Abstract
RecBCD is a DNA helicase/exonuclease implicated in degradation of foreign linear DNA and in RecA-dependent recombinational repair of chromosomal lesions in E. coli. The low viability of recA recBC mutants vs. recA mutants indicates the existence of RecA-independent roles for RecBCD. To distinguish among possible RecA-independent roles of the RecBCD enzyme in replication, repair, and DNA degradation, we introduced wild-type and mutant combinations of the recBCD chromosomal region on a low-copy-number plasmid into a ΔrecA ΔrecBCD mutant and determined the viability of resulting strains. Our results argue against ideas that RecBCD is a structural element in the replication factory or is involved in RecA-independent repair of chromosomal lesions. We found that RecBCD-catalyzed DNA degradation is the only activity important for the recA-independent viability, suggesting that degradation of linear tails of σ-replicating chromosomes could be one of the RecBCD’s roles. However, since the weaker DNA degradation capacity due a combination of the RecBC helicase and ssDNA-specific exonucleases restores viability of the ΔrecA ΔrecBCD mutant to a significant extent, we favor suppression of chromosomal lesions via linear DNA degradation at reversed replication forks as the major RecA-independent role of the RecBCD enzyme.
TWO-STRAND DNA lesions affect both strands of a DNA duplex opposite each other. In contrast to more common one-strand DNA lesions, two-strand DNA lesions threaten chromosomal integrity, block chromosomal replication, and interfere with chromosomal segregation; in this sense, they can be viewed as chromosomal lesions (Kuzminov 2001). Because of their two-strand nature, chromosomal lesions are recalcitrant to excision repair, which relies on the intact opposite strand to remove a one-strand DNA lesion. The only way to faithfully repair chromosomal lesions, and the major way employed by bacteria, is via homologous strand exchange with an intact sister duplex, promoted in Escherichia coli by the RecA protein (reviewed in Kuzminov 1999; Cox 2001). recA mutant cells are completely deficient in recombinational repair and therefore cannot mend chromosomal lesions. The two major configurations of chromosomal lesions are: (1) unfillable single-strand gaps (intermediates of a crosslink excision, daughter-strand gaps) and (2) double-strand ends (double-strand breaks, collapsed or broken replication forks; reviewed in Kuzminov 1999). Repair of both unfillable single-strand gaps and double-strand ends depends on RecA; however, it also depends on two independent sets of activities. RecF, RecO, and RecR proteins assist RecA in repairing single-strand gaps, whereas the RecBCD enzyme assists RecA in repairing double-strand ends (reviewed in Kuzminov 1999).
Consistent with the important role of homologous strand exchange in DNA damage repair, rec mutants are sensitive to DNA-damaging treatments. For example, recBC and recF mutants are sensitive to UV and show synergistic effect if combined in a double mutant, defining the two major independent pathways of recombinational repair (Horii and Clark 1973). In contrast, the recA mutation, which by itself makes cells extremely sensitive to UV (Clark and Margulies 1965), is epistatic to recBC or recF mutations for UV survival (Willetts and Clark 1969; Horii and Clark 1973), demonstrating the central role of RecA protein in recombinational repair. In addition to being sensitive to DNA-damaging treatments, recA mutant cultures exhibit lower viability, indicating the occurrence of endogenous chromosomal damage even during growth in laboratory conditions. The same epistatic interactions are observed between recA and recF in the case of endogenous chromosomal damage, assessed by the viability of cultures grown in the absence of exogenous DNA damage. However, the effects of recA and recBC mutations on viability are additive, rather than epistatic (Capaldoet al. 1974), suggesting that recBC genes define additional mechanisms relevant for viability.
The RecBCD enzyme is a multifunctional helicase-nuclease, also known as ExoV (reviewed in Taylor 1988; Kowalczykowskiet al. 1994; Kuzminov 1999). In vitro, RecBCD rapidly degrades linear duplex and single-strand DNA (ssDNA) in reactions dependent on ATP. The unusual ATP dependence of the DNA degradation by RecBCD is explained by the powerful DNA helicase activity of the enzyme (DNA unwinding requires ATP): apparently, RecBCD can degrade only those DNA strands that are passed through its helicase domain in an ATP-dependent manner (reviewed in Telander-Muskavitch and Linn 1981; Taylor 1988). Although the only nuclease domain of the holoenzyme is found in the C-terminal part of the RecB polypeptide (Yu et al. 1998a,b), mutational inactivation of the RecD subunit results in the RecBC enzyme that is completely deficient as a nuclease yet continues to be a potent helicase (Palas and Kushner 1990; Andersonet al. 1997) with in vitro rates of DNA unwinding ∼25% of RecBCD rates (Korangy and Julin 1994). Recombinational repair of double-strand breaks in recD mutants becomes dependent on ssDNA-specific exonuclease RecJ (Lloydet al. 1988; Lovettet al. 1988). The RecD subunit is reported to have no activity of its own in vitro (Mastersonet al. 1992) although it plays a regulatory role in vivo (Yuet al. 1998a; Amundsenet al. 2000).
Biochemical characterization of linear DNA processing by the RecBCD enzyme in the presence of RecA (Anderson and Kowalczykowski 1997) corroborated earlier proposals that RecBCD acts on double-strand DNA (dsDNA) ends to prepare them for RecA polymerization (Kuzminov 1996). According to the accepted models, the only role of RecBCD in E. coli is to help RecA in recombinational repair (Kowalczykowski 2000; Smith 2001). However, as mentioned above, the RecBCD enzyme must have an additional role in the chromosomal metabolism of E. coli, because additional inactivation of recBC genes reduces the viability of recA mutants from 50% to a mere 20% (Capaldoet al. 1974). In fact, the drop in viability due to the recBCD inactivation in wild-type cells is ∼3.5-fold, an effect much stronger than that of recA inactivation (Capaldoet al. 1974).
