The kfrA gene is the first in a tricistronic operon required for survival of IncP-1 plasmid R751

Sequencing of bacterial genomes has revealed not only that some chromosomes are very similar to megaplasmids in their replication and stable inheritance mechanisms, but also that most standard oriC-based chromosomes carry a range of stable inheritance functions, already studied in some detail on bacterial plasmids (Gerdes et al., 2000a). This includes active partitioning systems of the parAB family (Gerdes et al., 2000b; Bignell & Thomas, 2001), multimer resolution systems of the xerC/D family (Summers & Sherratt, 1984; Blakely et al., 1993) and a variety of post-segregational killing systems (Gerdes et al., 2000a; Zielenkiewicz & Ceglowski, 2001). Thus, studies on bacterial plasmids, particularly large, low-copy-number plasmids, can be of direct relevance for understanding key events of the bacterial cell cycle. Genomic approaches also reveal genes that are conserved between many systems and cluster with known replication and stable inheritance functions, but do not yet have a role assigned. One such gene family has as its archetype the kfrA gene of IncP-1α plasmid RK2 that we studied many years ago, but for which we did not establish a phenotype, although we speculated that it may be involved in active partitioning (Jagura-Burdzy & Thomas, 1992; Williams & Thomas, 1992). KfrA_RK2 is 308 aa in length, is a site-specific DNA binding protein, and is predicted to be almost 100 % α-helical, which is consistent with its circular dichroism (CD) spectrum (Jagura-Burdzy & Thomas, 1992). It is predicted to consist of an N-terminal, globular, DNA binding domain and an extended coiled-coil tail. Database searching revealed similarity to proteins of the myosin/kinesin family and, recently, to prokaryotic structural maintenace of chromosomes (SMC) and MukB-like proteins (Hirano, 1998; Soppa, 2001). However, there is no easily identified nucleotide-binding site in KfrA, so it seems unlikely to be able to provide the force-generating element of a molecular motor.

Recent accumulation of plasmid sequences has revealed other members of the kfrA family, located in or close to a cluster of stable inheritance functions. In both pFBAOT6 (Rhodes et al., 2004) and pIP02 (Tauch et al., 2002), this gene is close to stable inheritance functions very similar to the active partitioning/central control region of RK2 (Pansegrau et al., 1994; Thorsted et al., 1998). In other plasmids, the kfrA orthologue is clustered with replication functions (pMOL98, GenBank accession no. AJ 345055; pSa, Close & Kado, 1992), with partitioning functions (pXAC33, accession no. NC 003921; pKDSC50, accession no. NC 002638; Haneda et al., 2001), or a mixture of replication and stable inheritance determinants (pRA2, accession no. NC 005909; Kwong et al., 1998, 2000; pO157, accession no. NC 002128; Burland et al., 1998). This prompted us to re-examine the IncP-1 plasmids to see if we could establish a role for this gene in stable inheritance. The IncP-1 plasmids are, so far, the only system in which the operons encoding parA(incC) and parB(korB) homologues and the associated parS sequences (Motallebi-Veshareh et al., 1990; Williams et al., 1998) have been shown to require additional plasmid functions for full effectiveness (Thorsted et al., 1998). Thus, in both RK2 and R751, the central control region (consisting of central control operon (cco) korAincCkorB and orphan kfrA, upf54.8 and upf54.4 genes; Fig. 1) provides a significant partitioning phenotype, but this is increased markedly by the addition of the klc and kle regions, and it has been shown that the effect of these regions depends on an intact central control region (Thorsted et al., 1998). It is possible that the kfrA operon provides an independent stable inheritance mechanism or that it enhances that activity of another mechanism. This forms the basis for trying to establish what role, if any, the genes flanking the central control operon play in IncP-1 plasmid stability.

(14K): Fig. 1. Genetic organization of the studied region of R751. Comparison of the central control regions of RK2 (IncP-1α) and R751 (IncP-1β). ORFs are boxed, transcripts are indicated by arrows and broken lines, terminators are shown by the hairpin-loop symbol, and promoters for korA and kfrA are indicated as well as the binding sites for KorA (OA), KorB (OB) and KfrA (OK). The numbers in the central box show the level of DNA sequence identity between the corresponding genes from RK2 and R751 and predicted amino acid sequences of their products. A question mark in parentheses signifies a lack of data about operon organization before we started our analysis.

The stability of RK2 (IncP-1α) relies on active partitioning (Motallebi-Veshareh et al., 1990; Williams et al., 1998), assisted by kle and klc functions whose role is not understood (Macartney et al., 1997). It also encodes a multimer resolution system (parABC) (Roberts & Helinski, 1992; Easter et al., 1998) and a post-segregational killing system (parDE) (Easter et al., 1998; Roberts et al., 1993; Johnson et al., 1996). R751 (IncP-1β) lacks the whole parABCDE region. Thus, the stable inheritance of R751 is based only on the presence of the central control region, assisted by a more basic set of klc and kle operons (Thorsted et al., 1998). Its kle region is shorter than in RK2, because it does not contain the kleCD genes. On the other hand, the klc operon of R751 has a functional klcA gene which is inactivated in RK2 by Tn1 insertion (Pansegrau et al., 1994). Therefore, it seems likely that the central control region of R751 provides the basis for the definition of the minimal requirements for this region; the central control operon codes only for korA, incC and korB, and not for the two genes korF and korG that are present in IncP-1α. The upf54.4 ORF is also 50 % shorter, encoding a protein lacking the N-terminal 231 aa, as compared with RK2. Despite these differences, R751 is even more stably maintained than RK2, at least in Escherichia coli and Pseudomonas species, so we have chosen R751 as a simpler model than RK2 to analyse the role of the orphan genes kfrA, upf54.8 and upf54.4.

