Learn more: PMC Disclaimer | PMC Copyright Notice
Sister Chromatid Exchange Frequencies in Escherichia coli Analyzed by Recombination at the dif Resolvase Site
Abstract
Sister chromatid exchange (SCE) in Escherichia coli results in the formation of circular dimer chromosomes, which are converted back to monomers by a compensating exchange at the dif resolvase site. Recombination at dif is site specific and can be monitored by utilizing a density label assay that we recently described. To characterize factors affecting SCE frequency, we analyzed dimer resolution at the dif site in a variety of genetic backgrounds and conditions. Recombination at dif was increased by known hyperrecombinogenic mutations such as polA, dut, and uvrD. It was also increased by a fur mutation, which increased oxidative DNA damage. Recombination at dif was eliminated by a recA mutation, reflecting the role of RecA in SCE and virtually all homologous recombination in E. coli. Interestingly, recombination at dif was reduced to approximately half of the wild-type levels by single mutations in either recB or recF, and it was virtually eliminated when both mutations were present. This result demonstrates the importance of both RecBCD and RecF to chromosomal recombination events in wild-type cells.
Linear and circular chromosomes both present unique problems for replication. Due to the enzymatic limitations of DNA polymerases, for example, a problem is encountered when the ends of linear chromosomes are replicated. Eukaryotes have solved this problem by capping the ends of their chromosomes with specialized structures called telomeres, which can then be replicated by the telomerase enzyme (18). Because circular chromosomes do not have ends, no difficulty is encountered in replicating the entire molecule with conventional DNA polymerases. However, circular chromosomes are faced with a new problem when recombination occurs between the growing daughter chromosomes, a process we refer to as sister chromatid exchange (SCE). One exchange or any odd number of exchanges results in the formation of circular dimer chromosomes (24). Dimers must be resolved back to monomers prior to cell division in order to properly partition daughter chromosomes to daughter cells. This resolution process requires a compensating exchange, which occurs by site-specific recombination at the dif locus in the terminus of the Escherichia coli chromosome (39).
SCE is difficult to analyze in haploid organisms, such as Escherichia coli, due to the absence of genetically distinguishable homologous chromosomes. Previously, the only direct attempts to analyze SCE in E. coli used autoradiographic analyses to determine how often daughter cells each received parts of 3H-labeled chromosomes (14, 43). The difficulty and lack of sensitivity of those assays have limited their use, and no attempts have been made to extend these earlier studies in order to characterize the various mutations that might affect SCE. Efforts to identify factors contributing to SCE have generally involved genetic analyses of interactions between chromosomal regions present as direct or inverted repeats and which are either adjacent to or at some distance from each other (3, 9, 16, 29, 30, 31, 48, 49). The main problem with genetic assays of this kind is that recombination between chromosomal repeats can occur both intramolecularly as well as intermolecularly (true SCE), and it is not generally possible to distinguish which of these types of interactions predominates. Important concepts have certainly emerged from genetic analyses of SCE, but it is difficult to estimate whether it is valid to extrapolate these findings to events that occur between growing daughter chromosomes during their replication and partition.
Because dimer chromosomes can form only as a consequence of intermolecular recombination between daughter chromosomes, we reasoned that analysis of chromosome dimer resolution at the dif site provides a completely different method to study SCE. Previously, we reported that the frequency of recombination occurring at dif was increased above wild-type levels by a polA mutation, which is known by other assays to increase the frequency of chromosomal recombination events (22, 49). This result suggested that the frequency of SCE was higher in a polA mutant, that is, more dimers were formed and subsequently resolved at dif. Here we extend our analysis of chromosome dimer resolution at dif to other hyperrecombinogenic (hyper-rec) mutations and show that the dif recombination frequency is increased by these mutations as well.
