Different localization of SeqA-bound nascent DNA clusters and MukF–MukE–MukB complex in Escherichia coli cells

Authors


Abstract

MukF, MukE and MukB proteins form a complex that may participate in the organization of folded sister chromosomes in Escherichia coli. We have found that a MukB–GFPuv4 fusion protein is observed as discrete fluorescent foci, which are localized within cellular spaces occupied by nucleoids, but not at the constriction site of cell division in living cells. In contrast, MukB–GFPuv4 is distributed throughout the whole cell when either MukF or MukE is absent. Statistical analysis revealed that most newborn cells have two foci of mukB–gfpUV4 at one-quarter and three-quarter positions in the cell length and one focus of SeqA-bound nascent DNA at or near the middle of the cell. Subsequently, the single SeqA focus divides into two foci, and then these migrate to the one-quarter and three-quarter positions. Before cell division, most long cells have two SeqA foci and four MukB–GFPuv4 foci. In early stationary phase, SeqA foci disappear, but one or two foci of MukB–GFPuv4 remain. We discuss the reorganization and proper arrangement of folded sister chromosome in the cell quarter positions, which are performed after release from the long-time cohesion of sister chromosomes.

Introduction

The mukB gene is essential for proper partitioning of sister chromosomes in Escherichia coli cells (Niki et al., 1991). The mukB gene constitutes an operon together with mukF, mukE and the non-essential gene smtA in the order smtA–mukF–mukE–mukB (Yamanaka et al., 1996; see Fig. 1A). Null mutants of each muk gene share the same phenotypes: medium-dependent temperature-sensitive lethality, anucleate cell production and a defect in structure and proper positioning of the chromosome (Yamanaka et al., 1996). The mukB gene codes for a 170 kDa protein, MukB (Niki et al., 1992; Yamanaka et al., 1994). MukB has globular domains at both ends and forms a homodimer by means of an α-helical, coiled-coil rod structure in the centre of the protein (Niki et al., 1991; 1992). Biochemical studies revealed that purified MukB has ATP/GTP-binding, ATPase/GTPase and DNA-binding activities (Niki et al., 1992) and that the C-terminal globular domain participates in the DNA-binding activity (Saleh et al., 1996). Electron microscopy revealed that two MukB molecules form a homodimer in an antiparallel arrangement (Melby et al., 1998). MukF, MukE and MukB form a complex in vitro, and the C-terminal globular domain of MukB participates in complex formation with MukF and MukE (Yamazoe et al., 1999; see Fig. 1B). A truncated protein consisting of the first 342 amino acid residues of MukB hydrolyses ATP and GTP in the presence of Mg2+ ions (Lockhart and Kendrick-Jones, 1998a). The N-terminal domain of MukB binds the FtsZ polymer of E. coli and microtubules of eukaryotic cells in vitro in the presence of nucleotides (Lockhart and Kendrick-Jones, 1998a, b). The crystal structure of the N-terminal domain of MukB from E. coli has been described (van den Ent et al., 1999). Mutations of the topA gene encoding topoisomerase I suppress all phenotypes of mukF, mukE and mukB null mutants (Sawitze and Austin, 2000). The mukB null mutant is hypersensitive to novobiocin (Weitao et al., 1999). A null mutation of the seqA gene suppresses all phenotypes of the mukB null mutant (Weitao et al., 1999). In contrast, we reported that null mutations of seqA and/or dam genes partially suppress temperature-sensitive growth, but not anucleate cell formation and hypersensitivity to novobiocin of muk null mutants (Onogi et al., 1999).

Figure 1.

Composition of the mukFEB operon (A) and a proposed model for the structure of the MukFEB complex (B).

The chromosomal DNA of E. coli is methylated by DNA adenine methyltransferase (Dam methylase) at N6-adenine sites within GATC sequences (Bakker and Smith, 1989). Approximately 20 000 GATC motifs are unevenly distributed in the E. coli genome (Henaut et al., 1996). After initiation of DNA replication from the chromosomal origin (oriC), newly replicated DNA strands are hemimethylated at GATC sequences. The newly replicated hemimethylated oriC DNA segments are sequestrated to the cell membrane. The seqA gene is involved in the sequestration process, i.e. the proposed trapping of newly replicated hemimethylated oriC DNA into membranes, resulting in delayed access by Dam methylase (for a review, see Crooke, 1995). Purified SeqA protein preferentially binds hemimethylated oriC and other DNA segments containing multiple GATC sequences (Brendler et al., 1995; 2000; Slater et al., 1995; Brendler and Austin, 1999) and also binds fully methylated oriC DNA (Slater et al., 1995; Brendler and Austin, 1999), but does not bind unmethylated DNA.

