Focus on recombinational DNA repair

Workshop on Recombinational DNA Repair and its Links with DNA Replication and Chromosome Maintenance
Lorraine S. Symington

Author Affiliations

  • Lorraine S. Symington, 1 Department of Microbiology, Columbia University Medical Center, 701 W. 168th Street, New York, New York, USA

The Juan March workshop on ‘Recombinational DNA Repair and its Links with DNA Replication and Chromosome Maintenance’ was held in Madrid, Spain, between 13 and 15 December 2004. The workshop was organized by S. Kowalczykowski, S. West, A. Aguilera and J. Alonso.

Embedded Image


Homologous recombination (HR) has a crucial role in chromosome maintenance by promoting error‐free repair of DNA double‐strand breaks (DSBs) and single‐strand gaps. DNA strand breaks occur spontaneously during normal cellular processes, such as DNA replication, or when cells are treated with DNA damaging agents, and are essential intermediates in several programmed genome rearrangements. HR is also required for telomere maintenance in the absence of telomerase in eukaryotes. In Escherichia coli, most of the genes involved in HR were identified in screens for recombination‐defective mutants (rec genes). Similarly, screens for radiation‐sensitive mutants in budding yeast led to the discovery of the RAD52 epistasis group genes that define the main pathway for HR in eukaryotes. The RecA and Rad51 recombinases are central to the process of HR in prokaryotes and eukaryotes, respectively. Recombinases promote homologous pairing and strand exchange, which leads to recombination between the two interacting DNA molecules (Fig 1). However, the recombinases require several other proteins to accomplish strand exchange in vivo (Kowalczykowski, 2000; Krogh & Symington, 2004; West, 2003). The role of these accessory factors in facilitating the activity of the respective recombinase was one of the topics covered at the meeting. HR and nonhomologous end joining (NHEJ) are the main pathways for repairing DSBs in eukaryotic cells, but there are distinct sub‐pathways depending on the presence of one or two ends at the break site, or flanking homologies surrounding the DSB. The regulation of these pathways of DSB repair (DSBR) is an active area of investigation, as shown by the talks at this meeting.

Figure 1.

Models for recombinational repair of DNA double‐strand breaks in eukaryotes. The Mre11 complex is recruited to double‐strand breaks (DSBs) and may control 5′–3′ resection along with exonuclease 1 (Exo1). The resulting single‐stranded DNA (ssDNA) tails are coated by replication protein A (RPA), which removes secondary structures from the ssDNA. Radiation‐repair protein 52 (Rad52) recruits Rad51 to the RPA‐coated ssDNA, resulting in displacement of RPA and formation of the Rad51 nucleoprotein filament that is subsequently stabilized by the Rad51 paralogues. Rad54 facilitates a late step in homologous pairing with the donor duplex (grey lines), and the 3′ end of the strand invasion intermediate is extended by DNA synthesis. The invading end can be displaced from the donor duplex, possibly by the Sgs1/BLM (Bloom syndrome) helicase, and can then anneal to the other side of the break via the newly synthesized DNA. After gap filling and ligation, DSB repair is complete and there is no change to the donor duplex. The alternative fate of the strand‐exchange intermediate involves capture of the other side of the DSB by the displaced strand of the donor duplex and formation of a double Holliday junction (HJ) intermediate. This structure can be unwound by the Sgs1:Top3 (topoisomerase 3) complex to yield non‐crossover products, or resolved by a HJ resolvase to generate crossover or non‐crossover products.

It is now widely recognized that replication forks frequently collapse and that HR is crucial for origin‐independent replication restart. This meeting brought together investigators who use genetic, cell biological and biochemical approaches to unravel the mechanisms of HR, and investigators who study DNA replication with a view to understanding how HR solves replication problems to maintain genome integrity. The key issues discussed at the meeting were the causes and consequences of replication fork stalling, and how cells re‐establish replication after dissociation of the replisome at stalled forks.

