Autoregulation of the plasmid addiction operon of bacteriophage P1.

The P1 plasmid addiction operon increases the apparent stability of a plasmid that carries it by killing plasmid-free (cured) segregants. The operon consists of a gene encoding an endotoxin responsible for eath n uring (doc), preceded by a gene encoding a relatively unstable antidote that can revent ost eath (phd). When the copy number of the operon was increased, expression of a lacZ reporter fused to the promoter of the operon decreased, indicating that expression of the operon was stabilized by an autoregulatory circuit. Transcription of the lacZ reporter was repressed about 10-fold when phd, without doc, was expressed from an exogenous promoter. DNase I footprinting showed that Phd binds a perfect 10-base pair palindromic DNA sequence and, at higher concentrations, an adjacent, imperfect palindrome. The palindromic sites are located between the −10 region of the putative promoter and the start codon of phd. Electrophoretic mobility of DNA containing the promoter region was retarded in the presence of Phd and further retarded in the presence of Phd and Doc. When doc was co-expressed with phd, repression of the lacZ fusion was enhanced more than 100-fold. Thus, both products of the addiction operon participate in its autoregulation.

The P1 plasmid addiction operon increases the apparent stability of a plasmid that carries it by killing plasmid-free (cured) segregants. The operon consists of a gene encoding an endotoxin responsible for death on curing (doc), preceded by a gene encoding a relatively unstable antidote that can prevent host death (phd). When the copy number of the operon was increased, expression of a lacZ reporter fused to the promoter of the operon decreased, indicating that expression of the operon was stabilized by an autoregulatory circuit. Transcription of the lacZ reporter was repressed about 10-fold when phd, without doc, was expressed from an exogenous promoter. DNase I footprinting showed that Phd binds a perfect 10-base pair palindromic DNA sequence and, at higher concentrations, an adjacent, imperfect palindrome. The palindromic sites are located between the ؊10 region of the putative promoter and the start codon of phd. Electrophoretic mobility of DNA containing the promoter region was retarded in the presence of Phd and further retarded in the presence of Phd and Doc. When doc was co-expressed with phd, repression of the lacZ fusion was enhanced more than 100-fold. Thus, both products of the addiction operon participate in its autoregulation.
Bacteriophage P1, which lysogenizes Escherichia coli as an ϳ100 kilobase pair, low copy plasmid (1), is spontaneously lost at a frequency of about 10 Ϫ5 per generation (2). As is the case for other low copy number plasmids, this remarkable stability may be attributed to the combined effects of a partition system that ensures segregation of at least one plasmid to each daughter cell (reviewed in Refs. 3,4) and an addiction system that kills cells cured of the plasmid (5).
There are two P1-encoded addiction proteins. A 126-amino acid toxin, Doc, causes death on curing; an unstable 73-amino acid antidote, Phd, prevents host death while the plasmid is retained. The corresponding genes form an operon in which phd, the antidote gene, precedes doc, the toxin gene (5). Al-though the antidote is less stable than the toxin, due to degradation by the host-encoded ClpXP protease, it is also synthesized at a higher rate than the toxin, so as to ensure toxin neutralization. Upon loss of the P1 plasmid, proteolytic degradation of the unreplenished antidote unveils the toxic activity of Doc and causes post-segregational cell death (6).
Plasmid addiction elements functionally analogous to Phd/ Doc include CcdA/CcdB of F, the PemI/PemK of R100 (identical to Kis/Kid of R1), and ParD/ParE of RK2 (and RP4). Although their toxin targets may differ and homology among the analogous proteins is weak, the structure of the operons and the details of autoregulation are strikingly similar (7). In this work, we examine the role of Phd and Doc in the autoregulation of the P1 addiction operon and discuss the possible similarity of antidote proteins to each other and to well-studied DNA-binding proteins.

EXPERIMENTAL PROCEDURES
Media-Growth medium was LB broth or LB agar supplemented as needed with antibiotics (10) ampicillin, 100 g/ml, chloramphenicol, 25 g/ml, kanamycin, 30 g/ml, and spectinomycin, 40 g/ml, except that drug-resistant lysogens of P1 were grown in media containing 12.5 g/ml chloramphenicol or 15 g/ml kanamycin.