Indirect immunofluorescence microscopy of SeqA and fluorescence microscopy of a SeqA–green fluorescent protein (GFP) fusion revealed that SeqA is localized as discrete foci in exponentially growing wild-type cells of E. coli (Hiraga et al., 1998; 2000; Onogi et al., 1999; for a review, see Hiraga, 2000). Formation of visible SeqA foci depends on DNA methylation (Hiraga et al., 1998; Onogi et al., 1999) and ongoing replication of the chromosomal DNA (Hiraga et al., 2000). One SeqA molecule is able to interact with another SeqA molecule in cell extract (Onogi et al., 1999). These results indicate that SeqA molecules bind newly replicated hemimethylated DNA segments and form SeqA–DNA clusters within cells. A single SeqA focus localized in the middle of a wild-type cell initially separates into two foci. Thereafter, these foci migrate abruptly in opposite directions to the one-quarter and three-quarter positions in a stage of the cell cycle, when a clear FtsZ constriction ring is formed. These SeqA foci remain tethered there until the cell divides (Hiraga et al., 1998). Time-lapse analysis of the SeqA–GFP fusion in living cells indicates a rapid migration of the fluorescent foci, appearing as abrupt ‘jumping’ from mid-cell to the one-quarter and three-quarter positions (Onogi et al., 2000). SeqA foci are abnormal in size and localization in mukB null mutant cells (Hiraga et al., 1998; Onogi et al., 2000).

Using E. coli cells synchronized for the initiation of chromosome replication, it was shown that the replication origin oriC was replicated at mid-cell, and the resultant oriC sister copies are linked to each other for a substantial period of generation. The oriC copies then separate from each other and migrate in opposite directions towards both pole-proximal borders of the nucleoid after the bidirectional migration of SeqA foci from mid-cell to the cell quarter positions (Hiraga et al., 2000; Hiraga, 2000). Finally, the oriC copies are localized to pole-proximal borders of nucleoids at the cell division time (Niki and Hiraga, 1998; Niki et al., 2000). The oriC copy in newborn cells appears to migrate again to mid-cell before initiation of chromosome replication (Niki et al., 2000). The molecular mechanism of the long-time pairing of oriC copies is as yet unclear.

In the mukB null mutant, SeqA foci are abnormal in size and subcellular localization (Hiraga et al., 1998; Onogi et al., 2000). This strongly indicates that the MukB protein participates in the separation and bidirectional migration of the SeqA clusters. However, biochemical knowledge on MukFEB in vitro is insufficient to clarify the role of the MukFEB complex in vivo. It is therefore very important to analyse subcellular localization of the MukFEB complex and the SeqA protein in the cell cycle of E. coli.

In this work, we have succeeded in observing the localization of MukB using a bacterial strain producing a fusion protein of MukB with GFPuv4 (green fluorescent protein UV4). The MukB–GFPuv4 fusion protein is localized as discrete fluorescent foci in E. coli living cells and also fixed cells, which were analysed by indirect immunofluorescence microscopy. We have statistically analysed the localization of the MukB–GFPuv4 foci compared with that of SeqA foci and found that the localization pattern of MukB–GFPuv4 foci is quite different from that of SeqA foci. We discuss the mechanism of reorganizing sister chromosomes at the one-quarter and three-quarter positions after release from the long-time cohesion of sister chromosomes.

Results

Clustering of the MukB–GFPuv4 fusion protein in living cells

To investigate the subcellular localization of the MukFEB complex in living cells, we isolated a bacterial strain, KAT1, which expresses a MukB–GFPuv4 fusion protein. In the MukB–GFPuv4 fusion, a linker of three glycine residues links the carboxyl-terminal end of MukB to the amino-terminal end of GFPuv4, which was derived from a UV-excitable GFP (see Experimental procedures). The mukB–gfpUV4 fusion gene was substituted for the native mukB gene of the mukFEB operon in the KAT1 strain (see Table 1). Western blotting revealed that the MukB–GFPuv4 fusion protein in the KAT1 strain was expressed at the same level as the authentic MukB protein in cells of the parental muk+ strain YK1100 (data not shown). KAT1 grew normally with the same generation time as YK1100; 93 min of generation time at 30°C in MCAT2 medium. The KAT1 strain did not show temperature-sensitive colony formation and anucleate cell production like the parental strain YK1100. Thus, we concluded that the MukB–GFPuv4 fusion protein of KAT1 possesses full function as the native MukB protein.

