Corepression of the P1 Addiction Operon by Phd and Doc

Abstract

The P1 plasmid addiction operon encodes Doc, a toxin that kills plasmid-free segregants, and Phd, an unstable antidote that neutralizes the toxin. Additionally, these products repress transcription of the operon. The antidote binds to two adjacent sites in the promoter. Here we present evidence concerning the regulatory role of the toxin, which we studied with the aid of a mutation, docH66Y. The DocH66Y protein retained the regulatory properties of the wild-type protein, but not its toxicity. In vivo, DocH66Y enhanced repression by Phd but failed to affect repression in the absence of Phd, suggesting that DocH66Y contacts Phd. In vitro, a MalE-DocH66Y fusion protein was found to bind Phd. Binding of toxin to antidote may be the physical basis for the neutralization of toxin. DocH66Y failed to bind DNA in vitro yet enhanced the affinity, cooperativity, and specificity with which Phd bound the operator. Although DocH66Y enhanced the binding of Phd to two adjacent Phd-binding sites, DocH66Y had relatively little effect on the binding of Phd to a single Phd-binding site, indicating that DocH66Y mediates cooperative interactions between adjacent Phd-binding sites. Several electrophoretically distinct protein-DNA complexes were observed with different amounts of DocH66Y relative to Phd. Maximal repression and specificity of DNA binding were observed with subsaturating amounts of DocH66Y relative to Phd. Analogous antidote-toxin pairs appear to have similar autoregulatory circuits. Autoregulation, by dampening fluctuations in the levels of toxin and antidote, may prevent the inappropriate activation of the toxin.

Bacteriophage P1 lysogenizes Escherichia coli as a low-copy-number plasmid (24). Several mechanisms contribute to the maintenance of the P1 plasmid prophage. Replication (12), dimer resolution (6), and plasmid partition systems (1, 5) increase the probability that each daughter cell receives at least one plasmid.

The addiction operon, in contrast, increases segregational stability of the P1 plasmid by killing plasmid-free segregants. The operon encodes two products, a 126-amino-acid toxin (Doc) that kills plasmid-free segregants and an unstable 73-amino-acid antidote (Phd) that prevents host death while the plasmid is present (31). Plasmid loss prevents the synthesis of new antidote. Degradation of antidote by the host-encoded ClpXP protease liberates the toxin, which then poisons the plasmid-free segregant (33).

In addition to acting as antidote and toxin, Phd and Doc collaborate to autoregulate transcription of the addiction promoter. Expression of Phd partially represses transcription of the operon; coexpression of Doc with Phd enhances repression (35). In vitro, pure Phd binds to the promoter region, as indicated by electrophoretic mobility shift assays, and protects two adjacent palindromic sites in the promoter region from DNase I digestion (35). Mobility of the protein-DNA complex is further retarded in the presence of Doc, indicating that Doc may directly participate in the protein-DNA complex (35). Negative autoregulation of the operon may act homeostatically to prevent fluctuations in the levels of toxin and antidote (35). Without such regulation, fluctuations in the levels of antidote and toxin may inappropriately activate the toxin.

Analogous (although not necessarily homologous) antidote-toxin pairs (25, 51), such as CcdA-CcdB (16, 57, 58), Kis-Kid (identical to PemI-PemK) (49, 60), and HigA-HigB (59), also corepress transcription of their genes, indicating that transcriptional autoregulation involving both products may be a common feature of these small operons. The molecular mechanism of corepression has not been determined for any of these small operons.

Typically, coregulation of gene expression involves contacts between or among proteins that bind DNA or allosteric changes in the conformation of one such protein induced by another protein or by a small allosteric effector. For example, the P1 Bof protein stimulates operator binding by C1 and participates in the protein-DNA complex but does not bind to the DNA itself (32, 62), suggesting that Bof affects the conformation of C1. In a particularly well-studied example, the binding of cyclic AMP (cAMP) to the CRP protein induces a change in the conformation of CRP and thus increases its affinity for DNA (8, 19). Once bound to DNA, cAMP-Crp may engage in interactions with other DNA-binding proteins. An interaction between CytR and cAMP-Crp enhances the binding of CytR to DNA (27, 28) and thus excludes RNA polymerase (46). Binding of cytidine to CytR interferes with the contacts between CytR and cAMP-Crp and thus facilitates the release of CytR from the DNA (28). In the absence of CytR, RNA polymerase contacts cAMP-Crp, binds to the promoter, and initiates transcription (8, 61).

The interaction between two DNA-binding proteins can require extra factors. In some cases, a third adaptor protein links two DNA-binding proteins (20, 21, 30, 37, 44, 45, 56), whereas in other cases, an accessory protein affects the conformation of the DNA (41). Thus, IHF stimulates the binding of the P1 ParB protein to DNA but does not contact ParB. Rather, IHF bends DNA and thus facilitates cooperative interactions between flanking DNA-bound dimers of ParB (18, 23). Similarly, HU facilitates contacts between the flanking DNA-bound dimers of GalR (2, 3, 34). Thus, there is precedent for coregulatory mechanisms of great variety.

