Bacillus subtilis RecG branch migration translocase is required for DNA repair and chromosomal segregation

The absence of Bacillus subtilis RecG branch migration translocase causes a defect in cell proliferation, renders cells very sensitive to DNA-damaging agents and increases approximately 150-fold the amount of non-partitioned chromosomes. Inactivation of recF, addA, recH, recV or recU increases both the sensitivity to DNA-damaging agents and the chromosomal segregation defect of recG mutants. Deletion of recS or recN gene partially suppresses cell proliferation, DNA repair and segregation defects of DeltarecG cells, whereas deletion of recA only partially suppresses the segregation defect of DeltarecG cells. Deletion of recG and ripX render cells with very poor viability, extremely sensitive to DNA-damaging agents, and with a drastic segregation defect. After exposure to mitomycin C recG or ripX cells show a drastic defect in chromosome partitioning (approximately 40% of the cells), and this defect is even larger (approximately 60% of the cells) in recG ripX cells. Taken together, these data indicate that: (i) RecG defines a new epistatic group (eta), (ii) RecG is required for proper chromosomal segregation even in the presence of other proteins that process and resolve Holliday junctions, and (iii) different avenues could process Holliday junctions.     

The structure of Bacillus subtilis RecU Holliday junction resolvase and its role in substrate selection and sequence-specific cleavage

We have determined the structure of the enzyme RecU from Bacillus subtilis, that is the general Holliday junction resolving enzyme in Gram-positive bacteria. The enzyme fold reveals a striking similarity to a class of resolvase enzymes found in archaeal sources and members of the type II restriction endonuclease family to which they are related. The structure confirms the presence of active sites formed around clusters of acidic residues that we have also shown to bind divalent cations. Mutagenesis data presented here support the key role of certain residues. The RecU structure suggests a basis for Holliday junction selectivity and suggests how sequence-specific cleavage might be achieved. Models for a resolvase-DNA complex address how the enzyme might organize junctions into an approximately 4-fold symmetric form.     

Structure, flexibility, and mechanism of the Bacillus stearothermophilus RecU Holliday junction resolvase

Here we report a high resolution structure of RecU-Holliday junction resolvase from Bacillus stearothermophilus. The functional unit of RecU is a homodimer that contains a "mushroom" like structure with a rigid cap and two highly flexible loops extending outwards. These loops appear to be highly flexible/dynamic, and presumably are directly involved in DNA binding and holding it for catalysis. Structural modifications of both the protein and DNA upon their interaction are essential for catalysis. An Mg2+ ion is present in each of the two active sites in this homodimeric enzyme, and two water molecules are coordinated with each Mg2+ ion. Our data are consistent with one of these water molecules acting as a nucleophile and the other as a general acid. The identities of the general base and general acid involved in catalysis and the Lewis acid that stabilizes the pentacovalent transition state phosphate ion are proposed. A model for the RecU-Holliday junction DNA complex is also proposed and discussed in the context of DNA binding and cleavage.     

The stalk region of the RecU resolvase is essential for Holliday junction recognition and distortion

The Bacillus subtilis RecU protein has two activities: to recognize, distort, and cleave four-stranded recombination intermediates and to modulate RecA activities. The RecU structure shows a mushroom-like appearance, with a cap and a stalk region. The RuvB interaction and the catalytic residues are located in the cap region of dimeric RecU. We report here that the stalk region is essential not only for RecA modulation but also for Holliday junction (HJ) recognition. Two recU mutants, which map in the stalk region, were isolated and characterized. In vivo, a RecU variant with a Phe81-to-Ala substitution (F81A) was as sensitive to DNA-damaging agents as a null recU strain, and a similar substitution at tyrosine 80 (Y80A) showed an intermediate phenotype. RecUY80A and RecUF81A poorly recognize and distort HJs. RecUY80A cleaves HJs with low efficiency, and RuvB modulates cleavage. At high concentrations, RecUF81A binds to HJs but fails to cleave them. Unlike wild-type RecU, RecUY80A and RecUF81A do not inhibit RecA dATPase and strand-exchange activities. The RecU stalk region is involved in RecA interaction, but once an HJ is bound, RecU fails to modulate RecA activities. Our biochemical study provides a mechanistic basis for the connections between these two mutually exclusive stages (i.e., RecA modulation and HJ resolution) of the recombination reaction.     

