Regulatory Cross-talk in the Double par Locus of Plasmid pB171*

The double par locus of Escherichia coli virulence factor pB171 consists of two adjacent and oppositely oriented par loci of different types, called par1 and par2. par1 encodes an actin ATPase (ParM), and par2 encodes an oscillating, MinD-like ATPase (ParA). The par loci share a central cis-acting region of ≈200 bp, called parC1, located between the two par loci. An additional cis-acting region, parC2, is located downstream of the parAB operon of par2. Here we show that ParR of par1 and ParB of par2 bind cooperatively to unrelated sets of direct repeats in parC1 to form the cognate partition and promoter repression complexes. Surprisingly, ParB repressed transcription of the noncognate par operon, indicating cross-talk and possibly epistasis between the two systems. The par promoters, P1 and P2, affected each other negatively. The DNA binding activities of ParR and ParB correlated well with the observed transcriptional regulation of the par operons in vivo and in vitro. Integration host factor (IHF) was identified as a novel factor involved in par2-mediated plasmid partitioning.

Bacterial plasmids have been used extensively as model systems in the study of DNA segregation. This is because plasmids encode centromere-like loci, also called partitioning (par) loci, that ensure stable propagation of their replicons (1, 2) Plasmidborne par loci invariably consist of two proteins encoded by a bicistronic operon and one or more cis-acting centromere regions where the proteins act. The first gene in the operon (called parA, parF, or parM) encodes an ATPase. The second gene (called parB, parG, or parR) encodes an adaptor protein that binds to its cognate centromere and thereby forms the "partition complex" that, in turn, is recognized by the ATPase. Based on the ATPase, all par loci are divided into two types: Type I loci, which encode Walker box ATPases related to the MinD family, and Type II loci, which encode actin-like ATPases (3)(4)(5). Based on gene sizes and arrangement, Type I loci are subdivided in Type Ia and Ib. Type Ib ATPases are generally smaller than those of Type Ia, which include ParA of plasmid P1 and SopA of plasmid F. The Type Ib ATPases lack the DNAbinding helix-turn-helix (HTH) domain found in the N-termi-nal part of the longer Type Ia ATPases (5,6). Thus, contrary to the Type Ia ATPases, the Type Ib ATPases do not themselves specifically bind DNA. The molecular mechanism specified by Type II loci is well understood (1,7,8). By contrast, the molecular mechanism behind the common and more efficient Type I loci has been more difficult to understand (2,9).
The Escherichia coli virulence plasmid, pB171, has two par loci designated par1 (Type II) and par2 (Type I) with a peculiar genetic arrangement. The oppositely oriented par1 and par2 loci share a common cis-acting region, parC1, of Ϸ200 bp only (see Fig. 1, A and B) (10). parC1 contains 17 6-bp direct repeats (called B1 to B17) organized in two clusters. As described previously (10), parC1 expresses both par1-and par2-specific incompatibility, indicating that the full-length parC1 fragment contains centromere-like sites for both par loci. The parC2 region downstream of parB also expresses par2-specific incompatibility and contains 18 6-bp direct repeats (B18 -B35) that are related to the B repeats in parC1. In contrast, parC2 does not express par1-specific incompatibility (10). Based on sequence similarity, the B repeats could be divided into two subclasses (I and II) (see Fig. 1, B and C) (10). Individual repeats belonging to the two subclasses of B repeats are interspersed among each other in parC1 and parC2. In analogy with other Type I par loci (2,9), the genetic organization of parC1 and parC2 suggests that ParB binds to the B repeats. parC1 also harbors the P1 and P2 promoters that transcribe par1 and par2 (Fig. 1, A and B) (10).
ParA of par2 is a Walker box ATPase, which forms filaments that oscillate in spiral-shaped structures over the nucleoid (10,11). In the presence of par2, plasmids localize to midcell and/or cell quarters, a pattern that is dependent on ParA spiral formation and oscillation (11). Recently, we proposed a working model for how an oscillating and filament-forming protein can localize plasmids primarily at midcell and quarter-cell positions (12).