We also analyzed the effects on SCE of mutations in some of the major genes associated with recombination. Recombination in E. coli has traditionally been thought of in terms of recombination pathways. The three pathways originally defined by Clark (7) are the RecBCD, RecE, and RecF pathways. The RecBCD pathway predominates when one of the recombination substrates is a linear DNA molecule (23, 38). Since most recombination assays in E. coli (e.g., conjugation and transduction) involve linear molecules, the RecBCD pathway has generally been considered the major pathway of recombination in wild-type cells. In conjugation or transduction assays, the RecE and RecF pathways of recombination normally contribute very little to the overall frequency of recombination events (6). These pathways become active, however, and restore recombination proficiency to recB or recC cells when suppressor mutations are introduced. For example, sbcA mutations activate the RecE pathway of recombination, and sbcB sbcC mutations activate the RecF pathway (27).
The recE gene forms part of the cryptic Rac prophage located in the terminus of the E. coli chromosome (5). Since recE is absent in some strains and is not expressed in sbcA+ cells, it can be considered a minor, or cryptic, recombination pathway. RecF is not encoded by a cryptic gene, and although it is not normally involved in conjugational or transductional recombination, recF mutants are as sensitive to UV irradiation as recB cells are (20). However, the reason for this sensitivity is not completely clear. It has recently been suggested that the primary role of RecF may not be in recombination at all but rather in the reassembly of a replication holoenzyme at the site of a stalled replication fork (10). Although this might be one function of RecF, our data demonstrate a significant role for RecF in recombination between sister chromosomes. This conclusion is consistent with that recently reported by Galitski and Roth (16), who used a genetic assay to analyze SCE in Salmonella typhimurium.
MATERIALS AND METHODS
Strains.
All strains used in this study were E. coli K-12 strains and are listed in Table Table1.1. PK3772 is C600 thi-1 thr-1 leuB6 thyA deoB or deoC. PK3872 is PK3772 leu+ thr+. All mutations were introduced into PK3772 or PK3872 by bacteriophage P1 transduction.
TABLE 1
Strains
| Strain | Relevant genotype | Parent strain | Source or reference |
|---|---|---|---|
| PK3772 | C600 (thi-1 thr-1 leuB6) thyA deoB or deoC | Laboratory stock | |
| PK3777 | polA12 metE163::Tn10 | PK3772 | 32 |
| PK3798 | recB21 thyA748::Tn10 | PK3772 | 47 |
| PK3802 | recA56 srl-300:Tn10 | PK3772 | 11 |
| PK3821 | recB21 recF143 thy+ zid-501::Tn10 | PK3772 | 8, 47 |
| PK3827 | lexA3 zja-505::Tn10 | PK3772 | 33 |
| PK3872 | C600 (thi-1) thr+ leu+ thyA deoB or deoC | PK3772 | Derivative of PK3772 |
| PK3874 | recF143 zid-501::Tn10 | PK3872 | 8 |
| PK3882 | recB21 recF143 thy+ zid-501::Tn10 | PK3872 | 8, 47 |
| PK3984 | ΔuvrD288::kan | PK3872 | 46 |
| PK3987 | uvrA6 | PK3872 | 37 |
| PK3996 | recB21 thyA748::Tn10 | PK3872 | 47 |
| PK4001 | dut-1 zic-4901::Tn10 | PK3872 | 19 |
| PK4035 | Δfur::kan | PK3872 | 41 |
Media.
Glucose-minimal medium was M-9 salts (2) containing 0.2% glucose, 2 μg of thymine ml−1, 1 μg of thiamine ml−1, and 2 μCi of [3H]thymine ml−1 (Sigma). Rich medium was identical to glucose-minimal medium but contained in addition 0.2% Isogro (39). Acetate-minimal medium was identical to glucose-minimal medium except that sodium acetate was substituted for glucose at a final concentration of 40 mM. Heavy minimal medium for the initial growth of cultures contained the appropriate 13C-labeled carbon compound and 15NH4Cl at 99% atom purity. Heavy rich medium contained [13C]glucose, 15NH4Cl, and [13C-15N]Isogro, all at 99% atom purity. Heavy isotopes were obtained from Isotec, Inc., Miamisburg, Ohio.
Density label assay.