There are two competing models for the RecA-independent role of RecBCD in E. coli chromosomal metabolism. The first model suggests that the major nonrepair role of RecBCD is in removal of linear tails of the σ-replicating chromosomes, thus returning the chromosomes to θ-, replication (Figure 1, D to A; Horiuchi and Fujimura 1995; Uzestet al. 1995; Kuzminov and Stahl 1997). This is important, because a σ-replicating chromosome cannot produce a circular daughter chromosome (Figure 1, E to F)—a situation described as a “σ-replication trap” (Kuzminov 1999). According to this idea, the RecBCD-catalyzed linear DNA degradation in recA mutant cells is the only way out of the σ-replication trap. According to the other model, RecBCD acts to suppress chromosomal lesions at reversed replication forks by attacking the short linear duplex formed by extruded newly replicated DNA strands (Figure 1C) and thus eliminating the Holliday junction (Figure 1, C to B), which would otherwise be resolved by RuvABC, leading to replication fork breakage (Figure 1, C to D; Seigneuret al. 1998). We tested these and other conceivable ideas regarding RecA-independent functions of RecBCD by employing various alleles and combinations of genes in the recBCD chromosomal region.
MATERIALS AND METHODS
Media, growth conditions, and general methods: Cells were grown in Luria broth (LB; 10 g tryptone, 5 g yeast extract, 5 g NaCl, 0.25 ml 4 m NaOH/1 liter) or on LB plates (15 g agar/1 liter of LB). When cells were carrying plasmids, the media were supplemented with 100 μg/ml ampicillin. Other antibiotics, when required for strain construction, were used in the following concentrations: 100 μg/ml spectinomycin, 50 μg/ml kanamycin, 12.5 μg/ml chloramphenicol, or 10 μg/ml tetracycline. TM buffer is 10 mm Tris HCl pH 8.0, 10 mm MgSO4. Hershey (H) agar (13 g tryptone, 8 g NaCl, 2 g sodium citrate, 1.3 g glucose, 15 g agar/1 liter) was used for T4 plating (Carlson and Miller 1994). Top agar is prepared immediately before plating by mixing equal amounts of the bottom agar and the dilution buffer used in this experiment.
Bacterial strains: Bacterial strains are E. coli K-12 (Table 1). Individual recA, recBCD, and recF mutants were confirmed by their characteristic UV sensitivities. In addition, recBCD mutants were confirmed by their ability to plate T4 2 mutant phage (Silverstein and Goldberg 1976). Replacement of genes in the chromosome with either chloramphenicol or kanamycin resistance markers, as well as subsequent generation of in-frame deletions, was according to Datsenko and Wanner (2000). Deletion-replacement alleles of recBCD and recF were also confirmed by PCR.
Oligonucleotides used to replace recF with a cat insertion are: AATTCGACATCAACGTTTCTCGCTCATTTATACTTGG GTTTGTGTAGGCTGGAGCTGC and CAGAGCGCGGCTTATGTTGTCATGCCAATGAGACTGT AATCATATGAATATCCTCCTTAG.
Oligonucleotides used to replace the recC-ptr-recB-recD region with a kan insertion are: TGCGTAACACTCGTACGTCGCATCCGGCAATTACGTT TATTCCGTGTAGGCTGGAGCTGCTTC and GACCCGCCTGCATTGCCCGAATCGTCAGTAGTCAGGA GCCGCCATATGAATATCCTCCTTA.
Oligonucleotides used for insertion of a chloramphenicol-resistant gene between yajD and tsx at the position 430,320 bp on the E. coli chromosome are: CGCATCCGGCATGAACAAAGCACACGTTGTTAACAAT CAGAATGTGTAGGCTGGAGCTGC and CAACTTCTGATTATGAAAATGCCGGGATTTATTCCCG GCATATGAATATCCTCCTTAG.
In all cases, parts homologous to the chromosome are underlined.
Plasmids: A general description of the plasmids is in Table 1. Derivation of the plasmids built for this study is given below:
pAMP1: the 18.3-kbp recBCD region from pDWS2 cloned as the BamHI fragment into the BamHI site of pWSK29 in such orientation that the recC-ptr-recB-recD region is transcribed in the direction opposite to the lac promoter of the vector.
pAMP1B: as pAMP1, but the orientation of the BamHI insert is reversed.
pAMP2: 921-bp EcoRI deletion removes the promoter and the 5′ portion of the addB gene from pWSK2988 (Kooistra and Venema 1991; Kooistraet al. 1993).
pAMP3: the 11.7-kbp recC-ptr-recB region from pSA122 cloned into the BamHI site of pWSK29 in the same orientation as in pAMP1.
pAMP5: the 18.3-kbp recC-ptr-recB-recD region from pB1082CD cloned into the BamHI site of pWSK29 in the same orientation as in pAMP1. The presumed recBK1082Q mutation could not be confirmed by sequencing; since the non-wild-type behavior of the construct suggested an uncharacterized mutation, we designated the allele recB*CD.
pAMP7: the 18.3-kbp recC-ptr-recB-recD region from pMY330 cloned into the BamHI site of pWSK29 in the same orientation as in pAMP1.
pAMP8: the [XhoI-BamHI 4854 bp] fragment from pAMP1, containing recD and the 3′ part of recB, cloned at the XhoI and BamHI sites of pWSK29. The orientation of the recD gene in pAMP8 is the same as the orientation of the whole recBCD region in pAMP1.
pEAK1: the big EcoRI-HindIII fragment of pHGS415, carrying the bla gene and the replication origin, combined with the small EcoRI-HindIII fragment of pMTL21 (Chamberset al. 1988), carrying a multiple cloning site.
pK134: BamHI-cleaved pEAK1 first combined with the [BamHI-BamHI 11.7-kbp] fragment from pSA122 carrying the recC-ptr-recB region; the resulting plasmid was cleaved with SmaI, and the 14.8-kbp fragment was circularized, thus removing one of the BamHI sites and putting the recC gene under the cat promoter, left from pHSG415.
pK135: pK134, into the unique BamHI site of which the [BamHI-BamHI 3274 bp] fragment from pBEU14 containing the wild-type recA gene has been inserted such that recA is co-oriented with the recC-ptr-recB genes.
—The hypothetical RecA-independent roles of Rec-BCD in repair/removal of chromosomal lesions. The roles of the RecBCD enzyme are indicated by the numbered text in the grid of normal and abnormal chromosomal processes. DNA duplex is shown as a single line, the chromosome is shown as a circle, replication forks are marked by the solid circles at the branching points. (A-D) recA-RecBCD+; (E and F) recA-reBCD-. (A) A circular chromosome. (B) A θ-replicating chromosome. (C) The right replication fork of the θ-replicating chromosome has reversed, forming a Holliday junction from which the short linear duplex with an open end has extruded. (D) Resolution of the Holliday junction leads to breakage of the reversed replication fork. The chromosome is replicating in the σ-mode. (E) The σ-replicating chromosome initiates a new round of θ-replication from the origin. The old and new replication forks (including the broken one) are numbered. (F) Replication fork fusion lengthens the tail of the σ-structure (the σ-replication trap), dooming the cell.
