OUP user menu

Double-strand break repair in bacteria: a view from Bacillus subtilis

Silvia Ayora, Begoña Carrasco, Paula P. Cárdenas, Carolina E. César, Cristina Cañas, Tribhuwan Yadav, Chiara Marchisone, Juan C. Alonso
DOI: http://dx.doi.org/10.1111/j.1574-6976.2011.00272.x 1055-1081 First published online: 1 November 2011


In all living organisms, the response to double-strand breaks (DSBs) is critical for the maintenance of chromosome integrity. Homologous recombination (HR), which utilizes a homologous template to prime DNA synthesis and to restore genetic information lost at the DNA break site, is a complex multistep response. In Bacillus subtilis, this response can be subdivided into five general acts: (1) recognition of the break site(s) and formation of a repair center (RC), which enables cells to commit to HR; (2) end-processing of the broken end(s) by different avenues to generate a 3′-tailed duplex and RecN-mediated DSB ‘coordination’; (3) loading of RecA onto single-strand DNA at the RecN-induced RC and concomitant DNA strand exchange; (4) branch migration and resolution, or dissolution, of the recombination intermediates, and replication restart, followed by (5) disassembly of the recombination apparatus formed at the dynamic RC and segregation of sister chromosomes. When HR is impaired or an intact homologous template is not available, error-prone nonhomologous end-joining directly rejoins the two broken ends by ligation. In this review, we examine the functions that are known to contribute to DNA DSB repair in B. subtilis, and compare their properties with those of other bacterial phyla.

  • double-strand break
  • recombination functions
  • protein–protein interaction
  • protein localization
  • endprocessing
  • NHEJ


Faithful replication and maintenance of chromosome integrity and genetic variation are opposite forces that shape evolution. Genomes face constant challenges that can lead to DNA damage, from the intracellular genotoxic stress that arises as a result of normal cellular metabolism (e.g. free radicals, DNA replication errors) to a wide range of exogenous sources [e.g. ionizing radiation (IR), etc.]. DNA damage, which does not compromise nucleoid integrity, is removed by a set of dedicated error-free pathways (e.g. base excision repair, nucleotide excision repair, mismatch repair, etc.), each one specific for a given type or a group of lesions. These pathways have been reviewed previously (Sancar & Reardon, 2004; Friedberg, 2006; Iyer, 2006; Sanchez, 2007a; Ambur, 2009) and are beyond the scope of this review.

Chromosome replication is not a continuous process; it can be impaired by obstacles (as DNA-bound proteins) by ‘inactivation’ of a replication protein and/or by the presence of unrepaired DNA lesions or abortive repair of DNA by the dedicated error-free pathways (reviewed by Mirkin & Mirkin, 2007). Replication arrest can have dramatic consequences and replication defects are now recognized as a major source of genomic instability (Kuzminov, 1999; Cromie & Leach, 2000; Michel, 2004; Wyman, 2004; Kreuzer, 2005; Aguilera & Gomez-Gonzalez, 2008; Branzei & Foiani, 2008). Homologous recombination (HR) plays a central role in the re-establishment of these stalled or collapsed replication forks. In bacteria, DNA damage that does not compromise nucleoid integrity leads to replication fork stalling, which triggers the synthesis of a set of DnaA-, LexA- and RecA-dependent genes involved in DNA repair and metabolism, and increases the repair of single-strand (ss) DNA gaps via HR. RecA-mediated daughter-strand gap repair and DNA replication restart, which was one of the first proposed mechanisms for recombinational DNA repair in Escherichia coli (known as postreplication repair), has been reviewed recently and will not be discussed further (see Kogoma, 1997; Kuzminov, 1999; Cox, 2000; McGlynn & Lloyd, 2002a, b; Courcelle & Hanawalt, 2003; Marians, 2004; Michel, 2004; Kreuzer, 2005; Persky & Lovett, 2008). Those ssDNA gaps that fail to be repaired are ultimately converted into one-ended double strand breaks (DSBs) (Kuzminov, 1999; Cromie & Leach, 2000; Michel, 2004; Wyman, 2004; Kreuzer, 2005).

Different models have been proposed to explain how one-ended DSBs are generated and repaired in E. coli cells (Kuzminov, 1999; McGlynn & Lloyd, 2002b; Michel, 2007; Persky & Lovett, 2008). These proposals cannot be easily extrapolated to other bacteria, because the phenotypes observed in cells lacking some of the protein factors involved vary between different bacteria and are in some cases contradictory. Conspicuously, while the absence of both branch migration translocases, RecG and RuvAB, is synthetically lethal in Neisseria gonorrhoeae (representative of Betaproteobacteria class) or Bacillus subtilis (Firmicutes phylum) (Sanchez, 2005, 2007b, c; Sechman, 2006), it shows a synergistic defect in DNA repair in E. coli cells (Gammaproteobacteria class), but leads to a wild-type (wt)-like phenotype in Helicobacter pylori cells (Epsilonproteobacteria class) (Kang & Blaser, 2008; Rudolph, 2010). Inactivation of H. pylori RecG (RecGHpy), which does not show a significant defect in DNA repair, suppresses the recombination defect of the ruvBHpy mutations (Kang & Blaser, 2008), which suggests that in this bacterium, they may have opposing functions. This review is aimed at comparing the mechanisms used by bacteria to repair DSBs. In particular, it will focus on the recent cytological and biochemical information of HR proteins from Firmicutes.

Faithful repair of DSBs requires the recognition and processing of DSB ends and the interaction with an intact homologous DNA to produce genetic recombinants. The repair of one- and two-ended DSBs differs in the number of ends recruited into the repair reaction. The repair of multiended DSBs generated upon high doses of IR might follow a different and specialized avenue and hence it will be analyzed separately. To define a global bacterial response to DNA damage and the role of recombination proteins in one- or two-ended DSB repair, a search for the panrecombinosome, defined as a full and minimum set of genes required for recombinational DNA repair, and recombination-dependent replication restart, was undertaken. No panrecombinosome has been identified to date, in any reciprocal pairwise analyses, from >900 different bacterial genomes sequenced (identity cutoff of ∼25%, which is sufficient to indicate a structural homology; Sander & Schneider, 1991). A minimal bacterial panrecombination machinery emerges (Table 1) when endosymbiotic species are removed from the analysis and recombination functions are grouped into three distinct categories: those involved in the presentation and generation of the DNA substrate (presynapsis), the ones involved in the formation of heteroduplex DNA intermediates performed by the recombinase and its modulation (synapsis) and those involved in the generation of matured recombinant products (postsynapsis) (Table 1). Remarkably, this machinery is also conserved in large double-strand (ds) DNA viruses (Table 1). Such a broad distribution of recombination functions could result as a consequence of lateral gene transmission, perhaps involving bacterial viruses. For example, bacteria of the Firmicutes phylum, the earliest-branching bacterial phyla (Ciccarelli, 2006), encode a Holliday junction (HJ) resolvase of the RecU type, which belongs to the same superfamily as archaeal Hjc or virus T7 Endo I (with a PvuII-like fold), while non-Firmicutes bacteria encode a functional analogue, the RuvC type enzyme (with an RNase H fold) (Declais & Lilley, 2008; Ayora, 2011). However, despite the structural difference, B. subtilis RecU can complement the defect of E. coli ruvC (ruvCEco) cells (Sanchez, 2005, Table 1).

View this table:
Table 1

Panrecombinosome for DSB repair in bacteria and bacteriophages

Recombination stageBacterial productsViral products
PresynapticRecN, PNPase, 5′→3′ Exo (AddAB-like or RecJ-RecQ-like or RecJ-UvrD-like), SSB/SsbA5′→3′ Exo, Ssb-like
SynapticRecO, RecR, RecA, PcrA/UvrDSynaptase (UvsX, SSAP)
PostsynapticRuvAB, RecG, HJ resolvaseHJ Resolvase (e.g. RusA)
  • * A 5′→3′ exonuclease–helicase as the members of the AddAB-like family: AddA-AddB (AddAB), RecB-RecC-RecD (RecBCD) or AdnA-AdnB (AdnAB) (see Rigden, 2005; Unciuleac & Shuman, 2010) or a 5′→3′ ssDNA exonuclease in concert with a DNA helicase (as RecJ in concert with a RecQ- or a UvrD-like helicase) (see text).

  • The dash denotes the concerted action of more than one product, and forward slash equivalent functions.

  • A 5′→3′ exonuclease (Exo-like), as phage λ-encoded Redα exonuclease, that is specialized to work in concert with a recombinase (e.g. Redβ). There are two different classes of synaptases (recombinase), the ATP-dependent synaptases (e.g. bacterial RecA or T4-encoded UvsX) and the ATP-independent single-strand annealing protein (SSAP), as phage λ-encoded Redβ, and their mediators/modulators. The HJ resolvases either have an endonuclease-like fold (as Bacillus subtilis RecU or T7 Endo I), an RNaseH-like (Escherichia coli RuvC), a RusA-like or an HNH endonuclease fold (as T4 Endo VII) (Lilley & White, 2000; Sharples, 2001; Declais & Lilley, 2008).

In E. coli, which is considered the bacterial paradigm for the analysis of DNA repair via HR, mechanisms of daughter strand gap and DSBs repair and the re-establishment of the replication fork have been reviewed extensively (Kogoma, 1997; Kuzminov, 1999; Cox, 2000; Cromie & Leach, 2000; McGlynn & Lloyd, 2002b; Courcelle & Hanawalt, 2003; Marians, 2004; Michel, 2004; Kreuzer, 2005; Persky & Lovett, 2008). There is growing evidence, however, that not all bacteria use the same DNA repair functions or the same proteins to re-establish replication forks (Ambur, 2009). Indeed, in some bacteria (e.g. B. subtilis and Mycobacterium tuberculosis), two-ended DSB repair relies not only on error-free homology-directed HR but also on a simplified nonhomologous end-joining (NHEJ) mechanism (Bowater & Doherty, 2006; Pitcher, 2007; Shuman & Glickman, 2007; Ayora, 2011). We will discuss the different avenues that B. subtilis utilizes for the repair of DSB, comparing them with those of distantly related bacteria. Unless otherwise stated, the genes and products mentioned herein are of B. subtilis origin. The E. coli model will only be discussed when providing essential and different information. The comparison between the repair processes of evolutionary diverging bacteria may be particularly informative and will help us to identify the common repertoires of HR mechanisms in bacteria.

