Segregation of the Escherichia coli chromosome terminus

Summary

We studied the segregation of the replication terminus of the Escherichia coli chromosome by time-lapse and still photomicroscopy. The replicated termini lie together at the cell centre. They rapidly segregate away from each other immediately before cell division. At fast growth rate, the copies move progressively and quickly toward the centres of the new-born cells. At slow growth rate, the termini usually remain near the inner cell pole and migrate to the cell centre in the middle of the cell cycle. A terminus domain of about 160kb, roughly centred on the dif recombination site, segregated as a unit at cell division. Sequences outside this domain segregated before division, giving two separate foci in predivision cells. Resolution of chromosome dimers via the terminus dif site requires the XerC recombinase and an activity of the FtsK protein that is thought to align the dif sequences at the cell centre. We found that anchoring of the termini at the cell centre and proper segregation at cell division occurred normally in the absence of recombination via the XerC recombinase. Anchoring and proper segregation were, however, frequently disrupted when the C-terminal domain of FtsK was truncated.

Introduction

Escherichia coli contains a single circular chromosome. Replication begins at a specific origin (oriC) and proceeds bidirectionally to terminate within a broad region on the opposite side of the chromosome (Hill, 1996). The replication forks do not appear to move around the DNA. Rather, they are thought to be anchored at the cell centre, and the DNA moves through them (Koppes et al., 1999; Lemon and Grossman, 2000). The newly replicated DNA appears to be directed away from the anchored forks in such a way that two separate chromosome masses are formed as DNA replication proceeds (Dingman, 1974; Lemon and Grossman, 2001). The terminus region, being the last to be replicated, should be present at the cell centre during termination, and should be the last sequence to segregate away from the cell centre as separation of the sister chromosomes is completed. This has been confirmed by fluorescent labelling techniques showing that the terminus is at the cell centre in cells starting to divide (Gordon et al., 1997; Niki and Hiraga, 1998; Li et al., 2002). Evidence has been presented that the termini eventually segregate away from each other to the cell quarter positions, which will become the new cell centres in the next generation (Gordon et al., 1997; Li et al., 2002).

Because of the helical nature of the DNA, the terminated products are catenated. In addition, recombination between sister chromosome regions often causes the completed sister chromosomes to be covalently linked as a chromosome dimer. These linkages between sister chromosomes must be fully resolved before chromosome segregation can be completed. Decatenation requires the activity of topoisomerase IV (Deibler et al., 2001). Chromosome dimer resolution occurs by site-specific recombination at the dif site that is located in the terminus region, carried out by the XerC/XerD recombinase (Steiner and Kuempel, 1998a; Perals et al., 2001). Chromosome dimer resolution is obligatorily coupled to cell division (Steiner and Kuempel, 1998b). The FtsK protein is required for resolution via the dif site, and also may aid in clearing the resolved products from the plane of cell division (Steiner et al., 1999; Perals et al., 2001; Sawitzke and Austin, 2001; Corre and Louarn, 2002). As replication termination occurs at the cell centre, these resolution events occur there also (Barre et al., 2000; Perals et al., 2001; Aussel et al., 2002). FtsK forms a ring structure at the cell centre in dividing cells (Wang and Lutkenhaus, 1998). Segregation of the terminus appears to occur simultaneously with the physical separation of the daughter cells, suggesting a close coupling between cell division and final resolution of the two sister chromosomes (Corre and Louarn, 2002; Li et al., 2002).

Here, we followed the fate of the terminus by inserting a single parS site in or near it and visualized it using the parS binding protein GFP-Δ30ParB (Li et al., 2002). Like any means of sequence localization, it must be born in mind that the labelling method might itself influence DNA localization. However, this mutant protein–DNA complex does not appear to influence the localization of plasmids (Li and Austin, 2002a) and gives very different localization patterns for the chromosome origin and terminus (Li et al., 2002). Thus, it seems likely that the positions of the protein-parS foci reflect the normal positions of the marked sequences in the cell.