DNA double strand break repair and crossing over mediated by RuvABC resolvase and RecG translocase

DNA double-strand breaks threaten the stability of the genome, and yet are induced deliberately during meiosis in order to provoke homologous recombination and generate the crossovers needed to promote faithful chromosome transmission. Crossovers are secured via biased resolution of the double Holliday junction intermediates formed when both ends of the broken chromosome engage an intact homologue. To investigate whether the enzymes catalysing branch migration and resolution of Holliday junctions are directed to favour production of either crossover or noncrossover products, we engineered a genetic system based on DNA breakage induced by the I-SceI endonuclease to detect analogous exchanges in Escherichia coli where the enzymology of recombination is more fully understood. Analysis of the recombinants generated revealed approximately equal numbers of crossover and noncrossover products, regardless of whether repair is mediated via RecG, RuvABC, or the RusA resolvase. We conclude that there little or no control of crossing over at the level of Holliday junction resolution. Thus, if similar resolvases function during meiosis, additional factors must act to bias recombination in favour of crossover progeny.     

DNA repair defects ascribed to pby1 are caused by disruption of Holliday junction resolvase Mus81-Mms4

PBY1 continues to be linked with DNA repair through functional genomics studies in yeast. Using the yeast knockout (YKO) strain collection, high-throughput genetic interaction screens have identified a large set of negative interactions between PBY1 and genes involved in genome stability. In drug sensitivity screens, the YKO collection pby1Δ strain exhibits a sensitivity profile typical for genes involved in DNA replication and repair. We show that these findings are not related to loss of Pby1. On the basis of genetic interaction profile similarity, we pinpoint disruption of Holliday junction resolvase Mus81-Mms4 as the mutation responsible for DNA repair phenotypes currently ascribed to pby1. The finding that Pby1 is not a DNA repair factor reconciles discrepancies in the data available for PBY1, and indirectly supports a role for Pby1 in mRNA metabolism. Data that has been collected using the YKO collection pby1Δ strain confirms and expands the chemical-genetic interactome of MUS81-MMS4.     

The Holliday junction resolvase SpCCE1 prevents mitochondrial DNA aggregation in Schizosaccharomyces pombe

SpCCE1 (YDC2) from Schizosaccharomyces pombe is a DNA structure-specific endonuclease that resolves Holliday junctions in vitro. To investigate the in vivo function of SpCCE1 we made an Spcce1::ura4 ⁺ insertion mutant strain. This strain is viable and, despite being devoid of the Holliday junction resolvase activity that is readily detected in fractionated extracts from wild-type cells, exhibits normal levels of UV sensitivity and spontaneous or UV-induced mitotic recombination. In accordance with the absence of a nuclear phenotype, we show by fluorescence microscopy that a SpCCE1-GFP fusion localises exclusively to the mitochondria of S. pombe. In Saccharomyces cerevisiae the homologue of SpCCE1, CCE1, is known to function in the mitochondria where its role appears to be to remove recombination junctions and thus facilitate mitochondrial DNA segregation. A similar function can probably be attributed to SpCCE1 in S. pombe, since the majority of mitochondrial DNA from the Spcce1::ura4 ⁺ strain is in an aggregated form apparently due to extensive interlinking of DNA molecules by recombination junctions. Surprisingly, this marked effect on the conformation of mitochondrial DNA results in little or no effect on proliferation or viability of the Spcce1::ura4 ⁺ strain. Possible explanations are discussed. Received: 28 October 1999 / Accepted: 28 March 2000     

A pivotal role for the structure of the Holliday junction in DNA branch migration.

Branch migration of a DNA Holliday junction is a key step in genetic recombination that affects the extent of transfer of genetic information between homologous DNA sequences. We previously observed that the rate of spontaneous branch migration is exceedingly sensitive to metal ions and postulated that the structure of the cross-over point might be one critical determinant of the rate of branch migration. Other investigators have shown that in the presence of divalent metal ions like magnesium, the Holliday junction assumes a folded conformation in which base stacking is retained through the cross-over point. This base stacking is disrupted in the absence of magnesium. Here we measure the rate of branch migration as a function of Mg2+ concentration. The rate of branch migration increases dramatically at MgCl2 concentrations below 500 microM, with the steepest acceleration occurring between 300 and 100 microM MgCl2. This increase in the rate of branch migration coincides with the loss of base stacking in the four-way junction over this same interval of magnesium concentration, as measured by the susceptibility of junction residues to modification by osmium tetroxide and diethyl pyrocarbonate. We conclude that at physiological concentrations of intracellular Mg2+, base stacking in the Holliday junction constitutes one kinetic barrier to branch migration and that disruption of base stacking at the cross-over relieves this constraint.     

