The ratio between CcdA and CcdB modulates the transcriptional repression of the ccd poison–antidote system


  • Hassan Afif,

    1. Laboratoire de Génétique des Procaryotes, Institut de Biologie et de Médecine Moléculaires, Université Libre de Bruxelles, 12 Rue des Professeurs Jeener et Brachet, 6041 Gosselies, Belgium.
    Search for more papers by this author
  • Noureddine Allali,

    1. Laboratoire de Génétique des Procaryotes, Institut de Biologie et de Médecine Moléculaires, Université Libre de Bruxelles, 12 Rue des Professeurs Jeener et Brachet, 6041 Gosselies, Belgium.
    Search for more papers by this author
  • Martine Couturier,

    1. Laboratoire de Génétique des Procaryotes, Institut de Biologie et de Médecine Moléculaires, Université Libre de Bruxelles, 12 Rue des Professeurs Jeener et Brachet, 6041 Gosselies, Belgium.
    Search for more papers by this author
  • Laurence Van Melderen

    Corresponding authorSearch for more papers by this author


The ccd operon of the F plasmid encodes CcdB, a toxin targeting the essential gyrase of Escherichia coli, and CcdA, the unstable antidote that interacts with CcdB to neutralize its toxicity. Although work from our group and others has established that CcdA and CcdB are required for transcriptional repression of the operon, the underlying mechanism remains unclear. The results presented here indicate that, although CcdA is the DNA-binding element of the CcdA–CcdB complex, the stoichiometry of the two proteins determines whether or not the complex binds to the ccd operator–promoter region. Using electrophoretic mobility shift assays, we show that a (CcdA)2–(CcdB)2 complex binds DNA. The addition of extra CcdB to that protein–DNA complex completely abolishes DNA retardation. Based on these results, we propose a model in which the ratio between CcdA and CcdB regulates the repression state of the ccd operon. When the level of CcdA is superior or equal to that of CcdB, repression results. In contrast, derepression occurs when CcdB is in excess of CcdA. By ensuring an antidote–toxin ratio greater than one, this mechanism could prevent the harmful effect of CcdB in plasmid-containing bacteria.


Low-copy-number plasmids often possess poison–antidote systems (also called post-segregational killing, programmed cell death or addiction systems). These systems guarantee preferential growth of plasmid-carrying cells in a bacterial population by killing newborn bacteria that have not inherited a plasmid copy at cell division (post-segregational killing) (for reviews, see (Couturier et al., 1998; Engelberg-Kulka and Glaser, 1999; Gerdes, 2000). Such systems are composed of two genes organized in an operon and encoding two small products, a poison (protein) and an antidote counteracting its action (protein or RNA). Post-segregational killing of plasmid-free segregant relies on the differential stability of the two products, the toxin being stable whereas the antidote is unstable. When the antidote is a protein (proteic systems), it is degraded by a host-encoded ATP-dependent protease. Upon loss of a plasmid carrying a poison–antidote system, and thus in the absence of de novo synthesis of antidote and toxin, antidote degradation results in activation of the toxin and thus in the killing of plasmid-free bacteria.

Proteic poison–antidote systems are autoregulated at the level of transcription by binding of the antidote to the operator–promoter region of the operon (de Feyter et al., 1989; Ruiz-Echevarria et al., 1991; Magnuson et al., 1996; Smith and Rawlings, 1998). In most cases, the poison is a co-repressor enhancing antidote binding (Roberts et al., 1993;Gotfredsen and Gerdes, 1998; Magnuson and Yarmolinsky, 1998). However, in the case of the ω–ε–ζ operon, the ω protein is responsible for autorepression (de la Hoz et al., 2000). Autoregulation is thought to avoid inappropriate activation of the toxin in plasmid-containing bacteria.

The ccd system (control of cell death) of the F plasmid encodes two proteins, the CcdB toxin (101 amino acids) and the CcdA antidote (72 amino acids). The latter prevents CcdB toxicity by forming a tight CcdA–CcdB complex. The target of CcdB is the GyrA subunit of gyrase, an essential type II topoisomerase in Escherichia coli (Bernard and Couturier, 1992; Bernard et al., 1993). The CcdA antidote is degraded by the ATP-dependent Lon protease (Van Melderen et al., 1994; 1996). We and others have shown previously that in vivo transcriptional repression of the ccd operon requires CcdA and CcdB (Tam and Kline, 1989a,b;Salmon et al., 1994). Similar results were obtained in vitro: CcdA and CcdB specifically bind together to the ccd operator–promoter DNA region (Salmon et al., 1994). This region contains several binding sites spaced over 113 bp overlapping with the beginning of the ccdA open reading frame (ORF) (Tam and Kline, 1989b). However, the precise binding site(s) of the CcdA–CcdB complex have not yet been identified. Several studies have focused on the nature of the CcdA–CcdB complex involved in autoregulation and antikilling. Tam and Kline (1989b) isolated a 69 kDa CcdA–CcdB complex in a strain overproducing both proteins whether or not the ccd operator–promoter DNA region was present. Using native gel electrophoresis with the purified CcdA and CcdB, we found that CcdA and CcdB are able to form complexes of different stoichiometry depending on the CcdA–CcdB molar ratio (Van Melderen et al., 1996). In that context, a 1:1 and a 1:2 complex have been observed. The latter was isolated by gel filtration, and equilibrium sedimentation gave a molecular weight of 58 000, close to that expected for a complex of a tetramer of CcdB with a dimer of CcdA. An equimolar (CcdA)2–(CcdB)2 complex was isolated when CcdA was used to remove CcdB from the inactive GyrA–CcdB complex (Bahassi et al., 1999).

