likely to lead to a cessation of cell growth. It has also been found that, in gyrA+/gyrA462 merodiploid strains, the CcdB-sensitive phenotype is dominant over the resistance phenotype (Bernard and Couturier, 1992). This observation and other evidence obtained on the mode of action of CcdB, quinolones and the antibiotic microcin B17 (Heddle et al., 2001) indicate that these agents both trap and poison the cleavable DNA–gyrase complex. The ‘poison’ hypothesis, in which the stabilized gyrase–inhibitor cleavable complex blocks the passage of polymerases along the DNA, has drawn support from the findings in vitro that the CcdB-stabilized complex blocks transcription by E. coli RNA polymerase (Critchlow et al., 1997). A quinolone-stabilized complex similarly serves as a barrier to DNA replication by T7 DNA polymerase (Wentzell and Maxwell, 2000). A block in replication fork progression could lead to the SOS response that is observed during CcdB killing of E. coli (Karoui et al., 1983; Sommer et al., 1985). It remains to be determined whether the cell killing activity of ParE results from the inhibition of gyrase-catalysed formation of negative supercoils or from the formation of a cleavable complex with gyrase that serves as a barrier to vital DNA processes and/or gives rise to DNA breaks. Until more information is available at the in vivo level on the activity of ParE, the mechanism(s) of ParE cytotoxicity remain(s) to be determined.
ParD effectively prevents the inactivation of gyrase by ParE when present at the time of addition of ParE or when added after incubation of ParE with gyrase and supercoiled DNA. In addition to its ability to form a tight tetrameric complex with ParE in the absence of DNA (Johnson et al., 1996), this antitoxin protein also regulates the expression of the parDE operon in E. coli (Davis et al., 1992). The ability of ParD to reverse the ParE stabilization of the gyrase cleavable complex (Fig. 6) suggests that ParD releases ParE from the toxin–gyrase complex, as has been shown for CcdA and CcdB (Bernard et al., 1993), and restores gyrase catalytic activity. If ParD is present in excess of ParE before the addition of gyrase in vitro, it is possible that the formation of a tight complex between the two Par proteins prevents binding of ParE to gyrase. The effectiveness of the parDE PSK system in the stable maintenance of RK2 and heterologous plasmids is presumably the result of pref-erential loss of active ParD in cells that have lost the plasmid resulting in a release of ParE activity. Unlike the anti-toxins CcdA and Kid, which are degraded by the ATP-dependent protease Lon in the absence of the corresponding toxin (Tsuchimoto et al., 1992; Van Melderen et al., 1994), a mutation in the lon gene in E. coli that inactivates the Lon protein failed to affect the ability of the parDE PSK system to stabilize a mini-RK2 plasmid (Roberts et al., 1994). This suggests that another protease or another mechanism is responsible for the presumed differential decay rate of the ParD antitoxin.
Several plasmid-encoded proteic systems have now been described, including ParD and ParE of plasmid RK2 (Roberts and Helinski, 1992), CcdB and CcdA of plasmid F (Ogura and Hiraga, 1983; Miki et al., 1984), Kis and Kid of plasmid R1 (identical to PemI and PemK of R100; Bravo et al., 1988; Tsuchimoto et al., 1988) and phd/doc of P1 (Lehnherr et al., 1993). Unlike the ParE and CcdB toxins that inactivate gyrase, the Kid protein of plasmid R1inhibits DnaB-dependent initiation of DNA replication in vitro (Ruiz-Echevarria et al., 1995). A non-proteic PSK system, designated hok/sok, is also encoded by the R1 plasmid (Gerdes et al., 1985; 2000). The expression of the Hok toxin protein results in the attenuation of cell growth and the appearance of ‘ghost-like’ cells, possibly as a result of disruption of the bacterial membrane by the Hok protein. Given the vast number of plasmid elements in the wide range of bacteria residing in mammalian, terrestrial and aquatic environments and the likelihood that a high proportion of these plasmids encode a post-segregational killing system, it will not be surprising that the toxin proteins expressed by these systems bring about bacterial cell growth arrest or killing by a variety of mechanisms. In addition, the effectiveness of a PSK system undoubtedly depends upon the host cell in which the toxin is expressed. In comparative studies in which the parDE system, carried by plasmid RK2, was tested for its contribution to plasmid maintenance in different E. coli strains and five bacteria distantly related to E. coli, it was found that the effectiveness of the parDE system in promoting stable maintenance of the plasmid varied with the bacterial host (Sia et al., 1995; Easter et al., 1997; 1998). In another study, the RK2 parDE system was found to be considerably more effective in the maintenance of mini-R1 plasmids in E. coli than the F plasmid CcdA/CcdB system (Jensen et al., 1995). These host and PSK system differences in the stable maintenance of a plasmid are likely to result from a number of factors including differences in the expression of a PSK system in different hosts, internal environmental effects on the toxin– antitoxin interaction, differences in host chaperone activity, variation in antitoxin inactivation and differences in the sensitivity of the target in different bacteria. Stabilization of plasmid maintenance by parD/parE is of particular interest because these genes are on a plasmid that is stably maintained in a wide range of Gram-negative bacteria and the fact that this proteic system has been shown to be active in a number of distantly related bacteria. It is therefore of particular interest to confirm that the in vivo target of the ParE toxin is indeed gyrase and to understand the unique properties of this toxin that allow it to function in such a wide range of bacterial hosts.