The density label assay was performed as previously described (39). Net percentages of semihybrid density DNA for a given experiment (Table (Table2)2) were determined by subtracting the average percentage of semihybrid density DNA found in six independent density assays of xerC and xerD mutants, in which site-specific recombination at dif is prevented (39). For each experiment, the percentage of semihybrid density DNA in three, four, or five fractions was determined, and the average percentage of semihybrid density DNA in the same number of fractions from the experiments with xerC and xerD mutants was subtracted from this value. The number of fractions which gave the highest net value was then used to determine the average percentage of semihybrid density DNA for a given genotype. The use of either three, four, or five fractions was always found to give the highest net value in any given experiment. The background percentages of semihybrid density DNA from the xerC and xerD experiments were 7.0, 10.6, and 15.2% for three, four, and five fractions, respectively.
TABLE 2
Recombination frequencies at dif
| Straina | Genotype | Mediumc | No. of expts | Avg net % of semihybrid density DNAb |
|---|---|---|---|---|
| 3772/3872 | Wild type | Rich | 7 | 15.6 ± 2.6 |
| 3872 | Wild type | Glucose-min. | 3 | 16.0 ± 1.6 |
| 3872 | Wild type | Acetate-min. | 2 | 10.5 ± 1.1 |
| 3777 | polA12 | Rich | 3 | 24.3 ± 4.4 |
| 3984 | ΔuvrD288::kan | Rich | 2 | 23.3 ± 0.6 |
| 3984 | ΔuvrD288::kan | Glucose-min. | 1 | 29.3 |
| 4001 | dut-1 | Glucose-min. | 1 | 31.4 |
| 4035 | Δfur::kan | Rich | 2 | 24.5 ± 3.2 |
| 3802 | recA56 | Rich | 2 | 0.5 ± 0.5 |
| 3798/3996 | recB21 | Rich | 8 | 8.0 ± 1.4 |
| 3874 | recF143 | Rich | 6 | 8.9 ± 1.9 |
| 3821/3882 | recB21 recF143 | Rich | 4 | 2.3 ± 0.8 |
| 3827 | lexA3 | Rich | 1 | 15.1 |
| 3987 | uvrA6 | Glucose-min. | 1 | 15.8 |
RESULTS
Density label assay for dif recombination.
The results reported here were obtained by using the density label assay that we recently described (39) and which is shown in a simplified form in Fig. Fig.1.1. Exponentially growing cells uniformly labeled with heavy isotopes (15N and 13C) were shifted to light medium and grown for two more generations before DNA purification. All DNA was hybrid density after the first generation in light medium, and recombination between hybrid-density dif sites produces hybrid-density DNA in which the density of the strands changes at dif. Cell division is required for this recombination, and dimer chromosomes are resolved to monomers (39). An additional generation of growth and replication converts the hybrid-density dif sites into semihybrid-density sites (i.e., one-fourth heavy density), which are detected by the assay.
Density label assay for site-specific recombination at dif. Cells are grown in heavy medium, so that the DNA is uniformly labeled with 13C and 15N isotopes. Following a shift to light medium, the cells are allowed to grow for two more generations. Heavy DNA is indicated by solid lines, and light DNA is indicated by dashed lines. The dif site is in the middle of the molecule. If no recombination occurs at dif, equal amounts of hybrid and light DNA at dif are produced after two generations. If recombination occurs between dif sites in sister chromosomes after the first generation, recombinant hybrid DNA is produced, in which the density of both single strands switches from heavy to light at dif. After growth for a second generation in light medium, recombinant hybrid density DNA is converted to semihybrid density DNA, in which one-fourth of each duplex molecule is heavy density DNA. Semihybrid density DNA can be separated from hybrid and light density DNA in CsCl density gradients. See the text and reference 39 for further details. Rep., replication.
Recombination between dif sites can also occur in some cells after the second replication of the dif site. Two DNA doublings produce cells in which one dif site is hybrid density and the other is light density. Recombination between these sites at the time of cell division also produces semihybrid-density DNA. In the two generations of exponential growth that we used for the analysis, recombination of this type can occur only in cells that have not yet entered the D period (the interval between termination and cell division) at the time of the density shift (39).
