• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of geneticsGeneticsCurrent IssueInformation for AuthorsEditorial BoardSubscribeSubmit a Manuscript
Genetics. Nov 2005; 171(3): 873–883.
PMCID: PMC1456856

The RuvAB Branch Migration Translocase and RecU Holliday Junction Resolvase Are Required for Double-Stranded DNA Break Repair in Bacillus subtilis


In models of Escherichia coli recombination and DNA repair, the RuvABC complex directs the branch migration and resolution of Holliday junction DNA. To probe the validity of the E. coli paradigm, we examined the impact of mutations in ΔruvAB and ΔrecU (a ruvC functional analog) on DNA repair. Under standard transformation conditions we failed to construct ΔruvAB ΔrecG, ΔrecU ΔruvAB, ΔrecU ΔrecG, or ΔrecU ΔrecJ strains. However, ΔruvAB could be combined with addAB (recBCD), recF, recH, ΔrecS, ΔrecQ, and ΔrecJ mutations. The ΔruvAB and ΔrecU mutations rendered cells extremely sensitive to DNA-damaging agents, although less sensitive than a ΔrecA strain. When damaged cells were analyzed, we found that RecU was recruited to defined double-stranded DNA breaks (DSBs) and colocalized with RecN. RecU localized to these centers at a later time point during DSB repair, and formation was dependent on RuvAB. In addition, expression of RecU in an E. coli ruvC mutant restored full resistance to UV light only when the ruvAB genes were present. The results demonstrate that, as with E. coli RuvABC, RuvAB targets RecU to recombination intermediates and that all three proteins are required for repair of DSBs arising from lesions in chromosomal DNA.

IN all organisms, structural aberrations in the DNA template or strand breaks induce arrest or collapse of replication forks and their restoration relies on recombination functions (Haber 1999; Kuzminov 1999; Cox et al. 2000; Michel et al. 2004). In Escherichia coli, stalled forks can reverse to form a four-stranded Holliday junction (HJ) intermediate (Seigneur et al. 1998). Fork regression, which might also occur spontaneously, involves RecG or potentially RecA, the latter loaded onto single-stranded DNA (ssDNA) by the RecFOR complex (Robu et al. 2001, 2004; Singleton et al. 2001; McGlynn and Lloyd 2002a,b). Once formed, the HJ can be processed in a number of ways: (i) The extruded duplex end can be removed by either RecBCD or RecQ and RecJ to reset the fork; (ii) DNA synthesis on the extruded partial duplex end followed by restoration of the fork by RecG or RuvAB branch migration provides a means of translesion bypass; and (iii) branch migration away from a block or lesion and HJ resolution by RuvC generates a broken fork (as is the case when the replisome encounters a strand break). RecA then mediates invasion of this broken end into the intact chromosome arm to rebuild the replication fork (Haber 1999; Kuzminov 1999; Cox et al. 2000; Michel et al. 2004). Mechanisms for direct fork rescue, which do not invoke the formation of a HJ intermediate, have also been proposed and rely on the action of RecFOR, RecJ, and RecQ recombinases (Courcelle and Hanawalt 1999; Courcelle et al. 2001; Donaldson et al. 2004).

The models for recombination-dependent replication highlight the important role played by RecBCD, RecQ, RecJ, and RecFOR in processing the ends of collapsed forks and loading of RecA (Clark and Sandler 1994; Kowalczykowski and Eggleston 1994; Kuzminov 1999; Amundsen and Smith 2003; Michel et al. 2004). RecBCD preferentially degrades double-stranded DNA (dsDNA) ends to expose a 3′ single-stranded tail. Similar reactions can be catalyzed by unwinding the end using RecQ helicase coupled with strand removal by the RecJ 5′-3′ exonuclease (Courcelle et al. 2001; Amundsen and Smith 2003). RecA can be loaded directly onto this resected ssDNA by RecBCD (Anderson and Kowalczykowski 1997, 2000; Chedin and Kowalczykowski 2002; Amundsen and Smith 2003; Xu and Marians 2003) or by the RecFOR complex when the strand is coated with Single-stranded DNA binding (SSB) protein (Umezu and Kolodner 1994; Shan et al. 1997; Kantake et al. 2002; Ivancic-Bace et al. 2003). Formation of a RecA nucleoprotein filament allows homologous pairing and strand exchange between the broken end and its undamaged partner. Invasion of the homologous duplex by the processed 3′-tail creates a D-loop upon which the replication apparatus can be reloaded by PriA (Kowalczykowski 2000; Marians 2000; McGlynn and Lloyd 2002a,b). At this stage the chromosomes are still interconnected so further extension of the DNA joint to form a HJ is needed so that RuvABC resolution can complete the fork restoration process.

In Bacillus subtilis the recombination genes, other than recA, which is central to all pathways of recombinational repair, have been placed into six different epistatic groups: α [comprising recF, recL, recO, and recR (recFLOR) and recN], β (addA and addB), γ (recP and recH), ε (ruvA, ruvB, recD, and recU), ζ (recS), and η (recG) (Alonso et al. 1993; Fernandez et al. 1998, 1999, 2000; Chedin et al. 2000; Ayora et al. 2004; Carrasco et al. 2004). Throughout this article, unless stated otherwise, the indicated genes and products refer to those of B. subtilis origin. The recA, recF, recG, recJ, recN, recO, recQ, recR, ruvA, and ruvB genes each have a homologous counterpart in E. coli with the same gene designation. In addition, the addAB and recU genes encode functional equivalents of E. coli recBCD (recBCDEco) and ruvCEco genes, respectively (Fernandez et al. 2000; Ayora et al. 2004). However, several recombination genes (recL, recD, recH, and recP) have no known equivalent in E. coli and, along with recS (a RecQ-like helicase), remain uncharacterized (Fernandez et al. 1998). Hence products classified within the α, β, ε, and η groups have their E. coli counterparts in RecN-FOR, RecBCD, RuvABC, and RecG, respectively (Ayora et al. 2004; Carrasco et al. 2004; Kidane et al. 2004), while those within the γ and ζ epistatic groups have yet to be assigned a function in DNA repair and recombination (Fernandez et al. 2000). Additionally, genetic analysis has not been undertaken with RecJ and RecQ and so neither has been assigned to any of these groupings.

