Functional polarization of the Escherichia coli chromosome terminus: the dif site acts in chromosome dimer resolution only when located between long stretches of opposite polarity



In Escherichia coli, chromosome dimers are generated by recombination between circular sister chromosomes. Dimers are lethal unless resolved by a system that involves the XerC, XerD and FtsK proteins acting at a site (dif) in the terminus region. Resolution fails if dif is moved from its normal position. To analyse this positional requirement, dif was transplaced to a variety of positions, and deletions and inversions of portions of the dif region were constructed. Resolution occurs only when dif is located at the convergence of multiple, oppositely polarized DNA sequence elements, inferred to lie in the terminus region. These polar elements may position dif at the cell septum and be general features of chromosome organization with a role in nucleoid dynamics.


The circular Escherichia coli chromosome is replicated bidirectionally from one unique origin to a termination region half a chromosome away. The two processively replicated halves of the chromosome are termed arms or replichores (Blattner et al., 1997). First noticed at the gene level (Brewer, 1988), the symmetry of replication coincides with a bias in the orientation of genes and base sequence motifs (Lobry, 1996). In a recent study, Salzberg et al. (1998) reported that octamers belonging to several families are preferentially oriented in each replichore; that is, they show an orientation bias (or skew) that follows the direction of their replication. The orientation skew for most of these octamers inverts near the origin of replication and at a diametrically opposite position in the chromosome, the dif site, where dimer resolution occurs. These abundant skewed short sequences generate a polarization of each replichore.

Except perhaps for χ sites (Kuzminov, 1995), the functional significance of replichore polarization is unknown. Replication itself is not strongly sensitive to sequence orientation, as cells in which replication starts from the natural terminus are viable (Louarn et al., 1982; de Massy et al., 1987). The poor viability of cells with chromosomes carrying certain inversions had suggested the involvement of polarization in nucleoid organization (Rebollo et al., 1988). Cytological studies have consistently revealed a precise choreography of replication origin and terminus movements during the cell cycle (Webb et al., 1997; Niki and Hiraga, 1998), suggesting a nucleoid architecture facilitating replication and post-replicative remodelling. The present article supports this view. We report evidence that one final event of the cell cycle, the resolution of chromosome dimers, requires that the dif resolution site be located between regions displaying opposite polarity.

Site-specific recombination at dif resolves chromosome dimers arising from sister chromatid exchanges (Blakely et al., 1991; Kuempel et al., 1991). The XerC and XerD recombinases act co-ordinately to catalyse exchanges between copies of the 28 bp dif site (Blakely et al., 1997). Dimer resolution at dif takes place shortly before division, and its requirement for FtsK, a septum-associated protein, suggests that septum formation may control dif activity (Steiner and Kuempel, 1998a; Steiner et al., 1999). The frequency of dif usage for dimer resolution has been measured by Steiner and Kuempel (1998a, b). They concluded that, in growing wild-type bacteria, homologous recombination results in the formation of a chromosome dimer in about 15% of the cells. These dimers are resolved by XerCD/dif recombination, permitting normal cell division. Bacteria deficient in XerCD/dif recombination display poor viability when proficient for homologous recombination. Only 70% of such bacteria divide normally (Cornet et al., 1996), corresponding to the death of 15% of the cells at division. This frequency is similar to that of exchanges at dif in wild-type cells, suggesting that unresolved dimers at division lead to death. The phenotype of xerC/xerD/dif mutants is consistent with this view: low viability, apparent increase in doubling time and strong cell size heterogeneity with a majority of normal cells mixed with filaments and chains. The dif region is highly prone to homologous recombination (terminal recombination; Louarn et al., 1991, 19941994; Corre et al., 1997) when dif activity is inhibited (Corre et al. 2000), probably because failure to resolve dimers leads to frequent chromosome breakage in the terminus region.

A remarkable property of resolution by the XerCD/dif system is its regional specificity: dimer resolution is efficient only when dif is located within a 30 kb of its normal position, the dif activity zone or DAZ (Leslie and Sherratt, 1995; Tecklenburg et al., 1995; Cornet et al., 1996; Kuempel et al., 1996). Although replication terminates normally near dif (Louarn et al., 1994), compelling evidence indicates that dif resolution activity is not controlled by either termination of replication or timing of dif replication (Cornet et al., 1996; Kuempel et al., 1996). On the other hand, strains lacking dif and the entire DAZ region recover normal dimer resolution ability when dif is reinserted at the deletion junction point (Tecklenburg et al., 1995; Cornet et al., 1996; Kuempel et al., 1996). It has been proposed that the ability to resolve the dimer chromosome at dif is dictated by short, interchangeable elements in the immediate neighbourhood (Cornet et al., 1996; Kuempel et al., 1996).