Table 1. Bacterial strains.
StrainRelevant genotype
mukF mukE mukB Reference
  1. +, Wild-type gene.

YK1100+++ Yamanaka
et al. (1999)
KAT1++ mukBgfpUV4This paper
KAT37+ mukE::kan mukBgfpUV4This paper
KAT38 mukF::kan+ mukBgfpUV4This paper

The MukB–GFPuv4 fusion protein was localized as discrete fluorescent foci in living cells of KAT1 growing at 30°C in MCAT2 medium (Fig. 2A), whereas in a minority of cells, MukB–GFPuv4 was distributed more broadly, but not uniformly, in the nucleoid. In most short cells, MukB–GFPuv4 was localized as two foci, which were preferentially localized at the one-quarter and three-quarter positions in cell length. Long cells with a constriction at the mid-cell position had three or four foci of MukB–GFPuv4, which were never localized at the constriction site (Fig. 2G). To analyse MukB–GFPuv4 foci and nucleoids simultaneously in living cells, nucleoids were stained with DAPI (4′, 6-diamino-2-phenylindole). Foci of MukB–GFPuv4 were never localized at or near cell poles and between nucleoids (Fig. 2E and F). In contrast to KAT1, no significant fluorescent signal of GFPuv4 was detected in cells of the parental strain YK1100, which had the native mukB gene (data not shown).

Figure 2.

Subcellular localization of the MukB–GFPuv4 fusion protein in living cells. Cells were grown exponentially at 30°C or 22°C in MCAT2 medium. DAPI (10 µg ml−1) was added to a sample removed from the culture, and the living cells were observed with a fluorescence and phase-contrast microscope.

A. Strain KAT1 (mukF+mukE+mukB–gfpUV4) at 30°C.

B. Strain KAT1 (mukF+mukE+mukB–gfpUV4) at 22°C.

C. Strain KAT37 (mukF+mukE::kan mukB–gfpUV4) at 22°C. Arrows indicate anucleate cells.

D. Strain KAT38 (mukF::kan mukE+mukB–gfpUV4) at 22°C. Arrows indicate anucleate cells.

E and F. Strain KAT1 (mukF+mukE+mukB–gfpUV4) at 30°C. Left. DAPI-stained cells observed by a combined technique of fluorescence and phase-contrast microscopy.

Middle. Fluorescent foci of MukB–GFPuv4 fusion.

Right. A merged image of fluorescence from DAPI-stained nucleoids (red pseudocolour changed from blue colour by a computer) and fluorescence from MukB–GFPuv4 fusion (green).

G. KAT1 (mukF+mukE+mukB–gfpUV4) cells at different stage of the cell cycle at 30°C.

Dispersed distribution of the MukB–GFPuv4 fusion in mukE or mukF null mutants

MukB forms a complex with MukE and MukF in vitro (Yamazoe et al., 1999). To examine whether the absence of MukE or MukF affects the subcellular localization of MukB–GFPuv4, we disrupted the mukE or mukF gene of KAT1 cells. Strains KAT37 (mukF+mukE::kan mukB–gfpUV4) and KAT38 (mukF::kan mukE+mukB–gfpUV4) were derived from KAT1 (see Experimental procedures;Fig. 7C and E). Both KAT37 and KAT38 showed the characteristic phenotypes of muk null mutants: temperature-sensitive colony formation, anucleate cell production and abnormal structure and localization of the chromosome. KAT37 and KAT38 were able to grow at 22°C in MCAT2 medium, producing anucleate cells of normal size with a frequency as high as 10% of the total cells. The MukB–GFPuv4 protein was distributed throughout the whole cell, but never as discrete foci, in both nucleate and anucleate cells of KAT37 and KAT38 at 22°C (Fig. 2C and D). When plasmids carrying the native mukE or mukF gene were introduced into KAT37 or KAT38 cells, respectively, transformants grew normally as KAT1, and discrete fluorescent foci of MukB–GFPuv4 were restored in these transformants (data not shown). Thus, the formation of MukB–GFPuv4 foci was dependent on the presence of both MukE and MukF. These results indicate that MukB forms a complex with MukE and MukF in vivo as well as in vitro and that the MukFEB complex is localized as clusters in living cells.

Figure 7.

Structures of chromosomal mukFEB regions in the bacterial strains that produce the MukB–GFPuv4 fusion and structures of plasmids that were used for isolation of these bacterial strains.