Direct evidence for spontaneous branch migration in antiparallel DNA Holliday junctions

The Holliday junction is a central intermediate in genetic recombination. It contains four strands of DNA that are paired into four double helical arms flanking a branch point. In naturally occurring Holliday junctions, the sequence flanking the branch point contains 2-fold (homologous) symmetry. As a consequence of this symmetry, the junction can undergo a conformational isomerization known as branch migration, which relocates the site of branching. In the absence of proteins and in the presence of Mg(2+), the four arms are known to stack in pairs, forming two helical domains whose orientations are antiparallel. Nevertheless, the mechanistic models proposed for branch migration are all predicated on a parallel alignment of helical domains. Here, we have used antiparallel DNA double crossover molecules to demonstrate that branch migration can occur in antiparallel Holliday junctions. We have constructed a DNA double crossover molecule with three crossover points. Two adjacent branch points in this molecule are flanked by symmetric sequences. The symmetric crossover points are held immobile by the third crossover point, which is flanked by asymmetric sequences. Restriction of the helices that connect the immobile junction to the symmetric junctions releases this constraint. The restricted molecule undergoes branch migration, even though it is constrained to an antiparallel conformation.     

Conformational isomerization of the Holliday junction associated with a cruciform during branch migration in supercoiled plasmid DNA

The variable positions of a branch-migrating cruciform junction in supercoiled plasmid DNA were mapped following cleavage of the DNA with bacteriophage T7 endonuclease I. T7 endonuclease I specifically cleaved, and thereby resolved, the Holliday junction existing at the base of the cruciform in the circular bacterial plasmid pSA1B.56A. Cruciform extrusion of cloned sequences in pSA1B.56A (containing a 322 base-pair inverted repeat insert composed of poxvirus telomeric sequences) topologically relaxed the plasmid substrate in vitro. Thus, numerous crossover positions were identified within the region of cloned sequences, reflecting the range of superhelical densities in the native plasmid preparation. Endonuclease I-sensitive crossover positions, mapped to both strands of the viral insert following the T7 endonuclease I digestion of either plasmid preparations or individual topoisomers, were regularly separated by approximately ten nucleotides. The appearance of sensitive crossovers every ten nucleotides corresponds to a change in linking difference (delta Lk) of +/- 2 in the circular core domain of the plasmid during branch point migration. In contrast, individual topoisomers of a plasmid preparation differ in linking number in increments of +/- 1. Thus, the observed linearization of each individual topoisomer following enzyme treatment, as a result of resolution of the crossovers associated with each topoisomer, showed that branch point migration to sensitive crossover positions must have occurred facilely. T7 endonuclease I randomly resolved across either axis of the cruciform, though some discrimination (related to the sequence specificity of the enzyme) was observed. The ten-nucleotide spacing between sensitive crossover positions is accounted for by an isomerization of the cruciform junction on branch point migration. An hypothesis is that this isomerization was imposed upon the cruciform junction by the change in helix twist (delta Tw) in the two branches that compose the topologically closed, circular domain of the plasmid. T7 endonuclease I may discriminate between the various isomeric forms and cleave a sensitive conformation that appears with every turn of branch migration which leads to the extrusion, or absorption, of two turns of helix from the circular core.     

Interplay between Bacillus subtilis RecD2 and the RecG or RuvAB helicase in recombinational repair

Bacillus subtilis AddAB, RecS, RecQ, PcrA, HelD, DinG, RecG, RuvAB, PriA and RecD2 are genuine recombinational repair enzymes, but the biological role of RecD2 is poorly defined. A ΔrecD2 mutation sensitizes cells to DNA-damaging agents that stall or collapse replication forks. We found that this ΔrecD2 mutation impaired growth, and that a mutation in the pcrA gene (pcrA596) relieved this phenotype. The ΔrecD2 mutation was not epistatic to ΔaddAB, ΔrecQ, ΔrecS, ΔhelD, pcrA596 and ΔdinG, but epistatic to recA. Specific RecD2 degradation caused unviability in the absence of RecG or RuvAB, but not on cells lacking RecU. These findings show that there is notable interplay between RecD2 and RecG or RuvAB at arrested replication forks, rather than involvement in processing Holliday junctions during canonical double strand break repair. We propose that there is a trade-off for efficient genome duplication, and that recombinational DNA helicases directly or indirectly provide the cell with the means to tolerate chromosome segregation failures.     