Construction of lacZYA Transcriptional Fusion to the P1 Addiction Promoter-A plasmid, pG3, that contains the P1 addiction operon, including its promoter, was used as a template for a polymerase chain reaction with oligonucleotide primers HAL13 (5Ј-GGGAATTCTGAT-AGCCATCACCGTGA-3Ј) and HAL15 (5Ј-GGGGATCCGGATTGCATA-AACACCTCGTGTA-3Ј). The resulting DNA product is flanked by primer-encoded EcoRI and BamHI restriction sites (underlined in primer sequences) and contains the first three codons of phd, the predicted promoter, and additional upstream sequences (nucleotides 5 to 374 (5)). This DNA was digested with EcoRI and BamHI and cloned into the EcoRI-BamHI sites of the pRS415 transcriptional lacZYA fusion vector (11) to produce pRDM064. The pRDM064 plasmid was isolated in strain BR6568, in which the P1 addiction operon, provided in trans from the compatible plasmid pG3, repressed expression of this potentially toxic, multicopy operon fusion. The fusion was transferred by homologous recombination from the pRDM064 plasmid to RS45, and integrated into the bacterial chromosome, as described previously (11).
Construction of Plasmids That Express phd and doc-A DNA fragment containing the coding sequences of phd (nucleotides 366 -592 (5)) flanked by primer-encoded EcoRI and HindIII sites was produced by a polymerase chain reaction using pG3 template and oligonucleotide primers HAL20 (5Ј-GGGAATTCATGCAATCCATTAACTTCCGTA-3Ј) and HAL22 (5Ј-GGGAAGCTTCCTCATTATCGGTTACAGTT-3Ј). The DNA product was digested with EcoRI-HindIII and cloned into the EcoRI and HindIII sites of the pKK223-3 P tac expression vector to generate pHAL20.
Similarly, a DNA fragment containing the coding sequences of phddoc (366 -976 (5)) flanked by primer-encoded EcoRI and HindIII sites was produced by a polymerase chain reaction using pG3 template and oligonucleotide primers HAL20 (described above) and HAL23 (5Ј-GG-GAAGCTTGCCATTAATCTACTCCGCAGAA-3Ј). The DNA product was digested with EcoRI and HindIII and cloned into the EcoRI and * This work was supported in part by the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  HindIII sites of the pKK223-3 P tac expression vector to generate pRDM032.
The P tac expression constructs were isolated in the presence of lacI q in order to repress maximally the cloned genes. The structures of pHAL20 and pRDM032 were verified by digestion with restriction enzymes and by sequencing with the primers described or with oligonucleotide primers HAL27 (5Ј-GCGCCGACATCATAACGGTTCTGGCAA-3Ј) and HAL28 (5Ј-GGGACCACCGCGCTACTGCCGCCAGGCAA-3Ј) which prime on pKK223-3 sequences flanking the cloning sites.
␤-Galactosidase Assays-␤-Galactosidase assays were performed as described by Miller (10) on toluene-permeabilized cells of a RDM12 lysogen of MC1061 or derivatives of this lysogen. The prophage bears a transcriptional fusion of the P1 addiction promoter (Pr92) to lacZYA.
Cell Extracts-Cells were grown in LB broth containing antibiotic as indicated, pelleted by centrifugation, and resuspended in approximately 1 to 2 volumes of lysis buffer containing 50 mM Tris, pH 8.0, 10% sucrose, 1 mM EDTA, 1 mM dithiothreitol, and 0.1 mM phenylmethylsulfonyl fluoride (a protease inhibitor). Cells were lysed by sonication, and soluble proteins were separated from the membrane fraction by centrifugation at 12,000 ϫ g for 30 min. Total protein was quantitated by BCA protein assay (Pierce).
