|Home | About | Journals | Submit | Contact Us | Français|
Recombination between short linear double-stranded DNA molecules and Escherichia coli chromosomes bearing the red genes of bacteriophage λ in place of recBCD was tested in strains bearing mutations in genes known to affect recombination in other cellular pathways. The linear DNA was a 4-kb fragment containing the cat gene, with flanking lac sequences, released from an infecting phage chromosome by restriction enzyme cleavage in the cell; formation of Lac− chloramphenicol-resistant bacterial progeny was measured. Recombinant formation was found to be reduced in ruvAB and recQ strains. In this genetic background, mutations in recF, recO, and recR had large effects on both cell viability and on recombination. In these cases, deletion of the sulA gene improved viability and strain stability, without improving recombination ability. Expression of a gene(s) from the nin region of phage λ partially complemented both the viability and recombination defects of the recF, recO, and recR mutants and the recombination defect of ruvC but not of ruvAB or recQ mutants.
Efficient recombination involving the Escherichia coli chromosome takes place only when the recombining partner DNA is large and contains Chi sites to activate the recombination-promoting activities of RecBCD (for a review, see reference 20). Short linear DNA molecules in the cell generally are destroyed by RecBCD. Recombination between short linear DNA molecules and the host chromosome at a frequency high enough to be of practical use in making gene replacements has been observed with recD and recBC sbcBC mutant strains (12, 24). Still-higher frequency recombination is seen with E. coli strains in which the recBCD gene cluster is replaced by the red genes (gam, bet, and exo) of phage λ (19).
In addition to proceeding at high efficiency, Red-mediated recombination between a short linear DNA molecule and a circular homologue may represent a simpler recombination pathway than any of the previously characterized pathways for conjugational or transductional recombination. These properties of efficiency and (relative) simplicity recommend the hybrid phage-bacterial recombination system for research on general recombination mechanisms. In previous studies, we have shown that such recombination events require the activities of RecA, Exo, and Bet, as well as double-strand breaks (22). Murphy (19) found that the frequency is decreased by mutation of recA and recF and increased by mutation of recJ. The frequency of Red-mediated recombination is also elevated in a recG mutant strain. In the case of an event involving the insertion of substantial nonhomology (as in gene replacement), recombination in the recG host is apparently constrained to proceed through a pathway requiring RuvC resolvase (23).
In this study, we examined the dependence of Red-mediated gene replacement on several additional known E. coli recombination genes. Functional complementation between some of these genes and other phage λ genes was tested as well.
λ lac::cat819 nin5 has been described previously (23). Bacterial strains employed in this study are described in Table Table1.1. A number of them contain in vitro-assembled substitutions in which most of the coding sequence of a gene is replaced with a genetic element conferring resistance to either tetracycline or kanamycin or else with a short synthetic sequence that preserves the original reading frame. Details of these constructions will be described elsewhere (K. C. Murphy, K. Campellone, and A. R. Poteete, unpublished data).
Some of the strains described in Table Table11 contain an insertion in the galK gene of sequences from the nin region of the λ chromosome, fused to the promoter Ptac, along with a kanamycin resistance-conferring determinant derived from Tn903. The Ptac-nin fusion was constructed in four steps. (i) The EcoRI fragment of the chromosome of λ cI857 S7 bearing the nin region (bp 39168 to 44972) was ligated into the EcoRI site of pBR322, in the orientation in which the direction of transcription of the λ genes would be clockwise in the conventional map of pBR322. (ii) Sequences of the resulting plasmid between the PstI and SmaI sites, containing the N-terminal half of bla, were replaced by sequences between the PstI and PvuII sites of ptac12 (1) containing Ptac, with an XhoI linker (CCCTCGAGGG) inserted between the PvuII and SmaI ends. (iii) Sequences of the resulting plasmid, containing the transcriptional terminator tR2, were deleted by digestion with XhoI and StuI, filling in with DNA polymerase I large fragment, and ligation with BglII linkers (CAGATCTG). (iv) Sequences between the filled-in HindIII sites of the resulting plasmid were replaced with EcoRI linkers (GGAATTCC). An EcoRI fragment from the resulting plasmid, bearing the Ptac-nin fusion, was then ligated into the EcoRI site of a galK insertion vector. In the resulting plasmid, pTP878, the λ sequences are transcribed in the galK antisense direction. The galK insertion vector was constructed from pTP838 (Murphy et al., unpublished data) by digestion with ApaI and SacI, blunting the ends, and ligating EcoRI linkers between the flanking gal and kan sequences. The galK::Ptac-nin kan insertion of pTP878 was crossed into TP507 and TP554 (Table (Table1)1) by electroporation with PvuII-digested plasmid DNA, as described previously (19); kanamycin-resistant recombinants were selected.