In this study we used a nontoxic mutant of Doc to facilitate the analysis of how Doc enhances repression of the P1 addiction operon by Phd. We found that Doc is not a repressor in its own right but enhances repression by binding to Phd and mediating cooperative interactions between adjacent Phd-binding sites. Additionally, we found that several alternative protein-DNA complexes can be formed. The exact form of the complex and, consequently, the specificity of DNA binding and the strength of repression are sensitive to the ratio of Phd and Doc.

MATERIALS AND METHODS

Media.

Cells were grown on Luria broth (LB) or LB agar supplemented as indicated previously (38). Glucose was added to 1.0%, where indicated, to improve protein expression and allow growth to high cell density. Transcription from P_BAD (22) and P_tac (15) promoters was induced, where indicated, with arabinose or isopropyl-beta-d-thiogalactopyranoside (IPTG). Selection for plasmids was accomplished by the addition of ampicillin (100 μg/ml), chloramphenicol (30 μg/ml), or spectinomycin (40 μg/ml) as needed (54).

Phage.

Lambda phage and lysogens were constructed and manipulated by standard techniques (4, 55). Single-copy transcriptional fusions were constructed by the method of Simons, Housman, and Kleckner (55). λRDM12 (35) contains a transcriptional fusion of the addiction promoter to lacZYA. λRDM11 contains the addiction promoter and the full coding sequence of phd followed by lacZYA. The DNA fragment encompassing the addiction promoter and the full coding sequence of phd flanked by EcoRI and BamHI restriction sites was produced by PCR with the oligonucleotide primers HAL13 (GGGAATTCTGATAGCCATCACCGGTGA) and HAL03 (GGGGATCCTCATTATCGGTTAACCAG). This DNA fragment was cloned into pRS415 (55) and transferred by homologous recombination to λRS45 to produce λRDM11.

Bacteria.

E. coli strains were constructed, propagated, and stored by standard techniques (38). All E. coli strains were derived from MC1061 (11) unless otherwise indicated.

Isolation of mutations in doc that abolish toxicity.

We used PCR to introduce an amber mutation in the fifth codon of doc. This provided supD-dependent toxicity. The DNA fragment containing the coding sequence of docS5(Am) flanked by EcoRI and HindIII restriction sites was produced by PCR with the oligonucleotide primers HAL25 (GGGATTTC ATG AGG CAT ATA TAG CCG GAA C) and HAL23 (GGGAAGCTTGCCATTAATCTACTCCGCAGAA). This DNA fragment was cloned behind P_tac in pKK223-3 (7) to generate pRDM039, and the plasmid was mutagenized by growth in a mutD5 strain (LE30 [54]). Mutagenized plasmid was introduced by transformation into a temperature-sensitive suppressing strain (MX397 [43]) at the permissive temperature to select for nontoxic versions of Doc. Candidates were tested by retransformation. Twelve independent mutants were sequenced.

Regulatory activity of nontoxic mutations.

In λRDM11, lacZ⁺ is transcriptionally fused to the addiction operon downstream of phd⁺. Alleles of doc were screened for the ability to enhance repression of this lacZ⁺ fusion in a suppressing strain.

Construction of additional plasmids.

Oligonucleotides HAL24 (GGGAATTC ATG AGG CAT ATA TCA CCG GAA G) and HAL23 were used in a PCR to flank doc⁺ with EcoRI and HindIII sites. The resulting EcoRI-HindIII fragment containing doc⁺ was cloned into pKK223-3 to place doc⁺ under control of the IPTG-inducible P_tac promoter. The plasmid was introduced into cells containing a source of lacI^q (placIQ, obtained from R. Kolodner) to repress transcription of doc⁺ and a source of phd⁺ (λRDM11) to neutralize the toxin produced by the basal expression of doc⁺.

In several instances, e.g., docH66Y55(Am), we corrected the amber mutation by PCR with oligonucleotides HAL24 and HAL23. The EcoRI-HindIII fragment containing docH66Y was cloned into pKK223-3 (7) to place docH66Y under control of the P_tac promoter and into pMAL-c2 (New England Biolabs) to generate a malE-docH66Y fusion under the control of the P_tac promoter.

The EcoRI-HindIII fragments carrying doc⁺ and docH66Y were also cloned into pBAD24 to place doc⁺ and docH66Y under control of the arabinose-inducible P_BAD promoter. A HincII-HindIII fragment containing P_tac-phd⁺ was cloned into pGB2 (13) to provide a compatible source of Phd.

β-Galactosidase assays.

β-Galactosidase assays were performed as described by Miller (38) on toluene-permeabilized cells. β-Galactosidase activity per absorbance of the bacterial culture at 600 nm was used as a measure of the transcriptional activity of the addiction promoter.

Expression and purification of MalE-DocH66Y.