A. Procedures of isolation of bacterial strain KAT1 (mukF+mukE+mukB–gfp).

B. Plasmid pKAT4.

C. Bacterial strain KAT37 (mukF+mukE::kan mukB–gfpUV4).

D. Plasmid pKAT3.

E. Bacterial strain KAT38 (mukF::kan mukE+mukB–gfpUV4).

Aa, AatII; Ag, AgeI; B, BamHI; EV, EcoRV; P, PstI; Sa, SacI; cat, chloramphenicol resistance-conferring gene; kan, kanamycin resistance-conferring gene.

MukB–GFPuv4 foci and SeqA foci in different media or different growth phase

The average number of SeqA foci increases in rich media (Hiraga et al., 1998). We analysed the number of MukB–GFPuv4 foci and SeqA foci in growing cells in two different media. The generation time of KAT1 cells was 126 min and 93 min in M9–glucose–Trp medium and MCAT2 medium, respectively, at 30°C. The average numbers of MukB–GFPuv4 foci were 2.37 and 2.66 in M9–glucose–Trp medium and MCAT2 medium, respectively, in the exponential phase (Table 2). Thus, the average number of MukB–GFPuv4 foci was similar between the two media. On the other hand, the average numbers of SeqA foci were 1.74 and 2.28 in M9–glucose–Trp medium and MCAT2 medium respectively (Table 3).

Table 2. Proportion of cells with various numbers of foci of the MukB–GFPuv4 fusion in exponential phase and early stationary phase cells.
Number
of
MukB–GFP
foci
Proportion of cells (%)
M9–glucose–Trp mediumMCAT2 medium
Exponential
phase
Stationary
phase
Exponential
phase
Stationary
phase
  • The number of cells counted was 200–290.

  • a

    . Cells without significant foci.

0a0000
1872357
257254939
3253304
4100150
50030
Table 3. Proportion of cells with various numbers of fluorescent foci of the SeqA protein in exponential phase and early stationary phase.
No. of
SeqA
foci
Proportion of cells (%)
M9–glucose–Trp mediumMCAT2 medium
Exponential
phase
Stationary
phase
Exponential
phase
Stationary
phase
  • The number of cells counted was 200–230.

  • a

    . Cells without significant clear foci.

0a01000100
1380190
2520460
380230
420120

In early stationary phase, the average number of MukB–GFPuv4 foci in KAT1 decreased; the majority of cells had one focus at the mid-cell position, and most of the remaining cells had two foci at the cell quarter positions in both M9–glucose–Trp and MCAT2 media (Table 2). In contrast, all the cells lost SeqA foci in the early stationary phase (Table 3). The disappearance of SeqA foci in stationary phase is consistent with our previous results that the formation and maintenance of discrete SeqA foci depend on ongoing replication of the chromosomal DNA, suggesting that SeqA foci are clusters of SeqA-bound hemimethylated nascent DNA segments (Hiraga et al., 2000).

Statistical analysis of the localization of MukB–GFPuv4 foci

We analysed statistically the subcellular localization of fluorescent MukB–GFPuv4 foci in KAT1 cells that were grown exponentially at 30°C in M9–glucose–Trp medium with a 126 min generation time (55–60 min at 37°C). MukB–GFPuv4 foci were also observed clearly under the growth conditions. Cells were classified into four types according to the number of MukB–GFPuv4 foci as follows: type 1, cells with a single MukB–GFPuv4 focus; type 2, cells with two foci; type 3, cells with three foci; type 4, cells with four foci. Types 1, 2, 3 and 4 constituted 8%, 57%, 25% and 10% of the total cells respectively (Table 2). The average cell length in types 1, 2, 3 and 4 was 2.29, 2.72, 3.37 and 3.75 µm respectively. Subcellular localization patterns of MukB–GFPuv4 foci in each type of cell and in total cells are shown in Fig. 3B and A respectively. In type 2 cells, comprising the majority of total cells (57%), MukB–GFPuv4 foci were localized at the one-quarter and three-quarter positions. In contrast, short cells with one MukB–GFPuv4 focus (type 1) were only 8% of total cells. Long cells had three or four foci (type 3 or 4). Cells belonging to the type 2 category had a broad range of cell length. The length of the smallest cells in the type 2 category closely matched to the average length of type 1 cells (Fig. 3B). These results indicated that most newborn cells had two foci at the cell quarter positions, and a minority of newborn cells had one focus at the mid-cell position.

