Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Mol Cell. Author manuscript; available in PMC 2010 November 25.
Published in final edited form as:
PMCID: PMC2784892

Brh2 promotes a template-switching reaction enabling recombinational bypass of lesions during DNA synthesis


Accumulating evidence for Rad51-catalyzed DNA strand invasion during double-strand break repair features a 3′ single-stranded tail as the preferred substrate for reaction, but paradoxically the preferred substrate in model reactions in vitro is the 5′ end. Here we examined the Rad51-promoted 5′ end invasion reaction in the presence of Brh2, the BRCA2-family protein in Ustilago maydis. Using plasmid DNA and a homologous duplex oligonucleotide with 5′ protruding single-stranded tail as substrates, we found that Brh2 can stimulate Rad51 to promote formation of a four-stranded complement-stabilized D-loop. In this structure the incoming recessed complementary strand of the oligonucleotide has switched partners and can now prime DNA synthesis using the recipient plasmid DNA as template, circumventing a lesion that blocks elongation when the 5′ protruding tail serves as template for fill-in synthesis. We propose that template switching promoted by Brh2 provides a mechanism for recombination-mediated bypass of lesions blocking synthesis during DNA replication.

Keywords: BRCA2, Rad51, D-loops, recombination, DNA repair, lesion bypass, post replication repair, daughter strand gap repair, template switching


Multiple interdependent and autonomous repair mechanisms are in place to insure that blocks or obstacles to DNA replication can be removed or circumvented (Budzowska and Kanaar, 2009). Excision of lesions ahead of the replication fork or copying past them with translesion DNA polymerases provides a decisive strategy for removing or tolerating damage. Equally important is the homologous recombination system, which provides capabilities to bypass lesions and to reassemble DNA molecules broken during replication. Replication blocks encountered on the lagging strand template can be skipped over, allowing a new primer to be laid down ahead of the block so that the fork can continue, leaving behind the blocking lesion in a single-stranded DNA gap. A block on the leading strand template, however, would appear to present a more formidable problem due to the conception that synthesis is continuous, although there is now emerging evidence for leading strand priming and reactivation downstream of a block (Heller and Marians, 2006a). Most models imply that the leading-strand polymerase stalls while the helicase continues to unwind the fork for lagging strand synthesis, generating a single-stranded gap in the leading daughter strand ahead of the block (Heller and Marians, 2006b; Lehmann and Fuchs, 2006; Yao and O’Donnell, 2008). Fork regression coupled with template switching of the nascent strands with or without associated recombination provides means for the leading strand to extend past the block (Fig 1, schemes I and II). Alternatively, the daughter-strand gap can be repaired by recombination with the undamaged sister chromatid (Fig. 1, scheme III). This path requires homologous pairing coupled with some degree of branch migration ability to promote a template switch that will facilitate gap-filling synthesis. In prokaryotes it is thought that the RecFOR complex mediates loading of RecA recombinase into the gap for strand invasion and that RuvAB and/orRecG provide branch migration function (Kowalczykowski et al., 1994; Morimatsu and Kowalczykowski, 2003).

Figure 1
Circumventing a lesion blocking leading strand synthesis. A lesion (caret) in the leading strand template causes uncoupling of the replisome so that lagging strand synthesis continues in the absence of leading strand synthesis. For synthesis to continue ...

Eukaryotes use Rad51 for catalyzing homologous pairing. In restoring integrity of a chromosome that has suffered a double-strand break (DSB) during the course of replication, the current consensus model portrays the repair process occurring in several steps. These include preparation of DNA ends, Rad51-promoted strand invasion, and resolution of Holliday junction intermediates. A key initial step preparing the DNA for Rad51 action is the generation of 3′-single-stranded DNA tails through exonucleolytic degradation and duplex unwinding (Mimitou and Symington, 2009). Rad51 with the aid of a mediator binds to the 3′-single-stranded (ss) tails to form a nucleoprotein filament that performs strand invasion of an intact homologous duplex to form a D-loop (San Filippo et al., 2008). The 3′ end of the invading strand is then used to prime DNA synthesis on the recipient duplex template, and as a consequence replication can resume (Fig. 1, scheme IV).

An alternative conception for restoring replication fork integrity following cleavage of the fork due to encounter of a block during leading-strand synthesis is a 5′ end invasion pathway coupled with a template switch event to extend synthesis directed by the displaced strand of the D-loop (Fig. 1, scheme V). Although seemingly unorthodox when viewed in the framework of the 3′ invasion mode for DSB repair, this mode is in principle mechanistically identical to nascent daughter-strand gap repair by recombination (Fig. 1, scheme III). In each mode repair initiates with homologous pairing of the single-stranded stretch harboring the lesion and the undamaged sister homologous duplex. By branch migration and a switch in template, the recessed 3′ end of the damaged strand’s complement is repositioned and brought into register with the displaced strand of the D-loop for priming new DNA synthesis. These two steps, homologous pairing and template switching, constitute the essence of this recombination-based, lesion-bypass mechanism. Subsequently, the two modes differ only by topological disposition in that an obstacle to interwinding created by pairing of the internal gap (Fig. 1, scheme III) must be overcome before the nascent heteroduplex joint can be matured to the interwound form with Holliday junctions, while the 5′ protruding single-stranded end is unconstrained and can freely interwind to form a plectonemic joint and Holliday junction (Fig. 1, scheme V).

The evidence is unequivocal that 3′ ssDNA is generated after DSB formation in meiosis and in mitotic cells after cleavage with mega-endonucleases (Mimitou and Symington, 2009). However, in studies on DNA replication there are observations suggesting that lengthy single-stranded regions are formed ahead of or 5′ to damaged sites under conditions in which non-traversable lesions are encountered during synthesis of the nascent leading strand (Lopes et al., 2006; McInerney and O’Donnell, 2007; Pages and Fuchs, 2003). It has been established that Mus81-dependent double-strand breaks are formed in response to replication stress (Hanada et al., 2007). If the parental strand were cut near the fork, releasing the leading-strand arm, as has been demonstrated for Mus81 endonuclease with model substrates (Kaliraman et al., 2001; Matulova et al., 2009; Mazina and Mazin, 2008; Whitby et al., 2003), the recombination system would be presented with the problem of performing one-ended repair with a protruding 5′ tailed molecule, in contrast to the canonical situation in double-strand break repair of rejoining two ends with protruding 3′ tails. In the same vein, but independently of Mus81, if an interruption in the phosphodiester backbone were to be encountered during leading strand synthesis as the helicase passed over it, then the fork would collapse releasing the newly replicated leading strand chromatid. In this event the broken chromatid would quite possibly contain a protruding 5′-single-stranded tail representing the footprint of the replisome, which has been documented in the case of Xenopus to be 24 residues (Raschle et al., 2008). Other scenaros in which a 5′-single-stranded tail is produced can be easily imagined.

