Resolution of Joint Molecules by RuvABC and RecG Following Cleavage of the Escherichia coli Chromosome by EcoKI

DNA double-strand breaks can be repaired by homologous recombination involving the formation and resolution of Holliday junctions. In Escherichia coli, the RuvABC resolvasome and the RecG branch-migration enzyme have been proposed to act in alternative pathways for the resolution of Holliday junctions. Here, we have studied the requirements for RuvABC and RecG in DNA double-strand break repair after cleavage of the E. coli chromosome by the EcoKI restriction enzyme. We show an asymmetry in the ability of RuvABC and RecG to deal with joint molecules in vivo. We detect linear DNA products compatible with the cleavage-ligation of Holliday junctions by the RuvABC pathway but not by the RecG pathway. Nevertheless we show that the XerCD-mediated pathway of chromosome dimer resolution is required for survival regardless of whether the RuvABC or the RecG pathway is active, suggesting that crossing-over is a common outcome irrespective of the pathway utilised. This poses a problem. How can cells resolve joint molecules, such as Holliday junctions, to generate crossover products without cleavage-ligation? We suggest that the mechanism of bacterial DNA replication provides an answer to this question and that RecG can facilitate replication through Holliday junctions.


Introduction
Homologous recombination is used to repair DNA doublestrand breaks in E. coli. This reaction is catalysed by RecBCD and RecA proteins, which resect DNA ends and mediate strandexchange, respectively [1]. The products of strand exchange are understood to be joint molecules tethered to each other by Holliday junctions and replication forks. The Holliday junctions are then assumed to be migrated along the paired molecules by the RuvAB or RecG proteins and then resolved either by RuvCmediated cleavage in the presence of RuvAB, followed by ligation, or in some unknown way in the presence of RecG [2]. Two main classes of hypotheses have been proposed to explain resolution of Holliday junctions by RecG. First, RecG could operate with an unknown nuclease to cleave Holliday junctions. A specific example of such a mechanism involving nicked Holliday junctions has been suggested based on a model proposed for S. pombe meiotic recombination [3]. Second, RecG could branch migrate one Holliday junction into a DNA end [4] or into another Holliday junction as originally hypothesised by Thaler and Stahl for lambda phage recombination [5]. Synthesis-dependent strand annealing (SDSA) is another example of this second class of cleavage-ligation independent model [6]. A unifying feature of these cleavageligation independent models is that they do not lead to crossing over. According to the first class of hypotheses, evidence of Holliday junction cleavage-ligation should be detected. According to the second class of hypotheses evidence for crossing over should not be detected. We have therefore set out to obtain evidence for Holliday junction cleavage-ligation and for crossing over via the RecG pathway. We see no evidence of Holliday junction cleavageligation but do detect resolution to crossover products implying that neither the first nor the second class of hypotheses is correct and requiring a new model for the action of RecG.
Previously, we developed a system for generating DNA doublestrand breaks in the E. coli chromosome using the EcoKI restriction enzyme [7]. EcoKI is a type I restriction-modification complex that modifies hemimethylated DNA target sequences and cleaves fully unmethylated DNA target sequences. Its recognition sequence is AAC(N 6 )GTGC but cleavage occurs at a site distant from this sequence (reviewed in [8]. The restriction activity of EcoKI can be attenuated temporarily by the cell, a phenomenon referred to as restriction alleviation (RA). DNA damaging treatments that cause the formation of unmethylated target sites induce RA [9,10,11,12]. RA is also observed when the genes encoding a restriction-modification system are transferred into an E. coli cell lacking that system. RA is dependent on the protease specified by the clpX and clpP genes [13]. ClpXP protease alleviates restriction by degrading the HsdR subunit of EcoKI as the complex translocates along the DNA [14]. ClpXP is also responsible for restriction alleviation of cells treated with UV light, naladixic acid or 2-aminopurine [14]. Notably, it has been proposed that the original function of RA lies in protecting the chromosome when recombination generates unmethylated target sequences [15]. We have used 2-aminopurine treatment of clpX mutant cells to generate DNA double-strand breaks and observe DNA repair by homologous recombination [7]. This system has features that distinguish it from other systems for studying DNA double-strand break repair. Because the breaks are generated by a restriction endonuclease, it is expected that the damage will be more uniform than the damage generated by a DNA damaging agent such as X-or c-irradiation. However, in contrast to systems where a restriction endonuclease is induced in a cell, cleavage by EcoKI is expected on one sister chromosome only. Furthermore, since the cleaved target sequence is generated by DNA synthesis, the majority of cleaved sites are expected to lie in newly replicated DNA.