Bacterial strains, proteins, plasmids, and reagents
Escherichia coli C600 (thr, leu, thi, supE44, tonA) (Appleyard, 1954) was used in this study for the preparation of a crude extract (Fraction II) active in DNA replication. The RK2-encoded proteins ParE and ParD, purified as described previously (Roberts et al., 1993; Johnson et al., 1996), were generously provided by Dr E. Johnson. The His6-tagged mutant TrfA33 (254D/267L) version of the TrfA mutant protein, which contains two plasmid copy-up mutations and is fully functional in vivo and in vitro (Konieczny and Helinski, 1997a), was purified as described previously (Blasina et al., 1996). Protocols and bacterial strains used for the purification of E. coli gyrase were kindly provided by Dr N. Dixon (Australian National University). Purification of E. coli DnaC and the histidine-tagged versions of E. coli DnaA and DnaB proteins was carried out as described previously (Li and Crooke, 1999; Caspi et al., 2001). HU protein was a kind gift from Dr R. McMacken. pKD19L1 carrying the minimal RK2 origin region and plasmid pBSoriC carrying the oriC DNA origin fragment have been described previously (Yung and Kornberg, 1989; Doran et al., 1998). pYJ2 is an oriC plasmid constructed by the insertion of an EcoRI fragment containing the RK2 tetAR genes into pBSoriC. The E. coli SSB proteins were obtained from Promega, creatine kinase, bovine serum albumin (Fraction V), creatine phosphate and rNTPs from Sigma, dNTPs from Pharmacia, and [3H-methyl]-dTTP from ICN Radiochemicals.
Escherichia coli oriC DNA replication using crude extracts
Preparation of the E. coli crude Fraction II extract and reaction conditions was carried out essentially as described previously (Kittell and Helinski, 1991). The standard reactions contained 300 ng of pBSoriC template. Incubation was at 32°C for 60 min. Reactions were stopped by the addition of trichloroacetic acid, and total nucleotide incorporation (pmol) into DNA was measured by liquid scintillation counting, after filtration of the reaction mixture onto Whatman GF/C glassfibre filters.
Template unwinding assay (FI*)
Helicase unwinding assays (FI* formation) were performed as described previously (Konieczny and Helinski, 1997b). The plasmid pYJ2 (300 ng) containing the E. coli chromo-somal origin or plasmid pKD19L1 (300 ng) containing the RK2 replication origin were used as DNA templates. Unless noted otherwise, proteins were used in the following amounts: DnaA (320 ng), DnaB (600 ng), DnaC (120 ng), HU (5 ng), gyrase (120 ng) and SSB (230 ng). The oriV-dependent reaction was supplemented with RK2 replication initiation protein TrfA 254D/267L (500 ng). The indicated amounts of ParE and ParD proteins were added to the reaction mixtures. Reactions (25 μl) were assembled on ice in buffer containing 40 mM HEPES/KOH, pH 8.0, 25 mM Tris-HCl, pH 7.4, 80 μg ml–1 BSA, 4% sucrose; 4 mM dithiothreitol (DTT), 11 mM magnesium acetate, 2 mM ATP, 8 mM creatine phosphate and 20 μg ml–1 creatine kinase, and then incubations were carried out at 32°C for 30 min. After incubation, the reactions were stopped by the addition of 10 mM EDTA and 2% SDS (final concentration), followed by 2 min incubation at 65°C. Aliquots of 10% sucrose and 0.05% bromophenol blue (final concentration) were then added to the reaction mixture. The mixture was analysed on a 1% agarose gel in TBE buffer (0.09 M Tris-borate, 0.002 M EDTA). The samples were electrophoresed at 25 V for 20 h, and the gel was stained with ethidium bromide solution.
Gyrase inhibition assay
The standard reaction (25 μl) contained supercoiled pKDL19 DNA (300 ng), the indicated amounts of E. coli gyrase and, unless noted otherwise, ParE (350 ng), ParD (1000 ng) or oxolinic acid (40 ng). The reactions were carried out in the same buffer as that used for the template unwinding assay and incubated, terminated and analysed as described for the FI* assay.
We thank Dr Nick Dixon for bacterial strains and protocols for the purification of the gyrase enzyme, and Dr Aresa Toukdarian for critical reading of this manuscript. This work was supported by a National Institutes of Health Research Grant AI-07194 awarded to D.R.H.