Genetic requirements for Lac+ recombinant formation
Data presented in Table show that formation of Lac+ recombinants depends upon the same recombination functions as the Lac- recombinants: recA, recF, recO, recR, recQ, ruvAB, and ruvC. Deletion of recG increased the production of both Lac+ and Lac- recombinants. Lac+ recombinants accounted for 10–60% of the total chloramphenicol-resistant offspring of the crosses. These results strongly suggest that formation of the Lac+ recombinants takes place via homologous recombination, though they do not rule out the possibility that non-homologous end joining, or other varieties of "illegitimate" recombination, might be involved as well.
Genetic requirements of Lac+ recombinant formation
The reason for considering mechanisms other than homologous recombination in the formation of the Lac+ recombinants is because of the expectation that some of the phage-borne lac
sequences would be attached to phage sequences. Heitman et al. [5
] showed that double strand breaks generated by a restriction endonuclease in vivo
are rapidly repaired by DNA ligase. The results of previous physical studies with other substituted λ phages cut in vivo
by PaeR7 lead to the expectation that the λ lac
chromosome would be uncut, or only singly cut, much of the time in the infected cell [2
Preliminary characterization of the Lac+ recombinants formed in cells which were wild-type for all the recombination functions revealed that most, possibly all, were unstable. When streaked on plates containing chloramphenicol and X-Gal, they segregated Lac- (colorless) progeny at variable frequencies. Southern gel analysis of chromosomal DNA from unstable Lac+ recombinants revealed that some of them were multiploid for lacZ (not shown). When the lac::cat dsDNA was delivered into the cell by a λ vector bearing a tetracycline resistance determinant (Δ nin::tet859 – described below) neither the Lac+ nor Lac- recombinants acquired tetracycline resistance. These findings suggest that the process which formed the recombinants did not involve the entire phage chromosome.
A baseline level of Lac+ recombinants was to be expected in these crosses. The recombination event pictured in Figure , occurring in a cell with a pre-existing duplication of the lac
locus, will produce a Lac+, chloramphenicol-resistant recombinant. Data presented below, in which simpler lac
recombining substrates generate Lac+ recombinants at a frequency approximately 1000-fold lower than Lac-, suggest the frequency of spontaneous lac
duplications in Red+ but otherwise wild type E. coli
is approximately 10-3
, consistent with estimates of spontaneous duplication frequency in Salmonella
made by Roth et al. [7
]. The Lac+ recombinants formed in the crosses summarized in Table occurred at a much higher frequency – approaching 5% of the infected cells – suggesting that pre-existing chromosomal duplications were not involved in the generation of the majority of the Lac+ recombinants. Apparently, normal Red-mediated homologous recombination between the phage and bacterial chromosomes duplicated or amplified sequences in the bacterial chromosome.
The crosses described above were done by infecting log phase host cells grown in rich medium. Such cells contain multiple copies of their chromosome. Complications might arise if the lac::cat segment were to recombine with more than one chromosome at a time. Further complicating interpretation of the experiment, the cells were infected at a multiplicity of 10 phages per cell.
To reduce the complexity of the system, we switched to a cross procedure involving low multiplicity infection (0.1 phage per cell) of stationary phase cells. Presumably, under these conditions, most of the events producing chloramphenicol-resistant recombinants take place between single copies of the lac::cat segment and single copies of the bacterial chromosome. As shown in Table , the low copy infections produced substantial numbers of Lac+ recombinants, though fewer than the high copy infections (4–7% of the chloramphenicol-resistant progeny versus 28–40%). The Lac+ colonies produced in this way were still unstable, segregating Lac- progeny (colorless colonies) when restreaked, or suspended and plated, on medium containing chloramphenicol and X-Gal (not shown).