Many of these functions and pathways of recombination resemble those encountered in the E. coli system. In wild-type cells, the loading of RecA protein onto SSB-coated ssDNA (presynaptic step) relies on AddAB or RecN-RecFLOR proteins (Chedin et al. 2000; Kidane et al. 2004). Recently it has been shown that: (i) ~35% of the cells in a ΔrecA and ~5% in a ΔrecU mutant contain unrepaired DSBs under normal growth conditions (Kidane et al. 2004); (ii) the ruvA gene complements the defect of the recB2 mutation classified within the ε epistatic group [hence recB2 was renamed ruvA2 (Ayora et al. 2004)]; (iii) purified RecU protein binds preferentially to three- and four-strand junctions and cleaves Holliday junction substrates to produce nicked duplexes (Ayora et al. 2004); and (iv) in the absence of the RuvAB or RecG branch-migration activities, RecU and the poorly characterized RecD bias exchange toward crossovers (CO) (Carrasco et al. 2004).

To shed light on the importance of HJ branch migration and resolution in B. subtilis, we constructed a ruvAB null mutant (ΔruvAB) and analyzed its sensitivity to different DNA-damaging agents. The ΔruvAB, ΔrecG, and recF strains showed similar sensitivities to DNA damage, which were significantly increased when ΔruvAB was combined with addAB, recF, recH, or ΔrecJ. Previously it was shown that in the absence of RuvAB, RecU, or RecG a clear chromosomal segregation defect was observed (Carrasco et al. 2004). Under standard transformation conditions we failed to construct ΔruvAB ΔrecG, ΔruvAB ΔrecU, and ΔrecU ΔrecG double-mutant strains. Expression of RecU could replace the repair function of RuvCEco in the heterologous E. coli system if the RuvABEco complex were present. Formation of RecU foci on nucleoids also required the presence of RuvAB. Our data support the notion that RuvAB works in concert with the junction-resolving enzyme RecU, in a similar manner to the RuvABCEco resolvasome complex, and that RuvAB, RecD, and RecU play a vital role in DNA DSB repair.


Bacterial strains and plasmids:

All B. subtilis strains used in this study are listed in Table 1 and are isogenic to strain YB886 (rec+ control). A 2-kb six-cat-six cassette containing two directly repeated copies of the β-site-specific recombinase target site (six) surrounding the chloramphenicol acetyl transferase gene (cat) (Carrasco et al. 2004) was introduced within the coding sequence of recJ, recQ, recG, and ruvA ruvB to generate the recJ:six-cat-six, recQ:six-cat-six, recG:six-cat-six, and ruvAB:six-cat-six disruptions. These disruptions were transferred into the chromosome of wild-type cells to generate ΔrecJ, ΔrecQ, ΔrecG, and ΔruvAB strains, and expression of the β-gene provoked deletion of the cat gene. The null ΔrecU or ΔruvAB mutation was transferred into the isogenic rec-deficient derivatives and the double mutants generated by a double CO event as previously described (Alonso et al. 1993).

B. subtilis strains used in this study

Chromosomal DNA from ΔruvAB (ruvAB:six-cat-six), ΔrecU (recU:six-spc-six), or ΔrecS (recS:cat) strains were used to transform the wild type and the mutants ΔruvAB, ΔrecU, ΔrecG, or ΔrecJ strains with selection for chloramphenicol (conferred by the cat gene) or spectinomicyn (conferred by the spc gene) resistance. The ΔrecS mutation could be transferred into all transformed strains, showing that the mutant strains are competent for transformation. An equivalent amount of chromosomal DNA from ΔruvAB or ΔrecU mutants transforms wild-type cells with similar efficiency, but no bona fide transformants were obtained for the ΔruvAB, ΔrecU, ΔrecG, or ΔrecJ mutant strain. Few tiny colonies after 72 hr of incubation times were obtained, detailed analysis of several of these transformants suggested that some of them contained a single CO with one copy of the wild type and one of the mutant gene, and a few with a double CO contained suppressor mutations (see below).

E. coli K12 ruv mutants strains, SR2210 (ruvA200), N1057 (ruvB4), N2057 (ruvAB60::Tn10), GS1481 (ΔruvC::kan), CS85 (ruvC53 eda::Tn10), AM888 (ΔruvAC65 ΔrusA::kan), and N4454 (ΔruvABC::cat) are derivatives of the ruv+ wild-type strain, AB1157 (Sargentini and Smith 1989; Sharples et al. 1990; Mandal et al. 1993; Mahdi et al. 1996; Seigneur et al. 1998). Plasmid pCB564 was constructed by transferring the 1.2-kb BspEI-BamHI DNA segment containing the recU gene into pHP13 (pRecU). pCB593 contains the 4.9-kb StuI (ruvA) fragment inserted into pUC18 (pRuvA), pCB594 the 4.6-kb HindIII (ruvB) fragment inserted into pUC18 (pRuvB), and pCB559 the 5.5-kb BamHI-EcoRI (ruvAB) fragment inserted into pUC18 (pRuvAB). A pUC18 clone carrying RecU (pFC204) was obtained by polymerase chain reaction (PCR) from pCB564 using 5′-AGAATTCTAAGGAGGATGAGATAATGATTC-3′ and 5′-TCTGACATAGGATCCCAACCTTTCG and EcoRI and BamHI restriction endonucleases (underlined). To create a C-terminal fusion of RecU with YFP for single crossover integration into the chromosome, the 3′ region (500 bp) of recU was amplified by PCR using primers 5′-ATCGGGCCCTCGCGGAATGACCCTCG-3′ and 5′-CTAGAATTCACCTTTCGCACCAGATGATG-3′ and was cloned into ApaI and EcoRI sites of plasmid pYSG that carries yfp and cat genes and a xylose promoter for transcription of downstream genes (D. Kidane and P. L. Graumann, unpublished data), resulting in pYDK6. By transforming pyDK6 into PY79, we created the strain DK53. To move the recU-yfp fusion in different mutant backgrounds, DK53 was transformed with chromosomal DNA from the ΔrecN strain, giving strain DK55, and the ΔruvAB strain was transformed with chromosomal DNA of strain DK53, resulting in strain DK56. For the colocalization experiments, strain DK53 was transformed with chromosomal DNA from recN-cfp, generating strain DK54. Plasmid pGS739 was constructed by transferring the 1.7-kb BglII (SstII)-EcoRV fragment containing the ruvC gene (obtained from a derivative of pFB512; Benson et al. 1988) into pACYC184 cleaved with BamHI and HincII. This clone does not fully restore UV resistance to ruv mutants, possibly due to a slight negative effect on cell survival after UV exposure. Other plasmids used were pGS762, pGS711, and pPVA101 (Sharples et al. 1990; Sharples and Lloyd 1991).