The fragility of the chromosome when resolution is prevented, the existence of DAZ and the co-ordination of dimer resolution with cell division, taken together, suggest that cells with a dimer must position the nucleoid in a specific way at the end of the cell cycle and before the completion of septation. In this report, we examine the effect of dif transplacement and dif region rearrangement on resolution ability. The data reveal that dif activity requires the proper orientation of the immediately flanking sequences. This provides clear evidence that polarized elements of the chromosome may control the dynamics of nucleoid movements at division.


Measuring dif activity

We reported previously that transplacement of the dif site can impair dif activity (Cornet et al., 1996). Two assays were developed to quantify this position effect. The first assay measures chromosome dimer resolution indirectly, by the growth defect of cells with a dif activity defect. Growth rates of strains were compared by long-term co-culture (Experimental procedures). In this assay, the relative frequency of a dif deletion mutant (TcR; tetracycline resistant) dropped about 105-fold compared with a dif+ control strain over the course of 80 generations (Fig. 2A). This drop is consistent with 14% of divisions giving no viable progeny (Experimental procedures), a frequency in agreement with the direct measurements of dif site resolution made by Steiner and Kuempel (1998a, b). Importantly, this growth difference between difΔ and dif+ strains was eliminated by a recA mutation (Fig. 2A), as expected if the difference results from the ability to resolve dimers generated by homologous recombination. This assay was used to measure the efficiency of dimer resolution in strains with a variety of rearrangements of the dif region. All experiments included control competitions between difΔ and dif+ strains with and without a recA mutation.

Figure 2.

dif activity depends on location.

A. Quantification of the effects of dif transplacement using the co-culture assay. Strains carrying transplaced dif sites (Tcs) were grown in serial co-cultures with a Δdif2600 strain (Tcr) for 80 generations (Experimental procedures). The ratio of Tcs to Tcr bacteria (measured by plating every 20 generations) is plotted as a function of the number of generations. Examples shown are for strains LN3080 (Δdif2600::TcΩdif; black circles); LN3628 (Δdif2600::Tc zdc330Ωdif; black diamond) and LN3577 (Δdif2600::Tc zda192Ωdif; black square) co-cultivated with strain LN3079 (Δdif2600::Tc); and strain FC232 (recA56; open circles) co-cultivated with strain FC233 (Δdif2600::Ap recA56). In the latter case, the Aps/Apr ratio was followed in place of the Tcs/Tcr ratio.

B. Effects of dif transplacement on its activity. The activity of dif is plotted as function of its position of insertion (with respect to the natural position of dif, x-axis): from left to right, zda192, zdc310, zdc330, zdc338, zdd355, zdd360, zdd365, zdd370, zde406 and lacZ. Closed circles represent the Tcs/Tcr ratios after 80 generations (left scale) obtained using the co-culture assay for strains carrying the transplaced dif sites. Correspondence with the frequency of abortive divisions is indicated on the left (Experimental procedures). Open squares represent loss frequency per cell per generation (right scale) of the dif-Kn-dif cassette inserted at the same positions as above (Experimental procedures).

The second assay measures the frequency of recombination between two dif sites flanking, in direct order, a kanamycin resistance (KnR) gene (the dif-Kn-dif cassette, Fig. 1B). It was reasoned that action of the resolution system (XerC, XerD, FtsK) on the two tandem dif sites would generate progeny that had lost the KnR phenotype. This dif-Kn-dif cassette was inserted into the chromosome of a xerC mutant in which the KnR phenotype was stable. The XerC function was restored by the addition of a plasmid carrying the xerC+ gene, and resolution activity at dif was measured as the appearance of KnS bacteria (Experimental procedures). The structures of the substrate and products were confirmed by Southern blot analysis (data not shown). The frequency of loss of the KnR determinant was about 12% per generation when the cassette was inserted at the normal dif position. A detailed study using the dif-Kn-dif cassette demonstrates that the two assay systems give concurrent results (K. Pérals et al., in preparation).

Figure 1.

Strategy for analysis of regional requirements for the resolution activity of difA.

A. Relevant map of the dif region. Positions of insertions of the Tc fragments (Blattner et al., 1997) used in this study are indicated: zda192: 1438.8 kb; zdc310: 1553.5 kb; zdc330: 1573.6 kb; zdc338: 1581.8 kb; zdc346: 1588 kb; zdd347: 1589.7 kb; zdd355: 1597.6 kb; zdd360: 1602.1 kb; zdd365: 1606.6 kb; zdd370: 1612.0 kb; zde406: 1648 kb. The scale is relative to the dif position (position 0). The black and white square represents the 28 bp dif site (1588.8 kb).