Figure 3.

Statistical analysis of the subcellular localization of the MukB–GFPuv4 fusion. KAT1 cells were grown exponentially at 30°C in M9–glucose–Trp medium (generation time = 126 min).

A. Subcellular positions of MukB–GFPuv4 foci in the total population of cells. Coloured circles represent MukB–GFPuv4 foci according to various types of cells shown in (B). Position 0 in the horizontal axis indicates the mid-cell position.

B. Subcellular positions of MukB–GFPuv4 foci in various types of cells.

Comparison of localization patterns between MukB–GFPuv4 and SeqA foci in the KAT1 strain

The SeqA protein is also localized as discrete foci in a cell cycle-dependent manner in wild-type cells, but localized abnormally in mukB null mutant cells as described in the Introduction (Hiraga et al., 1998; Onogi et al., 2000). We therefore compared localization patterns of SeqA foci and MukB–GFPuv4 foci in KAT1 cells growing exponentially at 30°C in M9–glucose–Trp medium. The subcellular localization of SeqA protein was analysed by indirect immunofluorescence microscopy. Almost all the cells had discrete SeqA foci. Cells with one, two, three and four SeqA foci comprised ≈ 38%, 52%, 8% and 2% of total cells respectively (Table 3). In cells with one SeqA focus, the focus was localized at or near the mid-cell position (Fig. 4B). In cells with two SeqA foci, foci were preferentially localized at the one-quarter and three-quarter positions. On the other hand, in the same culture of KAT1, cells with one, two, three and four foci of MukB–GFPuv4 constituted 8%, 57%, 25% and 10% of total cells respectively (Table 2). These results indicated that the majority of short cells might have two foci of MukB–GFPuv4 localized at the one-quarter and three-quarter positions, whereas they might have a single SeqA focus localized at or near the mid-cell position. A minority of short cells might have one focus of MukB–GFPuv4 and one focus of SeqA at or near the mid-cell position. The majority of long cells had four MukB–GFPuv4 foci and two SeqA foci before cell division.

Figure 4.

Statistical analysis of the subcellular localization of SeqA foci. KAT1 cells were grown exponentially at 30°C in M9–glucose–Trp medium (generation time = 126 min), fixed with 80% methanol and stained for SeqA with the Cy3 compound by indirect immunofluorescence microscopic technique using anti-SeqA rabbit antibody.

A. Subcellular positions of SeqA foci in the total population of cells. Coloured circles represent SeqA foci according to various types of cells shown in (B). Position 0 in the horizontal axis indicates the mid-cell position.

B. Subcellular positions of SeqA foci in various types of cells.

To confirm the above observation, we simultaneously analysed KAT1 cells stained for both MukB–GFPuv4 and SeqA with indirect immunofluorescence microscopy using both anti-GFP monoclonal mouse antibody and anti-SeqA rabbit antiserum. The majority of short cells had two MukB–GFPuv4 foci at the cell quarter positions and one SeqA focus at the mid-cell position as expected (Fig. 5).

Figure 5.

Subcellular localization of MukB–GFPuv4 and SeqA foci. KAT1 cells were grown exponentially at 30°C in M9–glucose–Trp medium (generation time = 126 min), fixed with 80% methanol and double-stained for MukB–GFPuv4 and SeqA with the indirect immunofluorescence microscopic technique using both anti-GFP mouse monoclonal antibody and rabbit anti-SeqA polyclonal antibody raised in a rabbit. MukB–GFPuv4 was labelled with the Cy3 compound (red fluorescence), and SeqA was labelled with the Cy2 compound (green fluorescence).

Discussion

Our present results show that MukB–GFPuv4 molecules are localized as foci in living and fixed cells in the presence of both MukF and MukE. However, MukB–GFPuv4 molecules are distributed throughout the whole cell when either MukF or MukE is absent. These results indicate that MukB forms a complex with MukF and MukE in vivo, and the resultant MukFEB complex molecules are recruited into clusters in the cells. This is consistent with the previous result proving that most MukF, MukE and MukB molecules exist as the MukFEB complex in cell extracts of wild-type cells (Yamazoe et al., 1999).