Viability test:

B. subtilis recombination-deficient strains were plated and incubated in Luria broth (LB) medium overnight. At least six independent colonies from each strain were resuspended in fresh LB medium and shaken for 30 min to minimize aggregation. Appropriate dilutions were plated on LB and colony-forming units (CFUs) were counted or stained with membrane-permeable SYTO 9 and membrane-impermeable propidium iodide and subjected to conventional direct count of total cells. SYTO 9, which labels bacteria with green fluorescence, and propidium iodide, which stains membrane-compromised bacteria with red fluorescence, were purchased from Molecular Probes (Leiden, The Netherlands).

DNA repair survival studies:

Exponentially growing B. subtilis cells were obtained by inoculating overnight cultures in fresh LB media and grown to an A560nm of 0.4 at 37°. These were exposed to 10 mm methyl methanesulfonate (MMS) and the fraction surviving was determined with reference to an unexposed control plate. Alternatively, the sensitivity to MMS, 4-nitroquinoline-1-oxide (4NQO), or mitomycin C (MMC) was determined by growing cultures to an A560nm of 0.4 and spotting 10 μl of serial 10-fold dilutions (1 × 10−2 to 1 × 10−5) on LB agar supplemented with the indicated concentrations of the DNA-damaging agent and incubating overnight at 37°.

E. coli strains carrying appropriate clones were measured for UV resistance by growing cells in LB media to an A650nm of 0.4 and spotting appropriate dilutions onto agar plates. These were exposed to UV light at a dose rate of 1 J/m2/sec and the fraction surviving was determined with reference to an unirradiated control plate.

DNA lesions generated by MMS, which reacts with single reactive groups in adenine (N3-alkyladenine), guanine (N7-alkylguanine) and 4NQO, which is a potent mutagen that induces two main guanine adducts at positions C8 and N2 in damaged dsDNA or ssDNA. MMC results in the formation of interstrand crosslinks and UV light primarily induces pyrimidine dimers. All of these lesions act as DNA replication roadblocks, inducing replication fork arrest and DSBs.

Image acquisition:

Fluorescence microscopy was performed on an Olympus AX70 microscope. Cells were grown in minimal medium and were mounted on agarose pads containing S750 medium on object slides as described in Kidane et al. (2004). Images were acquired with a digital MicroMax CCD camera; signal intensities were measured using the META-MORPH 4.6 program. DNA was stained with 4′,6 diamidino-2-phenylindole (DAPI; final concentration 0.2 ng/ml) and membrane with FM4-64 (final concentration 1 nm).


Defects in the α, ε, and η epistatic groups render cells extremely sensitive to DNA-damaging agents:

To gain insight into the involvement of HJ processing in the repair of DNA damage, we constructed a null ΔruvAB mutant strain and analyzed its phenotype in parallel with mutations in the α (recF15, recL16, ΔrecO, ΔrecR), β (addA5 addB72, termed here addAB), γ (recH342), ε (recD41, ΔrecU, ΔruvAB), ζ (ΔrecS, ΔrecQ, ΔrecJ), and η (ΔrecG) epistatic groups as well as in the ΔrecA strain (Table 1).

The recombination-deficient cells, when present in an otherwise Rec+ strain, were exposed to the killing action of alkyl groups generated by MMS and their phenotypes were recorded. The ΔrecS, ΔrecQ, and ΔrecJ cells (epistatic group ζ) showed a similar degree of sensitivity to MMS; hence, only the former is shown. Figure 1 shows that ΔrecS, addAB, and recH342 cells displayed a moderate and/or sensitive phenotype to the killing action of 10 μm MMS when compared to the wild-type control.

Figure 1.
Survival of strains after exposure to MMS. The strains used, which are identified by the indicated relevant genotype, were exposed to 10 mm MMS for a variable time.

The recF15, recL16, ΔrecO, and ΔrecR cells (group α) showed a similar degree of sensitivity to MMS or 4NQO (Alonso et al. 1993); hence, only the sensitivity of the former mutant strain is shown. The recF15, ΔrecG, recD41, ΔrecU, ΔruvAB, and ΔrecA cells were extremely sensitive to the killing action of 10 mm MMS when compared to the wild-type control. The recF15 (group α), recD41, ΔrecU, ΔruvAB (group ε), and ΔrecG (group η) strains were less sensitive to 10 mm MMS than the ΔrecA strain (Figure 1).

Branch migration and resolution of Holliday junctions is essential for DNA repair:

Previously, it was shown that addAB, recH342, and ΔrecS mutations increased the sensitivity of ΔrecU cells to DNA damage (Fernandez et al. 1998). Furthermore, strains lacking RecF, RecU, or both are extremely sensitive to MMS and 4NQO (Alonso et al. 1993). These results indicate that RecA assembly factors (e.g., RecFLOR) and the RecU HJ resolvase facilitate repair of DSBs (Fernandez et al. 1998; Ayora et al. 2004). We therefore investigated whether RecU was required for repair of DSBs generated by different DNA-damaging agents. The ΔrecU null mutation was transferred into representatives from each of the epistatic groups (α, recF15 and ΔrecO; β, addA5 addB72; γ, recH342; and ζ, ΔrecS and ΔrecQ strains), but we were unable to recover the ΔrecU allele in ΔruvAB (epistatic group ε), ΔrecG (η), and ΔrecJ (ζ) backgrounds without the appearance of undesired mutations. In our attempt to construct ΔrecU ΔruvB, ΔrecU ΔrecG, and ΔrecU ΔrecJ double mutants, we obtained a few colonies after prolonged incubation. Analysis of several of these transformants suggested that they contained either single CO or suppressor mutations (e.g., ΔrecU ΔrecG sms; results not shown). To confirm that no other unselected mutations accumulate in these strains, DNA from a plasmid-borne recG:six-cat-six was used to transform B. subtilis BG501 (ΔrecU Δsms) competent cells selecting for chloramphenicol resistance. Using this approach, we succeeded in making a Δsms ΔrecU ΔrecG strain. This fits with the earlier observations that Δsms (also termed ΔradA) partially suppresses the ΔrecU defect (Carrasco et al. 2002).