Overexpression and Purification of Phd-Cells containing phd under the control of a repressed P tac promoter were grown in LB broth at 37°C to an absorbance of 0.5 at 600 nm. The expression of phd was then induced by the addition of isopropyl-␤-D-thiogalactopyranoside to a final concentration of 0.3 mM. The cells were further incubated for 1 h and then harvested by centrifugation for 10 min at 1,500 ϫ g at 4°C. The cell pellet was resuspended in 10 mM Tris/HCl, pH 7.4, 200 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol. Cells were broken by sonication; soluble proteins were separated from the membrane fraction by centrifugation, and the soluble fraction was loaded onto an FPLC S-Sepharose column equilibrated with 50 mM Tris/HCl, pH 7.4, 50 mM NaCl, 0.1 mM EDTA, and 1 mM dithiothreitol. Phd was eluted from the S-Sepharose column at approximately 600 mM NaCl in a linear gradient from 50 mM to 1 M NaCl. Fractions were electrophoresed in a denaturing 20% polyacrylamide gel with a Tricine 1 running buffer, which improves the resolution of small proteins (12). The resulting gels were stained with Coomassie Brilliant Blue in order to visualize Phd and contaminating proteins. Phd was precipitated from pooled fractions by adding ammonium sulfate to 70% of saturation. The precipitate was resuspended in 50 mM Tris/HCl, pH 7.4, 500 mM NaCl, 10% glycerol, 0.1 mM EDTA, and 1 mM dithiothreitol and loaded onto a Sepharose 12 HR 10/30 (FPLC) gel filtration column. The 8.1-kDa Phd protein eluted in a single peak. Quantitative total amino acid analysis, normalized for the composition of the protein, was used to determine the molar concentration of Phd. Samples of this preparation were used in DNA mobility shift and DNase I protection experiments. A small portion of this Phd preparation containing about 20 g of protein was resuspended in 6 M guanidine hydrochloride and further purified over a high performance liquid chromatography C18 reverse phase column. High performance liquid chromatography-purified Phd was collected in a single fraction and subjected to amino-terminal sequence analysis on a 477A protein sequencer (Applied Biosystems). The 23-amino acid amino-terminal sequence MQSINFRTARGNLSEVLNNVEAG determined by analysis matched the predicted amino-terminal sequence of Phd (5).
Radiolabeled DNA Fragments-Three DNA fragments were used in DNase I protection or DNA mobility shift experiments. A 403-bp DNA fragment, containing the P1 addiction promoter and additional sequences (nucleotides 5 to 391 (5)), and flanked by primer-encoded restriction sites, was produced by a polymerase chain reaction with primers HAL13 (5Ј-GGGAATTCTGATAGCCATCACCGTGA-3Ј) and HAL14 (5Ј-GGGGATCCGCGGTACGGAAGTTAATGG-3Ј). The DNA fragment was radiolabeled by cleaving the products with the restriction endonuclease EcoRI and subsequent filling in of the 5Ј-protruding end with Klenow polymerase, dTTP, and [␣-32 P]dATP.
Alternatively, one primer was radiolabeled with T4 polynucleotide kinase and [␥-32 P]ATP and then used with a second unlabeled primer in a polymerase chain reaction to generate a labeled DNA fragment. Using this technique, a 189-bp fragment (nucleotides 252-440 (5)), in which the P1 addiction promoter was centrally located, was generated with oligonucleotide primers ROY47 (5Ј-GATGCCCGGAGTGGAGAGTTT-GAT-3Ј) and ROY48 (5Ј-CTCTTCCCCGGCTTCAACATTGTT-3Ј). The radiolabeled DNA fragments were purified on 6% polyacrylamide gels in Tris-borate-EDTA buffer and eluted from gel slices in Maxam and Gilbert buffer (13). For mobility shift experiments and DNase I protection experiments, radiolabeled probe was used at a final concentration of approximately 0.1 and 3.0 nM, respectively.
DNA Mobility Shift Assays-DNA mobility shift assays were performed essentially as described previously (14,15). Cell extracts or purified Phd was incubated with approximately 0.1 nM radiolabeled DNA for 20 min at room temperature in 20 mM Tris/HCl, pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, 50 g/ml bovine serum albumin, 100 mM KCl, 31 g/ml calf thymus DNA, and 6% glycerol, electrophoresed on a 6 or 11% polyacrylamide gel in Tris/borate/EDTA buffer, which was then dried and autoradiographed.