Recombinant formation by λ lac::cat variants
Recombinant formation by single-cut variants of λ lac::cat
To examine the dependence of Lac- and Lac+ recombinant formation on the structure of the DNA substrate, we constructed variants of λ lac::cat, which are diagrammed in Figure . λ lac::cat819, the progenitor of this series, has two PaeR7 sites. Upon infecting a host cell bearing the Δ recBCD::Ptac-gam-bet-exo-pae-cI substitution, it injects its chromosome, which circularizes. Expression of its lytic genes, including those necessary for phage DNA replication, is blocked by the action of cI repressor. Cutting of its chromosome by the PaeR7 restriction endonuclease in the infected cell releases a 3.5 kbp linear dsDNA, consisting of the cat gene and 1.3 kbp flanks of lac sequences on the right and left. (In this and subsequent descriptions, "right" refers to the upstream, or 5' end of the lac operon). The ends of the flanking lac sequences match precisely their counterparts in the chromosome.
λ lac::cat930 and λ lac::cat931 have only single PaeR7 sites, on the right and left, respectively. Cutting by PaeR7 in the infected cell should produce large linear dsDNAs related to the lac::cat819 fragment, but with long tails of non-homologous DNA extending from the left and right sides, respectively. Interestingly, both of these single-cut phages produce recombinants more efficiently than the double-cut λ lac::cat819, in both recG+ and Δ recG backgrounds (Table ). The right-cut λ lac::cat930 produces relatively more Lac+ recombinants, whereas the left-cut λ lac::cat931 produces fewer. The frequency of Lac+ λ lac::cat931 recombinants, 0.2%, is roughly consistent with the expected frequency of pre-existing lac duplications in the population of infected cells, suggesting that λ lac::cat931 does not generate duplications, but λ lac::cat930 does. We cannot rule out the possibility that some of the 0.2% Lac+ recombinants made by λ lac::cat931 are cointegrates. Cointegrate formation is discussed further in the next section.
Recombinant formation by circular λ lac::cat
λ lac::cat929 has no PaeR7 site. Its chromosome should stay circular in the infected Δ recBCD::Ptac-gam-bet-exo-pae-cI cell. Surprisingly, it produces chloramphenicol-resistant recombinants only 10-fold less efficiently than λ lac::cat819. Unlike the PaeR7-cuttable λ lac::cat variants, λ lac::cat929 also efficiently produces weakly chloramphenicol-resistant microcolonies (not counted in the data presented in Table ). The provenance of the microcolonies is readily understandable. The uncuttable λ lac::cat929 chromosome cannot replicate; it also is not destroyed, and should produce chloramphenicol acetyl transferase in the cell in which it resides. The infected cell, and perhaps its descendants for one or two generations, might be able to divide in the presence of low-concentration chloramphenicol.
Approximately half of the strongly chloramphenicol-resistant recombinants generated by the uncuttable phage λ lac
are Lac- (Table ). This observation suggests that they are "legitimate" recombinants, in which the single chromosomal copy of lacZ
is replaced by lacZ
. These recombinants were unexpected. All Red-mediated recombination events are thought to involve at least one linear partner [6
]. How, then, can circular λ lac
recombine with the circular E. coli
chromosome? Three possible explanations were considered. (1) Some of the λ lac
chromosomes might be cut by the EcoK restriction endonuclease. The infected cells contained a functional EcoK restriction-modification system. λ lac
was grown in a similarly EcoK+ host, and therefore was presumably EcoK-modified. However, if the modification were not complete, there might be residual EcoK restriction activity, resulting in 10% of the phage chromosomes being cut. (The phage chromosome contains five EcoK sites. If each site were 98% methylated, then approximately one in ten phage chromosomes would have a single unmethylated site). (2) The production of the recombinants might not be Red-mediated. (3) The linear partner might be a broken bacterial chromosome.
To test the first explanation, we replaced the gene encoding the EcoK restriction endonuclease, hsdR, with a tetracycline resistance determinant. This replacement had little effect on the formation of recombinants by λ lac::cat929 (Table ).