Unlike Streptococcus pneumoniae in which the recU gene (TIGR SP0370) is apparently essential (Thanassi et al. 2002), a B. subtilis ΔrecU mutant is viable, although it grows poorly (Table 2) and accumulates suppressor mutations at a high frequency (Pedersen and Setlow 2000; Carrasco et al. 2002, 2004). Therefore, the ΔruvAB null mutation was transferred into representatives from the different epistatic groups (α, recF15 and ΔrecO; β, addA5 addB72; γ, recH342; ε, recD41; and ζ, ΔrecS, ΔrecQ, and ΔrecJ strains), the double and triple mutants were exposed to MMS, 4NQO, or MMC, and their phenotypes were recorded.

Viability of ΔrecA recombination-deficient mutants

In the absence of any DNA-damaging agent, the number of viable cells per colony of strains grouped in the α, β, γ, or ζ epistatic group was affected <1.5-fold when compared to wild-type cells (data not shown), whereas the ΔruvAB and ΔrecA strains showed a similar reduced number of viable cells per colony (4- to 5-fold) when compared to the wild-type strain (Table 2). Exponentially growing cells were stained with SYTO 9, and only ~4% of these wild-type cells were also stained with propidium iodide (an indicator of membrane-compromised “dead” bacteria). The proportion of ΔruvAB and ΔrecA cells stained with propidium iodide increased 3- to 4-fold when compared with wild-type cells (Table 2). A similar reduced number of viable cells per colony was observed when the ΔrecU or ΔrecG cells were analyzed (Table 2).

The ΔruvAB cells were extremely sensitive to 10 μg/ml of MMS, 0.75 μg/ml of 4NQO, or 12 ng/ml of MMC (Figure 2), whereas the wild-type strain showed a minimal defect in the presence of 250 μg/ml MMS, 24 μg/ml of 4NQO, or 150 ng/ml of MMC relative to an unexposed control (data not shown). The DNA damage sensitivity of ΔruvAB recD41 cells (epistatic group ε) was similar to that obtained with the ΔruvAB mutant strain (Figure 2). The recombination mutants classified within the β (addAB) and ζ (ΔrecS) groups marginally increased the sensitivity of ΔruvAB cells following exposure to 10 μg/ml MMS, 0.75 μg/ml of 4NQO, or 12 ng/ml of MMC (Figure 2), but they were slightly less sensitive than ΔrecA cells. It is likely that acting, in concert, RecQ and RecJ initiate DNA recombination in ΔruvAB cells.

Figure 2.
Survival of ΔruvAB mutants in combination with mutations from other epistatic groups after exposure to DNA-damaging agents. The strains used are identified by the indicated relevant genotype. A serial 10-fold dilution (10-μl sample) of ...

The recF15 and the poorly characterized recH342 mutation reduced the survival of ΔruvAB cells to a level comparable to the ΔrecA strain following exposure to 5 μg/ml MMS, 0.35 μg/ml of 4NQO, or 6 ng/ml of MMC (Figure 2). The ΔruvAB mutation did not increase the sensitivity of the ΔruvAB ΔrecA strain (Figure 2).

RecU resolves a HJ by endonucleolytic cleavage (Ayora et al. 2004), while it is believed that RuvAB, perhaps in concert with the unknown activity associated with RecD, recognizes and branch migrates HJs. The recU, recD, ruvA, and ruvB mutants all belong to the ε epistatic group. Since we failed to construct a ΔrecU ΔruvAB strain, yet successfully made ΔrecA ΔrecU (Carrasco et al. 2004), ΔrecA ΔruvAB (this work), and ΔrecA ΔrecG strains (H. Sanchez and J. C. Alonso, unpublished results), we propose that the ΔrecU ΔruvAB combination leads to accumulation of “toxic” recombination intermediates during strain construction.

RecU restores UV resistance to E. coli ruvC mutants:

Recently it was demonstrated that RecU protein binds three- and four-stranded DNA branches, resolves Holliday junctions, and promotes joint molecule and D-loop formation in vitro (Ayora et al. 2004). Furthermore, the structure of RecU, which shows a striking similarity to a class of resolvase enzymes found in archaea and members of the type II restriction endonuclease family, was determined (McGregor et al. 2005). To confirm the involvement of RuvAB and RecU in HJ processing, we examined their ability to replace the activities of their counterparts in the well-characterized E. coli system. Plasmid-borne recU, ruvA, ruvB, or ruvAB genes were introduced into various E. coli ruv mutant combinations and exposed to varying doses of UV light (Figure 3 and Table 3). Plasmids carrying RecU restored full UV resistance to strains (ΔruvCEco and ruvC53Eco) deficient in the RuvCEco HJ resolvase (Figure 3; Table 3; data not shown). RecU also conferred resistance to MMC at 0.2 and 0.5 μg/ml in these strains (data not shown). The results reveal for the first time that RecU functions as a HJ resolvase in vivo. Significantly, RecU only partially improved the UV sensitive phenotype in E. coli ruvA, ruvB, ruvAB, ruvAC, and ruvABC mutants (Figure 3; Table 3; data not shown). Because RecU is as effective as RuvCEco at promoting repair in a ruvCEco mutant, the results establish that, as with RuvCEco, RecU depends on RuvAB branch migration for efficient HJ resolution.

Figure 3.
Survival of UV-irradiated E. coli ruv mutants carrying the recU gene. The strains used were N2057 (ruvAB), GS1481 (ruvC), N4454 (ruvABC), and AB1157 (ruv+). Symbols for plasmids carrying RecU (pFC204), RuvCEco (pGS762), RuvABEco (pGS711), RuvABC ...
Effect of plasmids carrying RecU on the survival of UV-irradiated ruvEco mutants

The plasmid constructs carrying RuvA, RuvB, or RuvAB were unable to improve the UV sensitivity of the relevant ruvEco mutants (Table 3). Both RuvA and RuvAB clones produced an obvious negative effect on wild-type E. coli cells exposed to UV light. The results suggest that unlike RecU, RuvAB cannot replace the function of RuvABEco and that this may, in part, be due to a detrimental effect of RuvA expression. In fact, plasmids expressing high levels of RuvAEco are known to confer an extreme negative effect on the UV sensitivity of wild-type cells (Sharples et al. 1990). This is probably a consequence of RuvA binding HJ DNA and preventing access of alternative junction processing enzymes such as RuvCEco or RecGEco. These effects strengthen the argument that RuvAB cannot work properly with RuvCEco, rather than an expression and/or stability problem with the heterologous RuvAB complex.