The presence of Tn10, Tn5, and mini-Tn10-9(Kan) in various genes in strains described in Table Table11 was confirmed by production of appropriate-size DNA products in PCR with transposon- and chromosomal gene-specific primers. Cells from liquid cultures were pelleted, resuspended in water, and used directly as template. Primer tn5out (CCATGTTAGGAGGTCACATGGAAG) directs DNA synthesis from both ends of Tn5 outward; primer mtn10 (GATCATATGACAAGATGTGTATCCACCTT) does the same for Tn10 and mini-Tn10. Primers used for specific genes were as follows: edamU (CGCGGCCGGTATTCAGATTAACTG), recfU (ATCATCGAGCTCGAGATGGAAATGGTGGCACGTGTT), recfD (TC ATCAGAGCTCCGATTTCACCTCAGAAGAAACCAG), recjU (ATCATCGAGCTCAATTGACGTGTTGTTTCCCAGCCA), recjD (ATCATCGAGCTCTCCATCGCCTGTTTCTCGGCATTT), recoU (ATCATCGAGCTCCGCCGAACAGGCGTTGAAAAAACT), recoD (TCATCAGAGCTCGCTTTTGCTGCGGCTTCTTTCACA), recrU (ATCATCGAGCTCAAAGACTGGCTTCGGTCACCGAT), and recrD (TCATCAGAGCTCCTGGTGTACTCCTGCTTACCTTCA).
Crosses between λ lac::cat819 nin5 and bacterial strains were carried out as described previously (23). UV sensitivity was measured by plating bacteria on Luria-Bertani (LB) agar, exposing the plates to variable doses of UV (0, 10, 20, or 30 J/m2, measured with a Spectronics DM-254N shortwave UV meter), and incubating them at 37°C in the dark. Fractional survival was determined by colony counts relative to the unexposed control.
Experiments to measure recombination between a linear DNA fragment and the E. coli chromosome employed bacterial strains bearing a (recC-ptr-recB-recD)Δ::Ptac-gam-red-pae-cI substitution. In these strains, the E. coli recC, ptr, recB, and recD genes are replaced by the recombination genes of phage λ, the PaeR7 restriction-modification system, and the phage λ cI gene. Log-phase cultures of these bacteria are infected with λ lac::cat819 nin5. The injected phage chromosome circularizes but does not proceed through the lytic or lysogenic cycle because of the cI repressor present in the cell. The chromosomally encoded PaeR7 restriction endonuclease cuts the (unmodified) phage DNA at two sites, releasing a 4-kb linear DNA fragment consisting of the cat gene flanked by 1.5-kb lac sequences. Recombination between this fragment and the chromosome frequently results in gene replacement; recombinants are detected as white colonies on LB agar plates supplemented with chloramphenicol, IPTG (isopropyl-β-d-thiogalactopyranoside), and X-Gal (23).