The 41 carboxy-terminal residues of CcdA (CcdA41) are sufficient to antagonize the toxic activity of CcdB (Bernard and Couturier, 1991). However, the CcdA41–CcdB complex cannot mediate in vivo autoregulation or binding to the operator–promoter region of the ccd operon, suggesting that the amino-terminal part of CcdA may be required for DNA binding (Salmon et al., 1994).

The purpose of this study was to examine further the roles of CcdA and CcdB in binding to the ccd operator–promoter region and to determine the nature of the repression complex. Our results enable us to propose a model for the transcriptional autoregulation of the ccd operon.


CcdA, in contrast to CcdB, binds specifically to the ccd operator–promoter

Using electrophoretic mobility shift assays (EMSAs), the CcdA and CcdB proteins were tested separately for binding activity to the 113 bp ccd operator–promoter (O/P) region defined by Tam and Kline (1989b). We have reported previously that neither CcdA nor CcdB binds to the O/P fragment at protein concentrations ranging from 0.036 µM to 2.4 µM for CcdA or from 0.024 µM to 1.6 µM for CcdB (Salmon et al., 1994). Here, we tested higher concentrations and observed that CcdA does cause retardation of the O/P fragment at 2.8 µM or higher concentrations (Fig. 1A). At 2.0 and 2.4 µM CcdA, we did not observe any discrete mobility-shifted complex, but a smear. This could reflect some instability of the CcdA–O/P complex at these concentrations of CcdA. A first discrete mobility-shifted complex was observed from 2.8 to 3.6 µM CcdA. At higher concentrations than 3.6 µM, the O/P region underwent a second shift, suggesting the presence of more than one CcdA binding site in the O/P. A 120-fold excess of competitor DNA was unable to displace CcdA (at 4 µM) bound to the O/P fragment (data not shown). Thus, CcdA binds specifically to the O/P region in vitro even though we and others have never detected any repressor activity of CcdA on O/Pccd::lacZ fusions in vivo(Table 1, last line; Tam and Kline, 1989a; Salmon et al., 1994). This suggests that the binding of CcdA to the O/P region is probably too weak to be detected in the in vivo assays.

Figure 1.

Binding of CcdA and/or CcdB to the ccd operator–promoter region determined by EMSA. A 230 bp labelled DNA fragment containing the ccd O/P was incubated with increasing concentrations of CcdA (A), with increasing concentrations of CcdB (B) and with increasing concentrations of both proteins at a CcdA–CcdB ratio of 1.5 (C).

Table 1.   Autoregulation activity of CcdA and CcdB mutants.
PlasmidDescription LacZ expression
  1. LacZ expression was calculated as the ratio of β-galactosidase specific activity of the O/Pccd::lacZ measured for CSH50 (pULB2600/pULB2548 containing either the wild type or the ccdA and ccdB mutants) and the CSH50 (pULB2600/pKT279) control strain.

pULB2548Wild-type CcdA and CcdB0.03
ccdAR4C-ccdB CcdA Arg-4 in Cys and wild-type CcdB0.95
ccdAR4A-ccdB CcdA Arg-4 in Ala and wild-type CcdB0.90
ccdAR70K-ccdBCcdA Arg-70 in Lys and wild-type CcdB0.50
ccdAR70A-ccdB CcdA Arg-70 in Ala and wild-type CcdB0.12
ccdAK2A-ccdB CcdA Lys-2 in Ala and wild-type CcdB0.11
ccdAT6A-ccdB CcdA Thr-6 in Ala and wild-type CcdB0.03
ccdAT8A-ccdB CcdA Thr-8 in Ala and wild-type CcdB0.04
ccdAD10A-ccdB CcdA Asp-10 in Ala and wild-type CcdB0.05
ccdA-ccdBopa1 Wild-type CcdA1.20

CcdB was tested under the same conditions, and no binding of CcdB to the O/P DNA fragment was observed (Fig. 1B).