Different pathways vs. concerted action of functions during DSB repair in bacteria

A DSB is the most lethal type of DNA damage. In wt E. coli cells, DNA break processing and loading of RecA onto 3′-ssDNA ends is catalyzed mainly by the ‘RecBCD pathway’, while the ‘RecF or RecFOR pathway’ is normally required for daughter strand gap repair (Clark & Sandler, 1994; Kowalczykowski, 1994; Kuzminov, 1999; Amundsen & Smith, 2003; Michel, 2004). Such a division of labor between these pathways has considerably helped us to understand the DNA recombination mechanisms in E. coli, but such a simplified analysis cannot be assumed for other bacteria even of the same class. The E. coli RecFOR pathway can also process DSBs in a recB recC background upon activation by mutations on the SbcB and SbcCD nucleases (Clark & Sandler, 1994; Kowalczykowski, 1994; Kuzminov, 1999; Amundsen & Smith, 2003; Michel, 2004). In bacteria other than those of the Gammaproteobacteria class, avenues other than RecBCD play an active role in end processing. In fact, it has been shown that the SbcB (also termed ExoI, SbcB/ExoI) 3′→5′ ssDNA exonuclease is restricted to the Gammaproteobacteria and to some species of Betaproteobacteria, suggesting that RecFOR ‘activation’ through inactivation of SbcB/ExoI and SbcCD nucleases might not be a general requirement. Additionally, in some bacteria, the absence of the RecBCD pathway together with some RecFOR pathway factors (e.g. RecF) does not completely block DSB repair via HR (Rocha, 2005; Han, 2006), suggesting that other functions might perform this role. Hence, we favor a scenario where a highly complex multistep repair response is finely orchestrated towards different avenues by multiple redundant functions. They can be stimulated by factors such as the nature of the DNA substrate, the confinement of the homologous template and/or the number of chromosomal copies. Based on the data obtained in B. subtilis, this multistep repair response to DSBs can be arbitrarily divided into five distinct stages (Table 2): (1) recognition of DNA damage and increase of the local concentration of 3′-ssDNA ends; (2) end processing, which in turn is subdivided into two further steps: first, basal processing of the DNA ends and putative licensing of diverse DNA repair avenues, and second, long-range processing to give rise to a 3′-tail, which is bound by RecA to produce a RecA-ssDNA filament; (3) RecA-mediated homology search of multiple DNA ends at once in a crowded intracellular milieu, to generate the displacement loop (D-loop) recombination intermediate, necessary for replication fork resumption; (4) strand exchange processing, which can pursue three different avenues: helicase-dependent branch migration and resolvase-dependent processing of single or double HJs; topoisomerase-dependent decatenation and synthesis-dependent strand annealing (SDSA); and (5) faithful segregation of chromosomes. Many functions involved in these stages are conserved from bacteria to humans (Table 2).

View this table:
Table 2

Proteins involved in DSB repair via HR

Recombination stagesBacteriaEukaryotes
E. coliB. subtilisH. sapiens
(a) Damage recognitionRecN, PNPase?RecN, PNPaseMRN-CtIP
(b) End processingRecBCDAddAB
DSB coordinationRecN, PNPase?RecN, PNPaseMRN-CtIP
(c) Recombinase loadingRecBCDAddAB??
RecORRecO (R?)RAD52
Mediator/modulators proteinsRecFORRecFORRAD51BD-XRCC2
?HepA, YqhH?RAD54-RAD54B
(d) Replication fork re-establishmentPriAB-DnaTPriA-DnaDB?
PriC-PriA or RepPriA, PrcA??
Branch migration and HJ resolution protein(s)RecGRecG?
RecQ-Topo IIIRecQ/S-Topo III?BLM-TopoIIIα-RMI1
(e) Chromosomal segregationXerCD-FtsKRipX-CodV-SpoIIIE/SftA?
  • * In budding yeast, Mre11, Rad50 and Xrs2 (MRX) in concert with Sae2.

  • In the recBCsbcBsbcC background, RecQ and RecJ are involved in end processing.

  • Any of the two RecQ-like helicases (RecS and RecQ) in concert with RecJ process DNA ends in an otherwise wt B. subtilis strain.

  • § The human equivalents to budding yeast functions Sgs1-Dna2-top3-Rmi1 [Correction added 26 May 2011 after online publication: ‘fission’ has been changed to ‘budding’ in this sentence]. A question mark denotes postulated or unknown activity. A dash denotes different components of a given complex. A forward slash denotes alternative proteins

  • Any of the UvrD-like helicases (RecD2, PcrA/UvrD), in concert with RecJ, might process DNA ends.

Functional relationship and mechanistic differences between bacterial HR proteins

Genetic screenings of B. subtilis mutants that show decreased survival in response to DNA-damaging agents and/or altered rates of genetic recombination have led to the identification of at least 40 nonessential and ∼10 essential gene products involved in DSB repair and re-establishment of the replication fork. Recently, the conservation of these functions among different bacterial phyla has been reviewed (Ayora, 2011). Although most DSB repair proteins in B. subtilis work exclusively in either HR or NHEJ, some participate in both pathways (Alonso, 1988, 1991, 1993a, b; Ayora, 1996; Fernández, 1997, 1998, 1999; Sanchez, 2005, 2006, 2007b). In all free-living organisms, a dedicated recombinase (bacterial RecA, archaeal RadA or eukaryotic Rad51 or DMC1) catalyzes DNA strand invasion and strand exchange, as a central step in HR, making impossible the classification of the recA function into an epistatic group (see Fig. 1). Genes, other than recA, which are involved in HR, have been placed into 10 different epistatic groups. In addition, there are many essential (pcrA, smc-scpA-scpB, ssbA, dnaB, dnaD, topA, gyrAB, parCE, hbs/hupAB, etc.) and nonessential (cbfA/yhaM, helD, lrpC, mfd, mutS2, pnpA, priA, rarA, recD2, recK, polA, ripX-codV/xerCD/xerS], sms/radA, topB, ypcP, etc.) genes involved in HR, which are still to be classified (Fig. 1).

Figure 1

Schematic grouping of Bacillus subtilis RecA-dependent DNA recombinational repair genes into different epistatic groups. Because a recA mutation is epistatic with any representative mutation in the different epistatic groups (α–κ), it was placed in the center. The set of as yet unclassified essential or nonessential genes also involved in RecA-dependent or RecA-independent DNA repair are presented.

The complexity of DSB repair in bacteria is such that a precise molecular mechanism cannot be proposed for all bacteria. Indeed, there are genes that are essential in a given bacterial class, but dispensable in other classes. For example, within the Bacilli class, albeit with some exceptions, there are two paralogues with 5′→3′ exonuclease activity, DNA polymerase I (DNAP I, polA gene) and YpcP (ypcP), whose absence leads to synthetic lethality (Fukushima, 2007): in bacteria of the Clostridia class, only the polA gene is found, while bacteria of the Mollicutes class encode only the ypcP gene. The homodimeric chromatin-binding protein Hbsu is essential among bacteria of the Firmicutes phylum, but the heterodimeric HU is nonessential within bacteria of the Gammaproteobacteria class (Fernández, 1997). Additionally, an individual protein may have different biochemical activities in different bacterial genera. For example, RecR, of Firmicutes and Deinococcus-Thermus phyla, binds ssDNA and dsDNA (Alonso, 1993a; Ayora, 1997a, b; Lee, 2004), but its E. coli counterpart fails to do so (Umezu & Kolodner, 1994; Webb, 1995). There are mechanistic differences in the end-processing step catalyzed by the AdnAB, AddAB or RecBCD helicase–nuclease complex (Niu, 2009). Few bacteria, including those of the Mycobacterium genus, encode multiple copies of the multisubunit helicase–nuclease complexes, while obligate endosymbionts lack every component of the multisubunit helicase–nuclease complex (Montague, 2009). Bacteria from different phyla encode a large RecD helicase, RecD2 (Rocha, 2005), and a recD2 mutation is nonepistatic with the addA addB mutations, at least in bacteria of the Firmicutes phylum (C. E. César and C. Marchisone, personal communication). Finally, there are several recombinational repair proteins that, although conserved, seem to play a different role in different bacterial classes. As an example, the absence of the Sms/RadA protein, which plays an unknown, but active role in the stabilization and/or processing of branched DNA molecules (Carrasco, 2002, 2004), partially suppresses the DNA repair and segregation defect of B. subtilis ruvAB or recG cells (Carrasco, 2004; Sanchez, 2007b). By contrast, in E. coli cells, a mutation in the sms/radA gene shows a strong synergism with mutations in recG, and to a lesser extent, with defects in ruvA and ruvC. Furthermore, the absence of Sms/RadAEco enhances the conjugative recombination deficiency conferred by mutations in recGEco, ruvAEco and ruvCEco (Beam, 2002), whereas the absence of Sms/RadA in Firmicutes decreases the transformation frequency of otherwise wt cells (Carrasco, 2002; Claverys, 2009).

DNA damage response

DNA damage halts DNA replication and alters, in a DnaA-, LexA- and RecA-dependent manner, the expression of genes involved in DNA repair, among other processes, in order to increase the chances of survival (Sassanfar & Roberts, 1990; Messer, 2002; Friedberg, 2006; Erill, 2007; Patel, 2010). Not all bacteria, however, respond to DNA damage in an identical manner. This induction of DNA repair can proceed by different regulatory networks or by a combination of them, namely SOS response, oxidative stress, iron homeostasis, etc. (Courcelle, 2001; Zheng, 2001; Khil & Camerini-Otero, 2002; Sebastian, 2002; Campoy, 2003; Porwollik, 2003; Grifantini, 2004; Stohl, 2005; Sweetman, 2005; Jin, 2007; Jordan & Saunders, 2009). In some species of the Betaproteobacteria class (e.g. N. gonorrhoeae), the autoproteolysis of the LexA orthologue requires RecA in response to nonoxidative DNA damaging agents, but recA expression is not upregulated following treatment with different DNA damage-inducing agents (Schook, 2011). Furthermore, LexA is absent in bacteria of the Bacteroidetes and Chlorobi phyla, in the Epsilonproteobacteria class and in some species of the Bacilli and Mollicutes classes (Erill, 2007).

UV radiation generates ssDNA regions at stalled replication forks, to which RecA polymerizes, stimulating the LexA self-cleavage reaction that triggers the LexA-dependent response (Little, 1991; Cox, 2007b; Patel, 2010). In E. coli or B. subtilis, cleavage of LexA, which acts as a transcriptional repressor, increases the expression of roughly 30–50 SOS genes, including the recA and lexA genes (Fernandez de Henestrosa, 2000; Courcelle, 2001; Khil & Camerini-Otero, 2002; Au, 2005; Goranov, 2006; Erill, 2007). However, the genes induced are poorly or not conserved; from the primary recombinational repair genes, only lexA and recA gene induction is conserved among bacteria with a genuine SOS response (Erill, 2007). The SOS response has a cost, because a second DNA molecule and LexA share the same binding site within the RecA filament so that SOS response induction competes with RecA-mediated recombination (Harmon, 1996). These two RecA activities are indirectly regulated by ubiquitously present modulators (PcrA/UvrD, RecO, RecR, RecF and RecX) and by other less conserved ones (HelD, RecU, DinI and RdgC) that control the dynamics of the RecA·ssDNA filament (Cox, 2007a, b; Ayora, 2011).

When a DNA lesion collapses the replication fork compromising the nucleoid integrity, the DNA ends are resected and RecA is loaded onto the SsbA-coated ssDNA tails, resulting in altered expression of >500 genes in a RecA-dependent manner, from which <50 are directly regulated by LexA both in E. coli and in B. subtilis (Khil & Camerini-Otero, 2002; Goranov, 2006). Indeed, two-ended DSBs induced by IR cause a global SOS induction in virtually every E. coli cell, but only in a small fraction of B. subtilis cells (Simmons, 2009). Unlike E. coli cells (Sargentini & Smith, 1986), B. subtilis cells incapable of inducing the SOS response have nearly wt levels of survival rate in response to IR (Simmons, 2009). The roles of other stress-response transcriptional regulators, as the one recently described in N. gonorrhoeae, in the repair of DSBs are poorly characterized (Schook, 2011).