It was formerly reported that a recG mutant was highly sensitive to EcoKI mediated DNA cleavage while a ruvABC mutant was minimally sensitive [7]. However, further investigation revealed that several observations reported in that paper could not be reproduced. Notably, we were unable to reproduce the reported sensitivities of recF, recJ, recQ and sbcC strains, and were only able to detect a modest sensitivity of a recN strain. By contrast, significant sensitivity of a ruvABC mutant was detected and the sensitivities of recA, recB and recG mutants were confirmed [16]. The work reported here was designed to set the record straight with respect to the effects of the genes significantly implicated in the repair of EcoKI breaks and in particular to investigate the pathways of resolving recombination intermediates.

Repair of DNA double-strand breaks is required for cell viability
Cromie and Leach reported that recombination defective mutants of DclpX hsdR + strains survive poorly following treatment with 2-aminopurine (2-AP; [7]. We have improved their method of analysis by treating cells with 2-AP at 20 mg/ml and following their survival as a function of time ( Figure 1). Three rounds of replication are predicted to generate an unmethylated target: the first is required to incorporate 2-AP opposite cytosine, the second to incorporate thymine opposite 2-AP and the third to incorporate adenine opposite thymine. Consistent with the prediction that three rounds of DNA replication of about 30 minutes each are required to generate the unmethylated targets that are the substrates for cleavage by EcoKI, recA, DrecBCD, DrecG and DruvABC mutants showed no decrease in viability after 50 minutes but were affected after 100 minutes of 2-AP treatment ( Figure 1A). The kinetics of killing was similar in the recA, DrecG and DrecBCD mutant strains though the extent of killing was greater in DrecBCD. Interestingly, the DruvABC mutant showed continued killing at later times following treatment. On the other hand, the DrecG DruvABC strain was exquisitely sensitive to 2-AP treatment and already displayed killing at 50 minutes post treatment, suggesting a more rapid accumulation of unmethylated targets than the expected three rounds of DNA replication required in the other mutants. In all mutants apart from the DrecG DruvABC strain, there was little killing by 2-AP in an hsdR mutant strain ( Figure 1B). The DrecG DruvABC strain showed some killing by 2-AP in the absence of EcoKI, but substantially less than in the presence of EcoKI. All together, these data indicate that the majority of killing after the 2-AP treatment was caused by the EcoKI endonuclease.
The viability of different combinations of mutations was investigated to test the possible interactions between the ''early'' DrecA mutation and the ''late'' DrecG and DruvABC mutations. As shown in Figure 1C, the sensitivity of a recA DrecG double mutant strain to DNA double strand breaks induced by 2-AP was similar to the sensitivity of either recA or DrecG single mutant strain suggesting that RecG and RecA may be operating in the same pathway. By contrast, the sensitivity of a DruvABC mutant increased throughout the time course and at late times was greater than that of a recA mutant. At early times the sensitivity of a recA DruvABC double mutant was greater than that of a DruvABC mutant. These data suggest that at late times RuvABC contributes to a survival pathway independent of RecA and at early times RecA contributes to a survival pathway independent of RuvABC.
