MATERIALS AND METHODS

Bacteriophages:

The map of rII markers used in this study is presented in Figure 3. The origin of most of the phages was described earlier (Shcherbakov et al. 2002). The T4 dexA1 strain carrying an insertion in the DexA coding region (Gauss et al. 1987) and tsL42 (antimutator allele of T4 DNA polymerase) (Drake et al.1969) were kindly supplied by J. Drake; T4 strain D219A deficient in the 3′–5′ exonuclease of the DNA polymerase (an alanine substitution for aspartate 219) (Frey et al. 1993; Reha-Kranz and Nonay 1993) was a gift from L. Reha-Kranz.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.—

Map of rII mutations used in the present work. The numbers above the markers show their distance (the number of intervening bases) from the edge of the ets1 insertion.

Bacteria:

Escherichia coli strain BB permissive for rII mutants was used as a host for preparing phage stocks, for phage titration, as a host in phage crosses, and for measuring total phage yields. E. coli CR63(λ_h) was used for titration of recombinants with rII⁺ phenotype. The E. coli optA1 strain (Saito and Richardson 1981), which is nonpermissive for T4 3′–5′ exonuclease mutants including tsL42, was used for constructing multiple mutants bearing mutations in rII and in other T4 genes.

Standard phage cross procedure:

An aliquot of an E. coli BB overnight culture was diluted 100-fold in L broth and aerated at 37°. At the cell concentration of 1 × 10⁸/ml, the suspension was cooled to 0°; the cells were pelleted by centrifugation and resuspended in cooled L broth at the density of 4 × 10⁸ cells/ml (standard culture). An equivalent mixture of phage parents in the volume of 0.5 ml was added to 0.5 ml of the cooled BB suspension to provide total multiplicity of infection of 10 particles/cell. The infected cells were incubated for 10 min at 33°, diluted 1000-fold in prewarmed L broth, and incubated at the same temperature for another 80 min. Cell lysis was completed by adding 0.3 ml of chloroform to 5 ml of the diluted culture.

Determination of rII⁺ recombinant frequency and plating efficiency:

Recombinant frequencies were calculated by dividing the titer determined on a λ-lysogenic host by the total lysate titer. The resulting frequencies were corrected for plating efficiency that was determined as follows. An equivalent mixture of segC⁺ and segCΔ phage strains with genotypes of the expected recombinants was used to infect the standard culture of E. coli BB at the total multiplicity of 10 particles/cell. The infected cells were incubated and processed in the same way as those in the standard crosses. The lysates were plated on the E. coli BB and on the λ-lysogenic host CR63(λ) after proper dilutions. The ratio of the titer on BB to that on CR63(λ) was then used as a plating efficiency quotient to correct the titers of recombinants observed on the λ-lysogen. Typically, the measured values of the quotient were within the 15% range of unit.

RESULTS

It is generally accepted that extended 3′-single-stranded tails are obligatory for the initiation of DSB repair by homologous recombination (Paques and Haber 1999). It seems most probable that the very low recombinant frequencies in crosses of ets1 with the nearest rIIB markers N24 and 375 (12 and 33 bp distance, respectively) result from exonuclease degradation of the 3′-single-stranded termini (3′ resection). This is a negative (“counterproductive”) contribution of a 3′–5′ exonuclease activity to the recombinant output. Note, however, that the D-loop intermediate (see steps 4 and 5 in Figure 1) has the protruding nonhomologous single-stranded DNA, which must be removed before DNA synthesis can start. Hence, we may expect also a positive contribution of the 3′–5′ exonucleases to the process.