We attempted to construct a plasmid carrying all three B. subtilis HJ processing genes to test RuvAB and RecU functionality directly in an E. coli ruvABC-deficient strain. However, the clones obtained had suffered substantial deletions, indicating that this combination is highly deleterious. Similarly, we were unable to maintain pCB559 (RuvAB) and pCB564 (RecU) jointly in a strain lacking ruvAC (effectively a ruvABC mutant) and the cryptic HJ resolvase rusA (Mahdi et al. 1996; data not shown). It seems likely that too much RuvAB and RecU generates frequent lethal DSBs at regressed replication forks or additionally blocks their processing even in the absence of DNA-damaging agents.

RecU forms discrete foci on nucleoids after induction of DSBs and colocalizes with RecN:

Previously, it was shown that RecN, RecO, and RecF proteins accumulate in discrete foci following induction of DSBs (Kidane et al. 2004). RecN foci were detected 15–20 min after treatment with MMC, RecO foci were first visible 30 min after induction, while RecF foci were not observed until after ~60–90 min (Kidane et al. 2004; our unpublished results).

A RecU-GFP fusion strain was constructed and localization of the RecU protein was investigated in the presence or absence of MMC. The fusion was the sole source of RecU in the cell and fully supported repair of DNA following addition of MMC, showing that the fusion retained activity. In exponentially growing cells, RecU-GFP was present throughout the cells (Figure 4A), with fluorescence levels barely above background. However, after addition of 100 ng/ml of MMC, RecU-GFP formed discrete foci in up to 45% of the >500 cells analyzed (Figure 4C). RecU-GFP foci were always present on the nucleoids (see arrowheads in Figure 4C) and cells generally contained a single focus; only 2.7% of the cells contained two foci. One hour after the addition of MMC, clear foci were observed in only 1.5% of the cells (Figure 4B), with the highest number of foci observed 120 min after induction of DSBs (Figure 4C), and became increasingly fewer and fainter thereafter. The foci therefore occurred at a later point during DSB repair than RecN or RecO foci.

Figure 4.
Subcellular localization of RecU in live B. subtilis cells. (A) RecU-GFP in exponentially growing cells; (B) 60 min after addition of MMC; (C) 120 min after addition of MMC. Open arrowheads in C indicate RecU-GFP foci. (D) RecU-GFP in ΔruvAB cells, ...

To establish that RecU is recruited to the RecNOF DSB repair centers (RCs), we generated a RecU-YFP variant and combined it with a RecN-CFP fusion, such that both were simultaneously expressed within cells. Many dually labeled cells showed rather patchy areas on the nucleoids (shaded arrow, Figure 4F); only 10% of the cells showed clear RecU-YFP and RecN-CFP foci after MMC treatment (open arrowheads, Figure 4F), mostly because RecN-CFP fluorescence was extremely low. The formation of patches in many cells suggests that GFP labels on both proteins slightly interfere with the proper function of the proteins, although the single labels are fully functional. However, in all of the cells with clear foci, both RecU-YFP and RecN-CFP signals were coincident, and in most cells containing fluorescent patches these signals were likewise at similar places within the cells, showing that RecN and RecU colocalize within the DSB RCs.

To test if accumulation of RecU requires other proteins, we moved the RecU-YFP fusion into a ruvAB or recN mutant background. Only 0.4% of the cells showed RecU-YFP foci in the absence of ruvAB (Figure 4D), while 37% of the recN mutant cells contained RecU-YFP foci after addition of MMC (Figure 4E). Interestingly, 25% of the recN mutant cells contained two foci, rather than one (Figure 4E). As with RecU-YFP, RecN-GFP forms two foci in only 4% of the cells and a single focus in the remaining cells (Kidane et al. 2004). These experiments demonstrate that RecU is part of a dynamic response to DSBs in B. subtilis cells and is recruited into defined RCs at a late stage in a reaction dependent on RuvAB. RecU is recruited to RCs independently of RecN in agreement with data showing that different avenues can lead to the formation of crossovers that are the substrate for RecU. However, on the basis of our previous findings suggesting that several DSBs are recruited to and repaired within a single RC (Kidane et al. 2004), it is clear that RecN is a candidate for a factor combining different breaks into a single RC, because of the increase in the number of RecU-YFP foci in the absence of RecN.


This work provides evidence that inactivation of genes in epistatic group β (addAB) or in epistatic group ζ (recQ, recS, or recJ), when present in otherwise Rec+ cells, have rather modest effects on sensitivity to MMS. In contrast, elimination of those functions classified within the α (namely recF15, recL16, ΔrecO, or ΔrecR), ε (ΔruvAB, recD41, and ΔrecU) or η (ΔrecG) epistatic groups shows a dramatic reduction in ability to repair DNA damage mediated by these agents, showing only slightly more resistance than the recombination-defective ΔrecA strain. Previously it was shown that RecS shares 36 and 34% identity with RecQ and RecQEco proteins, respectively (Fernandez et al. 1998). This homology is significantly greater (43 and 40% identity) if only the regions containing the seven conserved DExH-box DNA helicase motifs (the first 330 residues of RecQ and RecS) are compared (Fernandez et al. 1998). It was shown that ΔrecS does substitute for the ΔrecQ defect as the double mutant is as sensitive as the single parent mutant (our unpublished results). RecQEco unwinds both partially dsDNA and fully duplex DNA with a 3′ to 5′ polarity, while RecJEco is a 5′ to 3′ ssDNA exonuclease, generating a 3′-terminated end subsequently coated with SSBEco (Kowalczykowski and Eggleston 1994; Courcelle and Hanawalt 1999; Amundsen and Smith 2003). We have therefore tentatively placed recJ and recQ within the ζ epistatic group, together with recS, in a recombination pathway that can generate 3′-tailed ssDNA at broken forks akin to the activities of AddAB or RecBCD.

