MOST of homologous recombination in wild-type Escherichia coli is initiated by a heterotrimeric RecBCD enzyme, which binds to blunt-ended DNA duplex, unwinds it, and concomitantly degrades both unwound strands. In vitro, RecBCD enzyme activities are regulated by its interaction with a χ-sequence, upon which the modified enzyme loses most of its nuclease activities (preserving only a weak 5′–3′ exonuclease activity), but gains the ability to facilitate RecA protein loading onto a thus created 3′-ending tail (reviewed in Kuzminov 1999). In this way a nucleoprotein filament, the central recombination intermediate, is created.

When the enzyme lacks its RecD subunit, as in recD mutants, the resulting RecBC enzyme is devoid of any detectable nuclease activity (ExoV⁻) and also of the ability to interact with χ sites, whereas its helicase and RecA loading activities are preserved (the latter being constitutive, focusing recombination exchanges to DNA ends) (reviewed in Kuzminov 1999). Although ExoV⁻, recD mutants are recombination and DNA repair proficient (Chaudhury and Smith 1984; Lovett et al. 1988). This proficiency was shown to rely (to a varying extent) on a single-strand-specific 5′–3′ exonuclease activity of RecJ nuclease (Lovett and Kolodner 1989) in most recombination assays (e.g., UV survival, Hfr conjugation, and P1 transduction) (Lloyd et al. 1988; Lovett et al. 1988). There is, however, a marked exception in RecJ requirement in recD mutants. Recombination of freely replicating λ red gam phages, which normally depends on host RecBC(D) functions (since λ's own recombination system is disabled), is unaffected by RecJ nuclease inactivation in recD mutants (Thaler et al. 1989). This finding [together with the one of the independence of λ-recombination on χ (Thaler et al. 1989)] had a great influence on the understanding of regulation of RecBCD enzyme functions in a cell as it led to a long-standing model according to which a χ-modified RecBCD is equivalent to the RecBC enzyme (suggesting that upon interaction with a χ site a RecD subunit is lost or inactivated) (Thaler et al. 1988, 1989). Yet, later biochemical studies showed that the 5′–3′ exonuclease activity of the χ-modified enzyme is absent in RecBC enzyme (Anderson et al. 1997). The paradoxical RecJ independence of λ red gam crosses in recD mutants has puzzled researchers for years, leading to several hypotheses. For instance, Amundsen and Smith (2003) proposed that recombination proficiency of a recJ recD mutant is due to single-stranded (ss)DNA substrates produced during λ-replication, i.e., that 5′–3′ exonuclease activity is not required for λ-recombination.

However, a recent report (Ðermić 2006) has shown that a phenotype of the recJ recD mutant does not reflect a need for a 5′–3′ exonuclease activity in a particular reaction; it rather represents a measure of the efficiency with which the lack of RecJ function is replaced by a function of exonuclease VII (ExoVII), another exonuclease with a processive 5′–3′ (and also 3′–5′) exonuclease activity on ssDNA (ssExo) (Chase and Richardson 1974). As that report also showed that the 5′–3′ exonuclease activity is essential for all events tested in recD mutants (UV and γ-survival, recombination in Hfr and P1 crosses, cell viability), it is conceivable that unaffected λ-recombination in the recJ recD mutant is due to activity of ExoVII. To test this hypothesis, we performed λ red gam χ⁺-lytic crosses in recD mutants devoid of activities of various exonucleases (Table 1 ).

As shown in Table 2 , RecJ inactivation left λ-recombination in the recD mutant background mostly unaffected, confirming earlier reports (Thaler et al. 1989; Jockovich and Myers 2001). This proficiency was indeed dependent on ExoVII since the triple RecD⁻ RecJ⁻ ExoVII⁻ mutant was recombination deficient, even more so than the recB null mutant. As the RecJ⁻ ExoVII⁻ derivative of the wild-type strain exhibited unaffected recombination proficiency, we may conclude that λ-recombination requires a 5′–3′ exonuclease activity either of χ-modified RecBCD in wild-type cells or of RecJ and ExoVII nucleases in recD mutants. This process therefore does not differ from the other DNA end-dependent recombination processes in E. coli and other organisms, which rely on resection of a 5′-ending tail (discussed in Ðermić 2006). The indispensability of the 5′–3′ exonuclease activity argues against models that posit that a helicase activity is sufficient for initiation of recombination [i.e., “split-end” models (Rosenberg and Hastings 1991), proposing that either of the unwound strands (Razavy et al. 1996) or both of them (reviewed in Myers and Stahl 1994; Smith 2001) invade the homologous DNA duplex; also a concept of RecBCD enzyme unwinding and rewinding DNA duplex, with just a nick at a χ site being required for recombination initiation (reviewed in Smith 2001)].