DNase I Footprinting-Probing with DNase I was essentially as described previously (16). Purified Phd and radiolabeled DNA were incubated together for 20 min at room temperature in a solution containing 20 mM Tris/HCl, pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, 50 g/ml bovine serum albumin, 100 mM KCl, and 31 g/ml calf thymus DNA. The formation of complexes was then probed by adding divalent  (17), was run alongside DNase I-treated DNA in order to localize the footprint. Sequence Alignments-Protein sequences were compared using the gap program of the Genetics Computer Group (version 8) which is based on the method of Needleman and Wunsch (18). Matches were scored with the default Dayhoff table normalized for use with this program. Default values of 3.0 and 0.1 were used for the gap creation and gap extension penalties, respectively. In order to assess the significance of the alignment scores, one protein in each pair was aligned to 100 randomized versions of the other protein. Scores were standardized by subtracting the mean randomized score from the real score and dividing the difference by the standard deviation, estimated from the randomized trials. For group analysis, we determined the mean standardized score of the group and calculated the standard error as reciprocal of the square root of the size of the group.

Transcriptional Regulation of the P1 Addiction Operon-The
promoter of the P1 addiction operon was fused to lacZYA, and a single copy of the fusion was integrated into the chromosome. This transcriptional fusion produced about 5200 Miller units of ␤-galactosidase in the absence of the P1 addiction proteins (Table II). A P1 prophage carrying the addiction operon repressed expression of the fusion approximately 40-fold, indicating that significant levels of repression are achieved in the natural context of the P1 addiction operon. A plasmid, pG3, which carries the P1 addiction operon and is maintained at ϳ8-fold higher copy number than a P1 prophage (1,19) repressed expression of the fusion approximately 400-fold, or roughly 10-fold more than was observed with the low copy P1 prophage.
In order to test whether both proteins were involved in repression, yet avoid the complications of autoregulation, we constructed plasmids in which Phd or Phd and Doc were provided from a repressed P tac promoter. Phd and Doc, supplied from such a construct, repressed expression of the transcriptional fusion to the addiction promoter more than 1000-fold (Table II). Phd alone, supplied from the same repressed P tac promoter, repressed the addiction promoter by about 10-fold. These results indicate that, although Phd is sufficient for repression, Doc dramatically enhances repression.
Binding of Phd and Doc to the P1 Addiction Promoter-In order to characterize the mechanism of repression, crude cell extracts of cultures expressing either no P1 genes, phd alone, or phd and doc, were assayed for binding activity to a 403-bp DNA fragment containing the addiction promoter. Control ex-tracts lacking P1 products did not interact with the labeled promoter fragment (Fig. 1A, middle lanes). DNA retardation was observed with extracts containing Phd (Fig. 1A, right lanes), consistent with the finding that Phd gives modest repression in vivo. Extract containing both Phd and Doc (Fig. 1A, left lanes) retarded the promoter fragment more than extracts containing only Phd. Furthermore, the Phd-and Doc-induced transition from unretarded to fully retarded DNA did not appear to involve a stable intermediate species, such as that produced by Phd alone. The absence of appreciable DNA⅐Phd complexes in extracts that presumably contain Phd in excess of Doc suggests that the DNA⅐Phd⅐Doc complex is more stable than the DNA⅐Phd complex. Stabilization of the repressive complex may be the mechanism by which Doc enhances repression.
Purification and Activity of Phd-Phd was overexpressed and purified by cation exchange chromatography, ammonium sulfate precipitation, and gel filtration chromatography, as described under "Experimental Procedures." Purified Phd was estimated to be approximately 90% pure by silver staining (Fig.  2), and total amino acid analysis was consistent with the predicted composition of Phd. Purified Phd had DNA binding activity as detected by gel mobility shifts (Fig. 1B).
DNase I Footprint of Pure Phd on the phd-doc Promoter-In order to identify the DNA sequences to which Phd bound, we tested the ability of Phd to protect the promoter from DNase I digestion. These footprinting experiments were performed with purified Phd and a 189-bp fragment of DNA in which the putative promoter was centrally located. DNase I footprints showed protection of a perfect 8 -10-bp palindrome at 2 and 10 nM Phd. At 50 nM Phd, a second imperfect palindrome was also protected (Fig. 3A). Similar protection patterns were observed on both DNA strands. The centers of the perfect and imperfect palindromes are separated by 13 bp. The two palindromic sites span the region between the Ϫ10 region of the putative promoter and start codon of phd (Fig. 3B).