B. Structure of a strain harbouring a transplaced dif site or an insertion of the dif-Kn-dif cassette (Experimental procedures). The dif fragment or the dif-Kn-dif cassette was inserted into chromosomal Tc fragments carried by strains harbouring a Δdif2600 deletion [deletion of the 2507 bp EcoRV (or EV) fragment]. Strains carrying insertions of the dif fragment are Δdif2600::Ap plus a TcΩdif insertion at the indicated position (with the exception of Δdif2600::Tc&~OHgr;dif for dif reinsertion at its natural position). Strains carrying insertions of the dif-Kn-dif cassette are xerC::Cm or Δdif2600::Ap plus TcΩdif-Kn-dif at the indicated position (except for Δdif2600::TcΩdif-Kn-dif). Examples are drawn for the zdd355 position. For simplification, these strains are designated Δdif2600::Tc zdd355Ωdif and Δdif2600::Tc zdd355Ωdif-Kn-dif in the following.

C and D. Generation of chromosomal rearrangements using the testek system. The tes and tek elements are two modified Tc fragments interrupted at two different positions by an ΩSp/Sm interposon (tes) or by a kanamycin determinant (tek) (François et al., 1987; Experimental procedures). Recombination between a tes and a tek element inserted on the chromosome allows the reconstruction of a selectable Tc fragment (Tcr). This generates either a deletion (C) or an inversion (D) of the intervening region, depending on the relative orientation of the tes and tek elements.

No sharp boundaries to the dif activity zone (DAZ)

By both the growth competition assay and the dif-Kn-dif cassette removal assay, the resolution activity dropped as the dif sequence was moved from its normal position. For the growth competition assay, we moved dif from positions ranging from −155 to +59 kb away from its normal position (Fig. 1A). All strains carried a dif deletion, Δdif2600::Ap, and a variably inserted 4.5 kb DNA fragment that included a dif sequence (Fig. 1B; Experimental procedures). The strains showed various Dif phenotypes (cell filamentation and growth defects), suggesting that some had a resolution ability intermediate between those of difΔ and dif+ strains (Cornet et al., 1996; data not shown). Each strain (TcS) was grown in competition with a dif deletion mutant (Δdif2600::Tc), and the ratio of TcS to TcR colonies was determined over 80 generations. Examples of this competition for dif inserted at the zdc330 and zda192 positions are shown in Fig. 2A. All results of the growth competition are graphed in Fig. 2B (black lines). It is apparent that strains with dif near the normal position have the greatest growth advantage over the reference dif deletion strain. Expressed in terms of inferred abortive divisions, the strains tested range from 0% (with dif at its normal position) to 13% abortive divisions (with dif furthest away).

The same results were obtained using the Kn cassette segregation assay. Here, the rate of cassette loss (recombination between dif sites) occurring per cell division was greatest when the cassette was located near the normal dif position and decreased as the cassette was moved in either direction. These results are plotted in Fig. 2B (dotted lines).

As is apparent in Fig. 2B, dif activity decreased progressively as dif was moved in either direction from its normal position; this was apparent for both the growth defect and the cassette segregation assays. Only dif sites inserted at the normal dif position or at the nearby position zdc338 displayed full activity. The resolution defect increases over about 20 kb to the right-hand side of the natural position of dif and over about the same distance to the left of position zdc338. The gradual decrease in dif activity on either side of the zdc338-dif segment indicates that the DAZ is not limited by discrete boundaries with an all-or-nothing effect on dif activity.

The DAZ lies at the junction between the left and right replichores in the terminus

It has been shown that the DAZ and its surrounding sequences contain no unique element required for dif activity (Tecklenburg et al., 1995; Cornet et al., 1996). The role of the left and right dif-flanking regions in the formation of the DAZ was examined further using a set of deletions in the dif region. Deletions were generated by homologous recombination between pairs of insertions of the Tc fragment (Fig. 1A) using the tes and tek system (Fig. 1C; François et al., 1987). Deletions between two insertions on the same side of dif had no effect on dif activity (data not shown). This remained true even for the deletion of segments ending inside the DAZ (e.g. deletion of the zdd347zde406 segment). On the other hand, when the deleted region contained dif, we found no exception to the rule that dif was fully active when reinserted into the Tc fragment located at the deletion join point. Two examples are shown in Fig. 3. The first deletion eliminates 155 kb to the left of dif and 0.9 kb to its right; the second deletion eliminates 0.8 kb to its left and 59 kb to its right. Strains carrying these deletions both regain a Dif+ phenotype after reinsertion of dif (Fig. 3). These data confirm that deletions in the dif region consistently lead to the formation of a de novo DAZ and suggest that the new DAZ always forms at the junction between sequences located to the left and to the right of the natural dif position (hereafter referred as the left and right terminus arms).