Cells were grown in LB supplemented with 1% glucose. Expression of the fusion protein was induced with 0.3 mM IPTG. Cells were incubated for 2 h after induction and then harvested and lysed by sonication. The fusion protein was bound to a DEAE Sepharose (Pharmacia) anion-exchange column and eluted in a linear gradient of NaCl in 20 mM Tris (pH 8.0). The eluted fusion protein was bound to an amylose Sepharose column (New England Biolabs), washed with 20 mM Tris (pH 7.4)–200 mM NaCl–1 mM EDTA, and eluted in the same buffer supplemented with 10 mM maltose (36). The fusion protein appeared to be greater than 90% pure, as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and staining with colloidal Coomassie G-250. DocH66Y was generated by digestion of MalE-DocH66Y overnight at room temperature with Factor Xa, a site-specific protease (40) that removes most of the MalE moiety, leaving only four extra residues fused to the N terminus of DocH66Y. Complete digestion was confirmed by SDS-PAGE.

Expression and purification of Phd.

Phd was expressed and purified essentially as previously described (35). For the electrophoretic mobility shift experiments presented here, we used a preparation of Phd, about 90% pure as judged by SDS-PAGE, from a Mono-S cation-exchange column cartridge (Bio-Rad). Material from the leading half of the peak gave a single retarded band on single-site DNA; material from the trailing half gave an additional, less retarded band as well, indicating some heterogeneity in the structure or conformation of Phd. For experiments reported here, we used material from the leading edge of the peak. The molar concentration of the purified proteins was estimated from the absorbance, measured at 230 or 280 nM, and from the extinction coefficient, deduced from the amino acid composition (9, 29).

DNA templates.

Oligonucleotides were radiolabeled by incubation with [γ-³²P]ATP and polynucleotide kinase (Promega) at 37°C for 30 min. Polynucleotide kinase was inactivated by heating to 95°C for 5 min. The labeled oligonucleotide was mixed with a twofold molar excess of a complementary oligonucleotide, heated to 100°C, and slowly cooled to 20°C to generate radiolabeled double-stranded DNA. The single Phd-binding site was generated by annealing oligonucleotide ROY59 (CCCGCGTGTACACTTCTTCACATGTGCCGCC) and the complementary oligonucleotide ROY60. The double Phd-binding site was generated by annealing oligonucleotide ROY57 (CCCGCGTGTACACTTCTTGTGTACACCCGCC) and the complementary oligonucleotide ROY58. Annealed double-stranded DNA was separated from [γ-³²P]ATP by gel filtration over G25 spin columns (5′/3′) in Tris-EDTA.

Electrophoretic mobility shift assays.

Electrophoretic mobility shift assays were performed essentially as previously described (10). Phd, Doc, and 1 nM radiolabeled DNA were mixed and incubated at 20°C for at least 20 min in 20 mM Tris (pH 7.6), 100 mM NaCl, 100 μg of bovine serum albumin per ml, and 5% glycerol. Samples were then loaded, at 200 V, onto a 1 by 150 by 150 mm polyacrylamide gel in 0.5× Tris-borate-EDTA. Gels were run at 200 V for 60 min, dried under vacuum at 80°C on Whatman 3MM paper, and autoradiographed 12 to 16 h at 20°C on Kodak Biomax film. For quantitative purposes, bands were imaged and analyzed with a phosphoimager (Fujix BAS 2000 and MacBas computer software [Fuji]).

Hill constants.

Approximate Hill constants were deduced from quantitative analysis of electrophoretic mobility shift experiments. The Hill constant (the slope of a plot of log [bound DNA/free DNA]) versus log concentration (of Phd, DocH66Y, or both) describes the steepness of the DNA-binding curve with respect to the concentration of the protein ligand or ligands. The Hill constant cannot exceed the number of ligand molecules participating in the complex. Thus, the stoichiometry of the complex determines the limit, but not the actual value, of the Hill constant.

Determination of half-lives.

After incubation for 30 minutes or more to permit complex formation, aliquots were mixed with a 1,000-fold excess of unlabeled DNA and analyzed at timed intervals by electrophoretic mobility shift assays. We also analyzed a sample before the addition of unlabeled DNA as a control for complex formation and a sample in which unlabeled DNA was mixed with labeled DNA before the addition of the proteins as an equilibrium control to show that the unlabeled DNA could in fact compete. In some cases, the amount of complex was undetectable or very near the equilibrium level immediately after the addition of competing DNA. Since approximately 1 min was required for the complex to enter the gel, in these cases we can conclude only that the half-life is much shorter than 1 min. In other cases, a portion of the complexes was lost quickly, before the first time point, but the remaining portion was more stable. The quick loss of a portion of the complexes might be attributed to the effects of mixing and loading the sample. Alternatively, the initial loss might reflect heterogeneity, with some complexes having a short half-life and some complexes having a longer half-life. In such cases, we report the half-life of the longer-lived moiety.

Nondenaturing gel electrophoresis.

Samples of purified Phd, MalE-DocH66Y, or both were mixed with 1 volume of loading buffer (600 mM Tris [pH 8.8]–20% glycerol–0.1% bromophenol blue [Novex]), incubated at 20°C for 60 min, and then electrophoresed on a 4 to 20% polyacrylamide gradient gel (Novex) in Tris-glycine buffer (24 mM Tris–192 mM glycine [pH 8.5]) at 125 V for 130 min. Proteins were detected by staining with colloidal Coomassie brilliant blue G-250 (42) and were photographed.

