Plasmid construction

We constructed a series of plasmids (Figure 1 and Table 1) encoding various truncated tetA genes to assay plasmid recombination frequencies using the strategies similar to those described previously [28, 35]. Restoration of a functional tetA gene via intra- or intermolecular recombination resulted in a change of the bacterial phenotype from tetracycline sensitive to tetracycline resistant, and served as a marker allowing us to measure the frequency of recombination events (Figure 2).

Plasmids pYA4463 and pYA4590 were constructed to test intraplasmid recombination (Figure 1 panel A). Plasmid pYA4463 carries two truncated tetA genes (5' end and 3'end), which have 466-bp of tandemly repeated sequence. An intramolecular recombination event can delete one of the repeats resulting in an intact tetA gene, thereby recreating the structure of plasmid pACYC184 (Figure 1 panel A). Theoretically, intermolecular recombination may occur between two pYA4463 molecules to form a plasmid dimer with a functional tetA gene (Figure 1 panel C). Plasmid pYA4590 contains a 602-bp tetA sequence duplication separated by a 1041-bp kan cassette. The intramolecular recombination product is equivalent to pACYC184. The intermolecular recombination product is a dimer plasmid containing an intact tetA gene (Figure 1 panel C). Plasmids pYA4464 and pYA4465 carry the 3'tet gene and 5'tet gene, respectively (Figure 1). The Rec⁺ Salmonella strain χ3761 carrying either plasmid individually was sensitive to tetracycline. There is 751-bp of tetA DNA in common between the two truncated tetA genes. Recombination between the two plasmids creates a hybrid plasmid containing an intact tetA gene (Figure 1 panel C).

Intraplasmid recombination products

To verify the recombination products, plasmid DNA was prepared from tetracycline resistant (Tc^R) single colonies derived from χ3761(pYA4463), χ3761(pYA4590) and χ3761(pYA4464, pYA4465). Plasmids extracted from Tc^R clones of χ3761(pYA4463) were digested with XbaI and SalI. Theoretically, XbaI/SalI digestion of pYA4463 will yield two fragments (3524 bp and 1187 bp), pACYC184 will yield two fragments (3524 bp and 721 bp) and pYA4463 dimer will yield four fragments (3524 bp, 3524 bp, 1653 bp and 721 bp). The results (Figure 3A) showed that digestion of all 16 Tc^R clones yielded a 721-bp band, indicating either a pYA4463 dimer or a plasmid equivalent to pACYC184. Three clones (lane 1, 5 and 10) yielded the pYA4463 dimer-specific 1653-bp band. Therefore, we conclude that the other 13 clones recombined to form the pACYC184-like structure. Of note, several clones (2, 13-16) also yielded the 1187-bp pYA4463-specific band, suggesting that the original plasmid (pYA4463) and its recombination product (pACYC184-like) could coexist in the same bacterial cell.

Plasmids extracted from Tc^R clones of χ3761(pYA4590) were digested with KpnI and EcoRI. Theoretically, plasmid pYA4590 will be digested into two fragments (3414 bp and 2474 bp), plasmid pACYC184 will be linearized (4245 bp) and the pYA4590 plasmid dimer will be digested into four fragments (4245 bp, 3414 bp, 2474 bp and 1643 bp). Examination of the restricted DNA (Figure 3B) showed that only one clone (lane 12) had the pYA4590 dimer-specific 1643-bp band. The most prominent band in the other lanes was a 4245-bp band expected for pACYC184-like recombination products. Nine clones contained a mixture of pACYC184 and pYA4590 (lane 1, 3-5, 8, 9, 14-16).

Interplasmid recombination products

Plasmids extracted from Tc^R clones of χ3761(pYA4464, pYA4465) were digested with NcoI and BglII. Both pYA4464 and pYA4465 are linearized into a DNA fragment about 4 kb. Therefore, in cells containing each or both monomeric plasmids, the digested product will be a single band. The pYA4464-pYA4465 hybrid will be cut into two fragments (5510 bp and 2481 bp). All four of the Tc^R clones we isolated and examined showed recombination product specific bands and the 4-kb band expected when each plasmid exists separately in the cell. Four tetracycline sensitive (Tc^S) isolates were examined and only a single band was observed, as expected (Figure 3C). These results suggest that interplasmid recombination occurred in the Tc^R cells and that both dimer and individual monomers corresponding to at least one of the two starting plasmids can coexist in the same bacterial cell. We performed a similar experiment in S. Typhi strain Ty2(pYA4464, pYA4465) and obtained identical results (data not shown).