To monitor the extent of DNA replication and the distribution of genomic DNA within the gradients, the cells were continuously labeled with [3H]thymine. The distribution of a 9.9-kb BglI dif fragment in the gradients was then determined by hybridization of the gradient fractions with a 32P-labeled dif probe. Figure Figure22 shows a typical result obtained when the density assay was performed with our wild-type strain.
Recombination at dif in wild-type (WT) cells. The density assay was performed with PK3872 grown in rich medium. (A) Analysis of 3H-genomic DNA shows equal amounts of hybrid and light DNA, indicating that two generations of growth and DNA replication had occurred. Since fractions are collected from the bottom of each gradient, higher fractions indicate lower densities. kdpm, 1,000 decays per minute. (B) Probing the same fractions shown in panel A with a 32P-labeled dif probe reveals a distinct semihybrid density DNA peak, which results from site-specific recombination at dif (39). y-axis units are arbitrary units from the phosphorimager.
Mutations increasing SCE.
A prediction from our previous results was that hyper-rec mutations should increase the frequency of SCE. The increased frequency of SCE should produce more dimer chromosomes, which in turn can be detected by the increased frequency of dimer resolution at dif. To test this correlation, we initially examined a polA mutant. polA codes for DNA polymerase I, and mutations in this gene were among the original hyper-rec mutations identified by E. B. Konrad (22). An increased frequency of SCE in a polA mutant background was also suggested by our earlier observation that polA dif double mutants exhibit increased filamentation and SOS induction, presumably due to the increased number of unresolvable dimer chromosomes (24). By use of the density assay, we observed that the level of recombination occurring at the dif locus was substantially increased in a polA mutant relative to that in its polA+ parent (Table (Table22).
To test further the correlation between dimer formation and dimer resolution at the dif site, we analyzed additional mutations originally identified by Konrad (22) as being hyper-rec. dut (also called dnaS) codes for dUTPase, and mutations in this gene result in increased incorporation of uracil into DNA. Uracil incorporated into DNA is rapidly removed by excision repair functions within the cell, and as a consequence, dut mutants tend to accumulate nicks in nascent DNA (21, 42). Similar to polA, the frequency of recombination at dif was also increased by a dut mutation (Fig. (Fig.3A;3A; Table Table2).2).
Recombination frequency at dif is sensitive to changes in SCE frequency. The density assay was performed with several different strains carrying mutations known to increase (dut and uvrD) or decrease (recA) genetic recombination. Only the distribution of DNA at dif in CsCl density gradients is shown for each experiment. (A) PK4001 (dut); (B) PK3984 (uvrD); (C) PK3802 (recA). The density assay of PK4001 (A) was performed in glucose-minimal medium; the other two experiments (B and C) were performed in rich medium. y-axis units in panel C indicate net counts obtained with an Ambis Systems imager.
Konrad also observed that a uvrD mutant (originally called mutU) was hyper-rec. Mutations in uvrD affect DNA helicase II, which functions in both excision and mismatch repair. Consistent with the results obtained with other hyper-rec mutants, a uvrD mutant also showed increased recombination at dif in the density assay (Fig. (Fig.3B;3B; Table Table2).2). The basis of the hyper-rec phenotype caused by uvrD mutations is not entirely clear, but in an assay based on recombination between tandem repeats, it appears to require RecF and SOS induction (3).
SOS induction has been shown previously to stimulate recombination in a different assay of SCE based on duplication formation due to unequal crossing over between direct repeats (13, 35). Because it was also involved in uvrD-stimulated recombination (3), we tested the possibility that some level of SOS induction was a general requirement of SCE. For this experiment, we analyzed a lexA3 mutant by the density label assay. Since lexA3 codes for a noncleavable repressor of the SOS regulon, cells carrying this mutation cannot be induced for SOS. Recombination at dif, however, was not significantly affected by lexA3 (Table (Table22).
It was also possible that the hyperrecombination seen in uvrD cells was a general property of cells affected in excision repair. Therefore, we tested the effect of another excision-deficient mutation, uvrA. SCE did not appear to be affected by this mutation (Table (Table22).
