The FtsK protein is required for septum formation in Escherichia coli and as a DNA translocase for chromosome processing while the septum closes. Its domain of action on the chromosome overlaps the replication terminus region, which lies between replication pause sites TerA and TerC. An extra Ter site, PsrA*, has been inserted at a position common to the FtsK and terminus domains. It is well tolerated, although it compels replication forks travelling clockwise from oriC to stall and await arrival of counter-clockwise forks. Elevated recombination has been detected at the stalled fork. Analysis of PsrA*-induced homologous recombination by an excision test revealed unique features. (i) rates of excision near PsrA* may fluctuate widely from clone to clone, a phenomenon we term whimsicality, (ii) excision rates are nevertheless conserved for many generations, a phenomenon we term memorization; their metastability at the clone level is explainable by frequent shifting between three cellular states – high, medium and low probability of excision, (iii) PsrA*-induced excision is RecBC-independent and is strongly counteracted by FtsK, which in addition is involved in its whimsicality and (iv) whimsicality disappears as the distance from the pause site increases. Action of FtsK at a replication fork was unexpected because the factor was thought to act on the chromosome only at septation, i.e. after replication is completed. Idiosyncrasy of PsrA*-induced recombination is discussed with respect to possible intermingling of replication, repair and post-replication steps of bacterial chromosome processing during the cell cycle.
Examples of bacterial memory which involve epigenetic systems or programmed genetic responses have been described (Casadesus and D'Ari, 2002). They often result in heritable phenotypic heterogeneity of a clonal population (Avery, 2006; Dubnau and Losick, 2006; Smits et al., 2006). We report here the discovery of a novel example of metastability in Escherichia coli. It concerns the frequency of recombination occurring at a replication fork arrested by a supernumerary replication pause site inserted into the terminus region.
When an extra replication pause site (Ter sequence; Hill, 1996) is inserted so that one of the two oppositely oriented replication forks must stop and wait for the arrival of the companion fork, homologous recombination may be triggered at the stalled fork. This Ter-induced recombination can be detected by an excision test (Louarn et al., 1994; Corre and Louarn, 2005). We noticed by chance that the frequency of this event fluctuates widely from clone to clone. The fluctuation is too large to have a single stochastic cause. We term this hyper-variability, whimsicality. We show here that the rates of Ter-induced recombination, although variable, may nevertheless be conserved for several generations. We term this conservation, memorization.
On the E. coli chromosome, a replication fork arriving at a Ter site occupied by the Tus protein is halted for some time when the pause site is appropriately oriented (Hill, 1996; Louarn et al., 2005). Ter site layout compels replication forks to collide somewhere between oppositely oriented TerA and TerC sites, which delimit a region of 275 kb known as the terminus or the replication fork trap (Fig. 1). TerC is equidistant from oriC in either direction and it has been shown that fork collision occurs near this site in the strain family used here (Louarn et al., 1994).
A supernumerary pause site, PsrA*, with the sequence and orientation of TerC, has been inserted in the replication fork trap, at about 220 kb – one tenth of a replication arm – to the left of TerC (Fig. 1). The extra site inhibits movement of rightward-moving forks (clockwise with respect to the circular genetic map), and termination occurs at or very close to PsrA* in the modified chromosome (Louarn et al., 1994). If replication proceeds at normal velocity from TerC to PsrA*, the replication time should increase by 10% (about 4 min at 37°C; Chandler et al., 1975) and the stalled fork should be immobilized for twice as long. No noticeable disadvantage during growth for 30 generations and no special requirement for homologous recombination were detected when PsrA* is present (not shown).
The recombination assay is based on positive selection for recombination between two direct repeats and loss of the intervening material. In essence, it involves the curing of a λ prophage derivative (about 36 kb of λ-specific material) flanked by direct repeats of 2.8 kb, as shown in Fig. 2. Excision always involves identical flanking sequences whatever the chromosomal insertion site. This is because initial prophage insertion occurred by homologous recombination between tet elements located in the phage DNA and chromosome. Excision is selectable because the loss of the prophage restores the ability for cured bacteria to form colonies at 42°C (lysogeny is maintained at low temperature only, owing to a temperature-sensitive phage repressor). The frequency of prophage loss at a given locus therefore reflects the frequency of recombination, or recombination density, at this locus. This is measured as the cured bacteria frequency (CBF), typically in clones derived from a single bacterium after about 30 generations of growth and consisting of about 109 cells.