It is certainly possible that the ends generated under any of these circumstances would be countermanded by DNA processing factors to remove the offending lesion, resect the broken DNA, and reveal 3′ ss tails. Yet there is one functional property of Rad51 that seems consistent with the 5′ end invasion pathway as depicted. Rad51 prefers 5′-rather than 3′-single-stranded tails in strand invasion reactions performed in vitro. In studies on strand exchange using yeast or human Rad51 with traditional plasmid length linear duplex and single-stranded circular molecules as substrates as well as with DNA oligonucleotide based systems, an end preference opposite that of prokaryotic RecA protein has been noted (Baumann and West, 1997; Mazin et al., 2000; Murayama et al., 2008; Sung and Robberson, 1995). This observation has prompted the suggestion that Rad51 filament formation proceeds with polarity opposite that exhibited by prokaryotic RecA protein, although this point remains unsettled. Notwithstanding the caveat that a failsafe mechanism might render any double-strand DNA end generated by a breakage to the 3′-tailed form that is suitable for the canonical Rad51 recombination pathway (Fig. 1, scheme IV), it could be imagined that the end-preference property of Rad51 is well-suited for recombinational repair of a break releasing the leading-strand chromatid as outlined above (Fig. 1, scheme V).

In the fungus Ustilago maydis the BRCA2 homolog Brh2 provides an essential role in homologous recombination. Mutants deleted of the structural gene are defective in DSB repair, highly sensitive to ionizing radiation, blocked in meiosis, and deficient in spontaneous mitotic allelic recombination (Kojic et al., 2002). Furthermore, the brh2 mutant is completely defective in induced allelic recombination in response to irradiation with ultraviolet (UV) light (Kojic et al., 2002). UV has long been known as a potent inducer of homologous recombination, but it does not directly introduce DSBs. Rather, it is thought to cause formation of single-stranded regions during DNA synthesis when replication across a UV lesion generates gaps in the newly synthesized daughter strands. Thus, repairing single-stranded gaps by recombination is an important aspect of the postreplication repair system (Rupp et al., 1971). brh2 and other mutants defective in homologous recombination in U. maydis, are extremely sensitive to killing by UV (Kojic et al., 2003), in contrast for example to homologous recombination defective mutants in budding yeast [e.g., ref. (Schiestl et al., 1990)], prompting the suggestion that in U. maydis postreplication recombinational repair is a preferred mode for cleansing the genome of UV-induced DNA lesions (Kojic et al., 2008). If so, the question of interest is how Brh2 contributes to the process.

Brh2 provides crucial service in homologous recombination through its interaction with Rad51 (Kojic et al., 2002; Kojic et al., 2005). By facilitating loading of Rad51 on 3′ single-stranded DNA tails coated with RPA, Brh2 is instrumental as a mediator in promoting Rad51 nucleoprotein filament formation for strand invasion in the canonical manner (Yang et al., 2005). Further studies, however, suggest that the role of Brh2 might be more complex. Besides mediating delivery of Rad51, Brh2 itself has an innate ability, albeit limited, to promote a strand invasion reaction (Mazloum et al., 2008). It preferentially binds to D-loop structures (Mazloum et al., 2007) and furthermore can facilitate annealing of a second complementary single strand to the displaced strand of a D-loop forming a complement-stabilized or duplexed D-loop (Mazloum and Holloman, 2009). This latter reaction emulates the second-end capture step envisioned in current conceptions of homology-directed repair of DSBs. The capability of Brh2 to promote intermolecular DNA transactions suggests the potential for a role(s) beyond the previously considered mediator function.

In previous work we showed that Brh2 promotes duplexed D-loop formation using superhelical plasmid DNA and homologous single-stranded oligonucleotides as substrates (Mazloum and Holloman, 2009). Here we examined duplexed D-loop formation with tailed DNA substrate in reactions coupled with DNA synthesis. We confirm the preference of U. maydis Rad51 for a 5′-protruding single-stranded tail in D-loop formation and present evidence that Brh2 promotes template switching of DNA strands within these complexes to form structures favorable for re-establishing replication of the leading strand after fork collapse.


Rad51 end preference in D-loop formation

We tested whether U. maydis Rad51 exhibited an end preference in strand invasion by monitoring D-loop formation using plasmid DNA and oligonucleotide substrates. Two different types of oligonucleotide substrates were tested, both based on plasmid pBluescript II DNA sequence (2961 base pairs). First, was tested single-stranded 100 mers (ss100mer) in which the 5′-end proximal or distal 50 residues were homologous (5′ hml or 3′ hml, respectively) to the plasmid, the remaining 50 residues being heterologous (from pUC19 DNA). With ss100mer substrates it was apparent that the level of D-loop product formed with the 5′ hml oligomer was the better substrate, though it was not as active as the completely homologous ss100mer (Fig. 2A). D-loop formation has been noted previously to be promoted by Brh2 alone (Mazloum et al., 2008) and was evident in reactions here with these same substrates although no end preference was apparent. Short duplexes of 49 base pairs with protruding 5′ or 3′ ss tails of 51 residues comprised the second type of substrate. In this case Rad51 was added at suboptimal levels so that the stimulatory effect of Brh2 would be manifest. Brh2 stimulated D-loop formation with each substrate but it was again apparent that a higher yield was obtained with the 5′ tailed molecules (Fig. 2B). With addition of RPA there was no change in bias and the yield of D-loops was nearly the same. These results show that the end preference of U. maydis Rad51 for strand invasion conforms to the paradigm set by the yeast and human proteins.

Figure 2
End preference in D-loop formation. The schematic depicts the design of reaction in which superhelical plasmid pBS II DNA forms D-loops with ss or tailed 32P-labeled (star) oligomers. A. Reaction mixes contained 32P-labeled ss100mer (+strand) that is ...