Figure 3.

Formation of a de novo DAZ at the junction between the terminus arms. The top line is a map of the dif region indicating the positions of dif (the black and white square) and the Tc insertions used. Deletions were obtained using the tes–tek system and are represented by open bars. The dif fragment was then inserted into the Tc fragments tagging the end-point of the deletions, and the resulting strains (Tcs) were grown in co-culture assays with their direct ancestors (with no dif fragment inserted, Tcr). The Tcs/Tcr ratio obtained after 80 generations of co-culture is given. The ratio obtained with a co-culture of LN3080 (Δdif2600::TcΩdif, taken as wild-type) and LN3079 (Δdif2600::Tc) is given for comparison.

Chromosome dimer resolution is prevented by inversions of segments of the dif region

The above data suggest a specific role for the left and right terminus arms in the formation of the DAZ. The two terminus arms might differ by specific elements or by opposite orientation of equivalent elements implicated in the formation of the DAZ. This led us to analyse the consequences for dif activity of inversions in regions surrounding the site. Inversions were generated by homologous recombination between pairs of insertions of the Tc fragment using the tes and tek system (Fig. 1D). The presence of the inversion was confirmed by Southern blot analysis for all inversions described (data not shown).

Inversions of some segments adjacent to dif caused a Dif phenotype (filament formation and increased generation time; Fig. 4; Table 1). Observations of DAPI-stained cells revealed frequent nucleoid alterations similar to those observed for mutants of the XerCD/dif system (Fig. 4B; data not shown). Consistent with this observation, the deleterious effect of inversions was not aggravated by mutations in the Xer system (Table 1). Thus, these inversions appeared to cause a serious defect in dimer resolution. This indicates that dif activity depends on the orientation of nearby DNA segments. These segments presumably contain oriented sequence elements that inhibit dif activity when inverted. Inversions provoking a filamentous cell phenotype are found on either side of dif (Fig. 4A). We infer that both terminus arms contain polar elements. More than one element is present on each terminus arm, as inversions of non-overlapping segments were deleterious (the zdc310–zdc330 and zdc330–zdc338 segments of the left terminus arm and the zdd347–zdd355 and zdd355–zdd370 segments of the right terminus arm). Although dif activity cannot be measured precisely owing to the relative instability of the inversion, the deleterious effect of the different inversions on dimer resolution appeared variable. We suspect that this depends in part on the sequence content of the inverted segment and in part on its location with respect to dif.

Figure 4.

Inversions in the dif region may inactivate dif.

A. The top line represents a map of the dif region indicating the positions of the Tc insertions and their distance (in kb) from dif (drawn as a black and white square). The inverted segments (generated using the testek system; Experimental procedures) are represented by horizontal lines ending in vertical arrowheads and are grouped with respect to the associated phenotype. The upper class includes inversion strains with frequent filaments, aberrant nucleoid distribution and poor growth (two examples are shown in (B) and Table 1; the lower class includes inversion strains with only normal cells and normal growth rate.

B. Micrographs of bacteria carrying deleterious inversions and/or xerC mutations. Strains were stained with DAPI and observed under combined fluorescence and phase-contrast microscopy (magnification × 2450; Experimental procedures). (a) LN3888; (b) LN4033; (c) LN3893; (d) LN4034; (e) LN3950; (f) LN4035; (g) LN3954; (h) LN4036. The relevant genotype of these strains is given in Table 1.

Table 1. Effect of inversion of two segments surrounding (but not including) dif on growth rate in xerC+ and xerC conditions.
StrainsaRelevant genotypesDoubling times (min)bFilamentsc
  1. a . Bacteria representative of these strains are shown in Fig. 4B.

  2. b . Fresh LB medium was inoculated with an aliquot of an overnight culture to OD 600 = 0.05. Bacteria were grown aerobically at 37°C, and OD600 was followed. The doubling time was deduced from the growth curve.

  3. c . Presence (+) or absence (–) of filaments in these strains (from Fig. 4B).