T4 2 mutant plating: A fresh overnight culture is diluted 1000-fold into 2 ml of LB (supplemented with 250 μg of ampicillin if the strain harbors a plasmid) and grown with shaking at 28° to 2 × 108 cells/ml. A total of 100 μl of the growing culture is mixed with 10 μl of either T4 wild-type or T4 2 mutant phages, diluted in TM buffer to produce approximately the same number of plaques on wild-type cells. After a 3-min incubation at room temperature, 1 ml of top H agar is added and the mixture is plated on a 60 × 15 petri plate with H agar and incubated at 37° for 16 hr. The number of plaques of T4 2 mutant phage is divided by the number of plaques of the T4 wild-type phages on the same strain the same day to determine the normalized T4 2 mutant plating.
UV survival: A fresh overnight culture is diluted 100-fold into 2 ml of LB and grown with shaking at 28° to 2 × 108 cells/ml. Tenfold serial dilutions are made in 1% NaCl and spotted by 10 μl in rows of six onto a square petri dish with LB agar. The plate is dried, partially covered with a screen, and exposed to a gradient of doses of UV light in the direction perpendicular to the dilution gradient, so that every dilution column (from 10-1 to 10-6 dilutions) receives its own dose. UV crosslinker (Amersham-Pharmacia), in which all the lamps except the central one are removed and 90% of the remaining lamp is shielded, is used to deliver precise doses of UV (measured by the internal UV sensor). Immediately after the exposure, the plate is covered with aluminum foil and incubated at 37° for 16-24 hr. The titer of the culture at the zero dose is used to determine the survival at various UV doses.
Quantitative P1 transduction: In a 1.5-ml microcentrifuge tube, 400 μl of a fresh overnight culture in LB is mixed with 500 μl of 30 mm MgCl2, 15 mm CaCl2, 200 μl of LB, and 30 μl of a P1 lysate of AM6 and incubated for 20 min at 28° with shaking. The cells are pelleted by 1 min centrifugation in a microcentrifuge, resuspended in 1.2 ml of LB supplemented with 20 mm sodium citrate, and incubated with shaking for 1 hr at 28°. A total of 100 μl of the culture is plated on LB + 20 mm sodium citrate plates with single selections, whereas cells from the remaining 1 ml are collected by centrifugation, resuspended in 100 μl of LB, and plated on an LB + 20 mm sodium citrate plate with the double (chloramphenicol + tetracycline) selection. The plates are incubated at 37° for 48 hr before transductants are counted.
Viability: A fresh overnight culture is diluted 100-fold into 2 ml of LB and grown with shaking at 28° to 1-5 × 108 cells/ml. Two 100-μl aliquots of the culture are taken at the same time: one aliquot is diluted 100-fold into 1% NaCl to stop growth, and the other one is mixed with 0.8 μl of 4 m NaOH to stop cell movement. A total of 5 μl of the NaOH-treated aliquot is used to count cells under the microscope in a Petroff-Houser counting chamber. The 1% NaCl-diluted aliquot is diluted further to achieve the final dilution of 10-5, 100 μl of the final dilution is combined with 3.5 ml of top LB agar, and the mixture is poured over an LB plate. The plate is incubated at 37° for 16-36 hr, and the titer of the colony-forming units is divided by the titer of the microscopic count to determine the viability.
Bacterial strains and plasmids
Field-inversion gel electrophoresis: Growing cultures of strains to be tested were normalized to OD600 = 0.5-0.6. Cells from 1.5 ml of the normalized cultures were pelleted by centrifugation and resuspended in 60 μl of TE buffer. After incubation at 37° for 2-10 min, 5 μl of 10 mg/ml Proteinase K was added, followed by 65 μl of 1.2% molten agarose in 0.2% sarcosyl, 10 mm Tris HCl pH 8.0, 5 mm EDTA, kept at 70°. After vortexing and mixing by pipetting, 110 μl of the cell suspensions were pipetted into plug molds and let solidify for 2-10 min. Each plug was then placed in a small glass tube containing 1 ml of 1% sarcosyl, 50 mm Tris HCl pH 8.0, and 25 mm EDTA and incubated for 1 hr at 60°. Plugs were inserted into a 1% agarose gel on 0.5× TBE buffer and run at room temperature in a regular agarose gel box with field inversion factor 3:1 at 4 V/cm for 5 hr with ramping 1-20 sec, then for 5 hr with ramping 20-60 sec, and then for 5 hr with ramping 60-100 sec. Gels were stained with ethidium bromide before being photographed in UV light.
The viability of E. coli strains carrying point mutations or deletions of recA, recBCD, and/or recF genes
RESULTS
The phenomenon and possible explanations: When grown under the standard laboratory conditions, recA mutants are only 46-48% viable (Table 2), indicating the frequency of chromosomal damage and demonstrating the importance of repairing it. The two pathways of recA-dependent repair in E. coli are controlled by the recBC and recFOR genes. A ΔrecF mutant has ∼71% viability (Table 2), which is consistent with inactivation of one of the two recombinational repair pathways. A ΔrecA mutation is epistatic to the ΔrecF mutation for viability (the ΔrecA ΔrecF double mutant has the viability of a single ΔrecA mutant), indicating that RecF has no effect on viability outside the RecA-dependent processes. In contrast, recBC mutants are only ∼28% viable, suggesting an additional role for RecBCD beyond the RecA-controlled recombinational repair. Indeed, compared with 46-48% viability of recA mutants, recA recBC combinations are only 18-19% viable (Table 2), supporting a RecA-independent role for RecBCD. recBCD inactivation is epistatic to recF inactivation (Table 2).
—Physical evidence for the chromosomal fragmentation in rec mutants. Field-inversion gel electrophoresis of chromosomal DNA isolated from exponential cells of the indicated genotype is shown. MWM, molecular weight markers (yeast chromosomes). The size of the MWM bands in kilobase pairs is shown on the left. The extent of chromosome fragmentation in various strains is reproducible; however, due to the inherent lack of linearity over a broad range of fluorescence intensities, comparison of the data from different gels is not appropriate. ΔrecA, JC10287; ΔrecBCD, JB1; ΔrecA ΔrecBCD, AM1; pRecBCD+, pAMP1.