Brh2 promotes formation of duplexed D-loops with 5′ tailed DNA

We showed previously that Brh2 could promote annealing of the displaced single strand of a D-loop to a second complementary strand oligonucleotide to form a complement-stabilized or duplexed D-loop (DD-loop) (Mazloum and Holloman, 2009). Annealing of the second ss oligomer to the displaced strand of the D-loop eliminates unpaired single-stranded DNA and so stabilizes the structure against dissociation by branch migration. Therefore, the diagnostic test for the duplexed D-loop is stable association of the invading single strand or the second complementary strand with plasmid DNA after endonucleolytic cleavage of the plasmid. Experimentally this can be demonstrated by showing that label on either the primary invading strand or on the second complementary strand remains stably associated with the recipient plasmid DNA after it has been cut to linear form (Mazloum and Holloman, 2009).

Using pBluescript II plasmid DNA and 5′ tailed oligomer composed of 32P-labeled ss49mer (− strand) annealed to ss100mer (+ strand) as substrate, we observed enhanced D-loop formation with Rad51 and Brh2 added together as expected (Fig. 3, lane g). But unlike the situation with simple D-loops containing a homologous single strand that dissociated rapidly after digestion of the plasmid DNA with restriction enzyme (e.g., Fig. 3, lanes l, n, p), a significant fraction of the complexes formed with the 5′ tailed oligomer by Rad51 and Brh2 together did not quantitatively dissociate after restriction enzyme digestion (Fig. 3, lane h). This indicates the presence of a complement-stabilized D-loop and implies that the 49mer (−) strand of the invading 5′-tailed molecule switched its partner (100mer + strand) and paired with the displaced D-loop strand to form a four-stranded duplexed D-loop structure. A likely mechanism is that following invasion of the single-stranded tail, the duplex portion of the 5′ tailed molecule becomes incorporated into the plasmid by branch migration with homologous sequence to form a four-stranded complex. Even in the absence of a motor to drive branch migration, formation of these duplexed D-loop structures was enabled by Brh2 alone, but was inefficient when only Rad51 was present. With a different set of substrates, namely gapped circular duplex and linear duplex DNA, it was demonstrated that Rad51 from human and fission yeast in the presence of a single-strand DNA binding protein has an innate but weak ability to promote formation of a four-stranded alpha-structure and drive branch migration with the same polarity that we find with respect to the single-stranded DNA (Murayama et al., 2008).

Figure 3
Brh2 promotes duplexed D-loop formation with tailed DNA. The schematic depicts the reactions in which 32P-ss100mer (+ strand) or 32P-labeled 49mer (−)/ss100mer (+) 5′ tailed DNA was used as substrate in D-loop reactions. After incubation ...

DNA synthesis primed from the complement-stabilizing strand

To investigate the notion that D-loops formed by Brh2-promoted uptake of 5′ tailed oligomers could be extended to a four-stranded, duplexed D-loop structure, we tested whether the recessed 49mer (−) strand from the duplex portion of the invading 5′ tailed oligomer could serve to prime DNA synthesis using the plasmid DNA as template. The rationale (Fig. 4A) was that if the 49mer (−) strand of the 5′ tailed oligomer was exchangeable with the complementary plasmid sequence, then DNA synthesized from the 49mer (−) strand as primer would have the potential for elongation to a length significantly greater than the 100mer (+) strand template. How much longer was not exactly clear, but we expected that synthesis would proceed until the plasmid DNA template could be unwound no further. The 3 kbp pBSII plasmid DNA prepared by our procedure contains 30–35 negative superhelical turns as determined by 2D electrophoresis in gels containing chloroquine (data not shown). Since untwisting DNA by 10.5 bp is compensated by a change in writhe of one superhelical turn in covalently closed circular DNA, we anticipated that a third strand of at least 300–350 residues in length could be fully accommodated in the heteroduplex region of a D-loop formed with pBSII plasmid upon removal of all the negative superhelical turns. Heteroduplex tracts longer than that might be able to form, but the length would be more and more restricted by the unfavorable introduction of positive superhelical turns and tract size would eventually reach a limit as the covalently closed circular DNA overwound to the utmost extent.

Figure 4
Extension of duplexed D-loops by DNA synthesis. A. The schematic illustrates the course of reaction in which D-loop DNA formed with 32P-ss100mer (+) or duplexed D-loop DNA formed with 32P-labeled 49mer (−)/ss100mer (+) 5′ tailed DNA was ...

To catalyze DNA synthesis, we used E. coli DNA polymerase I Klenow fragment (3′–5′ exo) (denoted as Pol) as a representative generic enzyme. When the third strand of a D-loop formed with 32P-labeled ss100mer (+ strand) was extended by Pol for a limited time in moderate salt, the mobility of the complex slowed to that expected for relaxed closed circular form (slightly faster than the trace of dimer form D-loops in the adjacent control lane lacking Pol) indicating that superhelical turns were being removed as synthesis extended (Fig. 4B, left panel). We determined that the product accumulated ranged from about 250 to 750 residues in length (Fig. 4B, note alkaline gel, right panel), suggesting that Pol was certainly capable of driving synthesis up to and even beyond the point at which all the negative superhelical turns would be removed from the plasmid template. With prolonged reaction there was even more extended DNA synthesis (Fig. 4C) with product reaching almost the full length of the plasmid. This newly synthesized DNA presumably exceeds the length that could be fully base paired with the plasmid DNA template and likely results from extrusion of the 5′ end of the elongating strand from the D-loop, viz., a migrating D-loop.