LN3888 zda192::tek zdc330::tes25
LN4033LN3888 xerC::Cm29.5+
LN3893LN3888 INV(zda192–zdc330)29+
LN4034LN3893 xerC::Cm29+
LN3950 zdd347::tes zde406::tek24
LN4035LN3950 xerC::Cm30.5+
LN3954LN3950 INV(zdd347–zde406)30+
LN4036LN3954 xerC::Cm30.5+

Other inversions in the terminus do not alter dimer resolution

Three classes of inversion in the dif region were harmless. The first class includes all inversions of segments that include dif (Fig. 4A). In strains carrying these inversions, the orientation of the inferred polar sequences with respect to dif is conserved. Thus, terminus arms can be exchanged without affecting dif activity. This suggests that the polar elements are not specific for the terminus arms but are interchangeable. The second class includes inversions of the zda192zdc310 and the zdd370zde406 segments. Strains carrying these inversions display normal generation times, although rare filaments appeared. For both these segments, the nearest end-point was more than 25 kb away from dif (Fig. 4A). These segments may contain no polar element. Alternatively, polar elements contained in these segments may be too far away from dif to have their inversion cause an inhibitory effect on DAZ. Lastly, inversion of the zdd338zdd346 segment had no effect. This small (6 kb) segment is directly adjacent to dif on its left (Fig. 4A). We suspect that no polar element is present in this segment. Consistent with this possibility, zdc338 is the only position at which an inserted dif is as active as at its natural position (Fig. 2B).

Predicting the formation of a new DAZ after inversion

A model illustrating the role of polar elements in formation of the DAZ is shown Fig. 5. This model predicts that inversion of a segment adjacent to dif might move the DAZ from one end of the segment to the other (Fig. 5, B4). Consistently, inversion of the segment between zda192, 150 kb left of dif, and zdd347, only 0.9 kb right of dif, was harmless for dif activity (Fig. 4A). The prediction was demonstrated further by showing that, after inversion of nearly the same segment zda192–Δdif2600, a dif site inserted at its right end was inactive, whereas a dif site inserted at the left end was active. This experiment is presented in detail in Fig. 6. This result demonstrates unambiguously that formation of the DAZ depends on the orientation of the regions flanking dif.

Figure 5.

A model for formation of the DAZ.

A. In a wild-type strain, each terminus arm flanking dif (the black and white square) harbours a series of identically oriented polar elements (the white and black arrows for the left and the right terminus arms respectively). The DAZ is the region where these series converge.

B. Effects of chromosome rearrangements on the DAZ. The end-points of the various rearrangements are indicated by vertical arrows, and deleted segments are indicated by dashed lines.

B1. Deletions in one of the terminus arms eliminate polar elements, but dif still lies in the convergence region of the remaining elements. Strains harbouring such deletions are Dif+. This is the case for all deletions on the same side of dif, such as the zdc310zdc330 segments (data not shown) or deletions from Cornet et al. (1996).

B2. In deletions removing dif, the end-points of the deletion lie systematically at the convergence region. Reinsertion of dif at the end-points of the deletion yields Dif+ strains. Examples come from Cornet et al. (1996) and from deletions reported in Fig. 3.

B3. Inversions including dif reshuffle the left and right terminus arms and their polar elements, but dif still lies at the convergence region. Strains harbouring these inversions are Dif+. For example, see Fig. 4A, the inversion of the zda192zdd347 segment.

B4. Inversion of segments entirely contained in one terminus arm (i.e. not containing dif) displace the convergence region to the end-point of the inversion furthest from dif. Thus, dif no longer lies in the competence region. Strains harbouring such inversions are Dif but regain a Dif+ phenotype when dif is transplaced to the new convergence region. Inversion of the zda192zdc338 segment is an example (Fig. 4A).

Figure 6.

Formation of a de novo DAZ after inversion. The top line represents the chromosome region left of dif with positions of the inversion end-points (Δdif2600 and zda192) indicated. The large arrow symbolizes the orientation of the intervening segment. Both end-points were primitively labelled by Tc insertions in either orientation. The tesdif element was inserted into the Δdif2600 end-point in either orientation (directed by the orientation of the Tc insertion), and the tek element was inserted at position zda192 in orientation opposite to the former (parental strains a and c; Experimental procedures). Similarly to the protocol described in Fig. 1D, inversions of the zda192–Δdif2600 segment were obtained by plating on tetracycline-containing medium and occurred by recombination between the internal parts of the tetA gene (the shaded area) carried by the tesdif and tek elements. Depending on the orientation of the elements, this yielded two strains differing only by the position of dif with respect to the inverted segment (strains b and d). The results of microscopic observations are reported on the right.


The dif activity zone (DAZ) is so far the only functionally defined domain of the E. coli chromosome (Cornet et al., 1996; Kuempel et al., 1996). The analysis of this domain has been challenging, as it is not defined by single border elements. The present study shows that formation of the DAZ is dictated by the orientation of multiple polar elements present in its flanking regions.