Figure 1 shows the map positions tested here. Most analyses have been performed on strains carrying the prophage inserted in the tyrR locus, at about 6 kb to the left of PsrA*. Other prophage insertion sites are zci233::Tn10 and zcj1378::Tn10 (respectively 24 kb and 13 kb to the left of PsrA*) and zcj152::Tn10 (7 kb to the right of PsrA*). In the absence of PsrA*, the density of recombination differs at these positions: CBF at tyrR and zcj152 are at least fivefold higher than at zci233 and zcj1378 (Corre and Louarn, 2005).
PsrA* and prophage insertion sites map within the domain of action of FtsK. FtsK manifests an activity in a region juxtaposed to, but larger than, the terminus region (Fig. 1), that we have named the FTSK domain (Corre and Louarn, 2005). This domain is characterized by a 10-fold increase of recombination density when the C-terminal domain of FtsK is inactive. It has been suggested that FtsK deficiency increases probability of DNA trapping by the septum which in turn results in DNA lesions. This ‘territorial’ recombination was discovered thanks to the same excision test used here. In an FtsK-deficient strain, prophage excision near PsrA* may thus be a combination of territorial and PsrA*-induced recombination.
Like the chromosome, λ DNA is polarized. When inserted in the FTSK domain, a λ prophage inhibits FtsK activity if its polarization is opposite to that of the surrounding region (Corre and Louarn, 2005). Only prophages allowing normal FtsK activity (termed λiso prophages) have been used here.
The frequency of cured bacteria may fluctuate widely in the presence of PsrA*
Forthcoming data show that the presence of PsrA* increases CBF (an effect termed PsrA*-induced excision) at positions to the left of the site. No increase was detected at zcj152 to the right of the pause site (not shown). Recombinogenic events at the stalled fork are thus restricted to the replicated region (Fig. 3). Figure 4 A shows CBFs at the tyrR locus in the presence of PsrA*, measured on 106 independent clones of strain LN4444 (Table 1) grown in identical conditions and each containing about 109 bacteria (Experimental procedures). CBF values were widely distributed, the average being close to 10−2. The significance of this large range of fluctuation became apparent when it was compared with CBF distributions generated in silico on the assumption that prophage loss occurs stochastically with a constant frequency per cell generation (Experimental procedures). Each distribution presented in Fig. 4B involved 1000 simulations to ensure statistical significance. Four prophage excision rates were tested: 10−4; 3 × 10−4; 10−3; 3 × 10−3. None of the simulations fits with the actual CBF distribution, whose width cannot therefore be explained by stochastic prophage loss occurring with a unique probability. The observed distribution corresponds better to an excision frequency per cell generation which fluctuates from cell to cell. Strain LN4444, although originating from a single cell, behaves as if it were a mix of several clonal subpopulations each characterized by a specific excision rate, as those presented in Fig. 4B. We give the term whimsicality to this large variability of excision rates.
Table 1. List of strains.
These notations, which indicate that the λTSKinv prophage has recombined into a Tn10 itself inserted into a certain chromosomal locus, have been simplified in ‘tyrR:: λiso’ or ‘zcj1378:: λiso’ through text and figures.
Insertion of the prophage confers resistance to kanamycin, streptomycin and spectinomycin and causes loss of the tetracycline resistance due to Tn10.
We took advantage of the strong linkage between recBC genes and thyA gene to cross the Δ(recBC::Ap) mutation in strains already Apr due to PsrA-Ap presence.