The unusual genetic arrangement of the double par locus of pB171 stimulated an analysis of parC1 and parC2. We show here that both ParB and ParR bind to parC1 and that ParB binds to parC2. ParB recognizes the 6-bp direct repeats located within parC1 and parC2. Interestingly ParR recognizes two 10-bp direct repeats located in parC1 just upstream of parM. The ability of ParB and ParR to bind to parC1 is consistent with their transcriptional regulation of the two par operons, i.e. ParR regulates the par1 operon and parB the par2 operon. Surprisingly, however, ParB represses the par1 operon as efficiently as ParR, suggesting that par2 is epistatic to par1. This is, to our knowledge, the first example of cross-talk regulation between two different par loci.
Construction of Plasmids Containing parC1-lacZ Fusions-Thirteen mini-R1 derivatives containing parC1 fragments with varying numbers of repeats cloned in front of lacZ in the transcriptional fusion vector pRBJ200 were constructed (see Table  1). In the pGE3100 series the P1 promoter reads into lacZ, and in the pGE3200 series, the P2 promoter reads into lacZ. The different truncated parC1 fragments were generated by PCR amplification using plasmid pB171 as the template (the names of the PCR primers used for the construction of the individual plasmids are shown in Table 1, and oligonucleotide sequences are given in Tables 2-4). All upstream primers contained an XhoI site, and the downstream primers contained a BamHI site. PCR fragments digested with BamHI and XhoI were cloned into the same sites of pRBJ200. A StuI-BamHI fragment of each pGE3000 series plasmid (harboring the cloned promoter fragment and the lacZYA genes) was subsequently cloned into the StuI-BamHI of plasmid pOU254, resulting in the plasmids used in the LacZ assays. The plasmids were transformed into the E. coli strain MC1000.
Construction of Plasmids Expressing ParR (pGE207) or ParB (pGE107)-DNA fragments containing parR and parB were PCR-amplified using pB171 as template. The primers used are listed in Table 1, and the DNA sequences of the oligonucleotides are given in Tables 2-4. A Shine-Dalgarno sequence (from parMR of R1, included in the upstream primer) was cloned in front of the parR and parB open reading frames. Upstream primers harbored a SacI site and downstream primers a HindIII site. These sites were subsequently used to clone the PCR fragments into the corresponding sites located downstream of the arabinose-inducible P BAD promoter of vector pBAD33. The pBAD33 derivatives were transformed into strains of MC1000 already harboring the pOU254 derivatives. The growth rate of almost all strains (except control strains containing the pBAD33 vector) decreased slightly upon arabinose addition, especially strains expressing ParB.
Construction of Plasmids Used for Overexpression of His 6 -ParR and His 6 -ParB-ParR and ParB were amplified from plasmid pB171 by using PCR with the following DNA primers: parB, B171-20 ϩ B171-21; parR, B171-22 ϩ B171-23 (sequences are given in Table 4). Upstream primers contained His 6 codons inserted between the start codon and the second codon of the ParR and ParB reading frames, respectively. The PCR products were digested with restriction enzymes BamHI and XhoI and cloned into the BamHI and SalI sites located downstream of the IPTG-inducible P A1/O4/O3 promoter of plasmid pMG25 thereby creating plasmids pGE121 (P A1/O4/O3 ::his 6 -parR) and pGE223 (P A1/O4/O3 ::his 6 -parB). DNA sequences of all PCR-amplified constructs were verified by DNA sequencing by using the CEQ2000 sequencing system provided by Beckman.