Recruitment of proteins to double‐strand breaks

A large number of genes are required for DSBR and their roles in this process are being further defined by sophisticated genetic and biochemical assays. The use of chromatin immunoprecipitation and live imaging of green fluorescent protein (GFP)‐tagged proteins has provided new insights into the assembly of recombination proteins at DSBs (Fig 2). By marking a chromosome break site with GFP, R. Rothstein (New York, NY, USA) showed co‐localization of several recombination proteins with the site‐specific DSB. Genetic dependencies and kinetic analysis of individual factors suggest an ordered assembly with the Mre11 complex (Mre11, Rad50 and Xrs2) arriving first, followed by replication protein A (RPA), Rad52, Rad51, Rad55 and then Rad54 (Lisby et al, 2004; Sugawara et al, 2003). The Mre11 complex is thought to tether DNA ends together and remove end‐blocking lesions by Mre11 nuclease activity, and to signal through the telomere length 1 (Tel1)/ataxia‐telangiectasia mutated (ATM) kinase to downstream effectors (Chen et al, 2001; de Jager et al, 2001; Krogh & Symington, 2004). Biochemical studies had suggested an early role for Rad54 in stabilizing the Rad51 filament, as well as later steps during homologous pairing (Mazin et al, 2003; Petukhova et al, 1998). However, studies by J. Haber (Waltham, MA, USA) using yeast strains, in which the two recombining sites were marked with GFP, showed efficient but protracted synapsis of the recipient and donor sites in rad54 mutants. This indicates that the essential function of Rad54 is late in recombination after homologous pairing has occurred. J. Alonso (Madrid, Spain) presented evidence for the ordered assembly of recombination proteins at DSBs to form discrete foci in Bacillus subtilis (Kidane et al, 2004). RecN, a component of the RecF pathway that is homologous to the structural maintenance of chromosomes (SMC) proteins, binds to ssDNA and is recruited to DSBs early, before RecF or RecA. This early recruitment of RecN is similar to the Mre11 complex, which raises the possibility that RecN tethers ends, as suggested for the Mre11 complex.

Figure 2.

Co‐localization of recombination proteins with double‐strand breaks. The upper panels show yellow fluorescent protein (YFP)–Rad51 foci in unirradiated and irradiated cells; the lower panels show differential interference contrast (DIC) images. Examples of spontaneous (S‐phase dependent) and ionizing radiation (IR)‐induced foci are indicated by the white arrowheads (C. Fung and L.S.S., unpublished data)

The repair of DNA lesions in vivo occurs in the context of chromatin, and there is considerable interest in the role of chromatin components and remodelling activities in DNA repair. M. Lichten (Bethesda, MD, USA) showed that the phosphorylated form of histone H2A in yeast (referred to as γ‐H2AX in human cells) is recruited rapidly to DSBs, and localizes to large chromosomal domains that flank the DSB, rather than at sequences that are immediately adjacent to the DSB (Shroff et al, 2004). By contrast, Mre11 and Rad51 localize to sequences that immediately flank the DSB. The failure to detect γ‐H2AX at sequences immediately adjacent to the break does not appear to be due to a lack of nucleosomes or to extensive resection, which suggests active chromatin remodelling around the break site. S. Gasser (Basel, Switzerland) showed recruitment of the Ino80 chromatin‐remodelling complex to DSBs and, in contrast to γ‐H2AX, recruitment was strongest close to the break site (van Attikum et al, 2004). Recruitment of the Ino80 complex is dependent on γ‐H2AX, is Rad52‐independent and appears to regulate resection of the DSB, which suggests that chromatin remodelling by this complex is important for end resection. Mutations in individual subunits of the Ino80 complex do not have as strong a DNA‐repair defect as the rad52 group mutants, which suggests that this step is not essential for repair or that there could be redundancy between the Ino80 complex and other chromatin‐remodelling complexes.

Sister‐chromatid interactions increase HR efficiency

The cohesin and condensin complexes include members of the SMC family of ATPases and are essential for sister‐chromatid cohesion and chromosome condensation to ensure proper segregation of chromosomes during cell division. T. Hirano (Cold Spring Harbor, NY, USA) discussed the role of ATP hydrolysis in the DNA‐binding cycle of SMC proteins (Hirano & Hirano, 2004). Lichten presented evidence that the cohesin complex co‐localizes with γ‐H2AX at DSBs and loading requires H2A phosphorylation and Mre11 (Unal et al, 2004). Lichten hypothesized that the cohesin‐binding sites, which are located about 40–50 kb apart in unperturbed cells, may be insufficient to support sister‐chromatid alignment for DSBR, and the additional cohesin loading may help to keep the ends of the DSB in close enough proximity for efficient repair. Although cohesion is necessary for efficient sister‐chromatid repair, it is not sufficient. Genetic studies indicate that the Mre11 complex (Mre11, Rad50 and Xrs2/Nbs1) is essential for DSB repair between sister chromatids. J. Petrini (New York, NY, USA) presented elegant genetic studies that showed the importance of Rad50 dimerization for the DNA repair, telomere maintenance and meiotic DSB functions of the Mre11 complex. The implication is that one function of the Mre11 complex is to tether DNA ends through the Rad50 hook domain to the sister chromatid for efficient DNA repair.