Construction of recdeletion strains

We constructed a series of strains for these studies carrying deletions in either recA, recF or recJ in S. Typhimurium UK-1, S. Typhi Ty2 and S. Paratyphi A (Table 2). We also constructed ΔrecAΔ recF and ΔrecJ Δ recF double mutants in S. Typhimurium. Deletion of recA, recF and recJ results in an increase in sensitivity to UV irradiation [36, 37]. To verify the presence of these deletions phenotypically in our strains, the UV sensitivity of the S. Typhimurium mutant strains was measured. The ΔrecF and ΔrecJ mutants showed significantly lower surviving fractions than the wild type strain after the same exposure dose (Figure 4). By contrast, after five seconds of UV exposure (16 J/m²) to 2.2 × 10⁹ CFU of the ΔrecA62 mutant (χ9833), we were unable to recover any surviving cells (not shown). UV resistance similar to the wild-type strain χ3761 was restored to S. Typhimurium ΔrecA and ΔrecF mutants strains after introduction of recA plasmid (pYA5002) or either recF plasmid (pYA5005/pYA5006), respectively. Transformation of either mutant strain with vector plasmid pYA5001 did not restore UV resistance (Figure 4 and data not shown for recA mutant).

Strain	Genotype* [parental strain]	Reference or source
S. Typhimurium UK-1
χ3761	wild type	[62]
χ9833	ΔrecA62 [χ3761]	This study
χ9070	ΔrecF126 [χ3761]	This study
χ9072	ΔrecJ1315 [χ3761]	This study
χ9081	ΔrecJ1315 ΔrecF126 [χ9072]	This study
χ9931	cysG494::(5'tet-kan-3'tet) [χ3761]	This study
χ9932	ΔrecF126 cysG494::(5'tet-kan-3'tet) [χ9070]	This study
χ9933	ΔrecJ1315 cysG494::(5'tet-kan-3'tet) [χ9072]	This study
χ9934	ΔrecA62 cysG494::(5'tet-kan-3'tet) [χ9833]	This study
χ9935	cysG493::(5'tet-kan) [χ3761]	This study
χ9936	ΔrecF126 cysG493::(5'tet-kan) [χ9070]	This study
χ9937	ΔrecJ1315 cysG493::(5'tet-kan) [χ9072]	This study
χ9938	ΔrecA62 cysG493::(5'tet-kan) [χ9833]	This study
χ9939	ΔrecF126 Δ recA62 [χ9070]	This study
S. Typhi Ty2
χ3769	wild type	[63]
χ11053	ΔrecF126 [χ3769]	This study
χ11134	ΔrecF1074 [χ3769]	This study
χ11159	ΔrecA62 [χ3769]	This study
χ11194	ΔrecJ1315 [χ3769]	This study
S. Typhi ISP1820
χ3744	wild type	D.M. Hone
χ11133	ΔrecF1074 [χ3744]	This study
S. Paratyphi A
χ8387	Plasmid pSPA1 was cured from wt isolate ATCC 9281	This study
χ11243	ΔrecA62 [χ8387]	This study
χ11244	ΔrecF126 [χ8387]	This study
χ11245	ΔrecJ1315 [χ8387]	This study
E. coli K-12
EPI300	F- mcrA Δ (mrr-hsdRMS-mcrBC) Φ80dlacZ Δ M15 Δ lacX74 recA1 endA1 araD139 Δ (ara, leu)7697 galU galK λ- rpsL nupG trfA dhfr	Epicentre
χ7213 (MGN-617)	thi-1 thr-1 leuB6 glnV44 fhuA21 lacY1 recA1 RP4-2-Tc::Muλpir ΔasdA4 Δzhf-2::Tn10	[55]

* kan: kanamycin resistance gene; 5'tet: 5' portion of the tetA gene together with its promoter; 3'tet: 3' portion of the tetA gene.