In the experiment with 5′ tailed substrates (Fig. 4D), we performed the DNA synthesis reaction for an extended period, the rationale being to provide a maximal window for Pol to access primer in the context of Brh2 and Rad51 filament. In a control reaction using Pol and 5′ tailed oligomer composed of 32P-labeled ss49mer (− strand) annealed to ss100mer (+ strand) as substrate, the labeled strand was extended the length of the 100mer template strand as evident from the shift of the 32P-label of the ss49mer to a slightly slower mobility in the alkaline gel (Fig. 4D, left panel, compare lanes a, b). Under experimental conditions in which D-loop formation with 5′ tailed oligomer and plasmid DNA was catalyzed by Rad51, followed by addition of Pol and dNTPs, there was scant extension of the 32P-labeled 49mer past the size that would be directed by the100mer (+ strand) template. However, in reactions also containing Brh2, a significant fraction of product was well-beyond 100 residues in length, implying that the 32P-labeled 49mer strand switched templates to prime synthesis off the complementary sequence in the plasmid. Products with lengths averaging 0.3–0.5 kb accumulated, but it was evident from the smear of 32P-label spreading upwards in the gel, that DNA synthesis could extend perhaps the entire circumference of the plasmid. This also appeared to be the case in reactions containing Brh2 alone (Fig. 4D, lanes n–q), although there was less total product formed by comparison. In a parallel set of simple D-loop controls in which the invading molecule was a homologous single strand (32P-labeled ss100mer instead of the 5′ tailed oligomer), it was clear that D-loops formed with Rad51 alone could effectively enable DNA synthesis (Fig. 4D, right panel). Again major product of 0.3–0.5 kb accumulated, but it was apparent that DNA synthesis could be extended the length of the plasmid. We noted that the 32P-labeled ss100mer in the absence of plasmid DNA could be extended to a length of about double in reactions with the DNA polymerase (e.g., Fig. 4D, right panel, lanes b, m). This is likely due to the 3′ end of the ss100mer molecule folding back to form a primer-template hairpin that can support DNA synthesis (denoted hp mer in the figure). In other controls it was determined that no 32P-labeled band with mobility equal to that of D-loops appeared in the absence of plasmid DNA or ATP (Fig. 4E). Taken together, these results suggest that Brh2 can promote a Rad51-catalyzed homologous pairing reaction with an invading 5′ tailed substrate to perform a four-way strand exchange, with the result that the recessed 3′ end switches templates and becomes available as a primer for extended DNA synthesis.

Brh2 promotes template switching to enable bypass of a block to replication

The idea that Brh2 could promote a template switch raised the notion that such an activity might potentially be of service in a recombination-based mechanism to enable bypass of leading strand replication blocks. As imagined in the scheme for 5′ end-initiated strand invasion (see Fig. 1, scheme V), if the recessed 3′ end of the broken arm switches partners and pairs with the displaced strand of the D-loop, DNA synthesis could ensue with this strand as template until the replication fork could be re-established. To test this idea experimentally we asked whether DNA synthesis on a 5′ tailed oligomer containing a synthesis-blocking lesion in the template strand could be extended by switching to a homologous template provided Brh2 and Rad51 were added (Fig. 5B). As a model lesion capable of blocking DNA synthesis, we chose to use an abasic nucleotide as a representative non-templating residue. In pilot studies, however, using oligomers with abasic residues, we noticed that translesion synthesis across abasic sites could be significant with the DNA polymerase we employed and that a single abasic site was not sufficient to stop accumulation of fully replicated product in bulk end-point determinations (data not shown). Therefore, we increased the abasic tract size to three and used a 5′ tailed oligomer designed so that the template 100mer (+) strand had the tract of abasic residues positioned at a point starting 8 nucleotides downstream from the 3′ end of the 49mer (−) primer strand (Fig. 5A). With this primer-template combination, DNA synthesis with Pol was effectively blocked at the abasic stretch as was apparent from the nearly quantitative conversion of the 5′-32P-labeled 49mer primer strand into 57mer product Fig. (5A), while with the control 5′ tailed molecule containing natural DNA sequence synthesis extended the full length of the 100mer template. Similarly, when the primer strand (−) was a 24mer and the abasic tract was positioned 33 nucleotides downstream from the 3′ end of the primer, DNA synthesis aborted at the abasic tract.

Figure 5
Lesion bypass during DNA synthesis enabled by Brh2. A. 5′tailed DNAs composed of 32P-labeled 49mer(−) or 24mer(−) annealed to 100mer(+) or abasic 100mer(+) were used in DNA synthesis reactions containing polymerase. The schematic ...

Having established that the abasic tract could effectively block DNA synthesis, the question posed was whether synthesis could be extended following recombination of the 5′ tailed molecule with a homologous duplex DNA (Fig. 5B). Therefore, D-loop formation was carried out with both Rad51 and Brh2 in reactions with plasmid DNA and the 5′ tailed oligomer consisting of 32P-labeled 49mer annealed to abasic or natural sequence 100mer. Then Pol and dNTPs were added and incubation continued. D-loop formation was evident in the reaction before Pol addition, and a shift in mobility of the radiolabeled D-loop band to the position corresponding to relaxed plasmid DNA after addition of Pol indicated supercoils were removed from the plasmid DNA, concomitant with DNA synthesis (Fig. 5C left panel). D-loop formation with the natural sequence 5′ tailed oligomer appeared more efficient compared to the abasic sequence (Fig. 5E, left panel) as might be expected due to the presence of three unpaired bases that could destabilize the product. When the DNA was examined under denaturing conditions (Fig. 5C, right panel) it was apparent that the bulk of the 49mer primer strand molecules had become slightly extended as expected due to aborted synthesis. But in addition it was evident that there was also extension product in the size range 0.3–0.5 kb and larger. In the absence of Brh2 and Rad51 there was no trace of extension product in this size range. These findings show that in D-loops formed by the action of Brh2 and Rad51, the 32P-labeled 49mer strand of the invading 5′ tailed molecule is capable of being extended well past the abasic tract, presumably by template switching of the 32P-labeled 49mer strand with its complementary sequence in the plasmid DNA. A similar outcome was found when the 5′ tailed molecule was prepared with a 32P-labeled 24mer strand annealed to the 100mer (Fig. 5D). In this case the abasic tract was located 33 residues downstream from the end of the primer strand. With this configuration D-loop formation was more efficient and extension product formation by DNA synthesis was about twice as efficient compared to that formed above in which the abasic tract was positioned only 8 residues from the primer 3′ end (Fig. 5E). Possibly a certain minimum length in the tract between the lesion and the 3′ end of the strand to be switched is important for establishing stability of the four-stranded junction structure. Thus, the position of the lesion with respect to the ss/ds junction could have influence on template switching of the recessed 3′ primer strand.


There are two conclusions to be drawn from this study. First, Brh2 can stimulate Rad51 to promote strand invasion of a partially duplex DNA molecule containing a 5′ single-stranded tail with a plasmid DNA recipient to form a four-stranded complement-stabilized D-loop complex in which the recessed complementary strand from the invading tailed duplex switches partners and base pairs with the displaced strand of the D-loop. Second, the switch in templates promoted by Brh2 provides a means for DNA synthesis to proceed past a lesion blocking replication, emulating recombination-mediated daughter strand gap repair.