Structure, dynamics, and branch migration of a DNA Holliday junction: a single-molecule fluorescence and modeling study

The Holliday junction (HJ) is a central intermediate of various genetic processes, including homologous and site-specific DNA recombination and DNA replication. Elucidating the structure and dynamics of HJs provides the basis for understanding the molecular mechanisms of these genetic processes. Our previous single-molecule fluorescence studies led to a model according to which branch migration is a stepwise process consisting of consecutive migration and folding steps. These data led us to the conclusion that one hop can be more than 1 basepair (bp); moreover, we hypothesized that continuous runs over the entire sequence homology (5 bp) can occur. Direct measurements of the dependence of the fluorescence resonance energy transfer (FRET) value on the donor-acceptor (D-A) distance are required to justify this model and are the major goal of this article. To accomplish this goal, we performed single-molecule FRET experiments with a set of six immobile HJ molecules with varying numbers of bps between fluorescent dyes placed on opposite arms. The designs were made in such a way that the distances between the donor and acceptor were equal to the distances between the dyes formed upon 1-bp migration hops of a HJ having 10-bp homology. Using these designs, we confirmed our previous hypothesis that the migration of the junction can be measured with bp accuracy. Moreover, the FRET values determined for each acceptor-donor separation corresponded very well to the values for the steps on the FRET time trajectories, suggesting that each step corresponds to the migration of the branch at a defined depth. We used the dependence of the FRET value on the D-A distance to measure directly the size for each step on the FRET time trajectories. These data showed that one hop is not necessarily 1 bp. The junction is able to migrate over several bps, detected as one hop and confirming our model. The D-A distances extracted from the FRET properties of the immobile junctions formed the basis for modeling the HJ structures. The composite data fit a partially opened, side-by-side model with adjacent double-helical arms slightly kinked at the four-way junction and the junction as a whole adopting a global X-shaped form that mimics the coaxially stacked-X structure implicated in previous solution studies.     

The E.coli RuvAB proteins branch migrate Holliday junctions through heterologous DNA sequences in a reaction facilitated by SSB.

During genetic recombination a heteroduplex joint is formed between two homologous DNA molecules. The heteroduplex joint plays an important role in recombination since it accommodates sequence heterogeneities (mismatches, insertions or deletions) that lead to genetic variation. Two Escherichia coli proteins, RuvA and RuvB, promote the formation of heteroduplex DNA by catalysing the branch migration of crossovers, or Holliday junctions, which link recombining chromosomes. We show that RuvA and RuvB can promote branch migration through 1800 bp of heterologous DNA, in a reaction facilitated by the presence of E.coli single-stranded DNA binding (SSB) protein. Reaction intermediates, containing unpaired heteroduplex regions bound by SSB, were directly visualized by electron microscopy. In the absence of SSB, or when SSB was replaced by a single-strand binding protein from bacteriophage T4 (gene 32 protein), only limited heterologous branch migration was observed. These results show that the RuvAB proteins, which are induced as part of the SOS response to DNA damage, allow genetic recombination and the recombinational repair of DNA to occur in the presence of extensive lengths of heterology.     

The AtRAD51C gene is required for normal meiotic chromosome synapsis and double-stranded break repair in Arabidopsis