Mechanisms of DNA DSBs repair

As we have already discussed, DSBs arise when a replication fork is collapsed (one-ended DSBs, Figs 2a and 3a) or directly through the action of DNA damaging agents, such as IR, that break the phosphate backbone, leading to two-ended (Figs 2b and 3b) or multiple-ended DSBs (McGlynn & Lloyd, 2002b; Courcelle & Hanawalt, 2003; Michel, 2004; Wyman, 2004). Both in eukaryotes and in prokaryotes, the information required for the repair of DSBs is readily available in the form of a an intact homologous template (McGlynn & Lloyd, 2002b; Courcelle & Hanawalt, 2003; McEachern & Haber, 2006; Llorente, 2008). To repair DSBs, the cell reacts by activating the synthesis and/or increasing the local concentration of recombination proteins, which are physically recruited to the DNA lesion site, in a step-wise manner. The recombination machinery has to bring the needed homologous sequence from the same DNA molecule (one-ended DSB) or from a different DNA molecule (two-ended DSBs) (Kidane, 2004; Lisby & Rothstein, 2004).

Figure 2

Early steps of DSB repair in Bacillus subtilis. (a) A physical or a chemical agent introduces an ssDNA nick that collapses the replication fork, leading to one-ended DSB. This can be repaired by HR at all stages of the cell cycle. (i) The RecN protein, in concert with PNPase, recognizes damaged ends and promotes basal end processing. If the ends already have an ssDNA region, SsbA binds to them. A RecN ‘backup system’, SbcEF, is present in some bacteria. (ii) The long-range resection of the 5′-ends can be catalyzed by RecJ, in concert with a RecQ-like (RecQ or RecS) or a UvrD-like (PcrA, RecD2?) helicase, with SsbA/SSB stimulating the nuclease and helicase activities. Alternatively, long-range resection is performed by the AddAB nuclease–helicase complex (AdnAB, RecBCD), which, under some circumstances, might also load RecA directly onto naked ssDNA (e.g. RecBCDEco). (iii) RecO alone or the RecOR(F) complex promotes the disassembly of the SsbA protein and the loading of RecA onto SsbA-coated ssDNA. RecA-polymerized onto SsbA/SSB-coated ssDNA promotes the search for a homologous template. RecA modulators (RecF, RecX, PcrA/UvrD, RecU, etc.) control RecA by modulation of RecA filament extension. (b) A physical or a chemical agent (in yellow) introduces a two-ended DSB that can be repaired by HR at all stages of the cell cycle or by NHEJ that functions primarily if HR is impaired or during the stationary phase. (i) RecN, in concert with PNPase, recognizes damaged ends and tethers them, promoting commitment to HR. If the ends already have an ssDNA region, SsbA binds to them. Steps ii and iii are similar to one-ended DSB. (iv) In the absence of HR during vegetative growth or in the absence of a sister chromatid (stationary phase), the NHEJ system is activated. (v) Long-range resection is blocked and the DNA ends are tethered by the action of the homodimeric Ku protein (bridged green ovals). (vi) The ends are ligated by LigD or are processed before ligation by redundant pathways consisting of nucleases (SbcCD, PNPase or DNAP X) and a UvrD-like DNA helicase (PcrA). This step is performed by as yet poorly characterized functions. For simplicity, only the names of B. subtilis functions are described.

Figure 3

Late steps of DSB repair in Bacillus subtilis. (a) Repair of one-ended DSBs. RecA-mediated strand invasion leads to the formation of a D-loop intermediate. The invading strand primes DNA synthesis using the intact homologous chromosome as a template to restore the lost genetic information. The replisome assembly machinery reloads the replicative helicase in the lagging strand template. In steps i–iii, replication continues up to encountering the invading strand forming an extended D-loop or a nicked HJ-like structure, which might be disrupted by MutS2 (absent in Alpha-, Beta- and Gammaproteobacteria) or any uncharacterized endonuclease. In steps iv–vi, by the action of a branch migration translocase (RecG or RuvAB), the D-loop is partially regressed to form a HJ intermediate that is resolved to a fork structure by RecU/RuvC. (b) Repair of two-ended DSBs. The steps i and ii are shared by one- and two-ended DSB repair. (iii–v) Capture of the second DNA end, derived from the other end of the DSB, by RecO (RecOR) leads to a double HJ that is branch migrated by RuvAB and resolved by RecU. The role of the RecG branch migration translocase in RecU-mediated resolution of HJs is poorly understood. Differential cleavage of the double HJs, gap filling and ligation leads to CO or NCO products. The unwanted dimers (CO products) are disentangled by SMC-Topo IV and converted into monomers by a tyrosine site-specific recombinase RipX-CodV in concert with one or two translocases (SpoIIIE or SftA). The double HJ formed in step iii could also be processed by a Topo III enzyme, in concert with a RecQ-like or a UvrD-like helicase and SsbA, by branch migration and strand decatenation that dissolve the double HJs, producing only NCO products. The 3′-invading end, extended by the replisome, can be displaced from the joint molecule and reannealed with the complementary strand of the other resected end of the break (steps viii and ix). This repair mechanism, which is termed SDSA, produces only NCO products. This putative avenue is poorly characterized in Bacteria. For simplicity, only the names of B. subtilis functions are described.

To understand the order of events during HR after DSB formation in bacteria, different experimental approaches have been adopted. These involve high-throughput methods to detect the functional, evolutionary and structural relationships between groups of interacting E. coli proteins (Su, 2008), and high-throughput yeast two-hybrid screens, as those used with Campylobacter jejuni and B. subtilis species (Noirot & Noirot-Gros, 2004; Parrish, 2007). To dissect how recombination and repair proteins are recruited and how they function at a DSB in living B. subtilis cells, genes involved in DNA repair were fused with a gene coding for a fluorescent protein (Lindow, 2002; Mascarenhas, 2002; Kidane, 2004; Kidane & Graumann, 2005a; Sanchez, 2005, 2006; Mascarenhas, 2006; Wang, 2006; Lecointe, 2007). The fully functional fused genes were used to replace the wt genes on their natural loci in order to examine the temporal order of protein recruitment to discrete foci, after treatment with the DNA damaging agent mitomycin C (MMC) (Kidane, 2004; Kidane & Graumann, 2005a; Sanchez, 2005, 2006; Mascarenhas, 2006; Krishnamurthy, 2010). In an alternative approach, the fused gene was expressed from an inducible promoter while the wt gene was expressed from its natural locus (Lindow, 2002; Kidane & Graumann, 2005b; Simmons, 2007). Not all protein fusions were informative. These cytological studies can be used to choreograph DSB repair into five acts in B. subtilis: (1) damage-induced RecN foci, which are dynamic structures typically containing a high local concentration of RecN at the broken DNA ends, recognize the broken ends that might modulate the initial response to DNA damage, (2) long range end-processing at the break (generation of 3′-ssDNA ends) and DSB ‘coordination’ to form a discrete repair center (RC), (3) loading of the strand exchange protein, RecA, onto the 3′-ssDNA ends at the RC (Fig. 2), (4) RuvAB-mediated loading of the RecU HJ resolvase and (5) disassembly of the recombination apparatus and replication restart (Fig. 3). From these defined steps, the recognition of the break site might be common for HR and NHEJ, followed by the commitment to DNA repair by HR, before end processing (Fig. 2b).

Break recognition and initial response to DNA DSBs

Different bacteria have a different number of SMC-like proteins; four are found in B. subtilis (SMC, SbcC, SbcE and RecN). The absence of either SbcC or SbcE renders cells sensitive and SMC or RecN very sensitive to MMC, H2O2 or IR, but moderately sensitive to UV radiation or 4-nitroquinoline (mimics UV damage), suggesting that they are mainly involved in DSB repair (Sargentini & Smith, 1986; Alonso, 1993b; Funayama, 1999; Dervyn, 2004; Kidane, 2004; Kosa, 2004; Mascarenhas, 2006; Stohl & Seifert, 2006; Guthlein, 2008; Krishnamurthy, 2010), and implying a central role for the RecN and SMC proteins and, to a lesser extent, for SbcC and SbcE in DSB repair in bacteria. The phenotype varied between different bacterial species; for example N. gonorrhoeae cells are very sensitive to UV irradiation (Skaar, 2002), but E. coli recN cells are almost fully resistant to it (Picksley, 1984). Furthermore, the acute sensitivity of E. coli recN cells to bleomycin-induced DSBs was comparable to that of recA, implying a central role for the RecN protein in the repair of bleomycin lesions (Kosa, 2004).

Exponentially growing B. subtilis wt cells in minimal medium at 22 °C show one discrete SMC-YFP focus adjacent to the oriC in the large majority of the cells (∼75%) and one SbcC-GFP or SbcE-YFP focus per nucleoid in ∼2% or ∼9% of cells, respectively, but RecN-YFP foci formation is rare (<0.1% of the total cells) (Mascarenhas, 2002; Mascarenhas, 2006; Krishnamurthy, 2010). However, exponentially growing B. subtilis addA, recU and recA cells accumulate RecN-YFP foci in ∼2%, ∼5% and ∼30% of the total cells, respectively (Kidane, 2004; Sanchez, 2005, 2006), suggesting that in the absence of AddAB, RecU or RecA, the replication fork might be more susceptible to breakage, and consequently subjected to RecN recognition.

About 15 min after the generation of a site-specific DSB or random DSBs (after exposure to different doses of IR, nalidixic acid or MMC), nucleoids tend to fuse, with only ∼4% of cells containing two nucleoids (∼21% of exponentially growing cells contain two nucleoids). SMC-YFP, SbcC-GFP and SbcE-YFP remain unaltered, but RecN-YFP relocalizes from a diffuse distribution to a discrete focus (or RC) per nucleoid in ∼35% of the total wt cells (Kidane, 2004; Mascarenhas, 2006; Shintomi & Hirano, 2007; Krishnamurthy, 2010). About 30 min after the addition of MMC, the SOS response is turned on (Sassanfar & Roberts, 1990; Messer, 2002; Friedberg, 2006; Erill, 2007; Patel, 2010), SMC-YFP retains its regular subcellular localization, the number of cells bearing SbcC-GFP or SbcE-YFP foci increases to ∼5% and ∼14% of the cells, respectively, and RecN-YFP is recruited into damage-induced foci in ∼75% of cells (Kidane, 2004; Kidane & Graumann, 2005a).

In the current model, SMC, SbcC and SbcE play a role in the maintenance of chromatin structure in actively replicating wt cells. This is consistent with the observation that SbcC and SbcE interact with the DnaG primase, and SbcCEco with the replicase (Noirot & Noirot-Gros, 2004; Darmon, 2007), and with the fact that SMC is recruited to the region adjacent to oriC to enhance faithful segregation (Gruber & Errington, 2009). RecN, which is among the first responders to DSBs, senses the DNA damage and recruits multiple DNA ends to form a single RC, i.e. two or more two-ended DSBs, generated concurrently, diffuse to coexist in space and time. RecN, which works as a concentrator of 3′-ends, bound to each DSB, undergoes a ‘large-scale motion’ (DNA ends tethering) and works as a scaffold that loads repair proteins (Fig. 4).