Comparison between DNA and amino acid sequences of R751 and RK2 in this region (Fig. 1) showed low conservation in the kfrA ORFs (50 % DNA identity, 38 % amino acid identity) but high conservation in the upf54.8 and upf54.4 of both representatives (70 % DNA identity, 68 and 64 % amino acid identity, respectively). Despite this unusually low conservation of the primary sequences of KfrA_RK2 and KfrA_R751, both proteins are highly α-helical and have an almost identical predicted secondary structure, with an N-terminal globular part and a long, coiled-coil α-helical tail (Jagura-Burdzy & Thomas, 1992; data not shown for KfrA_R751). The only function identified so far for the whole region (kfrA, upf54.8, upf54.4) is for KfrA_RK2 and concerns the autoregulation of transcription from kfrAp (Jagura-Burdzy & Thomas, 1992). The previous study of KfrA_RK2 by Williams et al. (1998) showed that an internal in-frame deletion of kfrA, in a pOG04 derivative (hybrid replicon P7pMB1 with stabilization functions of the central control region of RK2), did not affect the stability of the plasmid when it had a low copy-number in a polA1 strain. Although RNA polymerase binding sites have been identified in the upf54.8 and upf54.4 region of RP4 (=RK2) by electron microscopy (Pansegrau et al., 1994), DNA sequence analysis of both RP4 and R751 in this region has failed to identify strong candidate promoters, raising the possibility that all these genes constitute an operon expressed from the kfrA promoter. On the other hand, the sequence of a putative Rho-independent terminator, previously identified in RP4 between kfrA and upf54.8 (Pansegrau et al., 1994), is conserved in R751 (Supplementary data). In order to define better what role, if any, kfrA, upf54.8 and upf54.4 play in the stable inheritance of R751, we have addressed a number of questions in this paper. Do the three ORFs belong to the same operon and how is the expression of the operon regulated in R751? Can the genes in this region be inactivated and what effect will the mutations have on R751 biology? What is the function of the unusual α-helical KfrA and, possibly, the other two proteins? Do KfrA, Upf54.8 and Upf54.4 interact with each other to fulfil their biological function?

Bacterial strains and growth conditions.
E. coli strains used were K12 strains DH5α F(Φ80dlacZΔM15) recA1 endA1 gyrA96 thi-1 hsdR17(r_km_k⁺) supE44 relA1 deoR Δ(lacZYA-argF)U196, C600K thr-1 leu-6 thi-1 lacY1 supE44 ton21 galK (McKenney et al., 1981), NEM259 met, trpR, supE, supF, hsdR (W. Brammar, University of Leicester, UK), E. coli B strain BL21 F ompT hsdS_B (r_Bm_B) gal dcm (λ DE3) (Novagen), and E. coli C strains C2110 polA1 his rha (D. R. Helinski, University of California, San Diego, CA, USA) and C2110Rif^R (this work). Bacteria were generally grown in L broth (Kahn et al., 1979) at 37 °C. Some experiments were performed at 25 or 42 °C. L agar (L broth with 1.5 %, w/v, agar) was supplemented with antibiotics as appropriate: benzylpenicillin sodium salt (150 µg ml¹ in liquid media and 300 µg ml¹ in agar plates) for penicillin resistance; kanamycin sulphate (50 µg ml¹) for kanamycin resistance; tetracycline hydrochloride (1025 µg ml¹) for tetracycline resistance; and streptomycin sulphate (30 µg ml¹) for streptomycin resistance. Rifampicin was used at 25 µg ml¹. The L agar used for blue/white screening contained 0.1 mM IPTG and X-Gal at 40 µg ml¹.

Yeast strains and growth conditions.
Saccharomyces cerevisiae strain L40 MAT a, trp1, leu2, his3, ade2, LYS2 : : lexAHIS3, URA3 : : lexA-lacZ was used for transformation (Clontech) (Fields & Song, 1989). Yeast cells were grown in YPD medium (1 % yeast extract, 2 % Bacto peptone and 2 % glucose). Plasmid-containing yeast strains were grown in YNB medium (0.67 %, yeast nitrogen base and 3 %, glucose) supplemented with a mixture of appropriate nutrients lacking tryptophan (for selection of pBTM116 derivatives), leucine (for selection of pGAD424 derivatives) or both as required (for selection of double transformants). Agar was added to a concentration of 2 % (w/v) for the plates. To grow transformants for the β-galactosidase filter-lift assay, YNB agar was supplemented with 0.5 % instead of 3 % (w/v) glucose. For the β-galactosidase assay using liquid cultures, the yeast cells were grown to exponential phase in YNB medium with 0.5 % (w/v) glucose and then transferred to YNB medium with 3 % (w/v) ethanol for several hours. For testing the expression of the His reporter gene in strain L40, YNB plates were supplemented with a mixture of amino acids (Clontech) lacking tryptophan, leucine and histidine. Plates were incubated at 30 °C for 34 days.

Plasmids.
The plasmids used in this study are listed in Table 1.

Table 1. Plasmids used in this study

For expression of cloned ORFs, we used the high-copy-number Pn^R vector pGBT30 (Jagura-Burdzy et al., 1991), based on the pMB1 replicon with lacI^Q and tacp separated from λt_R1 by the multiple cloning site of pUC18 (Fig. 2C). A PstISalI fragment of pGBT30 derivatives, with the ORF of interest and all upstream control elements, was recloned into a broad-host-range Smr^R IncQ-based replicon, pGBT400 (Jagura-Burdzy et al., 1999b), or Tc^R broad-host-range pBBR1MCS-3 (Kovach et al., 1995). Sequencing of the initial PCR product corresponding to kfrA_R751 revealed an extra codon ACC (an extra Thr at amino acid position 246) (supplementary data) relative to the published sequence, but none of the studies performed here suggested that this was a mutant that had lost any activity. We sequenced two other independently amplified kfrA ORF clones and an identical change was found. It may be that the difference was due to an error in the original sequencing. The internal kfrA deletion (ΔαkfrA) was created by removal of an internal NotI fragment encoding 70 aa between A136 and A207. C- and N-terminal kfrA deletions were obtained using the unique NotI restriction site in pMAB424.11, a pGAD424 derivative with ΔαkfrA. The NotI linearized plasmid was treated with a DNA PolI Klenow fragment and then ligated either with an oligonucleotide carrying an SalI site and stop codons in all three frames (next section), to create ΔCkfrA, or with an oligonucleotide carrying an EcoRI recognition site, followed by an ATG codon in-frame with the rest of KfrA, to create ΔNkfrA. EcoRI digestion and religation removed the N-terminal part of kfrA. All constructions were checked by sequencing of the new junction point.

Table 1). The promoter activities detected, expressed as units of catechol oxygenase activity, are summarized in the first column on the right. The second column on the right shows which pairs of primers gave products of the expected size after RT-PCR with cDNA synthesized from RNA from E. coli NEM259(R751TetR), with the use of primer #1 annealing to the 3' end of upf54.4. ND, Not determined. (B) RT-PCR products of the kfrAupf54.8upf54.4 region obtained from cDNA from cells of strain NEM259(R751TetR). As positive controls, the same pairs of primers were used with R751 DNA as the template (dsDNA panel). M, 1 kb DNA ladder; , no template added. (C) Compatible plasmids used in the regulatory studies. C600K was first transformed with pMAB01(kfrApxylE) and then either one, or both, of the pGBT30 and pGBT400 derivatives were introduced to C600K(pMAB01). Both expression vectors had exactly the same transcriptional andtranslational signals to overproduce the regulatory proteins. They differed by copy number, so the experiments with two regulators in trans were done with reciprocal combinations of ORFs, on either medium- or high-copy-number vectors.