ruv mutants accumulate Holliday junctions while recG mutants do not To understand the role of RuvABC and RecG in processing branched DNA intermediates, we have carried out pulsed-field gel electrophoresis on the chromosomal DNA of DclpX mutants after treatment with 2-AP. The conditions used for gel electrophoresis allowed circular and branched molecules to be retained in the wells whereas linear DNA fragments of a wide spectrum of sizes (450 kb to 4.5 mb) migrated as a single band in the gel. As shown in Figures 2A1, treatment with 2-AP in the presence of EcoKI (hsdR + cells) induced the formation of linear DNA fragments above that observed in the absence of 2-AP ( Figure 3A1) or the absence of EcoKI ( Figure 3A2). Consistent with previous observations, DrecBCD DclpX mutant strains treated with 2-AP in the presence of EcoKI accumulated more linear DNA [17,18,19,20] and recA DclpX mutant strains treated with 2-AP in the presence of EcoKI showed loss of DNA from the wells, confirming the ''reckless'' DNA degradation previously observed in the absence of RecA after induction of DNA damage [21]. Surprisingly, the behaviours of DrecG DclpX and Druv DclpX mutants were very different from each other despite both mutants being sensitive to 2-AP. The DrecG DclpX strain's response to 2-AP treatment was similar to the rec + DclpX strain whereas the DruvAB DclpX and DruvC DclpX strains showed no detectable linear DNA ( Figure 2A1). Linear fragments were observed in the recA DruvAB DclpX, recA DruvC DclpX, DrecBCD DruvAB DclpX and DrecBCD DruvC DclpX mutant strains, suggesting that their absence in the DruvAB DclpX and DruvC DclpX mutant strains was caused by linear fragments trapped in the wells as they recombined with other DNA ( Figure 2A1). To investigate if the lack of linear DNA in the DruvAB DclpX and DruvC DclpX strains was due to chromosomal fragments tied together by structures such as unresolved Holliday junctions, a plasmid expressing the bacteriophage resolvase RusA was intro- duced into these strains. As shown in Figures 2A2, the presence of a plasmid encoding the RusA nuclease liberated linear DNA from both Druv DclpX strains; behaviour not observed in the presence of the plasmid vector lacking the rusA gene ( Figure 2A3).
The behaviour of these strains was further investigated by studying their chromosomal DNAs digested by the NotI restriction enzyme on pulsed field gels (Figures 2B1, 2B2 and 2B3) with respect to controls ( Figures 3B1. 3B2 and 3B3). Strikingly, the deficit of NotI fragments entering the gel from 2-AP treated DruvAB DclpX and DruvC DclpX strains suggests that joint molecules were connecting a substantial proportion of NotI cleaved DNA. Following NotI cleavage, a small increase in linear fragment DNA entering the gel was observed in DruvAB DclpX and DruvC DclpX mutant strains over DruvAB recG DclpX and DruvC DrecG DclpX mutant strains ( Figure 2B1), which contrasts with the absence of fragments visualised in ruv and ruv recG mutants without NotI cleavage. This finding suggests that RecG is carrying out a role in resolving joint molecules in the absence of RuvABC but RecG is not as efficient in liberating NotI fragments as is RuvABC. Finally, the role of RecG requires chromosome fragmentation with NotI to be visualised, which is consistent with the products of RecG action not including linear molecules.

Chromosome dimer resolution is required for cell viability in the presence and absence of RuvABC and RecG
In E. coli, crossing over can be assessed by its consequence on the segregation of the single circular chromosome [22]. Crossing over generates a single dimeric chromosomal structure, which is unable to segregate to the two daughter cells during cell division. Therefore, E. coli has evolved a dimer resolution pathway involving the XerCD proteins acting at the dif site, located close to the terminus of chromosome replication [23]. In xerC, xerD or dif mutants, dimers cannot be resolved back to monomers and the consequent segregation problem leads to cell death.