Whimsicality does not preclude memorization of PsrA*-induced excision rates
The in vivo data reported in Fig. 4 were collected in the course of experiments designed to detect whether or not wide variations of CBF between two clones of the same strain are accidental. If variation is purely stochastic, all bacteria from either clone should share the same excision probability, so that sublines of the clones that initially differed in CBF should exhibit the same average CBF. This test was applied to three strains harbouring the prophage inserted at tyrR, which varied by the absence of PsrA* (strain LN4446, Fig. 5A) or its presence, and by whether the site inhibits fork movement (Tus+ strain LN4444, Fig. 5C) or not (Δtus; strain LN4565, Fig. 5B). In the case of strains LN4446 and LN4565, in which PsrA* is absent or inactive (Fig. 5A and B), subclones from clones representing the extremes of CBF displayed the average CBF, indicating that probability of excision is the same in all cells of the population and that CBF variations are simply stochastic. In contrast, in strain LN4444, in which the rightward-moving fork must stop at PsrA*, groups of subclones (in general a ‘group’ contained six independent subclones) tend on average to display the same CBF as the parent clone, whether high or low (Fig. 5C). For example, the clone presenting the highest CBF in group 1 gave rise to subclones (group 2) whose average CBF was similarly high, and this average rate was maintained following further subcloning (group 3). We term this persistence, memorization.
Memorization coexists with whimsicality. Thus, starting from a clone with moderately high CBF, evolution to clones displaying low CBF could be observed (Fig. 5C, 0 to 3), as could evolution from low CBF to high (Fig. 5C. 3 to 6). The final range of fluctuation was large, from clones showing almost no effect of PsrA* to clones displaying a 20-fold higher CBF than in the absence of PsrA*. Groups of subclones displayed in general an average CBF similar to that of their parent clone, indicating that excision rates, high or low, are heritable and may be transmitted through the 60 generations that separate a bacterium at the origin of a clone from the bacteria tested in the group of subclones.
The number of excision rate classes might be limited to three. We could easily maintain and use in subsequent genetic analyses two subpopulations of LN4444 bacteria: LN4444L with subclones displaying an average low CBF of 0.13 10−2; LN4444H with subclones displaying an average high CBF of 1.7 10−2. Growth rates of both populations appeared similar, which indicates no obvious selective advantage of one state over the other. The third class, moderately high CBF, is exemplified in experiments reported Fig. 5C: starting from a clone with low CBF, and testing systematically subclones displaying the highest excision rates (Fig. 5C, 4 to 6), we obtained clones with moderately high CBF, which gave rise to a majority of clones with similarly moderate CBF (Fig. 5C, 4 and 5).
Whimsicality and memorization explained by metastability of excision rates
The data are consistent with coexistence of three bacterial states: L, M and H, displaying, respectively, low, medium and high excision probabilities, no one having a selective advantage over the others. Any state can shift stochastically to an adjacent one, which can be symbolized as:
Whimsicality and memorization of PsrA*-induced excision may result from these stochastic shifts. Various trials were performed as explained in Experimental procedures to simulate the experimental outcome. A state-shift probability of 10−2 per cell generation was the value retained. One hundred independent computations provided the following mean progeny composition after 30 generations: populations issuing from L bacteria contained 78–79% L, 18–19% M and 2–3% H bacteria; populations issuing from H bacteria displayed the same (but inverted) figure; populations issuing from M bacteria contained 63–65% M, 17–18% L and 17–18% H bacteria. Thus, when a clone issuing from an L or an H bacterium is subcloned, the phenotype of most subclones is that of the initial clone, representing memorization. The situation is different for M bacteria: with two state changes available, the progenies of these bacteria display a higher rate of phenotypic change, representing whimsicality. The model is valid to explain the situation in every strain showing whimsicality (see next sections), considering the limited number of clones analysed.
The value chosen for shift frequency is obviously too high to be compatible with gene mutation as a cause of metastability of excision rates. Even if the actual values differ somewhat from that chosen here, the simulation led us to suspect that excision rate is controlled by an epigenetic process or a programmed genetic rearrangement. The experiments now to be described indicate that whimsicality is due to cell to cell variation in a process involving FtsK. Analogous situations of heritable phenotypic heterogeneity have been referred to as bistability, when two alternative phenotypes coexist (Dubnau and Losick, 2006). As the number of alternative phenotypes is higher here, we adopt the term metastability to describe the present situation.