Results

The dynamics of terminus segregation at moderate growth rate

The GFP-Δ30ParB protein binds to its cognate P1 parS site and spreads out as multiple copies onto the surrounding DNA (Li and Austin, 2002a). If the parS site is introduced into the host chromosome at a defined site, a bright fluorescent focus is formed on the site that can be followed by fluorescence photomicroscopy (Li et al., 2002). A parS site at 33.7 min, near the dif site in the chromosome terminus, formed a single focus in 98% of the cells. Most foci were near the cell centre, with larger cells having the focus at the centre, over the constriction caused by nascent cell division (Li et al., 2002). These cells were grown at 30°C with a 55 min generation time where replication is always ongoing because a new round of replication begins before the previous one is terminated (Li et al., 2002; Fig. 1).

Figure 1. The state of the cell cycle. The state of the cell cycle was determined as previously described (Li and Austin, 2002b; Li et al., 2002). As the cytometer measures the cell size and DNA content of each individual cell, the relationship between the two can be plotted once the cell size output is calibrated against the length distributions of the cells as measured by microscopy. The graphs show the size distributions of cells with one (open diamonds), two (closed squares) and four (closed triangles), origins of replication. The arrows show the size classes of the average newborn (N) and dividing (D) cells in the population as determined by measurement of the cell images in photomicrographs as previously described (Li et al., 2002). The size of the average cell at initiation (I) is derived from the graphs and represents the size class in which 50% of the cells have duplicated their origins. A. Data for cells growing in M63 glycerol medium at 30°C (180 min generation time). From the graph, it can be deduced that the average cell at division contains two completed chromosomes (diagram below) such that the dividing cells have two origins (closed circles) and two termini (closed squares). Initiation in the majority of cells does not occur until just after cell division. B. Data for cells growing in AB glucose casaminoacids medium at 30°C (55 min generation time). The average dividing cell has two half-replicated chromosomes.

Figure 2A shows the result of time-lapse photography of a cell growing under the same conditions on an agarose plug on the microscope slide. The fate of the terminus focus could be followed for approximately two generations. The observed cell was initially well advanced in the cell cycle, with a slight central constriction and the terminus focus precisely placed at the cell centre. As the cell prepared to divide, the focus split and the two products moved away from each other progressively and rapidly to the centres of the daughter cells. The events were duplicated in the second cycle in the two daughter cells. However, in one of the daughters, one focus appeared to ‘hang up’ at the centre for a while and became elongated before releasing and occupying its new position (white arrow, Fig. 2A). Observation of seven other cells showed very similar results to those shown in Fig. 2A, with such ‘hang up’ events occurring in about one-third of the divisions. The initial rate of focus separation appeared to be quite rapid, as the majority of foci divided and separated by a micron or so between 2.5 min frame intervals. The second divisions in Fig. 2A are clear examples of this.

Figure 2. Time-lapse photomicroscopy of cell labelled at the terminus.

A. The cell was growing in AB minimal glucose medium with casaminoacids at 30°C. The chromosome was labelled at 33.7 min on the chromosome, near the dif site. The cells had a 55 min generation time in liquid medium, which was approximately maintained on the microscope slide as judged by the increase in cell length. The images were taken at 2.5 min intervals and the numbers indicate the elapsed time in minutes. The white arrow marks a focus that appears to hang up at the cell centre, lagging behind its sister before segregating.

B. The cell was grown in M63 minimal glycerol medium at 30°C with a 180 min generation time. After application of the cells to the microscope slide under the same conditions, the cell grew even more slowly. Numbers indicate elapsed time in minutes.

The precise timing of segregation is often difficult to define in time-lapse series because the ends of the divided cells frequently stay in contact with each other, obscuring the exact time of cell division. However, it is clear from the second cycle in Fig. 2A that cell division occurred at or soon after terminus segregation because, in these cases, the daughter cells moved away from each other. In addition, we can be confident that segregation occurs very close to cell division in all cells, based on the still images of the cell population: Almost all (98%) of the cells, including those that are close to cell division, have a single terminus focus placed at the cell centre (Li et al., 2002; Fig. 3). If terminus segregation occurs at or very near the time of cell division, very few cells will have two foci because division will immediately produce two one-focus cells. These single foci are usually at or near the cell centre (Fig. 3), because migration of the daughter foci from the old division plane to the new cell centres is rapid; it occupies less than 10% of the cell cycle (Fig. 2A).