Meiotic prophase I is a complex process involving homologous chromosome (homolog) pairing, synapsis, and recombination. The budding yeast (Saccharomyces cerevisiae) RAD51 gene is known to be important for recombination and DNA repair in the mitotic cell cycle. In addition, RAD51 is required for meiosis and its Arabidopsis (Arabidopsis thaliana) ortholog is important for normal meiotic homolog pairing, synapsis, and repair of double-stranded breaks. In vertebrate cell cultures, the RAD51 paralog RAD51C is also important for mitotic homologous recombination and maintenance of genome integrity. However, the function of RAD51C in meiosis is not well understood. Here we describe the identification and analysis of a mutation in the Arabidopsis RAD51C ortholog, AtRAD51C. Although the atrad51c-1 mutant has normal vegetative and flower development and has no detectable abnormality in mitosis, it is completely male and female sterile. During early meiosis, homologous chromosomes in atrad51c-1 fail to undergo synapsis and become severely fragmented. In addition, analysis of the atrad51c-1 atspo11-1 double mutant showed that fragmentation was nearly completely suppressed by the atspo11-1 mutation, indicating that the fragmentation largely represents a defect in processing double-stranded breaks generated by AtSPO11-1. Fluorescence in situ hybridization experiments suggest that homolog juxtaposition might also be abnormal in atrad51c-1 meiocytes. These results demonstrate that AtRAD51C is essential for normal meiosis and is probably required for homologous synapsis.     

The role of Holliday junction resolvases in the repair of spontaneous and induced DNA damage

DNA double-strand breaks (DSBs) and other lesions occur frequently during cell growth and in meiosis. These are often repaired by homologous recombination (HR). HR may result in the formation of DNA structures called Holliday junctions (HJs), which need to be resolved to allow chromosome segregation. Whereas HJs are present in most HR events in meiosis, it has been proposed that in vegetative cells most HR events occur through intermediates lacking HJs. A recent screen in yeast has shown HJ resolution activity for a protein called Yen1, in addition to the previously known Mus81/Mms4 complex. Yeast strains deleted for both YEN1 and MMS4 show a reduction in growth rate, and are very sensitive to DNA-damaging agents. In addition, we investigate the genetic interaction of yen1 and mms4 with mutants defective in different repair pathways. We find that in the absence of Yen1 and Mms4 deletion of RAD1 or RAD52 have no further effect, whereas additional sensitivity is seen if RAD51 is deleted. Finally, we show that yeast cells are unable to carry out meiosis in the absence of both resolvases. Our results show that both Yen1 and Mms4/Mus81 play important (although not identical) roles during vegetative growth and in meiosis.     

Bacillus subtilis RecO and SsbA are crucial for RecA-mediated recombinational DNA repair

Genetic data have revealed that the absence of Bacillus subtilis RecO and one of the end-processing avenues (AddAB or RecJ) renders cells as sensitive to DNA damaging agents as the null recA, suggesting that both end-resection pathways require RecO for recombination. RecA, in the rATP·Mg²⁺ bound form (RecA·ATP), is inactive to catalyze DNA recombination between linear double-stranded (ds) DNA and naked complementary circular single-stranded (ss) DNA. We showed that RecA·ATP could not nucleate and/or polymerize on SsbA·ssDNA or SsbB·ssDNA complexes. RecA·ATP nucleates and polymerizes on RecO·ssDNA·SsbA complexes more efficiently than on RecO·ssDNA·SsbB complexes. Limiting SsbA concentrations were sufficient to stimulate RecA·ATP assembly on the RecO·ssDNA·SsbB complexes. RecO and SsbA are necessary and sufficient to ‘activate’ RecA·ATP to catalyze DNA strand exchange, whereas the AddAB complex, RecO alone or in concert with SsbB was not sufficient. In presence of AddAB, RecO and SsbA are still necessary for efficient RecA·ATP-mediated three-strand exchange recombination. Based on genetic and biochemical data, we proposed that SsbA and RecO (or SsbA, RecO and RecR in vivo) are crucial for RecA activation for both, AddAB and RecJ–RecQ (RecS) recombinational repair pathways.     

Bacillus subtilis RecA and its accessory factors,RecF, RecO,RecR and RecX,are required for spore resistance to DNA double-strand break

Bacillus subtilis RecA is important for spore resistance to DNA damage, even though spores contain a single non-replicating genome. We report that inactivation of RecA or its accessory factors, RecF, RecO, RecR and RecX, drastically reduce survival of mature dormant spores to ultrahigh vacuum desiccation and ionizing radiation that induce single strand (ss) DNA nicks and double-strand breaks (DSBs). The presence of non-cleavable LexA renders spores less sensitive to DSBs, and spores impaired in DSB recognition or end-processing show sensitivities to X-rays similar to wild-type. In vitro RecA cannot compete with SsbA for nucleation onto ssDNA in the presence of ATP. RecO is sufficient, at least in vitro, to overcome SsbA inhibition and stimulate RecA polymerization on SsbA-coated ssDNA. In the presence of SsbA, RecA slightly affects DNA replication in vitro, but addition of RecO facilitates RecA-mediated inhibition of DNA synthesis. We propose that repairing of the DNA lesions generates a replication stress to germinating spores, and the RecA·ssDNA filament might act by preventing potentially dangerous forms of DNA repair occurring during replication. RecA might stabilize a stalled fork or prevent or promote dissolution of reversed forks rather than its cleavage that should require end-processing.     