RESULTS

Isolation of mutations in doc that abolish toxicity but not regulation.

To study the role of Doc in the absence of Phd, we isolated a nontoxic version of doc that retained its regulatory activity. Nontoxic versions of doc were selected as described in Materials and Methods. Twelve independent mutations were isolated and characterized by DNA sequencing, yielding eight different point mutations and two frameshift mutations. These nontoxic versions of doc were screened in the presence of phd⁺ for the ability to enhance repression of the addiction promoter (see Materials and Methods). Whereas most of the nontoxic versions of doc (L12P, A32fs, L82P, L84P, V89fs, and L118P) had lost regulatory activity, four point mutations (H66Y, H66R, D70N, and A76E) retained regulatory activity. One of these variants, docH66Y, which appeared to be most nearly like doc⁺ in its regulatory activity, was used in further experiments. For some experiments, we fused malE to docH66Y, thereby adding a proteolytically removable MalE tag to the N terminus of DocH66Y. The fusion improved expression of the protein, provided a tag for affinity purification, and assisted in the analysis of stoichiometry.

Effects of Doc and DocH66Y on repression.

To understand the role of Doc in repression, we provided Phd and Doc (or DocH66Y) from separate plasmids so that the effects of the proteins, separately and in combination, could be determined. In the presence of Doc and Phd, transcription was repressed 40-fold, while in the presence of Phd, transcription was repressed only 10-fold, indicating that under these conditions, Doc enhanced repression about fourfold. In the presence of Phd, DocH66Y enhanced repression as effectively as wild-type Doc (Table 1); yet, in the absence of Phd, DocH66Y had no effect on transcription of the addiction operon (Table 1), indicating that DocH66Y acts on Phd, not on DNA, to enhance repression. We tested these propositions in vitro.

TABLE 1.

Expression of the P1 addiction promoter^a

Relevant genotypeb	β-Galactosidase sp actc	Fold repression
lacIq PBAD	2,345	1
lacIq Ptac-phd+ PBAD	250	9
lacIq Ptac-phd+ PBAD-doc+	58	40
lacIq Ptac-phd+ PBAD-docH66Y	59	40
lacIq PBAD-docH66Y	2,319	1

^aExpression of the addiction promoter was measured with a transcriptional fusion to lacZ⁺. 

^bphd⁺ and doc⁺ or docH66Y were supplied on separate plasmids under the control of IPTG or arabinose-inducible promoters. Cells were grown in LB medium supplemented with ampicillin, kanamycin, spectinomycin, 0.2% arabinose, and 25 μM IPTG. 

^cβ-Galactosidase specific activity was measured as described in Materials and Methods. 

Binding of MalE-DocH66Y to Phd.

To look for an interaction between Doc and Phd, in the absence of DNA, we analyzed MalE-DocH66Y, mixtures of MalE-DocH66Y and Phd, and Phd alone, by nondenaturing PAGE. Protein was visualized with a colloidal Coomassie brilliant blue G-250 stain. As we increased the concentration of Phd (Fig. 1, lanes 1 to 9), we observed the diminution of the MalE-DocH66Y band and the appearance of a new band, corresponding to a complex of the two proteins. Only one new band was observed, given either excess MalE-DocH66Y (Fig. 1, lanes 4 to 6) or excess Phd (Fig. 1, lanes 8 and 9), suggesting either that the complex is unique or that alternative complexes are not resolved by this method. The observed complex appears to involve equimolar amounts of MalE-DocH66Y and Phd (Fig. 1, lane 7). Binding of Phd to Doc might be the physical basis for the capacity of Phd to neutralize the toxicity of Doc.

FIG. 1.

Electrophoretic mobility shift assay for protein-protein interactions. Phd, MalE-DocH66Y, and mixtures of Phd and MalE-DocH66Y were electrophoresed in nondenaturing conditions and stained with Coommassie blue as described in Materials and Methods. In nondenaturing gels, mobility is determined by the ratio of net charge and size, and consequently a complex typically migrates with a mobility intermediate to the mobility of its constituents.

Pure Phd protects two adjacent sites in the promoter region from DNase I digestion. In the presence of Phd and Doc, the mobility of promoter DNA was further retarded (35), yet only the two Phd-binding sites and no additional sequences were protected from DNase I digestion (data not shown). The apparent participation of Doc in the complex, combined with the failure to protect additional sequences or to repress transcription in the absence of Phd, suggested that Doc interacted directly with Phd but not with DNA. If so, then only the Phd-binding site or sites, but no additional flanking or intervening sequences, should be required for Doc to enhance DNA binding.

Effects of DocH66Y on the cooperativity and affinity with which Phd binds DNA.

