Double-strand (ds) breakage of chromosomal DNA is obviously a serious threat to cells because various activities of the chromosome – gene expression, replication and partition – depend on its integrity. However, recent experiments suggest that such chromosomal ds breakage may occur relatively frequently during "normal" growth in several organisms – in bacteria [1,2], yeast [3] and chicken cells [4].

In Escherichia coli, spontaneous breakage and degradation of the chromosome associated with a replication fork were predicted from early genetic analysis and were detected under various conditions of altered replication (for review, see [5]). DNA ds breaks play a key role in homologous recombination. From a DNA ds break, RecBCD enzyme starts degrading DNA (for review, see [6]). When it encounters a specific sequence called Chi, it promotes its pairing with a homologous DNA. Even in the absence of RecBCD enzyme, sbcA mutation confers other recombination pathway, called RecET pathway. The recE gene product of the Rac prophage converts dsDNA ends into 3' protruding single-stranded form and the recT gene product promotes recombination by annealing them with a homologous DNA in its vicinity (for review, see [7,8]). This recombination may result in one progeny DNA (non-conservative recombination) or two progeny DNAs (conservative double-strand break repair) [9]. In a recBC sbcBC background, a ds end stimulates homologous recombination that results in only one progeny DNA (non-conservative recombination) [10]. Analysis of the stimulation of recombination by replication (for review, see [11]) and analysis of altered chromosomal replication (for review, see [12]) led to the proposal that a chromosomal ds break formed during replication fork arrest triggers homologous recombination, which would reconstitute a replication fork (for review, see [5]).

Game and his colleagues have developed a sensitive means of detecting chromosomal ds breakage using a circular chromosome [3]. Under most conditions of pulsed-field gel electrophoresis, a circular yeast chromosome and circular bacterial chromosomes will not enter the gel, very likely because they are trapped by the branches of the network of agarose [3,13,14]. One double-strand break transforms this circular form into a linear form, which can now move slowly in the gel [3]. We used this procedure to detect double-strand breakage of a circular bacterial chromosome occurring spontaneously or after loss of a restriction-modification gene complex [15,16]. We found increased chromosome breakage in recBC-null and recC1002 mutants of E. coli under both conditions [1]. Michel and her colleagues used pulsed-field gel electrophoresis to detect degraded chromosomal DNAs arising spontaneously in recBC mutants and arising during replication fork arrest [2]. RuvABC proteins, which catalyze migration and cleavage of Holliday junctions, are responsible for the occurrence of the degraded DNAs following replication fork arrests [17].

In this work, we employed the pulsed-field gel electrophoresis procedure to measure large non-circular forms of the chromosome obtained from various recombination-defective mutants in rec⁺, recBC sbcA, and recBC sbcBC genetic backgrounds.

We found that large, non-circular forms of the chromosome accumulate in varying amounts in various recombination-defective mutants of Escherichia coli.

Our operational definition of the non-circular forms is their presence in an area just below the well in our pulsed-field gel, as marked by a bar in Figure 1. The molecular species in this area may not be limited to a simple linear form of varying lengths. If a chromosome carries multiple replication forks as usual at 37°C in rich media, more than one double-strand break may be necessary to form a non-circular, branched species, which should be able to move through the gel. Finding out macroscopic forms of these giant molecules would be a technical challenge (see [20], for example). We do not know why DNAs make two broad bands in this area (Figure 1, 4th lane, for example), either. Depending on the electrophoresis condition, one narrow band, a pair of two bands or one very broad band was observed (data not shown).

Abundance of these huge non-circular forms is expected to be affected by several factors, which might work potentially in opposite directions, such as: (i) breakage in the cell; (ii) degradation in the cell; (iii) repair in the cell; (iv) breakage and degradation out of the cell. Each term is, in turn, affected by other factors such as chromosome organization, number of the replication forks, speed of the replication forks, abundance of specific proteins, and so forth. Therefore, our finding of accumulation of more of the non-linear forms in a rich medium than in a poor medium (Figure 1) does not immediately allow us to conclude that starving conditions induce a chromosomal double-strand breakage.

Spontaneous DNA damages, repair and degradation are expected to be the key processes in interpreting our data. Spontaneous DNA damages may interfere with replication fork progression and produce chromosomal double-stranded breaks. This would lead to extensive exonucleolytic degradation. Complete repair at some of these steps would reconstitute a circular chromosome, which will stay in the well. On the other hand, further degradation of the huge, non-circular forms would result in shorter or no fragments, which will run faster in the gel. The presence of huge linear forms, therefore, probably indicates both the absence of complete repair and the absence of further degradation. The absence of the large linears could either mean the presence of complete repair or the presence of extensive degradation activity. Our control experiments demonstrated that restriction digestion of chromosome DNAs before the electrophoresis results in release of comparable amounts of DNA from the wells in all the strains examined (Figure 3). This result, at least, excludes the possibility that the absence of the large, non-circular chromosomes in some strains (Figure 2) reflects the absence of DNAs in the wells during the process or by extensive and general nuclease action. Of course, we cannot exclude the possibility that the broken chromosomes specifically have suffered extensive degradation.