Pulsed-field gel electrophoresis of undigested chromosomal DNA is employed to detect chromosomal fragmentation: under pulsed-field gel conditions, intact chromosomes stay at the origin, whereas linear subchromosomal fragments migrate into the gel, forming a smear (Michelet al. 1997; Thoms and Wackernagel 1998). Pulsed-field gel electrophoresis reveals a background level of chromosomal fragmentation in wild-type cells (Figure 2, lane a). Chromosomal fragmentation is visibly increased by inactivation of either recA or recBCD and is most evident in a recA recBCD mutant (Figure 2, lanes b-d). Thus, the chromosomal fragmentation results coincide with the viability results and suggest that the low viability of recA recBCD mutants is due to the failure to prevent or repair chromosomal fragmentation.
We conceived four possibilities for the RecA-independent role of RecBCD, numbered from 1 to 4 irrespective of their likelihood (Figure 1). Possibility 1 is that RecBCD is an integral part of a supramolecular complex at the replication factory (Figure 1, A to B), and that its absence destabilizes other components of the factory, negatively affecting chromosomal replication. Indeed, RecBCD is occasionally reported to purify from cells in complexes with replication enzymes (Hendleret al. 1975; Syvaoja 1987). Possibility 2 is that the RecBCD enzyme promotes RecA-independent recombinational repair of disintegrated replication forks (Figure 1, D to B). There are types of homologous recombination in the chromosome that are independent of RecA, although these either require inactivation of RecBCD as well (Wackernagel and Radding 1974; Weisberg and Sternberg 1974; Elliset al. 2001) or are independent of RecBCD (Lovettet al. 1993; Bierneet al. 1997). Possibility 3 is that the RecBCD-promoted linear DNA degradation is important in preventing replication fork breakage (Figure 1, C to B) as a chromosomal damage suppression mechanism (see Introduction). And last, possibility 4 is that RecBCD degrades linear tails of σ-replicating chromosomes (Figure 1, D to A) as part of a chromosomal damage removal mechanism (see Introduction).
The experimental system: To find the nature of the RecBCD effect on the cell’s viability outside the RecA-dependent recombinational repair, we transformed a ΔrecA ΔrecBCD strain with plasmids carrying various alleles of the E. coli recBCD genes or genes with similar functions from Bacillus subtilis (Table 1, plasmids; this study) and determined viability of the resulting strains, looking for the constructs that would restore viability to the ∼50% level of a single recA mutant. As controls, we used the same plasmids to transform a RecA+ RecBCD+ strain, as well as single ΔrecA or ΔrecBCD deletion mutants. As a vector for all our constructs, we used pSC101-derived pWSK29, which has a high stability and a low copy number (Wang and Kushner 1991).
To assess the functionality of the constructs, we determined to what extent defects of a ΔrecBCD mutant are complemented by the introduced constructs. As mentioned in the Introduction, the two major functions of the RecBCD enzyme are: (1) participation in the RecA-promoted recombinational repair and (2) degradation of linear duplex DNA (ExoV). To determine the recombinational repair capacity of ΔrecBCD mutants transformed with the constructs, we measured their resistance to UV irradiation. Relative to the wild-type cells, ΔrecBCD mutants are quite sensitive to UV due to their defect in repair of double-strand breaks (Wang and Smith 1983, 1986). Thus, restoration of UV resistance signals that the plasmid construct carries an enzyme proficient in recombinational repair. To evaluate the ExoV activity of the constructs, we measured the ability of recBCD mutant cells, transformed with the constructs, to plate a gene 2 mutant of bacteriophage T4. The gene 2 product binds to the ends of the linear T4 chromosome and protects them from exonuclease degradation inside the cell. T4 2 mutant phages plate with reduced efficiency on ExoV+ cells, but plate well on null recBCD mutants, which are deficient in ExoV (Silverstein and Goldberg 1976). Thus, poor plating of T4 2 mutant phages signals that the plasmid construct carries an enzyme proficient in linear DNA degradation.
To simultaneously evaluate both the recombination capacity and the ExoV activity of the constructs in functional interaction with each other, we measured the frequency of generalized P1 transduction in these strains. During P1 transduction, an ∼100-kbp piece of a donor chromosome, carrying a selectable marker, is introduced into a recipient cell, and recombinants, with the donor marker inserted in the recipient chromosome, are selected for. The formation of these recombinants depends on both RecA and RecBCD functions in the recipient cells (Willetts and Mount 1969; Zieg and Kushner 1977); the transduction is promoted by the recombination activity of RecBCD, while being somewhat inhibited by the DNA degradation activity (Chaudhury and Smith 1984). When incorporation of a single marker is selected for, the incorporated piece of the donor chromosome can be almost as short as the marker itself. However, when two separate donor markers are simultaneously selected for, the incorporated piece needs to be at least as long as the distance between the two markers (unless the recombinant is formed by two independent integration events, which is rare). Under these circumstances, the wild-type levels of linear DNA degradation should decrease the frequency of double transductants, while having little effect on the levels of single transductants. The two markers that we used were lacZ::Tn10(tet) and a cat insertion between yajD and tsx in the E. coli chromosome, placing the distance between the two markers at ∼64 kbp. By selecting for growth on tetracycline plus chloramphenicol after transduction, we were selecting for incorporation of the central 64 kbp of the injected 100 kbp (as well as for rare independent double transductants).