Phd Binding Sites-The footprinting experiments suggested that a single palindromic site would be sufficient for recognition by Phd. As expected, a 27-bp synthetic DNA fragment containing the perfect 10-bp palindrome, flanked by restriction site linkers, was retarded by Phd in the same manner as a 189-bp fragment containing the whole promoter region (Fig.  1B, 0 -21 nM Phd). To facilitate resolution of the small 27-bp DNA fragment, we used a higher percentage of acrylamide and increased the voltage at the beginning of the run. Under these conditions, the Phd⅐DNA complex appeared as a doublet (Fig.  1B, 21 nM Phd), which we interpret to be two alternative forms Pr92-phd doc, modest copy 9 390 placI q , pKK223-3 lacI q , control vector 4400 (1) placI q , pRDM032 lacI q , P tac -phd doc 2.2 2000 placI q , pHAL20 lacI q , P tac -phd 500 9 a ␤-Galactosidase assays were performed in a RDM12 lysogen of MC1061, or derivatives of this lysogen, bearing the indicated plasmids. The prophage in each strain bears a transcriptional fusion of the P1 addiction promoter to lacZYA. b ␤-Galactosidase-specific activity (Miller units) was averaged from four samples taken from cells growing exponentially in antibiotic-supplemented LB broth at 30°C. The standard deviation was less than 10% of the mean, except in the case of the strain containing pG3, in which case the standard deviation was 5 Miller units (55% of the mean).
c Relative repression is ␤-galactosidase-specific activity of the appropriate (unrepressed) control strain divided by the ␤-galactosidase-specific activity of the experimental strain.
of one molecular complex. Between 21 and 105 nM Phd, the 189-bp fragment containing the complete promoter region underwent an additional shift (Fig. 1B). This second shift probably reflects the filling of the second site, as observed in footprinting experiments. Consistent with this interpretation, the 27-bp fragment of DNA, which contains only a single palindromic site, did not undergo the second shift.

DISCUSSION
Physiology of Autoregulation-The plasmid addiction operon of bacteriophage P1 contains two genes, phd and doc, encoding, respectively, an antidote and a toxin. A chromosomally integrated, transcriptional lacZ fusion to the putative promoter of the P1 addiction operon was repressed about 400-fold when the addiction operon was furnished in trans on a moderate copy plasmid (Table II, pG3), indicating that transcription of the operon is negatively regulated by one or more products of the operon. Since the lacZ fusion was repressed approximately 40-fold in P1 lysogens, it appears that transcriptional repression occurs under physiological conditions.
In principle, autoregulation could buffer the cellular expression of the plasmid addiction operon against fluctuations in plasmid copy number. Evidence that this might be the case for P1 comes from the observation that when the copy number of the addiction operon was increased (from that of P1 prophage to that of pG3), expression of the operon decreased. The suggestion that levels of addiction proteins remain largely unaffected by variations in plasmid copy number, short of plasmid elimination, is consistent with the proposed post-segregational function of the operon. Autoregulation ensures that should plasmid copy numbers dip to levels that incur the risk of plasmid loss, the addiction system will remain fully primed.
Role of Phd in Autoregulation-Expression of phd, without doc, was sufficient to repress expression of a lacZ fusion to the addiction operon almost 10-fold (Table II). Consistent with this observation, purified Phd protected a region in the putative promoter from digestion with DNase I. The protected region, located between the Ϫ10 region of the putative promoter and the start codon of phd, encompasses a perfect and an imperfect palindrome (Fig. 3). These palindromic sites are 8 -10 bp in length and are centered 13 bp apart. The palindromic nature of the DNA sites protected from DNase I suggests that Phd, like many DNA-binding proteins, might bind as a dimer. DNA retardation and DNase I footprinting experiments show that a single perfect palindrome is sufficient for recognition by Phd (Fig. 1B, Fig. 3).