FtsK inhibits PsrA*-induced excision and is required for metastability
TyrR maps within the FTSK domain, which we had defined as the zone within which inactivation of FtsK translocase causes territorial stimulation of prophage excision frequency (Corre and Louarn, 2005). In this work, we had also noticed a very high CBF at tyrR in a PsrA*-harbouring derivative of the ftsK1 mutant. The extent and stochasticity of these stimulations are further investigated here. First, territorial recombination due to FtsK deficiency, which accounts for a 10-fold higher CBF in ftsK1 strain LN4454 (lacking the C-terminal domain of FtsK and PsrA*) than in the wild type (compare data of Fig. 6B with that of Fig. 5A), appeared to be subject to ordinary stochastic variation (Fig. 6B). Second, the ftsK1 mutation was introduced by phage P1 transduction into PsrA*-harbouring bacteria of the LN4444L and LN4444H populations, which displayed very different CBF (see above). Six independent ftsK1 transductants from each population were tested for prophage excision. As shown in Fig. 6A, all transductants obtained from LN4444L and LN4444H (collectively known as strain LN4453) displayed high CBF, close to 3.10−2 on the average, and subclones from the extremes also displayed this high CBF, indicating that variations were simply stochastic. Whimsicality observed in the FtsK+ strain was lost. It thus depends on FtsK.
A priori, recombination due to fork stalling at PsrA* should be independent of territorial recombination, a proposal further supported by results reported in the next section. Contributions to prophage excision at tyrR of territorial and PsrA*-induced recombination must then be additive. PsrA*-induced excision is thus estimated to be responsible for a CBF of 2 × 10−2 in clones of strain LN4453. This value coincides well with the maximum values observed in FtsK+ strain LN4444 (Figs 4 and 5). Thus, PsrA*-induced excision is blocked at its highest rate and displays no whimsicality when the C-terminal domain of FtsK is lacking. This suggests two conclusions: (i) FtsK decreases the occurrence of PsrA*-induced excision and (ii) this antagonistic activity is whimsical, from fully operative in L cells to nearly inactive in H cells.
Prophage excision at tyrR is RecBC-dependent when territorial but RecFOR-dependent when due to PsrA*
In E. coli, the RecBCD system is specialized in repair of double-stranded ends (DSE) while the RecFOR system is specialized in repair of DNA gaps (Kuzminov, 1999). Knowledge of the pathway involved in territorial and PsrA*-induced excision at tyrR could indicate whether these events, which are both counteracted by FtsK, are instigated by similar perturbations of DNA. The Δ(recBC)-ApR deletion-insertion mutation was crossed into female strains LN4446, LN4444L and LN4444H by conjugational transfer. Results of CBF measurements are shown in Fig. 7A and B. In the absence of PsrA*, removal of RecBC in strain LN4594 reduced CBF two- to threefold (compare with RecBC+ strain LN4446 in Fig. 5), indicating a contribution of the RecBCD pathway. In contrast, the frequency and whimsicality of PsrA*-induced excision were similar in RecBC– cells and in their Rec+ parents: in the strain collectively named LN4593, PsrA*-induced excision ranged from very low level in most recBC derivatives of LN4444L to high in most recBC derivatives of LN4444H (compare with the data of Fig. 5C). Analysis of cured clones (not shown) indicated that prophage excision results from bona fide homologous recombination, because the expected fraction, about 30%, has retrieved the TcR resistance gene (see legend of Fig. 2).
The recBC mutation was also crossed into ftsK1 derivatives. Viability of the double mutants was impaired, making CBF measurements less reliable. Nevertheless, two clear conclusions could be drawn (data not shown). First, in the absence of PsrA* (strain LN4596), prophage excision at tyrR was strongly reduced compared with that in the parental ftsK1 Rec+ strain LN4454 (see Fig. 6), and fell to the same level as in the FtsK+recBC strain LN4594. This suggests that territorial recombination in the FTSK domain is caused by DSE. Second, in the presence of PsrA* (strain LN4595), CBF remained high (about 10−2), although somewhat reduced compared with the parental ftsK1 Rec+ strain LN4453 (see Fig. 6). This is consistent with PsrA*-induced recombination, but not territorial recombination, remaining operative in this strain. Because they involve different pathways, PsrA*-dependent recombination and territorial recombination must be induced by distinct alterations to DNA.
