recombinational repair and restart of damaged replication forks - Nature

REVIEWS

RECOMBINATIONAL REPAIR AND RESTART OF DAMAGED REPLICATION FORKS Peter McGlynn and Robert G. Lloyd Genome duplication necessarily involves the replication of imperfect DNA templates and, if left to their own devices, replication complexes regularly run into problems. The details of how cells overcome these replicative ‘hiccups’ are beginning to emerge, revealing a complex interplay between DNA replication, recombination and repair that ensures faithful passage of the genetic material from one generation to the next.

Institute of Genetics, University of Nottingham, Queen’s Medical Centre, Nottingham NG7 2UH, UK. Correspondence to P.M. or R.G.L. e-mails: peter.mcglynn@ nottingham.ac.uk; bob.lloyd@nottingham. ac.uk doi:10.1038/nrm951

The ability of cells to replicate millions of base pairs of DNA every time they divide is one of the wonders of biology. The fidelity with which this replication is achieved is equally impressive, and ensures accurate transmission of the genetic code from one generation to the next. However, genome duplication, although inherently accurate and highly processive, relies on a close interplay with recombination and DNA repair. Indeed, the main function of recombination seems to be to underpin genome duplication. This is because the replication machinery often encounters ‘roadblocks’ that have the potential to stop replication in its tracks. These blocks take the form of lesions in, or on, the DNA template, and are potentially lethal if they prevent the completion of replication. But there can be problems in overcoming these blocks, and genetic defects, cancer and possibly the degenerative effects of ageing are the potential costs of any mistakes1,2. How often do replication-fork problems arise? A recent study in Escherichia coli using inactivation of the DnaC protein — which is thought to be involved only in assembly of replication forks — concluded that 18% of cells require replication-fork reloading during a single round of chromosome duplication in the absence of any exogenous DNA-damaging agents3. Other estimates4 put the figure as high as 50%. The frequency of replication-fork repair in eukaryotes is even less well defined, but, given the larger genome sizes and the

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

presence of multiple forks, replicative problems might occur at least as often as in bacteria. The recent appreciation that replication-fork problems arise more often than is generally realized has led to a reassessment of the roles of DNA recombination and repair proteins in maintaining chromosomal duplication. The essential role of recombination in the replication cycle of bacteriophage T4 (REF. 5) and its involvement in specific types of eukaryotic and prokaryotic DNA replication have been known for years6,7, but its importance in normal chromosomal duplication has not been fully understood. Genetic studies, mainly in bacteria and yeast, have cast new light on the interplay between DNA replication and recombination. Furthermore, the ability to reconstitute increasingly complex biochemical reactions in vitro has allowed the molecular details of these mechanisms to be analysed. Here, we discuss how — and when — replication-fork blockage can arise, what a blocked fork might look like and what the potential roles of recombination are in overcoming such blocks. The mechanisms by which replication is restarted, together with what little is known about how all these processes are coordinated, are also reviewed. Although bacterial and yeast systems have provided the greatest insights into these aspects of DNA metabolism, and form the basis for this review, emerging work on analogous mechanisms in higher eukaryotes is also mentioned.

VOLUME 3 | NOVEMBER 2002 | 8 5 9

© 2002 Nature Publishing Group

REVIEWS

a

DNA-base damage

b

Chemical crosslink

Nucleotide misincorporated Direction of replication

c

Nick in DNA backbone

d

Leading strand

Lesion on lagging-strand template

e

Free duplex DNA end

Replication continues Gap in lagging strand

RNA primer Lagging strand

Lesion on leadingstrand template

Replication eventually stops Gap in leading strand

Figure 1 | Potential types of replication-fork damage. a | Damage to DNA bases might provoke incorporation of the wrong nucleotide by the replicative polymerase, which allows replication to proceed but causes mutations to be fixed in the daughter molecule. b | Other lesions, such as chemical crosslinks formed between the two template DNA strands, might completely halt the replication machinery. c | Breaks in the template DNA backbone would block replication, but would also release a free double-stranded DNA end with the potential to recombine with homologous sequences. d | Lesions that are located in a single template strand and block the advance of the DNA polymerase might cause different replication problems. Lagging-strand template blocks might be bypassed by repriming lagging-strand synthesis downstream of the lesion, which allows replication to proceed, but leaves a gap in the lagging strand that has to be repaired later. e | By contrast, leadingstrand template lesions might halt the leading-strand polymerase at the replicative block, but the lagging-strand polymerase might possibly continue some way beyond the damage before halting. This would result in a stalled replication fork in which the leading strand is displaced from the branch point of the fork. 3′-ends of the leading and lagging DNA strands are depicted by arrowheads.

REPLISOME

A complex of DNA-replication enzymes that contains two DNA polymerases for leading- and lagging-strand synthesis, sliding clamps to maintain processivity, a sliding-clamp loader, a helicase to unwind the parental doublestranded DNA ahead of the advancing fork and a primase to synthesize RNA primers for discontinuous lagging-strand synthesis. OKAZAKI FRAGMENTS

Individual lagging strand that ranges from 40–300 base pairs (bp) in eukaryotes, and 1,000–2,000 bp in bacteria. PYRIMIDINE DIMER

Adjacent pyrimidine bases that are covalently linked in a DNA strand as a result of irradiation with UV light. Such lesions block normal replicative DNA polymerases.

860

Potential sources of replication-fork blockage

Direct chemical damage to the template DNA from both exogenous and endogenous sources is a chronic problem, and all cells have many systems that excise and replace damaged nucleotides8. These excision systems minimize the frequency with which the replication machinery encounters nucleotide damage. However, some lesions remain unrepaired prior to the arrival of a REPLISOME. Some base lesions cause the replicative polymerase to incorporate the wrong nucleotide, generating point mutations in the DNA (FIG. 1a). Other lesions, or gaps in the DNA backbone, can block progression of the replisome, and are lethal if they are not dealt with and replication restarted (FIG. 1b). A gap could also release one of the arms of the fork as a free duplex DNA end9 (FIG. 1c), which would be a substrate for recombination enzymes and might provoke inappropriate recombination, leading to chromosomal rearrangements. However, the frequency with which gaps are encountered by the replication machinery is unclear, and genetic studies in E. coli indicate that such events are uncommon in cells that have been irradiated with ultraviolet (UV) light10.

| NOVEMBER 2002 | VOLUME 3

Protein–DNA adducts associated with normal metabolism might also present obstacles to replication. Transcription normally occurs concurrently with replication and, given that replication is an order of magnitude faster than transcription, collisions are inevitable.A bacteriophage DNA polymerase can bypass stalled and transcribing E. coli RNA polymerases in vitro11, but polymerases that act at more complex replication forks might not have this capacity. Indeed, genetic evidence indicates that collisions do lead to pausing and breakdown of replication forks in both prokaryotes and eukaryotes10,12–15. The importance of removing RNA polymerases that are stalled at lesions in transcribed DNA is underlined by enzyme systems that displace RNA polymerases and facilitate repair of the original DNA damage16. Other physiologically normal protein–DNA complexes, such as nucleosomes and transcription factors17, bound ahead of the fork might also halt advance of the replication fork. Secondary structures, including triplet repeats, in the DNA could also inhibit fork progression18,19. What does a damaged replication fork look like?

Different blocks cause different types of damage to the replication fork; this is highlighted by the varied recombination- and repair-gene requirements for dealing with exposure to different DNA-damaging agents. A prime consideration could be the continuous versus discontinuous nature of leading- and lagging-strand synthesis (BOX 1). Blocks on the lagging-strand template might be bypassed by repriming OKAZAKI FRAGMENT synthesis downstream of the block20, leaving the lesion in a single-stranded gap to be repaired later by recombination with an intact sister duplex21,22 (FIG. 1d). Such a mechanism has become generally accepted, although convincing evidence is lacking. Lesions in the leading-strand template could present a greater hurdle, because the repeated synthesis of RNA primers for the priming of DNA synthesis is not thought to occur on the leadingstrand template. However, evidence of the inability to prime DNA synthesis on the leading strand repeatedly is provided largely by in vitro experiments, whereas several in vivo studies indicate that leading-strand synthesis might occur discontinuously, in a similar way to lagging-strand synthesis23,24. Nevertheless, the consensus is that any block to the leading-strand polymerase will halt replication owing to an inability to reprime DNA synthesis downstream of the block. In support of this view, blockage of both leadingand lagging-strand synthesis by human cell-free extracts on DNA that contains a PYRIMIDINE DIMER in the leadingstrand template has been observed25,26. However, lagging-strand synthesis became uncoupled from that of the leading strand, and continued beyond the end of the leading strand to generate a fork in which the leading strand was displaced from the branch point (FIG. 1e). Another study27 used the preprogrammed replicationfork block in the RIBOSOMAL DNA CLUSTER (rDNA) of Saccharomyces cerevisiae, in which forks that approach from one direction are halted at a specific sequence. Very little single-stranded DNA (ssDNA) was present at the

www.nature.com/reviews/molcellbio


REVIEWS

Box 1 | The mechanics of DNA replication

RIBOSOMAL DNA CLUSTER

(rDNA). Tandem repeats of genes that encode ribosomal RNA. TUS–TER

A complex formed by binding of the Tus protein to specific ter sequences in bacterial genomes. It blocks advancing replication forks, which allows the opposing replication fork to converge with the blocked fork to complete duplication of the circular bacterial chromosome. REPLICATIVE HELICASE

The helicase activity required to unwind the parental doublestranded DNA ahead of the advancing replication fork, which allows the unwound strands to be used as templates for leading- and lagging-strand synthesis. SLIDING CLAMPS

Also known as processivity factors. Protein dimers or trimers that encircle and slide along double-stranded DNA. They tether the replicative polymerase and prevent its rapid dissociation from the template DNA. HELICASE

A protein that uses the energy of ATP hydrolysis to disrupt hydrogen bonding between two nucleic-acid strands, therefore separating the strands.