Assembly of the RecA/Rad51 nucleoprotein filament

A universal requirement for HR is loading of the recombinase onto ssDNA. Binding of the ssDNA‐binding proteins, single‐stranded binding protein (SSB) or RPA, to ssDNA is an impediment to recombinase loading, which necessitates factors, often referred to as mediators, to facilitate loading of the recombinase onto SSB/RPA‐coated ssDNA (Fig 1). S. Kowalczykowski (Davis, CA, USA) presented evidence that the E. coli RecF, RecO and RecR proteins promote loading of RecA onto the 5′ junction of gapped DNA coated with SSB (Morimatsu & Kowalczykowski, 2003). This nucleation event results in the directional extension of RecA onto the ssDNA. The role of Rad52 in mediating the assembly of the Rad51 nucleoprotein filament was discussed by P. Sung (New Haven, CT, USA). The carboxy‐terminal region, which contains the Rad51 interaction domain, is known to be required for Rad52 mediator function. However, Sung showed that the C‐terminal domain alone is sufficient to promote this function in vitro. The role of the Rad51 paralogues (Rad55 and Rad57 in yeast; Rad51B, Rad51C, Rad51D, Xrcc2 (X‐ray cross complementing) and Xrcc3 in mammals) during HR remains controversial. L. Symington (New York, NY, USA) showed additive suppression of the radiation sensitivity of a rad57 null mutant by a rad51 gain‐of‐function allele that results in more stable binding of Rad51 to DNA (Fortin & Symington, 2002) and by deletion of SRS2, which encodes a DNA helicase that disrupts Rad51 filaments in vitro. This result suggests that the primary function of Rad55 and Rad57 is to nucleate or stabilize the Rad51 filament. Overexpression of Rad51 partially suppresses the DNA‐repair defect of vertebrate cell lines that are defective for the Rad51 paralogues, indicating that the mediator function of the Rad51 paralogues is conserved, but may not be their only function. S. West (London, UK) presented evidence for an additional late function for the mammalian Rad51 paralogues in the resolution of Holliday junctions (Liu et al, 2004).

HR occurs at high frequency during meiosis, and in most eukaryotes requires Rad51 and its meiosis‐specific homologue, Dmc1. There are additional factors induced during meiosis that function with Dmc1 and Rad51 to promote efficient DSBR, including the conserved homologous pairing 2 (Hop2) and meiotic nuclear divisions 1 (Mnd1) proteins (Petukhova et al, 2003). D. Camerini‐Otero (Bethesda, MD, USA) showed that Hop1 and Mnd1 interact to form a stable heterodimer that stimulates Rad51 or Dmc1‐promoted strand invasion.

Pathway choice during double‐strand break repair

DSBs can be repaired by HR, NHEJ or by single‐strand annealing (SSA) if repeated sequences flank the break site, but how cells choose between these pathways is unclear. One of the first steps in HR is the resection of the 5′ ends at the break site. Haber and M. Foiani (Milan, Italy) showed that the 5′–3′ resection of DSBs in yeast is cell‐cycle‐regulated with little resection occurring in G1, compared with asynchronous or G2 cells (Ira et al, 2004). The block to resection in G1 cells is due to cyclin‐dependent kinase (CDK) activity and results in greater efficiency of end‐joining repair of the DSB. M. Jasin (New York, NY, USA) has shown that HR or NHEJ efficiently repairs a chromosomal DSB, but when two DSBs are made on different chromosomes adjacent to Alu repeats, repair occurs by SSA or NHEJ, which results in chromosome translocations (Elliott et al, 2005). The choice between these pathways depends on the level of identity between the Alu repeats, with SSA predominating for identical repeats. To study homology‐dependent repair from one side of a DSB, Haber used model substrates in which only one end of a DSB generated by HO endonuclease has homology to the donor sequences. This forces cells to repair the break by extensive DNA synthesis that is initiated at the site of strand invasion and terminates at the telomere, a process known as break‐induced replication (BIR). Surprisingly, the initial primer extension step that follows strand invasion and leads to BIR is much slower than the primer extension that is associated with gene conversion (Malkova et al, 2005). Haber suggested that this regulation could be to prevent BIR at a DSB, which could result in the loss of heterozygosity for downstream markers.