In order to test whether recombination stimulated by EcoKI cleavage of the chromosome results in crossover products leading to chromosome dimer formation, we studied the sensitivity of DclpX DxerC and DclpX Ddif mutants to 2-AP in the presence or absence of DruvABC, DrecG, and recA ( Figure 4). DclpX DxerC and DclpX Ddif mutants were modestly sensitive to 2-AP (about 10 fold; Figure 4A). DruvABC DclpX DxerC and DruvABC DclpX Ddif mutants were significantly more sensitive to 2-AP than DruvABC DclpX mutants (about 100 fold; Figure 4B). Similarly, DrecG DclpX DxerC and DrecG DclpX Ddif mutants were significantly more sensitive to 2-AP than DrecG DclpX mutants (about 100 fold; Figure 4C). These results indicate that recombination in the presence or absence of RuvABC or RecG leads to the formation of a significant proportion of chromosome dimers that require XerCD action at dif for survival. Consistent with a requirement for recombination to produce dimers requiring XerCD and dif for resolution, a recA DclpX Ddif strain was no more sensitive to 2-AP than a recA DclpX strain ( Figure 4D).

Discussion
In this study, we have characterised the pathways of joint molecule resolution in E. coli following the generation of DSBs with EcoKI in a clpX mutant. Specifically, we have explored the roles of ruvABC and recG genes in the survival of EcoKI mediated DSBs, their roles in the formation of dimeric chromosomes and in resolving joint molecules detected on gels. Throughout this work, we have been careful to control for non-specific effects of 2-AP treatment by comparing the behaviour of hsdR+ and hsdR mutant strains. In all situations we attribute to double-strand break repair only the effects observed in the presence of EcoKI nuclease (hsdR+).

Survival of cells after EcoKI-mediated DNA double-strand breaks
The recA and recBCD genes, essential for the early stages of DNA double-strand break repair (DSBR), are required for survival of breaks generated by EcoKI. In addition both the recG and ruvABC genes, responsible for the resolution of Holliday junctions, are required for survival, though a DrecG DruvABC double mutant is significantly more sensitive to these breaks than are the single mutants. The high sensitivity of the DrecG DruvABC double mutant may arise from a combination of factors. First, there is a notable sensitivity of this strain to 2-AP even in the absence of EcoKI. Second, the early response to 2-AP treatment suggests that, in this mutant, the pathway of DSB formation may be different to the other strains. Third, there may exist a pathway of survival (e.g. via replication fork reversal) that requires resolution of Holliday junctions even in the absence of recombination (see below). The observations that recG and ruv single mutants are sensitive to 2-AP is reminiscent of the requirements for both recG and ruv for the repair of SbcCD-induced breaks at a DNA palindrome [24]. Both of these reactions are predicted to be DSB repair events occurring following DNA replication between one cleaved and one intact sister chromosome.
Surprisingly, a recA mutant is not more sensitive than any other single mutant tested here and is less sensitive at late times than a DruvABC mutant. This suggests the existence of a RecAindependent but RuvABC-dependent pathway for survival of double-strand breaks. The only known RuvABC-dependent, RecA-independent reaction is replication fork reversal (Seigneur et al. 1998) and it is possible that recovery of a small fraction of intact circular chromosomes could be mediated by a combination of RecBCD mediated degradation and RuvABC-mediated fork reversal as shown in Figure 5.