To test this, we used a radiolabeled synthetic operator containing two identical Phd-binding sites with the same spacing as in the natural promoter (double-site DNA, Fig. 2A). In the presence of Phd, we observed three electrophoretic species, corresponding to free DNA, DNA with one site occupied, and DNA with two sites occupied (Fig. 2B). At intermediate concentrations of Phd, all three species were observed at once, indicating that occupancy of the two sites was independent (53). The binding of Phd to a palindromic site was mildly cooperative. The symmetry of the palindromic Phd site and of the DNase I protection patterns suggests that each palindromic Phd site is bound by two molecules of Phd. In the absence of DocH66Y, the DNA-binding curve has a slope (Hill constant) between 1.5 and 2.0 (Fig. 2E and Fig. 3D and E). Since this Hill constant is greater than 1, it appears that at least two Phd species are involved in binding each site. Thus, we hypothesize that Phd is predominately monomeric at these concentrations and that two monomers of Phd bind cooperatively (as a dimer) to each palindromic site.

FIG. 2.

Equilibrium binding of Phd and DocH66Y to double-site DNA, as determined by electrophoretic mobility shift assays. (A) Sequences of natural-site, double-site, and single-site DNAs. (B) Binding of Phd to radiolabeled double-site DNA. (C) Binding of DocH66Y to radiolabeled double-site DNA. (D) Binding of Phd and DocH66Y to radiolabeled double-site DNA. (E) Quantitation of electrophoretic mobility shift assays, showing binding of Phd to radiolabeled double-site DNA in the absence (open circles) or presence (open squares) of 88 nM DocH66Y. Radiolabeled double-site DNA (at 1 nM) was incubated as indicated with Phd and MalE-DocH66Y and then electrophoresed and quantitated as described in Materials and Methods.

FIG. 3.

Influence of DocH66Y on DNA binding by Phd, as determined by electrophoretic mobility shift assays. (A) Electrophoretic species on single-site DNA, at various concentrations of DocH66Y. (B) Electrophoretic species on double-site DNA at various concentrations of DocH66Y. (C) Electrophoretic species on double-site DNA in the presence of excess single site at various concentrations of DocH66Y. (D) Quantitation of binding of single-site DNA by various mixtures of Phd and DocH66Y. (E) Quantitation of binding of double-site DNA by various mixtures of Phd and DocH66Y. (F) Effect of the DocH66Y/Phd ratio on binding of double-site DNA in the presence of excess single-site DNA. Radiolabeled DNA was incubated with Phd and DocH66Y as indicated, electrophoresed, autoradiographed, and quantitated as described in Materials and Methods.

Although it did not bind to the double-site DNA (Fig. 2C), DocH66Y dramatically enhanced the ability of Phd to bind the double-site DNA (Fig. 2D). Thus, the in vitro effects of DocH66Y on DNA binding were consistent with the in vivo effects of DocH66Y on repression. In the presence of DocH66Y (Fig. 2E), the DNA-binding curve was steeper (Hill constant, 3.0 to 4.0), indicating that Phd bound DNA more cooperatively, and the curve was shifted to the left, indicating that Phd bound to the DNA with greater affinity (Fig. 2E). Interestingly, when the amount of DocH66Y no longer exceeded the amount of Phd, further increases in the amount of Phd produced successive shifts to less retarded complexes and the binding curve was not so steep (Fig. 2D and E).

How does Doc enhance DNA binding? We hypothesized that DocH66Y improves binding to a single Phd-binding site, mediates interactions between a pair of Phd-binding sites, or both. To test these possibilities, we made a single Phd-binding site of the same size and composition as the double site (Fig. 2A) and tested the effect of DocH66Y on the binding of Phd to single-site and double-site DNAs.

Binding of single-site DNA.

To test the effect of DocH66Y on the binding of Phd to a single site, we used a concentration of Phd that gave some free and some shifted DNA and then varied the concentration of DocH66Y. At low concentrations, DocH66Y had no effect on the binding of Phd to a single site. At saturating concentrations of DocH66Y, the complex on the single site was supershifted (further retarded) and DNA binding was enhanced (Fig. 3A). To better characterize this effect, we varied the concentration of Phd or mixtures of Phd and DocH66Y and measured binding to radiolabeled single-site DNA. Low concentrations of DocH66Y had no effect on the affinity or cooperativity of binding the single site. High concentrations of DocH66Y increased the apparent affinity but not the cooperativity (Hill constant) of binding to a single site (Fig. 3D).

Binding of double-site DNA.

Similarly, using a fixed amount of Phd and varying the concentration of DocH66Y, we tested the effect of DocH66Y on the binding of Phd to double-site DNA. Although they had no effect on binding single-site DNA, low concentrations of DocH66Y significantly enhanced complex formation on the double site and supershifted the protein-DNA complex (Fig. 3B). As the concentration of Doc was increased, additional supershifts were observed, and binding was further enhanced (Fig. 3B). The predominant supershifted species changed with the Phd/DocH66Y ratio. More retarded species were obtained either by increasing the concentration of DocH66Y (Fig. 3B) or decreasing the concentration of Phd (Fig. 2D).

To better characterize these effects, we varied the concentration of Phd or mixtures of Phd and DocH66Y and measured binding to a radiolabeled double site. While low concentrations of DocH66Y had no effect on the binding of a single site (Fig. 3D), these concentrations of DocH66Y were sufficient to markedly increase the affinity and cooperativity of binding to the double site (Fig. 3E). Higher concentrations of DocH66Y further increased the apparent affinity for the double site but did not further increase the cooperativity of binding to the double site (Fig. 3E).