Figure 4

Temporal order of protein assembly at DSBs in Bacillus subtilis. Chromosome replication is not a continuous process. SMC, SbcC and SbcE form discrete foci during exponential growth (stage i). Upon induction of random one- or two-ended DSBs or a single site-specific two-ended DSB (time 0), RecN acts as a sensor and, in concert with PNPase, assembles on DNA, forming few repair centers (RCs) (stage ii). After basal resection, the 5′ ends are further resected either by the nuclease(s)–helicase(s) AddAB complex or by the RecJ ssDNA exonuclease, in concert with a RecQ-like (RecQ and/or RecS) or a UrvD-like DNA helicase (RecD2?, PcrA) in association with SsbA (stage iii). Stages ii and iii are closely related in vivo. The long-range end resection leads to SsbA binding and protection of the ssDNA and to RecN tethering of the DNA ends, leading to a discrete RC (15–30 min) and ‘DSB coordination’. Stages iv–vi are visualized by the temporal RecO, RecR and RecA recruitment (30–45 min). These proteins are directly or indirectly recruited to the RecN-promoted RC. At a later stage, RecF (min 60, stage vi) is recruited to the RecN-mediated RC and RecX modulates RecA threads (as probably PcrA/UvrD). Replisome assembly via PriA, DnaD, DnaB and DnaI might take place during 60 and 90 min after DNA damage (stage vii). The recruitment of RecU to a recombination intermediate, in the presence of RuvAB, is visualized at min 100–110 after DNA damage (stage vii). Concomitant with stages ii–viii, other processes such as SMC maintenance of the chromosome condensation and SbcCD- and SbcEF-dependent repair take place (stage ix). Once DNA replication is ended and cell division begins, the RipX-CodV/XerCD/XerS tyrosine recombinase, in concert with SpoIIIA/FtsK or SftA, resolves the accumulated dimeric chromosomes (stage x). Growth resumption in minimal medium at 22°C takes place ∼180 min upon DNA damage in wt cells.

Does RecN recognize stalled forks? A shift of dnaA or dnaB thermosensitive mutant strains to nonpermissive temperature causes stalling of the replication fork, and ultimately, replication fork collapse in a small fraction of cells (Graumann & Knust, 2009). Upon this temperature shift, only a fraction of the total cells (∼15%) contain discrete RecN or RecA foci per nucleoid (Mascarenhas, 2006; Simmons, 2007; Krishnamurthy, 2010), whereas upon induction by MMC discrete RecN-YFP foci are observed in ∼75% of the cells within the first 30 min, with RecA-GFP foci starting to accumulate in the 30–45-min interval (Kidane, 2004; Kidane & Graumann, 2005a). A possible explanation to reconcile this apparent contradiction is that a block in replication, generated by the inactivation of DnaA or DnaB, leads to replication fork stalling in the large majority of cells and perhaps to one-ended DSBs in a small fraction, with both RecN and RecA forming foci in ∼15% of cells (Simmons, 2007). Both the inactivation of DnaA or DnaB and the addition of MMC cause discrete RecN-YFP foci, observed in ∼75% of cells, and RecA foci in ∼15% of cells within the first 30 min. DNA damage-induced RecA foci and threads form later (30–120 min; see Fig. 4) (Kidane & Graumann, 2005a; Sanchez, 2006; Krishnamurthy, 2010). It is likely that when DNA replication is blocked and DNA DSBs are induced, two different DNA repair events (DNA gaps and DSBs) overlap and RecN recognizes DSBs rather than DNA gaps.

The confirmation that RecN specifically recognizes DSBs comes from experiments showing that RecN colocalizes with site-specific two-ended DSB induced at a specific region, close to oriC or terC (Kidane & Graumann, 2005a, b; Mascarenhas, 2006). Indeed, RecN assembles preferentially and directly at the site of a DSB upon induction of DNA damage, rather than at the replication factory, thus making RecN a very good marker for one- and two-ended DSB and the formation of RCs (Kidane & Graumann, 2005a).

Does a RecN backup system exist? SbcC and SbcE form foci in exponentially growing cells, and foci formation increases slightly upon MMC addition, the former colocalizing with the replication factory, but not the latter (Mascarenhas, 2006; Krishnamurthy, 2010). However, DNA damage-induced SbcE foci were observed in 36% of recN or 26% of sbcC cells within 30 min, after MMC addition, when compared with 14% in wt cells (Krishnamurthy, 2010), suggesting that SbcE may work as a RecN or an SbcC backup system. Indeed, the recN and sbcE or sbcC and sbcE mutations are nonepistatic, but the role of the Firmicutes SbcCD and/or SbcEF complexes in HR remains to be characterized (Mascarenhas, 2006; Krishnamurthy, 2010).

DSB recognition and the initial response to DNA damage are poorly characterized processes in other bacteria. In E. coli, the RecN concentration is very tightly regulated and is maintained at very low levels (Neher, 2006) before SOS induction (Finch, 1985; Courcelle, 2001), suggesting that RecNEco cannot act before end-processing and SOS induction (Meddows, 2005). Alternatively, RecNEco can be induced by other stress systems. Indeed, overexpressed GFP-RecNEco is found to localize in discrete foci on the nucleoid after the induction of DSBs by IR (Nagashima, 2006).

Several differences also exist at the biochemical level. The insolubility of RecNEco has hampered the efforts to purify the protein and has limited the possibility to compare the role of RecN in DSB repair in the two distantly related E. coli and B. subtilis bacteria. In vitro, RecN and Deinococcus radiodurans RecN (RecNDra) hydrolyze ATP with similar turnover rates, and such rates are stimulated four- to sixfold by DNA. The RecN ATPase activity is stimulated by ssDNA, whereas RecNDra rates are stimulated by dsDNA (Sanchez & Alonso, 2005; Reyes, 2010). However, RecN of Haemophilus influenzae RecN (RecNHin) and Aquifex aeolicus (RecNAae) have a weak ATPase activity, but Bacteriodes fragilis (RecNBfr) fails to hydrolyze ATP (Grove, 2009).

RecNHin, RecNAae and RecNBfr fail to bind DNA in vitro (Grove, 2009). RecN binds to discrete ssDNA ends and tethers them to increase the local end concentration, leading to rosette-like structures formed by these protein·ssDNA complexes (Sanchez, 2008), RecNDra performs a similar function on dsDNA (Reyes, 2010). It is worth noting that RecADra binds preferentially to dsDNA (Kim, 2002), whereas RecA from other bacteria bind preferentially to ssDNA (Cox, 2007a; Galletto & Kowalczykowski, 2007). The preference of both RecADra and RecNDra for dsDNA may be relevant for the accomplishment of DSB repair in Deinococcus (Kim, 2002; Reyes, 2010). Furthermore, the large RecN·ssDNA complexes are insensitive to SsbA, but disrupted by RecA (Sanchez & Alonso, 2005; Sanchez, 2008). Genetic data, however, suggest that RecNEco protects ssDNA during conjugative recombination (Lloyd & Buckman, 1995) and RecN, RecNNgo or RecNHpy protects ssDNA during natural transformation (Lloyd & Buckman, 1995; Skaar, 2002; Kidane & Graumann, 2005b; Wang & Maier, 2008; Kidane, 2009). Indeed, RecN interaction with incoming ssDNA favors the localization of RecN at the pole containing the competence machinery (Kidane & Graumann, 2005b).

Two-ended DSB repair by HR or NHEJ

Archaea, eukaryotes and some bacteria have evolved two different avenues to repair two-ended DSBs: HR and NHEJ. These avenues are largely distinct from one another and function in a completely different way. HR is the predominant mechanism used by bacteria to repair one- or two-ended DSBs during the transient diploidy generated by replication, before cell division, or by natural diploids or polyploids as described in some free-living or symbiotic bacteria (Hansen, 1978; Kitten & Barbour, 1992; Komaki & Ishikawa, 1999; Tobiason & Seifert, 2006; Tobiason & Seifert, 2010). Some bacteria also have functions that allow them to repair two-ended DSBs via error-prone NHEJ. Commitment to NHEJ in bacteria is still a poorly understood process and its physiological importance remains to be addressed. In mammalian cells, 50–70% of the DSBs created by the I-SceI endonuclease are repaired by NHEJ, with the remaining fraction being repaired by HR (Liang, 1998). Indeed, two-ended DSBs are repaired mainly by NHEJ during the M (G0) and the G1 (G1) phases, but when sister chromatids are adjacent, as in the S and G2 (G2) phases, DSBs are actively repaired by HR (Pâques & Haber, 1999; Rothkamm, 2003; Harper & Elledge, 2007; Aguilera & Gomez-Gonzalez, 2008; Huertas & Jackson, 2009; Mimitou & Symington, 2009a, b). Several lines of evidence suggest that end resection is needed for all HR processes in all living organisms, and resected DNA ends decrease NHEJ efficiency. DNA tethering by eukaryotic Rad50 or bacterial RecN and the recruitment of Rad51 or RecA mediators is also crucial in the commitment to HR, because end-processing and Rad51 or RecA loading are necessary for HR, but inhibitory for NHEJ (Aylon, 2004; Ira, 2004; Ayora, 2011). In B. subtilis cells, two-ended DSBs created by random breaks or by site-specific incision through HO endonuclease are mainly repaired by HR (>99.9%) (Bowater & Doherty, 2006; Mascarenhas, 2006; Cardenas, 2009). However, in the presence of a single chromosome copy or when HR is impaired at early stages, a minor fraction of two-ended DSBs are repaired by NHEJ, utilizing little or no homology to ligate DNA ends (Bowater & Doherty, 2006; Mascarenhas, 2006; Shuman & Glickman, 2007; Cardenas, 2009; Lieber, 2010). This is consistent with experiments showing that exponentially growing B. subtilis null recA mutants (ΔrecA) are extremely sensitive to DSB-inducing DNA agents, Δku cells are only marginally sensitive, if at all, when compared with wt cells, but ΔkuΔrecA cells are more sensitive to DNA damage than ΔrecA cells (Mascarenhas, 2006; Cardenas, 2009). In stationary-phase cells or in germinating spores, however, where there is only one intact chromosome template, ΔrecA cells become moderately sensitive to DSBs, and the double ΔkuΔrecA or ΔligDΔrecA mutants are significantly more sensitive than ΔrecA cells (Weller, 2002; Wang, 2006; Moeller, 2007, 2008). It is likely that under certain circumstances, NHEJ may serve as a backup system for DSB repair in exponentially growing cells (Mascarenhas, 2005; Cardenas, 2009; Gupta, 2011). It is believed that in the absence of basal end processing by the DNA recognition complex (RecN in concert with PNPase, the product of the pnpA gene) and before long-range end-processing, blunted or nearly blunted DNA ends accumulate and two-ended DSB repair occurs via NHEJ. This is consistent with the observation that in B. subtilis, ΔpnpA is epistatic with ΔrecN and Δku, which by themselves are nonepistatic (Cardenas, 2009). However, in Mycobacteria Ku protein protects DSBs from resection by the AdnAB helicases–nucleases complex (Sinha, 2009). Similarly, Ku70-Ku80 protects DNA ends from resection by Exo1 in eukaryotes (Tomita, 2003; Wasko, 2009), suggesting that the Ku protein might delay end resection by one of the end-processing avenues.