In contrast, inactivation of exonucleases of the opposite, 3′–5′ polarity [exonuclease I (ExoI), ExoVII, exonuclease X (ExoX), and SbcCD] did not reduce λ-recombination markedly, suggesting that it is largely independent of that activity. This is in accord with earlier studies that found that 5′–3′ exonucleases are more important than 3′–5′ ones in DNA repair reactions in both the wild-type (Viswanathan and Lovett 1998) and the recD backgrounds (Ðermić 2006). The strongest requirement for a 3′–5′ exonuclease was for ExoI (the strongest ssExo acting on a 3′-ending strand; Lehman and Nussbaum 1964) in the recJ recD mutant. Since the main role of exonucleases in recD mutants is apparently in preventing reannealing of the strands unwound by the RecBC enzyme (Ðermić 2006), it seems, therefore, that ExoVII resects a 5′-ending tail less efficiently than RecJ does, thereby increasing the requirement for degradation of the complementary strand. Moreover, inactivation of ExoI in the recJ recD mutant might further reduce the efficiency of ExoVII-catalyzed resection of the 5′-ending strand as it might enable the 3′-ending strand to compete for the ExoVII binding. Finally, the observed synergism between ExoI and RecJ ssExos (this study; Ðermić 2006) should not come as a surprise, considering that these two proteins interact (form a complex) in a cell (Butland et al. 2005), which may provide an additional function for both of them.

Overall exonuclease requirements for λ-phage recombination in recD mutants resemble those for P1 transduction in that background (Ðermić 2006), which is likely due to the similarity of DNA substrates delivered by the two phages [i.e., linear DNA duplex lacking substantial heterologous regions (discussed in Ðermić 2006)].

Total yield of λ-phage (MMS754) was in correlation with the recombination efficiency in most crosses (Table 2), which was expected since recombination is thought to be required for an efficient replication and packaging of λ-progeny (reviewed in Kuzminov 1999). An exception was a low total yield of crosses done in recombination-proficient mutants defective in multiple exonucleases of a 3′–5′ polarity (RecD⁻ ExoI⁻ ExoVII⁻ ExoX⁻ and RecD⁻ ExoI⁻ ExoVII⁻ SbcD⁻), which suggested that these exonucleases have a role in the λ-phage life cycle other than the recombinational one. They may participate in replication, as shown for some exonucleases in yeast (Tishkoff et al. 1997). A very low total yield in crosses involving the RecD⁻ RecJ⁻ ExoVII⁻ mutant may be due to its recombination deficiency (Table 2). But, as this mutant has an extremely low viability (<1%, Ðermić 2006), a low yield may also result from titration of infecting phage on cells that are unable to support its growth. Thus, a simple method was used to determine a fraction of RecD⁻ RecJ⁻ ExoVII⁻ mutant cells that enable phage-λ growth (Ðermić and Trgovčević 1999). Bacteria were lysogenized with a thermoinducible λ cI857 phage and its recombination-deficient red3 derivative (Arber et al. 1983). Cells were grown as for λ-crosses (except for a lower temperature: 31.5°) and, upon reaching an OD₆₀₀ of 0.25, they were serially diluted and mixed with indicator bacteria in soft tryptone agar, poured on tryptone plates, and incubated at 42° (to determine their plaque-forming ability). Compared to the RecD⁻ mutant, RecD⁻ RecJ⁻ ExoVII⁻ mutant cells produced just 31% less infective centers of λ cI857 phage, suggesting that although mostly incapable of giving rise to a colony, a majority of them were metabolically active and thus supported phage growth. However, this growth required recombination activity since only 15% of RecD⁻ RecJ⁻ ExoVII⁻ mutant cells developed λ cI857 red3 infective centers. This showed that recombination deficiency (of both bacteria and phages) is indeed the main reason for a low yield of λ-progeny in that mutant.

• Received May 8, 2006.
• Accepted May 11, 2006.

• Copyright © 2006 by the Genetics Society of America