Figure 3. The positions of foci in cells labelled at the terminus. Cells labelled at the terminus (33.7 min) were grown in M63 minimal glycerol medium (180 min generation time, A) or AB minimal glucose medium with casaminoacids (55 min generation time, B) at 30°C and the living cells photographed (above). The positions of the fluorescent foci of 200 of the resulting cell images were plotted as a function of measured cell length as shown below. Lines show the positions of the cell centres and the cell 1/4 and 3/4 cell lengths.

Terminus segregation at slow growth rate

Figure 3A shows a typical field of cells growing with a 180-min generation time at 30°C. These cells complete a round of replication before initiating a new one (Fig. 1A). Almost all of these slow growing cells had a single focus (245 of 253 scored), showing that segregation of the terminus occurred just before cell division. However, unlike faster growing cells, only the larger cells had the foci clustered around the cell centre (Fig. 3A). The results of a time-lapse experiment at this growth rate are shown in Fig. 2B. The overall pattern is similar to that at the faster growth rate (Fig. 2A). However, after the initial separation of the foci, they did not migrate rapidly to the centres of the new cells. Rather, they remained in the vicinity of the pole for some time. The foci eventually came to the new centres to start the next cycle. Similar results were obtained with time-lapse series of eight other cells; although the positions of the foci with respect to the pole varied somewhat, most foci (18 of 20) did not move to the new cell centres until at least half-way through the cell cycle. Thus, in slow growing cells there is a considerable delay in the migration of the foci to the cell centres after segregation. On average, the focus only reaches the vicinity of the centre at about half way through the cell cycle.

The position and extent of a chromosomal domain for terminus segregation

We constructed strains with nine different insertion positions for the parS site in the vicinity of the terminus (Fig. 4). The insertion points were numbered in order of their map positions with S4 being the 33.7 min dif– proximal site used in the experiments above (Fig. 4). Inspection of Fig. 4 shows that markers in the interval S2 through S7 behaved very similarly to S4. Examination of the population of the cells growing at the faster growth rate (55 min generation time at 30°C) showed very few cells with two foci (<2%). This indicates that all sequences within this region segregate very late in the cell cycle, just as the cells are dividing. We propose that this region constitutes a terminal segregation domain. It covers about 3.5 min of the chromosome, or approximately 160 kb of DNA. Outside of this interval, the number of cells with two foci increases at least five- to tenfold. The transition from the terminal segregation domain to the flanking sequences is illustrated in Fig. 5. The microscope images show that almost all dividing S7 cells have a single central focus (96%). However, most (∼ 60%) of deeply divided S8 cells have two foci. The chief difference in these two cell populations is the existence of a class of large, dividing cells in the S8 population that have already segregated the marked sequence (Fig. 5). Note that the hisC marker (45.1 min, Fig. 4) used by Gordon et al. (1997) to mark the vicinity of the terminus lies well outside the terminal domain defined here. Consistent with their results we found that hisC segregated well before cell division. We found that approximately 13% of all cells marked at hisC had two foci (data not shown).

Figure 4. The segregation of markers flanking the terminus. The positions of the parS marker inserts in the chromosome are shown on the map. The table shows the proportion of cells with two foci when cells with the given marker were grown at a 55 min generation time at 30°C. The presence of a high proportion of cells with two foci indicates that the relevant marker segregated well before cell division.

Figure 5. Cells labelled at positions S7 and S8. The S7 and S8 marker positions span the border of the terminus segregation domain. The cells were grown at the faster growth rate (55 min generation time at 30°C). S7 dividing cells have one focus, showing that segregation occurs very close to cell division. Most S8 dividing cells have two foci, showing that segregation occurs significantly before cell division. Examples of the dividing cells are shown, along with examples of a unique class of predivision cells that have two foci at the centre aligned approximately with the short axis of the cell. An analysis of the population based on scoring approximately 300 cells is shown below. The percentage of cells in each illustrated class is shown for S7 and S8 labelled cells. The values for cells labelled near the dif site (S4) are shown for comparison.