To explore further the origins of the recombinants formed between circular phage chromosomes and the bacterial chromosome, we constructed a series of bacterial strains bearing variants of the Δ recBCD
substitution lacking pae
. In these hosts, no cutting takes place at PaeR7 sites, and so all of the lac
-substituted phages should remain circular. The results of crosses in these strains are shown in Table . As in the Pae+ crosses, elimination of hsdR
in the Pae- background did not reduce the yield of recombinants. In the Pae- background, elimination of recA
reduced the yield of recombinants twenty-fold. The great majority of the residual recombinants were Lac+, suggesting that their formation was by a process not necessarily involving homologous recombination. Elimination of red
reduced the yield of recombinants three-fold, showing that most, but not all of the recombination events were Red-mediated. As in the Red+ cross, approximately half of the recombinants were Lac-, indicating that formation of both Lac- and Lac+ recombinants was Red-mediated in approximately the same proportions. Thus, the second hypothesis mentioned above, that the recombinants are formed by a Red-independent process, is not supported. The third hypothesis, that the linear partner might be a broken bacterial chromosome, is thus favored. It additionally seems reasonable, given that spontaneous double-strand breaks are frequent in E. coli
We sought to detemine what proportion of the circular phage-by-chromosome recombinants were cointegrates, by testing for co-inheritance of cat and a tetracycline resistance-conferring element present at a remote location in the phage chromosome. Among the Lac+ recombinants produced in the Pae+ host TP507 (Table ), only 2 out of 39 tested were found to be tetracycline-resistant; none of the 40 Lac- recombinants we tested was tetracycline-resistant. This result was consistent with our earlier finding that the great majority of recombinants formed in the high mulitiplicity infections of log phase cells acquired only the homology-flanked cat segment of the infecting λ chromosome. In contrast, in the Pae- host, 27 of 55 Lac- recombinants, and 23 of 25 Lac+ recombinants, were additionally tetracycline-resistant. These observations indicate that the most frequent Red-generated product of recombination between the uncut phage chromosome and the bacterial chromosome is a cointegrate.
Cointegrates could in principle be formed by four different single reciprocal recombination events between the circular λ lac::cat and bacterial chromosomes. These events could involve the right side lac sequences or the left side lac sequences, as diagrammed in Figure . A third event, not shown, involves recombination between the cI genes borne by both the phage and bacterial chromosomes (the latter at the substituted recBCD locus). A fourth event, also not shown, involves recombination between λ genes and homologues in cryptic prophages in the E. coli chromosome (discussed below). As suggested in the figure, cointegrates formed by recombination in the right-side lac flank are phenotypically Lac-, while recombination in the left-side lac flank (or other loci) forms Lac+ recombinants. To test this idea, we constructed variants of λ lac::cat missing either or both lac flanks; they are diagrammed in Figure . As shown in Table , in the non-cutting Pae- host, only the right flank-containing phage chromosome forms Lac- recombinants. The left-flank and no-flank phages form only Lac+ recombinants.
Figure 2 Cointegrates formed by recombination between circular λ cat and the bacterial chromosome. A. The cointegrate formed by recombination between the chromosome and λ cat989, which bears only the left-side lac flank, leaves the chromosomal (more ...)
The ability of the λ Red system to promote recombination events involving short sequence homologies makes it particularly useful for genetic engineering. Yu et al. [9
] reported that a linear cat
cassette with 1000 bp homologous flanks was only 10-fold more efficient than one with 40 bp flanks. Their experiments were done with cells which had been subjected to heat shock and electroporation. To test the efficiency of short-flank recombination under less extreme conditions, we constructed a short-flank λ lac
The fragment released by PaeR7 from the chromosome of λ lac::cat921 consists of the cat gene flanked by 40 bp sequences corresponding exactly to the terminal 40 bp at each end of the 1.3 kbp flanks of λ lac::cat819. As shown in Table , λ lac::cat921 exhibits a several hundred-fold lower efficiency of recombinant formation than its long-flank counterpart, in both recG+ and Δ recG backgrounds. λ lac::cat921 was similarly inefficient in log phase cells (data not shown). These observations suggest that short-flank recombination may not be a significant activity of the Red system in nature. However, they do not speak to the question of whether long flanks work better because they provide a larger homology target for synapsis, or because they provide non sequence-specific protection, perhaps delaying exonucleolytic degradation of the recombining sequences long enough to permit recombination to take place.
Recombination with inverted partners
The observation that Lac+ recombinants are formed efficiently by the right side-cut λ lac::cat930, but not by the left side-cut λ lac::cat931 (Table , discussed above) raised questions concerning the sequence determinants of this directionality. To explore these questions, we constructed locally inverted variants of the bacterial chromosome and the phages, as diagrammed in Figure .