We verified how the system works with ΔrecBCD cells transformed with an empty vector or the vector carrying a complete wild-type recBCD region of the chromosome. In the DNA degradation test (T4 2 mutant phage plating), the vector alone behaved essentially as the ΔrecBCD control strain, whereas the recBCD+ plasmid conferred nearly wild-type levels of DNA degradation capacity (Figure 3A). In the recombinational repair test (UV resistance) the vector alone did not provide any resistance over the ΔrecBCD levels, whereas the recBCD+ plasmid conferred close to wild-type levels of UV resistance at doses 27 J/m2 and lower (Figure 3B). With the combined recombinational repair/DNA degradation test, vector alone did not contribute to transduction over the ΔrecBCD levels, whereas the recBCD+ plasmid significantly increased transduction over the background levels, although it failed to bring it to the wild-type levels (Figure 3C). Finally, we measured the viability of wild-type cells, as well as single ΔrecA or ΔrecBCD deletions and of the double ΔrecA ΔrecBCD deletion strains, transformed with vector alone or with the recBCD+ plasmid. Vector alone did not significantly change the viability of the four strains (Figure 3D). In contrast, the recBCD+ plasmid, although having no effect on the viability of wild-type and ΔrecA cells, increased the viability of the ΔrecBCD mutant to the level intermediate between the mutant and the wild type and increased the viability of the ΔrecA ΔrecBCD mutant to the level of the ΔrecA mutant (Figure 3D). Subjecting chromosomal DNA to pulsed-field gels revealed that the recBCD+ plasmid decreases chromosomal fragmentation in both recBCD and recA recBCD mutant cells (Figure 2, lanes e and f). We conclude that (1) the tests for the RecBCD functions are highly sensitive, allowing a distinguishing power of two to three orders of magnitude; (2) the vector itself does not contribute to the RecBCD-attributed characteristics and to the viability; and (3) the recBCD-carrying plasmid does restore some characteristics to the levels of the wild-type cells and some other characteristics to intermediate levels (addressed in the discussion).
—pRecBCD+ plasmid fully complements the linear DNA degradation defect and partially complements the recombinational repair defect of recBCD mutant cells. The assays in A-C employ wild type, ΔrecBCD mutant cells, and ΔrecBCD mutant complemented with the vector alone or with the pRecBCD+ plasmid. The viability assay in D employs wild-type, ΔrecA, ΔrecBCD, and ΔrecA ΔrecBCD strains. (A) Linear DNA degradation capacity, as gauged by plating efficiency of T4 2 mutant phage. (B) Recombinational repair capacity, as gauged by survival of UV irradiation. (C) Recombination repair capacity vs. linear DNA degradation activity in functional interaction with each other, as gauged by a two-marker P1 transduction. (D) Viability expressed as colony-forming efficiency. The strains are: AB1157, the wild-type control; ΔrecA, JC10287; ΔrecBCD, JB1; ΔrecA ΔrecBCD, AM1. The plasmids are: vector, pWSK29; pRecBCD+, pAMP1.
RecBCD enzyme does not play a structural role in supramolecular assemblies: As elaborated above, RecBCD can boost the viability of E. coli independently of RecA in at least four possible ways (Figure 1): (1) by serving as a structural element at the replication factory, (2) by promoting RecA-independent recombinational repair of damaged chromosomes, (3) by suppressing chromosomal damage via degradation of abnormal replication forks, and (4) by returning σ-replicating chromosomes to θ replication via degradation of linear tails. First, we tested the idea that RecBCD could be an integral part of a supramolecular assembly involved in chromosomal replication/repair (replication factory) and that its absence destabilizes the assembly, negatively affecting the replication process.
To do that, we provided the four test strains with the AddAB enzyme from B. subtilis, the powerful exonuclease/helicase that plays the same role in Bacillus as RecBCD does in E. coli, but has little sequence homology with the E. coli RecBCD and therefore is unlikely to become a part of any supramolecular structure in E. coli. Somewhat surprisingly, the AddAB enzyme complemented not only the DNA degradation deficiency of the ΔrecBCD mutant (Figure 4A), but also restored its recombinational repair proficiency (Kooistraet al. 1993; Figure 4B). The combined test for DNA degradation/recombination revealed lowered proficiency of the AddAB enzyme, likely due to the high degradation power of this enzyme (Figure 4C; Kooistraet al. 1993). This hyperactive DNA degradase and a potent recombinase from a very different organism restore the viability of both the ΔrecBCD mutant (to the wild-type levels) and the ΔrecA ΔrecBCD mutant (to the recA mutant levels; Figure 4D), while slightly decreasing the viability of wild-type cells. Inactivation of addAB genes by deletion of the promoter region, as well as the very 5′-terminal part of the addB gene, eliminates all the complementation (Figure 4, A-C), elevating the viability of all three mutant strains somewhat (Figure 4D). This shows that the effect of the addAB clone is due to the functional AddAB enzyme, rather than to some sequences in the cloned region. The restoration of the ΔrecA ΔrecBCD mutant viability by the completely nonhomologous AddAB enzyme therefore contradicts the idea that the RecA-independent role of RecBCD is one of a structural element in a replication factory or another supramolecular complex involved in DNA replication.
As another test of this idea, we employed the RecB1080CD mutant enzyme of E. coli, which has a single-amino-acid change completely inactivating the only nuclease active center of the enzyme, situated on the RecB subunit. Although presumably structurally very similar to the wild-type enzyme, the mutant enzyme does not show any DNA degradation activity in vitro (Yuet al. 1998b). In our in vivo DNA degradation, recombinational repair, and combined DNA degradation/recombination tests, the recB1080CD allele behaves essentially as a null allele (Figure 5, A-C), confirming observations of others (Amundsenet al. 2000). If such a (presumably) structurally preserved but functionally inactive enzyme would restore the viability of ΔrecBCD mutant cells, this could be a strong indication for the structural role of RecBCD. However, the RecB1080CD mutant enzyme moderately increases the viability of recBCD mutant cells (and, surprisingly, of wild-type cells), but does not improve the viability of a recA single mutant and a recA recBCD double mutant (Figure 5D). These results with the RecB1080CD mutant enzyme argue against the structural role for the RecBCD enzyme in E. coli’s DNA replication process.
Wild-type levels of DNA degradation or DNA unwinding are important for viability of recA mutants: The RecBCD enzyme exhibits two groups of enzymatic activities: those important for recombination and those important for DNA degradation, both of which could contribute to the viability of recA mutants, as demonstrated by the effect of the analogous AddAB enzyme from B. subtilis. To see which activity of RecBCD is important for recA-independent survival, we employed the RecB*CD enzyme, which does not restore UV resistance and P1 transduction proficiency to the ΔrecBCD mutant cells, but does restore wild-type levels of DNA degradation, as measured by the T4 2 mutant survival (Figure 5, A-C). Thus, the recombination activities of RecBCD are selectively inactivated by the recB* mutation. If the RecB*CD mutant enzyme fails to restore the viability of the ΔrecA ΔrecBCD mutant cells, this would argue against the role of linear DNA degradation in recA-independent survival. However, we found that the RecB*CD mutant enzyme restores the viability of the ΔrecBCD mutant to wild-type levels and upgrades the viability of the recA recBCD mutant to the recA mutant level (Figure 5D), arguing that it is the DNA degradation activity of RecBCD that is important for its RecA-independent contribution to viability.