DNA replication is catalysed in a highly ordered Parental DNA manner by a multisubunit enzyme complex known as the replisome. The replisome shares significant functional and structural features between Replicative helicase DnaB prokaryotes and eukaryotes, with the hallmark of Primase γ-complex continuous DNA synthesis of the leading strand and DnaG clamp loader discontinuous synthesis of the lagging strand. Single strand Leading-strand– binding protein (SSB) In E. coli the replicative helicase DnaB acts to polymerase complex disrupt parental DNA and expose the two individual strands. One of these strands acts as the template for β-sliding clamp RNA primer synthesis of the leading strand by a DNA polymerase III complex, which proceeds uninterrupted in the 5′ Lagging-strand– to 3′ direction (arrows represent the 3′ ends of DNA polymerase complex strands). The anchoring of the leading-strand polymerase on this template strand by the β−slidingOkazaki fragment clamp allows synthesis of the leading strand to continue for many thousands of bases. By contrast, Leading strand Lagging strand because DNA synthesis occurs in the 5′ to 3′ direction, synthesis of the second DNA strand proceeds in segments using RNA primers made by the primase DnaG. This allows DNA synthesis to repeatedly initiate on the lagging strand template. Thus the lagging-strand–DNApolymerase-III complex continually associates and dissociates with the lagging-strand template to extend each RNA primer and form so-called Okazaki fragments. These Okazaki fragments are 1,000–2,000 base pairs (bp) in bacteria but only 40–300 bp in eukaryotes. β−sliding-clamps that are associated with the lagging-strand polymerase are also reloaded continually onto the lagging-strand template by the γ−complex clamp loader. A single continuous strand is formed from these discontinuous lagging strands by degradation of the RNA primers and the subsequent filling in of the resultant gaps by DNA polymerase I followed by ligation of the 5′ end of one Okazaki fragment with the 3′ end of the adjacent fragment.

blocked fork. Furthermore, the 5′ end of the lagging strand was extended three nucleotides past the 3′ end of the leading strand. However, this might be a specific property of the rDNA block. E. coli replication forks that are halted at the TUS–TER protein–DNA complex — a preprogrammed block that ensures replication forks converge within a small region of the circular bacterial chromosome on completion of replication — have a different structure28. Leading-strand synthesis halts adjacent to Tus–ter, whereas the final lagging-strand primers are located between 50 and 70 nucleotides upstream of the block, which leaves the lagging-strand template singlestranded at the fork. These examples highlight one possible distinction between the types of lesion — that a lesion affecting one template strand, but not the other, might block the polymerase but not the whole fork, possibly allowing progression of the second polymerase. However, the rDNA block and the Tus–ter complex can be thought of as ‘brick walls’ that cannot be passed by any part of the replication machinery. The fate of replication proteins at a stalled fork is equally important in understanding the nature of damaged forks. Any displacement of these proteins would necessitate their reloading. Indeed, if the DNA polymerases and the REPLICATIVE HELICASE become uncoupled25,26, reloading of the replisome might be essential. Conversely, SLIDING CLAMPS might remain bound to the daughter strands at damaged forks and prevent complete dissociation of the replisome. Which proteins require reloading at damaged forks remains to be


determined but, again, the picture might differ depending on the nature of the blockage. Possible solutions to replication-fork damage

Whatever its source, a replicative block has to be cleared or bypassed, and replication must restart. Clearance of a block implies that appropriate repair systems exist, but that they failed to deal with the lesion before the replisome arrived. In contrast, bypass of a block might be required for lesions that cannot be repaired, either because repair systems are not available, or because the repair systems cannot deal with the lesion in the context of a stalled replication fork. Removal of a protein block that is not associated with DNA damage might be achieved just as easily as in DNA that is not undergoing replication. This is because no loss of genetic information is involved, so an intact complementary strand is not needed to repair damage. So, the problem might be one of restarting replication. However, the replisome could merely pause — rather than stall irreversibly — and resume replication when the protein block has been cleared. Examples of this could include collisions between the replisome and RNA polymerases, nucleosomes or other physiologically normal protein–DNA complexes. We have little information on the nature or frequency of such events. This might be because the blocks are transient, and recombination or repair proteins are not needed to bypass them. However, such blocks might be the source of the slow-replication phenotype of E. coli rep mutants29, based on the ability of Rep HELICASE to



REVIEWS

Table 1 | Proteins implicated in repair of DNA replication forks Species

Protein

Type

Function

Eukaryotes

Mus81 complex

Endonuclease

Might cleave stalled replication forks or the Holliday junctions formed from such forks

Human

E. coli

Rad51

Homologous to RecA

Strand exchange

Translesion polymerases

Specialized DNA polymerases

Replication past lesions in the DNA template

BLM

RecQ family helicase

Mutation causes Bloom’s syndrome, with elevated levels of sister chromatid exchanges. Can unwind branched DNA structures

WRN


Mutation causes Werner’s syndrome, characterized by excessive chromosomal deletions. Can unwind branched DNA structures.

DnaB

Replicative helicase

Unwinds parental DNA strands at the head of the replication fork

DnaG

Primase

Synthesizes RNA primers for the initiation of lagging-strand DNA synthesis

PriA

Helicase

Catalyses reloading of the replisome onto branched DNA structures to allow replication to restart

RecA

Recombination enzyme

Strand exchange

RecG

Helicase

Catalyses the formation of Holliday junctions from stalled replication-fork structures. Can also move the branch point of a Holliday junction along the DNA

RecQ

Helicase

Can unwind various branched DNA structures. Reduces illegitimate recombination

Rep

Helicase

Facilitates progression of replication forks in vivo, possibly by removal of proteins bound ahead of the fork

RuvABC

Helicase–endonuclease complex

RuvAB forms a Holliday-junction-specific helicase and interacts with RuvC endonuclease, which cleaves Holliday junctions in two opposing strands and removes the Holliday junction

S. pombe

Rqh1


Mutants have a hyper-recombination phenotype

S. cerevisiae

Rrm3

Helicase

Promotes replication-fork progression

Sgs1


Mutants have a hyper-recombination phenotype. Can unwind branched DNA structures

E. coli, Escherichia coli; S. cerevisiae, Saccharomyces cerevisiae; S. pombe, Schizosaccharomyces pombe.

HETEROCHROMATIN

Chromatin — DNA packaged around nucleosomes — that is highly compacted as compared with most chromatin in a eukaryotic cell.

862

disrupt protein–DNA interactions in vitro30 (TABLE 1). The S. cerevisiae helicase Rrm3 is also implicated in the maintenance of fork progression through protein–DNA complexes31, and the WSTF–ISWI (Williams syndrome transcription factor–imitation switch protein) chromatin-remodelling complex in vertebrates might be required for the disruption of HETEROCHROMATIN to facilitate replication32. Chemical damage to the template DNA is potentially a far greater problem. Certain types of nucleotide damage — such as UV-induced pyrimidine dimers — might halt replication by preventing the incorporation of a nucleotide by the normal replicative polymerase. Others, such as interstrand crosslinks between the two template strands, could stall progression of the replicative helicase. If DNA damage is not repaired before the replication fork arrives, the lesion might also be masked by the replication machinery and compound the problem. Replication forks that are stalled in vitro retain the capacity to resume replication for several minutes33, and so access by excision-repair systems might be inhibited unless the replication enzymes are cleared from the damage. Similarly, transcription by RNA polymerases can also be inhibited by lesions in the transcribed strand34, so converting the DNA damage into a stable protein roadblock35 that might halt progression of the replication fork. How do cells clear or bypass such obstacles? Specialized DNA polymerases can incorporate nucleotides opposite sites of DNA damage that block


the normal replicative polymerases36. These ‘bypass’ — or ‘translesion’ — polymerases synthesize DNA strands to extend across nucleotide damage in the template, and then the normal replicative polymerases can reinitiate DNA synthesis. However, the misincorporation rates of these translesion polymerases are generally higher than those of the normal replicative polymerases37–39. Recombination is vital for the repair of damaged replication forks as it reduces the risk of mutagenesis by bypass polymerases. Recombination enzymes exploit the presence of two sister duplexes produced by the (damaged) replication fork to effect repair and restart. A key intermediate in restarting stalled forks has been postulated to be the four-stranded DNA (Holliday) junction. This junction is formed by unwinding the stalled fork so that the newly synthesized strands anneal and the parental strands reanneal (FIG. 2a). The popularity of models invoking Holliday junction intermediates stems from genetic and in vitro biochemical studies in bacteria and eukaryotes, and the fact that they provide attractive frameworks of how replication and recombination enzymes interact to maintain faithful chromosomal duplication. Holliday junctions formed from stalled replication forks have also been observed directly by electron microscopy40–42, although it is unclear how often such structures arise in wild-type cells under physiological conditions42. However, direct evidence for the formation of Holliday junctions from stalled forks in wild-type cells remains elusive. Indeed,



REVIEWS forks stalled at naturally occurring replication terminator sequences, as opposed to accidental stoppages, might not unwind to form Holliday junctions43. In E. coli, the formation of Holliday junctions is postulated to occur when replication is halted by defects in replicative enzymes44,45, or when cells are exposed to DNA-damaging agents such as UV light10. The subsequent processing of these junctions could be crucial in maintaining fork progression, as they might provide routes to the formation of DNA structures that allow subsequent loading of the replicative machinery. Although the molecular details of how Holliday junctions could facilitate replication through sites of DNA damage are still unclear, the genetics and enzymology of proteins that form and process Holliday junctions in E. coli are providing insights into how replication restart might occur (TABLE 1).

a

Holliday junction formation

b

RuvABC Holliday junction

e Holliday junction

D-loop

Free duplex end g

c

D-loop

Restart by Holliday junction formation

Holliday junctions formed by the unwinding of stalled replication forks are also referred to as ‘chicken feet’. These seem to be processed in one of two ways in E. coli — they are either cleaved in two opposing strands at the branch point of the junction by the RUVABC helicase/endonuclease complex44,46 and the nicks are subsequently resealed by DNA ligase, or they are processed in a manner that does not involve cleavage45,47,48. Cleavage of the Holliday junction seems to be necessary to restart replication when lesions affect both strands of the template DNA, whereas single-stranded lesions can be overcome without resorting to cleavage10.