Similarity of Phd to Other Antidote Proteins and to Other Known DNA-binding Proteins-Phd and other plasmid-encoded antidote proteins, such as CcdA, PemI, and ParD, exhibit certain functional and structural similarities to each other and to well-studied members of the ␤-sheet family of DNA-binding proteins (20,21), such as MetJ (22) and Arc (23). MetJ is a small dimeric repressor that binds cooperatively to two or more adjacent, palindromic 8-bp sites (24). Similarly, Arc dimers bind cooperatively to two adjacent (but asymmetric) sites (25). In both cases, specific DNA contacts of the dimer are mediated by a short, two-strand, antiparallel ␤-sheet which is inserted into the major groove of the DNA site. The antiparallel ␤-sheet of the dimer is formed by the amino-terminal regions of the constituent monomers.
Predictions of secondary structure (26,27) suggest that there is a short ␤-sheet region near the amino terminus of Phd, followed by two or three ␣-helical domains. Similarly, predictions of secondary structure of CcdA indicate the presence of one or two ␤-sheets at the amino terminus, followed by one or more ␣-helices (28). A truncated CcdA protein lacking the amino-terminal 31 amino acids loses autoregulatory activity (28) but retains antidote activity (29), indicating that the amino terminus of CcdA may be specifically involved in DNA binding. ParD of RP4 binds DNA and may also be structurally similar to members of the ␤-sheet family of DNA-binding proteins (30).
Four antidote proteins involved in autoregulation (Phd, CcdA, PemI, and ParD) and two ␤-sheet DNA-binding proteins (Arc and MetJ) were compared, in all pairwise combinations, using the method of Needleman and Wunsch (18). Alignment scores were standardized by subtracting the mean score from 100 alignments using randomized protein sequences of the same composition and dividing by the estimated standard deviation from the randomized trials. Only one pair of proteins, CcdA and PemI, is significantly similar, as previously noted (31). As a group, the antidote proteins are slightly similar to each other (p Ͼ 97%) and to ␤-sheet DNA-binding proteins (p Ͼ 96%). Alignment scores fail to reflect the structural similarity of Arc to MetJ. The involvement of the antidotes in transcriptional repression and their (weak) similarity to each other and to ␤-sheet DNA-binding proteins is consistent with the hypothesis that the antidote proteins can bind to DNA via ␤-sheet structures.
Role of Doc in Autoregulation-Although expression of Phd alone was sufficient to repress a lacZ fusion to the P1 addiction promoter, repression was enhanced more than 100-fold when Doc and Phd were co-expressed (Table II). We have not yet been able to determine whether the toxic Doc protein can repress the promoter in the absence of Phd. Extracts containing Phd and Doc caused a greater retardation of the electrophoretic mobility of DNA bearing the promoter than did extracts containing Phd without Doc. Most likely, Doc participates directly in the repressive complex by binding Phd or DNA or both.
Transcriptional Regulation of Operons Analogous to the P1 Addiction Operon-A number of large plasmids carry operons encoding antidote/toxin pairs that enhance plasmid stability by killing or arresting the growth of plasmid-free segregants.
Where studied, these operons have been shown to be under negative, transcriptional autoregulation. In the case of ParD/ ParE of RK2, ParD binds DNA at micromolar concentrations (32), but it is not clear whether the toxin, ParE, affects repression (30,33,34). In the case of CcdA/CcdB of F (28,(35)(36)(37) and PemI/PemK of R100 (38,39), both antidote and toxin are required for full transcriptional repression and DNA binding activity. At the concentrations tested, the antidote, CcdA, did not bind to its promoter region in the absence of CcdB but in vivo, CcdA, much like Phd, represses expression of its promoter 5-fold (35). In the case of PemI/PemK, repression and DNA binding are detected only in the presence of both toxin and antidote.
It seems possible, given the functional and structural similarity of these operons, that the ways in which they are autoregulated may be similar. A simple, unifying hypothesis is that 1) the antidote specifically interacts with DNA (shown in two out of four cases), possibly via ␤-sheet structures, and 2) the toxin enhances repression (shown in three out of four cases). How the P1 toxin contributes to repression is currently under study.