We have summarized the results of statistical analyses for the subcellular localization of MukB and SeqA in Fig. 6. The majority of newborn cells has two MukB foci localized at the one-quarter and three-quarter positions and one SeqA focus localized at or near the mid-cell position. The single SeqA focus separates into two close foci. Subsequently, these SeqA foci abruptly migrate in opposite directions to the one-quarter and three-quarter positions, where MukB foci exist in advance. In most cases, long cells have four MukB foci and two SeqA foci before cell division, yielding daughter cells with two MukB foci and one SeqA focus upon cell division. In a few cases, cells with two MukB foci and two SeqA foci are divided, producing daughter cells with one MukB focus and one SeqA focus. In this type of cell with a single MukB focus, MukFEB molecules appear soon to be recruited to the cell quarter positions to form two clusters.

Figure 6.

A schema of dynamic localization of MukEFB clusters and SeqA-bound hemimethylated DNA clusters in the cell cycle. Grey circles represent MukFEB clusters. Black circles represent SeqA-bound hemimethylated DNA clusters. Major cell cycle is shown in solid arrows. Minor cell cycle is shown with broken arrows.

In a recent model, sister oriC copies first migrate in opposite directions to near the cell poles or the one-quarter and three-quarter positions immediately after replication, and sister chromosomes are condensed around the oriC copies (for example, see Møller-Jensen et al., 2000). In contrast to the above model, we reported previously that oriC sister copies are linked to each other and stay in the middle position of the cell for a substantial generation time (Hiraga et al., 2000). Furthermore, we have found that sister chromosomal DNA segments are cohesive with each other throughout almost the whole chromosomal region for a long period. All the sister chromosome cohesion is removed all at once in a late replication period (Y. Sunako et al., unpublished data). These results indicate that the sister chromosomal DNA segments must be reorganized to folded chromosomal structures in the cell quarter positions after release from the long-time sister chromosome cohesion. It is likely that the MukFEB complex participates in the reorganization of sister chromosomes to form two separated folded chromosomes. Two clusters of the MukFEB complex may act as ‘reorganization centres’ of folded sister chromosomes to arrange them into proper positions after release from the sister chromosome cohesion.

Many questions remain to be resolved as follows. How are MukFEB molecules recruited to the cell quarter positions a long time before the bidirectional migration of SeqA clusters and release from the sister chromosome cohesion? What function do MukFEB molecules have before and after release from the sister chromosome cohesion? Is the MukFEB complex is required throughout the whole cell cycle or only in a limited period(s) of generation? Does the MukFEB complex itself play a role in sister chromosome cohesion?

It has been suggested that SeqA foci are clusters of SeqA-bound hemimethylated nascent DNA behind replication folks (Hiraga et al., 2000; Onogi et al., 2000). More recently, this was demonstrated conclusively by chromatin immunoprecipitation assay using anti-SeqA antibody (M. Yamazoe et al., unpublished data) and by indirect immunofluorescence microscopy, showing that SeqA foci were co-localized with foci of nascent DNA segments that were pulse labelled with 5-bromo-2′-deoxyuridine (M. Kohiyama et al. unpublished data). SeqA molecules might disassemble and reassemble in newly replicated hemimethylated DNA segments behind replication forks. The localization of two MukFEB clusters in the cell quarter positions may be an important event for the subsequent bidirectional migration of SeqA-bound hemimethylated DNA clusters.

In early stationary phase, most cells had one or two MukB–GFPuv4 foci, whereas all cells lost significant SeqA foci. After the end of chromosome replication in early stationary phase, all the chromosomal DNA segments should be fully methylated, and SeqA may be unable to bind the fully methylated DNA segments. In dam null mutant cells, SeqA was distributed throughout the whole cell (Hiraga et al., 1998; Onogi et al., 2000), but MukB–GFPuv4 was localized as foci (our unpublished data). Thus, the formation and maintenance of MukFEB clusters are independent of DNA methylation and ongoing replication of the chromosomal DNA in contrast to the case of SeqA clusters. In fact, the purified MukFEB complex binds, at the same affinity, to all the fully methylated, hemimethylated and unmethylated oriC DNA segments containing multiple GATC sequences (our unpublished data).