Effect of recdeletions on intraplasmid recombination

The results with plasmid pYA4590 were also variable among strains. The recombination frequency in Rec⁺ S. Typhimurium and S. Paratyphi A strains was approximately 2-3 × 10^-3 and in both S. Typhi strains, the frequency was 3-fold higher, at 1.16 × 10^-2 (Ty2) and 1.31 × 10^-2 (ISP1820). In S. Typhimurium and S. Typhi Ty2, the Δ recA and ΔrecF mutations reduced the recombination frequency of plasmid pYA4590 by 5-20-fold (P < 0.01; Table 3). The results were similar for S. Paratyphi A, though the ΔrecF mutation only led to 3-fold lower plasmid pYA4590 recombination (P < 0.01). The ΔrecJ mutation had no effect in S. Typhimurium and resulted in a 2-3-fold decrease in recombination in both S. Typhi Ty2 and S. Paratyphi A. Combining the ΔrecA ΔrecF mutations in S. Typhimurium led to a recombination frequency similar to the frequencies observed for both mutations individually, indicating no additive effect. In the complementation test, plasmid pYA5002, which encodes S. Typhimurium recA, was transformed into S. Typhimurium ΔrecA mutant χ9833(pYA4590) and S. Typhi ΔrecA mutant χ11159(pYA4590). Their respective recombination frequencies were 2.50 ± 0.42 × 10^-3 and 14.35 ± 2.44 × 10^-3, which were comparable to the corresponding wild type strains (P > 0.05) (Table 3). The recF-encoding plasmids pYA5005 and pYA5006 were transformed into recF mutant strains χ9070(pYA4590) and χ11053(pYA4590), respectively. The respective recombination frequencies were increased to 2.00 ± 0.24 × 10^-3 and 2.86 ± 0.59 × 10^-3.

Effect of recdeletions on interplasmid recombination

To evaluate interplasmid recombination, plasmids pYA4464 and pYA4465 were co-electroporated into the wild-type and rec deletion strains. Electroporants from each test strain were grown in LB broth containing both ampicillin and chloramphenicol to maintain selection for both plasmids. The frequency of recombination was determined as described in the Methods section. The interplasmid recombination frequency was 1-4 × 10^-3 for Rec⁺ S. Typhimurium, S. Typhi and S. Paratyphi A strains (Table 3). For Typhimurium and Paratyphi A, the ΔrecA and each ΔrecF mutation reduced the interplasmid recombination frequency by about 3-10-fold (P < 0.01). In contrast, the ΔrecA mutation had no effect on interplasmid recombination in S. Typhi Ty2. The ΔrecF mutations did not reduce interplasmid recombination in either of the Typhi strains. Surprisingly, introduction of the ΔrecF1074 mutation into S. Typhi Ty2 resulted in significantly higher interplasmid recombination (P < 0.01). Note that we performed this analysis in eight independent experiments and observed a higher recombination frequency of interplasmid recombination each time. The ΔrecJ mutation had no significant effect in S. Typhi, and a small (< 3-fold) but significant effect in S. Typhimurium and S. Paratyphi A. The recombination frequencies were also determined in S. Typhimurium strains ΔrecA ΔrecF and ΔrecF ΔrecJ double deletions. No additive effect between the two mutations was observed with respect to each single mutation.

Effect of recdeletions on chromosome related recombination

To examine plasmid integration, the 5'tet gene was introduced into the S. Typhimurium chromosome at cysG. The resulting strains were transformed with plasmid pYA4464 (3'tet) (Figure 1B). The 789 bp of overlapping sequence between 5'tet on the chromosome and the 3'tet on the plasmid could result in plasmid integration into the chromosome, generating an intact tetA gene (Figure 2B). Deletion of recA had a profound effect, reducing the integration frequency to less than 7 × 10^-10, which was below the limits of detection in this assay (P < 0.01), indicating a strict requirement for RecA in this process. Introduction of plasmid pKD46, which encodes the λ Red recombinase, into χ9938 (ΔrecA) carrying pYA4464 restored the integration frequency to the level of the Rec⁺ strain χ9935. Deletion of recF reduced the frequency of integration less than 3-fold (P < 0.01; Table 4) and the ΔrecJ deletion had no effect.