CcdB enhances CcdA binding to the ccd operator–promoter

Because in vivo repression of the ccd operon requires both CcdA and CcdB, we tested the effect of CcdB on CcdA binding to the operator–promoter. CcdA and CcdB were added together with the labelled O/P DNA. Increasing amounts of both proteins were added while the CcdA–CcdB ratio was kept constant at 1.5. Under these conditions, the O/P region appeared to be shifted at CcdA concentrations as low as 0.2 µM (compared with 2.8 µM for CcdA alone; Fig. 1C). This shows that CcdB enhances the binding activity of CcdA to the O/P region, even though it does not show any DNA-binding activity by itself. At higher CcdA and CcdB concentrations (0.4 and 0.27 µM respectively), two mobility-shifted complexes were observed. Although at CcdA and CcdB concentrations of 0.80 and 0.53 µM, almost all the probe is shifted, Fig. 1C (lanes 6 and 7) shows that it is possible to get a supershift when even more of CcdA and CcdB is added. This probably reflects multiple binding sites for the CcdA–CcdB complex. At least two binding sites have been described previously (Tam and Kline, 1989b), and we confirmed these results by dividing the O/P region into two parts, each of which was retarded by CcdA alone or by the complex as a single shift (data not shown).

Nature of the CcdA–CcdB complex that binds to the ccd operator–promoter

CcdA–CcdB complexes of different stoichiometry have been identified in vitro, depending on the CcdA–CcdB ratio: a hexameric (CcdA)2–(CcdB)4 complex, formed when CcdB is in excess of CcdA, and an equimolar (CcdA)2–(CcdB)2 complex, formed when CcdA is equal to or in excess of CcdB (Van Melderen et al., 1996). To determine the nature of the CcdA–CcdB complex that binds to the O/P region, we performed EMSAs in which the concentration of either CcdA or CcdB was fixed, and the concentration of the other protein was varied.

In a first experiment (Fig. 2A), we kept the CcdA concentration constant at 0.8 µM and varied that of CcdB. In the absence of CcdB, mobility of the O/P fragment was not altered (as shown in Fig. 1A). Increasing CcdB concentrations (0.2–0.6 µM) led to the appearance of mobility-shifted complexes that increased as a function of CcdB concentration. The persistence of free O/P independently of the amount of CcdB suggests that CcdA is the limiting factor for binding of the CcdA–CcdB complex. As the CcdB concentration increased further (0.8–1.6 µM), the amount of mobility-shifted complexes decreased sharply, and free O/P was released (Fig. 2A).

Figure 2.

Effect of the CcdA–CcdB ratio on DNA mobility. A 230 bp labelled DNA fragment containing the ccd O/P was incubated with CcdA (0.8 µM) and increasing concentrations of CcdB (A) and with CcdB (0.8 µM) and increasing concentrations of CcdA (B).

In the experiment depicted in Fig. 2B, we used a fixed concentration of CcdB (0.8 µM) and varied the concentration of CcdA. At CcdA concentrations ranging from 0 to 0.6 µM, mobility of the O/P fragment was not altered. At a CcdB concentration equal to that of CcdA (0.8 µM), the O/P region started to be shifted. As the amount of CcdA increased further (from 1.0 to 1.6 µM), the amount of mobility-shifted complexes also increased (Fig. 2B).

In both experiments, we observed mobility-shifted complexes when CcdA was in excess of CcdB. Under these conditions, it has been shown that the formation of the equimolar (CcdA)2–(CcdB)2 complex is favoured (Van Melderen et al., 1996), suggesting strongly that this complex has DNA-binding activity. In conditions that favour the formation of the hexameric complex (CcdB in excess of CcdA), mobility-shifted complexes are not detected. To test directly whether the hexameric complex has DNA-binding activity, we mixed CcdA and CcdB (CcdA–CcdB ratio 1:2) and isolated the CcdA–CcdB complex by gel filtration (data not shown). The estimated molecular weight of that complex was 59 000, which corresponds to the characterized (CcdA)2–(CcdB)4 complex (Van Melderen et al., 1996). We tested the ability of this complex to bind to the O/P fragment by EMSA and did not observe any significant mobility-shifted complexes at concentrations at which the equimolar complex interacts with the O/P fragment and produces mobility-shifted complexes (data not shown). This shows that the hexameric complex is unable to bind the O/P fragment.

An excess of CcdB destabilizes the (CcdA)2–(CcdB)2–DNA ternary complex

As the purified (CcdA)2–(CcdB)4 complex does not show any significant DNA-binding activity, we asked whether an excess of CcdB was able to destabilize the (CcdA)2–(CcdB)2–DNA complex. We first measured the stability of the repressor–DNA complex. CcdA (1 µM) and CcdB (0.8 µM) were incubated with 22 nM O/P DNA to allow complex formation, challenged with 11 µM unlabelled O/P DNA (500-fold excess) and analysed by EMSA at various time intervals. Controls confirmed the efficacy of complex formation and competition by unlabelled DNA. Under these conditions, the (CcdA)2–(CcdB)2–DNA complex had a half-life of at least 60 min (data not shown). This indicates that the repressor–DNA complex is very stable.