Protein Purification-E. coli strain Top10 harboring either pGE223 (P A1/O4/O3 ::his 6 -parB) or pGE121 (P A1/O4/O3 ::his 6 -parR) was grown in LB medium at 37°C to an A 450 of 0.8 before expression of recombinant protein was induced by the addition of isopropyl 1-thio-␤-D-galactopyranoside to a final concentration of 1 mM. Induction was continued for 4 h. All of the following steps of the purification procedure were performed on ice or in a 4°C cold room. Harvested cells were resuspended in 4 ml of lysis buffer/200 ml of culture. Egg LacZ Assay-Cells of strain MC1000 harboring one mini-R1 plasmid (pOU254 derivative) and one expression plasmid (pBAD33 derivative) were grown at 35°C in AB medium supplemented with 0.5% glycerol, 0.1% casamino acids, 0.1 g/ml thiamin, and antibiotics (50 g/ml chloramphenicol and 30 g/ml ampicillin). The generation time in this medium was about 50 min. Overnight cultures were diluted 50-fold into fresh medium with antibiotics and grown to an A 450 of ϳ0.04 -0.05. Then, 0.2% arabinose was added, and the cultures were grown for 3 h to an A 450 of ϳ0.4. Samples were collected for ␤-galactosidase activity measurements carried out according to Miller (14). The stability of the R1 plasmids containing the parC1-lacZ fusions were tested by plating on LA plates containing selection only for the pBAD33 derivatives in addition to 40 g/ml 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal). The R1 plasmids were found to be stable during the time course of the experiment. Prolonged growth of the cultures in the presence of arabinose resulted in reduced plasmid stability especially in strains expressing ParB.
Electrophoretic Mobility Shift Assay-DNA oligonucleotides were 32 P-end-labeled in a kinase reaction. 20 pmol of oligonucleotide was mixed with 1.5 l of 10ϫ polynucleotide kinase (PNK) buffer (1 M Tris-HCl, 100 mM MgCl 2 , 70 mM DTT, pH 8.0), 2 l of 10 Ci/l [␥-32 P]ATP (3000 Ci/mmol), 1 unit of PNK buffer, of PNK, and H 2 O to a total volume of 15 l. The reaction mixture was then incubated for 30 min at 37°C followed by 10 min at 70°C. End-labeled oligonucleotides were then used in a PCR reaction with an additional oligonucleotide designed to obtain the DNA fragment of interest.
Oligonucleotides used in construction of truncated parC1 fragments for the ParB electrophoretic mobility shift assay  Table 3).
Oligonucleotides used in construction of truncated parC1 fragments for the ParR EMSA are: fragment a, B171-PE2 ϩ  Table 2).
g/l sonicated salmon sperm DNA. The standard reaction mixture was mixed on ice followed by the addition of 32 P-endlabeled DNA fragments. Protein concentrations are given in the figure legends. We used a concentration of 5 nM 32 P-labeled DNA fragments throughout. The reactions were then incubated for 20 min at ambient temperature. After incubation, glycerol was added to a concentration of 5%, and reactions were analyzed by electrophoresis on a 0.5ϫ TBE (pH 7.5) 5% polyacrylamide gel in 1ϫ TBE running buffer (0.89 M Tris-base, 0.89 M boric acid, 0.02 M EDTA, pH 8.3) at 200 V for 1 1 ⁄ 2 h. The polyacrylamide gels were prerun at 200 V for 1 h, and the TBE running buffer was changed once every hour. Gels were dried on paper and exposed to a phosphor screen over night for imaging on a phosphorimaging device. In determination of the Hill coefficient, the amount (log 10 (X/(1 Ϫ X))) of bound parC1 (X) was plotted as a function of log 10 to the respective concentrations of ParB and ParR. Generation of Biotinylated Double-stranded parC1 Fragments for Surface Plasmon Resonance (SPR)-Biotinylated parC1-containing DNA fragments were produced by PCR using a 5Ј-end-biotinylated B171-39 clockwise primer and B171-40 counterclockwise primer (Table 3) with pGE2 as template DNA. B171-39 was biotinylated. The PCR product was purified using the GFX PCR DNA and Gel Band Purification Kit (Amersham Biosciences) as recommended by the manufacturer.