Effect of rec deletions on the virulence of S. Typhimurium

Discussion

We began our studies using information gathered in E. coli as a reference point. In E. coli, recA-dependent homologous recombination relies on the RecBCD pathway, the RecFOR pathway (originally designated the RecF pathway) and the RecE pathway [38]. The RecBCD pathway is important in conjugational and transductional recombination [39], and may also be involved in the recombination of plasmids containing one or more Chi sites [40]. Recombination in small plasmids lacking a Chi sequence is primarily catalyzed by the RecFOR pathway [41]. RecF, RecO, and RecR bind to gaps of ssDNA and displace the single-strand DNA binding proteins to allow RecA to bind [42, 43]. The RecJ ssDNA exonuclease acts in concert with RecFOR to enlarge the ssDNA region when needed. Strand exchange is then catalyzed by RecA [44]. Because of their prominent role in plasmid recombination in E. coli, we analyzed the effect of mutations in recF, recJ and recA on plasmid recombination in Salmonella.

Attenuated S. Typhi strains have been developed as antigen delivery vectors for human vaccine use. Due to the host restriction phenotype of S. Typhi, preliminary work is typically done in S. Typhimurium using mice as the model system to work out attenuation and antigen expression strategies. Recently, we have also been investigating attenuated derivatives of the host-restricted strain S. Paratyphi A as a human vaccine vector. Therefore, it was of interest to evaluate and compare the effects of rec mutations in these three Salmonella serovars. We selected S. Typhi strain Ty2 as exemplary of this serovar because most of the vaccines tested in clinical trials to date have been derived from this strain [45]. S. Typhi strain ISP1820 has also been evaluated in clinical trials [46, 47] and we therefore included it in some of our analyses. We found that, for some DNA substrates, the effects of ΔrecA and ΔrecF deletion mutations differed among Salmonella enterica serotypes. In particular, we found that deleting recA, recF or recJ in S. Typhi Ty2 and deleting recF in strain ISP1820 had significant effects (3-10 fold) on the recombination frequency of our direct repeat substrate, pYA4463 (Table 3). No or very limited effect (< 2 fold) was observed for our S. Typhimurium and S. Paratyphi A strains, consistent with results reported for E. coli indicating that recombination of this type of substrate is recA-independent [35]. In contrast, the ΔrecA and ΔrecF mutations resulted in lower interplasmid recombination in Typhimurium and Paratyphi A but not in Typhi strains. Deletion of recJ led to a reduction in intraplasmid recombination frequencies in S. Typhi, while no effect was seen in S. Typhimurium. The ΔrecJ mutation also affected plasmid recombination frequencies for two of the three substrates tested in S. Paratyphi A. Taken together, these results suggest that the recombination system in S. Typhi, or at least in strains Ty2 and ISP1820, is not identical to the recombination system in S. Typhimurium and S. Paratyphi A.

To investigate the mechanism responsible for the observed differences, we analyzed the genome sequences of S. Typhimurium UK-1 (Luo, Kong, Golden and Curtiss, unpublished whole genome sequence), S. Paratyphi A (NC_006511) [48] and S. Typhi Ty2 (NC_004631) [49]. No paralogs of the recA, recF and recJ genes were found in the three strains. The S. Typhimurium UK-1 has RecA, RecO and RecR protein sequences identical to Typhi Ty2, and RecF and RecJ protein sequences with over 99% identity. Plasmids expressing Typhimurium recF or Typhi recF complemented the ΔrecF126 mutation in Typhi, as evidenced by the UV sensitivity profile (Figure 4) and intraplasmid recombination of pYA4463 (Table 3). Therefore, the basis for these differences are not clear and indicates that there may be other genes or gene products involved. A more detailed analysis of this phenomenon is under investigation.

Plasmid recombination frequencies were higher in our Salmonella strains than those reported in E. coli. We observed intra- and interplasmid recombination frequencies on the order of 1 × 10^-3 in Rec⁺ Salmonella, whereas measurements made in E. coli strain AB1157 using a similar plasmid system (equivalent to our substrates pYA4590 and pYA4464 + pYA4465) revealed a basal frequency around 10-fold lower, approximately 1 × 10^-4 for both types of substrates [26]. Interestingly, the effect of a recF mutation in E. coli was to reduce the recombination frequency of intra- and interplasmid recombination approximately 30-fold, to roughly the same frequencies we observed for S. Typhimurium (Table 3). However, consistent with the results in E. coli, the effects of recA, recF, and recA recF mutations were similar, indicating that the mutations are epistatic.