Regulation of DNA repair

CDK activity is important for cell‐cycle progression, and recent studies indicate that CDKs regulate recombination. As described above, resection of the DSBs in G1‐arrested cells is regulated by CDK activity, and this promotes NHEJ repair in cells that lack a sister chromatid to template HR. West presented evidence that CDK‐dependent phosphorylation of the C‐terminal domain of the breast cancer susceptibility protein BRCA2 regulates the interaction with Rad51. Phosphorylation of BRCA2 is low in S‐phase, but increases during mitosis and prevents interaction with Rad51. However, treatment with ionizing radiation reduced phosphorylation and therefore restored the interaction between BRCA2 and Rad51, presumably to increase the efficiency of HR in cells with DNA damage (Esashi et al, 2005). The DNA‐damage checkpoint is important for delaying cell division and allowing time for repair to occur, but may also function more directly in the repair process. W. Heyer (Davis, CA, USA) showed that DNA‐damage‐induced phosphorylation of yeast Rad55, and a rad55‐SA phosphorylation site mutant was sensitive to genome‐wide genotoxic stress induced by methyl methanesulphonate (MMS).

What happens at a stalled replication fork?

The causes and consequences of replication fork stalling was one of the main topics of discussion at the meeting. B. Michel (Jouy en Josas, France) used conditional replication mutants that increase the frequency of fork stalling to delineate the pathways for origin‐independent replication fork restart (Michel et al, 2004). Replication fork stalling in E. coli replisome mutants results in a reaction called fork reversal in which the nascent strands pair to form a four‐way junction with a DNA double‐strand end (Fig 3). The resultant end is usually degraded by RecBCD for primasome assembly (Pri) protein‐dependent restart at the fork, but in the absence of RecBCD the junction is cleaved by the RuvC Holliday junction nuclease, which results in linearization of the E. coli chromosome. Surprisingly, Michel has found that the genetic requirements for fork reversal are different depending on which part of the replication apparatus is perturbed. Fork reversal in mutants that lack the DnaB replicative helicase (dnaB) is recA‐dependent and uvrD‐helicase‐independent, whereas fork reversal in polymerase (Pol) III holoenzyme‐defective mutants requires UvrD. Given this complexity for replication mutants, it seems likely that diverse pathways are used for replication restart depending on how the fork stalls—for example, by damage on the template strands or collision with a transcribing RNA polymerase. If the replisome disassembles at stalled forks, then reassembly is required and reloading of DnaB is thought to be the crucial step. Genetic studies indicate two pathways for origin‐independent replisome assembly, one of which is dependent on PriA, and the other on PriC. K. Marians (New York, NY, USA) reconstituted these two pathways in vitro and showed that the PriA pathway loads DnaB onto the lagging strand when the nascent leading strand terminates at the fork junction, whereas PriC is required to load DnaB onto the lagging strand when there is a gap on the leading strand (Heller & Marians, 2005).

Figure 3.

Role of recombination and lesion bypass in replication fork restart. The panel on the left illustrates replication fork collapse when the advancing fork encounters a nick on the template strand. The broken arm invades the homologous duplex and creates a new template for priming DNA synthesis. DNA damage on the leading strand stalls DNA synthesis (middle panel). If the replisome dissociates, the nascent strands can pair, which results in regression of the fork. The nascent lagging strand can then act as a template for the leading strand, and branch migration of the reversed fork results in bypass of the lesion. Because the lesion is in duplex DNA again after fork regression, it could also be repaired by excision repair. Alternatively, the regressed fork is cleaved by an HJ resolvase and then has to undergo strand invasion to restart the collapsed fork. The lesion on the template strand can also be bypassed by one of the lesion bypass polymerases. DNA damage on the lagging strand results in formation of ssDNA gap. This gap could be filled by a lesion bypass polymerase, or by strand exchange mediated by a recombinase.