Thus, DocH66Y had two distinct effects. At low concentrations of DocH66Y, binding of the double site was more cooperative than binding of the single site (Fig. 3D and E), indicating that DocH66Y was mediating cooperative interactions between Phd molecules bound to adjacent sites. Additionally, exclusively at high concentrations, DocH66Y increased the affinity with which Phd bound single-site DNA, perhaps by stabilizing the structure of Phd. At both high and low concentrations of DocH66Y, binding of the double site was more cooperative than binding of the single site (Fig. 3D and E), indicating that the second effect of DocH66Y (increased affinity for a single site) did not eliminate the first effect of DocH66Y (increased cooperativity of binding to a double site).

Effect of DocH66Y on specificity of DNA binding.

When the DNA-binding protein is limiting, occupancy of a particular site will be influenced by the presence of competing sites and by affinity for the site in question relative to that for competing sites. Since a Phd-binding site is approximately 8 bp (we do not know that all eight nucleotides are required), we expect the E. coli chromosome to contain a fair number of (nonspecific) single sites in addition to the pair of adjacent sites located in the addiction promoter of the P1 plasmid.

Given a mixture of single and double sites and limiting protein, increased affinity for the double site, observed at low concentrations of Doc, should increase binding to the double site at the expense of binding to the single site. Conversely, increased affinity for single sites, observed at high concentrations of Doc, might reduce binding to the double site.

In fact, we observed exactly this effect. In vitro, in the presence of excess unlabeled single-site DNA (Fig. 3C), as the concentration of DocH66Y was increased, the amount of free radiolabeled double-site DNA decreased, reached a clear minimum, and then increased sharply. Thus, in the presence of the single-site DNA, optimal binding of double-site DNA was observed with intermediate rather than maximal concentrations of DocH66Y, due to the changes in the relative affinities for single and double sites.

In order to verify that these changes in specificity were due to changes in the DocH66Y/Phd ratio rather than to changes in the absolute concentration of DocH66Y, we varied the concentrations of both Phd and DocH66Y and measured the efficacy of complex formation on a double site in the presence of excess unlabeled single-site DNA. For a given concentration of Phd, complex formation was maximal at a particular ratio of Phd and DocH66Y (Fig. 3F).

We found evidence of a similar effect in vivo (Table 2). In this experiment, phd⁺ was expressed from its natural promoter, and the activity of this promoter was indicated by expression of lacZ⁺, which was cloned immediately downstream of the addiction promoter and phd⁺. In the presence of doc⁺ or docH66Y, this transcriptional fusion was repressed approximately 10-fold more (with approximately 10-fold-less Phd). This enhanced repression can be attributed to the enhanced affinity and specificity of operator binding observed in vitro in the presence of docH66Y. Strikingly, higher levels of Doc, obtained by adding higher concentrations of IPTG, partially derepressed the operon and, at the highest concentration of IPTG, inhibited cell growth, indicating that the capacity of the antidote was exceeded under these conditions. Similar or greater derepression, without deleterious effects on growth, was observed after induction of docH66Y. The loss of repression observed in vivo at saturating concentrations of Doc or DocH66Y can be attributed to the loss of specificity for the operator DNA observed in vitro with saturating concentrations of DocH66Y relative to Phd.

TABLE 2.

Expression of the addiction promoter^a

Relevant genotypeb	β-Galactosidase sp actc with indicated IPTG concn (μM)
	1.6	5	16	50
lacIq Ptac	1,121	1,175	1,154	1,170
lacIq Ptac-doc	126	186	395	250d
lacIq Ptac-docH66Y	144	166	399	424

^aphd⁺ and lacZ⁺ were expressed from the addiction promoter. 

^bdoc⁺ or docH66Y was expressed from the IPTG-inducible P_tac promoter. 

^cβ-Galactosidase specific activity was determined 2.5 h after the addition of IPTG. 

^dInduction of doc⁺ with 50 μM IPTG arrested cell growth. 

Stoichiometry of protein-DNA complexes.

We used DocH66Y and MalE-DocH66Y to examine the stoichiometry of Doc in some of the protein-DNA complexes. Increasing amounts of DocH66Y relative to Phd gave increasingly retarded protein-DNA complexes (Fig. 3B). By using low concentrations of DocH66Y and MalE-DocH66Y, we optimized conditions for the formation of the least-supershifted complex on the double site (Fig. 4). Under these conditions, MalE-DocH66Y and DocH66Y had no effect on the binding of single-site DNA (Fig. 4). MalE-DocH66Y gave a complex that was similar but less-mobile than that of DocH66Y, indicating that the toxin physically participates in the protein-DNA complex. A mixture of MalE-DocH66Y and DocH66Y yielded both of the parental complexes, but no new complexes, indicating that the least-supershifted species contained a single unit or molecule of DocH66Y (or MalE-DocH66Y) (Fig. 4). By extrapolation, we expect that more-retarded complexes (Fig. 3B), observed at higher concentrations of DocH66Y, contain two, three, or possibly four molecules of DocH66Y. Higher concentrations of MalE-DocH66Y also yielded higher-order complexes, but MalE-DocH66Y was somewhat less effective than DocH66Y in forming these species (data not shown), indicating that the MalE moiety may hinder their formation.