Figure 3 Construction of inversions. A. A plasmid containing the lac operon and some flanking sequences was constructed. The SphI site was converted to a BsrG1 site, the large BsrG1 lac segment was inverted, and the inverted operon was crossed into the bacterial (more ...)
The strategy for inverting the chromosomal lac operon consisted of five steps (detailed in the Methods section and Table ): (1) deletion of the entire operon from the chromosome, replacing it with the cat gene; (2) cloning the cat gene from the deletion mutant, along with flanking chromosomal sequences; (3) using the cloned, plasmid-borne flanking sequences, and Red-mediated gap repair, to clone the lac operon in a plasmid; (4) inverting the lac operon, relative to its flanking sequences, in the plasmid; (5) replacing the Δ lac::cat chromosomal allele with the plasmid-borne lac-inv allele.
Bacterial strains used in this study
The single-cut lac::cat930 and lac::cat931 alleles were constructed by replacing the left-side and right-side PaeR7 sites, respectively, of pTP819, with XbaI sites. To produce inverted derivatives of these two alleles, the plasmids pTP930 and pTP931 were digested with XbaI and XhoI (a PaeR7 isoschizomer); lac::cat inserts and bla-ori backbone fragments from the two plasmids were exchanged. The inverted alleles were then crossed into phage λ, as described in the Methods section.
In constructing the chromosomal inversion, we switched genetic backgrounds, from the AB1157-derived strains with which most recombination studies have been done, to MG1655, the sequenced wild type E. coli
]. In the MG1655 background, the same directionality of Lac+ recombinant formation was observed. The efficiencies of recombination of the MG1655 derivatives with both λ lac
and λ lac
, and the efficiency of Lac+ recombinant formation by λ lac
, were slightly elevated relative to those in the corresponding AB1157-derived strain.
The eight combinations of normal and inverted phages and bacteria (left- and right-cut phages, and their inverted counterparts, in normal and lac-inverted E. coli) were tested for Lac+ recombinant formation. The results are shown in Figure . Only two of the eight combinations produced large numbers of Lac+ recombinants: lac::cat930 (right-cut, normal orientation) in lac-wild type, and lac::cat1033 (left-cut, inverted orientation) in lac-inv.
Figure 4 Normal and inverted phage-by-chromosome crosses. Phages bearing the indicated lac::cat alleles were crossed with wild type (TP829) and lac-inv (TP894) bacteria. Sequence segments are designated as shown in Figure 3. Bacterial sequences are colored blue, (more ...)
Recombination with electroporated linear DNA species
The linear DNA produced by cutting of the λ lac
phages at a single site is large and complex relative to the lac
segment itself. To reduce the complexity, we generated linear DNA species which correspond to shortened versions of the single-cut phage chromosomes. The DNA species were generated by transferring, into a conditionally-replicating vector, parts of the plasmids previously used to introduce the lac
substitutions into λ. The vector can replicate only in a host which supplies the plasmid R6K Pir protein [11
]. Details of the plasmid constructions are given in the Methods section. Plasmids bearing the cloned lac
and flanking sequences from λ were digested with restriction enzymes, and the DNA fragments were introduced into bacteria by electroporation. The results of some of these crosses are shown in Figure . The electroporated DNAs faithfully mimicked their single-cut phage counterparts: only the two crosses corresponding to the high Lac+ producer crosses of Figure generated high proportions of Lac+ recombinants.
Figure 5 Normal and inverted linear DNA-by-chromosome crosses. Plasmids bearing the indicated lac::cat alleles (corresponding to the phages in Figure 4; the numbers of the plasmids bearing these alleles are given in the Methods section) were digested with BamHI (more ...)