In complement, the role of the RecFOR pathway has been investigated. The recF::CmR mutation has been transferred by phage P1-mediated transduction in strains LN4446 (−PsrA*) and LN4444 (+PsrA*). Average CBF value in strain LN4446 was 0.68 × 10−3 (Fig. 5A). RecF– derivatives of this strain (strain LN4597 in Table 1) displayed an average CBF of 0.57 × 10−3. Thus, in the absence of PsrA*, the loss of RecF function does not modify palpably the frequency of prophage excision at tyrR. A different situation was observed in the presence of PsrA*. The LN4444 culture used in the transduction experiment contained mostly bacteria in M/H states, giving rise to subclones showing high CBF (average for 12 subclones: 1.35 × 10−2). The study of 12 independent recF transductants (strain LN4598 in Table 1), reported in Fig. 7C, revealed an average CBF reduced to 2.1 × 10−3, with no whimsicality (a possible consequence of the focusing on bacteria in M/H states). Thus, inactivation of recF downgrades severely (by a factor 8), although not totally, PsrA*-induced excision. The RecFOR pathway clearly favours this recombination at the tyrR position.
PsrA*-induced excision is rarely associated with chromosome dimer formation
As FtsK is needed for resolution of chromosome dimers by Xer recombination at dif (Steiner et al., 1999), the rate of PsrA*-induced excision in the ftsK1 mutant could be underestimated if this recombination is frequently associated with generation of a chromosome dimer: in this case, recovery of viable excisants requires proficiency of the dimer resolution system (see Fig. 3). To test this possibility, we analysed PsrA*-induced excision in a xerC mutant unable to perform chromosome dimer resolution (Blakely et al., 1993). Results presented in Fig. 8 show that PsrA*-induced excision displays the same rate and whimsicality in the xerC strain as in the Xer+ parent (compare with Fig. 5C). Chromosome dimers must be rarely formed as the prophage is excised at the stalled fork. Their fate cannot influence measurements of excision rates in the ftsK1 background.
Effect of distance from PsrA* on metastability
We analysed PsrA*-induced excision at two prophage positions further from PsrA* than tyrR, zci233 and zcj1378, respectively, 24 and 13 kb to the left of PsrA*. Results are presented in Fig. 9 for zcj1378. In FtsK+ bacteria, CBF was 10-fold more frequent in the strain harbouring PsrA* (LN4505, Fig. 9A), than in the strain without it (LN4480, Fig. 9B), clearly showing PsrA*-induced excision. Two differences from excision at tyrR were noticed: (i) no clone losing the prophage as infrequently as in the control strain was found and (ii) test of subclones provided no convincing evidence for memorized whimsicality of excision rates.
PsrA*-induced excision at zcj1378 is enhanced by the ftsK1 mutation. CBF was about sevenfold higher in the ftsK1 strain than in the corresponding FtsK+ strain (Fig. 9C and B). Its value, 1.8 × 10−2, is higher than the summed contributions of PsrA*-induced excision in the FtsK+ strain (about 3 × 10−3, Fig. 9B) and territorial excision in the ftsK1 background (about 2 × 10−3; Corre and Louarn, 2005). At position zcj1378, FtsK reduces the frequency of PsrA*-induced excision to one fifth of that observed in ftsK1 cells, and this without whimsicality.
Excision at position zci233 displayed similar characteristics. It is therefore possible that FtsK antagonizes excision without whimsicality at all positions remote from PsrA*. FtsK probably interferes with PsrA*-induced excision throughout the whole region where the event is detected (about 40 kb to the left of PsrA*; Corre and Louarn, 2005), while this interference is whimsical only near the pause site.
The peculiarities of PsrA*-induced excision illustrate the intricate interactions between processes of chromosome replication, repair, organization, partition and cell division. They deserve further comment.
The recombination process at the stalled fork
PsrA*-induced recombination has been detected exclusively on the replicated DNA, and is not performed by the RecBC pathway, a strong indication that features favouring recombination at the arrested fork are not DSE. We suspect that some DNA gaps persist until replication is completed, as long as 20% of the replication time, and allow RecA filament assembly. As the RecFOR proteins are known to increase the probability of entry of RecA at DNA gaps (Lusetti et al., 2006), our observation that the recF mutation downgrades PsrA*-induced excision provides support to this model. The absence of DSE at the stalled fork may also characterize the normal termination process, because both situations ultimately involve collision with a fork migrating in the opposite direction.