To verify this conclusion, we employed a recBC+ allele; deletion of recD should selectively inactivate the DNA degradation activities of the RecBCD enzyme, leaving the DNA helicase activity (important for recombinational repair) intact (Amundsenet al. 1986; Korangy and Julin 1994; Churchillet al. 1999). Indeed, our in vivo tests show selective removal of the bulk of DNA degradation activity of the enzyme (Figure 6, A-C, pRecBC variant): the RecBC-supplemented ΔrecBCD mutant cells behave as wild-type cells in the recombinational repair (UV survival) test, are grossly defective in linear DNA degradation (50% permissivity for T4 2 mutant), and hyper-rec in the combined DNA degradation/recombination test (two-marker P1 transduction), reflecting their defect in linear DNA degradation. If the wild-type levels of DNA degradation were important for the recA-independent contribution to viability by RecBCD, the RecBC enzyme would not contribute significantly to the viability of recA recBCD mutants. The plasmid expressing only recBC genes does not influence the viability of the wild-type or recA mutants cells, but, remarkably, it almost doubles the viability of recBCD mutant cells and restores the viability of recA recBCD mutant cells to the levels of recA mutants (Figure 6D). As a control we used a plasmid that expresses only the RecD polypeptide. As expected, this plasmid did not restore any RecBCD-specific activity of the ΔrecBCD mutant cells, nor did it change the viability of the three mutants (Figure 6). Characteristically, the RecD-producing plasmid lowered the viability of the wild-type cells, probably due to the lack of regulation of the DNA degradation, observed in strains overproducing the RecD subunit (Brcic-Kosticet al. 1992).
—The effect of the B. subtilis AddAB nuclease. The four assays are the same as in Figure 3. The assays in A-C employ ΔrecBCD mutant cells complemented with various plasmids. The viability assay in D employs wild-type, ΔrecA, ΔrecBCD, and ΔrecA ΔrecBCD strains (identified in the legend to Figure 3). The plasmids are: vector, pWSK29; pRecBCD+, pAMP1; pAddAB+, pWSK2988; pAddAB-, pAMP2.
—The effect of two RecBCD enzymes with point mutations. The four assays are the same as in Figure 3. The assays in A-C employ ΔrecBCD mutant cells complemented with various plasmids. The viability assay in D employs wild-type, ΔrecA, ΔrecBCD, and ΔrecA ΔrecBCD strains (identified in the legend to Figure 3). The plasmids are: vector, pWSK29; pRecBCD+, pAMP1; pRecB1080CD, pAMP7; pRecB*CD, pAMP5.
DNA unwinding restores viability only if ssDNA-specific degradation is available: At face value, the result with the RecBC enzyme contradicts the previous result, arguing that the wild-type level of DNA degradation is not as important for viability in the absence of RecA as the recombination-relevant DNA helicase activity of the RecBC(D-) mutant enzyme is. To see if the RecBC enzyme or the complete RecBCD enzyme could promote recombinational repair in the absence of RecA, we verified whether the corresponding plasmids confer any degree of UV resistance or P1 transduction proficiency to the recA recBCD mutant. We found no changes in UV resistance and a very low level of P1 transduction promoted by the RecBCD or RecBC enzymes in the absence of RecA (Figure 7). Thus, the only possible explanation for the surprising restoration of the viability of the ΔrecA ΔrecBCD mutant by the RecBC-producing plasmid is that the attenuated DNA degradation, catalyzed by the combination of RecBC helicase and ssDNA-specific nucleases (Figure 8A; Rinkenet al. 1992; Korangy and Julin 1994), is enough to restore the viability of recA recBCD mutants to the level of recA mutants. The prediction is that inactivation of the ssDNA-specific exonucleases in the RecBC-complemented ΔrecA ΔrecBCD mutant would diminish the degree to which the viability is restored.
—The effect of expressing the RecBC enzyme vs. the RecD subunit. The four assays are the same as in Figure 3. The assays in A-C employ ΔrecBCD mutant cells complemented with various plasmids. The viability assay in D employs wild-type, ΔrecA, ΔrecBCD, and ΔrecA ΔrecBCD strains (identified in the legend to Figure 3). The plasmids are: vector, pWSK29; pRecBCD+, pAMP1; pRecBC, pAMP3; pRecD, pAMP8.
To test this idea, we constructed a ΔrecA ΔrecBCD variant, with the two major single-strand DNA-specific exonucleases, ExoI (gpxonA) and RecJ, inactivated. To accomplish this, we complemented the strain with a temperature-sensitive plasmid, carrying both recA+ and recBCD+ genes, introduced either a ΔxonA or a recJ mutation by P1 transduction, and, finally, lost the complementing plasmid. The plasmid was easily lost from the ΔrecA ΔrecBCD ΔxonA mutant, was lost with some difficulty from the ΔrecA ΔrecBCD recJ mutant, but could not be lost, even after repeated attempts, from the ΔrecA ΔrecBCD ΔxonA recJ mutant, suggesting that this combination is inviable (in fact, even the ΔrecA ΔrecBCD recJ mutant has an extremely low viability; Figure 8C). Since the double ΔxonA recJ mutant was inviable in the ΔrecA ΔrecBCD background, we had to do the experiment in ΔrecA ΔrecBCD ΔxonA and ΔrecA ΔrecBCD recJ mutant strains.
—RecBCD or RecBC enzymes are unable to promote recombinational repair in the absence of RecA protein. The assays employ wild-type, ΔrecA ΔrecBCD mutant cells, and the ΔrecA ΔrecBCD mutant complemented with the pRecBCD+ plasmid or with the pRecBC+ plasmid. (A) Survival of UV irradiation. (B) One-marker P1 transduction. The strains are: AB1157, the wild-type control; ΔrecA ΔrecBCD, AM1. The plasmids are: pRecBCD+, pAMP1; pRecBC+, pAMP3.