Recombination at dif and the XerC recombinase are not required for positioning of the terminus at the cell centre

Chromosome dimer resolution occurs at the cell centre and is coupled to cell division. It appears to involve recognition of sequences (RAG motifs) which flank the dif site over an extended region (Corre and Louarn, 2002). These motifs are highly biased to being in one orientation on one side of dif and the opposite orientation on the other side. They may act as guide sequences to direct the dif site to the recombination complex (Corre and Louarn, 2002). Thus, the positioning of the terminus segregation domain at the cell centre, and the timing of segregation to correspond to cell division, might reflect the handling of the terminus by the dif recombination machinery. We therefore asked if known components of this machinery were required for terminus positioning and segregation. A deletion mutant of xerC was introduced into the strain containing a parS insert near dif (S4), and the positions of the marked region observed (Fig. 6). As previously described, some cells formed irregular filaments (Fig. 6), presumably because they contain chromosome multimers that prevent the completion of cell division (Perals et al., 2001). The filaments often contained more than one terminal focus, though usually not as many as might be expected from the number of partially divided cells present (Fig. 6). This may reflect the guillotining of the DNA and subsequent loss of the terminus sequences as previously described for xer mutants (Prikryl et al., 2001).

Figure 6. Cells defective for XerC and the FtsK C-terminus. Typical fields of cultures of xerC ^– and ftsK1::cat cells cells, labelled near dif (S4), are shown. The graphs show the positions of the single foci in cells of normal length (<4 µM). Open symbols; foci in non-dividing cell. Closed symbols; foci in cells with a division constriction. Examples of dividing cells of normal length from the two mutant cultures are shown at the bottom. The growth rates, as measured by optical density, were ∼ 55′ for both wild-type and mutant cells.

Approximately 10% of the xerC mutant population formed filaments. The remaining 90% of the cells were of normal size and most contained a single terminus focus (Table 1). Very few of the normal length cells contained two foci (Table 1). The single foci were clustered around the cell centre in all but the smallest cells in the population (Fig. 6). Thus, most Xer-deficient cells behave just like their wild-type progenitors with respect to terminus segregation. Typical dividing cells with a single central focus are shown at the bottom of Fig. 6. We conclude that the XerC protein is not essential for positioning the terminus at the cell centre, and that segregation of the terminus is restrained until late in the cell cycle in the presence or absence of recombination at the dif site.

Table 1. . Proportions of cell types in cultures of xerC and ftsK1 cells.^a

Marker
^a The positions of foci and developing septa were recorded in ∼ 300 cell images from each culture. Only those cells under 4.6 µM in length were scored. This excludes the filaments that form in the mutant cultures. All images containing foci corresponded approximately to one of the five classes shown above. The proportion of cells in each class (as a percentage of the total) is shown.
S4	88	<1	1	10	0.5
xerC	83	<1	0.6	15	1
ftsK1	67	5	0.8	18	5

The FtsK C-terminus is important for the precise positioning of the chromosome termini

In the ftsK mutant population (Fig. 6), many more cells formed filaments than in xerC mutant cells (∼ 50%). There were also filamentous and normal length cells without any foci. These cells contained chromosomal DNA as judged by DAPI staining (see Experimental procedures) and may have lost their termini by breakage and degradation as has previously been shown for Xer recombinase deficient mutants (Prikryl et al., 2001).

Of the focus-containing cells of normal length (i.e. within the 1.8–4.6 µM size range of wild-type cells), 85% had a single focus (Table 1). Some cells had foci at or near the cell centre and some dividing cells had a focus at the point of constriction (Table 1). However, many were broadly distributed with respect to the long axis of the cell (Fig. 6). There were also a significant number of normal length cells that had two foci (5% of the normal length cells, Table 1). We conclude that the precise positioning of the termini is disrupted in many of the cells lacking the FtsK C-terminus. Figure 6 shows examples of deeply septated cells that have two foci, or, as is seen occasionally, one focus not in the middle of the cell. We conclude that ftsK1::cat cells frequently misalign their termini such that a large number (approximately half) of the normal sized cells lack a single focus aligned at the cell centre before division.

Discussion

Termination of replication occurs well before cell division; the time between the two being the D period of the cell cycle (Bremer and Dennis, 1987). The replicated termini appear as a single focus right up to the time of cell division. Thus, the two completed termini must be held together at the cell centre for some time before terminal segregation occurs. What restrains the termini to this position? It is probable that the sequences are recognized by a structure at the cell centre, and held there, sufficiently close together that they appear as one focus.