A plasmid for KfrA purification was constructed by inserting the kfrA allele into a derivative of pET28a Km^R (Studier & Moffatt, 1981; Novagen), modified with the help of the synthetic oligomer, so that a His₆-tag and thrombin cleavage site preceded, almost directly, the EcoRI site (Lukaszewicz et al., 2002). Purified His-tagged KfrA was N-terminally extended by the sequence MGSSH₆SSGLVPRGSHSEF.

E. coliS.cerevisiae shuttle plasmids used in the yeast two-hybrid (YTH) system, based on vectors pBTM116 (Invitrogen) and pGAD424 (Fields & Song, 1989; Ma & Ptashne, 1987; Clontech Matchmaker), fused the polypeptides encoded by EcoRISalI inserts to the C terminus of the LexA binding domain (BD) and GAL4 activation domain (AD), respectively. The system provided Ap^R selection in E. coli and LEU⁺ (pGAD424) or TRP⁺ (pBTM116) selection in the yeast cells.

Plasmid DNA isolation, analysis, cloning and manipulation.
Plasmid DNA was isolated by standard procedures (Birnboim & Doly, 1979). Digestion of plasmid DNA with restriction enzymes was carried out under conditions recommended by suppliers, and digested DNA was run on 0.82 % (w/v) agarose gels. Standard PCR reactions (Mullis et al., 1986) were performed as described by Jagura-Burdzy & Thomas (1994), with the pairs of primers listed in Table 2.

Table 2. Primers for PCR reactions Restriction enzyme recognition sites are underlined and start/stop codons are in bold type.

The PCR reaction to amplify the ORF for kfrA was performed with a hot start denaturation step (98 °C for 5 min) and 35 cycles of denaturation (94 °C for 30 s), annealing (52 °C for 30 s) and elongation (72 °C for 90 s). The reaction ended with a final elongation step (72 °C for 7 min). The PCR reaction mixture was supplemented with N,N-dimethylformamide to a final concentration of 2 % (v/v).

PCR products were usually cloned into the pGEM-T vector (Promega) and later recloned as EcoRISalI fragments into appropriate vectors. All PCR-derived clones were analysed by DNA sequencing to check their fidelity.

Sequencing and computer sequence analysis.
DNA sequencing was performed using an Applied Biosystems 377 automated DNA sequencer (Alta Bioscience, School of Biochemistry, University of Birmingham, UK) or a Pharmacia ALF automatic sequencer (IBB, PAS) using dye terminator kits supplied by the manufacturer. DNA and amino acid sequence analysis was carried out using the GCG Wisconsin Package version 8.0. Sequences were compared to GenBank/EMBL databases and studied using the programs BLASTN and BLASTP (NCBI), BLAST (Swiss server) and Coil tool (ExPASy).

E. coli transformation and PCR-mediated screening of transformants.
Competent cells of E. coli were prepared by the standard CaCl₂ method (Sambrook et al., 1989). To screen for the recombinant plasmids, each transformant colony picked from the selective plate was resuspended in 200 µl water and boiled for 2 min. After centrifugation, 510 µl supernatant (depending on the plasmid copy number) was added to a 25 µl PCR reaction mixture. Alternatively, bacterial colonies were directly transferred into 25 µl PCR reaction mixture supplemented with 0.5 µl N,N-dimethylformamide.

Bacterial conjugation.
Overnight cultures of the donor and recipient strain were grown in selective medium. One millilitre of each culture was drawn into a syringe mounted on a Millipore filter. The bacterial suspensions were filtered through a sterile nitrocellulose membrane, which was then placed face-up onto an L agar plate and incubated at 37 °C overnight. The filter was transferred to a tube containing 1 ml L broth and vortexed to resuspend the bacteria. Serial dilutions of the suspension were plated out on L agar with appropriate selection and allowed to grow for 48 h at 37 °C.

Yeast transformation.
Yeast transformation was performed using the PEG/LiAc standard method (Chen et al., 1992). Either single or double transformations were conducted and transformants selected and stored on minimal YNB agar.

Determination of LacZ activity in yeasts.
β-Galactosidase activity was monitored by the filter-lift assay with X-Gal as a substrate and by the quantitative liquid culture assay (Clontech) using ONPG as substrate. The protein concentration was estimated using the Bradford assay (Bradford, 1976). The specific activity of β-galactosidase activity was expressed in standard units where one unit represents the amount of enzyme that hydrolyses 1 nmol ONPG (mg protein)¹ min¹ at 30 °C.

Determination of catechol 2,3-oxygenase activity (XylE) in bacteria.
Catechol 2,3-oxygenase activity (the product of xylE) was assayed in exponentially growing bacteria (Zukowski et al., 1983) and normalized to the same plasmid level as the control. One unit of catechol 2,3-oxygenase is defined as the amount needed to convert 1 µM catechol in 1 min under standard conditions. Protein concentration was determined using the Bradford method (Bradford, 1976).

Purification of His₆-tailed KfrA.
Exponentially growing E. coli BL21(λDE3)(pMAB28.1) was induced with 0.5 mM IPTG at a cell density of ∼2x10⁸ c.f.u. ml¹ and grown for an additional 23 h with shaking at 37 °C (Arnold, 1991). The cells from 300 ml culture were harvested by centrifugation and sonicated. The cell lysate/Ni²⁺-nitrilotriacetic acid/agarose mixture was stirred for 60 min at 42 °C and then transferred to a column. The purification procedure of His₆-tailed KfrA was monitored by SDS-PAGE using a Pharmacia Phast gel system.

Preparation of anti-KfrA antiserum.
Purified His₆KfrA protein (1 mg ml¹) was injected into a rabbit. The blood collected from the rabbit was allowed to clot at room temperature before removing the antiserum. Anti-KfrA antibodies were affinity-purified on His₆KfrA bound to Affi-gel 15 (Bio-Rad), using a method described by Reznekov et al. (1996) with modifications introduced by Bignell et al. (1999). The antibodies were stored at 20 °C in 50 % (v/v) glycerol.

Estimation of KfrA concentration in the cells carrying R751Tet^R by Western blotting.
Growth of E. coli NEM259(R751Tet^R) was monitored by OD₆₀₀. The culture was diluted and plated on L agar to establish c.f.u. ml¹. Bacteria were harvested from an appropriate volume depending on the phase of growth to collect similar numbers of cells. The cells were washed in water and resuspended in 200 µl sonication buffer (50 mM phosphate buffer, pH 8.0, 300 mM NaCl). Bacteria were kept on ice and disrupted by sonication in short bursts for 2 min. Crude extracts were cleared by centrifugation at 4 °C in a microfuge for 15 min at 12 000 g. Protein concentration was measured using the Bradford method (Bradford, 1976). Proteins from the cleared extracts were separated on 12.5 % (w/v) SDS-PAGE gels (Laemmli, 1970) and electroblotted (semi-dry transfer unit) for 12 h onto nitrocellulose membranes (Protron; Schleicher & Schuell). Probing of the blots was carried out using the amplified alkaline phosphatase goat anti-rabbit immunoblot assay (Promega) with anti-KfrA antiserum at a dilution of 1 : 100 000 as the primary antibody. The cell extracts were compared to a range of concentrations of purified His₆KfrA on the same gel. Band intensity on Western blots was determined using Image Quant (Molecular Dynamics).