To confirm the ssDNA-exonuclease defects, we introduced the RecBCD+ or the RecBC+ plasmids into the ssDNA-exonuclease-deficient ΔrecA ΔrecBCD cells and measured the level of linear DNA degradation in these cells by T4 2 mutant plating. As expected, T4 2 mutant plating was high on the ΔrecA ΔrecBCD mutant and its ΔxonA or recJ derivatives (Figure 8B). T4 2 mutant plating decreased to the same low level in all three mutants supplemented by the RecBCD+ (Hel+/Exo+) plasmid (RecBCD degrades dsDNA without the help of ssDNA-specific exonucleases) and was intermediate in the presence of the RecBC+ (Hel+/Exo-) plasmid in the ΔrecA ΔrecBCD cells (Figure 8B). However, compared with the ΔrecA ΔrecBCD strain, T4 2 mutant plating in the presence of pRecBC+ plasmid was four times higher on ΔrecA ΔrecBCD ΔxonA and seven times higher on ΔrecA ΔrecBCD recJ mutant cells, corroborating their defect in ssDNA-specific exonucleases (Figure 8B).
As expected, the RecBCD+ plasmid completely restored the viability of the ssDNA-exonuclease-deficient ΔrecA ΔrecBCD cells to the viability levels of recA mutants (Figure 8C). RecBC+ plasmid also restored the viability of the ΔrecA ΔrecBCD control strain almost to the recA mutant level. However, the viability of the ΔrecA ΔrecBCD ΔxonA mutant, and especially of the ΔrecA ΔrecBCD recJ mutant, although improved, was still two- to threefold below the viability of the same mutants, complemented with the complete RecBCD+ region (Figure 8C). In other words, the effects on the two graphs (Figure 8, B vs. C) were exactly reversed: the higher T4 2 mutant survival (indicating lower DNA degradation levels) translated into the lower viability of the strain. Thus, ssDNA-specific exonucleases become essential for viability if the complete RecBCD enzyme is replaced with the exonuclease-deficient but helicase-proficient RecBC enzyme. Therefore, the relatively high viability of recA mutants relies on the high affinity of the RecBCD or RecBC enzymes toward double-strand ends and depends on some DNA degradation from these ends.
DISCUSSION
RecBCD of E. coli is a large protein complex with DNA helicase and exonuclease activities, which are implicated in degradation of linear DNA and in recombinational repair of double-strand breaks in the chromosome. recA null mutants are 50% viable, whereas recA recBCD mutants are only 20% viable, indicating a role for the RecBCD enzyme in a recA-independent survival. We conceived four possibilities for such a role (Figure 1): (1) RecBCD is a structural element in the replication factory assembly; (2) RecBCD has RecA-independent recombinational repair activity (uses its DNA helicase activity to reassemble disintegrated replication forks); (3) RecBCD is a chromosomal damage suppressor (uses its high affinity to double-strand ends to attack abnormal replication structures whose alternative processing would lead to chromosomal lesions); and (4) RecBCD has chromosomal damage-removal activity (uses its powerful exonuclease activity to degrade long linear tails of σ-replicating chromosomes, thus returning chromosomes to θ-replication).
—The viability of recA recBCD mutant cells complemented with the RecBC+ enzyme depends on the ssDNA-specific exonucleases. The assays employ recA recBCD (ΔΔ), recA recBCD xonA (ΔΔ xonA), and recA recBCD recJ mutants (ΔΔ recJ), either uncomplemented or complemented with pRecBCD+ or pRecBC+ plasmids. (A) The scheme of linear DNA degradation by the RecBCD nuclease or by the combination of the RecBC helicase and ssDNA-specific exonucleases ExoI and RecJ. (B) Linear DNA degradation capacity, as gauged by plating efficiency of T4 2 mutant phage. (C) Viability expressed as colony-forming efficiency. The strains are: ΔrecA ΔrecBCD, AM1; ΔrecA ΔrecBCD recJ, AK132; ΔrecA ΔrecBCD xonA, AK133. The plasmids are: pRecBCD+, pAMP1; pRecBC+, pAMP3.
To identify the RecA-independent role for the RecBCD enzyme in the viability of E. coli cells, we employed several constructs, carrying the recBCD chromosomal region on a low-copy-number plasmid (Table 3). The constructs were characterized in ΔrecBCD mutant cells for the DNA degradation capacity (T4 2 mutant survival) and for recombinational repair proficiency (UV resistance) as well as in a combined test for both DNA degradation and recombination (two-marker P1 transduction). Some constructs restored both the DNA degradation and the recombinational repair capacity of the ΔrecBCD mutant, some other constructs restored selectively either DNA degradation or recombinational repair, while yet other constructs restored neither property. In addition, the introduced enzymes were judged either structurally similar or dissimilar to the wild-type RecBCD enzyme on the basis of the sequence homology or the nature of the mutation (Table 3). We then introduced these different constructs into a ΔrecA ΔrecBCD mutant and determined viability of the resulting strains. We found that the RecBCD mutant enzyme lacking any activity due to a point mutation in the active site cannot restore the viability of the ΔrecA ΔrecBCD mutant, which argues against the idea that RecBCD plays the role of a structural element at the replication factory. The enzyme proficient only in DNA degradation restores the viability, which argues for the importance of DNA degradation in prevention/removal of chromosomal lesions. Surprisingly, the enzyme proficient only in DNA unwinding can also restore the viability (Table 3), suggesting recA-independent recombinational repair. However, the constructs that restore the recombinational repair capacity to the ΔrecBCD strain do not restore the recombinational repair capacity to the ΔrecA ΔrecBCD strain, suggesting that the only RecA-independent role of the RecBC enzyme is still in linear DNA degradation. The inability of the helicase-proficient but exonuclease-deficient RecBC enzyme to fully restore the viability of ΔrecA ΔrecBCD strains, deficient in either one of the two major ssDNA-specific exonucleases, confirms this idea. We conclude that the low viability of recA recBCD mutants is due to their loss of the capacity to target linear DNA for limited degradation.
Summary of the findings
We introduced our constructs on a pSC101-derivative plasmid, which is reported to have a copy number of five to seven per chromosome (Manen and Caro 1991). The constructs retain natural promoters, and since no regulation is known for any gene of the recBCD operon, we may assume that the actual level of the enzymes produced from these constructs is five to seven times higher than the wild-type level. Therefore, the incomplete complementation of the ΔrecBCD mutant phenotype by the pRecBCD and pRecBC plasmids could be attributed to this severalfold overproduction of the enzyme from plasmid. On the other hand, the recBCD genes in all our plasmids are cloned in opposite orientation relative to the direction of transcription from the truncated lacZ gene, making it possible that the recBCD genes actually are underexpressed. However, when we inverted the wild-type recBCD fragment in the pAMP1 construct, the viability of the ΔrecBCD deletion strain, complemented with the “inverted” construct (pAMP1B), was the same as the viability complemented by the “direct” construct (data not shown), suggesting that the possible lacZ expression does not interfere with the expression of recBCD. The incomplete complementation is unlikely to be due to the plasmid instability, because (1) plasmids are relatively stable in recBC mutants (Bassett and Kushner 1984) and (2) the pSC101 replicon has a dimer resolution site and therefore always exists in monomeric form, which increases plasmid stability (Cornetet al. 1994). Few examples of incomplete complementation notwithstanding, the combined results provide a coherent and compelling picture, allowing one to distinguish among the competing hypotheses with confidence.