Since the DNA-damaging agents used in this report act as replication roadblocks, inducing replication fork arrest and single-strand nicks or DSBs, we considered the possibility that replication restart in ΔruvAB cells relies on the processing of DNA ends by the action of AddAB (Chedin et al. 2000) or by the combined action of the RecQ or RecS helicase and the RecJ ssDNA exonuclease. RecA protein could be loaded on the 3′-ssDNA by the AddAB enzyme (Chedin et al. 2000) or RecN-RecFLOR complex (Kidane et al. 2004). This is consistent with the observation that the ΔrecU ΔrecJ strain did not seem to be viable and that RecN forms RCs at DSBs in concert with RecO and RecF (Kidane et al. 2004). RecA bound to ssDNA promotes homologous pairing, D-loop formation, and strand exchange between one or both of the broken ends with an intact DNA molecule to generate HJs. RecG or RuvAB (alone or in concert with RecD) branch migrates these junctions for RecU resolution. This model fits with the observations that: (i) recF addAB cells are impaired in DNA repair and genetic recombination to the level of recA cells and showed a similarly reduced viability in the absence of external damage, together with extreme sensitivity to the killing action of MMS, 4NQO, or MMC (Alonso et al. 1993), and (ii) recF or ΔrecO mutations reduce the viability of ΔruvAB cells to a greater extent than do addAB mutations (see Figure 2; our unpublished results).

To confirm the functionality of RuvAB and RecU in HJ processing, we studied their ability to complement the DNA repair defect of E. coli ruv mutants. We found that expression of the recU gene restores UV-light resistance to ruvCEco strains to a level similar to that of clones carrying RuvCEco. In contrast, RecU conferred only a slight improvement in UV survival of ruvAEco, ruvBEco, ruvACEco, ruvABEco, or ruvABCEco mutants. Since the E. coli ruv system is well defined, we can conclude that RecU does indeed function as a HJ resolvase as demonstrated by in vitro data (Ayora et al. 2004; McGregor et al. 2005). The improvement in resistance to UV when RecU is present in strains lacking ruvABEco indicates that it can function to some extent in the absence of RuvAB. However, this may be artificially high due to overexpression of RecU, since clones carrying RuvCEco also improve the survival of ruvABEco strains following exposure to UV light. The dependence on RuvAB for full DNA repair activity does suggest that RecU normally functions together with the branch migration complex as is the case with E. coli RuvABC (Zerbib et al. 1998; van Gool et al. 1999). Any contacts that stabilize a RuvABCEco or a RuvAB-RecU complex must be conserved between these heterologous systems, if indeed they are important for stability of the tripartite complex. Consequently, the resolvasome model, where the resolution endonuclease (either RecU or RuvCEco) scans the junction for preferred target sequences, appears to be widely conserved in bacteria.

RuvABEco or RecGEco catalyze replication fork regression in vivo and play a critical role in promoting the recovery of replication when it is blocked by DNA damage (Bolt and Lloyd 2002; Gregg et al. 2002; Meddows et al. 2004). Other studies, however, indicate that RuvABEco- or RecGEco-catalyzed fork regression is not essential for DNA synthesis to resume following arrest by UV-induced DNA damage in vivo (Donaldson et al. 2004). In this work, we also show that it is possible to visualize the place of action of RecU in live cells. We have found that RecU forms a single discrete center on the nucleoid upon induction of DSBs, as previously observed with RecNOF proteins (Kidane et al. 2004). RecN is the first to form the RCs, within 15–20 min with foci visible at defined DSBs in live cells (Kidane et al. 2004). RecU is recruited into RCs, since it colocalizes with RecN. Consistent with a role in resolution of HJs, RecU accumulated within the RCs after the formation of RecN, RecO, or RecF foci; RecU foci were clearly visible 120 min after induction of DSBs. These data indicate that repair of DSBs occurs over a long period of time during which several sequential processes take place. Recruitment of RecU was dependent on RuvAB proteins, strengthening the view that these proteins form a resolvasome complex. Interestingly, the number of RCs was increased in the absence of RecN protein, supporting our suggestion that RecN might organize different recombination events within a single center (Kidane et al. 2004).

Under standard transformation conditions we failed to construct ΔruvAB ΔrecG, ΔruvAB ΔrecU, and ΔrecU ΔrecG mutant strains. Previously we reported the construction of ruvA2 recU40 double-mutant strains (Alonso et al. 1992) and here report the construction of the recD41 ΔruvAB strain. However, the recU40 strain is proficient in plasmid transformation and shows a doubling time similar to the wild-type strain, whereas a ΔrecU is significantly impaired in plasmid transformation and shows a marked growth defect (Fernandez et al. 1998; Pedersen and Setlow 2000; Carrasco et al. 2002; Table 2). It is likely that the recU40 allele may encode only a partially defective HJ resolvase. Very little information is available concerning the recD41 strain. These two pieces of apparently conflicting data argue that RecU resolves HJ intermediates branch migrated by either RuvAB(RecD) or RecG DNA helicases. The improvements in UV resistance conferred upon ruvABEco mutants by clones carrying RuvCEco and RecU support this idea, confirming that both these HJ resolvases can function in vivo without RuvAB branch migration. Alternatively, the apparent lethality of ΔrecG ΔruvAB, ΔrecG ΔrecU, and ΔruvAB ΔrecU double mutants arises from accumulation of “toxic” recombination intermediates (Gangloff et al. 2000). This is consistent with the observation that ΔrecA ΔrecU, ΔrecA ΔrecG (Carrasco et al. 2004), and ΔrecA ΔruvAB were viable, albeit with a 14- to 25-fold reduced plating efficiency (see Table 2). In E. coli, recG ruvAB, recG ruvC, and ruvABC mutants are viable, the latter indistinguishable from single-mutant strains (Lloyd 1991; Mandal et al. 1993). There are clearly important differences between Gram-negative and Gram-positive recombinational repair processes despite the apparent similarities in coordination of HJ resolution by RuvAB-RecU and RuvABCEco. This serves to highlight the importance of having more than one model system to evaluate the mechanics of complex repair, replication, and recombination processes.


We are grateful to Begoña Carrasco for the communication of unpublished results. This work was supported by grants BMC2003-00150 and BIO2001-4342-E from Dirección General de Investigación (J.C.A.), from the Deutsche Forschungsgemeinschaft (P.L.G.), and from the Biotechnology and Biological Sciences Research Council (G.J.S.).