The activity of transplaced dif sites was assessed by two different assays, one measuring the frequency of recombination between tandem dif sites and the other measuring the growth defect associated with transplacement. The results obtained using these two assays were strongly correlated and indicated that inhibition of recombination at dif was the cause of the viability defect displayed by strains carrying transplaced dif sites. Levels of activity of transplaced dif sites depend on their location with respect to the natural position of dif (Fig. 2B). These data confirm the existence of a dif activity zone (DAZ; Cornet et al., 1996; Kuempel et al., 1996) and define its extent. The positions in which dif displays full activity include zdc338 and the natural position of dif but do not include the flanking zdc330 or the zdd355 positions. Thus, the size of the optimal DAZ is between 7 and 25 kb. If one assumes that dif lies at the centre of the DAZ, the DAZ could be 16 kb long. For about 30 kb on either side of the DAZ, the activity of dif decreases progressively to reach a basal level. Thus, the DAZ is not limited by single elements having an all-or-nothing effect on dif activity.

No deletion adjacent to dif or including dif has removed DAZ. Chromosomes with deletions of more than 150 kb around dif (Cornet et al., 1996; Kuempel et al., 1996) or 59 kb on one side of dif and 155 kb on the other side (Fig. 3) still possess a DAZ. Each deletion resulted in the formation of a new DAZ at the position that juxtaposed sequences from the left and right terminus arms (the DNA sequences from the left and the right of dif on a linear map of the terminus region). This suggests that a variety of sequences from each terminus arm can combine to dictate formation of the DAZ. The crucial information came from inversion studies. Inversions of segments adjacent to the normal dif position prevented its activity. These segments are entirely contained in one of the terminus arms and do not include dif, suggesting that each terminus arm contains polar elements that generate the DAZ. This suggestion is strongly supported by the finding that the orientation of the dif-flanking regions dictates the positioning of the DAZ (Fig. 6). Each terminus arm must contain more than one polar element, as each contains at least two non-overlapping segments whose inversion inactivates dif. Inversion of just the central region of the DAZ (the zdc338zdc346 segment adjacent to dif) has no effect on dif, suggesting that the DAZ itself is devoid of polar elements. Inversions that include dif are also harmless, indicating that sequences from both terminus arms may be exchanged without damage to dif activity. Thus, the active elements are not specific for either terminus arm but, rather, are interchangeable, provided that their orientation with respect to dif is conserved. These observations are summarized in Fig. 5 as a model for formation of the DAZ by oppositely oriented arrays of polar sequence elements. Consistent with this model, we have shown recently that a λ prophage inserted on one side of dif inhibits dif in one orientation but not in the other. When inserted on the other side of dif, the harmless and deleterious orientations of the prophage are inverted (Corre et al. 2000). This observation strongly supports the hypothesis that the regions flanking dif must display opposite polarity if dif is to function normally.

The 60 kb fragment between the zdc310 and zdd370 positions contains the information necessary for formation of the DAZ, as inversion of the zda192zdc310 and zdd370zde406 segments, relatively remote from dif, have virtually no effect on dif activity (Fig. 4A). Thus, about 25 kb of polarized material on each side of the DAZ may be sufficient to dictate its formation. However, the terminus arms may harbour polar elements over a much greater distance than that needed to generate the natural DAZ, as deletions of sequences between 150 kb to the left of dif and 56 kb to its right still allow formation of DAZ at the deletion join points. The orientation-dependent inhibition of dif activity by a λ prophage (Corre et al. 2000) suggests that a polarization related to that involved in the formation of the DAZ may be found in other regions.

What could be the molecular nature of the polar elements? No extensive homology was detected by alignment of DNA segments thought to contain the inferred polar elements. Polarization may thus be caused by a number of short and/or degenerate sequences. Several sequence motifs have been identified with a skewed distribution in each replichore (Salzberg et al., 1998; J. Lobry, personal communication). For instance, the RRRAGGGY motif (R = purine; Y = pyrimidine) is distributed with an average skew of about 70% in favour of the leading strand of each replichore. Interestingly, we noticed that this skew reaches 90% within 100 kb on either side of dif, and the polarity switch point coincides with dif. The same sequences are polarized in λ DNA sequence in a way that could explain its ability to inhibit dif activity (Corre et al. 2000). These distributions are consistent with a model in which the functional polarity of the terminus arms is determined by specific sequences. Experiments are under way to identify the polar elements.