Immunofluorescence microscopy.
The fixation and permeabilization of cells and subsequent staining for immunofluorescence were carried out as described by Bignell et al. (1999). Ten microlitres of affinity-purified anti-KfrA and anti-KorB antibodies were used as the primary antibodies (1 : 5000 dilution for anti-KfrA and 1 : 2000 dilution for anti-KorB antibodies in 2 %, w/v, BSA/PBS) followed by 10 µl anti-rabbit IgG FITC-conjugate solution (6.9 µg ml¹ in 2 %, w/v, BSA/PBS) (Sigma). The coverslip was placed onto a microscope slide in a solution of 4',6-diamidino-2-phenylindole (DAPI)/Vectorshield mounting medium (Vector Laboratories) (ratio of 1 : 4 DAPI 1 µg ml¹ to Vectorshield). Cells were studied with an Olympus IX70 inverted reflected light fluorescence microscope fitted with a Sensys charge-coupled device (CCD) camera (Photometrics). Images were captured and manipulated on a Macintosh G3 with the Smartcapture I program (Digital Scientific).

Introduction of mutant alleles into R751Tet^R backbone by reverse genetics.
E. coli strain NEM259(R751Tet^R) was transformed with suicide vector pAKE600 derivatives (El-Sayed et al., 2001). Double transformants were conjugated with the recipient strain C2110Rif^R in which the suicide vector based on pMBI ori was unable to replicate. Putative conjugants with R751Tet^R-suicide vector cointegrated plasmids were selected on L agar with rifampicin, penicillin and tetracycline. After double restreaking, Rif^RPen^RTet^R colonies were used to inoculate L broth containing 10 % (w/v) sucrose and tetracycline to facilitate the removal of the suicide vector insertion from R751Tet^R. Survivors were plated on L agar with 10 % (w/v) sucrose and tetracycline and checked for the Pen^S phenotype. Suc^RTet^RPen^S colonies were analysed by colony PCR to determine whether allele exchange had occurred and, where necessary, the PCR products were gel-purified and digested with SalI (ΔCkfrA, Δupf54.8). To confirm the knockout of the kfrA gene in the R751Tet^R backbone protein, extracts were analysed by Western blotting with anti-KfrA antibodies.

Determination of plasmid stability.
Plasmid stability tests were performed as described by Williams et al. (1998) and Macartney et al. (1997). E. coli strain NEM259 was transformed with plasmid R751Tet^R (or its derivatives) and transformants were selected on L agar plates containing tetracycline. The transformants were grown in selective medium and then diluted 10⁵-fold into non-selective L broth. Samples of the culture were collected every 12 h (at which point the cultures were diluted 10⁵-fold and transferred into fresh medium), diluted and plated out onto L agar and then transferred onto L agar with tetracycline. The c.f.u. were counted to determine the proportion of bacteria retaining the plasmid.

RT-PCR reactions.
One microgram of total RNA (isolated by Qiagen RNeasy kit), suspended in diethyl pyrocarbonate-treated water, was mixed with 2.5 pM gene-specific downstream primer #1 annealing to the 3' end of upf54.4, and the reverse transcriptase reaction was performed with 200 U SuperScript RT. RNA was removed by treatment with RNase and cDNA was purified using GlassMAX spin cartridges as recommended by the manufacturer (GibcoBRL). Purified cDNA was used as a template for PCR with primer pairs as in Fig. 2(B, C).

kfrA, upf54.8 and upf54.4 constitute a transcriptional unit
From the original studies on IncP-1α plasmid RK2 it appears that the kfrA gene forms an autogenously regulated, monocistronic operon, leaving completely open the benefit that it may provide for RK2 (Thomas et al., 1990). However, there are no experimental data confirming termination of transcription after kfrA or independent transcription of downstream genes upf54T.8 and upf54.4. To test the possibility that all three genes form a single operon (and may have a common function), reporter gene fusions were first constructed with xylE and subsegments of this region (Fig. 2A and supplementary data for primer positions). Strong transcriptional activity was detected for the region containing the predicted kfrAp (pMAB01; nucleotides 22012597, supplementary data), whereas activity was only very weak for a fragment spanning the end of kfrA and the start of upf54T.8 (pMAB03.4; nucleotides 36664062) and hardly detectable for another fragment running from upf54.8 to upf54.4 (pMAB05.6; nucleotides 34323691). When the cloned segment started upstream of kfrA and ended within upf54.8 (pMAB08.5; nucleotides 22013691), much higher transcriptional activity was observed compared to the upf54.8 fragment alone (pMAB03.4), indicating that transcription originating within, or upstream of, kfrA was responsible for most of the upf54.8 transcription. To further locate this activity, a 408 bp internal deletion in kfrA was created, running from nucleotides 3025 to 3432 (pMAB08.51). This resulted in a further fourfold increase in transcription into upf54.8, confirming that the promoter responsible for upf54.8 transcription must be upstream of nucleotide 3025, close to the beginning of kfrA at nucleotide 2467 (Supplementary data). Although this deletion did not abolish KfrA autorepressor activity (see below), the increase in XylE activity in pMAB08.51 versus pMAB08.5 suggests that the promoter activity measured was under the control of KfrA. No obvious promoters, other than kfrAp, were found in the remaining R751 DNA present in pMAB08.51, and also no similarities to the putative KfrA binding site identified in kfrAp. Therefore, the simplest explanation is that transcription from kfrAp ran through into upf54.8. This may have added to the weak promoter activity detected in the intergenic region between kfrA and upf54.8.

Independent evidence for co-transcription of kfrA, upf54.8 and upf54.4 came from isolation of total RNA from strain NEM259(R751Tet^R) and PCR on cDNA obtained with reverse transcriptase and primer #1 (supplementary data) complementary to the 3' end of upf54.4. PCR products of expected sizes were obtained with primer pairs spanning both the upf54.4upf54.8 and upf54.8kfrA junctions, as well as for the whole of kfrA, indicating that the cDNA, and hence mRNA, covered the whole region (Fig. 2A, B). As a negative control, we performed RT-PCR on the korBkfrA intergenic region, which contained a predicted strong transcriptional terminator preceding kfrAp (primers 7/17) and on the kfrA promoter region (primers 7/8), and obtained no products. All these primer pairs amplified the appropriate fragments from the R751 DNA template in parallel PCR reactions. We therefore conclude that while upf54.8 and upf54.4 may have been weakly expressed independently of kfrAp, the majority of their transcription was part of a tricistronic unit starting with kfrA.