Disturbance in the progression of replication forks is a problem faced by all dividing cells. Given the crucial importance of precision and accuracy in transmitting genetic information to daughter cells and in maintaining genome stability it is not surprising that numerous mechanisms would be in place to insure that replication can re-initiate from a stalled or broken replication fork. Indeed, accumulating evidence indicates that a variety of recombination-independent and -dependent mechanisms contribute to the progression and reinitiation of DNA synthesis. A lesion on the leading strand template blocking fork progression presents a challenging situation in the event that the locus cannot be traversed by translesion polymerases. In this event three general operational methods are imagined for how synthesis could proceed past the lesion apart from the possibility that repriming could take place downstream. If fork movement is stalled but fork integrity maintained, then replication could be reestablished following fork regression or daughter-strand gap recombination. If the fork is broken then replication could be re-established by homologous recombination.

In the fork regression mode, a lesion on the leading strand template could cause the replisome to stall with the result that lagging-strand synthesis becomes uncoupled from leading-strand synthesis. By regression of the fork and further extension of this process after template switching, the nascent leading strand could be extended by use of the nascent lagging strand as template (Higgins et al., 1976). Then after reversal of the regressed structure by branch migration the fork could be re-established with the 3′ terminus of the nascent leading strand positioned beyond the site of the lesion (Fig. 1, scheme I). Alternatively, the 5′ end of the regressed structure could be resected to reveal a 3′ ss tail, which could invade the duplex upstream of the lesion (Michel, 2000) by the canonical Rad51 pathway (Fig. 1, scheme II). After dissolution of the double Holliday junction by the RecQ helicase-topoisomerase III-Rmi complex (Mankouri and Hickson, 2007) replication could be re-established.

In the daughter-strand gap repair mode the distinguishing features are the homologous pairing of the single-strand in the gap with the sister duplex and the template switch that enables DNA synthesis to extend past the position of the lesion. In Escherichia coli daughter-strand gap repair involves the RecF pathway. The RecFOR complex mediates loading of RecA in the ssDNA gap, RecG serves as a motor to promote branch migration, RecQ is a helicase providing unwinding activities, and RecJ provides exonucleolytic processing activity (Handa et al., 2009). Analysis of components that might serve in daughter-strand gap repair in eukaryotes is less extensive (Fig. 1, scheme III), but there are numerous structural homologues and functionally related proteins that parallel the prokaryotic functions. In budding yeast experimental evidence has been obtained for a direct role for Rad51 and two other homologous recombination proteins in daughter-strand gap repair (Gangavarapu et al., 2007). One is Rad52, which mediates Rad51 loading on single-stranded DNA coated with RPA (Sung and Klein, 2006), and Rad54, an ATP-dependent DNA translocase, which enhances Rad51 function and clears Rad51 from double-stranded DNA (Heyer et al., 2006).

In the fork cleavage mode, following uncoupling of lagging and leading strand synthesis at the lesion, the leading strand sister chromatid is cut off by the action of a branch-specific endonuclease to yield a protruding 5′ ss tail. By a failsafe mechanism the 5′ end could be trimmed back and the offending lesion removed. After resection to reveal a 3′ protruding tail, Rad51 could promote strand invasion and the replication fork could be re-established (Fig. 1, scheme IV). Alternatively, as demonstrated in this study, by the preferential activity of Rad51 the 5′ end could directly invade the homologous duplex. Then Brh2 could promote the recessed 3′ end to switch templates and prime new DNA synthesis (Fig. 1, scheme V), a mechanism that is functionally equivalent to daughter-strand gap repair (Fig. 1, scheme III) except that a plectonemic rather than paranemic joint is made. In this regard it is interesting to note that Brh2 has about a 5-fold preference for binding to 3′ branch structures (the uncoupled fork in Fig. 1) compared to 5′ tails (the product after fork cleavage in Fig. 1) (Mazloum et al., 2007). A difference between the fork restoration events in scheme IV versus scheme V is that the replication-blocking lesion would be retained in the DNA, but would now be recombined into the lagging-strand sister chromatid. Having switched templates, the recessed complementary strand could prime DNA synthesis on the sister molecule, which could then extend past the locus corresponding to the site of the lesion present in the original configuration.

Ongoing studies with a number of eukaryotic systems have revealed an astonishing array of overlapping, alternative, and backup activities capable of performing and/or contributing to the key reactions in the various schemes outlined in Fig. 1. Fork regression has been demonstrated for a variety of helicases (Blastyak et al., 2007; Gari et al., 2008a; Ralf et al., 2006; Sun et al., 2008) and catalysis of branch migration activity, expected for the reversal of regressed forks has been shown for certain helicases (Bugreev et al., 2008; Gari et al., 2008b; Karow et al., 2000), the translocase Rad54 (Bugreev et al., 2006), and although weakly, even for Rad51 itself in combination with a single strand binding protein (Murayama et al., 2008). Cleavage of model forks has been shown for several structure-specific endonucleases (Fricke and Brill, 2003; Whitby et al., 2003), and processing broken DNA ends to expose 3′ ss tails as substrate for Rad51 has been established for several helicases and exonucleases (Mimitou and Symington, 2009). Holliday junction processing by dissolution (Mankouri and Hickson, 2007) or by cleavage through a number of activities has also been shown (Ip et al., 2008; Osman et al., 2003; Rouse, 2009). This veritable arsenal of activities is testament to the importance of circumventing replication-blocking obstacles.

Viewed with this perspective it might not be considered surprising that Rad51 would be endowed with a certain degree of plasticity in how strand invasion is initiated. Presented with a DNA end prepared in the orthodox manner revealing a 3′ ss tail, the repair machinery could be directed to load Rad51 for strand invasion by the canonical pathway. Alternatively, as we have shown here in Brh2-coupled reactions, strand invasion catalyzed by Rad51 through the non-canonical 5′ end pathway could be harnessed for advantage when a switch in template by the recessed 3′ end would enable DNA synthesis to extend past the site of a replication block. Aside from the absolute requirement for Brh2 in UV-induced allelic recombination (Kojic et al., 2002), which is consistent with a role in daughter-strand gap repair, the caveat should be kept in mind that there has not yet been a direct in vivo demonstration for the template switch mechanism proposed for Brh2. However, if our contention is correct that the 5′ end invasion-template switch mode of postreplication repair is equivalent to daughter strand gap repair, then it is expected that Brh2 function would be indispensable for repair of discontinuities in nascent DNA strands synthesized after UV irradiation. Such a role for Brh2 would expand its repertoire of activities and could account for the observed loss in stabilization of stalled replication forks in the absence of BRCA2 (Lomonosov et al., 2003).