Reversed forks have only been detected in yeast cells that were treated with hydroxyurea (HU) in the absence of the checkpoint kinase Rad53/Chk2 (Sogo et al, 2002). This suggests that replisome dissociation is rare in wild‐type cells and that one important function of the checkpoint is to maintain the replisome at stalled forks. Foiani suggested that regressed forks could give rise to sister‐chromatid fusions at telomeres, thus contributing to the genetic instability in checkpoint mutants. Foiani presented evidence that the Exo1 exonuclease associates with stalled forks and resects newly synthesized strands to counteract fork reversal in the rad53 mutant (Cotta‐Ramusino et al, 2005). Gasser analysed replisome stability by retention of polymerases at stalled forks in the presence of HU and found significant reductions in polymerase α and polymerase ϵ in the absence of the ATR (ataxia‐telangiectasia and Rad3‐related)/Mec1 (yeast homologue of ATR) kinase or the Sgs1 helicase (Bloom syndrome (BLM) helicase in humans; Bjergbaek et al, 2005). sgs1 mutants are known to have increased genome instability, which could be due to disassembly of the replisome and subsequent inappropriate recombination.

Recombinogenic structures at stalled forks

High levels of transcription have been shown to stimulate HR, and one suggested mechanism for this stimulation is replication fork stalling caused by collision of the replication fork with RNA polymerase. To test this model, A. Aguilera (Sevilla, Spain) constructed a direct repeat recombination reporter downstream of a strong promoter, in two orientations with respect to a replication origin (Prado & Aguilera, 2005). When transcription was directed towards the origin, high levels of recombination were observed and two‐dimensional gel analysis revealed a replication fork block. By contrast, no block was observed when transcription proceeded in the same direction as the fork and recombination frequencies were reduced. It is unclear whether the high levels of recombination that result from stalled forks are initiated by ssDNA or a DSB that results from fork collapse.

DNA damage on the template strand at stalled forks can be bypassed by recombination or by translesion polymerases (Fig 3). Because lesion bypass using translesion polymerases is potentially mutagenic, there is likely to be a hierarchy in the assembly of HR proteins and polymerases at forks that are arrested by DNA damage. F. Hanaoka (Osaka, Japan) discussed the substrate specificity for Polη and suggested that Polη and the Rev1 deoxycytidyltransferase are sequentially recruited to damaged replication forks through protein‐protein interactions. Foiani described the accumulation of X‐shaped DNA structures in sgs1 mutants when grown in the presence of MMS (Liberi et al, 2005). Although formation of these intermediates is Rad51‐dependent, they do not appear to be Holliday junctions and instead may be multiply catenated structures that are formed by template switching. These structures are not detected in wild‐type cells, even when treated with MMS, which suggests that Sgs1 rapidly resolves these intermediates and that their accumulation in sgs1 cells probably contributes to the observed genetic instability.

The chemotherapeutic drug camptothecin traps the covalent linkage between topoisomerase 1 (Topo1) and DNA, which forms a ssDNA break that can be converted to a DSB during S‐phase. Studies by K. Caldecott (Brighton, UK) showed that camptothecin causes strand breaks in nondividing cells that require the combined activity of tyrosyl phosphodiesterase (TDP1) and Xrcc1 for repair. It is possible that redundant end‐processing pathways function to repair camptothecin‐induced lesions in dividing cells, but no redundant pathway exists in nondividing cells. This results in spinocerebellar ataxia with axonal neuropathy (SCAN1), a severe neuronal defect in cells that lack TDP1.

Concluding remarks

This is an exciting time in the recombination field. Although many components are known, the identity of two key factors in eukaryotic HR, the nuclease that processes DSBs to generate 3′ ssDNA tailed intermediates and the Holliday junction resolvase, remain unknown. Understanding how recombination proteins function at the level of chromatin and how they are regulated post‐translationally to control protein interactions, cellular distribution and turnover is still in its infancy. There are many other unresolved questions, such as the events that lead to fork stalling, the structure of stalled forks and the appropriate use of HR to restart replication. These questions are likely to keep this growing field occupied for years to come.


I thank the Juan March Foundation and the organizers for the excellent and timely workshop. I thank the speakers for allowing their work to be cited and regret that space limitations prevented coverage of all the exceptional work discussed at the conference. Studies in the author's lab are supported by the National Institute of General Medical Sciences.