RecBC independence of PsrA*-induced recombination does not actually contradict previous observations that recombination induced by replication fork stalling at an extra Ter site involves the RecBCD pathway (Horiuchi and Fujimura, 1995; Bidnenko et al., 2002). In these experiments the extra pause site was located in the middle of a replication arm and the stalled fork was presumably disassembled by a second fork on the same arm, yielding a DSE and triggering the RecBCD pathway.
Because excision is a splice event based on the joining of two regions initially separated by the lost sequence, every such event involving both nascent sister chromatids would result in formation of a chromosome dimer (Fig. 3C), and would depend on the dif/Xer resolution system to be recovered as viable excisants. Because inhibition of chromosome dimer resolution has no effect on PsrA*-induced excision frequency (Fig. 8), the elevated frequency of excision caused by fork stalling at PsrA* does not stem from recombination between sister chromatids. Even if PsrA*-induced recombination is restricted to the zone replicated by the stalled fork, this does not favour sister chromatid exchanges. Sister chromatid exchange might be limited because chromatids segregate soon after replication. The observation implies that most RecA-DNA complexes formed at the stalled fork are not involved in recombination events, because the sister chromatid is the only place where sequences in single copy could find a target for homologous recombination.
FtsK at the stalled fork: adding a new role to a known activity
The activity of FtsK which antagonizes PsrA*-induced as well as territorial excision is probably the translocase, whose deficiency is the major alteration of the ftsK1 protein (Bigot et al., 2004). However, not only substrates but also tempos of action appear to differ in the two phenomena.
The FtsK translocase is known to control the DNA movements involved in dimer resolution (reviewed in Bigot et al., 2004; 2005). Our present finding that territorial recombination requires RecBC indicates that FtsK deficiency triggers DSE formation in the FtsK domain. As for the DSEs observed near dif when Xer recombination is inhibited (Louarn et al., 1994; Corre et al., 2000; Prikryl et al., 2001), most should be generated from unprocessed dimers or catenation loops, after trapping of FTSK domains by the septum in the last phase of cell cycle (Corre and Louarn, 2005).
At the stalled fork, FtsK probably interferes with the initiation phase of excisive recombination. If so, this FtsK activity should anticipate termination, because the suspected recombinogenic gaps should be filled as soon as replication is completed by the counter-clockwise fork. Even if PsrA* delays the timing of termination, as reckoned above, this step should be achieved long before cell division, because the D period between normal termination and division is almost as long as the generation time in fast-growing wild-type cells (Cooper and Helmstetter, 1968).
Thus, the involvement of FtsK in PsrA*-induced excision might reveal that this protein performs some functions in the cell cycle long before it processes chromosome dimers. FtsK could be an accessory factor at the replication fork, which would be consistent with the presence of skewed Kops sequences along the whole of the chromosome arms (Bigot et al., 2005). Alternatively, it is possible that the FtsK molecules active in PsrA*-induced excision belong to the ring structure that anticipates the septum, detectable at or soon after termination (Aarsman et al., 2005; Wang et al., 2005).
Whimsicality helps to orient the predictions on FtsK action at the fork
FtsK might modulate chromosome replication, or organize DNA at the stalled fork. In the first hypothesis, FtsK might impinge on the time required for replication of each chromosome arm. If replication of the left arm is delayed with respect to that of the right arm (refer to Fig. 1), fork stalling at PsrA* could be too brief or infrequent to cause formation of recombination substrates. Conversely, if replication of the left arm is the faster, the resulting long fork arrest at PsrA* would favour recombination. Because high PsrA*-induced excision characterizes the FtsK– phenotype, FtsK might favour slow replication of the left arm, and whimsicality might reflect metastability of this activity. However, this model does not match with the observation that PsrA*-induced excision is whimsical only at the position closest to the pause site.