It seems likely that Brh2 exerts an effect on DNA structure such that the duplex region at ss/dsDNA junctions is partially opened in the presence of a homologous DNA molecule to allow template switching. The annealing activity of Brh2 and its binding preference to D-loop DNA as well as to Holliday junctions (Mazloum et al., 2007) coupled with its limited strand exchange activity (Mazloum et al., 2008) likely provide it with capability to promote template switching. Such template switching activity might well jumpstart the innate but weak activity of Rad51 to promote polar branch migration (Murayama et al., 2008), and could well be potentiated if coupled with a branch migration motor such as Rad54 (Bugreev et al., 2006). This latter has been established as a multi-tasking machine capable of enhancing Mus81-promoted cleavage of 3′ branch structures (Matulova et al., 2009; Mazina and Mazin, 2008) as might be generated by the uncoupling of lagging- from leading-strand synthesis, and able to sweep Rad51 from DNA (Solinger et al., 2002) making the 3′ end of a invading strand accessible to prime synthesis (Li and Heyer, 2009). Interplay and concerted action of these proteins might well be intrinsic to daughter-strand gap repair and 5′ end invasion.



Rad51, RPA, and Brh2 in complex with Dss1 were purified after overproduction in E. coli as described previously (Mazloum et al., 2007). Plasmid pBluescript II SK+ (pBSII) DNA (2961 base pairs) was purified without an alkaline denaturation step as described (Mazloum et al., 2008). All DNA concentrations are expressed as moles of molecules rather than nucleotide, unless indicated otherwise. Oligonucleotides were 5′-end-labeled using [γ-32P]ATP and T4 polynucleotide kinase. Oligonucleotides were synthesized by Integrated DNA Technologies and purified by electrophoresis in 8% polyacrylamide gels. Oligonucleotide sequences used in the tailed substrates were based on pBSII residues 2–101. A. 100mer (−): 5′-TAA ATT GTA AGC GTT AAT ATT TTG TTA AAA TTC GCG TTA AAT TTT TGT TAA ATC AGC TCA TTT TTT AAC CAA TAG GCC GAA ATC GGC AAA ATC CCT TAT A-3′ B.100mer (+): 5′-TAT AAG GGA TTT TGC CGA TTT CGG CCT ATT GGT TAA AAA ATG AGC TGA TTT AAC AAA AAT TTA ACG CGA ATT TTA ACA AAA TAT TAA CGC TTA CAA TTT A-3′ C. 49mer (−): 5′-AA ATT GTA AGC GTT AAT ATT TTG TTA AAA TTC GCG TTA AAT TTT TGT TA-3′ D. 49mer (+): 5′-TAA CAA AAA TTT AAC GCG AAT TTT AAC AAA ATA TTA ACG CTT ACA ATT T-3′. E. abasic 100mer (+): 5′-TAT AAG GGA TTT TGC CGA TTT CGG CCT ATT GGT TAA AAA ASpSp SpGC TGA TTT AAC AAA AAT TTA ACG CGA ATT TTA ACA AAA TAT TAA CGC TTA CAA TTT A-3′, where Sp indicates abasic residues; F. 24mer (−) 5′-TAA ATT GTA AGC GTT AAT ATT TTG-3′. 100mer oligonucleotides designed with 50 residues of terminal sequence homologous to pBSII and 50 heterologous residues (pUC19 sequence) were as follows: 5′ homologous (5′ hml) 100mer 5′-GTT CCA AAC TGG AAC AAC ACT CAA CCC TAT CTC GGT CTA TCT TTT GAT TAC TCA TAC TCT TCC TTT TTC AAT ATT ATT GAA GCA TTT ATC AGG GTT ATT G 3′ 3′ homologous (3′ hml) 100mer 5′-GAC GCT GCC GAA TTC TGG CTT GCT AGG ACA TCT TTG CCC ACG TTG ACC CGA ATC AGC TCA TTT TTT AAC CAA TAG GCC GAA ATC GGC AAA ATC CCT TAT A-3′. Tailed DNA substrates were prepared by thermal annealing of 32P-labeled ss49mer (10 pmoles) and unlabeled ss100mer (12 pmoles) then purified by electrophoresis on an 8% polyacrylamide gel in neutral buffer.

D-loop reactions

For D-loop reactions 6 nM 32P-labeled oligomer was pre-incubated with Rad51 (800 nM unless otherwise indicated) at 37° in a buffer containing 25 mM Tris-HCl, pH 7.5, 20 mM KCl, 1 mM DTT, 2 mM CaCl2, 1 mM ATP, bovine serum albumin 100 μg per ml. Separately, 18 nM pBSII DNA was pre-incubated in the same buffer with Brh2 (400 nM unless otherwise indicated). After 15 min reactions were started by mixing the oligomer and pBSII cocktails and incubation continued for an additional 30 min at 37°. These conditions were established in previous studies to be optimum for D-loop formation using the DNA substrates employed here (Mazloum and Holloman, 2009; Mazloum et al., 2008). To measure formation of complement-stabilized, or duplexed D-loops, MgCl2 was brought to 10 mM and restriction endonuclease (EcoRI) (500 units per ml) was added for 15 min (Mazloum and Holloman, 2009). For reactions involving DNA synthesis after D-loop formation was performed as above (30 min incubation), KCl and MgCl2 were brought to 50 mM and 10 mM, respectively, then a mixture of all four deoxynucleoside triphosphates (50 μM each) and 250 units per ml E. coli DNA polymerase I Klenow fragment (3′–5′ exo, New England BioLabs) was added (concentrations refer to the final levels in the reaction). After incubation for 15 min, reaction was stopped by addition of 10 mM EDTA, 1.2% sodium dodecyl sulfate and 1.7 mg per ml proteinase K. After an additional incubation for 15 min tracking dye was added and DNA products were resolved by electrophoresis on 1% agarose gels in either neutral (pH 7.6) or denaturing (30 mM NaOH) running buffer. For analysis of primer extension products using as substrate 32P-labeled 24mer or 49mer oligonucleotide annealed to 100mer template containing abasic residues, after reaction formamide was added to 50%, the mixture brought to 95° for 2 min, then products separated by electrophoresis in gels of 8% polyacrylamide containing 7M urea. For visualization, gels were dried onto Whatman DE81 paper, exposed to phosphor storage screens (Molecular Dynamics) and processed with a Typhoon 9400 PhosphorImager (Amersham Biosciences). Quantification was performed using ImageQuaNT software (Molecular Dynamics).