What is the nature of chromosomal lesions suppressed or removed by the RecBCD-promoted DNA degradation? Pulsed-field gel electrophoresis reveals significant chromosome fragmentation in recA mutants, which is roughly doubled in recBCD and recA recBCD mutants (Figure 2; Michelet al. 1997; Gromponeet al. 2002). Therefore, the chromosomal lesions must be double-strand ends: either double-strand breaks or disintegrated replication forks. Our original idea was that the main targets of the RecBCD-promoted degradation were long linear tails of σ-replicating chromosomes (Figure 1, D → A; Horiuchi and Fujimura 1995; Uzestet al. 1995; Kuzminov and Stahl 1997). The speed with which the RecBCD enzyme can unwind DNA in vitro [1000 bp/sec (Roman and Kowalczykowski 1989)— and presumably even faster in vivo] is higher than the speed of replication fork propagation in E. coli (600-900 bp/sec; Bremer and Dennis 1996), so, theoretically, RecBCD can overcome the remaining replication fork in the σ-replicating chromosome. However, our finding that a more modest DNA degradation activity, due to the combined action of the RecBC helicase plus ssDNA-specific exonucleases ExoI and RecJ (Rinkenet al. 1992; Korangy and Julin 1994), restores the viability of the recA recBCD mutant almost to the same level that the wild-type RecBCD enzyme does, argues against the σ-replicating chromosome as the most frequent target of RecBCD. Two additional observations suggest that degradation of linear tails of σ-structures cannot be the sole RecA-independent pathway of RecBCD-promoted chromosomal repair: (1) in polA recA, lig recA, and dam recA mutants the chromosome is degraded by RecBCD, but the degradation does not make these mutants viable (Monk and Kinross 1972; Marinus and Morris 1975; Morseet al. 1976) and (2) there is no more chromosomal degradation in rep recA or holD recA mutants than in recA single mutants, but the single rep or holD mutants are dependent on RecBCD, while the double mutants (with recA) are even dependent on RecD (Seigneuret al. 1998; Floreset al. 2001). Apparently, there are situations when RecBCD does not have to degrade long linear tails to contribute to viability. This raises a question about the chromosomal structure, in which a short linear piece of DNA makes the difference between a healthy chromosome and a chromosome with a lesion.
It was proposed that inhibited replication forks can reverse, extruding the newly synthesized DNA strands into a new duplex and forming a Holliday junction (Figure 1C; Scudiero and Strauss 1974; Higginset al. 1976; Morgan and Severini 1990; Louarnet al. 1991). Formation of such reversed replication forks has been inferred from chromosomal fragmentation dependent on the Holliday-junction-resolving enzymes (Seigneuret al. 1998; Floreset al. 2001; Gromponeet al. 2002), as well as observed by electron microscopy in preparation for replicating chromosomes from yeast checkpoint mutants, where DNA before isolation was crosslinked in vivo to preserve the original topology (Sogoet al. 2002). It was further proposed that the open double-strand end of the newly formed duplex is attacked by the RecBCD enzyme (Louarnet al. 1991) and that the duplex is either completely degraded, eliminating the Holliday junction (Figure 1, C → B; Kuzminov 1995), or recombined with the parental duplex ahead of the fork (Seigneuret al. 1998). Without such degradation, the Holliday junction is eventually cut by the RuvABC resolvasome, which breaks the replication fork (Figure 1, C → D; Seigneuret al. 1998). Interestingly, the chromosome fragmentation in recB recA mutants is suppressed to the recA levels by inactivation of the RuvABC resolvasome (Michelet al. 1997; Gromponeet al. 2002), suggesting reversed replication forks as one possible chromosomal lesion suppressed by the RecBCD-promoted DNA degradation. However, in our hands, inactivation of RuvABC did not significantly improve the viability of recA recBCD mutants (data not shown), so there must be other chromosomal lesions, prevented or removed by the RecBCD-promoted degradation.
RecBCD is not the only enzyme hypothesized to be involved in both the formation and subsequent repair/removal of chromosomal lesions. As detailed above, RuvABC is another enzyme involved at both stages. The difference is that, while both RecBCD and RuvABC help to repair chromosomal lesions, RecBCD also suppresses their formation, whereas RuvABC promotes their formation. Still other recombinational repair enzymes, RecG and PriA helicases, are recognized by their in vitro affinity to branched DNA structures as potential early players around inhibited replication forks (McGlynnet al. 1997; Whitby and Lloyd 1998; McGlynnet al. 2001; Gregget al. 2002). RecG and PriA recently have been proposed to repair stalled forks without breakage or recombination (Gregget al. 2002), but we found no predicted decrease in viability caused by recG inactivation in the double recA recBCD mutants (data not shown).
In summary, we propose that recBCD mutant cells are unable (1) to suppress breakage of inhibited replication forks independently of RecA and (2) to degrade the linear tails of σ-replicating chromosomes in the absence of RecA. We propose that this double defect is the reason why the viability of recA recBCD mutants is significantly lower than that of recA mutants. If we add to these two defects the deficiency in repair of broken replication forks, in which RecBCD is involved together with RecA, it becomes clear why recBCD mutants have such a low viability.
Acknowledgments
We are grateful to Doug Julin, Ichizo Kobayashi, Richard Kolodner, Sidney Kushner, Sue Lovett, Gerry Smith, Frank Stahl, and Gerard Venema for bacterial strains and plasmids. Elena Kouzminova, David Schlesinger, and two anonymous reviewers offered helpful comments on the manuscript. This work was supported by grant MCB-0196020 from the National Science Foundation.
Footnotes
-
Communicating editor: S. Lovett
- Received November 8, 2002.
- Accepted January 6, 2003.
- Copyright © 2003 by the Genetics Society of America