In a second experiment, we preformed the proteins–DNA complex and increased the CcdB concentration. Figure 3 (lane 2) shows the mobility-shifted complexes (RC) in the presence of CcdA at 1 µM and CcdB at 0.8 µM. When extra CcdB was added (0.4–1.6 µM), the amount of mobility-shifted complexes decreased, and free O/P was released. This shows that an excess of CcdB with respect to CcdA leads to destabilization of the preformed (CcdA)2–(CcdB)2–DNA complex.

Figure 3.

Effect of an excess of CcdA or CcdB on the repressor–DNA complex. Mobility-shifted complexes were preformed with CcdA at 1 µM and CcdB at 0.8 µM (RC). CcdB or CcdA were added as indicated to the preformed protein–DNA complexes (RC).

In contrast, increasing CcdA concentration had no effect on binding of the (CcdA)2–(CcdB)2 complex to the O/P fragment (Fig. 3).

In vivo isolation of mutants deficient in autoregulation

To identify functional domains of the CcdA and CcdB proteins involved in autoregulation, we randomly mutagenized the ccdA and ccdB genes.

The pULB2548 plasmid carrying both genes under the control of their own promoter (pKTccdA-ccdB) was mutated in vitro by treatment with 2-hydroxylamine (see Experimental procedures). To screen for ccdA or ccdB mutants affected in autoregulation, we introduced the mutagenized plasmids into a Δ(pro-lac) CcdB-resistant strain harbouring a pKT-compatible plasmid carrying an O/Pccd::lacZ transcriptional fusion (CSH50 gyrA462/pULB2600). Transformation mixes were plated on MacConkey lactose plates with the appropriate antibiotics, and we looked for Lac+ colonies (unable to perform autoregulation) among the Lac colonies. Thirty Lac+ colonies were obtained among several thousand transformants. Plasmid DNAs were extracted and used to transform the same strain in order to confirm the Lac+ phenotype. They were also introduced into isogenic strains carrying different amber suppressors (supE, supF and supP). Among the 30 clones, 11 showed a Lac phenotype in the suppressor strains, indicating that each of these clones carried an amber mutation in either the ccdA or the ccdB gene. Of the clones still showing a Lac+ phenotype in the suppressor strains, 16 were sequenced. Eleven were found to carry a mutation in the ccdB gene, all of which were opal mutations (stop codon) at positions 1 or 61 in the CcdB protein sequence. These clones could thus produce CcdA alone or CcdA and a truncated form of CcdB (lacking the 41 C-terminal amino acids of wild-type CcdB). The ccdA-ccdBopa1 plasmid (carrying the opal mutation at position 1) was unable to repress LacZ expression from the ccd promoter (Table 1). This confirms that CcdA alone cannot mediate in vivo autoregulation under these conditions. The five other sequenced clones were found to carry a mutation in the ccdA gene sequence. Two of them carried a point mutation changing the CGT codon coding for an arginine at position 4 to a TGT codon coding for cysteine (R4C). The other three clones each carried three mutations, two point mutations changing an AGG codon coding for arginine at position 70 to an AAA codon coding for a lysine (R70K), and an amber mutation changing the TGG codon coding for a tryptophan at position 72 to an amber codon (TAG). The R70K and ΔW72 mutations were separated by site-directed mutagenesis, and only the R70K mutation was found to affect the autoregulatory property of CcdA. The ΔW72 had no effect on either the antikilling or the autoregulatory properties of CcdA (data not shown).

We also analysed the two CcdA mutants affected in autoregulation. Both mutants can still counteract CcdB toxicity, so they must still be able to bind CcdB (data not shown). We measured autoregulation of the ccd O/P by these mutant proteins in the presence of CcdB by assaying LacZ expression from the O/Pccd::lacZ fusion in a wild-type strain. Mutation of R4 in C in CcdA completely abolishes repression of the O/Pccd::lacZ fusion (Table 1). The R70K mutation causes a twofold decrease in repression (Table 1). Using EMSA, we tested CcdAR70K alone or in the presence of CcdB (data not shown). Whether CcdB was present or not, two- to threefold greater concentrations of CcdAR70K protein were required to produce the same shifts as for the wild-type CcdA protein. This is consistent with the diminished capacity of CcdAR70K to repress the O/Pccd::lacZ in vivo in the presence of CcdB. When the R70 is changed to an alanine (CcdAR70A), only a 10% decrease in repression was observed (Table 1). This suggests that the positive charge at position 70 is not essential for the DNA-binding activity of CcdA.