NHEJ repairs DSBs by short processing of the ends, followed by ligation, without any major diffusion of DNA ends, because both ends are available in the neighborhood. To perform these reactions, the eukaryotic NHEJ machinery relies on many proteins, including structural stabilization factors, as well as DNA degradation, polymerization and ligation functions, whereas the bacterial NHEJ system relies only on a small number of proteins (reviewed by Daley, 2005; Pitcher, 2007; Shuman & Glickman, 2007; Weiner, 2009, Table 3; Lieber, 2010). Nevertheless, all NHEJ reactions require a core machinery that is composed of three discrete sets of functions: in bacteria, the DSB recognition homodimeric protein Ku, which binds and bridges DNA ends, recruits the DNA ligase, LigD and DNA end cleaning proteins to DNA ends (Weller, 2002, Fig. 2b). In B. subtilis, there is a set of proteins with potential activity in basal end processing (e.g. SbcCD, SbcEF, PNPase, DNAP X), but their role in NHEJ remains to be documented (Mascarenhas, 2006; Sinha, 2007; Baños, 2008; Cardenas, 2009; Ayora, 2011, Table 3). LigD is a multidomain enzyme consisting of a C-terminal ATP-dependent ligase domain fused to an upstream polymerase module (Akey, 2006; Brissett, 2007). It was proposed that after basal end processing, Ku, in the absence of RecN, binds to the DNA ends and LigD seals them (Weller, 2002; Mascarenhas, 2006; Shuman & Glickman, 2007; Baños, 2008; Cardenas, 2009). Indeed, ablation of PNPase, RecN, SbcC and UvrD1, together with Ku, sensitizes Bacillus and Mycobacterium cells to IR (Mascarenhas, 2006; Sinha, 2007; Cardenas, 2009). Whether RecQ, RecS or PcrA/UvrD DNA helicases play a specific role in NHEJ remains to be demonstrated. Because many bacteria are capable of conducting NHEJ, although this quality is not present in other closely related bacteria, it is likely that NHEJ acquisition or maintenance is a stress-driven process that cannot be inferred by phylogeny (see Weiner, 2009).

View this table:
Table 3

Proteins involved in NHEJ in various systems

ActivityB. subtilisM. tuberculosisH. sapiens
End processingRecN/PNPaseRecN/PNPase?MRN
SbcCD, SbcEF
End bindingKuKuKU70-KU80
kinase/phosphataseLigD?LigDPNK and others
Terminal processingPcrA/RecQ?/RecS?UvrD1WRN, BLM
DNA polymeraseLigDLigDDNAP μ and DNAP λ
5′ flap endonucleaseDNAP X??FEN1

In the absence of a homologous donor and after RecBCD-dependent end processing, two-ended DSBs can also be joined upon exposure of short stretches of homology or microhomology, at least in E. coli and Mycobacteria (Chayot, 2010; Gupta, 2011). This single-strand annealing (SSA) avenue, which resembles eukaryotic SSA (Ivanov, 1996; Mortensen, 1996), was detected upon inactivation of RecA and Ku in Mycobacteria (Gupta, 2011). The proteins that may catalyze strand annealing, flap removal, gap filling and end ligation in bacteria remain to be characterized.

DNA end-processing and end recruitment promote DSB repair by HR

If the DNA replication fork encounters a nick in one of the DNA strands, this interruption is converted to a one-ended DSB with a variable region of unreplicated 3′-tailed duplex DNA if the nick is in the lagging strand or to a blunt or nearly blunted end if the nick is in the leading strand. On the other hand, chemical or physical damage can generate two-ended DSBs. In both cases, the DNA ends are resected in a 5′→3′ manner to yield a recombinogenic 3′-tailed DNA that is bound by the recombinase (RecA or Rad51), aided by mediator factors, to form a nucleoprotein filament, hence committing to DSB repair by HR rather than by NHEJ (Table 2). In all living organisms, end resection is characteristically a two-step process: in the first step, basal resection generates the substrate for long-range processing, while the second step involves processive resection by redundant pathways consisting of nucleases, DNA helicases and associated proteins (Michel, 2004; Sanchez, 2007a, Fig. 4, Gravel, 2008; Mimitou & Symington, 2008; Zhu, 2008). In bacteria, processive end resection is achieved by either variations or combinations of two otherwise alternative avenues, by either one of them, as a solo orchestrated avenue, or by a pseudo-solo pathway, when a main solo avenue is impaired and a secondary avenue has to be activated. Some bacteria can also use, under certain conditions, an end resection avenue hijacked from lysogenic viruses. For example, the SKIN prophage, which lysogenizes B. subtilis 168 cells, encodes an end-processing exonuclease (RecE), an SSA protein (RecT) and an HJ resolvase (RusA) (Ayora, 2011). In other cases, these sets of functions are provided by different cryptic prophages, as E. coli K12 lysogenized by Rac (RecE/SbcA, RecT) and DLP12 (RusA) defective phages (Clark & Sandler, 1994; Sharples, 2001). These functions alone, or in concert with other cell factors, help cells to cope with DSBs via the SbcA pathway (Clark & Sandler, 1994; Kowalczykowski, 1994; Sanchez, 2007c). This DSB repair avenue (see Table 1), which proceeds by a RecA-independent recombination, is necessary for viral recombination-dependent DNA replication (Ayora, 2011), and will not be discussed.

Two alternative avenues

In bacteria other than those of the Gammaproteobacteria class, which usually lack the SbcB/ExoI function, the activation of the RecFOR avenue might not be required. In these bacteria, DNA end resection involves two alternative avenues or variations of one of them, with bacteria of the Firmicutes phyla the prototype. They consist of a complex of helicase–exonucleases (one helicase, two nucleases) as the AddAB enzyme, present also in bacteria of the Alpha- and Betaproteobacteria classes or an ssDNA exonuclease (RecJ) that works in concert with a DNA helicase [of the RecQ-like (RecQ, RecS) or the UvrD-like (RecD2, PcrA) family] and with the SsbA protein that stimulates the helicase and exonuclease activities. Similarly, eukaryotic single-strand binding protein, RPA, plays an active role during end processing (Nimonkar, 2011)

In B. subtilis, genetic studies revealed that the addAB and ΔrecJ double mutant strain has a synergistic effect, with survival after DNA damage being reduced to the levels seen in ΔrecA cells (Sanchez, 2006). Here, basal end processing may be carried out by the poorly characterized PNPase enzyme, because this bacterium lacks the majority of all the 3′→5′ specific ssDNA exonucleases described for E. coli (e.g. ExoI/SbcB, ExoIX, ExoX, ExoXI and perhaps ExoVII) (Cardenas, 2009). These results suggest that in B. subtilis, there are two major parallel end-processing avenues, the AddAB- and the RecJ-dependent, which contribute equally to accumulate 3′-tailed duplex DNA (Ayora, 2011). The AddAB enzyme consists of one N-terminal UvrD-like motor (AddA), which translocates DNA with a 3′→5′ polarity, and two distinct C-terminal nucleases. A second inactive helicase motor is present in the AddB subunit (Yeeles & Dillingham, 2010). Long-range end processing performed by the AddAB avenue is similar to the activity of other nuclease–helicase complexes of the AddAB family (namely RecBCDEco or AdnABMtu), although the implication of AddAB in the recruitment of RecA onto naked ssDNA, as it has been reported in the E. coli case, remains to be documented (Dillingham & Kowalczykowski, 2008; Yeeles, 2009; Unciuleac & Shuman, 2010; Ayora et al., 2011).

A RecQ-like DNA helicase (RecQ or RecS) in concert with RecJ, unwinds duplex DNA with blunt or 3′-overhangs, degrades the 5′-terminated strand and protects the ssDNA ends, resulting in a 3′-ssDNA overhang that is bound by SsbA that further stimulates the helicase and exonuclease activities (Sanchez, 2006; Handa, 2009). Although both avenues are operatively and functionally active, Firmicutes might have other end-processing avenues. There are some proteins (e.g. RecD2 and PcrA/UvrD DNA helicases) whose roles in end processing have not been addressed in B. subtilis.

Pseudo-solo orchestrated avenue

This avenue, which was initially described in E. coli (of the Gammaproteobacteria class), is the best-characterized end resection avenue. As in all cases, end resection occurs in two steps: a basal resection by a single-strand specific exonuclease and the second step, consisting of a long-resection performed by the highly processive helicases–nuclease RecBCDEco complex (Amundsen & Smith, 2003; Dillingham & Kowalczykowski, 2008; Niu, 2009, Fig. 2, Yeeles & Dillingham, 2010). Basal resection is needed in order to blunt the ends before RecBCD-mediated end processing, because 5′ or 3′ overhangs on duplex DNAs of ∼25 nt or longer strongly impede the action of RecBCDEco (Dillingham & Kowalczykowski, 2008; Yeeles & Dillingham, 2010, and references therein). It has been proposed that a 3′→5′ ssDNA exonuclease (e.g. SbcB/ExoIEco) and/or an enzyme with more complex activities (e.g. SbcCDEco) play a role in DNA end blunting (basal end-processing) (Connelly & Leach, 2002; Thoms, 2008). RecBCDEco, which is a heterotrimer complex composed of two helicase subunits (RecB and RecD) that translocate in parallel and with opposite polarities, and one nuclease module (located at the C-terminus of RecB), which excises the single strands displaced by the helicases, is the only contributor to processive end resection of one- or two-ended DSBs in an otherwise wt strain (Dillingham & Kowalczykowski, 2008). The RecBCDEco complex resects both DNA ends, which are differently degraded during translocation, until a chi sequence is encountered. Then, the RecBCDEco 3′→5′ exonuclease activity is attenuated, while the enzyme continues to unwind the DNA duplex and degrades the 5′→3′ strand (Dillingham & Kowalczykowski, 2008; Yeeles & Dillingham, 2010, Fig. 2). The net result is a duplex molecule with a 3′-ssDNA tail starting at the chi sequence (Dillingham & Kowalczykowski, 2008; Niu, 2009; Yeeles & Dillingham, 2010). That in E. coli, the RecBCD pathway is the unique actor in DSB repair is consistent with the observation that the DSBs generated by the SbcCDEco hairpin DNA-cleaving nuclease are dependent on DNA replication, and require RecBCDEco, RecAEco, RuvABCEco, RecGEco and PriAEco, but not RecFOREco (Eykelenboom, 2008). The RecBCDEco complex actively loads RecAEco onto naked ssDNA (Dillingham & Kowalczykowski, 2008; Yeeles & Dillingham, 2010).

Escherichia coli cells, in the recB, recC, sbcB/exoI and sbcC background, activate an alternative avenue: the RecFOR pathway. This suggests that 3′ ssDNA can be produced in the absence of RecBCDEco, but it is unstable and subsequently resected by the SbcB/ExoIEco or SbcCDEco nucleases (Kogoma, 1997; Kuzminov, 1999; Cox, 2000; Cromie & Leach, 2000; McGlynn & Lloyd, 2002b; Courcelle & Hanawalt, 2003; Marians, 2004; Michel, 2004; Kreuzer, 2005; Persky & Lovett, 2008). In the RecFOR pathway, RecJ, in concert with RecQ, degrades the 5′ ssDNA end, producing 3′-ssDNA tails with the help of SSB (SsbA) that stimulates the helicase and exonuclease activities and protects the ssDNA ends (Clark & Sandler, 1994; Amundsen & Smith, 2003; Michel, 2004; Handa, 2009; Yeeles & Dillingham, 2010). In a reconstituted in vitro assay, RecJEco can degrade one strand of duplex DNA, in the absence of RecQEco (Handa, 2009).

Not all bacteria of the Gammaproteobacteria class follow this pseudo-solo orchestrated end processing. For example Acinetobacter baylyi, which lacks recQ and helD genes, is nonviable for a null recJ mutation when additional mutation(s) in recBCD or recD are introduced (Kickstein, 2007). It is likely that in this bacterium, RecBCDAba and RecJAba represent separate active end-processing pathways, and that RecJAba, in concert with an UvrD-like helicase (RepAba, PcrAAba) or any other DNA helicase, catalyzes end resection even in the presence of RecBCDAba (Kickstein, 2007).