In spite of these potential complexity and essential ambiguity, our measurements provided a unique clue to the action of recombination-associated enzymes in the chromosome metabolism. Indeed, some of our observations in the mutants can be readily related to the established properties of the affected enzyme.

Accumulation of the huge linear DNAs in the recBC null mutants can be interpreted from the known properties of RecBCD enzyme in a straightforward way. These null mutant enzymes cannot degrade DNA from a ds break nor can they repair DNA by recombination [6]. We assume that they cannot repair the broken chromosomes to form intact circular chromosomes and that they cannot degrade them into smaller pieces. The recD mutant does not show nuclease activity but is recombination-proficient and able to repair the broken DNA molecules [6]. This explains why it does not accumulate the huge linear forms. The other non-null recBCD mutants (recC1001, recC1002, recC1003, recC1004, recC2145, recB2154, and recB2155) are all nuclease positive [21,22]. They would be expected to degrade the huge linears. They retain some to nearly complete recombination proficiency [21,22], which may contribute to repair of the large linears into circles. The other recombination-defective mutants, in otherwise rec⁺ background, did not accumulate the huge linears probably because the DNA was degraded by active RecBCD enzyme or was not produced.

Partial suppression of the accumulation of the huge linears in a recBC null allele by a recF mutation (Figure 2B) leads to several possible explanations. For example, RecF-mediated homologous recombination may transform a circular chromosome, possibly with a spontaneous damage, into some type of non-circular forms. This is expected because RecF-mediated recombination is non-conservative in the sense that it generates only one progeny DNA molecule from two parental DNA molecules [10]. Alternatively, RecF function may somehow help generation of broken chromosomes or maintenance of break to load RecA [23].

In the recBC sbcA and the recBC sbcBC backgrounds, the absence of RecBCD nuclease may prevent faster degradation of the large non-circular DNAs. However, we see only little accumulation of the broken forms. One might expect that the accumulation of the huge linears may correlate with the capacity for recombination repair that reconstitutes a circular form. Indeed, the effects of recA, recJ and several other rec mutations on accumulation of the huge linear chromosomes in these two recBC backgrounds (Figures 2C and 2D) were similar to their negative effects on conjugational recombination [19] with interesting exceptions (see next paragraph). This accords with the concept that a huge linear fragments of the chromosome is involved in recombination following conjugation. However, any of the recombination mutants that lead to accumulation of linear DNA could affect the probability of breaks occurring in the first place.

The mutations in Holliday-structure-processing enzymes – RecG, RuvAB, and RuvC – did not result in accumulation of the huge linears even in the recBC-minus background. The complex intermediate forms accumulating in these mutants may be trapped in the agarose gel (see [24,25]). An alternative interpretation could be that these enzymes may be involved in generation of double-strand breaks as hypothesized by Seigneur et al. [17].

The accumulation by the recJ mutation in the recBC sbcA background is suppressed by a recQ mutation (Figure 2C). Kusano et al. [26] found that both sensitivity to DNA damaging agents and decreased association of crossing-over with double-strand break repair in a recBC sbcA recJ strain are suppressed by mutant recQ alleles. Such suppressing relationship was interpreted to suggest that RecQ acts prior to or concurrently with RecJ. Pulsed-field gel electrophoresis analysis of chromosomes after ultraviolet irradiation has revealed extensive chromosome degradation dependent on uvrA incision enzyme [27]. A report [28] showed that RecQ and RecJ proteins process nascent DNA at replication forks blocked by ultraviolet irradiation prior to the resumption of DNA synthesis (see also [29]).

The accumulation of the non-circular, broken chromosomes correlated with the growth rate or DNA damage response in most of the recBC-minus background [30]. The recB or recC null mutation showed low viability even in the absence of exogenous DNA damage [31,32]. A simple interpretation of these data is that RecA, RecFOR, and RecQJ functions (and RecET functions for the sbcA background) repair chromosome breakage and/or prevent generation of the breakage. The major contradiction observed here is the phenotype in ruv mutants. The ruv mutants in all the background did not show any accumulation of the broken chromosome. This may suggest that the possible role of Ruv protein is making a break into dsDNA [33].

Our sensitive measurements of the large non-circular forms of the chromosome – which should be able to detect one ds break out of 4 million bp – provided unique sets of data that would help in further elucidating the mechanisms of chromosome double-strand break repair. A simplest interpretation of our data is that RecBCD enzyme is involved in repair and degradation of broken chromosomes, and that RecA, RecFOR, RecQJ and RecET functions are involved in prevention and/or repair of the breakage. Interaction was observed between a recC mutation and a recF mutation and between a recQ mutation and a recJ mutation. ruvABC mutants and a recG mutant did not accumulate broken chromosomes. Further molecular analysis would bring about interpretation of the present data in detailed molecular terms.

NH carried out all of the experiment and drafted the manuscript. IK supervised the work and edited the manuscript. All authors read and approved the final manuscript.

ds, double-strand. E. coli, Escherichia coli. Amp, ampicillin. Cml, chloramphenicol. Kan, kanamycin. Tet, tetracycline.