The existence of such a mechanism has previously been proposed to account for the late resolution of dimer chromosomes by the dif-XerCD recombination system. Using a novel recombination assay, Corre and Louarn (2002) concluded that the FtsK protein guides the regions flanking the dif sites in dimeric chromosomes towards the in-growing septum so that synapsis and recombination can occur. It was proposed that the alignment involves recognition of multiple sequences (Rag motifs) which flank dif site in polarized orientations (Corre and Louarn, 2002). As judged by this assay, precise alignment of the dif region was lost when the C-terminal domain of FtsK was deleted, although some mechanism for restraining the termini to the general vicinity of the cell centre remained (Corre and Louarn, 2002). Steiner and Kuempel (1998b) found that recombination at dif occurred well after replication of the dif locus and that it required cell division. They proposed that the replicated dif sites are sequestered from each other until cell division, so that recombination is restricted to that time. Thus, a structure at the cell centre, probably involving the FtsK protein, may capture and hold the termini of dimeric chromosomes apart for some time before cell division, at which time they synapse and recombine at the dif site.

Here, we used direct visualization of the terminus to investigate its segregation. The fates of termini were followed in all cells: not just those that contain chromosome dimers. If we assume FtsK recognizes and sequesters the termini not only in dimeric chromosomes, but in all completed chromosome pairs, our direct physical evidence appears to correlate well with the deductions of Steiner and co-workers and Corre and co-workers (Steiner and Kuempel, 1998b; Corre and Louarn, 2002). We show that a broad region, centred on the dif site, acts as a domain for terminus segregation, such that all replicated sequences within it are held together precisely at the cell centre until just before cell division. At that time, the two sister regions segregate as a unit to the daughter cells. It seems likely that the terminus segregation domain can be equated with the extensive region of polarized Rag sequences that is thought to be held in alignment with its sister region for recombination (Corre and Louarn, 2002). In the absence of the FtsK C-terminal domain, the alignment fails in a high proportion of the cells.

However, our data do not necessarily imply a direct role for FtsK in terminus alignment. It is possible, though unlikely, that this misalignment is not related to the other topological problems documented for the truncated FtsK protein, but rather is a result of indirect effects of the ftsK1::cat insertion on other genes. More likely, misalignment may be a secondary consequence of the known defects caused by the mutation. For example, the termini may initially be aligned properly in ftsK1 cells, but, because of the very high level of failure of dimer resolution and/or decatenation, many cells are unable to resolve their chromosomes. In these cells, cell division is delayed or blocked. This might lead to subsequent terminus misalignment due to a breakdown of the anchoring of the termini as cell division is aborted. In support of this latter interpretation, there is an approximately twofold excess of dividing cells in the ftsK1::cat population over the wild-type and xerC populations (Table 1). Moreover, there are just as many dividing cells with single central foci in the ftsK1 population as in the wild type (Table 1). Thus, successful cell divisions in ftsK1::cat cells may generally proceed via a wild-type terminus capture and segregation pathway. The excess cells that show terminus misalignment may all be products of chromosome resolution errors that block cell division. Thus, the C-terminal domain of FtsK may not be required for terminus alignment per se. Perhaps this role is played by the N-terminal segment of the protein, or by some other cell component as yet not defined. Time-lapse studies of FtsK-deficient cells might resolve these issues. However, it has not yet proved possible to obtain such data because of the poor growth of the cells.

Although terminus segregation occurs very close in time to cell division, the localization of the termini at the cell centre clearly occurs much earlier and persists for some time. It seems probable that the completed sister termini are initially held close, but separate, from each other at the cell centre, perhaps at different positions on the FtsK ring. Completion of the septum would then bring the sequences together, so that, if necessary, they can recombine at dif just as the cell divides. Finally, the completion of septation might disrupt the FtsK anchor, thus releasing the termini and allowing segregation to proceed.

Cells labelled within the terminus domain include a unique class of predivision cells which have two foci closely apposed at the cell centre and aligned with the short axis of the cell (Fig. 5). It is possible that these are cells in which the two sister termini are held separately at the centre by FtsK. However, this cannot require the FtsK C-terminus, because they are also seen in ftsK1::cat cultures (data not shown).