A solo orchestrated avenue

Bacteria of the Epsilonproteobacteria class (e.g. H. pylori), which might be devoid of the RecQ function (Rocha, 2005), encode a helicase–nucleases complex (AddAB) (Cromie, 2009). Here, the inactivation of RecOHpy makes cells moderately sensitive to IR-induced DSBs, but the absence of the AddABHpy complex makes cells as sensitive as a recAHpy mutant. The recOHpy or recRHpy mutation does not increase the sensitivity does of addABHpy cells (Marsin, 2010), suggesting that the RecFORHpy pathway cannot act as a backup system of AddABHpy long-range end processing, and that in this bacterium, only one avenue (AddABHpy) is active (Marsin, 2010).

Bacteria of the Actinobacteria phylum are devoid of the RecJ function and possess a helicase–nuclease complex, AdnAB, of the AddAB family. Some species of the phylum also encode for a RecBCD complex (Cromie, 2009), but they are a nonfunctionally redundant helicases–nuclease(s) complex, because AdnAB is involved in HR and the RecBCD complex required for SSA (Gupta, 2011). In Mycobacteria, basal end processing is carried out by poorly characterized enzymes, whereas long-range processing is catalyzed by the AdnAB complex (Unciuleac & Shuman, 2010). The resection process consists of two N-terminal UvrD-like motors that translocate in tandem with a 3′→5′ polarity and two C-terminal nucleases that form the heterodimeric AdnABMtu complex (Unciuleac & Shuman, 2010). AdnABMtu resects both DNA ends, which are differentially degraded by the two exonuclease subunits, until a chi sequence is encountered (Niu, 2009; Sinha, 2009; Unciuleac & Shuman, 2010). Upon chi cleavage, the 3′→5′ exonuclease activity is blocked in AdnABMsm, while the complex continues to unwind the DNA duplex and degrades the 5′→3′ strand (Chedin, 2006; Dillingham & Kowalczykowski, 2008; Yeeles, 2009; Unciuleac & Shuman, 2010, Fig. 2). The net result is a duplex molecule with a 3′-ssDNA tail starting at the chi sequence and coated by SSB/SsbA (Dillingham & Kowalczykowski, 2008; Niu, 2009; Yeeles, 2009; Unciuleac & Shuman, 2010). The implication of AdnABMtu, which is required for RecA-dependent HR, in RecAMtu loading onto naked ssDNA by a direct protein–protein interaction remains to be documented (Dillingham & Kowalczykowski, 2008; Yeeles, 2009; Unciuleac & Shuman, 2010). Mycobacterial RecBCD is not involved in HR, rather in RecA-independent SSA (Gupta, 2011).

A solo orchestrated avenue and its variations

In Bacteria of the Deinococcus-Thermus phylum, devoid of helicase(s)–exonuclease(s) complexes of the AddAB/RecBCD/AdnAB family (Rocha, 2005; Han, 2006), the resection process involves an ssDNA exonuclease and a DNA helicase of either the RecQ-like or the UvrD-like family. In these bacteria, poorly characterized enzymes carry out basal end processing. RecJ ssDNA exonuclease, in concert with a RecQ- or a UvrD-like (RecD2, PcrA/UvrD) helicase, degrades the 5′-terminated ssDNA strand, thereby producing 3′-ssDNA tails (Fig. 2). This is consistent with the observation that: (1) RecJ is an essential protein in D. radiodurans (Bentchikou, 2010; Cao, 2010), (2) Thermus thermophilus is naturally devoid of the RecQ and RecD2 helicases (Rocha, 2005) and (3) D. radiodurans cells lacking either recQ or recD2 have a wt DNA repair capacity, whereas uvrD cells show a markedly decreased radioresistance, strongly suggesting the active role of UvrDDra in recombinational DNA repair (Bentchikou, 2010). UvrDDra, RecQDra or another uncharacterized DNA helicase unwinds DNA with a 3′→5′ polarity and uses both blunt-ended and 3′-tailed dsDNAs as substrates, whereas RecD2Dra unwinds DNA with a 5′→3′ polarity and uses both blunt-ended and 5′-tailed dsDNAs (Umezu & Nakayama, 1993; Wang & Julin, 2004; Singleton, 2007; Shereda, 2008; Sinha, 2008). It is therefore appealing to think that the in vivo DNA helicase ‘chosen’ is dictated by the nature of the DNA substrate, although the presence of a specific DNA structure or a binding partner could also play a role in the election of the helicase.

End resection in eukaryotes is also a two-step process: the MRN/X complex (constituted by Mre11, Rad50 and Nbs1 in mammals or Xrs2 in budding yeast MRN/X) in concert with CtIP (in mammals) or Sae2 (in budding yeast) catalyzes basal resection (Table 2) (reviewed by Mimitou & Symington, 2009a, b; Niu et al., 2009). This initial resection is followed by redundant pathways consisting of nucleases, helicases and associated proteins that perform processive resection by the Dna2 helicase–exonuclease with 5′→3′ polarity as part of the BLM/Sgs1-TopoIIIa/Top3-RMI1/Rmi1-DNA2/Dna2 complex (Cejka, 2010a; Niu, 2010) or by the BLM-Exo1, in concert with MRN and RPA, to form a long 3′ ssDNA tail (long-range end processing) (Table 2) (Mimitou & Symington, 2009a; Niu, 2009; Nimonkar, 2011).

End recruitment

Cytological studies have revealed that long-range end resection is necessary for end-tethering or redistribution from two or more B. subtilis RecN foci into a discrete RC per nucleoid within the first 30 min after DSB induction, because in the absence of AddAB and RecJ, RecN-mediated end diffusion (a discrete RecN focus/nucleoid) and RecA-mediated homology search (RecA threads) are not seen (Sanchez, 2006). Concomitant with end-processing of the 5′ end, RecN binds to the ssDNA tail of the duplex molecule at the DSB, protecting the 3′-OH end and facilitating the tethering of these DNA ends (Kidane, 2004; Kidane & Graumann, 2005a; Sanchez & Alonso, 2005; Sanchez, 2006). We thus hypothesize that: (1) end-processing takes place after damage recognition by RecN and before RecN-mediated concentration of DNA ends towards a unique focus per nucleoid (end recruitment), and RecA colocalization with RecN-induced RC and (2) basal resection and/or long-range end-processing commit to repair the DSBs by HR rather than by NHEJ.

End-tethering by the RecN ATPase could also restrict chromosome diffusion by keeping several DNA ends in close proximity so that the slow long-range motion of topological domains, including the intact target template DNA, is spatially confined to facilitate genome-wide homology search of multiple DNA ends by the RecA nucleofilament (Postow, 2004; Barzel & Kupiec, 2008; Weiner, 2009). This is consistent with the observation that: (1) B. subtilis recN and ku mutations are nonepistatic, (2) the absence of cohesion (Δsmc cells) renders cells very sensitive to DSB-producing DNA damaging agents and (3) DNA partners of one-ended DSB repair are confined into spatial proximity, while such close proximity is not obvious during two-ended DSB repair (Dervyn, 2004; Mascarenhas, 2006; Cardenas, 2009). In eukaryotic cells, end tethering and DSB coordination are catalyzed by Rad50, as part of the MRN/X–CtIP/Sae2 complex (Hopfner & Tainer, 2003; Mimitou & Symington, 2009b; Niu, 2009).

Recombinase loading

In Firmicutes, RecA hydrolyzes dATP preferentially over rATP, and supports more efficiently the DNA strand exchange reaction in the presence of dATP (Lovett & Roberts, 1985; Steffen & Bryant, 1999; Carrasco, 2008). RecAEco, under identical experimental conditions, shows a preference for rATP (Lovett & Roberts, 1985; Steffen & Bryant, 1999). Saturating amounts of SsbA/SSB, independent of the order of addition, reduce the ssDNA-dependent dATPase activity and block the ATPase activity of RecA (Lusetti & Cox, 2002). Hence, SsbA/SSB must be displaced from the ssDNA to assemble the RecA presynaptic filament. Alternatively, SsbA/SSB recruits RecA mediators [RecO (RecOREco) or RecOR (RecFOREco)] onto ssDNA, and the mediators facilitate RecA loading (Lohman & Ferrari, 1994; Shereda, 2008). Genetic studies have demonstrated that a recF, recO or recR defect is partially overcome by common suppressors in both B. subtilis and E. coli, (e.g. a recA mutant allele, etc.), suggesting that RecA recruitment and/or RecA filament modulation by RecF, RecO and RecR might occur in a concerted manner (single step) (Alonso & Luder, 1991; Wang, 1993; Carrasco, 2001). Cytological studies in B. subtilis cells support a stepwise assembly, with RecO and RecR colocalizing with damage-induced RecN foci, just before or together with RecA, within the first 30–45 min after MMC addition (see Fig. 4). In exponentially growing cells, RecA localizes throughout the nucleoids, whereas this pattern of localization changes dramatically upon induction of DSBs. RecA forms discrete foci that colocalize with damage-induced RecN foci, 30–45 min after MMC addition. Then, damage-induced RecA foci form threads that emanate from RecN-RecOR-RecA foci (Kidane & Graumann, 2005a). RecF foci formation, which is strictly dependent on the presence of RecO, colocalizes with RecN later, concomitant with RecA threads formation (Kidane, 2004; Sanchez, 2006, Fig. 4). In vitro studies have been carried out to understand how RecFOR works. They reveal that both B. subtilis and E. coli RecO, which interacts with SsbA/SSB, alone or in concert with RecN-RecR, load RecA onto SsbA/SSB-coated 3′-ssDNA tails (Umezu, 1993; Umezu & Kolodner, 1994; Manfredi, 2008, 2010). Recently, it has been shown that RecFEco plays an accessory role in loading RecAEco·ATP·Mg2+ onto linear ssDNA (Handa, 2009). Once the RecA nucleofilament has been formed, it catalyzes a search for homology by random and transient interactions, followed by strand invasion and synapsis, resulting in the generation of extensive heteroduplex regions, where strand transfer by branch migration occurs (Cox, 2007a, b; Galletto & Kowalczykowski, 2007; Ayora, 2011).

Biochemical data show that the RecBCDEco complex, upon interaction with the chi site, activates RecBEco to facilitate the loading of RecAEco onto the exposed naked ssDNA (Anderson, 1999; Dillingham & Kowalczykowski, 2008; Yeeles & Dillingham, 2010). The activated RecBEco domain also interacts with the RecAEco nucleoprotein filament (Dillingham & Kowalczykowski, 2008; Lucarelli, 2009; Yeeles & Dillingham, 2010). It is unknown whether the RecBCDEco paradigm for RecAEco loading onto naked ssDNA, upon DSB and concomitant end-resection, can be extended to other members of the family (AddAB or AdnAB).