SPR Measurements-SPR was carried out using a Biacore 3000 instrument (Biacore AB, Uppsala, Sweden) with streptavidin-coated SA sensor chips (Biacore AB). Biacore measures the binding of molecules to ligands immobilized on sensor chips as real-time changes in the optical SPR phenomenon that is caused by changes in the refractive index near the sensor chip. The refractive index changes are linearly dependent on mass binding to the sensor chip, and hence the amount of analyte that binds the immobilized ligand can be calculated. Because the injections of analyte supply constant concentrations, the binding kinetics is monitored. When the flow is changed to buffer without analyte, the dissociation kinetics is monitored, as bound analyte leaves the immobilized ligand. For the binding studies, 150     the addition of 1 l of DNase (1 unit) and were then incubated an additional 15 min at 37°C. Reactions were stopped by the addition of formamide loading buffer and boiled for 10 min at 99°C followed by denaturing PAGE on a 5% polyacrylamide gel. The gel was dried on paper and exposed to a 32 P-sensitive screen overnight for imaging on a PhosphorImager. Numbers in the two following columns represent repression -fold when parR and parB were present in trans in p15 replicons (pGE107 and pGE207, respectively). Cells of MC1000 carrying the appropriate plasmids were grown at 35°C in glycerol minimal medium containing antibiotics and 0.2% arabinose to induce the pBAD promoter upstream of the parR and parB genes. FEBRUARY 2, 2007 • VOLUME 282 • NUMBER 5

JOURNAL OF BIOLOGICAL CHEMISTRY 3139
Oligonucleotides Used for PCR-These primer sequences are listed in Tables 2-4.

RESULTS
The Divergent P1 and P2 Promoters Are Negatively Coupled-As described above, parC1 contains the B1-B17 repeats located in two clusters that are potential operators for ParB binding (Fig. 1B). Further inspection of the parC1 region revealed two 10-bp direct repeats, R1 and R2, separated by 31 bp (corresponding to three helical turns of the DNA helix). R1 and R2 are located between the P1 promoter and parM, and the distance between R2 and B1 is only 9 bp (Fig. 1B). To investigate whether the R and B repeats are involved in regulation of the divergent P1 and P2 promoters in parC1, we constructed two series of lacZ transcriptional fusions containing either the entire parC1 region or truncations thereof (Fig. 2). In the absence of ParR and ParB, P1 and P2 had comparable, strong activities (Fig. 2, Aa and Ba). In the P1 series, deletion of P2 resulted in increased P1 activity (Fig. 2A, c and d). Likewise, deletion of P1 from the P2 series also increased P2 activity (Fig. 2B, c-e). Deletion of the Ϫ10 and Ϫ35 boxes of P1 or P2 resulted in loss of promoter activity in both cases, consistent with previous primer extension analysis of P1 and P2 (10).
P2 Is Regulated by ParB, whereas P1 Is Regulated by both ParR and ParB-ParR or ParB were donated in trans, and changes in P1 and P2 activities were determined. ParR repressed P1 activity 3.5-fold (Fig. 2Aa). Repression of P1 by ParR was not seriously affected by the progressive deletion of B repeats ( Fig. 2A,  b-e), and the construct containing only R1, R2, and B1-B4 was fully repressible (Fig. 2Ae). A promoter-less lacZ fusion containing R1 and R2 only did not respond to ParR (Fig. 2Ag). ParB in trans repressed P1 to a similar extent ( Fig. 2A, a-e), and the activity from the promoter construct containing only R1, R2, and B1-B4 was still repressed. These results are consistent with the proposal that ParR represses P1 via binding to R1 and R2 and that ParB represses P1 via binding to the B repeats, notably B1-B4. This conjecture was supported by in vitro experiments described later.