Alternatively, the ability of FtsK protein to wrap around DNA and to mediate its polarized mobilization may reduce formation or stability of recombination substrates at the stalled fork, or limit the search for a recombination partner 40 kb distant. The following scheme is possible: in the FtsK+ context, when FtsK is absent from the fork or inactive, the probability of excision is high; when FtsK is present or active, the probability of excision is low; in addition, the alternative FtsK states of presence/activity and absence/inactivity are conserved through multiple generations. Shifts between states of excision frequency would involve an epigenetic control on FtsK presence/activity at the fork. To account for the lack of whimsicality at positions remote from PsrA*, we suggest that an FtsK complex is always present at the stalled fork and functions uniformly in all cells at the remote positions, but becomes whimsical when it comes closer to the site. As FtsK mobilizes the PsrA* region from left to right (referring to Fig. 1), according to the orientation of Kops motifs (Bigot et al., 2005), FtsK movement might cease at some distance from the stalled replication machinery. This distance might be the primary whimsical parameter.
Hints of metastability of other chromosome-processing events
Intriguing variations of terminus region movements have been observed during the late phase of the cell cycle, which may be interpreted in terms of heritable phenotypic heterogeneity. Wang et al. (2005) reported that in some cells both copies of a terminus marker stay at mid-cell where the septum forms, while in other cells segregation of these copies is asymmetrical. In this case, one copy stays close to the encroaching septum and the other migrates towards a cell pole. A similar marker on the other replichore also behaves in this way except that the mobile copy migrates towards the other cell pole, and choice of poles is remembered. These authors also observed rare examples of symmetrical migration of sister terminus markers towards opposite cell poles. Earlier reports (Niki et al., 2000; Li et al., 2003) found sister terminus copies to stay near the closing septum. Protocol differences may explain these discrepancies but, because cytological observations are made on a few hundred cells only, it is also possible that different but temporarily memorized patterns of terminus positioning have been observed in different studies.
Two final comments may be added. First, whether similar phenomena occur at TerC in wild-type chromosomes is not yet known. The formation of an FtsK ring on a replicating chromosome could be an unusual phenomenon that PsrA* makes frequent by increasing replication time. Second, losing a prophage is an act of evolution for a bacterium. The uncommon situation analysed here may be considered as an evolutionary process spontaneously changing speed in a genetically stable population submitted to no external stress. This could have been of some significance in evolution of the terminus region.
Bacteriophage and bacterial strains
Most E. coli K12 strains (Table 1) derive from CB0129, an F– W1485 thi leu thy deoB or C supE (Bird et al., 1972), except the tus-deleted strain which derives from TH838. In this MG1655 Δtus805 strain, a gift from Tom Hill, replication pause sites are inactive (Henderson et al., 2001). All marker positions are given in kilobases with reference to the published E. coli K12 sequence (Blattner et al., 1997). The PsrA* insertion (TerC sequence) is tagged by an ampicillin resistance determinant (Ap) and introduces a 3.4 kb chromosomal deletion between positions 1393.4 and 1396.8 (Louarn et al., 1994). Tn10 insertions have been previously described (Corre and Louarn, 2005). The ftsK1 mutation, a chloramphenicol resistance (Cm) insertion, codes for an FtsK protein devoid of part of the linker domain and of the entire C-terminal domain (Diez et al., 1997). The xerC mutation is a deletion-insertion tagged by a gentamicin resistance (Gen) determinant (Aussel et al., 2002). The recBC mutation is a deletion-insertion tagged by an ampicillin resistance determinant (Louarn et al., 1994). The recF mutation is an insertion of a Cm determinant at the Eagl site located in the middle of the gene (a gift from Marc Bichara, unpublished). All gene combinations were generated by bacteriophage P1 transduction or conjugational transfer, as described in Miller (1992). The relevant genotype of strains used in this work is given in Table 1. Bacteriophage λTSKInv has been previously described (Corre et al., 2000). It carries a tet determinant with the tetA gene inactivated at two positions by kanamycin resistance (Knr) and streptomycin/spectinomycin resistance (Sm/Spr) determinants (the TSK element; Fig. 2). The phage lacks of its specific integration-excision system and harbours a cI857 ind mutation that makes the repressor thermosensitive and SOS-insensitive.
Media and bacterial growth
Bacteria were grown in L-broth medium and plated on the same L-agar medium (Miller, 1992). Sodium citrate (0.8 10−3 M) was added when needed to avoid reinfections by phages P1 or λ. The following antibiotic concentrations were used: ampicillin: 20 μg ml−1; chloramphenicol: 20 μg ml−1; gentamicin: 7.5 μg ml−1; kanamycin: 20 μg ml−1; streptomycin: 25 μg ml−1; tetracycline: 15 μg ml−1.