To reveal a possible role of T4-encoded 3′–5′ exonucleases in the DSB repair process, we analyzed the DSB-promoted genetic recombination in 3′–5′ exonuclease-deficient strains. Series of rII mutant crosses of the type i × ets1 were performed, using the wild-type (except rII) T4 strain, tsL42, dexA1, and D219A mutants, and a double dexA1 D219A mutant. Single-base-substitution mutations in rIIB and rIIA located at different distances, 12–2040 bp to the left of the ets1 insertion, were used as i-markers. All the i-markers used are of low-recombination (LR) type; i.e., the recombinational mismatches i/+ are not repaired and thus do not contribute to the apparent recombinant frequencies (Shcherbakov and Plugina 1991). The results of the crosses at short distances (up to 103 bp) are presented in Table 1. They are shown in Figure 4 in a more vivid form as a ratio of rII⁺ recombinant frequencies observed in the mutant crosses to those observed in the wild-type crosses (R_mut/R_wt). A general observation is the significant increase in the recombination frequency in crosses at the shortest distances, 12 and 33 bp.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.—

Effect of tsL42, D219A, dexA1, and dexA1 D219A mutations on genetic recombination in crosses i × ets1. The ratio of recombinant frequency in the mutant cross (R_mut) to that in the wild-type cross (R_wt) is shown. The data are from Table 1.

In dexA1 crosses, recombinant frequency was greatly increased (about threefold) in crosses with N24 and 375 mutants; but in crosses with distant markers (up to 2040 bp), the recombinant frequency did not differ from that in the wild-type crosses (data not shown).

In the case of D219A mutants (absence of functional 43Exo), the recombinant frequency was significantly increased in all crosses. The observed value of recombinant frequency in the cross X504 × ets1, (29.7 ± 0.8)%, is significantly higher than the level of the intermediate plateau predicted by the SPC model (Figure 1) and experimentally observed in the wild-type crosses (Figure 2). More extended recombination data (Figure 5) demonstrate that the increased recombinant output takes place at all tested distances, up to 2040 bp. We suggest that a general change in the DSB repair pathway occurs in the absence of functional 43Exo (see discussion). In crosses of the double dexA1 D219A mutant, the preservation of 3′ termini was maximal. The recombinant frequency for the N24 × ets1 cross (12 bp) reached 13% (a sevenfold increase in comparison to the value for the wild-type strain). The combined effect of dexA and D219A in crosses at the shortest distances was additive, implying functional independence of the corresponding nucleases. Nevertheless, even in the double-mutant background, the recombinant frequencies at the shortest distances did not reach the intermediate plateau. This demonstrates that some, though small, loss of the 3′-terminal sequences occurs in the double mutants. Crosses at longer distances (Figure 5) demonstrate a frequency/distance relationship that differs significantly from that in the wild-type strain. The distinctive features of this relationship are increased recombination within the range of distances from 12 to 427 bp and somewhat decreased recombination at longer distances. The dexA1 D219A curve differs also from that in D219A crosses: the recombinant frequency was higher in the single mutant D219A in all the crosses except those at the shortest distances. It seems reasonable to think that the reduced ability to remove the nonhomologous 3′ terminus impedes the initiation of replication in the D-loop intermediate and causes an accidental decay of the junction. Such an impediment may be of general significance in the phage DNA metabolism: the dexA1 D219A strain does not multiply as well as the wild-type strain or the single dexA and D219A mutants and produces small plaques on E. coli BB or CR63.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.—

Effect of D219A and dexA1 D219A mutations on the frequency/distance relationship in two-factor crosses i × ets1.

In crosses of the tsL42 mutant (antimutator DNA polymerase) overall frequency/distance relationships did not differ from those in the wild-type crosses (data not shown) except the crosses at 12 and 33 bp, where recombinant frequencies were about twofold higher than in the corresponding wild-type crosses (Table 1 and Figure 4).