In the intact parC1 fragment, ParB, repressed P2 activity 3.5-fold (Fig. 2Ba). The P2 activity of constructs carrying progressive deletions into the B1-B13 repeats were still repressible by ParB (Fig. 2B, b-e), and the construct that had B14 -B17 only was also repressed (Fig. 2Bf ). Somewhat unexpectedly, the constructs in which R1 and R2 were deleted were more responsive to repression by ParB. The reason for this behavior of the parC1 deletion mutations is unknown. In any case, these results show that ParB regulates P2 activity, most likely via binding to the B repeats (see below). ParR did not affect P2 activity (Fig. 2B).
In Vitro Analysis of P1 and P2 Regulation-DNA fragments encoding P1 or P2 (hatched bars in Fig. 3A) were used in in vitro transcription reactions with E. coli RNA polymerase. P1 and P2 generated run-off transcripts of the expected sizes (Fig. 3B). The addition of ParR repressed P1 but not P2. By contrast, ParB repressed both P1 and P2 (Fig. 3B). These results are consistent with the transcriptional regulation observed in vivo and furthermore show that host-encoded factors are not required for regulation of P1 and P2. As a negative control the in vitro tran-scription from the relBE promoter from E. coli chromosome was analyzed. Neither ParR nor ParB had any effect on promoter activity (data not shown).
ParB Binding to parC1 and parC2 Is highly Cooperative and Yields High Molecular Weight (HMW) Complexes-ParB binding to parC DNA was analyzed directly using gel-shift analysis. Addition of ParB to a radioactively labeled parC1 fragment resulted in the formation of three distinct complexes even at low ParB concentrations ( Fig. 4Ab; Bc indicates ParB-parC complexes). Bc1 and Bc2 exhibited retarded migration but entered the gel matrix. In contrast, the Bc3 complex hardly entered the gel. This HMW complex was seen even at the lowest ParB concentration used here. Gel-shift analysis with a parC2 DNA fragment yielded two separate ParB/parC2 complexes (Fig. 4Ac). Again, a HMW complex appeared, also at the lowest ParB concentration. Control DNA bound ParB at a much higher concentration, and a HMW band was not observed (Fig. 4Aa). To try to resolve the HMW complexes into discrete bands, we also used low percentage agarose gels. In these experiments also, the HMW complexes were observed. These results show that ParB binds avidly to parC1 and parC2. The Hill coefficient for ParB binding to parC1 was determined to 4.9, indicating that ParB binding to parC1 is highly cooperative (data not shown).
ParR Also Binds Cooperatively to parC1-Similarly, we assessed the ability of ParR to bind to parC1 and parC2. As seen from Fig. 4B, ParR shifted parC1 DNA at a much lower concentration than that required to shift parC2 or control DNA (Fig. 4, Bc and Ba, respectively). At the highest ParR concentration, two shifted bands were seen (Rc1 and Rc2 in Fig.  4Bb), indicating the formation of distinct higher order com- plexes. The HMW complexes seen with ParB binding were not observed with ParR, perhaps reflecting a biological difference in the mode of action of the proteins. The Hill coefficient of ParR binding to parC1 was 2.7, which indicates cooperativity.
Comparison of ParB and ParR DNA Binding Affinities-To compare the binding affinities of ParB and ParR for parC1, the percentage of shifted parC1 fragment was plotted against the respective protein concentrations (Fig. 4C). ParB bound to parC1 at much lower concentrations than ParR. Furthermore, at all protein concentrations analyzed, the amount of parC1 bound was much higher for ParB than for ParR. As the concentration of substrate DNA, and consequently of protein-DNA complexes, was considerably inferior to total protein concentrations in the gel-shift experiments, it is reasonable to assume that the amount of free protein was approximately equal to that of total protein. This allows for the estimation of the concentration of DNA-binding protein i.e. ParB and ParR, required for 50% complex formation, which is a measure of the apparent dissociation constant, K d . The K d values for binding of ParB and ParR to parC1 were determined to 0.5 and 20 M, respectively, showing that ParB binds with an ϳ40-fold higher affinity than ParR (Fig. 4C).