RecF has been shown previously to play a role in recombinational repair of chromosomal DNA in response to DNA damaging agents [50], including a major role in homologous recombination between direct repeats in the chromosome of S. Typhimurium. In our study, we did not observe any effect of recF on intrachromosomal recombination, although it did have an effect on the frequency of plasmid integration (Table 4). This discrepancy can be explained by the fact that we did not use DNA damaging agents in our study. These agents lead to single stranded stretches of DNA that represent substrates for recF (and recA). Our observation that recF did affect plasmid integration may reflect the presence of stretches of ssDNA in the plasmid, presumably due to supercoiling effects.

To induce strong primary and memory immune responses, Salmonella delivery vectors should be sufficiently invasive and persistent to allow antigen expression in targeting organs, while maintaining a high degree of safety. This requires the use of mutations that attenuate the Salmonella vector without impairing its antigen delivery ability. Many attenuating mutations impair invasion and colonization ability. In our study, we confirmed a previous report that recF is not required for S. Typhimurium virulence in mice [51], indicating that the recF mutant remains invasive and replicates well in colonized organs. Therefore, including a ΔrecF mutation in a Salmonella vaccine strain is unlikely to affect its immunogenicity. Our results with the S. Typhimurium ΔrecA strain are consistent with two previous, independent studies showing that recA mutations reduce Salmonella virulence [51, 52]. To evaluate the potential effect of ΔrecA mutation on immunogenicity, mice inoculated with the recA mutant were challenged with a lethal dose of virulent wild-type S. Typhimurium. All the challenged mice survived, indicating that a ΔrecA mutant retains immunogenicity and therefore may be suitable for use in a vaccine. However, since it does not affect virulence, inclusion of a ΔrecF mutation into a Salmonella vector that has been attenuated by other means to reduce the frequency of intra- and interplasmid recombination, may be more desirable than a ΔrecA mutation. Studies are currently underway to investigate these possibilities.

Our data show that ΔrecA and ΔrecF mutations resulted in reduced frequencies of intraplasmid recombination in all Salmonella strains tested, which included three serovars, when there was an intervening sequence between the direct duplications (Table 3). Our results also show that it is likely that deletions in recA, recF or recJ will not be useful for reducing interplasmid recombination in S. Typhi vaccine strains, since we did not observe any reduction in interplasmid recombination frequency. This result was disappointing, since the majority of human trials with live Salmonella vaccines have focused on S. Typhi. In the case of S. Typhi, it appears that the best approach to preventing interplasmid recombination will be in the careful design of each plasmid, avoiding any stretches of homology. However, for vaccines based on S. Typhimurium or S. Paratyphi A, introduction of a ΔrecF mutation into attenuated Salmonella vaccine strains carrying multiple plasmids is a useful approach to reduce unwanted plasmid/plasmid or plasmid/chromosome recombination without further attenuating the strain or negatively influencing its immunogenicity. The ΔrecA mutation had a similar or more pronounced effect on reducing various classes of recombination and it clearly had an effect on virulence. We did not examine the effect of a ΔrecA mutation on the immunogenicity of a vectored antigen. Based on its effect on virulence, it may affect the immunogenicity of the vectored antigen in some attenuation backgrounds and therefore may not be applicable for all attenuation strategies.

Bacterial strains and media

E. coli K-12 strain EPI300™ was used for cloning and stable maintenance of plasmids. All Salmonella strains used in this work were derived from Salmonella enterica serovar Typhimurium wild-type (wt) strain χ3761 (UK-1), serovar Typhi strains Ty2 and ISP1820 or serovar Paratyphi A strain χ8387. Their origin and relevant genotypes are presented in Table 2. Bacteria were grown in LB broth [53].

Plasmid construction

To construct plasmid pYA4463 (Figure 1 panel A), a XbaI-HincII fragment containing the tetA promoter and 568 bp of the 5' end of tetA, was excised from pACYC184 and ligated into XbaI-EcoRV digested pACYC184.