RUVABC

A Holliday-junction-specific multisubunit helicase (RuvAB) and endonuclease (RuvC) that act in concert to move and then cleave the branch point of Holliday junctions. RECA

A recombination enzyme that catalyses strand exchange between a single strand of DNA and a homologous doublestranded DNA. D-LOOP

A recombination intermediate that is formed by the base pairing of a single-stranded DNA with a homologous double-stranded DNA; a structure in which one strand of the dsDNA is displaced to make way for the invading strand is formed. PRIA

Recognizes and loads the replication machinery at branched DNA structures in Escherichia coli.

Restart by cleavage. Two general models have been proposed to explain the order of cleavage and recombination events at a Holliday junction formed from a stalled replication fork. One possibility is that the end of the double-stranded DNA (dsDNA) that is spooled out from the junction is recombined back into the chromosome by the bacterial recombination protein 44,45 RECA to form a D-LOOP (FIG. 2b). This D-loop could then be targeted by PRIA, which, through a series of protein–protein interactions, is able to reload an active replisome onto the DNA49. Cleavage by RuvABC — of the original Holliday junction that was formed at the stalled fork and of the Holliday junction formed at the D-loop — would then remove the connection between the two sister duplexes behind the reconstituted replisome (FIG. 2d). Alternatively, the Holliday junction at the stalled fork could be cleaved by RuvABC before the DNA end is recombined back into the chromosome10 (FIG. 2e). This would release one of the sister dsDNA arms as a free DNA end (FIG. 2f), which might then be recombined back into the chromosome through D-loop formation (FIG. 2g) as above. Cleavage of the Holliday junction formed at the D-loop and reloading of the replication machinery would restore an active replication fork (FIG. 2h). In both models, the structure of the RuvABC complex might also ensure that cleavage of the Holliday junctions is biased, minimizing chromosome-segregation problems upon completion of replication50,51.


f

Free duplex end

RuvABC

h RuvABC

d

Figure 2 | Restarting DNA replication by junction cleavage. a | Formation of a Holliday junction from a replication fork stalled at a lesion (pink rectangle) might allow processing by the RuvABC helicase–endonuclease complex in one of two ways. b | First, the double-stranded DNA (dsDNA) end that is spooled out from the Holliday junction might be recombined with the homologous sequences in the reannealed parental strands to form a D-loop intermediate that is linked to the original Holliday junction at the fork. A second Holliday junction would also be formed at the Dloop. c | Assuming that the original block could be removed, cleavage of both Holliday junctions by RuvABC would generate a forked DNA structure (d) onto which the replication machinery could be reloaded. e | Second, RuvABC might cleave the Holliday junction at the stalled fork before recombination from the free dsDNA end has occurred, while formation of the Holliday junction might facilitate unmasking and subsequent removal of the block (f). g | The released dsDNA end would then be recombined back into the homologous duplex to form a D-loop. h | Cleavage of the Holliday junction formed at the D-loop by RuvABC and reassembly of the replication machinery at the fork would reconstitute an active replisome. 3′-ends of DNA strands are shown with arrowheads, cleavage of Holliday junctions by RuvABC is shown by pink arrowheads and the replication machinery is depicted as grey ovals.

In both models, the requirement for recombination of the dsDNA end that is spooled out from the fork indicates that inappropriate recombination between the dsDNA end and repeated sequences elsewhere in the genome might be a risk. However, the models differ in the timing of cleavage of the Holliday junction that is formed by unwinding of the fork. The first model (FIG. 2b–d) does not predict that one of the sister chromatids would be released as a free dsDNA, whereas the second model does (FIG. 2e–h). Release of a free dsDNA end might increase the possibility of inaccurate recombination. Unfortunately, as these models are based on analyses of mutant strains, it is not clear whether either, or both, mechanisms are used in wild-type cells.



REVIEWS

CHECKPOINT

An enzyme system that ensures events, such as DNA replication, are completed before progression to the next stage of the cell cycle in eukaryotes. RECG

A helicase that can unwind forked DNA to generate Holliday junctions, and that can also move the branch point of a Holliday junction along the DNA. EXCISION REPAIR

The removal of damaged nucleotide residues in DNA and their replacement with the correct residue.

864

Regardless of the timing of Holliday junction cleavage, it is not clear how recombination through the formation of a Holliday junction at the stalled fork can be used to repair or bypass replicative blocks. Genetic analysis indicates that RuvABC is important for repairing replication forks halted by damage to both strands of the template (such as interstrand crosslinks) or by stalled RNA polymerases10. In other words, RuvABC seems to be needed to deal with blocks that halt replication of both the leading and lagging strands simultaneously. In both of the mechanisms described above, the recombination event (FIG. 2b,g) occurs upstream of the original block. So if the block remains, the reconstituted replisome could be halted at the block a second time. One possibility is that unwinding of the stalled fork allows the replication machinery to be cleared away from the block to facilitate subsequent repair. In this model, the recombination event would function only to provide a D-loop structure for reloading of the replisome by PriA. It is also possible, however, that recombination directly facilitates bypass of the block. Although E. coli has a single circular chromosome, there are often many copies of the chromosome in a cell. So the dsDNA end that is formed via the Holliday junction might undergo strand exchange with another copy of the chromosome. This might allow replication of the damaged chromosome to continue by using an unadulterated copy as a template to replicate past the site of the block. However, this proposal raises the problem of having a replication fork reconstituted on a different copy of the chromosome. Somehow, the replicated dsDNA would have to be reincorporated into the original chromosome to form an intact, circular copy of the genome. Could such mechanisms be used in eukaryotes? Holliday junction structures have been detected at the rDNA locus in S. cerevisiae during S phase52, and they could be formed by the unwinding of replication forks that are stalled at the known replication-fork block in the rDNA53. The formation of these structures also correlates with enhanced levels of recombination54. S. cerevisiae cells that are defective in the replication CHECKPOINT response also accumulate Holliday junctions at replication forks42, and this correlates with fork stalling and chromosome breakage55. However, although these correlations are persuasive, in eukaryotes there is no direct evidence that stalled replication forks are repaired through the formation and cleavage of Holliday junctions. However, a Holliday junction-specific helicase/endonuclease activity has been detected in mammalian cells56. So, eukaryotes might have an activity that is analogous to that of E. coli RuvABC. Replication forks can also assemble at recombination intermediates in eukaryotes, presumably through the formation of D-loops57,58. The Mus81 complex, which is conserved from yeast to humans, has also been implicated in the cleavage of Holliday junctions formed from stalled forks59. Partially purified Mus81 complexes from Schizosaccharomyces pombe and humans can cleave synthetic Holliday junctions in vitro59,60. However, other studies61,62 indicate that


Mus81 might cleave stalled fork structures directly, rather than targeting Holliday junctions formed from such forks. Regardless of whether Holliday junctions or fork structures are the in vivo substrates of Mus81, this enzyme facilitates the survival of S. pombe cells after irradiation with UV light. Mus81 also promotes the survival of cells that have defective DNA polymerases, which supports a role for it in the repair of stalled replication forks63. Restart avoiding cleavage. Cleavage of — and recombination from — replication forks through the formation of Holliday junctions risks promoting inappropriate recombination between the dsDNA end formed at the fork (FIG. 2) and homologous sequences elsewhere in the genome45. This could be especially true in higher eukaryotes such as humans, where repetitive or duplicated sequences represent 5–7% of the genome64. There are, however, alternative methods of processing stalled replication forks — at least in E. coli. The free dsDNA end that is generated by the formation of a Holliday junction at a stalled fork (FIG. 2a) might be digested by exonucleases, so that a fork structure is directly reformed to allow reloading of the replisome44,45. This direct resetting of the fork would avoid the risk of inappropriate recombination. Damage to a single strand of the template DNA during replication might also be repaired or bypassed, and replication then restarted, through targeting of the stalled fork by the helicase RECG10. This is thought to be achieved without any accompanying cleavage of Holliday junction DNA, and so it might also minimize the risk of aberrant recombination. Single-stranded damage might present a different problem to double-stranded damage, as the lesion could initially affect only one of the polymerases at the replication fork. Thus, one of the newly synthesized DNA strands might be longer than the other. This is generally assumed to be a particular problem when blocks occur on the leading-strand template, causing the lagging-strand end to be extended some way beyond the leading strand25,26 (FIG. 1e). There are two potential problems with such stalled forks. First, the lesion would be located in ssDNA, so there would be no complementary strand to act as a template for EXCISION REPAIR of the damage. And second, the 3′ end of the leading strand would be displaced from the branch point of the fork, possibly by several hundred nucleotides26,65. Assuming that leading-strand synthesis cannot be primed by generating an RNA primer as per lagging-strand synthesis, the lack of a 3′ leadingstrand end at the fork would inhibit the repriming of leading-strand synthesis. How RecG might deal with blocked forks of this nature has been revealed by in vitro studies, which showed that RecG can specifically unwind forked DNA structures to form Holliday junctions10,66,67 (FIG. 3a). Furthermore, the preferred forked DNA substrate in vitro contains a lagging strand but no leading strand at the branch point47. Thus, RecG might target those forks where lagging-strand synthesis has continued some way beyond a block in leading-strand synthesis (FIG. 1b).