Two-stage Binding of ParB to parC1-To characterize further the in vitro interactions between ParB and parC1, a series of SPR experiments were accomplished (see "Experimental Procedures"). A DNA fragment encoding parC1 was immobilized on a sensor chip, and subsequently ParB in various concentrations was injected over the chip.
ParB bound avidly to parC1 (Fig. 5). At 100 nM ParB, a response corresponding to 34 molecules of ParB per parC1 or one ParB dimer per direct repeat in parC1 was detected (Fig.  5A). Furthermore, at this concentration of ParB, the binding rate dropped suddenly to almost zero at a mass corresponding to 17 ParB dimers bound.
Next, ParB was injected for 350 s at different concentrations (Fig. 5B). At ParB concentrations above 100 nM, the binding kinetics exhibited a biphasic pattern; the initial binding rate was relatively high and clearly depended on the ParB concentration but reached the slower phase when binding levels corresponded roughly to 17 ParB dimers bound per parC1 molecule (Fig. 5B) The possible reasons for the biphasic response are addressed under "Discussion." Identification of ParB Binding Sites-To further analyze the requirements for ParB and ParR binding to parC1, we performed gel-shift experiments with a number of truncated parC1 DNA fragments (Fig. 6). As described above, the addition of ParB to the longest parC1 fragment resulted in a large shift of the parC1 fragment to a position in the top of the gel (Fig. 6, fragment a). When truncated in the direction of P2, ParB binding was not affected by deletion of the first 68 bp of parC1 containing the R1 and R2 repeats (Fig. 6, fragments a-c). These results endorse those obtained in vivo, where deletion of R1 and R2 had no effect on the ability of ParB to regulate the promoter activity of P2. However, when truncation led to deletion of 9 or more of the 17 6-bp iterons, ParB binding was clearly affected, resulting in a less shifted band, which became more and more intense as the number of iterons was reduced (Fig. 6, fragments e and f). Importantly, ParB was still able to bind to the shortest parC1 fragment (Fig. 6, fragment g) containing only four 6-bp iterons.
Next, we analyzed the effect on ParB binding when parC1 was truncated in the direction of P1 (Fig. 6,  fragments h-m). Again, as the number of 6-bp iterons was reduced, the slowly migrating band in the top of the gel gradually disappeared, and a less shifted band became evident. The intensity of a shifted band entering the gel increased as more and more iterons were removed (Fig. 6, fragments j-l). Importantly, ParB bound to the fragment containing B1-B4 repeats that overlaps with the P1 promoter (Fig. 6, fragment l). ParB was not able to bind to the 68-bp fragment containing R1 and R2, supporting the notion that the binding sites for ParB are the 6-bp iterons (Fig. 6, fragment m). Again, these results are consistent with the in vivo P1 and P2 promoter analysis (Fig. 2).
Identification of ParR Binding Sites-Deletion of the region harboring R1 and R2 completely abolished ParR binding to the parC1 DNA (Fig. 7A). Thus R1 and R2 may be ParR recognition sites. Therefore, a more detailed truncation analysis was performed on the region encoding R1 and R2 using successive deletions of 10 bp. ParR-binding was not affected by deletion of the region 5Ј of the R1 site (Fig. 7B, fragments a-c). For fragments a and b, one shifted band was observed, whereas two were observed for fragment c (denoted as Rc1 and Rc2 in Fig. 7B). When the 5Ј 5 bp of R1 were deleted, only one specific band with a migration corresponding to Rc1 was observed (Fig. 7B, fragment d). Thus, partial deletion of R1 clearly affected ParR binding. Deletion of the region between R1 and R2 did not abolish ParR binding (Fig. 7B, fragments e-g). By contrast, deletion of the 5Ј 5 bp of R2 completely abolished ParR binding (Fig. 7B, fragments h and i).