We also tested whether the frequency of SCE was influenced by oxidative DNA damage, as suggested by Galitski and Roth (16). For this experiment, we analyzed a fur mutant, which lacks a protein regulating the uptake of iron into the cell. fur mutants incorporate iron constitutively, and the excess iron in the cell stimulates the production of hydroxyl radicals and other DNA-damaging agents (41). The fur mutant we tested showed substantially increased recombination at the dif site (Table (Table2),2), which is consistent with DNA damage being a cause of SCE. That the increased levels of SCE occur as a result of repair in fur mutants is indicated by the fact that fur recA or fur recB double mutants are not viable under aerobic conditions (41).
Recombination pathways involved in SCE.
As mentioned above, the recombination-related genes involved in SCE have previously been studied by genetic assays based on recombination between repeated regions (9, 16, 48). Since our experiments demonstrated a correlation between recombination at dif and SCE elsewhere in the chromosome, we have also utilized the density assay as an independent test to identify recombination functions required for SCE. We initially tested a recA mutant, which would be expected to block all recombination and undergo no SCE. Consistent with this expectation, no recombination was observed at dif (Fig. (Fig.3C).3C). Since recombination at dif itself is RecA independent (24), this result demonstrates that SCE does not occur in a recA mutant. It also is consistent with the ability of recA mutations to suppress the Dif phenotype: in the absence of SCE, dimer chromosomes are not formed, and resolution of dimers at dif is not required (24).
Since the RecBCD pathway is usually considered the major route for recombination and chromosome interactions in wild-type E. coli, we tested whether resolution at dif was affected in a recB mutant. Figure Figure4A4A shows a typical result obtained with such strains. Although recombination was reduced by the recB mutation (51% of wild-type levels), it still clearly occurred. This was a somewhat surprising result, because recB mutations, as well as recA mutations, had previously been shown to suppress the Dif phenotype (24). We originally proposed that the suppression of the Dif phenotype seen in recB mutants was due to a lack of dimer chromosome formation caused by this mutation. Since SCE obviously still occurred in recB mutants, the ability of recB to suppress the Dif phenotype was likely due to another effect of this mutation. We have subsequently observed that the lack of filamentation of dif recB mutants is due to the inability to induce the SOS response (39a).
RecBCD and RecF provide alternate routes to SCE. The density assay was performed with strains carrying mutations in recB, recF, or both of these genes. The recombination frequency at dif was reduced to approximately the same extent by either single mutation, and it was virtually absent when both mutations were present. (A) PK3798 (recB); (B) PK3874 (recF); (C) PK3882 (recB recF). y-axis units in panel A indicate net counts obtained with an Ambis Systems imager.
Since recombination at dif was still observed in a recB mutant, it suggested that SCE could also occur by a pathway other than the RecBCD pathway. Therefore, we tested a recF mutant, and the result was similar to that observed in the recB strain (Fig. (Fig.4B;4B; Table Table2).2). When the recB recF double mutant was tested in the density assay, recombination at dif was reduced even further and only a small, residual amount of resolution still occurred at dif (Fig. (Fig.4C).4C). These results demonstrate that RecBCD and RecF both contribute to SCE and the formation of dimer chromosomes.
Frequency of SCE for cells grown in different media.
Most of the experiments described above were performed on cells grown in rich medium, in which the generation time was 33 min. In some cases, however, it was necessary to grow the cells in glucose-minimal medium. For example, PK4001 (dut) grew very poorly in rich medium, and minimal medium was used for the experiment shown in Fig. Fig.3A.3A. To determine whether the growth medium itself affected SCE frequency, we also conducted experiments on our wild-type strain (PK3872) grown in glucose-minimal or acetate-minimal medium. In glucose-minimal medium, the generation time was increased to approximately 63 min (39). At this growth rate, the frequency of recombination observed in cells grown for two generations in light medium was essentially unchanged (16%) (Table (Table2).2). When PK3872 was grown in acetate-minimal medium, the generation time was increased to approximately 130 min. Under these conditions, a slight decrease in recombination was observed at dif (10.5%) (Table (Table2).2). These data suggest that the SCE frequency may decrease slightly with a decreasing growth rate, but the effect is not substantial.