  • Alonso, J. C., R. H. Tailor and G. Luder, 1988. Characterization of recombination-deficient mutants of Bacillus subtilis. J. Bacteriol. 170: 3001–3007. [PMC free article] [PubMed]
  • Alonso, J. C., G. Luder and T. A. Trautner, 1992. Intramolecular homologous recombination in Bacillus subtilis 168. Mol. Gen. Genet. 236: 60–64. [PubMed]
  • Alonso, J. C., A. C. Stiege and G. Luder, 1993. Genetic recombination in Bacillus subtilis 168: effect of recN, recF, recH and addAB mutations on DNA repair and recombination. Mol. Gen. Genet. 239: 129–136. [PubMed]
  • Amundsen, S. K., and G. R. Smith, 2003. Interchangeable parts of the Escherichia coli recombination machinery. Cell 112: 741–744. [PubMed]
  • Anderson, D. G., and S. C. Kowalczykowski, 1997. The translocating RecBCD enzyme stimulates recombination by directing RecA protein onto ssDNA in a chi-regulated manner. Cell 90: 77–86. [PubMed]
  • Arnold, D. A., and S. C. Kowalczykowski, 2000. Facilitated loading of RecA protein is essential to recombination by RecBCD enzyme. J. Biol. Chem. 275: 12261–12265. [PubMed]
  • Ayora, S., B. Carrasco, E. Doncel, R. Lurz and J. C. Alonso, 2004. Bacillus subtilis RecU protein cleaves Holliday junctions and anneals single-stranded DNA. Proc. Natl. Acad. Sci. USA 101: 452–457. [PMC free article] [PubMed]
  • Benson, F. E., G. T. Illing, G. J. Sharples and R. G. Lloyd, 1988. Nucleotide sequencing of the ruv region of Escherichia coli K-12 reveals a LexA regulated operon encoding two genes. Nucleic Acids Res. 16: 1541–1549. [PMC free article] [PubMed]
  • Bolt, E. L., and R. G. Lloyd, 2002. Substrate specificity of RusA resolvase reveals the DNA structures targeted by RuvAB and RecG in vivo. Mol. Cell 10: 187–198. [PubMed]
  • Carrasco, B., S. Fernandez, K. Asai, N. Ogasawara and J. C. Alonso, 2002. Effect of the recU suppressors sms and subA on DNA repair and homologous recombination in Bacillus subtilis. Mol. Genet. Genomics 266: 899–906. [PubMed]
  • Carrasco, B., M. C. Cozar, R. Lurz, J. C. Alonso and S. Ayora, 2004. Genetic recombination in Bacillus subtilis 168: contribution of Holliday junction-processing functions in chromosome segregation. J. Bacteriol. 186: 5557–5566. [PMC free article] [PubMed]
  • Ceglowski, P., G. Luder and J. C. Alonso, 1990. Genetic analysis of recE activities in Bacillus subtilis. Mol. Gen. Genet. 222: 441–445. [PubMed]
  • Chedin, F., and S. C. Kowalczykowski, 2002. A novel family of regulated helicases/nucleases from Gram-positive bacteria: insights into the initiation of DNA recombination. Mol. Microbiol. 43: 823–834. [PubMed]
  • Chedin, F., S. D. Ehrlich and S. C. Kowalczykowski, 2000. The Bacillus subtilis AddAB helicase/nuclease is regulated by its cognate chi sequence in vitro. J. Mol. Biol. 298: 7–20. [PubMed]
  • Clark, A. J., and S. J. Sandler, 1994. Homologous genetic recombination: the pieces begin to fall into place. Crit. Rev. Microbiol. 20: 125–142. [PubMed]
  • Courcelle, J., and P. C. Hanawalt, 1999. RecQ and RecJ process blocked replication forks prior to the resumption of replication in UV-irradiated Escherichia coli. Mol. Gen. Genet. 262: 543–551. [PubMed]
  • Courcelle, J., A. K. Ganesan and P. C. Hanawalt, 2001. Therefore, what are recombination proteins there for? BioEssays 23: 463–470. [PubMed]
  • Cox, M. M., M. F. Goodman, K. N. Kreuzer, D. J. Sherratt, S. J. Sandler et al., 2000. The importance of repairing stalled replication forks. Nature 404: 37–41. [PubMed]
  • Donaldson, J. R., C. T. Courcelle and J. Courcelle, 2004. RuvAB and RecG are not essential for the recovery of DNA synthesis following UV-induced DNA damage in Escherichia coli. Genetics 166: 1631–1640. [PMC free article] [PubMed]
  • Fernandez, S., A. Sorokin and J. C. Alonso, 1998. Genetic recombination in Bacillus subtilis 168: effects of recU and recS mutations on DNA repair and homologous recombination. J. Bacteriol. 180: 3405–3409. [PMC free article] [PubMed]
  • Fernandez, S., Y. Kobayashi, N. Ogasawara and J. C. Alonso, 1999. Analysis of the Bacillus subtilis recO gene: RecO forms part of the RecFLOR function. Mol. Gen. Genet. 261: 567–573. [PubMed]
  • Fernandez, S., S. Ayora and J. C. Alonso, 2000. Bacillus subtilis homologous recombination: genes and products. Res. Microbiol. 151: 481–486. [PubMed]
  • Gangloff, S., C. Soustelle and F. Fabre, 2000. Homologous recombination is responsible for cell death in the absence of the Sgs1 and Srs2 helicases. Nat. Genet. 25: 192–194. [PubMed]
  • Gregg, A. V., P. McGlynn, R. P. Jaktaji and R. G. Lloyd, 2002. Direct rescue of stalled DNA replication forks via the combined action of PriA and RecG helicase activities. Mol. Cell 9: 241–251. [PubMed]
  • Haber, J. E., 1999. DNA recombination: the replication connection. Trends Biochem. Sci. 24: 271–275. [PubMed]
  • Ivancic-Bace, I., P. Peharec, S. Moslavac, N. Skrobot, E. Salaj-Smic et al., 2003. RecFOR function is required for DNA repair and recombination in a RecA loading-deficient recB mutant of Escherichia coli. Genetics 163: 485–494. [PMC free article] [PubMed]
  • Kantake, N., M. V. Madiraju, T. Sugiyama and S. C. Kowalczykowski, 2002. Escherichia coli RecO protein anneals ssDNA complexed with its cognate ssDNA-binding protein: a common step in genetic recombination. Proc. Natl. Acad. Sci. USA 99: 15327–15332. [PMC free article] [PubMed]