The observation that only DNAs with sequences designated "T-9" on the left side generated many Lac+ recombinants (Figure ) led us to try a linear species in which T was detached by digestion with a restriction enzyme (KpnI). Removal of the T segment greatly reduced the ability of the linear DNA to generate Lac+ recombinants (Table ). A BLAST search of the E. coli
genome for sequences related to T revealed four loci in MG1655 with close matches to T's leftmost 200 bp, which constitute the C-terminal third of the λ tfa
gene. The four bacterial sequence segments are located in cryptic prophages: ybcX
and an unnamed gene fragment in prophage DLP12, tfaR
in prophage Rac, and tfaQ
in prophage Qin. The locations and orientations of the first three of these are such as to permit a linear DNA species to recombine both with them and with the lac
locus. Such an event is diagrammed in Figure , in which the three recombining bacterial loci are designated I, II, and III; closely related events have been extensively documented in other studies [7
]. In the pictured cross, recombination between the linear DNA and the bacterial chromosome at lac
and I generates a recombinant bearing intact lac
, a duplication of all sequences between lac
and I, a smaller duplication of the left lac
flank, and an insertion of λ sequences unrelated to the cryptic prophages (see Figure ).
Recombinant formation by variant linear DNA species
Figure 6 Model for Lac+ recombinant formation in linear DNA-by-chromosome crosses. (a) Recombination between the B sequence segments generates two broken chromosome arms. In the diagram, which is not to scale, I, II, and III designate the cryptic prophage tfa (more ...)
The hypothesis that the high-frequency Lac+ chloramphenicol-resistant recombinants are generated by homologous recombination with cryptic prophage sequences in the chromosome predicts that such recombinants would not be generated in an E. coli
strain in which the cryptic prophages are deleted. To test this prediction, we electroporated linear lac
DNA into TP750, a Red+ derivative of MDS12, the reduced-genome E. coli
strain constructed by Kolisnychenko et al. [12
]. In this cross, Lac+ recombinants constituted only 0.06% (average of 6 measurements) of the total chloramphenicol-resistant progeny, a frequency no higher than expected for spontaneous pre-existing duplications in the bacterial chromosome.
The hypothetical Lac+ chloramphenicol-resistant recombinant pictured in Figure has some predicted properties which were confirmed experimentally. First, it is predicted to segregate Lac- recombinants at low frequency, and chloramphenicol-sensitive recombinants at high frequency, the results of recombination between the short and long duplicated segments, respectively. Overnight cultures of three Lac+ recombinants grown in the absence of selection were found to include Lac- chloramphenicol-resistant clones at an average frequency of 0.03%, and Lac+ chloramphenicol-sensitives at 30%. Second, it should be possible to demonstrate specific duplication junctions in each of the three types of recombinants by the use of PCR (see Figure ). Primers were designed to this end, and employed in colony PCR with a collection of 12 Lac+ recombinants. Each of the 12 was found to have one of the three predicted junctions: seven were type I, two were type II, and three were type III. Examples are shown in Figure .
Figure 8 PCR products indicating specific duplication junctions. Colonies of Lac+ recombinants were tested by PCR as described in the Methods section. Lanes labeled S are standards (1 kb ladder, Invitrogen). Lanes 1, 2, and 3 are type I, II, and III recombinants, (more ...)
We constructed several other linear DNA substrates to test the generality of the model in Figure . The abilities of these substrates to generate Lac+ recombinants are shown in Table . The first of these was a derivative of lac
lacking its χ site. The χ site is located in the right lac
flank, which is labeled "B" in Figures 3–6, close to the cat
gene. The orientation of this χ site is such that it would be expected to interact productively with RecBCD enzyme approaching from the PaeR7- or Xho-generated right end of lac
. While no activity of χ in these crosses was expected, as the bacteria lack RecBCD, the directionality of Lac+ recombinant formation was reminiscent of the directionality of χ-RecBCD interaction (see [13
] for a review). The Δχ substrate was just as active as the χ + version, showing that χ does not contribute significantly to Lac+ recombinant formation. Similarly, a lac
derivative (pTP1052) lacking its sequences derived from phage P22 was equally proficient at generating Lac+ recombinants. A derivative missing the left lac
flank made Lac+ recombinants almost exclusively; the small number of Lac- recombinants in this case probably represent cointegrates made by uncut plasmid DNA in the fragment preparation.
Plasmids pTP1055 and 1056 were constructed to test whether duplications of the type pictured in Figure could be generated by Red-mediated recombination involving arbitrarily chosen sequences on either side of lac in the chromosome. As indicated in Table , linear DNAs from these plasmids also generated Lac+ recombinants almost exclusively. The structures of the plasmids, chromosome, and expected recombinants are diagrammed in Figure . That the expected duplications were in fact generated was demonstrated by the production of duplication-spanning products in PCR using the primers Ddi and Edi, described in the Methods section and represented schematically in Figure . Examples are shown in Figure .