DISCUSSION
Assay for SCE.
The density assay that we have used here to analyze SCE differs considerably from the genetic assays previously used for this purpose. The dif locus reduces chromosome dimers, which are the products of SCE, to monomers by providing a compensating exchange. Therefore, the frequency of recombination at dif should reflect the frequency of recombination that occurs throughout the rest of the chromosome, and our results with known hyper-rec and hyporecombinogenic mutations are consistent with this interpretation (Fig. (Fig.3;3; Table Table2).2). In contrast to genetic assays, an advantage of the density assay is that it uses the entire bacterial chromosome as the target for SCE. Furthermore, all regions of the chromosome are at their normal locations and undergo the normal interactions inherent in replication, repair, and recombination.
The density assay does, of course, have some disadvantages. One disadvantage is that the assay itself is labor-intensive. Also, differences of orders of magnitude cannot be detected. The assay works well when the frequency of SCE falls between 10 and 15% per generation, as with the wild-type strain we used (39), but very low levels of SCE would be difficult to reliably detect. Increases in SCE can be readily observed (Fig. (Fig.3),3), but the maximum level of recombination and semihybrid density DNA that can be detected is 50%. This arises since even numbers of SCE events will not produce dimer chromosomes, and at very high frequencies of SCE, even and odd numbers of exchanges will become equally frequent. Consequently, the density assay rapidly loses sensitivity as the frequency of recombination approaches 50%.
Since our assay is quite different from those previously used to analyze SCE, it is interesting to compare our results for the frequency of SCE (∼16% for wild-type cells) with those recently reported by Galitski and Roth (16). Their assay was based on recombination between tandem repeats and consequently used a much smaller target (∼40 kb). They estimated a frequency of recombination events per generation of 0.49%, and when this value is extrapolated to the size of the entire chromosome, an SCE frequency of 58% results. Considering the differences between the two procedures (recombination detected by physical rejoining versus sectored colonies) as well as the species difference (E. coli versus S. typhimurium), the results are rather similar. A possible source of the higher estimate by their assay is that in addition to scoring events due to unequal crossing over, which is similar to SCE, their assay might also score events due to intrachromosomal recombination and looping out of the scored loci. It should be noted that if the difference is due to looping out, this implies that SCE and looping out have similar genetic requirements.
Two pathways of SCE.
Perhaps the most interesting result that we report here is that both RecBCD and RecF are involved in producing SCE. The data presented in Fig. Fig.44 and Table Table22 suggest that each contributes more or less equally to SCE, and each operates independently of the other. Although this might seem consistent with current interpretations of recombination pathways, it should be stressed that sbcB sbcC mutations were not required to observe RecF-dependent recombination in our assay. Consequently, SCE occurred in the absence of any mutations that stabilize 3′ ends of DNA, so it did not involve the classical recF recombination pathway. Except for some possible unknown effect of the recB mutation that was necessary to conduct the assay, the RecF recombination should be occurring at wild-type levels. These results are consistent with those recently reported by Galitski and Roth (16), who also observed that RecF contributed substantially to chromosomal recombination events in wild-type cells.
Given that both RecBCD and RecF contribute to SCE, what does this tell us about the nature of the recombination substrates present within the cell? recB mutant cells show substantially reduced viability in comparison to recB+ cells (4), so the RecBCD nuclease obviously plays an important role in the cell apart from its role in conjugational and transductional recombination. Since RecBCD is a very potent double-stranded DNA exonuclease, degradation of DNA by this enzyme could be quite damaging to a cell unless it contains other copies of the degraded genes. Keeping this in mind, a current model for the normal role of RecBCD is in the repair of broken replication forks (1, 26). When a replication fork breaks and releases the end of one of the growing daughter chromosomes, it presents a double-strand end, which is required for the loading of the RecBCD nuclease (40). Degradation would proceed backwards, towards the origin of replication (oriC), and the other nascent daughter chromosome would contain copies of the degraded genes. The combined actions of RecBCD and RecA proteins on the broken arm of the chromosome would facilitate its invasion into the intact daughter chromosome. This establishes a Holliday junction, which is thought to promote restoration of the collapsed replication fork. Since the Holliday junction can be resolved in either of two orientations, a dimeric chromosome should be produced in half of these events (for example, see Fig. 8 and 9 in reference 1).