Formation and resolution of Holliday junctions
In order to understand the contributions of RecG and RuvABC to the resolution of Holliday junctions, we analysed by pulsed field gel electrophoresis the genomic DNA of cells in which EcoKI breaks had been generated. Long linear DNA molecules are able to enter a pulsed field gel whereas long branched and circular molecules cannot (Nakayama et al., 1994). Because the E. coli chromosome is circular it does not enter the gel. As shown in Figure 2A1 and 2A3, linear DNA is generated by treatment of a DclpX mutant with 2-AP implying that cleavage with EcoKI produces some linear DNA fragments. Recombination of such linear fragments will produce branched molecules that will not enter the gel and, if these branched molecules include linear molecules joined together by Holliday junctions, cleavage-ligation of the junctions can regenerate linear DNA fragments (see Figure 6). DruvAB DclpX and DruvC DclpX mutants do not produce a detectable level of linear DNA, suggesting the presence of Holliday junctions tying DNA molecules together. Consistent with this explanation, expression of the bacteriophage resolvase RusA in the DruvAB DclpX and DruvC DclpX mutants liberates linear fragments ( Figure 2A2). Surprisingly given its sensitivity to 2-AP, the DNA of a DrecG DclpX mutant behaves similarly on a pulsed field gel to the DNA of a DclpX mutant following treatment with 2-AP. Within the limits of detection of this methodology, our data suggest that the RuvABC resolvase is able to act in the DrecG DclpX mutant and resolve a substantial proportion of the Holliday junctions tying the DNA linear molecules together. Nevertheless, this action of RuvABC is not sufficient to prevent sensitivity of the DrecG DclpX mutant to 2-AP. We suggest therefore, that the sensitivity of the DrecG DclpX mutant to 2-AP is either explained by the presence of some critical unresolved joint molecules despite the ability of RuvABC to visibly resolve the Holliday junctions in our gel assay or by the action of RecG in a step other than the resolution of Holliday junctions.
RecG and any proteins working with it are unable to produce a detectable level of linear DNA fragments in the absence of RuvABC ( Figure 2A1 and 2A3). The generation of these fragments as a function of the presence of RuvABC, or RusA in the absence of RuvABC, suggests the junction of two DNA molecules by one or more Holliday junctions linking two DNA ends ( Figure 6). Cleavage-ligation of Holliday junctions has the potential to generate linear DNA depending on the plane of resolution of the junctions and the number of junctions present ( Figure 6A and 6C). Resolution can also occur by branch migration if the ends are joined by an even number of Holliday junctions ( Figure 6D). However, resolution of junctions by branch migration will never generate linear DNA if the ends are joined by an odd number of junctions ( Figure 6B). As observed in the ruv mutants, the RecG pathway does not result in linear DNA, which implies that no Holliday junction cleavageligation can be detected. If RecG can resolve joint molecules simply by branch migration, our result implies either that this reaction is too weak to liberate DNA fragments or that there are predominantly odd numbers of Holliday junctions between the DNA ends. Following cleavage with NotI, a small increase in the liberation of linear DNA fragments is observed in ruv mutants over what is observed in ruv recG mutants ( Figure 2B1 and 2B3). This suggests that RecG is capable of resolving some joint molecules within the context of products that require NotI cleavage for visualisation. This is consistent with the proposal that RecG may resolve some critical junctions. Formation and resolution of chromosomal dimers Mutations in xerC and dif confer modest sensitivity to EcoKI mediated breaks. This suggests that a proportion of these breaks are repaired by a mechanism that yields dimeric crossover products. Interestingly, DxerC and Ddif mutations confer a greater sensitivity to EcoKI breaks in DruvABC or DrecG mutants. A similar sensitivity of both recG and ruv mutants to the inactivation of the XerCD/dif system has been observed after induction of DSBs using the rare-cutting endonuclease I-SceI [6]. The increase in sensitivity of the DrecG mutant can be due to crossover products generated by cleavage-ligation of Holliday junctions using RuvABC. However, the reason for the increase in sensitivity of the DruvABC mutant is not easy to explain, as no Holliday junction nuclease is known to act in conjunction with RecG and the pulsed field gel electrophoresis presented here shows that no cleavageligation of Holliday junctions can be detected in the presence of RecG and absence of RuvABC. The simplest implication of this work is that resolution of intermediates by RecG results in crossing over without cleavage-ligation of Holliday junctions.

Conclusions
Here we show that, following induction of DNA double-strand breaks by EcoKI in a clpX mutant strain, RuvABC and RecG work very differently from each other and cannot be considered simply as catalysing steps in redundant pathways. RuvABC behaves as predicted for a protein that can resolve Holliday junctions by cleavage and this cleavage, followed by ligation, can lead to crossing over. Resolution by RecG also leads to crossing over but we can detect no evidence of Holliday junction cleavageligation via this pathway. This creates an apparent contradiction since all the standard cleavage-ligation independent models for recombination, such as SDSA, do not lead to crossing over.