FIG. 4.

Stoichiometry of DocH66Y in a protein-DNA complex. DocH66Y (D), MalE-DocH66Y (MD), and Phd were mixed as indicated, incubated with radiolabeled single- or double-site DNAs, electrophoresed, and autoradiographed as described in Materials and Methods.

Half-lives of protein-DNA complexes.

Equilibrium binding can be increased by increasing the rate of association, decreasing the rate of dissociation, or both. Since low concentrations of DocH66Y enhanced binding to the double site, but not to a single site, we hypothesized that Doc mediated positive interactions between adjacent Phd-binding sites. We expected that a positive interaction between sites would dramatically increase the half-life of the protein-DNA complex. We measured the half-lives of the various electrophoretically distinct complexes that were formed on single and double sites. Proteins were incubated with 1 nM labeled DNA to allow complex formation, challenged with 1,000 nM unlabeled DNA, and analyzed by electrophoretic mobility shift assays at time intervals. Controls confirmed the efficacy of complex formation and competition by the unlabeled DNA. Phd complexes on single sites, with or without DocH66Y, had half-lives much shorter than 1 min. Phd complexes on the double site also had half-lives much shorter than 1 min (Table 3). In the presence of DocH66Y, however, we observed longer-lived complexes on the double site. The lowest supershifted complex, containing a single molecule of DocH66Y, had a half-life of approximately 2 min (Fig. 5 and Table 3). The next supershifted complex, presumably containing two molecules of DocH66Y, had a half-life of 60 to 100 min. Larger complexes, presumably containing three or four molecules of Doc, were not appreciably more or less stable than complexes containing two molecules of DocH66Y (Table 3). The ability of DocH66Y to increase the half-life of double-site complexes without similarly increasing the half-life of single-site complexes indicates that DocH66Y mediates interactions between the adjacent sites.

TABLE 3.

Half-lives of protein-DNA complexes

DNA template	Interpretation of complex	Half-life of complex (min)
Single Site	2 Phd	<1
Single Site	2 Phd, 2 Doc	<1
Double Site	2 Phd	<1
Double Site	4 Phd	<1
Double Site	4 Phd, 1 Doc	1.8–3.1
Double Site	4 Phd, 2 Doc	60–100
Double Site	4 Phd, ≥3 Doc	50–100

FIG. 5.

Half-lives of protein-DNA complexes. Radiolabeled DNA (1 nM) was incubated with Phd and DocH66Y for at least 20 min at room temperature. Unlabeled DNA was then added to a final concentration of 1,000 nM, and the samples were incubated for the indicated time, electrophoresed, autoradiographed, and analyzed as described in Materials and Methods. In the precompetition control (P), unlabeled DNA was not added. In the equilibrium control (E), labeled and unlabeled DNA were mixed before the addition of Phd and DocH66Y.

DISCUSSION

Phd and Doc form a complex in solution.

Phd binds MalE-DocH66Y (Fig. 1), as indicated by native gel electrophoresis. Complex formation between Phd and Doc may be the physical basis for the neutralization of the toxicity of Doc. Interestingly, complex formation could also provide signal amplification; small changes in the total concentration of the antidote can produce large changes in the concentration of free toxin (Fig. 1; 16 and 30 pmol of Phd). Thus, given modest differences in the concentration and stability of toxin and antidote, high-affinity complex formation can provide a unified means to neutralize the toxin, sense small changes in the total antidote concentration, and generate large changes in the amount of free toxin.

Effects of DocH66Y on binding of Phd to DNA.

DocH66Y increased the half-life of a protein-DNA complex on a double-site DNA by greater than 60-fold (Table 3), yet it increased the affinity for double-site DNA by only about 10-fold (Fig. 2E and 3E), indicating that in the presence of DocH66Y the protein-DNA complexes may be slower to assemble as well as slower to disassemble. DocH66Y, by increasing the size and number of species participating in DNA binding, may also increase the steric and statistical costs of assembling the protein-DNA complex. In general, the comparison of protein-DNA half-lives on different DNA templates may be a particularly sensitive way to detect cooperative interactions.

Model for the architecture of the repressive complexes.

Prior work showed that Phd bound DNA. The symmetry of the Phd site and the Hill constant for binding of Phd support the hypothesis that two molecules of Phd cooperatively bind to a single Phd-binding site. Thus, in the model (Fig. 6), a monomer of Phd, represented by a right triangle, contacts a second monomer of Phd, and both monomers of Phd contact DNA, represented by a horizontal line. Electrophoretic mobility shift experiments indicate that Doc binds Phd but not DNA. Thus, Doc, represented by a rectangle, contacts Phd but not DNA. Since these three interactions are likely to be both direct and strong, they are drawn with long interfaces.

FIG. 6.