DISCUSSION

The essence of DSB repair is recombination of the broken chromosome with an intact homolog. The recombination is brought on by the DNA ends. Examination of this recombination is a promising approach to study mechanisms of DSB repair in vivo. Under the conditions of our crosses [equal and high multiplicity of the breakable (ets1) and unbreakable (i) phage parents and virtually complete cleavage of ets1-DNA by SegC endonuclease], 42 and 50% of rII⁺ recombinants in progeny phages are expected according to the s-SPC and fc-SPC models, respectively (Shcherbakov et al. 2002). Measurement of recombinant frequency in a series of crosses of ets1 with a set of i-markers located at various distances (12–2040 bp) from ets1 reveals a complex four-phase dependence of rII⁺ recombinant frequency on the distance (Figure 2). The expected maximal values (final plateau) are reached only after 1000 bp. This means that the DNA sequences adjacent to the end of the broken chromosome are partially lost during the end processing. It is generally accepted that 5′ resection resulting in the formation of 3′-terminal single-stranded DNA is a necessary step in any recombination pathway. This 5′ resection (i.e., removing one of the two DNA strands) accounts for a 50% loss of the DNA of the breakable parent and, correspondingly, for a twofold decrease in the recombinant formation (intermediate plateau, Figure 2). The extremely low recombinant frequencies observed in crosses of ets1 with the closest markers N24 and 375 (12 and 33 bp, respectively) clearly demonstrate that 3′-terminal sequences are also lost during the end processing. We studied the DSB-promoted genetic recombination in T4 mutants dexA1 and D219A deficient in the phage 3′-5′ exonucleases DexA and 43Exo, respectively. The recombinant frequency was greatly elevated in both single mutants in crosses with N24 and 375 (12 and 33 bp, respectively). The double D219A dexA1 mutant has demonstrated additive interaction of the two mutations (Table 1 and Figure 4). This confirms that the low recombinant frequency observed in crosses of the wild-type strain at these distances (phase 1 in Figure 2) results from the loss of the terminal sequences of the 3′-single-stranded ends of the broken chromosome, with both DexA and 43Exo contributing to this loss.

The observed additivity implies that the two exonucleases act in parallel ways and do not compete with each other. The 3′ termini could be attacked during different stages of the process (structures 2–5 in Figure 1), and it is tempting to think that the two enzymes have different substrate preferences. For 43Exo, the D-loop intermediate with the protruding whisker may be its natural substrate, and the 43Exo attack on the whisker may be the easiest way to initiate DNA replication in the D-loop (Shcherbakov et al. 1995).

The data obtained with the tsL42 mutant (antimutator allele of T4 DNA polymerase) further support this suggestion. At first glance, the significant increase in recombinant frequency in tsL42 crosses (Table 1 and Figure 4), even if limited to the shortest distances, may look unexpected, since the balance between the nucleotide incorporation and the proofreading exonuclease activity is shifted toward exonuclease in antimutator variants of DNA polymerase (for review see Reha-Kranz 1998). Nevertheless, a simple explanation could be offered for this observation. After removal of the unpaired terminus, 43Exo excises several correctly paired nucleotides before switching to polymerization (Shcherbakov et al. 1995), the number of excised paired nucleotides being ∼25 and 15 in 43⁺ and tsL42 crosses, respectively, due to a faster switching from hydrolysis to polymerization in the antimutator DNA polymerase. This implies that the probability for N24⁺ and 375⁺ sites on the 3′ termini of the broken DNA (12 and 33 bp from ets1, respectively) to be removed by 43Exo must be substantially decreased in tsL42 crosses, resulting in an increase in rII⁺ recombinant frequency.