Regulation of transcription from kfrAp
The DNA sequence of kfrAp_R751 contains highly conserved sequences, known to bind the global regulators of IncP-1 plasmids KorB (O_B) (Bignell et al., 1999; Williams et al., 1993; Jagura-Burdzy et al., 1999a; Kostelidou & Thomas, 2000) and KorA (O_A) (Theophilus et al., 1985; Jagura-Burdzy & Thomas, 1995) located in tandem upstream of the predicted 35 region (Supplementary data). Here, O_B and O_A are separated by only 3 nt (16 nt between their centres of symmetry). While similar close spacing is also observed at klcAp_R751, previous studies on simultaneous binding and cooperativity of KorA and KorB have been confined to promoters where the centre-to-centre distance between O_B and O_A is 3234 nt. To determine the sensitivity of kfrAp_R751 to KorA and KorB, reporter gene assays were performed with the regulatory genes, inducible by IPTG, provided in trans on compatible plasmids (Fig. 2C). Both KorA and KorB alone showed only two- to fourfold repression but gave about 100-fold repression (Table 3) when present together, giving a cooperativity index of 10, which indicates a 10-fold higher repression than expected from the product of the repression indices of both proteins individually. Thus, close spacing of O_A and O_B did not prevent their cooperative interaction and also did not cause them to interfere with each other. It may be of significance that the centres of symmetry of these operator sequences are actually located 16 bp apart and therefore should be on opposite faces of the DNA helix (Bingle et al., 2005).

Table 3. Effects of different regulators on expression of kfrApR751

We previously showed for RK2 that kfrAp is autoregulated and that the target for KfrA_RK2 is an operator, O_K, overlapping the 10 region (Jagura-Burdzy & Thomas, 1992; Thomas et al., 1990). The long inverted sequence is also present in kfrAp of R751 and this was tentatively named putative KfrA_R751 operator O_K (Thorsted et al., 1998). The kfrA_R751 ORF was inserted into pGBT30 and pGBT400 derivatives under tacp control, and the effect of KfrA_R751 overproduction on the activity of kfrAp_R751 (kfrApxylE transcriptional fusion in pMAB01) was analysed in double transformants. Table 3 shows that when KfrA_R751 was expressed in trans from tacp, significant repression was observed even without induction by IPTG (0.28 U XylE in the presence of pMAB30.1 in trans to pMAB01 in comparison to 3.55 U in the presence of vector pGBT30 under the same conditions). In contrast, no effect of KfrA_RK2 on kfrAp_R751 was seen [E. coli C600K(pMAB01)(pWS131)], confirming that the O_K sequences changed autoregulator specificity. The regulatory effects of Upf54.8 (pMAB30.4) and Upf54.4 (pMAB30.5) were tested in the same way, but with negative results (Table 3). The presence of Upf54.8 or Upf54.4 did not increase the repression of kfrAp exerted by KfrA from pMAB40.1 (data not shown).

To determine how KfrA_R751, KorA _R751 and KorB _R751 might interact at kfrAp_R751, we provided korA or korB at the same time as kfrA. Surprisingly, under uninduced conditions, when either KorA or KorB was present with KfrA, the activity of kfrAp was two- to fourfold higher than when KfrA was present alone (Table 3). The effect seemed not to depend on the level of korA/korB gene dosage versus kfrA gene dosage (tacpkfrA was expressed either from medium-copy-number plasmid pMAB40.1 or high-copy-number plasmid pMAB30.1). These results suggest some sort of interference between these repressors. Western blotting extracts from the cultures did not indicate any significant fluctuations in the level of KfrA production in the presence or absence of other regulators. The presence of three repressors, KorA, KorB (pPDB3.6) and KfrA (pMAB30.1), together completely shut off expression from kfrAp.

Repression by deletion derivatives of KfrA
From the previous section it can be seen that repression of transcription is a key property of KfrA. The predicted secondary structure of KfrA_R751, 343 aa, revealed a distinct N-terminal region that might form a globular head linked to a long, coiled-coil, α-helical structure (data not shown). No typical DNA binding motifs were detected. This is very similar not only to RK2 KfrA but also to the protein TlpA of pLT2, a protein involved in regulating temperature-responsive gene expression (Koski et al., 1992; Hurme et al., 1994, 1996, 1997). To determine which parts of KfrA are needed for DNA binding, mutant alleles of kfrA were created with extensive deletions: ΔαkfrA, encoding KfrA, lacking 70 aa between A136 and A207 from the long α-helical tail (KfrAΔ137206); ΔCkfrA, encoding the N-terminal 136 aa (KfrA1136); ΔC₂kfrA, encoding the N-terminal 186 aa linked in-frame to the C-terminal 20 aa (KfrAΔ187323); and ΔNkfrA, encoding the C-terminal 137 aa (KfrA207343).

To test the possibility that KfrA might control a temperature-dependent switch, the deletion alleles of kfrA were expressed from tacp in trans to the kfrApxylE reporter plasmid (pMAB01) at 25, 37 and 42 °C. At all temperatures, WT KfrA showed very strong repression and this activity appeared to be enhanced even at 42 °C. The C-terminal fragment KfrA207343 showed no repression at any temperature. At 25 °C, both KfrA1136 and KfrAΔ137206 were able to repress the krfA promoter, although KfrA1136 demonstrated a significant drop in repression compared to wild-type (WT) KfrA or KfrAΔ137206. At the higher temperatures, KfrA1136 lost most of its repressor activity. Interestingly, KfrAΔ137206 showed quite clear temperature sensitivity at low levels of induction, with significant levels of repression at 37 °C but not at 42 °C. However, at higher levels of induction, KfrAΔ137206 still caused repression at high temperature and, indeed, was an even more potent transcriptional repressor than WT KfrA, at least at 37 °C (Table 4). This indicates that the DNA binding domain is in the protein segment amino acids 1136, but that the rest of the protein contributes to stabilization of the active form of the protein, so that inactivation can occur at higher temperature in the mutants. The kfrA allele producing KfrAΔ187323 with 137 aa removed was also tested (pPDB30.14) and found to have repressor activity comparable to WT KfrA when overproduced (data not shown).

Table 4. Influence of temperature on autorepression ability of KfrA and its truncation products For key to symbols, see footnotes to Table 3.