A new model is required that can allow the maturation of a Holliday junction intermediate to a crossover product without junction cleavage-ligation. We suggest here that one way in which a Holliday junction can be matured to a crossover product without cleavage-ligation is if two new replication forks run through the junction. This would not be possible in most eukaryotic cells where new rounds of DNA replication are not initiated until after cell division. However, in bacteria new rounds of replication are normally initiated prior to cell division so it would be normal for an unresolved Holliday junction to act as a potential barrier to the passage of new forks. We suggest that RecG may facilitate the passage of the replication forks across the junction as detailed in Figure 7.

Bacterial strains
All the E. coli strains described in Table 1 are derivatives of BW27784 with the following genotype: lacI q , rrnB3, DlacZ4787, hsdR514, DaraBAD567, DaraFGH, DrhaBAD568, W(DaraEpP CR18 -araE) [25]. This strain allows homogeneous expression from the P BAD promoter thanks to a deletion of the genes encoding the AraFGH transporter and expression of the low-affinity, highcapacity AraE transporter from the constitutive promoter P CP18 .
The DclpX 1789 , DrecG 2472 , DruvAB 2757 , and DruvC 2731 mutations were generated by plasmid-mediated gene replacement (PMGR) using pTOF24 derivative vectors, carrying homology arms spanning the gene of interest [26]. These homology arms were generated by crossover PCR using primers described in Table 2 and inserted into pTOF24 using SalI and PstI restriction enzymes. The mutations xerC Y17 ::cat (derived from strain DS984 obtained from D. Sherratt), DruvABC 65 eda 51 ::Tn10 (derived from strain N4155 obtained from R. Lloyd), DrecG 263 ::kan (derived from strain N3793 obtained from R. Lloyd), difD6::kan (derived from strain GR47 obtained from D. Sherratt), recA::cat (derived from strain DB1318 obtained from D. Botstein) and DrecBCD::kan (derived from strain JJC1086 obtained from B. Michel) were introduced by P1 transduction. HsdR + derivatives were made by bacteriophage l insertion and excision as described by Arber and collaborators [27] using l NM1048 containing the wild type hsdR gene [28]. The HsdR phenotype was tested using methylated and unmethylated derivatives of l clear and l virulent. The plasmids pBAD18 [29] and its derivative pBAD-rusA, constructed by V. Bidnenko, were obtained from B. Michel.

Standard DSBR assay
This assay was used to obtain the viability curves presented in Figures 1 and 2

Pulsed field gel electrophoresis (PFGE)
Overnight cultures grown in LB at 37uC of strains to be tested were diluted to an O.D. of A 600 = 0.02 in LB and grown to O.D. 0.2 at 37uC under agitation. At that time, when appropriate, 100 mg/ml of 2-AP was added and the cultures were incubated for a further 2.5 hours (strains carrying the pBAD-rusA plasmid were always cultured in the presence of 0.002% arabinose to induce expression of RusA). At this point, 5 ml samples of cultures were extracted and spun down for 10 minutes at 3,500 rpm before resuspending the pellet in TEE solution (10 mM Tris, 100 mM EDTA, 100 mM EGTA, pH 8.5) to give an O.D. of 0.9. 350 ml of cells was mixed with 350 ml of 2% low melting point agarose (GIBCO) and cooled to 55uC. The mixture was immediately pipetted into disposable Biorad CHEF plug moulds and refrigerated until set. The plugs were then removed from the moulds and each set of ten incubated in 10 ml of lysozyme solution at 37uC with gentle agitation for two hours. Plugs were incubated overnight at 55uC in 5 ml of proteinase K solution and then rinsed in 10 ml TE buffer (10 mM Tris, 1 mM EDTA, pH 7.5) for 3X 1 hour. Plugs were then washed in 10 ml 1mM PMSF solution in TE buffer for 2 X 1 hour and then rinsed in 10 ml TE buffer for 2X 30 minutes. All TE and PMSF wash steps were carried out at room temperature under gentle agitation. The plugs were stored in TE buffer at 4uC and used within two months. The gels presented are representative