To generate plasmid pYA4590 (Figure 1 panel A), the 5' end of tetA gene together with its promoter was amplified from pACYC184 with primers P1 and P2, which contain engineered XbaI and KpnI restriction sites, respectively. The resulting PCR fragment was digested with XbaI and KpnI. The kan gene was amplified from plasmid p15A-PB2-kan, a pACYC184 derivative carrying a influenza virus PB2 gene and a kan cassette, with primers P3 and P4, which were engineered to contain KpnI and BamHI sites, respectively. The resulting PCR fragment was digested with KpnI and BamHI. The two digested PCR fragments were ligated into pACYC184 digested with XbaI and BamHI. The resulting plasmid, pYA4590, contains the tetA promoter and 891 bp of the 5' end of tetA, a 1041-bp fragment encoding kan and its promoter followed by 902 bp of the 3'end of tetA.

To construct plasmid pYA4464 (Figure 1 panel B), plasmid pACYC184 was digested with XbaI and EcoRV to remove the 5' 102 bp of the tetA gene and the tetA promoter. The cohesive ends were filled using the Klenow large fragment of DNA polymerase and the linear plasmid was self-ligated to yield plasmid pYA4464.

To construct plasmid pYA4465 (Figure 1 panel B), the 5' 853 bp of tetA together with its promoter was amplified from pACYC184 using primers P5 and P6, which were engineered with SmaI and BglII sites, respectively. The resulting PCR fragment was digested with SmaI and BglII, and ligated to EcoRV and BglII digested pBAD-HisA.

Creation of recdeletions

The recA62 deletion, which deletes 1062 bp, encompassing the entire recA open reading frame, introduced into the bacterial chromosome using either λ Red recombinase-mediated recombination [54], or conjugation with E. coli strain χ7213(pYA4680) followed by selection/counterselection with chloramphenicol and sucrose, respectively [55]. The cat-sacB cassette was amplified from plasmid pYA4373 by PCR with primers P7 and P8 to add flanking sequence. The PCR product was further amplified with primer P9 and P10 to extend the flanking sequence. Those two steps of amplification resulted in the cat-sacB cassette flanked by 100 bp of recA flanking sequences at both ends. The PCR product was purified with QIAquick Gel Extraction Kit (QIAGEN) and electroporated into Salmonella strains carrying plasmid pKD46 to facilitate replacement of the recA gene with the cat-sacB cassette. Electroporants containing the cat-sacB cassette were selected on LB plates containing 12.5 μg chloramphenicol ml^-1. From S. Typhimurium chromosome, a 500-bp sequence upstream recA gene was amplified with primers P11 and primer P12 and a 500-bp sequence downstream recA gene was amplified with primers P13 and P14. Primers P12 and P13 were engineered with a KpnI site. The two PCR fragments were digested with KpnI, ligated and amplified with primers P11 and P14. The resulting PCR product was digested with isocaudarner SpeI and XbaI and ligated into XbaI-digested pRE112 to yield plasmid pYA4680. In addition, undigested, agarose-gel purified PCR product was electroporated into the cat-sacB Salmonella strains carrying plasmid pKD46 and spread onto LB plates containing 5% sucrose to select for deletion of the cat-sacB cassette. Chloramphenicol-sensitive isolates were verified as ΔrecA62 by PCR using primers P15 and P16 (ΔrecA62: 1360 bp; wt: 2412 bp). S. Typhimurium strains χ9833 and χ9939 were constructed by this method (Table 2). For construction of a ΔrecA62 mutant of S. Typhi, wild-type strain Ty2 was mated with E. coli strain χ7213(pYA4680). Transconjugants were selected on LB plates containing chloramphenicol, followed by counterselection on sucrose plates as described above. The resulting ΔrecA62 strain was designated χ11159. The S. Paratyphi A strain χ11243 was generated from wild-type strain χ8387 using the same strategy.