KfrA, Upf54.8 and Upf54.4 interact with each other
The YTH system was used to test for interactions between KfrA_R751 and other proteins encoded by the same operon as KfrA, and with proteins encoded by the adjacent central control operon and involved in the active partitioning of IncP-1 plasmids. To study relevant proteinprotein interactions, kfrA, incC1, korB, upf54.8 and upf54.4 were cloned into the shuttle vectors pBTM116 and pGAD424 to create translational fusions with the C termini of LexA and the GAL4 activation domain (GAL4_AD). S. cerevisiae L40 was used to detect proteinprotein interactions on the basis of activation of the lacZ and HIS genes under control of GAL4-dependent promoters, with LexA binding sites upstream.

The YTH data demonstrated that KfrA was able to dimerize in vivo. In liquid cultures of double transformants of strain L40(pMAB116.1)(pMAB424.1), a high level (90 U) of β-galactosidase activity was detected, in comparison with <0.2 U in transformants with control plasmids, suggesting very strong proteinprotein interactions. Experiments with glutaraldehyde cross-linking of purified His₆KfrA confirmed the ability of KfrA_R751 to form dimers and other higher-order complexes (data not shown).

No interactions were detected between KfrA and either IncC1 or KorB, which are core components of the active partitioning system encoded by the adjacent operon. These therefore provide negative controls (<0.2 U β-galactosidase). However, the YTH system did show that there were weak interactions between KfrA and Upf54.8 (25 U β-galactosidase) but not between KfrA and Upf54.4 (data not shown). Both Upf proteins self-interacted (Upf54.4 dimerization gave rise to ∼30 U β-galactosidase, whereas Upf54.8 was much weaker with 35 U) and interacted with each other in the YTH system (3090 U β-galactosidase, depending on translational fusion). It is therefore possible that although KfrA did not interact with Upf54.4 directly, all three proteins may have formed a complex, with Upf54.8 being the linker between the other two components.

Mutations in KfrA and Upf54.8 lead to plasmid instability
To establish the role of intact KfrA in R751 biology, mutant kfrA alleles were introduced into the R751Tet^R backbone as described in Methods. In essence, for each mutation, 400 bp arms flanking the mutation were generated by PCR, introduced together into a suicide plasmid, which was then forced to integrate by homologous recombination into R751 (Fig. 3A). Finally excisants were selected and screened by PCR and Western blotting with anti-KfrA antibodies for those that had retained the mutant allele. This was attempted with a total deletion of kfrA as well as two previously characterized deletion alleles, ΔαkfrA and ΔCkfrA. The procedure worked smoothly for the ΔαkfrA allele and nine out of 10 clones after the last step had exchanged WT kfrA for the mutation. However, for the ΔCkfrA allele, only one in 50 analysed clones demonstrated the loss of WT kfrA and, while further analysis indicated the presence of a deletion in kfrA, this did not exactly match the allele which was supposed to recombine, producing a larger polypeptide on SDS-PAGE than was expected (M_r 23 000 rather than 17 000 estimated from the product of plasmid pMAB30.12) (Fig. 3B). PCR amplification followed by DNA sequencing showed that the derivative had a deletion from 187Q to 323L, linking the C-terminal 20 aa to the first 186 aa. This derivative was designated R751ΔC₂kfrA. The regulatory ability of KfrAΔ187323 was tested by cloning the PCR product, under tacp, into pPDB30.14, and was shown to give a similar autorepression effect as WT KfrA (data not shown). The difficulty in obtaining R751ΔCkfrA suggested that the allele we tried to introduce, encoding KfrA1136 with significantly impaired regulatory properties, was somehow disadvantageous for R751. This conclusion is also strengthened by the fact that we were unable to construct an R751ΔkfrAnull plasmid; none of the 100 derivatives screened had lost kfrA completely.

(18K): Fig. 3. Stability of R751 mutant derivatives. (A) Schematic representation of mutated kfrA and upf54.8 alleles introduced by reverse genetics into an R751TetR background. The expressed parts of the encoded proteins are marked in black. (B) Western blot of extracts from E. coli NEM259(R751TetRΔC2kfrA) and NEM259(R751TetR) with anti-KfrA antibodies. Similar numbers of cells (2x108, collected at OD600 0.6) were loaded. Column M shows biotinylated molecular mass markers. (C) Stabilitydata for R751TetR and its three derivatives. Overnight cultures of E.coli NEM259(R751TetR or its mutant derivatives), grown on tetracycline, were diluted 105-fold and left to grow without selection for 12 h (∼20 generations), and then diluted into fresh medium. At the time of dilution, the appropriate samples were plated on L agar plates to obtain single colonies, which were subsequently checked for the ability to grow on plates with tetracycline. Plasmid retention was calculated as the percentageof TetR clones in relation to the total number of colonies.

The two R751Tet^R derivatives that were obtained (R751Tet^RΔαkfrA and R751Tet^RΔC₂kfrA) were introduced into E. coli NEM259 and checked for their stability phenotype. The strains were grown for approximately 60 generations without tetracycline selection. Every 20 generations, the cultures were diluted by a factor of 10⁵ into fresh medium. The results of this test are presented in Fig. 3(C), demonstrating the loss of R751 Tet^RΔC₂kfrA, but not R751Tet^RΔαkfrA. The presence of R751Tet^RΔC₂kfrA in the host did not inhibit the growth of bacteria more than R751Tet^R (data not shown). Therefore, it seems unlikely that, in the absence of antibiotic selection, the overgrowth of plasmid-free segregants, relative to the cells with R751Tet^RΔC₂kfrA, was the only reason for the observed loss of plasmid stability (especially since neither R751Tet^R itself nor R751Tet^RΔαkfrA demonstrated significant segregation).

The instability of R751 carrying a mutation in kfrA could have been due to a need for KfrA in a complex directly involved in stable inheritance/plasmid survival or an indirect effect due to disrupted regulatory circuits. To distinguish between these possibilities, the production of the truncated KfrA from R751ΔC₂kfrA was analysed by Western blotting. Standardized samples of bacteria carrying R751 and R751ΔC₂kfrA showed lower production of KfrAΔ187323 than KfrA (as expected from regulatory studies) (Fig. 3B), indicating that the level of expression of the kfrAupf54.8upf54.4 operon was not dramatically disturbed in the R751ΔC₂kfrA mutant, and that plasmid instability was therefore most likely due to a direct role of KfrA.

A mutant allele for upf54.8 was also constructed by introducing a stop codon after the tenth codon into the coding sequence of upf54.8 (pMAB6.41). Using the suicide technique, the wild-type upf54.8 in R751Tet^R was successfully exchanged for the mutated allele at the expected frequency with no problems. The stability test on R751Tet^RΔupf54.8 showed that the mutated plasmid became unstable to a similar extent as with the impaired kfrA allele (ΔC₂kfrA) (Fig. 3C). The upf54.8upf54.4 region in RK2 has previously been suggested to be a part of the tra region (Krishnapillai, 1988) and occasionally referred to as traNtraO. The constructed mutant R751Tet^RΔupf54.8 was highly proficient in conjugative transfer between E. coli strains, as well as between E. coli and Pseudomonas aeruginosa. Attempts to construct an upf54.4 deletion allele have not been successful.