The recombinational effect of dexA was largely limited to the short distances, while the general form of the frequency/distance relationship (data are not shown) retained the essential features of the wild-type relationship. It might mean that the only action that DexA carries out during the DSB repair process is the removal of some stretch of the single-stranded DNA from the 3′ termini. In D219A crosses, a significant increase in the recombinant frequency was observed at all the tested distances, from 12 to 2040 bp (Figure 5). While the short-range effect is adequately explained by the preservation of 3′ tails, the enhancement of recombination in the crosses with distant markers requires a different interpretation. The most meaningful deviation from the SPC model prediction is the absence of the intermediate plateau: beginning from the cross X504 × ets1 (103 bp), the recombinant frequency reliably exceeds the 25% level, the theoretical maximum predicted by SPC models for the i-markers located in zone 2. Zone 2 (see Figure 1) results from the 5′ resection during processing of the ends of the broken chromosome and, in the wild-type crosses, it stretches up to 400 bp (see Figure 2). Formally, the observed >25% recombinant frequency within this range implies a reduction in the 5′ resection. It is hard to perceive why the mutation in 43Exo may cause the reduction in 5′ resection. A plausible explanation is offered by the model of synthesis-dependent strand annealing (SDSA) (Mueller et al. 1996) illustrated in Figure 6. According to this model, invasion of the uncut chromosome by the single-stranded tail initiates leading-strand synthesis accompanied by migration of the bubble and displacement of the newly synthesized single-stranded DNA from the complex. Such synthesis is of low processivity, and the replicative complex decays soon. The released single-stranded DNA is then annealed to the single-stranded tail of the other half of the broken chromosome. DNA polymerase and ligase fill the gaps in the annealed product. The gap filling is in fact a restoration of the DNA sequences that were removed during the 5′-resection stage. This presents a good explanation for the >25% recombinant frequency observed in crosses with the z2-located markers. It also explains the enhanced recombinant frequency in D219A crosses at the longest distances, because, as can be deduced from Figure 6, the maximal recombinant frequency predicted by the SDSA model is 50%, which corresponds exactly to the experimentally observed value. If this explanation is correct, we have to suggest that D219A DNA polymerase, devoid of functional 43Exo, is imperfect in the formation of a full replication fork, and it frequently initiates the bubble-migration DNA synthesis. Note, however, that such synthesis does not lead to chromosome replication and hence cannot be the major pathway for the RDR in the T4 life cycle. We should add that the progeny DNA in the SDSA pathway contains only splices (single exchanges) and no patches. The patches, however, are produced in a nearly wild-type proportion in D219A crosses (our unpublished data). Therefore, SDSA may be only a minor pathway of DSB repair even in the D219A background.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.—

SDSA model for DSB repair (Mueller et al. 1996). Designations are as in Figure 1.

The increase in the maximum value of recombinant frequency in the D219A background could be a consequence of more frequent than in the wild-type background simultaneous formation of replication forks on both sides of the DSB (fc-SPC pathway). The 43Exo attack on the protruding 3′ terminus may be a way of loading the DNA polymerase on the D-loop with the consequent formation of a replication fork. A delay in the replication initiation may enhance the probability for the formation of replication forks on both sides of the DSB, i.e., shifting from the s-SPC to the fc-SPC. The fc-SPC predicts a higher recombinant frequency than the s-SPC does: 50% vs. 42% for the final plateau (Shcherbakov et al. 2002).

In the double-mutant dexA D219A crosses, the recombinant frequency is higher than in the wild-type crosses in the range of distances up to 427 bp. At longer distances, the recombinant frequency is a bit lower than in the wild-type crosses. We interpret these data as evidence for the triple effect of the DexA + 43Exo deficiency: (1) the preservation of 3′-terminal sequences at the ends of the broken chromosome greatly enhances recombination at the shortest distances; (2) contribution of the SDSA mechanism accounts for the enhanced recombination within the intermediate (zone 2) range; and (3) the absence of both 3′–5′ exonucleases creates a difficulty with the removal of the nonhomologous whisker from the D-loop intermediate, thus lowering the general efficiency of the DSB repair process that is overtly seen in crosses at the longest distances.

The in vitro data (Reha-Kranz and Nonay 1993) demonstrated that the D219A enzyme has extremely low exonuclease activity, and the dexA1 mutant was shown not to produce the DexA protein (Gauss et al. 1987). A possible contribution of certain host 3′–5′ exonucleases to the 3′ resection during the intron homing in T4 was tested by Huang et al. (1999) and was not observed. A relatively small reduction in the maximum recombination level observed in the double-mutant background raises a question about the functional role of 3′–5′ exonucleases in DSB repair, recombination, and RDR: Are they essential participants of the pathways or just accidental interferences? The 3′–5′ resection as such would not look crucial for the repair process, if it were not for the necessity to remove the nonhomologous 3′ termini. One cannot exclude that some unidentified endo- and/or exonuclease activities help to get rid of the heterology. B family DNA polymerases that contain both polymerase and proofreading exonuclease in a single polypeptide and share homology with T4 gp43 are known in other phages, viruses, yeast, and animals (Ho and Braithwaite 1991; Shcherbakova et al. 2003). Thus, DNA replication initiation in a D-loop via an attack of the proofreading nuclease on the protruding whisker may occur rather ubiquitously. For example, the 3′–5′ exonuclease of yeast DNA polymerase δ, in evocative resemblance to 43Exo, effectively removes 3′-nonhomologous tails <30 nucleotides during DSB repair, whereas the excision repair endonucleases Rad1p/Rad10p and the mismatch repair proteins Msh2p/Msh3p are needed for the removal of lengthy heterologies (Paques and Haber 1997; Colaiacovo et al. 1999).