The ΔrecF deletion strains were constructed using suicide vectors pYA3886 and pYA4783. From the S. Typhimurium chromosome, a 397-bp sequence upstream of the recF gene was amplified with primers P17 and P18, which were engineered with XbaI and KpnI sites, respectively. The downstream 296-bp sequence (including 78 bp from the 3' ORF of recF) was amplified with primers P19 and P20 containing KpnI and SphI sites, respectively. The two fragments were digested and inserted into XbaI-SphI digested pRE112, resulting in plasmid pYA3886. The corresponding deletion was designated ΔrecF126. Strains χ9070, χ9081 and χ11244 were generated by conjugation using E. coli strain χ7213(pYA3886). Phage P22HTint mediated transduction was used to construct Typhi strain χ11053 [56]. The ΔrecF126 deleted 996 bp from the 5'end of recF in serovars Typhimurium and Paratyphi. The upstream flanking sequence of S. Typhi is different with the other serotypes. To construct a serovar Typhi-specific ΔrecF mutation, we constructed a new suicide vector. The recF upstream flanking sequence in plasmid pYA3886 was replaced with the corresponding DNA sequence (447 bp) from S. Typhi Ty2. Primers P21 and P22 were used for this modification. The resulting plasmid was designated as pYA4783. The Typhi-specific ΔrecF1074 mutation was introduced into S. Typhi strains ISP1820 and Ty2 by conjugation with E. coli strain χ7213(pYA4783) to yield strains χ11133 and χ11134, respectively. Primers P23 and P24 were used to verify the recF126 and recF1074 deletions.

Similar strategies were used to construct the Δ recJ1315 deletion with suicide vector pYA3887. From the S. Typhimurium chromosome, 330 bp upstream of the recJ gene was amplified with primers P25 and P26, which were engineered with XbaI and KpnI sites, respectively. The 299-bp downstream sequence was amplified with primers P27 and P28, engineered with KpnI and SphI sites, respectively. The two fragments were digested and ligated with XbaI-SphI digested pRE112. The resulting plasmid was designated pYA3887 and the corresponding deletion was named ΔrecJ1315. Strains χ9072 and χ11245 were generated by conjugating the parental strains with E. coli strain χ7213(pYA3887). Strain χ11194 was constructed by phage P22HTint mediated transduction. The ΔrecJ1315 mutation is a deletion of the entire recJ gene (1734 bp). Primers P29 and P30 were used to verify the recJ1315 deletion (ΔrecJ1315: 736 bp; wt: 2461 bp).

To test chromosome-related recombination, the 5'tet and 3'tet fragments were inserted into the cysG gene of each S. Typhimurium strain using the λ Red system. The 460-bp fragment of the cysG gene was amplified using primers P31 and P32 that were engineered with HindIII and BglII sites, respectively. The PCR product was digested with HindIII and BglII. A 480 bp adjoining fragment of cysG was amplified with primers P33 and P34. Primer P33 was engineered with BglII and PstI sites and primer P34 was engineered with a SacI site. The PCR product was digested with BglII and SacI. The two digested PCR fragments were ligated into HindIII and SacI digested pYA4518, deleting green fluorescent protein (GFP) gene. The resulting plasmid pYA4518-cysG has BssHII and PstI sites between the two cysG-fragments. This plasmid was digested with BssHII, followed by treatment with the Klenow large fragment. The linear plasmid was further digested with PstI for insertion of truncated tetA genes. The 5'tet-kan-3'tet cassette was amplified from pYA4590 with primers P35 and P36. Primer P36 was engineered with a PstI site. The PCR product was digested with PstI and inserted between the cysG fragments in pYA4518-cysG to yield plasmid pYA4689. The 5'tet-kan cassette was amplified from pYA4590 with primers P35 and P37. Primer P37 was engineered with a PstI site. The PCR product was digested with PstI and inserted into treated pYA4518-cysG to obtain plasmid pYA4690. The 5'tet-kan-3'tet cassette, together with cysG flanking sequences, was amplified from pYA4689 using primers P31 and P34. The PCR product was electroporated into strains χ3761(pKD46), χ9070(pKD46), χ9072(pKD46) and χ9833(pKD46) with selection on LB plates containing 25 μg/ml chloramphenicol. After growth at 37°C to cure plasmid pKD46, the resulting strains containing chromosomal copies of the 5'tet-kan-3'tet cassette in cysG were designated χ9931 (Rec⁺), χ9932 (ΔrecF), χ9933 (ΔrecJ) and χ9934 (ΔrecA), respectively. Primers P38 and P39 were used to verify insertion in the cysG gene. The 5'tet-kan cassette together with cysG flanking sequences was amplified from pYA4690 with primers P31 and P34. Using the same strategy, the PCR product was electroporated into pKD46 transformants of strains χ3761, χ9070, χ9072 and χ9833 to yield strains χ9935 (Rec⁺), χ9936 (ΔrecF), χ9937 (ΔrecJ) and χ9938 (ΔrecA), respectively, each containing the 5'tet-kan cassette inserted into cysG. These strains were transformed with plasmid pYA4464 to test plasmid integration based on the 789-bp of tetA sequence common to both the plasmid and the bacterial chromosome.