We thank laboratory members Qingwen Zhou, Ninghui Mao, and Milorad Kojic for generosity in providing reagents, help, and advice. WKH extends thanks to Lorraine Symington, Columbia University for stimulating discussions, and to Mike O’Donnell, Rockefeller University, for comments on the manuscript. Support for this work was provided by grants GM42482 and GM79859 from the National Institutes of Health. Neither author received any financial interest or benefit from the results or interpretation in this report.


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


  • Baumann P, West SC. The human Rad51 protein: polarity of strand transfer and stimulation by hRP-A. EMBO J. 1997;16:5198–5206. [PubMed]
  • Blastyak A, Pinter L, Unk I, Prakash L, Prakash S, Haracska L. Yeast Rad5 protein required for postreplication repair has a DNA helicase activity specific for replication fork regression. Mol Cell. 2007;28:167–175. [PMC free article] [PubMed]
  • Budzowska M, Kanaar R. Mechanisms of dealing with DNA damage-induced replication problems. Cell Biochem Biophys. 2009;53:17–31. [PubMed]
  • Bugreev DV, Brosh RM, Jr, Mazin AV. RECQ1 possesses DNA branch migration activity. J Biol Chem. 2008;283:20231–20242. [PubMed]
  • Bugreev DV, Mazina OM, Mazin AV. Rad54 protein promotes branch migration of Holliday junctions. Nature. 2006;442:590–593. [PubMed]
  • Fricke WM, Brill SJ. Slx1-Slx4 is a second structure-specific endonuclease functionally redundant with Sgs1-Top3. Genes Dev. 2003;17:1768–1778. [PubMed]
  • Gangavarapu V, Prakash S, Prakash L. Requirement of RAD52 group genes for postreplication repair of UV-damaged DNA in Saccharomyces cerevisiae. Mol Cell Biol. 2007;27:7758–7764. [PMC free article] [PubMed]
  • Gari K, Decaillet C, Delannoy M, Wu L, Constantinou A. Remodeling of DNA replication structures by the branch point translocase FANCM. Proc Natl Acad Sci U S A. 2008a;105:16107–16112. [PubMed]
  • Gari K, Decaillet C, Stasiak AZ, Stasiak A, Constantinou A. The Fanconi anemia protein FANCM can promote branch migration of Holliday junctions and replication forks. Mol Cell. 2008b;29:141–148. [PubMed]
  • Hanada K, Budzowska M, Davies SL, van Drunen E, Onizawa H, Beverloo HB, Maas A, Essers J, Hickson ID, Kanaar R. The structure-specific endonuclease Mus81 contributes to replication restart by generating double-strand DNA breaks. Nat Struct Mol Biol. 2007;14:1096–1104. [PubMed]
  • Handa N, Morimatsu K, Lovett ST, Kowalczykowski SC. Reconstitution of initial steps of dsDNA break repair by the RecF pathway of E. coli. Genes Dev. 2009;23:1234–1245. [PubMed]
  • Heller RC, Marians KJ. Replication fork reactivation downstream of a blocked nascent leading strand. Nature. 2006a;439:557–562. [PubMed]
  • Heller RC, Marians KJ. Replisome assembly and the direct restart of stalled replication forks. Nat Rev Mol Cell Biol. 2006b;7:932–943. [PubMed]
  • Heyer WD, Li X, Rolfsmeier M, Zhang XP. Rad54: the Swiss Army knife of homologous recombination? Nucleic Acids Res. 2006;34:4115–4125. [PMC free article] [PubMed]
  • Higgins NP, Kato K, Strauss B. A model for replication repair in mammalian cells. J Mol Biol. 1976;101:417–425. [PubMed]
  • Ip SC, Rass U, Blanco MG, Flynn HR, Skehel JM, West SC. Identification of Holliday junction resolvases from humans and yeast. Nature. 2008;456:357–361. [PubMed]
  • Kaliraman V, Mullen JR, Fricke WM, Bastin-Shanower SA, Brill SJ. Functional overlap between Sgs1-Top3 and the Mms4-Mus81 endonuclease. Genes Dev. 2001;15:2730–2740. [PubMed]
  • Karow JK, Constantinou A, Li JL, West SC, Hickson ID. The Bloom’s syndrome gene product promotes branch migration of holliday junctions. Proc Natl Acad Sci U S A. 2000;97:6504–6508. [PubMed]
  • Kojic M, Kostrub CF, Buchman AR, Holloman WK. BRCA2 homolog required for proficiency in DNA repair, recombination, and genome stability in Ustilago maydis. Mol Cell. 2002;10:683–691. [PubMed]
  • Kojic M, Mao N, Zhou Q, Lisby M, Holloman WK. Compensatory role for Rad52 during recombinational repair in Ustilago maydis. Mol Microbiol. 2008;67:1156–1168. [PubMed]
  • Kojic M, Yang H, Kostrub CF, Pavletich NP, Holloman WK. The BRCA2-interacting protein DSS1 is vital for DNA repair, recombination, and genome stability in Ustilago maydis. Mol Cell. 2003;12:1043–1049. [PubMed]
  • Kojic M, Zhou Q, Lisby M, Holloman WK. Brh2-Dss1 interplay enables properly controlled recombination in Ustilago maydis. Mol Cell Biol. 2005;25:2547–2557. [PMC free article] [PubMed]
  • Kowalczykowski SC, Dixon DA, Eggleston AK, Lauder SD, Rehrauer WM. Biochemistry of homologous recombination in Escherichia coli. Microbiol Rev. 1994;58:401–465. [PMC free article] [PubMed]
  • Lehmann AR, Fuchs RP. Gaps and forks in DNA replication: Rediscovering old models. DNA Repair (Amst) 2006;5:1495–1498. [PubMed]
  • Li X, Heyer WD. RAD54 controls access to the invading 3′-OH end after RAD51-mediated DNA strand invasion in homologous recombination in Saccharomyces cerevisiae. Nucleic Acids Res. 2009;37:638–646. [PMC free article] [PubMed]