The smc gene coding for a homologue of eukaryotic SMC proteins has been found in some bacterial and archaeal genomes (for reviews, see Hirano, 1999; Holmes and Cozzarelli, 2000). Bacillus subtilis mutant cells lacking the smc gene show very similar phenotypes to mukB null mutant cells of E. coli (Britton et al., 1998; Moriya et al., 1998). The B. subtilis Smc protein has a similar tertiary structure to MukB and also forms an antiparallel homodimer (Melby et al., 1998). However, no accessory subunits for Smc, such as MukF and MukE for MukB in E. coli, have been found yet, whereas all the SMC proteins found in eukaryotic cells form a complex with accessory subunit proteins (Hirano, 1999). Subcellular localization of the Smc protein in B. subtilis cells was reported previously using two different strategies (Britton et al., 1998; Graumann et al., 1998). We have found that E. coli related bacteria belonging to a limited group (Escherichia, Klebsiella, Salmonella, Yersinia, Vibrio, Actinobacillus, Haemophilus and Pasteurella) have homologues of mukF, mukE, mukB, dam, seqA and mutH (coding for hemimethylated GATC recognition endonuclease involved in mismatch repair) genes, but not of the smc gene. Presumably, a common ancestor of the E. coli related bacteria acquired these genes, but lost the smc gene during evolution (Hiraga et al., 2000; Hiraga, 2000). The Smc protein of B. subtilis failed to suppress the temperature-sensitive growth of E. coli mukB null mutant cells (our unpublished data). It is not yet clear why the MukFEB complex was substituted for the Smc protein in the ancestor.

Experimental procedures

Bacterial strains and plasmids

All bacterial strains used were derivatives of E. coli K-12: YK1100 (trpC9941) (Yamanaka et al., 1996) and JM109 (Yanisch-Perron et al., 1985). Plasmids pUC18 (Vieira and Messing, 1982), pUC4K (Vieira and Messing, 1982), pHSG399 (Takeshita et al., 1987) and pHSG415s (Hashimoto-Gotoh et al., 1981) were used. Plasmid pGFPgcn4 encoding a GFP derivative, GFPuv4, which shows much stronger fluorescence than the original GFP, was gifted by Dr Y. Ito (Ito et al., 1999).

Media and chemicals

M9 medium (Miller, 1972) was supplemented with 1 mM MgSO4, 0.1 mM CaCl2, 5 µg ml−1 thiamine, 0.2% glucose and 40 µg ml−1l-tryptophan. This medium was named M9–glucose–Trp medium. An M9–glucose–Trp medium was additionally supplemented with 0.2% casamino acids (Difco) to make MCAT2 medium. L medium has been described previously (Hiraga et al., 1998). Kanamycin, chloramphenicol and ampicillin were added to media at concentrations of 25, 15 and 25 µg ml−1, respectively, when necessary. To construct plasmid pKAT2, two chromosomal DNA segments were amplified by polymerase chain reaction (PCR) using the pair of primers 5′-GAATGATTGAACGCGGTAAATTTCGCT CAC-3′ and 5′-CGCCGCCGCCACTCGCCTGAGAAGGCG CTTCGTCAGTTCC-3′ and the pair of primers 5′-CGAA GCGCCTTCTCAGGCGAGTTAGAATTCAGCAGCAATGCC GCC-3′ and 5′-CGCCAGAATTCAGCAGCTTTGCCCGATC ATC-3′. Enzymes and reagents used for DNA manipulation were purchased from Takara Shuzo or New England Biolabs.

General methods of DNA manipulation

The following procedures used were as described by Maniatis et al. (1982): DNA digestion with restriction enzymes, agarose gel electrophoresis, purification of DNA fragments from agarose gels, and alkali–SDS mini-DNA preparation. Ligation and blunting reactions of DNA fragments were performed using a DNA ligation kit version 1 or 2 and a DNA blunting kit (both Takara Shuzo), respectively, according to the manufacturer's instructions. E. coli pulser (Bio-Rad) was used for bacterial transformation (Dower et al., 1988).

Construction of a bacterial strain carrying the mukB–gfpUV4 fusion gene on the chromosome

We isolated the E. coli strain KAT1, in which the mukB gene of the chromosome was replaced with a fusion gene encoding MukB–GFPuv4 as shown in Fig. 7A. Plasmid pKAT2 is a derivative of plasmid pHSG415s, which codes for the temperature-sensitive Rep protein for replication and the ble gene conferring ampicillin resistance. Plasmid pKAT2 carries the following DNA segment: a C-terminal half of the mukB gene (named ′mukB) fused to the gfpUV4 gene in frame (a derivative of a UV-excitable gfp gene; Ito et al., 1999) with a (GGC)3 sequence coding three glycine residues as a linker and a downstream region of the mukB gene in which the cat segment is inserted (Fig. 7A). The plasmid pKAT2 was introduced into the cells of the E. coli muk+ strain YK1100, and gene replacement between the plasmid and the chromosome was performed according to the method of Hamilton et al. (1989) as follows. Chloramphenicol-resistant transformants with pKAT2 were first isolated at 30°C. Subsequently, clones in which the plasmid was integrated into the bacterial chromosome were selected at 42°C. One of the clones was picked up, inoculated in non-selective L medium and incubated overnight at 30°C. The cells grown to stationary phase were inoculated again in fresh L medium and incubated overnight at 30°C. From the culture, chloramphenicol-resistant colonies were isolated at 30°C on L agar plates containing chloramphenicol. Independent colonies were tested for ampicillin sensitivity at 42°C and 30°C. Chloramphenicol-resistant and ampicillin-sensitive colonies were isolated at 30°C. The genomic structure of the mukB gene and its flanking regions in the cells was analysed by PCR. The resultant bacterial strain termed KAT1 has intact mukF and mukE genes followed by the mukB–gfpUV4 fusion gene on the chromosome as expected (Fig. 7A). The KAT1 strain can grow normally as the parental muk+ strain YK1100, suggesting that the MukB–GFPuv4 fusion protein functions normally in vivo.