In addition to the white colonies formed by gene replacement, blue (Lac+) chloramphenicol-resistant recombinants were formed as well. The numbers of Lac+ recombinants were variable, amounting to 5 to 60% of the total chloramphenicol-resistant population, and were found in crosses with all mutant hosts (data not shown). In experiments involving electroporation with pure DNA fragment preparations, such Lac+ recombinants were formed only when the DNA fragment bore one or two nonhomologous flanks (unpublished observations). The Lac+ recombinants formed after infection with λ lac::cat819 nin5, therefore, presumably reflect recombination reactions that took place between the bacterial chromosome and phages that were cut only once by the PaeR7 endonuclease. The Lac+ recombinants are the subject of ongoing investigation. Only Lac− recombinant production is considered below.
In the experiments reported below, defects in the production of recombinants are due to defects in the process of recombination, not in any of the steps involved in the delivery of the linear double-stranded DNA recombination substrate into the cell. All the strains that formed recombinants at reduced efficiency were found to plate Pae-modified λ imm22 and λ h80 imm22 at high efficiency (data not shown). This observation demonstrates that adsorption and injection of phage DNA are not impaired in these mutants. (It also demonstrates that these strains, many of which have poor viability, are as capable as the wild type of becoming infective centers.) In addition, all were found to plate unmodified λ h80 imm22 with an efficiency of less than 0.001 (data not shown). This observation indicates that PaeR7 cutting is efficient in all of them.
Mutant alleles of genes recA, recF, recJ, recO, recQ, recR, ruvA and ruvB, and ruvC were introduced into TP507 (AB1157 recBCDΔ::Ptac-gam-red-pae-cI822 ) and a derivative, TP554, in which the recG gene had been deleted (Table (Table1).1). Measurements of the recombination proficiencies of most of these strains are indicated in Table Table2.2.
As reported previously, or as expected from the results of similar experiments, recombination in the recG+ background was greatly decreased by mutation of recA, mildly decreased by mutation of ruvC, and stimulated by mutation of recJ (19, 23). Loss of ruvAB additionally caused a slight decrease in recombination rate. The recQ1802::Tn3 allele reduced recombination approximately 5-fold, while a deletion-substitution allele reduced it approximately 20-fold.
In the recGΔ background, the recombination rate was elevated and it was reduced more by mutation of ruvC than it was in the recG+ strains, as previously reported (23). In the recGΔ background, loss of ruvAB function led to a slightly greater loss of recombination proficiency than in the recG+ strain, but it is not clear that the difference is significant. As in the recG+ strain, recombination was strongly dependent upon recA, reduced 5-fold by recQ::Tn3 and 20-fold by recQΔ. In comparing the recG+ and recGΔ strains, the biggest difference in the dependence of recombination on specific E. coli genes was seen in the case of recJ. Loss of recJ function, while increasing recombination in the recG+ strain, decreased it 10-fold in the recGΔ strain. While this observation might seem to indicate an interesting mechanistic relationship between RecG and RecJ proteins, it is unclear that the effect of the double mutant is specific. The recGΔ recJ strain is only marginally viable (data not shown), and its recombination deficiency might be secondary to other defects.
The recF, recO, and recR mutants in both recG+ and recGΔ backgrounds exhibited even lower viability than did the recGΔ recJ strain, on the order of 1 CFU per 1,000 countable cells in log-phase culture (data not shown). Cultures of these mutants exhibited low, but highly variable, rates of recombination, possibly reflecting the outgrowth of revertants or pseudorevertants. We found that the viability of recF, recO, and recR mutants could be improved by deletion of sulA (an SOS-inducible gene, formerly known as sfiA, whose product is a cell division inhibitor; see reference 7) and so conducted tests in this background (Table (Table2).2). The sulA mutation itself has no effect on recombination frequency. Mutation of recF, recO, or recR substantially decreased recombination in both the sulAΔ recG+ and sulAΔ recGΔ backgrounds.
We used the sulAΔ strain to test whether induction of the SOS regulon would increase Red-mediated recombination activity by increasing the intracellular levels of recombination functions. Introduction of the lexA71::Tn5 mutation, however, slightly reduced recombination, in both the recG+ and recGΔ backgrounds (Table (Table22).
