The results of experiments examining Red-mediated recombination between linear and circular partners suggest two main interpretations concerning the roles of RecG and RuvC. (i) the main pathway to recombinant formation involves RuvC and (ii) RecG impedes this main recombination pathway. However, RecG may also stimulate recombination via a RuvC-independent pathway.
In the presence of RecG function, the degree of dependence of Red-mediated recombination on RuvC varied substantially among the three types of crosses described in the preceding section. The dependence was greatest in crosses between nonreplicating phage chromosomes in non-plasmid-bearing cells (Fig. ). In the other cases, recombination appeared to be nearly independent of RuvC.
The contribution of RecG to Red-mediated recombination was less variable among experiments: in all cases, elimination of RecG function increased the frequency of recombination. In both types of crosses that were carried out in non-plasmid-bearing cells—between nonreplicating λ chromosomes and between short linear DNA fragments and the bacterial chromosome—the resulting higher-frequency recombination was also more dependent on RuvC (except when recombinant formation could in theory proceed via branch migration, heteroduplex formation, and mismatch repair, as pictured in Fig. ). The simplest interpretation of these observations is that RecG opposes the main pathway of Red-mediated recombination, which is RuvC dependent. The data in Table suggest that RecG may also promote an alternative, RuvC-independent recombination pathway: recombination frequencies in the ruvC mutant were slightly higher than in the ruvC recG double mutant.
An alternative interpretation of the effects on Red-mediated recombination of the recG258
allele is that they do not result from inactivation of recG
but rather from a function expressed by the mini-Tn10
used to generate the allele (22
). This interpretation is ruled out by the observation that a simple deletion allele of recG
exhibits the same stimulation of Red-mediated recombination as recG258
Our description of the roles of RecG and RuvC is similar to a model proposed by Whitby and coworkers (39
), in which RecG actually aborted genetic exchanges resulting from RecA-mediated strand invasion but then allowed RecBCD to catalyze exchanges at the ends of the incoming DNA by an unspecified mechanism. Although these researchers found biochemical evidence for RecG’s ability to act in this way, the fact that a recG
mutant is recombination deficient relative to wild type apparently led them to favor other interpretations of RecG’s activity in recombination (1
The generation of late-arising revertants of lacZ
mutants under conditions of selection is another cellular activity that apparently is impeded by RecG (10
). This activity, sometimes called “adaptive mutability,” is dependent upon recombination functions (6
). Its dependence upon recombination functions may be a consequence of the involvement of gene amplification in the process (3
). Harris et al. (13
) proposed a model for recombination associated with adaptive mutation, in which RecG aborts 3′-end invasion and promotes 5′-end invasion.
The initial steps of Red-mediated recombination are perhaps less complex than those of RecBCD-mediated recombination or of RecBCD-dependent adaptive mutation. The λ exonuclease specifically and processively degrades the 5′-ended strand of double-stranded DNA (20
), leaving exclusively 3′-ended single strands for synapsis and strand invasion. Production of these 3′-ended single-strand tails in λ-infected cells has been observed directly (14
). The observation that RecG inhibits Red-mediated recombination that is constrained to proceed via strand invasion favors models in which RecG tends to push out invading 3′ ends.
The model diagrammed in Fig. and accounts reasonably well for recombination between linear and circular partners in a red-expressing cell lacking recBCD and recG functions: double-stranded ends are channeled nearly exclusively through a pathway that involves the creation of 3′-ended single-strand tails, which invade an unbroken homologous duplex. If and only if branch migration is impeded by a significant nonhomology, recombinant formation is dependent upon nucleolytic resolution of the resulting Holliday junction by RuvC. In the absence of RuvC, presumably, a recombination intermediate something like the structure diagrammed in Fig. b accumulates. We have not observed such an intermediate, but the methods employed in the extraction and restriction enzyme digestion of DNA from λ-infected cells might be expected to favor its dissociation by spontaneous branch migration.
The model diagrammed in Fig. and does not account so well, by itself, for what happens in a recG+
cell. A more complete model would account for three questions raised by the data in Fig. and Tables and . (i) Why was recombination between nonreplicating λ chromosomes, with no substantial nonhomologies, dependent upon RuvC in one set of crosses (Table , non-cat
-bearing phage cut) and independent in another (Fig. )? One possible explanation is that the interacting DNA sequences in question were not identical in the two experiments and branch migration could proceed all the way past the restriction site polymorphism in the face of opposition by RecG in one case but not in the other. (ii) The crosses diagrammed in Fig. (left side) and Fig. both involve recombination of a linear cat
-bearing chromosome with a circular chromosome. Why was recombinant formation dependent upon RuvC in the former case and independent in the latter? A key difference between the two crosses is that replication of both partners is blocked in the λ cross (Fig. ) whereas replication of only the linear partner is blocked in the linear-by-host chromosome cross (Fig. ). The passage of a replication fork may potentiate alternative pathways for the resolution of recombination intermediates. (iii) How does RecG promote, rather than impede, RecBCD-mediated recombination? One possibility is that RecBCD-mediated recombination proceeds primarily via 5′-ended strand invasion, but extensive studies of the activities of RecBCD, both in vitro (2
) and in vivo (11
), favor the view that RecBCD generates invading 3′-ended strands. Together, these observations are consistent with the idea that RecBCD may function in the processing or resolution of recombination intermediates that it participates in generating (5
). Perhaps the normal role of RecG is to facilitate this function of RecBCD.