Isolation of mukF or mukE null mutant strains derived from the strain KAT1

To construct a mukE null mutant strain expressing the MukB–GFPuv4 fusion protein, a SacI–AatII chromosomal fragment containing the mukE gene and its flanking regions was excised from pAX850 (Yamanaka et al., 1996) and blunted. This fragment was inserted into a blunted EcoRI–BamHI site of the temperature-sensitive replication plasmid pHSG415s. A 1.3 kb kan DNA fragment prepared from pUC4K (Vieira and Messing, 1982) was inserted into the AgeI site existing in the middle of the mukE gene of the plasmid, yielding pKAT4 (Fig. 7B). Plasmid pKAT4 was introduced into cells of strain KAT1 at 30°C in the presence of kanamycin. One of the kanamycin-resistant transformants was grown at 30°C and, subsequently, the culture was spread on L agar medium containing kanamycin and incubated at 42°C to isolate clones that had both the native mukE gene and the mukE::kan gene in tandem arrangement on the bacterial chromosome. One of the colonies grown was isolated and grown at 22°C, and then the culture was diluted, spread on L agar plates containing kanamycin and incubated at 22°C. Each colony grown was tested for ampicillin sensitivity at 22°C. Subsequently, ampicillin-sensitive clones were tested for temperature sensitivity at 42°C. Temperature-sensitive clones were tested for the phenotype of anucleate cell production. One of the clones showing anucleate cell production was named KAT37. It was confirmed that KAT37 had the intact mukF gene and a disrupted mukE::kan gene followed by the mukB–gfpUV4 fusion gene on the chromosome (Fig. 7C).

Using a similar strategy, we prepared plasmid pKAT3 (Fig. 7D) and isolated bacterial strain KAT38 (Fig. 7E), which had a disrupted mukF::kan gene and the intact mukE gene followed by the mukB–gfpUV4 fusion gene on the chromosome. Both strains KAT37 and KAT38 showed the same phenotypes as muk null mutants: temperature-sensitive growth, anucleate cell formation and abnormal localization of the chromosome (Yamanaka et al., 1996).

Fluorescence microscopy of living cells and indirect immunofluorescence microscopy

Bacterial cells were grown at 30°C to avoid inactivation of the fluorescent GFPuv4 protein. Living cells stained with DAPI (10 µg ml−1) for nucleoids were observed with a combined fluorescence and phase-contrast microscope. Indirect immunofluorescence microscopy of SeqA protein has been described previously (Hiraga et al., 1998). A fluorescence and phase-contrast microscope (Eclipse E800; Nikon) had a UV-2E/C filter for DAPI fluorescence, a B-2E/C filter for GFPuv4fluorescence and a Cy3 filter for Cy3 compound fluorescence. The microscope was connected with a colour chilled CCD camera C5810 (Hamamatsu Photo). Images were transferred directly to a Power Macintosh and processed using Adobe photoshop 5.5-J software. To measure cell length and distance between a fluorescent focus and a cell pole, we used an image analysing software, macscope version 2.5.5 (Mitani). Statistical analysis was carried out using statview 5.0-J software (SAS Institute).

Acknowledgements

We thank Dr Yoichiro Ito for the kind gift of the plasmid GFPuv4. We also thank Chiyome Ichinose, Yuki Kawata, Noriko Fukuda and Mizuho Yano for assistance in the laboratory. This work was supported by a grant of the Grant-in-Aid for Scientific Research (A) on Priority Areas and other grants from the Ministry of Education, Science, Sports, and Culture of Japan and CREST.

Ancillary