REVIEWS a Fork

RecG

Holliday junction

Lagging strand

c

d

Exonuclease

Leading-strand lesion

Replisome reloading

RecG

b

g

Recombination by D-loop formation via RuvABC

Polymerase decoupling

e f

Extension of leading strand

Replisome reloading + lesion bypass

Figure 3 | Restarting DNA replication without cleavage. a | RecG unwinds forked DNA structures to form Holliday junctions in vitro through coupled unwinding of the leading and lagging strands and reannealing of the parental strands. b | In vivo, RecG might preferentially unwind stalled forks blocked on the leading strand. Holliday junction formation would spool out the extended lagging strand annealed to the shorter leading strand. c | Exonuclease digestion of the single-stranded lagging strand followed by rewinding of the Holliday junction back to form a fork would generate a structure with the 3′-end of the leading strand positioned at the branch point ready to act as a primer for renewed leading-strand synthesis (d). Assuming the lesion is repaired, reloading of the replication machinery would allow replication to continue. e | Alternatively, the Holliday junction that is spooled out from such a damaged fork might sustain replication of the truncated leading strand using the extended lagging strand as a template. f | Rewinding of the Holliday junction to form a fork would locate the 3′-end of the leading strand at the fork for continued replication, while simultaneously bypassing the original lesion. g | Crosstalk between mechanisms of replication repair might also occur by RuvABC targeting Holliday junctions formed by RecG. 3′-ends of daughter strands are depicted with arrowheads, lesions are depicted by pink triangles and the replication machinery is depicted as grey ovals.

NEGATIVE SUPERCOILING

Almost all DNA molecules are negatively supercoiled. The DNA is twisted about itself in the direction opposite to that of the two strands of the double helix. POSITIVE SUPERCOILING

Positively supercoiled DNA, like negatively supercoiled DNA, has a higher energy than relaxed DNA and this energy can alter local DNA structure. However, the DNA is twisted on itself in the same direction as the two strands of the double helix.

Unwinding of a fork with a truncated leading strand would spool out the extended lagging strand and, eventually, the recessed leading strand would be reached. Continued unwinding of both lagging and leading strands by RecG would result in annealing of the two strands to form a Holliday junction (FIG. 3b), and evidence from genetic studies indicates that this Holliday junction is processed to restore a fork structure that is suitable for reloading of the replisome48. Exonucleasemediated digestion of the extended lagging strand might remove the discrepancy between the leading and lagging strands (FIG. 3c), such that if the Holliday junction is unwound far enough back to form a fork then the 3′ end of the leading strand would be repositioned at the fork, ready for priming of leading-strand synthesis (FIG. 3d). NEGATIVE SUPERCOILING in the template DNA might promote spontaneous rewinding of the Holliday junction to reform the fork66,68 (see below). Alternatively, exonucleases might continue to digest the leading and lagging strands back to the base of the fork, removing the Holliday junction without the need to unwind it. However, although unwinding of the stalled fork to form a Holliday junction might


facilitate access to (and repair of) the lesion, the original block must still be removed before replication can proceed (FIG. 3d). An alternative strategy known as template switching40 might bypass the block. The 3′ end of the leading strand in a Holliday junction formed by RecG would be annealed to the extended lagging strand. This lagging strand could be used as a template for extension of the leading strand (FIG. 3e), possibly by the DNA-damageinducible DNA polymerase II69. If the formation of the Holliday junction is reversed, the reformed fork would then have the 3′ end of the leading strand at the branch point ready for priming of renewed leading-strand synthesis (FIG. 3f). Furthermore, the leading strand would now be extended past the original block so that replication could restart, leaving the lesion to be repaired later. Unfortunately, direct evidence for template switching is lacking, and it remains to be seen whether this elegant solution is used by any organism. The formation of Holliday junctions by RecG also raises the possibility that RuvABC targets these junctions, resulting in their cleavage and recombination (FIG. 3g). However, the genetic evidence for crosstalk between RecG and RuvABC has to be interpreted with caution. Both RuvAB and RecG can unwind Holliday junction structures in vitro70,71, so both might target Holliday junctions in vivo regardless of their origin. Both helicases might, therefore, be involved in several recombination processes in which Holliday junctions arise, such as the possible repair of gaps in the lagging strand (FIG. 1d). Furthermore, the synergistic increase in the DNA-repair defect that is observed in recG ruv double-mutant strains, compared with either single mutant, indicates that RecG and RuvABC provide two alternative mechanisms of dealing with DNA damage48,72. The cell could, therefore, have at least two specialized helicase systems to deal with different types of replication-fork damage. Alternative ways to unwind stalled forks

If Holliday junctions that are targeted by RuvABC are not generally formed by RecG, what other mechanisms might promote the formation of Holliday junctions from stalled replication forks? In vitro, POSITIVE SUPERCOILING in the template DNA has been shown to drive the unwinding of stalled replication forks to form Holliday junctions66,68. In vivo, regions of positive supercoiling do occur transiently— for example, ahead of an advancing RNA polymerase73 — so head-on collisions between the replication and transcription machinery might lead to stalling of the replication fork in positively supercoiled DNA. However, chromosomal DNA is generally maintained in a net negatively supercoiled state. RecG can still promote unwinding of forks in negatively supercoiled DNA66, but negative supercoiling inhibits the spontaneous formation of Holliday junctions. So, the extent to which positive supercoiling promotes unwinding of stalled replication forks in vivo is unclear. There might be other, enzymatic means of unwinding stalled replication forks. The formation of Holliday junction substrates for RuvABC from stalled forks by



REVIEWS inhibition of the replicative helicase DNAB depends on recA in vivo74. Furthermore, in vitro, at a fork in which there is a gap in the leading strand (FIGS 1e,3b), RecA catalyses strand exchange between the single-stranded leading-strand template and the lagging-strand duplex, resulting in Holliday junction formation75. Helicases other than RecG might also catalyse the processing of stalled forks. RuvAB can unwind forked DNA in the right direction for the formation of a Holliday junction in vitro, although it preferentially unwinds in the opposite direction76. However, processing of stalled forks would not necessarily involve the formation of Holliday junctions if unwinding of only the leading or the lagging strand was used to facilitate repair and the restart of replication. One candidate for such a process is the RecQ helicase, which is involved in minimizing ILLEGITIMATE RECOMBINATION in E. coli77 and can unwind various branched DNA structures in vitro78. Invivo-labelling studies indicate that RecQ unwinds the lagging strand at stalled forks to promote the specific degradation of this strand by the RecJ exonuclease79. This degradation has been suggested to facilitate the stabilization of a stalled replication fork by promoting binding of RecA at the fork together with the RecA accessory factors RecF and RecR80. Unwinding stalled forks in eukaryotes?

DNAB

The replicative helicase of Escherichia coli that unwinds the parental double-stranded DNA ahead of the replication fork by translocating along the laggingstrand template. ILLEGITIMATE RECOMBINATION

Recombination between DNA sequences that share little or no homology. TOPOISOMERASE III

An enzyme that alters the degree of supercoiling in doublestranded DNA by the transient introduction of nicks in a single strand of the DNA. DNAG

A specialized Escherichia coli RNA polymerase, or primase, that transiently interacts with DnaB and synthesizes short stretches of RNA on the laggingstrand template. These RNAs prime lagging-strand DNA synthesis.

866

Homologues of recG have been detected in plants, but not in the available genome sequences of other eukaryotes81, and fork-unwinding activities analogous to that of RecG have yet to be detected in eukaryotes. However, the presence in eukaryotes of the RecA homologue Rad51 (REF. 82) raises the possibility that Rad51-catalysed strand exchange at stalled forks could promote formation of Holliday junctions in the same manner as is proposed for RecA. A family of helicases that are related to E. coli RecQ have also been implicated in the processing of damaged forks. In humans, disruption of these helicases causes genetic disorders, a hallmark of which are chromosomal rearrangements that can lead to a predisposition to cancer or premature ageing1. Two of these gene products — BLM and WRN (disruption of which causes Bloom’s and Werner’s syndromes, respectively) — are known to unwind branched DNA structures, including Holliday junctions, in vitro83–85. They might also be associated with the replicative apparatus in vivo86,87. Furthermore, the S. pombe RecQ homologue, Rqh1, has been implicated in the bypass of DNA damage by replication forks, on the basis of the sensitivity to UV light of rqh1 mutant cells and the requirement for Rqh1 in cells with defects in DNA synthesis88. This has led to the proposal that the apparently antirecombinogenic function of BLM might involve winding back Holliday junctions that were formed from damaged forks to re-form the fork83. This might prevent cleavage of the Holliday junction by an activity analogous to bacterial RuvC56, and so inhibit any inappropriate recombination from the free dsDNA end that would be released by this cleavage. Partial complementation of the DNA repair and mitotic defects of S. pombe rqh1