of at least two gels run using plugs prepared from at least two independent cultures. In these experiments a concentration of 100 mg/ml 2-AP was used Figure 7. Model for the generation of chromosome dimers without Holliday junction cleavage-ligation. Bacteria, such as E. coli, have circular chromosomes and at normal growth rates reinitiate DNA replication before the previous round of replication has completed and before cell division takes place. This means that an unresolved Holliday junction is a potential barrier to the passage of the next set of replication forks. We propose here that replication through the Holliday junction may be possible and that this may be facilitated by the branch migration protein RecG. A. Chromosome in which a Holliday junction (HJ) has formed following the passage of a replication fork (RF1). A second pair of replication forks (RF2) are shown approaching the Holliday junction. The first two chromosomes to be produced by this replicating structure are labelled C1 and C2. The two chromosomes destined to be made from C1 are labelled C1.1 and C1.2 and the chromosomes destined to be made from C2 are labelled C2.1 and C2.2. The DNA strands that have exchanged to form the Holliday junction (and strands templated on these) are shown in red whereas the DNA strands that have not exchanged (and strands templated on these) are shown in blue. The four double-stranded molecules formed by the second pair of replication forks are shaded in light blue and pink. B. Two new forks (RF2) approaching the Holliday junction. When the pair of RF2 forks approaches the Holliday junction, the positive supercoiling ahead of the forks is predicted to push the junction ahead of them. At some point the forks are likely to stall, presumably because the Holliday junction impedes their progression. At this point, branch migration of the Holliday junction to the fork will lead to a swapping of newly synthesised sister chromosome arms. Chromosomes C1.1 and C2.1 will be connected to the unreplicated arm of C1 while chromosomes C1.2 and C2.2 will be connected to the unreplicated arm of C2. C. Formation of two monomeric and one dimeric chromosome. The replication machinery is reassembled on the two hybrid RF2 forks and replication continues. The figure illustrates the point where the RF2 forks have passed through the Holliday junction leaving the red strands crossed over (CO). The RF1 forks have completed their replication and no longer exist. When the RF2 forks complete replication and meet at the terminus, two monomeic blue chromosomes (C1.1 and C2.2) will have formed as well as one dimeric red (crossover) chromosome (C1.2-C2.1). The shading of the molecules formed by the RF2 replication forks illustrates that the red double strands are crossed over whereas the blue double strands are not. doi:10.1371/journal.pone.0006542.g007 to ensure visualisation of chromosome fragmentation in the light of the experience of Cromie and Leach [7].
When agarose embedded DNA was required to be digested by NotI, single plugs were equilibrated in 1 ml of the appropriate 1X restriction buffer for one hour at room temperature. Then, the buffer was replaced with 350 ml reaction buffer containing 30-50 units of restriction enzyme and incubated for 4 hours at 37uC. Following digestion, plugs were used immediately for PFGE.
Plugs were halved and placed on the comb of the PFG apparatus. 100 ml of 1% (w/v) high-strength agarose (AquaPor TM ) was made up fresh in 0.5X TBE and cooled to 55uC. 0.5 mg/ml of ethidium bromide was added and the agarose carefully poured around the plugs attached to the comb. The gel was left to set at 4uC and the same agarose solution used to pour the gel was pipetted into the gaps left by the comb. The gel was run in 0.5X TBE using CHEF-DR TM II (Biorad) PFGE equipment. PFGE was carried out using the following parameters: initial switch time 5 seconds; final switch time 30 seconds; run time 17 hours; voltage 5 V/cm and temperature 4uC. Gels were viewed using a UV trans-illuminator.