Intracellular concentration of KfrA
Polyclonal rabbit antibodies raised against KfrA were applied to estimate KfrA concentrations in the cells with plasmid R751. The specificity of anti-KfrA antibodies was tested against E. coli extract. A single protein band was detected only when the strain contained R751. There was only very insignificant cross-reactivity between anti-KfrA_R751 antibodies and KfrA_RK2. No signal was observed with an extract from E. coli (RK2) (data not shown) and only a very weak signal with highly overproduced KfrA_RK2 from pWS131 (tacpkfrA_RK2), as demonstrated in Fig. 4(A). Overproduction of KfrA_RK2 was confirmed by Coomassie staining (data not shown). Thus, the 40 % homology at the primary amino acid sequence level was not sufficient for anti-body cross-reactivity. The purified KfrA_R751 demonstrated a range of multimers, even without the presence of cross-linking agent. Western blotting was applied to estimate the level of KfrA_R751 in bacteria from logarithmic and stationary-phase cultures, to determine whether levels were high enough for detection by immunofluorescence, and to compare these levels with those of other regulatory proteins that have already been studied. Previous studies on RK2 proteins (KorA and KorB) showed dependence of the protein level on the stage of culture growth (G. Jagura-Burdzy & C. M. Thomas, unpublished observations). Extracts from E. coli NEM259(R751Tet^R) demonstrated that actively growing cells contained up to 7000 monomers of KfrA per cell (the number of cells estimated on the basis of c.f.u.), whereas stationary-phase bacteria contained about 500 monomers of KfrA per cell (Fig. 4B). This drop in number of molecules per cell cannot be explained completely by the smaller size of stationary-phase bacteria, which had decreased approximately twofold, as estimated from the micrographs. This indicates that either production of KfrA was switched off as bacteria approached the end of the exponential phase, or KfrA was degraded. However, no products of KfrA proteolysis were observed, suggesting that there was a transcriptional block in KfrA production in the transition-phase cells. Thus, the fall in KfrA level may have been due to dilution of the protein after production was reduced in bacteria that continued to divide.

(21K): Fig. 4. Intracellular level of KfrAR751. (A) Western blot with anti-KfrAR751 antibodies. Similar numbers of cells from logarithmic cultures (108) were loaded on the gel in all tracks, with the exception of track NEM259(R751TetR), marked S, which corresponds to 2x109 cells from the stationary culture (track L represents 108 cells of the same strain but from a logarithmic culture). Cells with plasmids overproducing KfrAR751 (pMAB30.1) and KfrARK2 (pWS131) were induced for 3 h with 0.5 mM IPTG. Purified His6KfrA (1.28 µg) and biotinylated molecular mass standards (track M) were also run on this 12 % SDSpolyacrylamide gel. (B) A culture of strain NEM259(R751TetR) was diluted 100-fold and grown on L broth at 37 °C. Growth was monitored by OD600 and c.f.u., and samples were collected, sonicated and analysed by SDS-PAGE with dilutions of purified His6KfrA as standards. This gel is representive of many experiments. The samples from logarithmic cultures correspond to 108 cells and that from the stationary culture to 5x108 cells. The OD600 at which the cells were collected is indicated above the tracks. The stationary culture had OD600 2.0.

KfrA forms foci in the presence of R751
To determine possible localization of KfrA_R751 in particular parts of the bacterial cell, and to obtain more information on the defect caused by KfrAΔ187323 (R751ΔC₂kfrA), we used immunofluorescence with anti-KfrA antibodies to visualize KfrA_R751 protein. In earlier studies, use of anti-KorB antibodies showed that IncP-1 plasmids form one to four foci, dependent on the stage of cell growth, despite the fact that the plasmid copy number per cell is seven to 15 (Bignell et al., 1999; Pogliano et al., 2001). In our experiments, we used anti-KorB and anti-KfrA antibodies. No signals were detected in bacteria without plasmid R751, while very strong and specifically localized foci were observed in bacteria with R751 and the KfrA single binding site (Fig. 5A). Ninety percent of cells contained one to four KfrA foci per cell. A similar distribution of foci was observed with anti-KorB antibodies, suggesting co-localization of KfrA with R751 plasmid clusters.

(34K): Fig. 5. Localization of KfrA in cells with R751 by immunofluorescence. (A) Representative cells of E. coli NEM259(R751TetR) stained by DAPI and incubated with the primary anti-KfrA or anti-KorB rabbit antibodies and secondary FITC-conjugated, anti-rabbit antibodies. The top panel shows the cells of strain NEM259(R751TetR) incubated with the primary anti-KfrAR751 rabbit antibodies. The bottom panel shows the cells from the same culture incubated with the primary anti-KorBRK2 rabbit antibodies. (B) Cells of strain NEM259(R751ΔC2kfrA) were DAPI-stained and incubated with the primary anti-KfrA or anti-KorB rabbit antibodies and then FITC-conjugated, anti-rabbit antibodies. Cells were collected from logarithmic culture grown for only 3 h without selection, so no plasmid loss should have been observed. Representative small and large cells are shown, stained either with DAPI alone or with DAPI and FITC combined. (C) Magnification of part of panel (B).

The visualization experiments were repeated with exponentially growing cultures of E. coli NEM259(R751Tet^RΔC₂kfrA). The culture was grown without selection for five to six generations, so the R751 derivative was still expected to be present in most of the bacteria. KorB and KfrA foci were visible in almost all cells, but two different conclusions could be drawn. First, most of the cells were of normal size or only slightly elongated and, in those cells, the KfrA and KorB foci were much smaller, more numerous and localized close to the pole (Fig. 5B, C). Second, there was a significant percentage of long filamentous cells (forming up to 3 % of the population) which had regularly distributed nucleoids and large DNA-free zones between nucleoids filled with KorB/KfrA/plasmid molecules. In this paper we provide evidence that three R751 genes, kfrA, upf54.8 and upf 54.4, form a unit that is transcribed from kfrAp and regulated by KorA, KorB and KfrA. The kfrA_RK2 gene was originally deduced to be monocistronic (Thomas et al., 1990), mainly on the basis of the presence of a predicted mRNA secondary structure downstream of kfrA that may constitute part of a transcriptional terminator, but this picture needs to be reviewed in the light of the studies on R751. While the low level of transcription into upf54.8_R751 is consistent with there being a transcriptional terminator between kfrA and upf54.8, the latter gene clearly does not have a strong promoter of its own and transcription from kfrA is able to pass through into the downstream gene. Circumstantial evidence for such operon organization has already been provided by in vitro transcription/translation studies on this region from RK2 as a DNA template: much higher expression of upf54.8 and upf54.4 is observed if kfrA is mutated (D. R. Williams & C. M. Thomas, unpublished observations). Further work will be necessary in both IncP-1α and β systems to establish whether there is any transcriptional termination after kfrA and before upf54.8 which would result in higher mRNA levels for kfrA relative to the two downstream genes. Our reporter gene assays suggest that any promoter immediately upstream of upf54.8 is very weak, but it is conceivable that even a weak promoter may be significant under some circumstances. Nevertheless, because of the organization, coregulation and physical interaction between the proteins and the phenotypes reported in this paper, we conclude that the upf genes are part of the coregulated plasmid backbone functions of the IncP-1 plasmids, and are probably part of the machinery that ensures the efficient survival of the plasmid as a mobile genetic element. In recognition of this, we suggest that the upf54.8 and upf54.4 genes should be renamed kfrB and kfrC, respectively.