We show that the immediate fate of the termini after separation at cell division depends on the growth rate. We suggest that this is due to the different replication states of the chromosomes at the different growth rates. Figure 1 illustrates the determined chromosome cycles. It is likely that newly synthesized DNA is added to the outside borders of the replicating nucleoid (Sawitzke and Austin, 2001). At the slow growth rate, the termini would be connected to the DNA that is expected to lie at the inner border of the nucleoid mass at the time of cell division (Fig. 7). At the faster growth rate, the new round of replication, overlapping the first, would produce a mass of newly replicated material at the borders of the nucleoid, with the segregating termini connected to the old DNA which is in the nucleoid interior (Fig. 7). We suggest that the initial phase of terminus segregation consists of DNA condensation. This gathers the free terminus to the bulk of the unreplicated sequences with which it is immediately linked. Thus, the terminus segregates with the inner border of the nucleoid at the slow growth rate, but would go rapidly to a more central position at the higher growth rate (Fig. 7). Further segregation would be a result of the progression of the unreplicated portion of the chromosome toward the centre of the nucleoid as replication proceeds, and, ultimately, to the drawing of the terminus into the replication machinery at the new cell centre as termination occurs.

Figure 7. Terminus segregation pattern varies with growth rate. As previously proposed (Sawitzke and Austin, 2001), the newly replicated DNA (thick filament) is added to the outside faces of the old DNA (thin filament), which is being fed into the replication factory (triangIes). A. At slow growth rate, the cell divides with complete chromosomes. The termini (white squares) are held at the cell centre (perhaps by the by the FtsK protein) until cell division, when they are released. Because of DNA condensation (Sawitzke and Austin, 2001), they segregate with the inside face of the nucleoid to which they belong. They progress toward the centre later, when replication pushes the old DNA toward the cell centre. B. At faster growth rates, replication of the new DNA is well underway at cell division. The termini are held at the cell centre and are released at division. Because they are attached to the old DNA, condensation rapidly displaces them to the vicinity of the cell centre. At both growth rates, replication ultimately draws the terminus into the factory, where it is duplicated. The products are captured to the new cell centre, restarting the cycle.

Experimental procedures

Strains construction by λ Red-recombination

Escherichia coli strains containing P1 parS sites in the host chromosome were derived from strain DY330, using the λ Red-promoted recombination method of Yu et al. (2000). Primers were used to amplify the P1 parS sequence linked to a kanamycin resistance cassette as previously described (Li et al., 2002). The primers used are listed in Table 2. Transformants were selected on LB plates at 30°C with 30 µg ml⁻¹ of kanamycin. The structure of the recombinants was confirmed by PCR analysis. Strain CC4713, which has the parS integrated into the chromosome 20kb from dif site, was described previously (Li et al., 2002). The xerC strain used in this study was CC4724. It was produced by P1 transduction of the xerC::cat insertion mutation from DS941 (Chatwin and Summers, 2001) into CC4713. Similarly, CC4725 was produced by introducing ftsK1::cat into CC4713 from WM974 (Yu et al., 1998).

Table 2. . Primers used to insert the P1parS site at various points in the chromosome.^a