Arginine 4 of CcdA is essential for operator–promoter binding

As the CcdR4C mutant was completely defective for in vivo autoregulation, we characterized it further. The mutant protein was purified, and its ability to bind to the ccd O/P region was assayed in vitro. CcdAR4C showed no DNA-binding activity alone or in the presence of CcdB (data not shown). In the course of purification, it appeared as a stable dimer when subjected to SDS denaturation (data not shown). As wild-type CcdA is a dimer in solution (Van Melderen et al., 1996), it seems likely that CcdAR4C is stabilized by disulphide bond formation between cysteines on different monomers. To rule out any such effect on CcdA–CcdB complex formation, we mutated R4 to alanine by site-directed mutagenesis. In vivo, the R4A mutant showed the same phenotype as R4C, i.e. it was unable to mediate autoregulation in the presence of CcdB (Table 1). CcdAR4A was purified and found to be a dimer by gel filtration (Fig. 4). Figure 5 shows the gel electrophoretic profile of the complex formed with different ratios of CcdAR4A and CcdB. As the amount of CcdB was increased, the band in the CcdAR4A position disappeared, and a band of lower mobility appeared. Free CcdB was not observed until the molar ratio of CcdA to CcdB was greater than 1:2 (0.2 nmol CcdA−0.6 nmol of CcdB). These observations are consistent with our previous data on CcdA–CcdB complex formation (Van Melderen et al., 1996) and strongly suggest that the CcdR4A mutant and CcdB are able to form the equimolar and the hexameric complex. Formation of both complexes was confirmed further by an isothermal titration calorimetry experiment (data not shown). In EMSA, CcdAR4A displayed no DNA-binding activity, whether CcdB was present or not (Fig. 6). The fact that both the R4C and the R4A mutation completely abolished the ability of CcdA to bind to the O/P fragment suggests that R4 is essential for DNA binding. These results also confirm that CcdA is the DNA-binding element of the CcdA–CcdB repressor.

Figure 4.

Determination of the native form of the CcdAR4A by gel filtration. The standard curve of log molecular weight versus elution volume was constructed using the following standard proteins: albumin (66 000); carbonic anhydrase (29 000) and cytochrome c (12 400) (open diamonds). Samples of 60 µg of CcdA or CcdAR4A were loaded on the gel filtration column. The positions of CcdA (filled circle) and CcdAR4A (open triangle) are indicated on the standard curve. The mutant protein, like the wild-type CcdA, shows an apparent molecular weight of about 26 000 as described previously (Van Melderen et al., 1996).

Figure 5.

Analysis of the CcdAR4A–CcdB complexes by native gel electrophoresis. Gel electrophoretic profile of the complex formed with different ratios of CcdAR4A and CcdB. Separated mixtures containing 0.2 nmol of CcdAR4A and increasing amounts of CcdB were applied to the gel lanes as indicated. In the last lane, CcdAR4A is omitted. Samples were prepared at room temperature, mixed with sample buffer and loaded on a non-denaturing polyacrylamide gel.

Figure 6.

DNA-binding activity of CcdAR4A alone or in the presence of CcdB. A 230 bp labelled DNA fragment containing the ccd O/P was incubated with increased concentrations of CcdAR4A (A) and with increased concentrations of CcdAR4A and CcdB at a CcdAR4A–CcdB ratio of 1.5 (B).

Because a CcdA mutant lacking the 31 amino-terminal amino acids (CcdA41) is unable to bind the O/P DNA in vitro or to perform autoregulation in vivo in the presence of CcdB (Salmon et al., 1994), we wondered whether or not other amino-terminal amino acids might be important for autoregulation. To answer this question, we constructed several alanine-substituted mutants and tested them in vivo for antikilling and autoregulation in the presence of CcdB. The mutants tested were CcdAK2A, CcdAT6A, CcdAT8A and CcdAD10A. All can still counteract CcdB toxicity, showing that they are still able to interact with CcdB. Only the K2A mutant was slightly deficient in autoregulation (Table 1).


The ccd operon of plasmid F contains two genes, ccdA and ccdB, encoding the CcdA antidote and the CcdB toxin respectively. In vivo autoregulation of the operon requires both proteins. In vitro studies have shown that CcdA and CcdB interact and form complexes of different stoichiometry depending on the ratio of CcdA to CcdB (Van Melderen et al., 1996). When this ratio is equal or exceeds one, an equimolar (CcdA)2–(CcdB)2 complex is preferentially formed. When the CcdA–CcdB ratio is smaller than one, the stoichiometry of the complex shifts to a hexameric (CcdA)2–(CcdB)4 complex. Using EMSAs, we observed mobility-shifted complexes when the CcdA–CcdB ratio was equal to or exceeded one. This strongly suggests that the (CcdA)2–(CcdB)2 complex or multiple forms of this complex is the repressor complex. We were unable to test directly the DNA-binding activity of the purified (CcdA)2–(CcdB)2 complex because this complex precipitates in solution (Van Melderen et al., 1996). However, in our EMSAs, the proteins–DNA ternary complex did produce discrete mobility-shifted complexes, suggesting that the equimolar complex is soluble in the presence of O/P DNA. In contrast, we failed to observe any DNA-binding activity of the purified (CcdA)2–(CcdB)4 complex.