The ACF1 complex is required for DNA double-strand break repair in human cells

DNA double-strand breaks (DSBs) are repaired via?nonhomologous end-joining (NHEJ) or homologous?recombination (HR), but cellular repair processes remain elusive. We show here that the ATP-dependent chromatin-remodeling factors, ACF1 and SNF2H, accumulate rapidly at DSBs and are required for DSB repair in human cells. If the expression of ACF1 or SNF2H is suppressed, cells become extremely sensitive to X-rays and chemical treatments producing DSBs, and DSBs remain unrepaired. ACF1 interacts directly with KU70 and is required for the accumulation of KU proteins at DSBs. The KU70/80 complex becomes physically more associated with the chromatin-remodeling factors of the CHRAC complex, which includes ACF1, SNF2H, CHRAC15, and CHRAC17, after treatments producing DSBs. Furthermore, the frequency of NHEJ as well as HR induced by DSBs in chromosomal DNA is significantly decreased in cells depleted of either of these factors. Thus, ACF1 and its complexes play important roles in DSBs repair.     

XRCC1 is required for DNA single-strand break repair in human cells

The X-ray repair cross complementing 1 (XRCC1) protein is required for viability and efficient repair of DNA single-strand breaks (SSBs) in rodents. XRCC1-deficient mouse or hamster cells are hypersensitive to DNA damaging agents generating SSBs and display genetic instability after such DNA damage. The presence of certain polymorphisms in the human XRCC1 gene has been associated with altered cancer risk, but the role of XRCC1 in SSB repair (SSBR) in human cells is poorly defined. To elucidate this role, we used RNA interference to modulate XRCC1 protein levels in human cell lines. A reduction in XRCC1 protein levels resulted in decreased SSBR capacity as measured by the comet assay and intracellular NAD(P)H levels, hypersensitivity to the cell killing effects of the DNA damaging agents methyl methanesulfonate (MMS), hydrogen peroxide and ionizing radiation and enhanced formation of micronuclei following exposure to MMS. Lowered XRCC1 protein levels were also associated with a significant delay in S-phase progression after exposure to MMS. These data clearly demonstrate that XRCC1 is required for efficient SSBR and genomic stability in human cells.     

The Bacillus subtilis chromatin-associated protein Hbsu is involved in DNA repair and recombination

The Bacillus subtilis hbs gene encodes an essential chromatin-associated protein termed Hbsu. Hbsu, the counterpart of the Escherichia coli HU protein, binds DNA in a non-specific way but has a clear preference for bent, kinked or altered DNA sequences. To investigate the role of Hbsu in DNA repair and DNA recombination we have constructed a series of site-directed mutants in the hbs gene and used these mutant genes to substitute the wild-type chromosomal hbs gene. The hbs47 mutation, which codes for a mutant protein in which residue Phe-47 has been replaced by Trp, does not cause any discernible phenotype. Additional substitution of residue Arg-55 by Ala (hbs4755 mutation) rendered cells deficient in DNA repair, homologous recombination and (i protein-mediated site-specific recombination. We have also tested the effect on DNA repair of the hbs4755 mutation in combination with mutations in different functions of homologous DNA recombination (recA, recF, recG, recti and addAB). The hbs4755 mutation did not modify the sensitivity of recH and addAB cells to the DNA-damaging agents methylmethane sulphonate (MMS) or 4-nitroquinoline-1-oxide (4NQO), and it only marginally affected recF and recG cells. The hbs4755 mutation blocked intermolecular recombination in recH cells and markedly reduced it (20- to 50-fold) in recF and recG cells, but had no effect on addAB cells. Taken together, these data indicate that the Hbsu protein is required for DNA repair and for homologous DNA recombination.