The sensitivity of the bacterial strains to UV radiation was tested. Results are shown in Table Table2.2. Mutations in all of the recombination genes tested significantly increased UV sensitivity, whether they decreased or increased the efficiency of recombination.
Proteins encoded by the recombination genes tested in this study have been extensively characterized. The RecF, RecO, and RecR proteins of E. coli form a complex that is thought to function in recombination by modulating the DNA binding of RecA (10, 32, 33). RecQ protein is a helicase which can cooperate with RecA and SSB proteins to initiate recombination-like events in vitro (9, 30). The RuvA and RuvB proteins form a complex that catalyzes the branch migration of crossover structures in DNA (17, 18). The complex of RuvAB with DNA is thought to direct the action of RuvC, which resolves Holliday junctions by making strand-specific cuts (31).
Two genes in the nin region of phage λ, orf and rap, function in homologous recombination. The orf gene can substitute for recF, recO, and recR in phage recombination mediated by the host system in recBC sbcBC cells in the absence of the λ Red system (26). The rap gene encodes an endonuclease which specifically cleaves branched DNA structures, including Holliday junctions, that are thought to be generated during recombination and which hypothetically might substitute for RuvC (11, 28). We tested whether λ nin genes could functionally replace any of the E. coli genes needed for Red-mediated gene replacement, by inserting a Ptac fusion of orf and other λ genes downstream of it into the galK gene. Details of the construction are given above.
Complementation activity was seen with both recG+ and recGΔ backgrounds but more strongly in the recGΔ strain. As shown in Table Table2,2, in the nin+ recGΔ background, loss of recF, recO, or recR function reduced recombination, but less than in the recGΔ strain lacking nin functions. The smaller effects of recF, recO, and recR mutations in this background occurred in spite of the fact that the alleles used in this case were deletions. Partial complementation of the defects of recF, recO, and recR mutants by nin was also evident in the viability of these mutant strains, which was better than that of their nin-less counterparts (data not shown). The nin genes also were seen to compensate for loss of ruvC in gene replacement recombination. In contrast, nin did nothing to remedy the recombination defects of recA, recQ, or ruvAB mutants. The nin genes did not compensate for loss of any of the recombination genes in providing resistance to the lethal effects of UV radiation (Table (Table22).
A hypothetical mechanism for Red-mediated recombination leading to gene replacement in E. coli recGΔ, based on an earlier scheme (23), on results shown in Table Table2,2, and on the research on recombination proteins cited above, is shown in Fig. Fig.1.1. Results presented above do not distinguish which λ proteins encoded by nin genes contribute to complementation of the recombination defects of mutant strains lacking RecFOR or RuvC. However, research by others on orf and rap strongly suggests that these are the relevant functions (11, 26, 28). The mechanism shown in Fig. Fig.11 accounts for the main pathway to recombinant formation in a recG mutant cell. In a recG+ cell, other pathways, frequently not leading to recombinant formation, are apparently more prevalent (23).
Results described above and in previous work (19, 23) indicate the importance of recA, recF, recO, recR, recQ, ruvAB, and ruvC, and the inhibitory influence of recJ and recG, in Red-mediated gene replacement in E. coli lacking RecBCD function. These observations distinguish the hybrid phage-bacterial recombination pathway from the classical RecF and RecE pathways for conjugational and transductional recombination, in that the latter are partially dependent upon both recJ and recG (for a review, see reference 13). However, strains of E. coli bearing chromosomal substitutions of the λ red genes for the recC-ptr-recB-recD cluster are particularly analogous to recB recC sbcA strains, in which RecBCD is functionally replaced by the induced recombination functions recE and recT of the cryptic lambdoid prophage Rac (3–6, 8). A more direct comparison would be needed to determine whether the Red pathway (with or without expression of other λ recombination genes) is mechanistically different from the RecE pathway.
We thank Steven Sandler, Kenan Murphy, and Kenneth Campellone for strains, helpful advice, and discussions.
This work was supported by Public Health Service grant GM51609 from the National Institutes of Health.