Strain	Location (min)	Primers
a . Underlined sequences in primers are complementary to pALA1073 (Li et al., 2002), a template plasmid containing the P1 parS site.
ParS1 (CC4722)	27.5 5′CGATGCATTGTCAAACGACGAACTGCGCTGGGAACCGTAACGATAAAAAGCCGAAGCCTTAAAC 5′GGAAGCAGCGGTATGGGTGCATCAGGAGCGCATTTTCTGACACAGCTTTAGAGCGTTTTGCGAT
ParS2 (CC4721)	32 5′GTCCATCGATATACATAGGATGTTGAACGGGTACTGACATCGATAAAAAGCCGAAGCCTTAAAC 5′GGTGCTCATCGAGGCGATGGAAAGTCGCTCGTTACGTTAACACAGCTTTAGAGCGTTTTGCGAT
ParS3 (CC4720)	33 5′GTGAGGTAAAAGCATCGACGTGGTACACCTGCGGTTTCATCGATAAAAAGCCGAAGCCTTAAAC 5′CTTTACTCGTCGATATCAAATTGTGCGTTCGCGGCGATAACACAGCTTTAGAGCGTTTTGCGAT
ParS4 (CC4713)	33.7	(Li et al., 2002)
ParS5 (CC4716)	34.7 5′ACTGTGCACCGGGAAAATCGCGACGATTTAGTGTTTTCACCGATAAAAAGCCGAAGCCTTAAAC 5′CCGTACCAGCACTATGCGTTACGGTGAGGAAGTTGTCTGACACAGCTTTAGAGCGTTTTGCGAT
ParS6 (CC4717)	35.4 5′TACCTGCTGCTTGTGCAAGTTTTGTTTTTGAACCGAAATACGATAAAAAGCCGAAGCCTTAAAC 5′ACCTGATGCGTTCGCCGATAGTTAAATTTTTTGTTTCCATCACAGCTTTAGAGCGTTTTGCGAT
ParS7 (CC4718)	36.3 5′TTTGATCATTCCACGGCTGCACCTGTATGTTGCAGACTAGCGATAAAAAGCCGAAGCCTTAAAC 5′GCCAGAACAGATGGTCGGCAGTATGAAAGCCGGGCGCTAGCACAGCTTTAGAGCGTTTTGCGAT
ParS8 (CC4719)	41 5′ACTGAAAATGATGGATCTGATCAGCAAAATCGATAAGTAACGATAAAAAGCCGAAGCCTTAAAC 5′CGATAAATACCAGCACAATTTGAAGAGTGGTAATCTCCATCACAGCTTTAGAGCGTTTTGCGAT
ParS9 (CC4723)	43.1 5′TCATTAAGAACTCCTGGTAGTCAAAGTTAAAGTGCGGCATCGATAAAAAGCCGAAGCCTTAAAC 5′GGCTATTAAACGGTTACGCACTAAACTGGGTAAGTTATAACACAGCTTTAGAGCGTTTTGCGAT

Growth of cells

Cells were grown in AB minimal medium (Clark and Maaloe, 1967) supplemented with 0.2% glucose, 1 µg ml⁻¹ thiamin, 1 µg ml⁻¹ biotin, 50 µg ml⁻¹ casamino acids and 50 µg ml⁻¹ ampicillin, or in M63 minimal medium (Miller, 1972) containing 1 µg ml⁻¹ thiamin, 1 µg biotin and 0.2% glycerol. Expression of the fusion protein was induced by adding 100 µM IPTG to the medium for 180 min at 30°C. The cells reached an OD₆₀₀ of 0.1 and were then used directly for microscopy and flow cytometry. The doubling time of the cells at 30°C in AB glucose casaminoacids medium was ∼ 55 min and approximately 180 min in M63 glycerol medium.

Fluorescence microscopy of cell populations

Living cells were observed and measured as previously described (Li et al., 2002). In some cases, the cell were fixed and stained with DAPI (4′,6-diamidino-2-phenylindole) in order to visualize the DNA as described (Sun et al., 1998).

Flow cytometry and analysis of the cell cycle

The number of origins per cell was determined by the replication run-off method (Skarstad et al., 1995) as previously described (Li et al., 2002). The data was normalized to the cell size distribution of the culture as determined by measurements of the micrographs of the same cell population (Li et al., 2002). The length of the average newborn cell (L₀) can be determined from the average cell length (L_A) by the formula L₀ = L_A/2 log_n 2. The length of the average cell at division is twice this value. The length of the average cell at initiation was determined as the point at which 50% of cells had doubled there number of origins as determined by flow cytometry (Li et al., 2002).

Time-lapse microscopy

Cells grown at 30°C were spread and immobilized on agarose-coated microscope slides as described (Glaser et al., 1997). Two microlitres of the appropriate medium containing 100 µM IPTG and 1.2% agarose was pipetted onto a microscope slide. The coated slide was left to solidify and dry in the open air at room temperature for 8 min. A 5 µl drop of cell suspension was placed on the surface of the agarose and immediately covered with a coverslip, sealed with petroleum jelly, and observed at 30°C for up to 15 h. Images were recorded at defined intervals as previously described (Li and Austin, 2002b).