  • Kidane, D., H. Sanchez, J. C. Alonso and P. L. Graumann, 2004. Visualization of DNA double strand breaks repair in live bacteria reveals dynamic recruitment of Bacillus subtilis RecF, RecO and RecN protein to distinct sites on the nucleoids. Mol. Microbiol. 52: 1627–1639. [PubMed]
  • Kowalczykowski, S. C., 2000. Initiation of genetic recombination and recombination-dependent replication. Trends Biochem. Sci. 25: 156–165. [PubMed]
  • Kowalczykowski, S. C., and A. K. Eggleston, 1994. Homologous pairing and DNA strand-exchange proteins. Annu. Rev. Biochem. 63: 991–1043. [PubMed]
  • Kuzminov, A., 1999. Recombinational repair of DNA damage in Escherichia coli and bacteriophage lambda. Microbiol. Mol. Biol. Rev. 63: 751–813. [PMC free article] [PubMed]
  • Lloyd, R. G., 1991. Conjugational recombination in resolvase-deficient ruvC mutants of Escherichia coli K-12 depends on recG. J. Bacteriol. 173: 5414–5418. [PMC free article] [PubMed]
  • Mahdi, A. A., G. J. Sharples, T. N. Mandal and R. G. Lloyd, 1996. Holliday junction resolvases encoded by homologous rusA genes in Escherichia coli K-12 and phage 82. J. Mol. Biol. 257: 561–573. [PubMed]
  • Mandal, T. N., A. A. Mahdi, G. J. Sharples and R. G. Lloyd, 1993. Resolution of Holliday intermediates in recombination and DNA repair: indirect suppression of ruvA, ruvB, and ruvC mutations. J. Bacteriol. 175: 4325–4334. [PMC free article] [PubMed]
  • Marians, K. J., 2000. Replication and recombination intersect. Curr. Opin. Genet. Dev. 10: 151–156. [PubMed]
  • McGlynn, P., and R. G. Lloyd, 2002. a Genome stability and the processing of damaged replication forks by RecG. Trends Genet. 18: 413–419. [PubMed]
  • McGlynn, P., and R. G. Lloyd, 2002. b Recombinational repair and restart of damaged replication forks. Nat. Rev. Mol. Cell Biol. 3: 859–870. [PubMed]
  • McGregor, N., S. Ayora, S. Sedelnikova, B. Carrasco, J. C. Alonso et al., 2005. The structure of Bacillus subtilis RecU Holliday junction resolvase and its role in substrate selection and sequence specific cleavage. Structure 13: 1341–1351. [PubMed]
  • Meddows, T. R., A. P. Savory and R. G. Lloyd, 2004. RecG helicase promotes DNA double-strand break repair. Mol. Microbiol. 52: 119–132. [PubMed]
  • Michel, B., G. Grompone, M. J. Flores and V. Bidnenko, 2004. Multiple pathways process stalled replication forks. Proc. Natl. Acad. Sci. USA 101: 12783–12788. [PMC free article] [PubMed]
  • Pedersen, L. B., and P. Setlow, 2000. Penicillin-binding protein-related factor A is required for proper chromosome segregation in Bacillus subtilis. J. Bacteriol. 182: 1650–1658. [PMC free article] [PubMed]
  • Robu, M. E., R. B. Inman and M. M. Cox, 2001. RecA protein promotes the regression of stalled replication forks in vitro. Proc. Natl. Acad. Sci. USA 98: 8211–8218. [PMC free article] [PubMed]
  • Robu, M. E., R. B. Inman and M. M. Cox, 2004. Situational repair of replication forks: roles of RecG and RecA proteins. J. Biol. Chem. 279: 10973–10981. [PubMed]
  • Sargentini, N. J., and K. C. Smith, 1989. Role of ruvAB genes in UV- and gamma-radiation and chemical mutagenesis in Escherichia coli. Mutat. Res. 215: 115–129. [PubMed]
  • Seigneur, M., V. Bidnenko, S. D. Ehrlich and B. Michel, 1998. RuvAB acts at arrested replication forks. Cell 95: 419–430. [PubMed]
  • Shan, Q., J. M. Bork, B. L. Webb, R. B. Inman and M. M. Cox, 1997. RecA protein filaments: end-dependent dissociation from ssDNA and stabilization by RecO and RecR proteins. J. Mol. Biol. 265: 519–540. [PubMed]
  • Sharples, G. J., and R. G. Lloyd, 1991. Resolution of Holliday junctions in Escherichia coli: identification of the ruvC gene product as a 19-kilodalton protein. J. Bacteriol. 173: 7711–7715. [PMC free article] [PubMed]
  • Sharples, G. J., F. E. Benson, G. T. Illing and R. G. Lloyd, 1990. Molecular and functional analysis of the ruv region of Escherichia coli K-12 reveals three genes involved in DNA repair and recombination. Mol. Gen. Genet. 221: 219–226. [PubMed]
  • Singleton, M. R., S. Scaife and D. B. Wigley, 2001. Structural analysis of DNA replication fork reversal by RecG. Cell 107: 79–89. [PubMed]
  • Thanassi, J. A., S. L. Hartman-Neumann, T. J. Dougherty, B. A. Dougherty and M. J. Pucci, 2002. Identification of 113 conserved essential genes using a high-throughput gene disruption system in Streptococcus pneumoniae. Nucleic Acids Res. 30: 3152–3162. [PMC free article] [PubMed]
  • Umezu, K., and R. D. Kolodner, 1994. Protein interactions in genetic recombination in Escherichia coli. Interactions involving RecO and RecR overcome the inhibition of RecA by single-stranded DNA-binding protein. J. Biol. Chem. 269: 30005–30013. [PubMed]
  • van Gool, A. J., N. M. Hajibagheri, A. Stasiak and S. C. West, 1999. Assembly of the Escherichia coli RuvABC resolvasome directs the orientation of Holliday junction resolution. Genes Dev. 13: 1861–1870. [PMC free article] [PubMed]
  • Xu, L., and K. J. Marians, 2003. PriA mediates DNA replication pathway choice at recombination intermediates. Mol. Cell 11: 817–826. [PubMed]
  • Yasbin, R. E., P. I. Fields and B. J. Andersen, 1980. Properties of Bacillus subtilis 168 derivatives freed of their natural prophages. Gene 12: 155–159. [PubMed]
  • Zerbib, D., C. Mezard, H. George and S. C. West, 1998. Coordinated actions of RuvABC in Holliday junction processing. J. Mol. Biol. 281: 621–630. [PubMed]

Articles from Genetics are provided here courtesy of Genetics Society of America
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...