Figure 7 Red-generated duplications to the left and right of lac. Recombination between electroporated linear dsDNA species ACF (left) and GCB (right) generate duplications of chromosomal sequences to the left and right of lac, respectively, through a sequence (more ...)
Structures of the λ lac::cat Lac+ recombinants
The major class of Lac+ chloramphenicol-resistant recombinants formed in crosses involving infection by λ lac::cat819 (cut left and right) and λ lac::cat930 (cut right) behave like their counterparts generated by electroporation of linear dsDNAs into Red+ cells: they segregate Lac- chloramphenicol-resistant clones at low frequency, and Lac+ chloramphenicol-sensitives at high frequency (data not shown). In addition, their formation also depends upon the presence of cryptic prophages in the chromosome: λ lac::cat819 was found to produce Lac+ chloramphenicol-resistant recombinants as only 0.13% (average of six measurements) of the total chloramphenicol-resistant progeny in crosses with TP750. These properties suggested both kinds of recombinant might have the same structures as well, but PCR tests of 50 of the phage-generated recombinants with the primers used to demonstrate type I, II, and III recombinants (Figure ) were negative (data not shown).
Further computer analysis revealed a cryptic prophage sequence which could recombine with the phages, but not with the shorter, plasmid-derived linear DNAs, producing recombinants by the mechanism drawn in Figure : a 3.5 kbp patch of DLP12 closely matching sequences in the vicinity of the λ cos site. This site is located immediately to the left of the tfaD locus in the chromosome. Recombinants formed by crossing over at this site and at the right lac flank would be expected to contain duplications of bacterial sequences identical to those of type I recombinants (pictured in Figure ), but to contain a larger part of the phage chromosome as well. PCR tests of six λ lac::cat819 and six λ lac::cat930 recombinants with primers designed to demonstrate cos recombinant junctions showed that all twelve had them. An example is shown in Figure .
The predominance of cos
recombinants over tfa
recombinants among the phage-generated Lac+ recombinants is to be expected. First, λ Exo, traveling in from the end of the long left-side non-homologous tail of λ lac
therefore presumably has a kinetic advantage. Second, the bacterial cos
homology patch is significantly larger than the tfa
homologies. A third possible advantage of cos
is that, at least in some parts of the phage lytic cycle, it is the site of a double-strand break for the DNA encapsulation step of phage assembly. We expected that cos
would be unbroken almost all the time in our crosses. The λ chromosome is linear at the time of injection, but is rapidly circularized by annealing and ligation [14
]. The other parts of the phage lytic cycle are inhibited by the presence of cI repressor in the infected cells. Even so, the hypothesis that Red sometimes manages to gain access to cos
ends in these crosses remains plausible; particularly so because Red is present in the cell at all times, whereas, in a normal infection, Red is not present until its genes are expressed from the phage chromosome. An unexpectedly significant presence of cos
ends in the infected cells could also help explain the otherwise surprisingly high frequency of chloramphenicol-resistant recombinants seen in crosses involving phage chromosomes not cut by restriction endonucleases, described above.
One aspect of Lac+ recombinant formation by λ lac::cat819 is not explained by the model of Figure . The λ lac::cat819 contains two PaeR7 sites. Cutting by PaeR7 should release a linear dsDNA with lac flanks and no attached non-lac DNA. This DNA species did not generate Lac+ recombinants (above the background of pre-existing duplications) when electroporated directly into cells (Table ). How does it apparently do so in the infected cells? As discussed above, it is expected that the λ lac::cat819 chromosome would be uncut, or only singly cut, much of the time in the infected cell. If it recombined at a time at which it was cut only on the right, the type of recombinant pictured in Figure could be formed. In this recombinant, an uncut and unmodified PaeR7 site would sit in the chromosome. Presumably, the presence of this site would make the recombinant unstable, initially; but the PaeR7 site might eventually become modified, that is, escape restriction. The pae-expressing strains employed in these experiments restrict plaque formation by single PaeR7 site-bearing λ phages only approximately 5-fold (unpublished data).