Assembly and disassembly of the RecA filament is in turn regulated not only by dATP and ATP (Lovett & Roberts, 1985; Steffen & Bryant, 1999; Carrasco, 2008) but also by the interaction with many proteins (review by Cox, 2007a, b). There are a number of RecA nucleofilament modulators such as DinI, RdgC or RecU restricted to certain bacterial classes (Cox, 2007b; Cañas, 2008). The RecXEco or RecXNgo protein inhibits RecAEco/RecANgo filament extension, leading to net filament disassembly (Stohl, 2003; Drees, 2004; Gruenig, 2010). RecFEco antagonizes RecXEco function and stimulates RecAEco-mediated joint molecule formation in a reconstituted early step of DSB repair (Lusetti, 2006; Handa, 2009; Gruenig, 2010). The RecU HJ resolvase modulates RecA activities by enhancing RecA-promoted D-loop formation and inhibiting DNA strand exchange (Ayora, 2004; Carrasco, 2005, Fig. 3, McGregor, 2005; Cañas, 2008). The isolation of the separation of function mutants, i.e. proficient in HJ cleavage, but deficient in RecA modulation, has been reported (Cañas, 2008). Whether other HJ resolvases modulate RecA functions remains to be determined. In eukaryotes and archaea, the Rad54 protein stimulates Rad51-mediated DNA strand exchange by directly interacting with the Rad51 nucleoprotein filament and remodelling the topology of duplex DNA (reviewed by San Filippo, 2008). The widely conserved HepA1 and HepA2 proteins, which belong to the Swi2/Snf2 family of helicases, as Rad54, might play such a role in HR (Haseltine & Kowalczykowski, 2009; Ayora, 2011, Table 2). A number of proteins have been identified that exert an antirecombinase activity, as the PcrA/UvrD DNA helicase that promotes RecA/RecAEco nucleoprotein filament dissociation (Veaute, 2005; Anand, 2007). Finally, proteins of the MutSII subfamily (bacterial MutS2 and eukaryotic MSH4 and MSH5) are involved in HR, probably acting as modulators. The MutS2Hpy protein inhibits RecA-mediated DNA strand exchange, whereas MutS2Tth, which binds and processes branched structures (D-loop, nicked HJ), may suppress HR through the resolution of early recombination intermediates (Pinto, 2005; Fukui, 2008).

Strand exchange, replication fork restart, branch migration and HJ processing

One-ended DSB repair, stable DNA replication and eukaryotic break-induced replication resemble recombination-dependent replication used by bacteriophages to generate concatemeric DNA substrates (Viret, 1991; Kogoma, 1997; Ayora, 2002; Krogh & Symington, 2004; Martinez-Jimenez, 2005; McEachern & Haber, 2006). RecA bound to the 3′-tailed end promotes invasion of the homologous duplex, causing the displacement of one of the two DNA strands and the formation of a D-loop intermediate (Fig. 2a). The 3′ end from the broken chromosome is used to prime leading strand DNA synthesis with the donor duplex as a template, followed either by D-loop disruption or HJ resolution (Fig. 3a). In the former case, the disruption of a D-loop or a nicked HJ displaces the invaded strand to reconstitute a replication fork (Fig. 3a, steps i–iii), whereas in the latter case, upon branch migration, the other end of the break interacts with the displaced strand to form a HJ (Fig. 3a, steps iv–vi). RuvAB loads the RecU/RuvC resolvase and the resolution of the HJ allows the displacement of the invading strand. A processive replication fork is established and DNA synthesis proceeds to the end of the chromosome. In two-ended DSB repair, the second 3′-tailed end is captured by a RecA-independent event, leading to a double HJ (Fig. 3b). The processing of the two possible recombination intermediates will be discussed in the next section.

In D. radiodurans, the multi-ends are repaired by a RecA-dependent mechanism, using overlapping homologies, to form a full genome by extended SDSA (ESDSA), followed by cross-over (Zahradka, 2006). In ESDSA, the chromosomal fragments of the multigenome copies are used both as primers and as templates for the massive synthesis of complementary single strands, as occurs in a single round of a multiplex PCR (Zahradka, 2006).

Replication restart

Some features of the initiation of DNA replication in E. coli are conserved throughout Bacteria. In E. coli cells, the replication apparatus (replisome) is loaded onto the ssDNA region formed by DnaA·ATP-mediated helix opening at the replication origin (oriC) by a direct interaction of the replisome organizer, DnaA·ATP, with the replicative hexameric helicase, complexed with its helicase loader, DnaB·ATP·DnaC (Baker & Bell, 1998). In a second step, DnaB·ATP bound to ssDNA loads the primase, DnaG, and by interaction with the τ (DnaX) subunit of DNAP III loads the DNAP III holoenzyme (Baker & Bell, 1998; O'Donnell, 2006). In a third step, DnaG provides the 3′-OH end to be elongated by DNA synthesis. After the initiation reaction, DnaA and oriC are inactivated by different mechanisms in different bacteria, to prevent reinitiation in the same round of the cell cycle (Katayama, 2010). Hence, replication fork reactivation requires oriC-independent reloading of the replicative hexameric helicase (in a D-loop intermediate). In E. coli, this process can follow three different avenues, PriA-PriB-DnaT, PriC-PriA or PriC-Rep, depending whether the template for the recombination intermediate is in the leading or the lagging strand (Sandler, 2000; Heller & Marians, 2005; Gabbai & Marians, 2010). At a D-loop, DnaG and the replicase (DNAP III) are recruited via interactions with DnaB. DnaG, upon interaction with the DnaB helicase, initiates RNA synthesis on the lagging strand, and under certain circumstances, forms new primers also on the leading strand by binding the DnaB helicase on the lagging strand (Gabbai & Marians, 2010). This allows the replication fork to continue passing damaged sites in the leading and lagging strand template, leaving the lesion behind in the form of an ssDNA gap.

In B. subtilis cells, three replication proteins, DnaD, DnaB and DnaI, load the replicative hexameric helicase DnaC (different from the DnaCEco helicase loader) at the oriC region during replication initiation and at stalled replication forks during replication restart (Bruand, 2001, 2005; Marsin, 2001; Velten, 2003; Rokop, 2004; Ioannou, 2006; Smits, 2010). Indeed, there are two loading systems for the hexameric replicative helicase at the oriC region: the DnaDB- and DnaI-dependent mechanisms (Velten, 2003; Ioannou, 2006; Nuñez-Ramirez, 2007; Smits, 2010). DnaC bound to ssDNA loads DnaG and by interacting with the stated τ(DnaX) subunit the replicases DNAP C and DNAP E are loaded (Bruck & O'Donnell, 2000; Dervyn, 2001; Noirot, 2007; Sanders, 2010). Finally, DnaG provides 3′-OH end for the leading strand synthesis, while DnaG and DnaE provide the 3′-OH end to be elongated in lagging strand DNA synthesis (Sanders, 2010). The loading of the helicase via two stable intermediates and the use of two DNA polymerases, DNAP C (Pol ɛ-like) and DNAP E (Pol α-like), resemble the replication apparatus of eukaryotes (Velten, 2003; Sanders, 2010). Replication fork restart in B. subtilis shares some features with the E. coli system, although the presence of the PriB, PriC and DnaT assembly factors is not obvious (Noirot, 2007). Here, PriA, in concert with DnaB, DnaD and SsbA, load the replication apparatus at the recombination intermediate (Bruand, 2001, 2005; Polard, 2002; Noirot, 2007; Smits, 2010, Fig. 4). From the primosome components, only PriA is ubiquitous among free-living bacteria. The restart components of the PriA, DnaD and DnaB type or of the PriA-PriB-DnaT, PriC-PriA or PriC-Rep type are not observed in many bacterial endosymbionts (Sharples, 2009). For example in species classified in the Mollicutes class PriA, DnaD and DnaB are not observed, and the DnaB protein is not observed in species classified within the Clostridia class (Petit, 2005; Sanchez, 2007c). The role of a hypothetical PcrA pathway in replication fork restart has not been addressed experimentally. Additionally, DNAP I and/or YpcP (a DNAPol I paralogue), type II DNA topoisomerases and a DNA ligase are needed to facilitate the closing of the DNA ends (reviewed by Kuzminov, 1999). Then, branch migration and resolution proteins resolve the recombination intermediates, once the complete replisome has been loaded, in order to fully reconstitute the active replication fork (see Fig. 3a).

Cytological studies have shown that SbcC foci colocalize with DnaX, which is a bona fide replicase marker, 120 min in ∼40% of the cells upon addition of MMC, whereas RecN and SbcE foci, on the other hand, colocalize with DnaX in less than 15% or 20% of cases, respectively (Mascarenhas, 2006; Krishnamurthy, 2010, Fig. 4). This is consistent with the observation that SbcC and SbcE interact with the DnaG primase, and SbcCEco interacts with the replicase (Noirot & Noirot-Gros, 2004; Darmon, 2007). However, whether SbcC acts in any replication-restart pathway remains unknown.

Branch migration and HJ processing

The RecG and RuvAB families of branch migration translocases are ubiquitous in free-living bacteria (McGlynn & Lloyd, 2002a; Wen, 2005). The RecU/RuvC HJ resolvase (Table 2) forms a single, discrete focus on the nucleoid 90–120 min after the induction of DSBs, in a RuvAB-dependent manner, and it colocalizes with RecN at the RC, which confirms that the repair of DSBs is a sequential process (Sanchez & Alonso, 2005, Fig. 4; Sanchez, 2005, 2006).

There are at least two possible avenues to process the one-ended recombination intermediates: in one of them, branch migration translocases such as RecG or RuvAB catalyze the regression of the extended heteroduplex so that the D-loop becomes now a HJ intermediate (Fig. 3a, step v). Then, strand cleavage of the single HJ by RuvAB and RecU, in Firmicutes, or RuvABC in other bacteria, and concomitant strand ligation generate the substrate that leads to replication fork reactivation. Depending on the orientation of the cleavage, cross-over (CO) products, where strands are exchanged relative to the initial chromosome organization (dimeric chromosomes), or non-crossover (NCO) products (monomeric chromosomes), where no strand exchange occurs, can be generated (West, 2003, Fig. 3a, Sanchez, 2007c; Ayora, 2011). ΔrecU cells display defective resolution of HJs and when SMC is present in limiting amounts, 25% of the cells are anucleated. In the absence of SMC, however, the removal of RecU is synthetically lethal (Pedersen & Setlow, 2000), suggesting that additional functions act in parallel to RecU/RuvCEco. In the second model, RecG branch-migrates the D-loop to allow heteroduplex extension (Fig. 3a, step iii). Then, the D-loop or the nicked HJ is resolved by the action of the MutS2 disruptase (Fukui, 2008), and concomitant strand ligation generates the substrate that leads to replication fork reactivation (Fig. 3a, step iii). This HJ-independent one-ended DSB repair resembles break-induced replication (Kogoma, 1997; Pâques & Haber, 1999; Krogh & Symington, 2004; Marians, 2004).

The two-ended DSB repair model (Szostak, 1983) differs from the one-ended DSB models in that the D-loop intermediate can be processed by three different mechanisms and that there are two HJs rather than one (Fig. 3b). In two of these mechanisms, extension of the strand exchange reaction after a second-end capture leads to the formation of a double HJ (Fig. 3b, steps iii–vii), in which none (dissolution) or two of the four DNA strands (resolution) are cleaved (Holliday, 1967; Szostak, 1983; Wu, 1999; Lopez, 2005; Bachrati & Hickson, 2008; Klein & Symington, 2009; Bzymek, 2010; Cejka, 2010b). In the third mechanism, the presence of a double HJ is not obvious (Fig. 3b, steps viii–ix). The three mechanisms will be briefly described:

HJ resolution

Once the genetic information lost at the break is restored, by copying the homologous template, the RecU/RuvCEco HJ resolvase, with the help of the RuvAB branch migration translocase, cleaves the recombination intermediate (Fig. 3b, step v). As predicted by the canonical DSB repair model, the orientation of resolution of the double HJ results in either monomeric, NCO, or dimeric, CO, chromosomes with a 1: 1 ratio (Szostak, 1983). However, molecular mechanisms that introduce a strong bias for the direction of HJ processing towards NCOs have been described both in E. coli and in B. subtilis cells (West, 1997; Carrasco, 2004). The generated CO can then be resolved by the site-specific recombination machinery coupled to DNA translocation before cell division.