Our results show that CcdA alone binds with low affinity to the O/P region, whereas CcdB alone has no binding activity. In the presence of an equimolar amount of CcdB, the affinity of CcdA for DNA is significantly increased. The addition of extra CcdB to the preformed ternary (CcdA)2–(CcdB)2–DNA complex, and thus most probably the formation of the hexameric complex, destabilizes the repressor–DNA complex. These results support the conclusion that, although CcdA is the DNA-binding element of the repressor complex, CcdB can either activate or inhibit the DNA-binding activity of CcdA, depending on the stoichiometry of CcdB present in the CcdA–CcdB complex. CcdA is a rather flexible protein that undergoes conformational changes upon binding to its partners, i.e. CcdB (Van Melderen et al., 1996) or the O/P DNA (Dao-Thi et al., 2000). Structural modifications induced by the formation of different complexes might render the DNA-binding motif of CcdA more or less accessible or buried, thus making the complex able or unable to interact with DNA.

Our results led us to propose a model in which the amount of free CcdB in the cell regulates the repression state of the ccd operon (Fig. 7). When CcdA is in excess with respect to CcdB, the equimolar complex is formed and is able to bind the ccd operator–promoter region (Fig. 7A). When the CcdB concentration becomes higher than the concentration of CcdA, formation of the hexameric complex destabilizes the (CcdA)2–(CcdB)2–DNA complex (Fig. 7B). In vivo, one can imagine that, under certain conditions, uncontrolled degradation of the antidote by the Lon protease could modify the CcdA–CcdB ratio. Excess of the toxin would favour formation of the hexameric complex and thereby alleviate repression of the operon. This would lead to subsequent CcdA and CcdB synthesis until the proper ratio is reached.

Figure 7.

Model for transcriptional autoregulation of the ccd operon. The genetic organization of the ccd operon is shown in (A). The ccdA gene and its product are represented in yellow, the ccdB gene and its products in orange and the operator–promoter region of the operon in green. Both the CcdA and the CcdB proteins are dimers and are represented as a single oval for clarity. In vivo, we observed that the concentration of CcdA is greater than that of CcdB (our unpublished data). This is also the case for other systems, such as pem, phd/doc and relBE. Under these conditions, the (CcdA)2–(CcdB)2 equimolar complex is formed and represses the ccd operon. In (B), the ATP-dependent Lon protease is represented in purple, and the gyrase in blue. Our model proposes that, in the case of the appearance of free CcdB (as a result of CcdA degradation by Lon), there is a preferential binding of CcdB to the repressor–DNA complex. This has two consequences: first, it prevents CcdB interacting with gyrase and, secondly, it leads to derepression of the ccd operon. This mechanism restores the CcdA–CcdB ratio to a value greater than one.

Interestingly, a high toxin-to-antidote ratio was also found to result in a loss of repression and/or DNA-binding activity in the case of phd/doc and parDE (Johnson et al., 1996; Magnuson and Yarmolinsky, 1998). Alleviation of repression when the toxin is in excess of the antidote may be a common mechanism in all poison–antidote systems to prevent harmful activation of toxin in plasmid-containing bacteria.

CcdA alone can bind to the ccd O/P region. What is the motif in CcdA responsible for DNA binding? In previous work, we observed that CcdA mutants affected in their amino-terminal part (R4G point mutation or deletion of the 31 amino-terminal amino acids) were unable to mediate autoregulation in the presence of CcdB (Bernard and Couturier, 1991; Salmon et al., 1994). These results suggest that the amino-terminus of CcdA might be specifically involved in DNA binding. In our in vivo screen, the only point mutation we isolated in ccdA that abolishes repression completely is located in that same region. CcdAR4C or R4A mutations abolish both in vivo repression and binding of the complex to the O/P DNA in vitro. This confirms that binding of the CcdA–CcdB complex to O/P DNA occurs via CcdA. Predictions of CcdA secondary structure indicate one or two putative β-sheets at the amino-terminus, followed by α-helices. As pointed out by Magnuson et al. (1996), CcdA and other plasmid-encoded antidotes such as PhD, PemI and ParD, share structural similarities with members of the ribbon–helix–helix family of DNA-binding proteins, such as Arc, MetJ and others. These repressors bind to specific DNA sequences through the amino-terminal β-sheet (Somers and Phillips, 1992; Suzuki, 1995). Among the mutations in CcdA so far isolated or constructed, only mutations of the arginine at position 4 completely abolish CcdA binding to DNA. This strongly suggests that the amino-terminal part of CcdA, especially arginine 4, is directly involved in DNA binding.