The polar elements inferred here may participate in precise positioning of the dif region vis-à-vis the division septum. The fact that dif activity depends on formation of the division septum and on the FtsK protein suggest that dif sites must be localized near the septum to allow recombination (Steiner et al., 1999). Consistent with this proposal, the terminus region localizes near the division plane at the time of cell division (Webb et al., 1997; Niki and Hiraga, 1998; Niki et al., 2000). Thus, dimer resolution appears to be achieved by an integrated process. The dif sites, the cell division apparatus and the recombinases must co-localize within a small volume for dimer resolution to occur, as depicted in Fig. 7. We propose that the polar elements aid in chromosome positioning subsequent to replication. The terminus region (maybe the entire chromosome) could be submitted to a dragging process, which ensures that no DNA is trapped by the division septum. This movement could be unidirectional along a fraction of each replichore and directed by interactions of a molecular motor anchored at the polar elements. When acting on a chromosome dimer, the process would extend two DNA tracts of the segregating nucleoids and place dif at the septum plane.

Figure 7.

Model for the control of chromosome dimer resolution. The drawing figures the central part of a dividing cell, the growing division septum and the dif surrounding regions of a chromosome dimer with their polar elements (arrows) required for formation of the DAZ. The dif site is represented as a black and white square bound by the XerCD recombinases (the open circles). The FtsK protein is represented inserted into the membrane of the division septum, with its C-terminal domain (the black circle) pointing towards dif (Wang and Lutkenhaus, 1998; Yu et al., 1998). The grey area represents the small volume in the division septum vicinity where we suspect that both dif sites must lie for interaction with FtsK allowing XerCD/dif recombination to occur (at the time of cell division or shortly before; Steiner et al., 1998a). The role of the functional polarization disclosed in this study may be to ensure that the two dif sites are correctly positioned at this time, at least in cells harbouring a chromosome dimer.

A. In a wild-type cell, the different partners required for resolution of the dimer co-localize at the division septum, and resolution can occur.

B. A chromosomal inversion near dif has reversed the orientation of some polar elements and displaced the convergence region (formation of a de novo DAZ, case B4 in Fig. 5). The region localized at the division septum no longer contains dif, and chromosome dimer resolution cannot occur.

Experimental procedures

Bacterial strains

All bacterial strains are derived from strain CB0129 (F W1485 leu thyA deoB or C supE;Bird et al., 1972), strain LN2666, a streptomycin-resistant spontaneous mutant of CB0129 (Cornet et al., 1994), or strain LN3061, which is LN2666 rendered Δdif2600::Ap (Cornet et al., 1996).

Media and general methods

Bacteria were grown in LB (Luria–Bertani) medium (Miller, 1992). Antibiotic concentrations used were: ampicillin (Ap): 25 µg ml−1; chloramphenicol (Cm): 20 µg ml−1; kanamycin (Kn): 25 µg ml−1; spectinomycin (Sp): 40 µg ml−1; streptomycin (Sm): 200 µg ml−1 (rpsL) or 15 µg ml−1 (Sp/Sm); tetracycline (Tc): 15 µg ml−1. DNA purification and manipulation, cloning, Southern blot hybridization and P1 transduction were performed according to standard procedures (Sambrook et al., 1989; Miller, 1992). For microscopic observation, bacteria were grown to OD600 = 0.5, stained with DAPI (4,6-diamino-2-phenylindole; 2 µg ml−1) for 30 min at 37°C, concentrated 50 times by centrifugation, mounted on microscope slides and observed using a Leica DMRB microscope. Images were obtained using a Photonic Science CCD camera and the visiolab 1000 software.

Chromosomal Tc insertions

The Tc fragment is the 2775 bp BglII fragment from Tn10, which confers resistance to tetracycline (Tcr). To obtain chromosomal insertions of the Tc fragment, chromosome fragments were first cloned on the integration-excision vector pLN135 (Cornet et al., 1996). The Tc fragment was then inserted into the chromosome fragment, and the resulting plasmid was used to instal it on the chromosome. The detailed procedure and the construction of most of the Tc insertions used have been described previously (Cornet et al., 1996; 1997). Additional insertion of the Tc fragment, at positions zdd360[1602.1 kb on the Blattner et al. (1997) sequence] and zdc346 (1588 kb), were constructed for this study. Insertions were moved between strains by P1 transduction.

Transplacement of dif

To obtain strains carrying transplaced dif sites, the dif fragment, which consists of the 1715 bp EcoRV chromosomal fragment containing dif, was inserted into the chromosomal Tc fragments (2775 bp; total insertion 4.49 kb) carried by strain LN3079 (Δdif2600::Tc) or by derivatives of strain LN3061 (Δdif2600::Ap) harbouring Tc insertions at the indicated positions. This was achieved using the pFC72 plasmid as described previously (Cornet et al., 1996). The resulting strains are sensitive to tetracycline.