mutant cells by expression of a bacterial Holliday junction resolvase indicates that Holliday junctions formed from stalled forks might be removed by Rqh1-catalysed unwinding back to a fork. In the absence of Rqh1, however, removal of these Holliday junctions through cleavage by a bacterial resolvase can partially compensate for the inability to remove them by Rqh1. The inviability of S. pombe cells that lack both the Mus81 endonuclease complex and the Rqh1 helicase also supports the idea of two mechanisms of replication-fork repair59,62,63. By contrast, it has been suggested that the S. cerevisiae RecQ homologue — Sgs1 —unwinds stalled forks in the opposite direction to Rqh1 (REF. 61) . Unwinding of forks by Sgs1 could allow repair enzymes to access the site of a blocking lesion, whereas direct cleavage of the fork by Mus81 might release one arm of the fork as a free DNA end that can then be recombined with the sister chromatid once the original damage has been repaired. These contrasting models for the activities of Mus81 and Rqh1/Sgs1 in S. pombe and S. cerevisiae clearly depend on the nature of the target for Mus81 endonuclease, highlighting the importance of identifying the true in vivo substrate of Mus81. The proposed balance between Mus81 and Sgs1 activities in S. cerevisiae has parallels with E. coli RuvABC-dependent repair via cleavage of stalled forks, and RecG-dependent repair without cleavage. Indeed, any fork unwinding by Sgs1 might result in the formation of a Holliday junction, especially given that Sgs1 can unwind various branched DNA structures in vitro90, and could conceivably be used for template switching. However, helicases of the RecQ family, including Sgs1, seem to act as part of a complex with TOPOISOMERASE III (REFS 91–93). A role for these RecQ–TopoIII complexes in the disruption of illegitimate recombination intermediates, rather than processing of damaged replication forks, has been postulated92. Restarting replication in prokaryotes

Initiation of replication occurs routinely at pre-programmed origins through a series of specific protein –protein interactions. These events are precisely controlled so that each origin fires only once per cell cycle, and at the appropriate time94,95. Reloading of the replication machinery at sites of replication-fork damage must subvert these normal controls so that replisomes can be assembled away from normal origins and in a manner that is not subject to the same cell-cycle regulation. In bacteria, this is achieved by PriA. E. coli priA mutants are sensitive to UV light, are recombination deficient, grow slowly and have reduced viability compared with wild-type cells96,97. This highlights the importance of replication-fork reloading. PriA achieves this reloading by binding to forked DNA structures and recruiting, via a series of protein–protein interactions, the replicative helicase DnaB onto the lagging-strand template, and subsequent association of the DNAG primase with DnaB49,98. This ‘primosome’ can then unwind the parental dsDNA and synthesize primers for laggingstrand synthesis.



REVIEWS However, priming of leading-strand synthesis is also required at stalled forks. DnaG probably cannot operate on the leading-strand template, so the 3′ end of a pre-existing leading strand might be required for the repriming of leading-strand synthesis. This requirement seems to be met, as the preferred substrate for PriA in vitro is forked DNA in which there is a leading strand positioned at the branch point48,99 (FIG. 4a). So, PriA facilitates priming of both leading- and laggingstrand synthesis at sites away from the normal origin of replication, oriC, thereby circumventing the cellcycle controls that regulate the activity of the initiator protein DnaA at oriC. PriA also has a helicase function that can unwind any lagging-strand dsDNA that is present at the branch point of a stalled fork98,100. This is necessary for loading of DnaB at such a fork, as DnaB requires about 20 nucleotides of ssDNA for stable binding (FIG. 4b). However, priA mutants that specifically lack helicase function complement the recombination, viability and growth-rate defects of a priA null mutant97,101,102. This implies that replication restart in vivo can be achieved through intermediates that do not have a lagging strand at the fork. As D-loops do not have a lagging strand at the branch point, D-loop formation (FIG. 2) might be an efficient mechanism of replication restart, at least in the absence of wild-type PriA helicase activity49. However, genetic evidence indicates that stalled replication forks with a lagging strand at the branch point do arise in vivo, and that PriA and RecG helicase functions are both required for replication fork reassembly at these intermediates48. These findings indicate that there are at least two recombinational mechanisms for replication restart in E. coli48 (FIG. 5). One requires RuvABC-directed cleavage of the fork and the primosome-assembly function — but not the helicase function — of PriA (FIG. 2). The other mechanism requires the primosome-assembly and helicase functions of PriA together with RecG helicase, but it does not need RuvABC-dependent cleavage of the stalled fork (FIG. 3). The ability of PriA-helicase-deficient proteins to complement priA-null mutants could be explained by the ability of RuvABC to process damaged forks that are normally targeted by the RecG and PriA helicase activities. In the absence of PriA or RecG helicase function, forks with a lagging strand at the branch point could still form Holliday junctions, which could then be cleaved by RuvABC and recombined to form D-loops (FIG. 3g). Reloading of the replisome through helicasedeficient PriA could then occur at these D-loops. Potential overlaps between RecG- and Ruv-dependent replication restart might also explain why both ruv and recG mutant cells have a modest DNA-repair defect when compared with ruv recG cells, which are extremely sensitive to DNA-damaging agents48,72. The requirement for at least two overlapping methods of replication restart via recombination enzymes could reflect the variety of replicative blocks that cells face, and the importance of efficient and accurate replication of imperfect DNA templates.


a

PriA

Leading strand DnaB Single-stranded lagging-strand template

b

Reassembly of the replication fork

Leading strand

Reassembly of the replication fork

Lagging strand

Figure 4 | PriA-directed reloading of the replication machinery. The PriA helicase preferentially binds to forked DNA structures that have a leading strand at the branch point. Through a series of protein–protein interactions (not shown), PriA facilitates loading of the replicative helicase DnaB onto the lagging-strand template. a | At forks without a lagging strand, the single-stranded lagging-strand template can be bound directly by DnaB. b | At forks with a lagging strand, this strand must be unwound by PriA helicase to generate a singlestranded binding site for DnaB loading. Loading of DnaB is the key step in replication-fork assembly, and the preference of PriA for binding to forks with a leading strand at the branch point might ensure the availability of a leading strand 3′OH for priming of leading-strand synthesis.

The severe growth defects and sensitivity to DNAdamaging agents of E. coli cells that lack PriA indicate that PriA is the main initiator for the reassembly of damaged replication forks. However, there could be alternative restart mechanisms that are independent of PriA103. These mechanisms might depend on the Rep helicase103, indicating that Rep might also unwind some feature of damaged replication forks to facilitate replisome reloading in the absence of PriA. Restarting replication in eukaryotes

DNA damage and protein roadblocks might present similar problems for eukaryotic replication, so there could be analogous systems for replication restart to those in prokaryotes. Indeed, replication forks can be assembled at recombination intermediates in eukaryotes57,58. However, eukaryotes could also use the presence of multiple origins of replication on each chromosome, so that a fork from an adjacent origin might converge on a damaged fork to complete replication of that region. Thus, restart of replication from stalled forks might not be crucial, although the original blocking lesion would have to be removed to allow the second fork to complete replication. In support of this idea, aberrant replication-fork structures, including Holliday junctions and regions of ssDNA, arise much more frequently in yeast cells that are deficient in the checkpoint response than in wild-type cells. This indicates that replication-fork unwinding might be a pathological, rather than a physiological, response to replicative damage in eukaryotes42. Unfortunately, such a model does not explain how damaged forks that emanate from origins located near the ends of chromosomes, and which do not converge



REVIEWS

Damage to the DNA template

Translesion synthesis

Cleavage

No cleavage

Replication restart

Figure 5 | Potential mechanisms of replication restart. Recombinational repair of damaged replication forks might occur either by cleavage of the DNA at, or near, the site of the damaged fork, or by avoiding cleavage, as already shown for RuvABC and RecG, respectively. Both general mechanisms allow high-fidelity repair of damaged replication forks. The third pathway for replication-fork reactivation uses low-fidelity ‘translesion’ polymerases to replicate past the lesions in the template; this allows replication to continue in the face of extensive DNA damage, but at the cost of elevated mutation rates. Several overlapping mechanisms allow replication to proceed, despite the panoply of blocks to replication-fork progression that are thought to exist.

with another fork, could be rescued. Could the transcriptional silencing of subtelomeric DNA104 be one mechanism to reduce the potential number of replicative blocks in the form of stalled RNA polymerases at chromosome ends, perhaps to counter any problems due to the absence of a converging fork? Coordination of replication and repair processes

SOS RESPONSE

The induction of expression of a series of genes, many of which encode proteins that are involved in DNA-damage tolerance mechanisms, in response to elevated levels of DNA damage.