  • Lomonosov M, Anand S, Sangrithi M, Davies R, Venkitaraman AR. Stabilization of stalled DNA replication forks by the BRCA2 breast cancer susceptibility protein. Genes Dev. 2003;17:3017–3022. [PubMed]
  • Lopes M, Foiani M, Sogo JM. Multiple mechanisms control chromosome integrity after replication fork uncoupling and restart at irreparable UV lesions. Mol Cell. 2006;21:15–27. [PubMed]
  • Mankouri HW, Hickson ID. The RecQ helicase-topoisomerase III-Rmi1 complex: a DNA structure-specific ‘dissolvasome’? Trends Biochem Sci. 2007;32:538–546. [PubMed]
  • Matulova P, Marini V, Burgess RC, Sisakova A, Kwon Y, Rothstein R, Sung P, Krejci L. Cooperativity of Mus81.Mms4 with Rad54 in the resolution of recombination and replication intermediates. J Biol Chem. 2009;284:7733–7745. [PMC free article] [PubMed]
  • Mazin AV, Zaitseva E, Sung P, Kowalczykowski SC. Tailed duplex DNA is the preferred substrate for Rad51 protein-mediated homologous pairing. EMBO J. 2000;19:1148–1156. [PubMed]
  • Mazina OM, Mazin AV. Human Rad54 protein stimulates human Mus81-Eme1 endonuclease. Proc Natl Acad Sci U S A. 2008;105:18249–18254. [PubMed]
  • Mazloum N, Holloman WK. Second-end capture in DNA double-strand break repair promoted by Brh2 protein of Ustilago maydis. Mol Cell. 2009;33:160–170. [PMC free article] [PubMed]
  • Mazloum N, Zhou Q, Holloman WK. DNA binding, annealing, and strand exchange activities of Brh2 protein from Ustilago maydis. Biochemistry. 2007;46:7163–7173. [PMC free article] [PubMed]
  • Mazloum N, Zhou Q, Holloman WK. D-loop formation by Brh2 protein of Ustilago maydis. Proc Natl Acad Sci U S A. 2008;105:524–529. [PubMed]
  • McInerney P, O’Donnell M. Replisome fate upon encountering a leading strand block and clearance from DNA by recombination proteins. J Biol Chem. 2007;282:25903–25916. [PubMed]
  • Michel B. Replication fork arrest and DNA recombination. Trends Biochem Sci. 2000;25:173–178. [PubMed]
  • Mimitou EP, Symington LS. Nucleases and helicases take center stage in homologous recombination. Trends Biochem Sci 2009 [PubMed]
  • Morimatsu K, Kowalczykowski SC. RecFOR proteins load RecA protein onto gapped DNA to accelerate DNA strand exchange: a universal step of recombinational repair. Mol Cell. 2003;11:1337–1347. [PubMed]
  • Murayama Y, Kurokawa Y, Mayanagi K, Iwasaki H. Formation and branch migration of Holliday junctions mediated by eukaryotic recombinases. Nature. 2008;451:1018–1021. [PubMed]
  • Osman F, Dixon J, Doe CL, Whitby MC. Generating crossovers by resolution of nicked Holliday junctions: a role for Mus81-Eme1 in meiosis. Mol Cell. 2003;12:761–774. [PubMed]
  • Pages V, Fuchs RP. Uncoupling of leading- and lagging-strand DNA replication during lesion bypass in vivo. Science. 2003;300:1300–1303. [PubMed]
  • Ralf C, Hickson ID, Wu L. The Bloom’s syndrome helicase can promote the regression of a model replication fork. J Biol Chem. 2006;281:22839–22846. [PubMed]
  • Raschle M, Knipsheer P, Enoiu M, Angelov T, Sun J, Griffith JD, Ellenberger TE, Scharer OD, Walter JC. Mechanism of replication-coupled DNA interstrand crosslink repair. Cell. 2008;134:969–980. [PMC free article] [PubMed]
  • Rouse J. Control of genome stability by SLX protein complexes. Biochem Soc Trans. 2009;37:495–510. [PubMed]
  • Rupp WD, Wilde CE, 3rd, Reno DL, Howard-Flanders P. Exchanges between DNA strands in ultraviolet-irradiated Escherichia coli. J Mol Biol. 1971;61:25–44. [PubMed]
  • San Filippo J, Sung P, Klein H. Mechanism of eukaryotic homologous recombination. Annu Rev Biochem. 2008;77:229–257. [PubMed]
  • Schiestl RH, Prakash S, Prakash L. The SRS2 suppressor of rad6 mutations of Saccharomyces cerevisiae acts by channeling DNA lesions into the RAD52 DNA repair pathway. Genetics. 1990;124:817–831. [PubMed]
  • Solinger JA, Kiianitsa K, Heyer WD. Rad54, a Swi2/Snf2-like recombinational repair protein, disassembles Rad51:dsDNA filaments. Mol Cell. 2002;10:1175–1188. [PubMed]
  • Sun W, Nandi S, Osman F, Ahn JS, Jakovleska J, Lorenz A, Whitby MC. The FANCM ortholog Fml1 promotes recombination at stalled replication forks and limits crossing over during DNA double-strand break repair. Mol Cell. 2008;32:118–128. [PMC free article] [PubMed]
  • Sung P, Klein H. Mechanism of homologous recombination: mediators and helicases take on regulatory functions. Nat Rev Mol Cell Biol. 2006;7:739–750. [PubMed]
  • Sung P, Robberson DL. DNA strand exchange mediated by a RAD51-ssDNA nucleoprotein filament with polarity opposite to that of RecA. Cell. 1995;82:453–461. [PubMed]
  • Whitby MC, Osman F, Dixon J. Cleavage of model replication forks by fission yeast Mus81-Eme1 and budding yeast Mus81-Mms4. J Biol Chem. 2003;278:6928–6935. [PubMed]
  • Yang H, Li Q, Fan J, Holloman WK, Pavletich NP. The BRCA2 homologue Brh2 nucleates RAD51 filament formation at a dsDNA-ssDNA junction. Nature. 2005;433:653–657. [PubMed]
  • Yao NY, O’Donnell M. Replisome dynamics and use of DNA trombone loops to bypass replication blocks. Mol Biosyst. 2008;4:1075–1084. [PMC free article] [PubMed]