A key part of this newly defined operon is the autoregulation by KfrA. The putative target for KfrA is probably an extended inverted repeat overlapping the 10 region of kfrAp (Supplementary data). One could therefore expect KfrA_R751 to be at least dimeric. This is consistent with the self-association demonstrated using the yeast two-hybrid system and confirmed by chemical cross-linking studies (data not shown). The predicted structure of KfrA_R751 implicates the whole of the extended α-helical, coiled-coil structure in dimerization. The removal of an internal stretch of 70 aa (KfrAΔ136207) or even 137 aa (KfrAΔ187323) from the α-helical tail did not impair the autoregulatory properties of KfrA_R751, at least at 37 °C (when overproduced), in the dissected kfrApxylE reporter gene assays with tacpwtkfrA or mutants in trans. Nevertheless, deletion of 137 internal aa did have a slight effect in pMAB08.51, which was consistent with it causing fourfold derepression of transcription from kfrAp. The N-terminal part of KfrA (KfrA1136) is sufficient for DNA binding and a basal level of repression, but its temperature sensitivity indicates decreased stability relative to full-length KfrA or derivatives containing more of the α-helical, coiled-coil domain.

The introduction of the ΔαkfrA allele (KfrA136207) into R751 had no effect on plasmid stability. A similar observation has been made previously (Williams et al., 1998) for a short in-frame deletion of kfrA_RK2 in the P7pMB1 hybrid derivative, which was the basis of the conclusion that KfrA_RK2 did not play any role in the stability of the test plasmid. Attempts to introduce a null kfrA allele into R751 were unsuccessful. However, the reverse genetic method used to introduce the kfrAΔ137343 allele led to formation of a new deletion allele, kfrAΔ186323, in R751 (R751ΔC₂kfrA). The failure to introduce the kfrAΔ137343 mutation into the R751 background may be due to either loss of repressor activity (it causes a 30-fold decrease in repression when in excess) or some other function that is essential for plasmid or cellular survival. However, almost all the extra DNA that was present in mutant R751ΔC₂kfrA was missing from the ΔαkfrA allele which was successfully introduced into R751. It therefore seems more likely that the failure to recombine kfrAΔ137343 into R751 was due to the effect of this deletion on gene expression. As already mentioned, the kfrAΔ187323 allele, which arose spontaneously in R751ΔC₂kfrA, seems to increase transcription from kfrA into kfrBC when tested in the promoter-probe vector (pMAB05.81). In the natural context of R751, this mutation does not seem to have a drastic effect on production of truncated KfrA, suggesting that the presence of KorA and KorB, which also repress kfrAp, suppresses this slight defect. Nevertheless, the inability to completely inactivate kfrA in R751 implies that, despite kfrAp being subject to repression by KorA and KorB, KfrA is the lead regulator and is necessary to maintain correct expression levels.

Our estimate of intracellular KfrA concentration is about 7000 monomers per cell, which would give a concentration well in excess of 1 µM. Fluorescence microscopy indicated that the majority of the these KfrA molecules were present in the same type of plasmid DNAprotein foci as previously identified for KorB of the IncP-1α plasmids (Bignell et al., 1999). Since there are no additional copies of the putative KfrA binding sequence elsewhere in the genome, other than at kfrAp, this implies either proteinprotein interaction or presence of multiple, less specific, binding sites. So long as the protein is distributed evenly around the genome, there should still be a small fraction of the promoter unoccupied at any time, thus allowing enough expression to replenish the levels of KfrA. The very weak promoter for kfrB and kfrC may also ensure synthesis of these proteins under steady-state conditions.

The nature of the KfrADNA complex formed should be of interest and will be the subject of future work. One of the functions of the KfrABC complex appears to be to aid in the formation of clusters containing multiple copies of the plasmid since, in its absence, the observed foci of KorB bound to O_B sites in the IncP-1 genome and KfrA foci split into multiple, smaller signals. This could reflect the fact that individual molecules are no longer sticking together. The structure of KfrA is reminiscent of SMC-like proteins (Soppa, 2001; Rhodes et al., 2004) and, as such, it may be important in linking proteinplasmid DNA complexes together, perhaps helping them to condense into more compact units and this, in turn, may be essential for their correct localization within the bacterial cell.

Thus KfrA may act as a specific plasmid nucleoid organizer that is important for stable inheritance and plasmid survival. This would be consistent with the instability phenotype observed in the R751ΔC₂kfrA and R751ΔkfrB mutants constructed, and the importance of kfrC for the plasmid may be reflected by our inability to create a mutation in this gene. The kinetics of R751ΔC₂kfrA and R751ΔkfrB plasmid loss may not be typical for defects in the stability determinants. The strains with R751Tet^RΔC₂kfrA and R751Tet^RΔkfrB do not have an impaired growth rate when compared to the strain with stable R751Tet^R and the R751Tet^RΔαkfrA derivative. The ΔC₂kfrA mutation does not seem to disturb significantly the level of expression of the kfrA operon, although we cannot exclude an effect on the general level of expression of other genes, as it leads to the defect in plasmid clustering and, putatively, DNA condensation. Introduction of the kfrA allele on a medium-copy plasmid in trans to E. coli NEM(R751Tet^R) led to 90 % loss of R751 after 10 generations (data not shown), so the attempt to complement the ΔC₂kfrA mutation by supplying extra WT KfrA was declined. Previous studies suggest that the IncP-1 active partitioning system requires additional functions to work efficiently (Thorsted et al., 1998). It is conceivable that KfrABC could be part of this apparatus, but we were unable to demonstrate any direct interactions between K