Analysis of recombination frequency

To examine plasmid recombination and plasmid integration, plasmid(s) containing truncated tetA genes were introduced into Salmonella strains with or without rec mutations. The resulting strains were inoculated into 3 ml of LB broth supplemented with 100 μg/ml ampicillin and/or 25 μg/ml chloramphenicol, as needed. After 8 h growth at 37°C, bacteria were serially diluted in 10-fold steps. 100 μl of the 10^-2, 10^-3 or 10^-4 dilution were spread onto LB-agar plates supplemented with 10 μg tetracycline ml^-1 and 100 μl of the 10^-5, 10^-6 or 10^-7 dilutions were spread onto LB-agar plates with or without the addition of antibiotics, as needed. Plates were incubated overnight at 37°C. The ratio of tetracycline resistant colonies to total colonies was calculated as the recombination frequency. The average mean frequency was calculated using the frequencies obtained from 3-10 assays for each strain. Following one-way ANOVA, the Dunnett's test was used to compare multiple groups against the control. The Student's t-test was used to analyze two independent samples.

Complementation of recmutation

Plasmid pYA5001 has a pSC101 ori, a gentamicin resistance marker and a prokaryotic green fluorescent protein (GFP) gene cassette flanked by two AhdI sites. A linearized T vector for cloning PCR products can be obtained by removing the GFP cassette by AhdI digestion. The recA genes from S. Typhimurium and S. Typhi were amplified using their respective chromosomal DNAs as template with primers P40 and P41. The recF genes were amplified similarly using primers P42 and P43. The forward primer P42 was engineered to include the S. Typhimurium lpp promoter sequence ttctcaacataaaaaagtttgtgtaatact (the -35 and -10 boxes are underlined). Amplified DNA fragment were treated with Taq DNA polymerase in the presence of dATP to add 3' A overhangs. Then the treated PCR products were cloned into pYA5001-derived T vector to yield recA plasmids pYA5002 (Typhimurium) and pYA5004 (Typhi), and recF plasmids pYA5005 (Typhimurium) and pYA5006 (Typhi). The recA plasmids, recF plasmids or empty vector plasmid pYA5001 were transformed into S. Typhimurium recA or recF mutants, respectively for complementation studies. The recA and recF plasmids were also introduced into Salmonella strains carrying pYA4590 or pYA4463 to complement the rec mutation and measure the plasmid recombination frequency.

UV sensitivity test

Quantitative UV killing curves were measured as described previously [57]. Briefly, cells were grown in 3 ml of LB broth at 37°C with vigorous shaking to mid-log phase. The cells were then 10 fold serially diluted in buffered saline with gelatin (BSG) and spread on LB agar plates. Multiple dilutions were exposed to 254 nm UV in a dark room at each designated dose. Then the plates were wrapped with aluminum foil and placed at 37°C overnight. The 10^-6 dilutions were not exposed to UV to determine the total bacterial cell numbers present in the culture. Surviving fractions were calculated as the CFU remaining after UV exposure/total CFU present.

Virulence determination of the recmutants

Eight-week old BALB/c female mice were purchased from Charles River Laboratories (Wilmington, MA). Mice were held in quarantine for 1 week before use in experiments. Food and water were deprived 6 h before administration of bacteria. Each mouse was orally inoculated with 20 μl of Salmonella suspended in buffered saline with gelatin (BSG) by pipet feeding. Food and water were returned 30 min after inoculation. All mice were observed for a month to record mortality. The 50% lethal dose (LD₅₀) was determined via the Reed and Muench method [58]. Surviving mice were challenged orally with wild-type Salmonella χ3761 two months after the first inoculation.