Experimental procedures

Bacterial strains and plasmids

The strains used in this work are CSH50 (Δ(lac-pro), ara, thi) and the isogenic strain CSH50 gyrA462 containing a mutation that confers resistance to CcdB (CSH50, zei298::Tn10, gyrA462) (Salmon et al., 1994). The O/Pccd::lacZ transcriptional fusion is carried by a pACY–C184 derivative pULB2600 plasmid (Salmon et al., 1994). The pULB2548 plasmid is a pBR322 derivative resistant to tetracycline and kanamycin (pKT279) containing the ccd operon (pKTccdA-ccdB) (our laboratory collection).

Media, antibiotics and recombinant DNA methodology

LB broth, LB agar and MacConkey media were prepared as described previously (Miller, 1972). TB medium was prepared as described by Sambrook et al. (1989). Most routine manipulations of DNA were performed as described previously (Sambrook et al., 1989). Selection for plasmids was accomplished by the addition of ampicillin (100 µg ml−1), chloramphenicol (20 µg ml−1) and kanamycin (50 µg ml−1).

β-Galactosidase assays

CSH50/pULB2600 (O/Pccd::lacZ) containing the pULB2548 (pKTccdA-ccdB) or the ccdA or ccdB mutant derivatives was grown in LB containing the appropriate antibiotics at 37°C to an OD600 of 0.3–0.5. β-Galactosidase assays were performed as described previously (Miller, 1972) on 0.1 ml of SDS–chloroform-permeabilized cells. Specific activities used to construct Table 1 are an average of at least three experiments.

Electrophoretic mobility shift assays (EMSAs)

The DNA fragment used for the DNA-binding assay is a 230 bp DNA fragment containing the 113 bp operator–promoter of the ccd operon. This fragment was obtained by polymerase chain reaction (PCR) using pULB2601 (pACY-C 184 derivative carrying the regulatory region of the ccd operon; Salmon et al., 1994) as a template with 5′ digoxigenin (DIG)-labelled primers OPccd1 (5′-GATTACGAATTCGAGCTCGG) and OPccd2 (5′-CAGTGCCAAGCTTGCATGCC). Binding reactions (15 µl) were carried out at 30°C for 20 min with purified proteins, 50 ng of probe (22 nM) and 1 µg of sonicated salmon sperm DNA in 30 mM Tris-HCl, pH 7.5, 200 mM NaCl, 15 mM MgCl2, 3 mM EDTA, 150 µg ml−1 bovine serum albumin and 0.3 mM dithiothreitol (DTT). Electrophoresis was performed in Tris borate–EDTA buffer at 80 V in a 5% polyacrylamide gel. The gel was pre-electrophoresed for 1 h at 4°C. After electrophoresis, the protein–DNA complexes were transferred onto a nylon membrane. Blots were developed with anti-DIG antibody conjugated with alkaline phosphatase using the CSPD chemiluminescent substrate (Boehringer Mannheim).

In vitro mutagenesis of plasmid DNA by 2-hydroxylamine

Randomly mutated pULB2548 was obtained according to the 2-hydroxylamine mutagenesis procedure as described previously (Isackson and Bertrand, 1985). Briefly, 20 µg of purified plasmid DNA was incubated in 1 ml of 0.8 M hydroxylamine-HCl, 50 mM sodium phosphate, pH 6, and 1 mM EDTA for 48 h at 37°C. The mutagenized DNA was diluted with five volumes of water and dialysed against 20 mM Tris-HCl, pH 8, 20 mM NaCl and 1 mM EDTA at 4°C. After dialysis, the DNA was ethanol precipitated, suspended in water and kept at 4°C for further use.