The dif-Kn-dif cassette

Construction and use of the dif-Kn-dif cassette will be described elsewhere (K. Pérals et al., in preparation). Plasmid pKP16, which carries a TcΩdif-Kn-dif segment, was used to insert the dif-Kn-dif cassette into the Tc insertions using the general procedure for use of this family of integration-excision vectors (Cornet et al., 1994; 1996). The dif-Kn-dif cassette was installed in xerC::Cm (xerC::MudIIPR13 Y17 allele; Colloms et al., 1990) derivatives of the strains used for dif transplacement.

The tes, tek and tes′dif elements

The tes and tek elements are derivatives of the Tc fragment in which the tetA gene is interrupted at different positions by a ΩSp/Sm interposon encoding resistance to streptomycin and spectinomycin (Fellay et al., 1987) and by a fragment from transposon Tn5 encoding resistance to kanamycin respectively. Substitution of the chromosomal Tc fragments for tes or tek elements was achieved using bacteriophage λTSK as described previously (François et al., 1987). The tesdif element was obtained by cloning of the ΩSp/Sm interposon into one of the EcoRV sites flanking the dif fragment of plasmid pFC72 (Cornet et al., 1996). The resulting plasmid, pYM2, was used to substitute the Δdif2600::Tc deletion for a Δdif2600::tesdif deletion (Fig. 6) using the general procedure described for this family of integration-excision vectors (Cornet et al., 1994, 1996).

Co-culture assay and quantification of cell viability

Strains carrying transplaced dif sites (Tcs) were grown in co-culture with strain LN3079 (Δdif2600::Tc) (Tcr). A 1:1 mixture of the two strains was grown in serial cultures in LB medium at 37°C for up to 80 generations. Every 20 generations, the relative numbers of Tcr and Tcs bacteria were determined by plating (Fig. 2A). The percentage of abortive divisions was calculated as follows. If k is the probability of generating a non-viable cell at division and τ is the generation time of viable cells, the number (N) of viable cells in a culture growing exponentially for time t increases according to N = N0 × [2(1–k)]t. Letting R0 and Rg be the initial and final Tcr/Tcs ratios after g generations, k1 and k2 be the probabilities for Tcrdif) and Tcs (dif-transplaced) strains, respectively, and assuming that the generation time of viable cells, τ, is the same for a pair of isogenic strains, then Rg/R0 = [2(1–k1)]g/[2(1–k2)]g. It follows that k2 = 1–[(1–k1) × 1/(Rg/R0)1/g]. The value of k was taken as 0 for the wild-type strain.

Measurement of the frequency of resolution of the dif-Kn-dif cassette

Strains carrying chromosomal insertions of the dif-Kn-dif cassette were transformed with the XerC expression vector pFC225, which carries a Sp/Sm resistance cassette and a lacZp–xerC fusion (K. Pérals et al., in preparation). After phenotypic expression, a sample was plated on spectinomycin (Sp)-containing plates to measure the transformation efficiency, and 200 µl of the same cells (containing 1000–2000 transformants) was used to inoculate 5 ml of LB medium containing 40 µg ml−1 spectinomycin. Bacteria were grown for 16 h to stationary phase, which corresponds to 20–22 generations, and suitable dilutions were plated on Sp-containing medium. The percentage of Kn-resistant bacteria was measured by replica plating, and the frequency of loss of the Knr determinant per cell per generation was deduced from the ratio of Knr to total bacteria and from the number of generations.

Construction of chromosomal rearrangements

Deletions and inversions were obtained by in vivo recombination between tek and tes elements (Fig. 1C and D; François et al., 1987). The same procedure was followed for inversions between the tek and tesdif elements (Fig. 6). Briefly, a derivative of strain CB0129 carrying a tes and a tek insertion in the suitable relative orientation for the desired rearrangement was constructed by P1 transduction. Cells containing the rearrangement were selected by plating on tetracycline-containing medium. Deletions were found in Tcr, Kns and Sp/Sms clones, and were ultimately transferred to strain LN2666 by P1 transduction (e.g. for reinsertion of the dif fragment using plasmid pFC72). Inversions were found among Tcr, Knr and Sp/Smr clones. A sample (typically four) of these clones was analysed by Southern blotting; most of them carried the selected inversion.


We are grateful to Agammenon Carpoussis, David Lane and John Roth for critical reading of the manuscript. We thank J. Corre-Louarn and J. Patte for constant co-operation. This work was supported by the Association pour la Recherche sur le Cancer (ARC contract no. 9823). K.P. and F.C. were funded by ARC during part of this work.