868

Given the complex nature of replication, and of the emerging mechanisms of replication repair, it is tempting to assume that the enzymes responsible for these processes are organized into higher-order structures. Replication in both prokaryotes and eukaryotes might occur in so-called ‘replication factories’, in which template DNA is spooled through stationary replication forks clustered into foci105,106. Recombination enzymes that are required for replication-fork repair might be colocalized with these replication factories in eukaryotes, with each focus of recombination enzymes able to repair many lesions107. However, this co-localization might be only transient, with the dynamic association and dissociation of repair enzymes occurring at sites of DNA damage when required107,108. It remains to be seen whether specific protein-targeting mechanisms exist that allow the rapid repair of damaged replication forks in eukaryotes, or whether protein diffusion in the nucleus is sufficient109. It will be interesting to see whether, in bacteria, enzymes such as PriA and RecG colocalize with replication forks in distinct foci, even in the absence of exogenous DNA damage. Another question is how a particular repair mechanism is chosen to overcome a particular replicative block. E. coli coordinates some aspects of replication-fork repair; for example, exposure of cells to UV light causes increased transcription of translesion polymerase genes through the SOS RESPONSE110. In this way, a potentially mutagenic fork-repair mechanism is limited to conditions of


increased DNA damage69,111. So, translesion synthesis might be a last-ditch attempt to restart replication of extensively damaged DNA when high-fidelity recombination repair mechanisms cannot cope (FIG. 5). Eukaryotes seem to coordinate replication and repair through signal-transduction cascades that sense the presence of DNA damage or replication stress and which regulate DNA repair, cell-cycle progression and, if all else fails, apoptosis112,113. The DNA-replication checkpoint in S. cerevisiae has also been implicated in stabilizing damaged forks, which might facilitate the resumption of replication from stalled forks after the block has been cleared114,115. Indeed, the lack of an active checkpoint leads to accumulation of Holliday junctions and ssDNA at replication forks42 and to the appearance of chromosomal breaks in slowly replicating regions of chromosomes55. Checkpoint proteins might, therefore, facilitate replication-fork progression to minimize the formation of abnormal replication intermediates such as Holliday junctions, possibly by the phosphorylation of enzymes that are involved in replication or nucleotide metabolism116–118. The checkpoint response might, therefore, be more than a damage-sensing mechanism, and could be involved directly in the normal progression of replication forks55. In this view of checkpoint control, only cells with a critical number of damaged forks trigger cell-cycle arrest or apoptosis42. Cells with fewer damaged forks would either continue replication while the stalled forks are repaired or await the arrival of the converging replisome. The finding that checkpoint proteins can both bind to63 and phosphorylate119 recombination enzymes indicates that checkpoint proteins might also have a direct role in regulating recombination activities at stalled replication forks120. Interactions between recombination and replication enzymes might also directly influence the activities of each other121. So, a complex interplay between replication, recombination and checkpoint enzymes might be required for efficient chromosomal duplication, even in the absence of exogenous DNA-damaging agents. Recombination enzymes = replication factors?

We are just beginning to scratch the surface of how damage to replication forks is dealt with in prokaryotes. In vitro reconstitution experiments are required to test the various models that have been proposed, and in vivo studies might help to ascertain the relative importance and co-ordination of different repair mechanisms. The picture is even hazier in eukaryotes. What is emerging is that replication-fork blockage does occur and, when it does, error-free repair or bypass of the block is required, followed by the restart of replication. Recombination proteins are crucial to these processes. Indeed, the main function of recombination on a frequency-of-use basis has been proposed to be replication-fork repair4. The generation of genetic diversity — the textbook view of recombination — could therefore be a mere sideshow that arose by the hijacking of replication repair enzymes during evolution. Thus, paradoxically, their main function might not be the promotion of genome diversity, but the maintenance of genome stability.



REVIEWS 1.

2. 3.

4. 5. 6.

7.

8. 9. 10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

Karow, J. K., Wu, L. & Hickson, I. D. RecQ family helicases: roles in cancer and aging. Curr. Opin. Genet. Dev. 10, 32–38 (2000). Venkitaraman, A. R. Cancer susceptibility and the functions of BRCA1 and BRCA2. Cell 108, 171–182 (2002). Maisnier-Patin, S., Nordstrom, K. & Dasgupta, S. Replication arrests during a single round of replication of the Escherichia coli chromosome in the absence of DnaC activity. Mol. Microbiol. 42, 1371–1382 (2001). The first direct measurement of the frequency with which replication forks stall in E. coli. Cox, M. M. et al. The importance of repairing stalled replication forks. Nature 404, 37–41 (2000). Kreuzer, K. N. Recombination-dependent DNA replication in phage T4. Trends Biochem. Sci. 25, 165–173 (2000). Kogoma, T. Stable DNA replication: interplay between DNA replication, homologous recombination, and transcription. Microbiol. Mol. Biol. Rev. 61, 212–238 (1997). Paques, F. & Haber, J. E. Multiple pathways of recombination induced by double-strand breaks in Saccharomyces cerevisiae. Microbiol. Mol. Biol. Rev. 63, 349–404 (1999). Lindahl, T. & Wood, R. D. Quality control by DNA repair. Science 286, 1897–1905 (1999). Kuzminov, A. Collapse and repair of replication forks in Escherichia coli. Mol. Microbiol. 16, 373–384 (1995). McGlynn, P. & Lloyd, R. G. Modulation of RNA polymerase by (p)ppGpp reveals a RecG-dependent mechanism for replication fork progression. Cell 101, 35–45 (2000). Identification of RecG as a helicase that generates Holliday junctions from damaged replication forks to assist fork progression. Liu, B. & Alberts, B. M. Head-on collision between a DNA replication apparatus and RNA polymerase transcription complex. Science 267, 1131–1137 (1995). Vilette, D., Ehrlich, S. D. & Michel, B. Transcription-induced deletions in Escherichia coli plasmids. Mol. Microbiol. 17, 493–504 (1995). Krasilnikova, M. M., Samadashwily, G. M., Krasilnikov, A. S. & Mirkin, S. M. Transcription through a simple DNA repeat blocks replication elongation. EMBO J. 17, 5095–5102 (1998). Chavez, S. et al. A protein complex containing Tho2, Hpr1, Mft1 and a novel protein, Thp2, connects transcription elongation with mitotic recombination in Saccharomyces cerevisiae. EMBO J. 19, 5824–5834 (2000). Brewer, B. J. When polymerases collide: replication and the transcriptional organization of the E. coli chromosome. Cell 53, 679–686 (1988). Park, J. S., Marr, M. T. & Roberts, J. W. E. coli transcription repair coupling factor (mfd protein) rescues arrested complexes by promoting forward translocation. Cell 109, 757–767 (2002). Fan, Q., Xu, F. & Petes, T. D. Meiosis-specific double-strand DNA breaks at the HIS4 recombination hot spot in the yeast Saccharomyces cerevisiae: control in cis and trans. Mol. Cell. Biol. 15, 1679–1688 (1995). Usdin, K. & Woodford, K. J. CGG repeats associated with DNA instability and chromosome fragility form structures that block DNA synthesis in vitro. Nucleic Acids Res. 23, 4202–4209 (1995). Samadashwily, G. M., Raca, G. & Mirkin, S. M. Trinucleotide repeats affect DNA replication in vivo. Nature Genet. 17, 298–304 (1997). Meneghini, R. & Hanawalt, P. T4-endonuclease V-sensitive sites in DNA from ultraviolet-irradiated human cells. Biochim. Biophys. Acta 425, 428–437 (1976). Rupp, W. D., Wilde, C. E., Reno, D. L. & Howard-Flanders, P. Exchanges between DNA strands in ultraviolet-irradiated Escherichia coli. J. Mol. Biol. 61, 25–44 (1971). Identification of exchanges between sister duplexes after replication of DNA in cells exposed to UV light, which indicates that single-stranded gaps left in the replicated DNA at UV-induced pyrimidine dimers might be repaired by recombination with the intact sister duplex. West, S. C., Cassuto, E. & Howard-Flanders, P. Mechanism of E. coli RecA protein directed strand exchanges in postreplication repair of DNA. Nature 294, 659–662 (1981). Gottesman, M. M., Hicks, M. L. & Gellert, M. Genetics and function of DNA ligase in Escherichia coli. J. Mol. Biol. 77, 531–547 (1973). Johnston, L. H. & Nasmyth, K. A. Saccharomyces cerevisiae cell cycle mutant cdc9 is defective in DNA ligase. Nature 274, 891–893 (1978). Svoboda, D. L. & Vos, J. M. Differential replication of a single, UV-induced lesion in the leading or lagging strand by a human cell extract: fork uncoupling or gap formation. Proc. Natl Acad. Sci. USA 92, 11975–11979 (1995). Cordeiro-Stone, M., Makhov, A. M., Zaritskaya, L. S. &

27.

28.

29.

30.

31.

32.

33.

34. 35.

36.

37. 38.