Site-directed mutagenesis by PCR

Site-directed mutagenesis based on PCR was performed as described previously (Ho et al., 1989). This procedure requires two primers (one forward and one reverse) carrying the desired mutation and two external primers overlapping the 5′ and 3′ regions of the ccd operon (ccd-For, 5′-CGATGATAAGCTGTCAAAC; and ccd-Rev, 5′-CATTGCTGCAGACTGGC). The primers containing the desired mutations are the following: ccdA2A-For (5′-GGCGCAGCGTATTACAGTG) and ccdA2A-Rev (5′-CACTGTAATACGCTGCGCC) for construction of the CcdAK2A mutant; ccdA4A-For (5′-GCAGGCTATTACAGTG) and ccdA4A-Rev (5′-CACTGTAATAGCCTGC) for construction of the CcdAR4A mutant; ccdA70K-For (5′-CGAGAACAAAGACTGG) and ccdA70K-Rev (5′-CCAGTCTTTGTTCTCG) for construction of the CcdAR70K mutant; ccdA70A-For (5′-CGAGAACGCGGACTGG) and CcdA70A-Rev (5′-CCAGTCCGCGTTCTCG) for construction of the CcdAR70A mutant; ccdA6A-For (5′-GAAGCAGCGTATTGCAGTG) and ccdA6-Rev (5′-CTGCCACTGTAATACGCTG) for construction of the CcdAT6A mutant; ccdA8A-For (5′-CAGCGTATTACAGTGGCAG) and ccdA8A- Rev (5′-CACTGTAATACGCTGCGCC) for construction of the CcdAT8A mutant; and ccdA10A-For (5′-CGTATTACAGTGACAGTTGCCAGCG) and ccdA10A-Rev (5′-CGCTGGCAACTGTCACTGTAATACG) for construction of the CcdAD10A mutant. In a first step, two fragments that partially overlap and contain the desired mutation were generated using either ccd-For and mutants-Rev primers or ccd-Rev and mutants-For primers. The two PCR products were gel purified and used as template DNA in a second PCR reaction with the two external primers (ccd-For and ccd-Rev). Final PCR products were analysed by agarose electrophoresis and cloned in pULB2548 plasmid. All the final constructs were sequenced to confirm the presence of the expected mutations.

Purification of CcdA, CcdA mutants and CcdB

The CcdA and mutant proteins were produced in SG22623 containing a pKK223-3 (Brosius and Holy, 1984) derivative carrying the ccdA or ccdAR4A genes under the control of the tac promoter. Mutant genes were PCR amplified with pULB2548 as a template using the following primers: ccdAR4A-EcoRI (5′-TTGTGAATTCTATGAAGCAGGCTATTACAG) and ccdA-PstI (5′-AGTCTCTGCAGTCACCAGTCCCTGTTCTC) to construct the CcdAR4A-overproducing plasmid. The wild-type or mutant PCR fragments were cloned into EcoRI and PstI sites of pKK223-3.

Bacteria containing the various overproducing plasmids were grown in TB broth at 37°C for 4 h. Expression of CcdA was then induced by the addition of 0.5 mM IPTG (isopropyl-1-thio-β-D-galactopyranoside). After 3 h of induction, bacteria were harvested and stored at −70°C until used. Frozen cells were suspended in 50 mM Tris-HCl, pH 8, 2 mM EDTA, 10% glycerol, 150 mM NaCl, 0.1 mg ml−1 AEBSF (4-(2-aminoethyl)benzenesulphonyl fluoride, HCl) and 1 µg ml−1 leupeptin. Cells were broken in a French pressure cell at 20 000 psi. After centrifugation for 1 h at 13 000 r.p.m., the supernatant was treated with 40% saturated ammonium sulphate. After centrifugation, the pellet was suspended in 50 mM Tris-HCl, pH 8, and applied to a Mono-Q anion exchange column equilibrated with the same buffer. Proteins were eluted using a linear gradient of NaCl. Fractions containing CcdA or mutant proteins that were eluted at 0.3 M NaCl were precipitated with 40% saturated ammonium sulphate. The pellet was suspended in 50 mM Tris-HCl, pH 8, 2 mM EDTA, 200 mM NaCl and 10% glycerol and applied to a Superdex 75 PG (16/90) gel filtration column. The column was run at 0.2 ml min−1. The wild-type CcdA and mutant proteins appeared homogeneous on SDS gel electrophoresis stained with Coomassie brilliant blue.

The CcdB protein was purified as described previously (Steyaert et al., 1993).

Gel filtration analysis

Analytical gel filtration was performed with a Superdex 75 PG (16/90) column using a buffer containing 50 mM Tris-HCl, pH 8, 2 mM EDTA, 200 mM NaCl and 10% glycerol. The flow rate was 0.25 ml min−1. Molecular weight markers used were cytochrome c (12 400), carbonic anhydrase (29 000) and albumin (66 000).

Native gel electrophoresis

Native electrophoresis was performed using the Laemmli buffer system, but SDS was omitted. The protein mixtures were mixed 3:1 (v/v) with sample buffer (50 mM Tris-HCl, pH 6.8, 10% glycerol and 0.1% bromophenol blue) and loaded on a 10% polyacrylamide gel.


We are grateful to C. Y. Szpirer and G. Maenhaut-Michel for discussions and comments, and to S. Gottesman for critical reading of the manuscript. We thank J. Backman for helping us in ITC experiments. This work was supported by grants from the Fonds National de la Recherche Scientifique (FNRS) and Action de Recherches Concertées. H.A. was supported by the Université Libre de Bruxelles (ULB), the Fondation Brachet and the Fonds Van Buuren. M.C. is Chercheur Qualifié at the FNRS. L.V.M. is Chargé de Recherches at the FNRS.