When truncated in the opposite direction, ParR binding was not affected by deletion of the region 3Ј of R2 (Fig. 7C, fragment j-l). For fragments j-l, two distinct shifted bands, Rc1 and Rc2, were observed. When the 5Ј 5 bp of R2 were deleted, only one specific band with a migration corresponding to Rc1 was observed (Fig. 7C, fragment m). Thus, partial deletion of R2 affected ParR binding. Again, deletion of the region between R1 and R2 did not affect ParR binding (Fig. 7C, fragments n-p). However, deletion of 5 bp of R1 completely abolished ParR binding (Fig. 7C, fragments q and r). The results described above show that ParR specifically recognizes the R1 and R2 sequences and that one binding site is sufficient for efficient ParR binding. These results are also consistent with the in vivo analysis of P1 regulation (Fig. 2).
IHF Binds to Two Sites within par2-Inspection of the DNA sequence revealed two potential IHF binding sites within par2. Both sites exhibit 100% identity with the minimum consensus IHF recognition sequence (15). The two sites were designated ihf1 and ihf2, respectively. ihf1 is located within parA on the (Ϫ)-strand, 504 bp downstream of the parA start codon. ihf2 is located on the opposite strand and overlaps with 2 bp of the second stop codon of parB (Fig. 1, A and D). Both ihf1 and ihf2 exhibit similarity to other strong IHF binding sites (data not shown), hence indicating that ihf1 and ihf2 might act as bona fide IHF binding sites (Fig. 1D). 32 P-end-labeled DNA fragments of ϳ200 bp, containing either ihf1 or ihf2, were synthesized and used in gel-shift experiments with purified IHF. As a positive control, parS, the cis-acting site of the P1 par locus, which contains a known IHF binding site, was included in the analysis. Because parC1 contains no sequence homology to known IHF binding sites, it was used as a negative control. As expected, IHF bound to parS from P1 (Fig. 8D) but not to parC1 of pB171 (Fig. 8C). Interestingly, IHF bound to both the ihf1-and ihf2-containing DNA fragments (Fig. 8, A and B). Consistently, par2-carrying plasmids were moderately but reproducibly less stable in an ihf deletion strain as compared with an isogenic wildtype strain (data not shown).

DISCUSSION
We show here that ParR regulates the parMR operon via binding to the two 10-bp R repeats upstream of parM and that ParB regulates the parAB operon via binding to the B repeats. This type of transcriptional regulation is typical for Type Ib and Type II loci (2,9). Surprisingly, however, ParB regulates the parMR operon as efficiently as ParR. The B repeats in parC1 and parC2 bound ParB highly cooperatively in vitro (Fig. 4). The SPR analysis is consistent with one dimer of ParB binding to each of the 17 direct repeats in parC1 (Fig. 5A), in agreement with ParB behaving as a dimer in solution (16). At ParB concentrations higher than 100 nM, the SPR analysis yielded a biphasic response, indicating a two-stage binding process (Fig.  5B). The reason for the two-stage binding can only be speculated upon. One possibility is that ParB binds to the two types of B repeats ( Fig. 1C) with different affinities. Another explanation that we favor assumes that all repeats are rapidly saturated and that ParB then extends from the repeats covered with ParB via ParB-ParB interactions. The formation of such higher order nucleoprotein complexes is consistent with the HMW bands seen in the gel-shift analysis (Fig. 4A). The gel-shift and SPR analyses are also consistent with the observation that ParB regulates P2 in vivo (Fig. 2) and in vitro (Fig. 3). The deletion analysis in Fig. 2B shows that a DNA fragment containing four B repeats is sufficient for ParB binding. We conclude that ParB recognizes the B repeats in parC1 and parC2 and thereby generates a nucleoprotein complex that functions in P2 promoter repression and likely plasmid partitioning. It was more surprising to learn that ParB also represses P1. However, this observation is consistent with the fact that the Ϫ10 and Ϫ35 sequences of P1 are embedded in the B repeats (Fig. 2). Thus, ParB binding to the B1-B4 repeats may prevent RNA polymerase to access P1 by steric hindrance.