The source of the broken replication forks that are repaired by RecBCD could be spontaneous breaks in the template strands. If a replication fork proceeds through a region containing such a nick, a double-strand break would be generated. It can be estimated that chromosomes contain about 10 single-strand breaks per template (17). If only 1.5% of these single-strand breaks resulted in a broken replication fork, about 15% SCE per generation would be produced.
What is the role of RecF in SCE? Breakage and reformation of forks by a RecF-mediated process seem unlikely by current understanding of these mutants, especially in a strain that is recBC+ sbcBCD+. RecF, therefore, presumably operates on substrates that do not involve double strand-breaks. RecF has been shown to be of major importance in the repair of daughter strand gaps left on chromosomes following UV irradiation of cells (44, 45). When a replication fork encounters a lesion in the template, such as a thymine dimer, it can reinitiate replication downstream in a process known as translesion synthesis (36). This leaves a damaged single-stranded region in one chromosome that can be repaired only with information from the undamaged sister chromosome. Invasion of the single-stranded region of the damaged chromosome into the undamaged chromosome establishes a Holliday junction which can migrate across the lesion. The damaged DNA can then be excised by repair enzymes within the cell and resynthesized from information in the undamaged strand (28). Resolution of the Holliday junction, again, can occur in either of two orientations, and a dimer chromosome will result half of the time. Resumption of synthesis at the stalled replication fork could require RecF, as proposed by Courcelle et al. (10).
Sources of SCE.
As mentioned above, one possible source of SCE is single-strand breaks in the template strand. These breaks could arise from the replication process itself. This is suggested by the increased level of SCE in polA and dut mutants, which increase the frequency of single-strand breaks in newly synthesized DNA (21, 25, 34, 42). If some of these breaks are not closed before they appear as template DNA, a double-strand break would occur. The increased frequency of SCE in mutants affected in sealing breaks indicates that at least some of the spontaneous SCE arises from a background level of breaks of this type.
It can be argued that the distribution of SCE events might not be uniform around the chromosome. Louarn and coworkers have observed that hyperrecombination occurs in the terminus region of the chromosome and that the frequency of recombination in that region increases more than 3 orders of magnitude (29). The hyperrecombination is RecBCD dependent, which indicates that it is derived from double-strand breaks or from nicks leading to double-strand breaks (9). Louarn and coworkers have demonstrated that this terminal hyperrecombination is not related to the meeting of replication forks in this region (29), and they suggest that it is due to the postreplication reconstruction of nucleoid organization that occurs prior to cell division. Decatenation and other events associated with this reconstruction are proposed to lead to double-strand breaks, which are the basis of the hyperrecombination.
Another source of SCE is oxidative damage to DNA, which occurs spontaneously in cells grown under aerobic conditions (12, 16). As a test of this hypothesis, we examined a fur mutant, which shows increased amounts of oxidative DNA damage (41). Table Table22 shows that SCE was substantially increased by this mutation. Oxygen by-products can affect DNA in a variety of ways, including the generation of single-strand breaks and damage to DNA bases (15). As discussed above, the breaks would ultimately be repaired by RecBCD nuclease, while repair of any gaps left by translesion synthesis would involve the action of RecF. As mentioned by Galitski and Roth (16), oxidative DNA damage could contribute to a large part of the spontaneous level of SCE observed in wild-type cells.
ACKNOWLEDGMENTS
The research was supported by National Institutes of Health grant GM32968.
We thank Heather Szerlong and Andrias Hojgaard for assistance with experiments and helpful discussions.