39. 40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

Griffith, J. D. Analysis of DNA replication forks encountering a pyrimidine dimer in the template to the leading strand. J. Mol. Biol. 289, 1207–1218 (1999). Gruber, M., Wellinger, R. E. & Sogo, J. M. Architecture of the replication fork stalled at the 3′ end of yeast ribosomal genes. Mol. Cell. Biol. 20, 5777–5787 (2000). Hill, T. M. & Marians, K. J. Escherichia coli Tus protein acts to arrest the progression of DNA replication forks in vitro. Proc. Natl Acad. Sci. USA 87, 2481–2485 (1990). References 25–28 provide the only descriptions of the structures of stalled replication forks. Yancey-Wrona, J. E. & Matson, S. W. Bound Lac repressor protein differentially inhibits the unwinding reactions catalyzed by DNA helicases. Nucleic Acids Res. 20, 6713–6721 (1992). Lane, H. E. & Denhardt, D. T. The rep mutation. IV. Slower movement of replication forks in Escherichia coli rep strains. J. Mol. Biol. 97, 99–112 (1975). Ivessa, A. S., Zhou, J. Q. & Zakian, V. A. The Saccharomyces Pif1p DNA helicase and the highly related Rrm3p have opposite effects on replication fork progression in ribosomal DNA. Cell 100, 479–489 (2000). Bozhenok, L., Wade, P. A. & Varga-Weisz, P. WSTF–ISWI chromatin remodeling complex targets heterochromatic replication foci. EMBO J. 21, 2231–2241 (2002). Marians, K. J., Hiasa, H., Kim, D. R. & McHenry, C. S. Role of the core DNA polymerase III subunits at the replication fork. α is the only subunit required for processive replication. J. Biol. Chem. 273, 2452–2457 (1998). Tornaletti, S. & Hanawalt, P. C. Effect of DNA lesions on transcription elongation. Biochimie 81, 139–146 (1999). Selby, C. P., Drapkin, R., Reinberg, D. & Sancar, A. RNA polymerase II stalled at a thymine dimer: footprint and effect on excision repair. Nucleic Acids Res. 25, 787–793 (1997). Pham, P., Bertram, J. G., O’Donnell, M., Woodgate, R. & Goodman, M. F. A model for SOS-lesion-targeted mutations in Escherichia coli. Nature 409, 366–370 (2001). Goodman, M. F. & Tippin, B. The expanding polymerase universe. Nature Rev. Mol. Cell Biol. 1, 101–109 (2000). Tissier, A., McDonald, J. P., Frank, E. G. & Woodgate, R. polι, a remarkably error-prone human DNA polymerase. Genes Dev. 14, 1642–1650 (2000). Baynton, I. & Fuchs, R. P. Lesions in DNA: hurdles for polymerases. Trends Biochem. Sci. 25, 74–79 (2000). Higgins, N. P., Kato, K. & Strauss, B. A model for replication repair in mammalian cells. J. Mol. Biol. 101, 417–425 (1976). The proposal of template switching as a mechanism of replication restart. Viguera, E., Hernandez, P., Krimer, D. B., Lurz, R. & Schvartzman, J. B. Visualisation of plasmid replication intermediates containing reversed forks. Nucleic Acids Res. 28, 498–503 (2000). Sogo, J. M., Lopez, M. & Foiani, M. Fork reversal and ssDNA accumulation at stalled replication forks owing to checkpoint defects. Science 297, 599–602 (2002). The direct observation of Holliday junctions and regions of ssDNA at replication forks in checkpointdeficient yeast mutants. Bidnenko, V., Ehrlich, S. D. & Michel, B. Replication fork collapse at replication terminator sequences. EMBO J. 21, 3898–3907 (2002). Seigneur, M., Bidnenko, V., Ehrlich, S. D. & Michel, B. RuvAB acts at arrested replication forks. Cell 95, 419–430 (1998). A key study, which shows that damaged replication forks form Holliday junctions in vivo. Flores, M. J., Bierne, H., Ehrlich, S. D. & Michel, B. Impairment of lagging strand synthesis triggers the formation of a RuvABC substrate at replication forks. EMBO J. 20, 619–629 (2001). Zerbib, D., Mézard, C., George, H. & West, S. C. Coordinated actions of RuvABC in Holliday junction processing. J. Mol. Biol. 281, 621–630 (1998). McGlynn, P. & Lloyd, R. G. Rescue of stalled replication forks by RecG: simultaneous translocation on the leading and lagging strand templates supports an active DNA unwinding model of fork reversal and Holliday junction formation. Proc. Natl Acad. Sci. USA 98, 8227–8234 (2001). Gregg, A. V., McGlynn, P., Jaktaji, R. P. & Lloyd, R. G. Direct rescue of stalled DNA replication forks via the combined action of PriA and RecG helicase activities. Mol. Cell 9, 241–251 (2002). Liu, J., Xu, L., Sandler, S. J. & Marians, K. J. Replication fork assembly at recombination intermediates is required for bacterial growth. Proc. Natl Acad. Sci. USA 96, 3552–3555 (1999). Cromie, G. A. & Leach, D. R. Control of crossing over. Mol. Cell 6, 815–826 (2000).


51.

52.

53.

54.

55.

56.

57.

58.

59. 60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

The realization that the orientation of the cleavage of Holliday junctions by RuvABC is not random and that the actions of RuvABC at damaged replication forks (and other forms of DNA repair) might be biased against crossover formation, therefore avoiding chromosome segregation problems. Michel, B., Recchia, G. D., Penel-Colin, M., Ehrlich, S. D. & Sherratt, D. J. Resolution of Holliday junctions by RuvABC prevents dimer formation in rep mutants and UV-irradiated cells. Mol. Microbiol. 37, 180–191 (2000). Zou, H. & Rothstein, R. Holliday junctions accumulate in replication mutants via a RecA homolog-independent mechanism. Cell 90, 87–96 (1997). Defossez, P. A. et al. Elimination of replication block protein Fob1 extends the life span of yeast mother cells. Mol. Cell 3, 447–455 (1999). Sinclair, D. A. & Guarente, L. Extrachromosomal rDNA circles — a cause of aging in yeast. Cell 91, 1033–1042 (1997). Cha, R. S. & Kleckner, N. ATR homolog Mec1 promotes fork progression, thus averting breaks in replication slow zones. Science 297, 602–606 (2002). The correlation of chromosome breakage with stalling of the replication forks in slowly replicating zones of checkpoint-deficient yeast mutants indicates that checkpoint proteins might aid in the maintenance of normal replication fork progression. Constantinou, A., Davies, A. A. & West, S. C. Branch migration and Holliday junction resolution catalyzed by activities from mammalian cells. Cell 104, 259–268 (2001). Holmes, A. M. & Haber, J. E. Double-strand break repair in yeast requires both leading and lagging strand DNA polymerases. Cell 96, 415–424 (1999). Malkova, A. et al. RAD51-independent break-induced replication to repair a broken chromosome depends on a distant enhancer site. Genes Dev. 15, 1055–1060 (2001). Boddy, M. N. et al. Mus81–Eme1 are essential components of a Holliday junction resolvase. Cell 107, 537–548 (2001). Chen, X. B. et al. Human Mus81-associated endonuclease cleaves Holliday junctions in vitro. Mol. Cell 8, 1117–1127 (2001). Kaliraman, V., Mullen, J. R., Fricke, W. M., Bastin-Shanower, S. A. & Brill, S. J. Functional overlap between Sgs1–Top3 and the Mms4–Mus81 endonuclease. Genes Dev. 15, 2730–2740 (2001). Doe, C. L., Ahn, J. S., Dixon, J. & Whitby, M. C. Mus81–Eme1 and Rqh1 involvement in processing stalled and collapsed replication forks. J. Biol. Chem. 277, 32753–32759 (2002). References 59–62 identified the Mus81 complex as a branched-DNA-specific endonuclease with a role in the repair of replication forks. Boddy, M. N. et al. Damage tolerance protein Mus81 associates with the FHA1 domain of checkpoint kinase Cds1. Mol. Cell. Biol. 20, 8758–8766 (2000). Johnson, M. E. et al. Positive selection of a gene family during the emergence of humans and African apes. Nature 413, 514–519 (2001). Meneghini, R., Cordeiro-Stone, M. & Schumacher, R. I. Size and frequency of gaps in newly synthesized DNA of xeroderma pigmentosum human cells irradiated with ultraviolet light. Biophys. J. 33, 81–92 (1981). McGlynn, P., Lloyd, R. G. & Marians, K. J. Formation of Holliday junctions by regression of nascent DNA in intermediates containing stalled replication forks: RecG stimulates regression even when the DNA is negatively supercoiled. Proc. Natl Acad. Sci. USA 98, 8235–8240 (2001). Singleton, M. R., Scaife, S. & Wigley, D. B. Structural analysis of DNA replication fork reversal by RecG. Cell 107, 79–89 (2001). The structure of RecG revealed how a single polypeptide could unwind both the leading and the lagging strands, and promote reannealing of the parental strands, to allow Holliday junction formation at stalled replication forks. Postow, L. et al. Positive torsional strain causes the formation of a four-way junction at replication forks. J. Biol. Chem. 276, 2790–2796 (2001). Rangarajan, S., Woodgate, R. & Goodman, M. F. Replication restart in UV-irradiated Escherichia coli involving pols II, III, V, PriA, RecA and RecFOR proteins. Mol. Microbiol. 43, 617–628 (2002). Parsons, C. A., Tsaneva, I., Lloyd, R. G. & West, S. C. Interaction of Escherichia coli RuvA and RuvB proteins with synthetic Holliday junctions. Proc. Natl Acad. Sci. USA 89, 5452–5456 (1992). Lloyd, R. G. & Sharples, G. J. Dissociation of synthetic Holliday junctions by E. coli RecG protein. EMBO J. 12, 17–22 (1993).