Recently, it has been proposed that pathological PriA-dependent replication in E. coli cells, rather than the accumulation of RecA-dependent recombination intermediates, is primarily responsible for the phenotype of recG cells (Rudolph, 2010). Conversely, the absence of proteins involved in processing and resolving of HJ intermediates in B. subtilis, like RecG, RuvAB or RecU, leads to chromosomal segregation defects that are suppressed in the ΔrecA background (Sanchez, 2005, 2007b). It is likely that the accumulation of toxic intermediates, in the absence of RecG, and defective resolution of HJs, in the absence of RuvAB or RecU, impair the segregation of daughter chromosomes, and hence preventing recombination suppresses the segregation defect (Carrasco, 2004; Sanchez, 2007b). In the absence of any externally added DNA damaging agent, the inactivation of RecG or RuvAB reduces the plating efficiency to ∼20% of that of the parental strain, as it also channels recombination intermediates into the CO pathway, leading to the accumulation of toxic intermediates (Carrasco, 2004; Sanchez, 2007b). A ΔrecG mutation renders cells very sensitive to DNA damaging agents and it is synthetically lethal with a ΔrecU mutation (Sanchez, 2005, 2007b). Thus, it is possible that the segregation of newly replicated chromosomes involves redundant pathways.

HJ dissolution

In the absence of branch migration translocases (RuvAB or RecG) and/or the HJ resolvase (RecU/RuvCEco), the HJs can be dissolved by a poorly defined avenue. It was proposed that Topo III, in concert with a RecQ-like (RecQ or RecS) or a UvrD-like (RecD2, UvrD/PcrA) DNA helicase and SsbA/SSB, may dissolve double HJs (Wu, 1999, Fig. 3b, step vi, Table 1, Lopez, 2005), as shown for eukaryotic cells (Bachrati & Hickson, 2008; Bzymek, 2010; Cejka, 2010b). However, direct evidence for HJ dissolution in bacteria has not been reported.

Nonreciprocal transfer of DNA via SDSA

Because ΔrecGΔrecU and ΔrecGΔrecQ double mutants exhibit a synthetic growth defect (Sanchez, 2007b) and ΔrecGΔripX cells are severely impaired in chromosomal segregation, it is possible that in vivo HJs intermediates are mainly resolved or dissolved. However, two-ended DSB repair in the absence of double HJ formation cannot be ruled out (Fig. 3b, steps viii–ix). Here, RecG translocates the D-loop that migrates with the replicase so that the helicase activity is coupled to DNA synthesis (Pâques & Haber, 1999; Krogh & Symington, 2004; Meddows, 2004, Fig. 3b, steps viii–ix). After 3′-extension of the invading DNA end, the joint molecule is disrupted by unwinding (via bacterial PcrA/UvrD, RecQ? or eukaryotic Srs2) of the D-loop intermediate. The invading and extended 3′-end anneals via eukaryotic Rad52 or bacterial RecO to a complementary sequence at the other side of the break, leading to NCO products (Pâques & Haber, 1999; Krogh & Symington, 2004; McEachern & Haber, 2006; Manfredi, 2010). The formation of NCOs via SDSA in bacteria remains to be documented (see Petit & Ehrlich, 2002; Veaute, 2005), but in ΔrecG cells, defects in chromosomal segregation occur with a>150-fold frequency (Carrasco, 2004; Sanchez, 2007b). The presence of an alternative mechanism for SDSA that includes the action of the MutS2 disruptase cannot be discarded (see Fukui, 2008). MutS2 recognizes branched DNA structures, including HJ structures, and cleaves them in vitro (Fukui, 2008), but evidence for HJ disruption in vivo is missing (Pinto, 2005, Fig. 2b, step viii). Such an activity is also proposed in Fig. 3a (steps i–iii) to reconstitute the replication fork after one-ended DSBs. Furthermore, there is no evidence in bacteria, other than E. coli, that the asymmetrically extended D-loop intermediate or nicked HJ (Fig. 2b, step viii) is processed by a Mus81-Eme1/Mms4-like structure-specific endonuclease, as reported for eukaryotic cells and proposed for E. coli cells (Osman, 2003; Hollingsworth & Brill, 2004; Meddows, 2004). This is consistent with the observation that RuvCEco might cleave nicked HJs (Osman, 2009).

Disassembly of the recombination apparatus and chromosomal segregation

Concomitant with the dissipation of the RecA threads (Kidane & Graumann, 2005a), the RecU foci are dismantled 120 min after DSBs induction (Sanchez, 2005), and before resumption of growth, complete removal of entanglement between sister DNAs created in the process of DNA replication (i.e. decatenation) and resolution of CO products (chromosomal dimers in circular genomes, Fig. 3b) take place. The former step is accomplished by the concerted action of the SMC/MukB condensin in concert with the heterotetrameric Topo IV (Tadesse, 2005; Hayama & Marians, 2010; Li, 2010), whereas the latter step is accomplished when the preassembled RipX-CodV/XerCDEco/XerSLla site-specific tyrosine-recombinase interacts with the septum-located SpoIIIE/FtsKEco or SftA translocases (Sherratt, 2003; Le Bourgeois, 2007; Ptacin, 2008, Fig. 4; Kaimer, 2009). The tyrosine recombinase, whose action is regulated and facilitated by the DNA translocases, acts at the dif recombination sites, which are located close to the terminus region that is segregated last, just before cell division, so that site-specific recombination takes place when chromosome dimers need to be resolved. The role of the SpoIIIE/FtsKEco and/or SftA translocase in the reaction is to place the two dif sites into close proximity (Le Bourgeois, 2007; Ptacin, 2008) and to activate site-specific recombinase to catalyze the dimer to monomer conversion. Upon resolution, the translocase(s) directionally translocate the chromosomes through the division septum. Growth resumption takes place ∼180 min after damage induction (Kidane, 2004, Fig. 4).

Conclusions and perspectives

DSBs are dangerous lesions that need to be efficiently repaired in order to prevent the accumulation of chromosomal aberrations. This effect is particularly magnified in cells that are deficient in DSB repair (Sanchez, 2007a, c). Genetic, cytological and biochemical data allow us to propose that, in B. subtilis, RecN recognizes pre-existing ssDNA tails on duplex DNA to form an RC (Fig. 4). SsbA binds and coats the ssDNA concomitantly to RecN binding to the 3′-end. In this process, the Hbsu/HU, LrpC chromatin-associated proteins and the SMC–ScpA complex, in concert with Topo IV, contribute to the organization, condensation and disentangling of the nucleoid (Sanchez & Alonso, 2005; Tadesse, 2005; López-Torrejon, 2006; Shintomi & Hirano, 2007; Cardenas, 2009; Hayama & Marians, 2010; Li, 2010). As with eukaryotic or archaeal cells, end processing can be divided into two distinct steps: first, a 3′→5′ ssDNA exonuclease (e.g. PNPase in concert with RecN or SbcBEco SbcCDEco) promotes the basal removal of dirty ends or blunting of the lagging strand tails (Michel, 2007; Cardenas, 2009; Mimitou & Symington, 2009b). Second, long-range resection process involves solo or redundant avenues consisting of nucleases and DNA helicases, as the nuclease(s)–helicase(s) AddAB/RecBCD/AdnAB complex and/or the nuclease RecJ in concert with a RecQ-like or UvrD-like helicase and SsbA/SSB, to resect the 5′-ends generating the 3′-tailed duplex substrate needed for RecA filamentation (Dillingham & Kowalczykowski, 2008; Hopkins & Paull, 2008; Huertas, 2008; Zhu, 2008; Montague, 2009; Niu, 2009, 2010, Fig. 2; Cejka, 2010a; Ayora, 2011). By a RecN·RecN interaction, different 3′-OH ends are tethered together to concentrate the DNA ends, forming a rosette-like structure in vitro or a discrete focus per nucleoid or RC in vivo (Kidane, 2004; Sanchez, 2008, Fig. 2). RecO alone or in concert with RecN and RecR partially displaces SsbA from ssDNA and promotes RecA loading onto SsbA-coated ssDNA (Kidane, 2004; Sanchez, 2006, 2008, Fig. 2; Carrasco, 2008; Manfredi, 2008). In E. coli, however, the loading of RecA onto the 3′-ssDNA ends relies on RecBCD or on RecOR(F) onto SSB-coated ssDNA in the absence of the former avenue (Umezu & Kolodner, 1994; Shan, 1997; Hobbs, 2007; Sanchez, 2008, Fig. 2). RecA partially disassembles the RecN rosette-like structures and forms discrete threads or filaments that search for homology along the sister chromosome located in the other cell half (two-ended DSBs) or in the neighborhood (one-ended DSBs) (Kidane & Graumann, 2005a; Sanchez, 2006, 2008). When a homologous segment is found, the RecA nucleoprotein filament, in the presence of RecA modulators, promotes DNA strand invasion (D-loop intermediate) (Carrasco, 2005; Carrasco, 2008; Manfredi, 2008). This substrate is then used by the replisome assembly proteins to load the hexameric replicative helicase. The activated helicase recruits DnaG and the replicase(s) (Bruand, 2001, 2005; Rokop, 2004; Zhang, 2005). At two-ended DSB, after second end capture, which could be mediated by RecO and DNA synthesis, a genuine double HJ is formed (Fig. 3). RecU modulates RecA activities by promoting RecA-catalysed strand invasion and by inhibiting RecA-mediated branch migration (Ayora, 2004; Carrasco, 2005, 2009; Cañas, 2008). The branch migrating enzymes (RecG or RuvAB) promote strand exchange, and RecU/RuvC catalyzes the resolution of the HJ formed by recombination, leading to the full re-establishment of the replication fork (West, 2003; Robu, 2004; Carrasco, 2009). Alternatively, the double HJs are dissolved by Topo III in concert with a RecQ-like or a UvrD-like DNA helicase and the SsbA/SSB protein (Fig. 3). These functions are required to prevent the accumulation of chromosome dimers and/or dead-end recombination intermediates (Carrasco, 2004). If DSB repair cannot be completed, the cell undergoes unprogrammed cell death, thus eliminating the possibility of passing on unrepaired genetic material. Indeed recG, recU and ruvAB cells show a fivefold reduction in their plating efficiency, while recA cells show a 10-fold reduction (Sanchez, 2007a, c). In the absence of the RecA protein, limited processing by the NHEJ machinery can catalyze the joining of the two-ended DSBs (Fig. 2b) (Mascarenhas, 2006). However, NHEJ can be mutagenic during vegetative growth. In the absence of a homologous DNA template, the role of NHEJ in two-ended DSB repair is more relevant, as occurs during the stationary phase (Weller, 2002) or in spores (Wang, 2006).


This work was supported by grants BFU2009-02907 and CSD2007-00010 and BFU2009-07167 from Ministerio de Ciencia e Innovación-Dirección General de Investigación to S.A. and J.C.A., respectively. We are grateful to Peter L. Graumann for the communication of unpublished results.


  • Editor: Martin Kupiec


View Abstract