The recombination assay
Local propensity to recombination was measured by an excision test involving the curing of a λTSKInv prophage (Corre et al., 2000) inserted by homologous recombination into a chromosomal Tn10 or tet insertion as shown in Fig. 2A. The phage repressor, coded by a cI857 ind gene, is temperature-sensitive (Ts), so that lysogens are killed by induction of the lytic cycle when incubated at 42°C. Integrative recombination between positions of KnR and Sm/SpR insertions was systematically retained, giving rise to KnR Sm/SpR TcS Ts bacteria. Prophages (36 kb of λ DNA) inserted by this mechanism are flanked by direct repeats of a total length of 2.8 kb. Bacteria able to form colonies at 42°C (cured bacteria) result from excision by homologous recombination and loss of the prophage (Fig. 2B), which must have occurred prior to the shift at 42°C. CBFs reported here were measured after about 30 generations of growth in L-broth from a single bacterium. Typically, a 24 h-old colony grown at 30°C on L-agar medium is picked up, suspended in 0.5 ml of L-broth and incubated for 24 h at 30°C, so that the stationary phase is reached. Convenient dilutions, chosen to get about 500 colonies per plate, are then plated on L-agar and incubated at 30°C or 42°C. No repeat was performed. In spite of this, the distribution of number of bacteria per clone was satisfactorily narrow: more than 90% of Rec+ FtsK+ Xer+ clones analysed contained between 5 × 108 and 1.1 × 109 bacteria. This shows a good reproducibility of growth conditions and viable cell counting.
Expected clone-to-clone variation of CBF. The theoretical model to account for stochastic prophage loss is based on the following assumptions: (i) when excision occurs, the next division yields a cured bacterium plus a lyzogenic one, (ii) excision is irreversible, (iii) frequency of excision p is constant over generations, (iv) presence or absence of λ prophage does not affect growth rate and (v) generation time is constant. To simulate the evolution of the frequency of cured bacteria in a clone issuing from a single lyzogenic bacterium and determined after a limited number of generations, a computer program was written in Perl language. The program simulates synchronized cell division as a recursive function. Each cell is characterized by its type (A = lyzogenic bacterium, B = cured bacterium) and its age (in generation time). According to the model, B bacteria give rise to B bacteria. An A bacterium gives rise to a progeny determined by a random drawn from uniform distribution. If the random number is less than p (probability of λ prophage excision), one daughter is an A cell and the other a B cell, otherwise both daughters are A cells. Bacterial growth is stopped at a user-defined number of generations, Ng, fixed at 30. In order to reduce the computational time, only half of the cells (randomly chosen) are allowed to pursue growth after a user-fixed generation number, Nthr. Nthr was chosen in function of the number of B bacteria in the population at this generation, which must be high enough to avoid sampling-caused drift. This artificial selection maintains the number of cells around 2Nthr in the next generations, so that the evolution of the population could be followed for more generations without exorbitant calculation times. Nthr was generally fixed at 20. Simulations with different values of Nthr allowed us to check that means and medians of cured clone distributions are unaffected, but that dispersion around these values increases for too low values of Nthr (data not shown).
Metastability of excision rates. A modification of the above program was set up to predict population evolution assuming that the rate of prophage excision may spontaneously shift between different values along successive generations in a clonal population. We decomposed the A cell state in three types displaying alternative excision rates: medium (M), low (L) and high (H). Transitions are allowed and reversible between L and M and between M and H types, all with the same probability. With this set-up, the mother excision rate is transmitted to daughters except when a transition occurs. When the number of generations increases, the computation shows that frequency of each cell types converges to 1/3, as expected. These programs are available on demand (contact: firstname.lastname@example.org).
We thank warmly Jacqueline Corre for her participation in preliminary phases of this work, and David Lane for his interest and the time spent in correcting the manuscript in its successive forms. J.M.L. thanks Jean-Pierre Claverys for welcoming him as an emeritus professor in the LMGM, and François Cornet for making available laboratory facilities. This work received no specific support from any funding agency, and for this reason is dedicated to the memory of the Unknown Toothpicker.