REVIEWS 72. Bolt, E. L. & Lloyd, R. G. Substrate-specificity of RusA resolvase reveals the DNA structures targeted by RuvAB and RecG in vivo. Mol. Cell 10, 187–198 (2002). 73. Liu, L. F. & Wang, J. C. Supercoiling of the DNA template during transcription. Proc. Natl Acad. Sci. USA 84, 7024–7027 (1987). 74. Seigneur, M., Ehrlich, S. D. & Michel, B. RuvABC-dependent double-strand breaks in dnaBts mutants require recA. Mol. Microbiol. 38, 565–574 (2000). 75. Robu, M. E., Inman, R. B. & Cox, M. M. RecA protein promotes the regression of stalled replication forks in vitro. Proc. Natl Acad. Sci. USA 98, 8211–8218 (2001). 76. McGlynn, P. & Lloyd, R. G. Action of RuvAB at replication fork structures. J. Biol. Chem. 276, 41938–41944 (2001). 77. Hanada, K. et al. RecQ DNA helicase is a suppressor of illegitimate recombination in Escherichia coli. Proc. Natl Acad. Sci. USA 94, 3860–3865 (1997). 78. Harmon, F. G. & Kowalczykowski, S. C. RecQ helicase, in concert with RecA and SSB proteins, initiates and disrupts DNA recombination. Genes Dev. 12, 1134–1144 (1998). 79. Courcelle, J. & Hanawalt, P. C. RecQ and RecJ process blocked replication forks prior to the resumption of replication in UV-irradiated Escherichia coli. Mol. Gen. Genet. 262, 543–551 (1999). 80. Courcelle, J., Crowley, D. J. & Hanawalt, P. C. Recovery of DNA replication in UV-irradiated Escherichia coli requires both excision repair and recF protein function. J. Bacteriol. 181, 916–922 (1999). 81. Sharples, G. J., Ingleston, S. M. & Lloyd, R. G. Holliday junction processing in bacteria: insights from the evolutionary conservation of RuvABC, RecG, and RusA. J. Bacteriol. 181, 5543–5550 (1999). 82. Aboussekhra, A., Chanet, R., Adjiri, A. & Fabre, F. Semidominant suppressors of Srs2 helicase mutations of Saccharomyces cerevisiae map in the RAD51 gene, whose sequence predicts a protein with similarities to procaryotic RecA proteins. Mol. Cell. Biol. 12, 3224–3234 (1992). 83. Karow, J. K., Constantinou, A., Li, J. L., West, S. C. & Hickson, I. D. The Bloom’s syndrome gene product promotes branch migration of Holliday junctions. Proc. Natl Acad. Sci. USA 97, 6504–6508 (2000). 84. Constantinou, A. et al. Werner’s syndrome protein (WRN) migrates Holliday junctions and co-localizes with RPA upon replication arrest. EMBO Rep. 1, 80–84 (2000). 85. Mohaghegh, P., Karow, J. K., Brosh, R. M. Jr, Bohr, V. A. & Hickson, I. D. The Bloom’s and Werner’s syndrome proteins are DNA structure-specific helicases. Nucleic Acids Res. 29, 2843–2849 (2001). 86. Yan, H., Chen, C. Y., Kobayashi, R. & Newport, J. Replication focus-forming activity 1 and the Werner syndrome gene product. Nature Genet. 19, 375–378 (1998). 87. Brosh, R. M. Jr et al. Replication protein A physically interacts with the Bloom’s syndrome protein and stimulates its helicase activity. J. Biol. Chem. 275, 23500–23508 (2000). 88. Murray, J. M., Lindsay, H. D., Munday, C. A. & Carr, A. M. Role of Schizosaccharomyces pombe RecQ homolog, recombination, and checkpoint genes in UV damage tolerance. Mol. Cell. Biol. 17, 6868–6875 (1997). 89. Doe, C. L., Dixon, J., Osman, F. & Whitby, M. C. Partial suppression of the fission yeast rqh1(-) phenotype by expression of a bacterial Holliday junction resolvase. EMBO J. 19, 2751–2762 (2000). 90. Bennett, R. J., Keck, J. L. & Wang, J. C. Binding specificity

870

determines polarity of DNA unwinding by the Sgs1 protein of S. cerevisiae. J. Mol. Biol. 289, 235–248 (1999). 91. Gangloff, S., McDonald, J. P., Bendixen, C., Arthur, L. & Rothstein, R. The yeast type I topoisomerase Top3 interacts with Sgs1, a DNA helicase homolog: a potential eukaryotic reverse gyrase. Mol. Cell. Biol. 14, 8391–8398 (1994). 92. Harmon, F. G., DiGate, R. J. & Kowalczykowski, S. C. RecQ helicase and topoisomerase III comprise a novel DNA strand passage function: a conserved mechanism for control of DNA recombination. Mol. Cell 3, 611–620 (1999). 93. Wu, L. et al. The Bloom’s syndrome gene product interacts with topoisomerase III. J. Biol. Chem. 275, 9636–9644 (2000). 94. Kato, J. & Katayama, T. Hda, a novel DnaA-related protein, regulates the replication cycle in Escherichia coli. EMBO J. 20, 4253–4262 (2001). 95. Nguyen, V. Q., Co, C. & Li, J. J. Cyclin-dependent kinases prevent DNA re-replication through multiple mechanisms. Nature 411, 1068–1073 (2001). 96. Nurse, P., Zavitz, K. H. & Marians, K. J. Inactivation of the Escherichia coli priA DNA replication protein induces the SOS response. J. Bacteriol. 173, 6686–6693 (1991). 97. Kogoma, T., Cadwell, G. W., Barnard, K. G. & Asai, T. The DNA replication priming protein, PriA, is required for homologous recombination and double-strand break repair. J. Bacteriol. 178, 1258–1264 (1996). References 96 and 97 showed the crucial links between replication and recombination, and the central role of PriA. 98. Jones, J. M. & Nakai, H. Duplex opening by primosome protein PriA for replisome assembly on a recombination intermediate. J. Mol. Biol. 289, 503–516 (1999). 99. Nurse, P., Liu, J. & Marians, K. J. Two modes of PriA binding to DNA. J. Biol. Chem. 274, 25026–25032 (1999). 100. Jones, J. M. & Nakai, H. Escherichia coli PriA helicase: fork binding orients the helicase to unwind the lagging strand side of arrested replication forks. J. Mol. Biol. 312, 935–947 (2001). 101. Zavitz, K. H. & Marians, K. J. ATPase-deficient mutants of the Escherichia coli DNA replication protein PriA are capable of catalyzing the assembly of active primosomes. J. Biol. Chem. 267, 6933–6940 (1992). 102. Sandler, S. J., Samra, H. S. & Clark, A. J. Differential suppression of priA2::kan phenotypes in Escherichia coli K-12 by mutations in priA, lexA, and dnaC. Genetics 143, 5–13 (1996). 103. Sandler, S. J. Multiple genetic pathways for restarting DNA replication forks in Escherichia coli K-12. Genetics 155, 487–497 (2000). 104. Baur, J. A., Zou, Y., Shay, J. W. & Wright, W. E. Telomere position effect in human cells. Science 292, 2075–2077 (2001). 105. Lemon, K. P. & Grossman, A. D. Movement of replicating DNA through a stationary replisome. Mol. Cell 6, 1321–1330 (2000). 106. Cook, P. R. The organization of replication and transcription. Science 284, 1790–1795 (1999). 107. Lisby, M., Rothstein, R. & Mortensen, U. H. Rad52 forms DNA repair and recombination centers during S phase. Proc. Natl Acad. Sci. USA 98, 8276–8282 (2001). 108. Essers, J. et al. Nuclear dynamics of RAD52 group homologous recombination proteins in response to DNA damage. EMBO J. 21, 2030–2037 (2002).


109. Phair, R. D. & Misteli, T. High mobility of proteins in the mammalian cell nucleus. Nature 404, 604–609 (2000). 110. Courcelle, J., Khodursky, A., Peter, B., Brown, P. O. & Hanawalt, P. C. Comparative gene expression profiles following UV exposure in wild-type and SOS-deficient Escherichia coli. Genetics 158, 41–64 (2001). 111. Wagner, J. et al. The dinB gene encodes a novel E. coli DNA polymerase, DNA pol IV, involved in mutagenesis. Mol. Cell 4, 281–286 (1999). 112. Zhou, B. B. & Elledge, S. J. The DNA damage response: putting checkpoints in perspective. Nature 408, 433–439 (2000). 113. Caspari, T. & Carr, A. M. Checkpoints: how to flag up double-strand breaks. Curr. Biol. 12, R105–R107 (2002). 114. Tercero, J. A. & Diffley, J. F. Regulation of DNA replication fork progression through damaged DNA by the Mec1/Rad53 checkpoint. Nature 412, 553–557 (2001). 115. Lopes, M. et al. The DNA replication checkpoint response stabilizes stalled replication forks. Nature 412, 557–561 (2001). References 114 and 115 showed that the checkpoint response can stabilize stalled replication forks to prevent their collapse, facilitating the eventual completion of replication. 116. Brush, G. S., Morrow, D. M., Hieter, P. & Kelly, T. J. The ATM homologue MEC1 is required for phosphorylation of replication protein A in yeast. Proc. Natl Acad. Sci. USA 93, 15075–15080 (1996). 117. Pellicioli, A. et al. Activation of Rad53 kinase in response to DNA damage and its effect in modulating phosphorylation of the lagging strand DNA polymerase. EMBO J. 18, 6561–6572 (1999). 118. Zhao, X. & Rothstein, R. The Dun1 checkpoint kinase phosphorylates and regulates the ribonucleotide reductase inhibitor Sml1. Proc. Natl Acad. Sci. USA 99, 3746–3751 (2002). 119. Bashkirov, V. I., King, J. S., Bashkirova, E. V., SchmuckliMaurer, J. & Heyer, W. D. DNA repair protein Rad55 is a terminal substrate of the DNA damage checkpoints. Mol. Cell. Biol. 20, 4393–4404 (2000). 120. Carr, A. M. Checking that replication breakdown is not terminal. Science 297, 557–558 (2002). 121. Aguilera, A. Double-strand break repair: are Rad51/RecADNA joints barriers to DNA replication? Trends Genet. 17, 318–321 (2001).

Acknowledgements Work in the authors’ laboratories is supported by the MRC (P.M. and R.G.L.) and the Wellcome Trust (R.G.L.). P.M. is a Lister Institute–Jenner Research Fellow.

Online links DATABASES The following terms in this article are linked online to: Entrez: DnaB | DnaC | DnaG | PriA | RecA | RecF | RecG | RecJ | RecQ | RecR | resolvase | Rqh1 | Rrm3 | RuvC | Sgs1 OMIM: Bloom’s syndrome | Werner’s syndrome Swissprot: ISWI | Mus81 | Rad51 | WSTF Access to this interactive links box is free online.