ParR bound to the R1 and R2 repeats in parC1 in vitro (Figs. 6 and 7). ParR binding was also cooperative although considerably less avid than ParB binding to the B repeats. ParR did not bind to the B repeats (Fig. 4Bc), and ParB did not bind to the R repeats (Fig. 6). ParR regulated P1 in vivo, consistent with the R operator sites located between P1 and the start of parM. ParR of plasmid R1 binds to 10 direct repeats in the parC locus (7) and thereby facilitates DNA pairing and repression of transcription (17,18). It is likely that ParR of pB171 have similar functions.
Many divergent promoters have been described. In most cases, transcriptional interference between divergent promoters is positive (19,20), and only few examples of negative interference are known (21). The deletion analysis presented in Fig.  2 shows that the P1 and P2 promoters are negatively coupled, i.e. one promoter affects negatively the activity of the other. Our analysis thus adds a new example to the short list of divergent promoters exhibiting negative interference. The mechanism of transcriptional interference is not known. Unexpectedly, deletion of the R repeats resulted in a highly increased, ParB-mediated repression-fold of P2 (Fig. 2B). The reason for this effect can only be speculated upon, but it is possible that RNA polymerase binding to the full-length fragment competes with ParB binding to B1-B4 and thereby reduces P2 repression.
Interestingly, IHF binds to two sites in par2 (Figs. 1, A and D,  and 8). In plasmid P1, IHF binds to parS and induces a strong bend in the DNA (22,23). The bend is likely to facilitate interactions between P1 ParB molecules bound to sequences in parS flanking the IHF site (23). Plasmid stability measurements showed that IHF had a moderate positive effect on par2 effi-ciency (data not shown). We speculate that binding of IHF to ihf1 and ifh2 might induce a bend in the DNA such as to spatially join parC1 and parC2 in a higher order partition complex.
Both par loci of pB171 are active, i.e. par1 stabilizes mini-R1 by Ϸ15-fold, whereas par2 stabilizes mini-R1 by Ϸ150-fold, showing that par2 is significantly more efficient than par1 (10). The higher efficiency of Type I par loci is correlated with a much broader phylogenetic distribution (5). Furthermore, ParB repressed par1 transcription as efficiently as ParR (Figs. 2 and 3). This result is consistent with the hypothesis that par2 is epistatic to par1.
It was pointed out by Eugene Koonin (24,25) that almost all biological functions have evolved independently at least twice (the classical example is aminoacyl-tRNA synthetases with identical substrate specificities but unrelated tertiary folds) and that the corresponding analogous genes can replace each other in a process called nonorthologous gene displacement. The Walker box and actin partition ATPases belong to ancient, nonorthologous gene families and therefore must have evolved independently. This fact raises the possibility that Type I and Type II loci undergo nonorthologous gene displacement. The higher prevalence of Type I loci is correlated with a higher efficiency of plasmid stabilization. The double par locus of pB171 may thus represent an evolutionary snapshot revealing one par locus in the process of replacing another.
Even though most plasmids have one par locus only, plasmid pB171 is not the only plasmid that has two. Thus, plasmid R27 also has two par loci, par1 (Type I, encoding a Walker box ATPase) and par2 (Type II, encoding an actin ATPase) located several kilobases apart (26). As in the case of pB171, the Type I locus of R27 positioned plasmids at mid-and quarter-cell positions and was more efficient than the Type II locus. These observations are also consistent with the nonorthologous gene displacement hypothesis. To our knowledge, plasmids carrying two par loci of the same type have yet to be identified.
In conclusion, we show here that ParB binds to the B repeats in parC1 and parC2 and thereby is able to form the partition complex at which ParA acts. Likewise, ParR binds to the R repeats in parC1, thereby probably forming a paired plasmid complex at which ParM can act. Surprisingly, however, ParB repressed transcription of par1, indicating that par2 is epistatic to par1 and might possibly replace par1 by nonorthologous gene displacement.