In our SegC/ets1 system (Figure 1), the terminal portion of the 3′-recombinogenic tail (∼30 nucleotides) is not homologous to the intact i-chromosome, so one may wonder if the classical D-loop (like structure 4 in Figure 1) can be formed without first removing the nonhomologous terminus. In illustrative recombination schemes, the strand invasion step is usually depicted as an insertion of the 3′ tip into the double-stranded partner. But the “tip” is just the point of no use in the search for homology. In reality, the strand exchange is a multistage process involving a rather extended single-stranded stretch, not even obligatory terminal, although the complete strand exchange is thought not to occur in the absence of a free homologous end (for review see Griffith and Harris 1988). The processes related to the strand transfer were studied in vitro in numerous works. Probably, a presynaptic filament first forms a paranemic joint with the double-stranded DNA, and in a second reaction it is converted to a plectonemic joint (Riddles and Lehman 1985). The mechanism of the paranemic–plectonemic transformation is not clear, but it looks plausible that the transitory intermediates with the unpaired 3′ terminus like the one shown in Figure 1 (structure 3) arise during DSB repair (and RDR) as an initial stage even if the recombinogenic single-stranded ends are fully homologous. They may also reappear from the “whiskerless” D-loops as a result of random branch migration. It was demonstrated by Salinas and Kodadek (1995) that the 41-helicase, not the strand transferase, drives polar branch migration during the homologous strand exchange. The direction of this branch migration corresponds to the pairing progressing from the double-stranded/single-stranded edge to the 3′ tip, not in the opposite direction. The double-stranded/single-stranded DNA transition regions serve as DNA effectors for DNA-dependent ATPases (Muthuswami et al. 2000). Therefore, it is likely that these regions, not the tips, are active in joint formation, whereas the tip protruding from the D-loop is recognized by the proofreading exonuclease with the consequent initiation of DNA synthesis (Shcherbakov et al. 1995).

It is interesting to compare our present results with those obtained by Huang et al. (1999). They studied the effects of the same T4 mutations, dexA1 and D219A, on the transfer of group I intron I-TevI from a chromosomal donor to a plasmid recipient. The mechanism of intron homing is identical to recombinational DSB repair. The observed influence of the exonuclease mutations on the homing efficiency and coconversion frequencies was in general agreement with our recombination data. The only definite exception was that the combined effect of dexA1 and D219A was additive in our crosses (indicating independent action of the two enzymes), whereas Huang et al. found the homing efficiency in the double mutant to be intermediate between those in the single mutants (indicating action in the same pathway). The apparent discrepancy may arise from the fact that the homing efficiency and the recombinant frequency are not equivalent parameters. In the standard crosses of the i × ets1 type at the multiplicity of infection of five particles of each parent per cell, the efficiency of DSB repair could be assumed to be maximal. The recombinant frequency (R) that we measured here is related to coconversion (Co) (R = 1 − Co), and the coconversion frequency was not measured by Huang et al. in the dexA1 D219A mutant because of too low phage production. The triple consequence of the double 3′–5′ exonuclease deficiency discussed above may help to reconcile this seeming discrepancy: DexA and 43Exo are interchangeable in the removal of the heterologous single-stranded DNA (a step that is obligatory in the homing process also), while they act independently in the homologous 3′-termini hydrolysis.