(A) Model of the addiction operon. The antidote, Phd, prevents host death and is subject to degradation by the host-encoded ClpXP protease (33). The toxin, Doc, is responsible for death on curing (plasmid loss). Both Phd and Doc participate in the transcriptional autoregulation of the addiction promoter (P). (B) Schematic interpretation of alternative protein-DNA complexes. DNA is represented by a horizontal line, Phd is represented by right triangles, and Doc is represented by rectangles. The addiction promoter contains two adjacent Phd-binding sites. Each palindromic site is bound by two molecules of Phd. Doc binds Phd but not DNA. Doc mediates cooperative interactions between molecules of Phd bound at adjacent sites.

Stoichiometry (Fig. 4) and titration experiments (Fig. 3E) indicate that a single molecule of Doc is sufficient to mediate cooperative interactions between the two Phd-binding sites. Thus, either Doc contacts two molecules of Phd, thereby directly bridging the two sites, or induces an allosteric change in Phd, thereby indirectly bridging the two sites. The first and simpler possibility is illustrated (Fig. 6B). Doc, represented by a rectangle, is shown contacting two molecules of Phd. The first contact, drawn with a long interface, is inferred from the observation that MalE-DocH66Y binds to Phd in solution. The second contact, drawn with a short interface, is inferred from the ability of a single molecule of Doc to mediate interactions between two Phd-binding sites (Fig. 4).

A second molecule of Doc further increased the half-life of the complex. For simplicity and symmetry, the drawing suggests that the interior molecules of Doc mediate the interactions between the Phd-binding sites and that the second molecule acts in the same way as the first molecule of Doc. Although this possibility is the easiest to draw, these simplifying assumptions are unsupported, and no alternative arrangement can be excluded. As drawn, a Doc-Doc contact might also contribute to the half-life of this complex. At present, however, we have no direct evidence for dimerization of Doc.

At (unphysiological) saturating concentrations, Doc increased the affinity of binding to a single site. We suggest that saturating amounts of Doc may stabilize the structure of Phd and may thus (rather passively) improve binding to single site. In contrast to the increased affinity for the double site, the increased affinity to the single site was not accompanied by an increased cooperativity of binding or by a (detectably) increased half-life of the protein-DNA complex.

The increased affinity for a single site was associated with decreased specificity for the double site in vitro (Fig. 3C), and with partial derepression in vivo (Table 2). Ordinarily, we would not expect to find saturating concentrations of Doc in vivo, while the addiction operon is still present (31). It is possible, however, that derepression (or partial failure to enhance repression), observed at saturating or near-saturating concentrations of Doc, may be a fail-safe mechanism that facilitates recovery from occasional upward fluctuations in the Doc/Phd ratio.

Are all of these interactions and effects mutually compatible? Doc did not inhibit binding of Phd to a single site; thus, it appears that the Phd-DNA, Phd-Phd, and primary Phd-Doc interactions are all mutually compatible. Since the participation of more than two molecules of Doc did not adversely affect the half-life of the complexes, it appears that the primary and secondary Phd-Doc contacts are also compatible. The model provides a plausible explanation of the observations and provides a conceptual framework for the genetic and biochemical dissection of the various interactions.

Function of autoregulation.

We have previously proposed that autoregulation has a homeostatic function. A loss of expression of the addiction operon, such as that which occurs upon loss of the operon, results in death. Smaller changes in the expression of the operon might also activate the toxin. By ensuring that the concentration of Phd and Doc is maintained within narrow limits, regardless of stochastic or environmental changes, autoregulation may prevent inappropriate activation of the toxin (35). By increasing the cooperativity of DNA binding, Doc makes repression more sensitive to changes in the concentration of Phd and Doc and thus ensures a more constant level of Phd and Doc.

Derepression, observed when the concentration of Doc exceeded the concentration of Phd, might help correct occasional variations in Phd/Doc ratio. A similar phenomenon has been observed in vitro with ParD and an altered form of ParE (26). Further experiments may clarify the significance of this property.

In a number of cases, e.g., Phd-Doc (Fig. 1), Kis-Kid (52), CcdA-CcdB (58), and ParD-ParE (26), there is evidence that toxin and antidote interact in solution and in the repressive complex. For Phd (35), Kis (49), HigA (59), and ParD (14, 17, 48), there is also evidence that the antidote can repress transcription or bind DNA in the absence of toxin. The similarity of these findings supports the hypothesis that the autoregulation of these small operons may be mechanistically similar.

Antidote and toxin appear to be synthesized in a fixed ratio, such that the synthesis of antidote exceeds the synthesis of toxin (31). Translational coupling of antidote and toxin, as demonstrated in the case of Kis and Kid (50), ensures that at least one molecule of antidote is synthesized immediately prior to the synthesis of each molecule of toxin.

Divergence.

The similarities in the gross structure and function of these operons suggests that they are derived from a common ancestor (25), yet their sequences are quite divergent, suggesting multiple origins, great age, or a fast rate of change. Theoretical analysis of the contribution of such elements to fitness indicates that they may be very useful in winning competitions between competing incompatible plasmids (39) (or between competing chromosomes, as with meiotic drive). In this role, the utility of the element decreases with its frequency. Consequently, these operons may be under positive selection for diversity, as has been proposed for colicins (47). In such a case, the general architecture and regulatory properties of the operons may be conserved, even as their primary structure diverges.