INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

To avoid genetic catastrophe caused by double-strand breaks, prokaryotic and eukaryotic cells maintain biochemical systems that can efficiently repair these breaks (10, 44, 45, 54). A connection between double-strand-break repair and recombination has been established (45, 48), and a number of models have been proposed to explain the genetic consequences of double-strand-break repair (2, 7, 14, 24, 38, 43, 45, 48, 50, 57). In vitro systems have aided our understanding of double-strand-break repair mechanisms in bacteria and in mammalian cells (11, 31). The specific mechanism used to repair a double-strand break is likely to depend upon the nature and location of the break and upon the particular constellation of enzymes available to facilitate recovery and rescue of one or both of the partial genomes formed by the break. In no case is the biochemistry of double-strand-break repair fully understood. Studies from our laboratory investigated genetic rearrangements associated with double-strand breaks in bacteriophage T7 (18, 26, 56). Those experiments dealt primarily with the effects of double-strand breaks upon homologous recombination and deletion mutagenesis but did not consider explicitly the process by which double-strand breaks are repaired. The present paper focuses on the repair processes themselves and upon the relative contributions that different mechanisms make to the overall efficiency of repair. Several aspects of double-strand-break repair in T7 warrant further investigation. Although considerable progress has been made in understanding double-strand-break repair in Escherichia coli and in phages lambda and T4 (1, 8, 15, 32, 49, 51), it is not clear that T7 uses the same repair mechanisms. In E. coli and in lambda, double-strand breaks appear to be repaired primarily by recombination with intact DNA molecules (1, 15, 22, 31, 51). Recombination proceeds at a very high rate in T7 (47), but details of the mechanism of recombination in this phage remain obscure. The enzyme system employed by the phage during recombination is distinct from that of the host. T7 does not use E. coli RecA protein as a recombinase (36). The RecBCD protein, a major enzymatic component of one of E. coli's recombination pathways, is inactivated by T7 soon after infection to allow the phage genome to escape devastation from RecBCD's exonuclease V activity (53). Recently it has been found that T7 single-strand-DNA binding protein and the helicase encoded by gene 4 play major roles in annealing complementary DNA strands together (16, 19, 20). Thus, formation of single-stranded DNA and annealing of those strands to form heteroduplexes may figure prominently in T7 recombination (42). The uncertainty regarding the recombination process in T7 underscores the need for more information regarding what happens at a double-strand break in T7 that causes increased recombination and increased deletion when the break forms between a pair of direct repeats. Bacteriophage T7 has provided a powerful tool with which to study DNA replication and offers several advantages as an experimental system. The genetics of T7 are well developed, and a considerable amount is known about the biochemistry of T7 enzymes involved in DNA metabolism (5, 6, 16, 40, 46). Efficient in vitro systems for DNA replication and DNA packaging have been developed (9, 21, 29). In vitro systems have also been used to study recombination in T7 (28, 41, 42). Thus, further in vitro studies of double-strand-break repair in T7 have the potential to tell us more about recombination mechanisms in T7 and to allow comparison between T7 and other prokaryotic double-strand-break repair systems.

The in vitro system employed in this study was developed to study DNA repair and replication in T7 (29). One or more complete rounds of DNA synthesis are completed during the initial incubation, and the product from that reaction is encapsulated in an in vitro packaging system to produce viable T7 phage (21). The high yield of phage recovered from these reactions allow detection of replication and repair errors at frequencies as low as one error per million nucleotides (nt) incorporated (4). Under normal conditions, with undamaged exogenous DNA substrates, the error frequency of the in vitro system is about the same as what is measured for normal in vivo T7 growth (4). This system has been used to study repair and mutagenesis associated with various types of DNA damage and has proved especially useful in investigating deletion mutagenesis in vitro (17). To assay the frequency of deletion, we constructed T7 genomes with inserts of DNA that had a random sequence located between directly repeated sequences, one of which was in the insert and the other was in the T7 genome. The presence of the insert inactivated T7 gene 1.3 (phage ligase), while deletion between the direct repeats restored normal gene function. During the course of these experiments, we found that a double-strand break located between the repeats increased the frequency of deletion between those repeats (18). The deletion frequency increased as a function of the length of the direct repeat, reaching a value over 1% when the repeat lengths were 20 bp long. Other insert designs allowed measurement of deletion during intermolecular exchanges. When a partial genome was generated by a double-strand break, exchanges between repeated sequences in the partial genome and an intact genome could also cause deletion of an insert (56). Thus, double-strand breaks located between repeated sequences were found to be instrumental in increasing deletion by either intramolecular (18) or intermolecular (56) interactions. We also found that this in vitro system repaired double-strand breaks and that this repair was frequently associated with recombination of genetic markers located near the break site (26).

At least two molecular mechanisms might explain the observation that, in T7, deletion between a pair of direct repeats is stimulated by formation of a double-strand break between those repeats. One possibility is found in the single-strand-annealing model and its variants, which explain repair of double-strand breaks in mammalian cells and Saccharomyces cerevisiae (7, 24, 43). Digestion of a break in the T7 genome by a 5'  3' exonuclease, such as the product of T7 gene 6 (12), may expose single-stranded complementary copies of direct repeats located on either side of the break. Annealing between the repeats may then rejoin the genome at the expense of deletion of the region between the repeats (18). A second possibility is that the double-strand break may induce homologous recombination (26), and misalignment errors during strand transfer may occasionally produce deletions (56). This mode of homologous recombination might also be initiated by invasion of single-stranded tails, or if recombination in T7 depends upon relatively extensive DNA synthesis, errors during the DNA synthesis step may also lead to deletion between direct repeats. To gain additional insight into the mechanisms responsible for double-strand-break repair in T7, we employed an in vitro system capable of replicating and repairing T7 DNA (4, 17, 22). The work reported here serves to further characterize the T7 in vitro system as a tool with which to study repair.

In the experiments described below we compared the relative contributions of repair via direct rejoining of the broken ends, annealing of complementary sequences on either side of the break, and recombination with DNA segments that are intact in the region where the double-strand break formed in the damaged genome (see Fig. 2). (For convenience, we refer to these third DNA molecules as donor DNA.) The data show that effective repair of a broken T7 genome is greatly enhanced by the presence of donor DNA, which suggests that in this biological system recombination between partial and intact genomes is the major mechanism of double-strand break repair. We found that perfect homology between the break site and the donor DNA was not a major factor in the efficiency of repair but that the distance between the site of the break and the end of the donor DNA molecule was important. Our data suggest that ends generated by double-strand breaks are rapidly widened to gaps of between 150 and 500 nt before the break is repaired and show that gaps as large as 1600 nt can be efficiently repaired.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Bacteria and bacteriophages. Strains of E. coli used in this study included wild-type strain W3110, strain 011' (supE), and strain N2668 [lig-7(Ts)]. T7 phage was from the collection of F. W. Studier (46). T7 DNA sequence information is from the work of Dunn and Studier (6). Amber mutants of T7 used in this study included T7 with an am29 mutation in gene 3 (T7 endonuclease), T7 with an am28 mutation in gene 5 (the major subunit of T7 DNA polymerase), and T7 with an am147 mutation in gene 6 (T7 exonuclease). In the text mutants are designated with a superscript minus sign beside the number of the gene (i.e., T7 with an amber mutation in gene 3 is referred to as T7 3). The T7 A mutation is a deletion extending from the promoter of gene 1.3 to the promoter of gene 1.5 (34). This phage completely lacks T7 ligase (gene 1.3). T7X is a derivative of wild-type T7 with a unique XhoI site engineered at position 6663 by site-directed mutagenesis (34). The XhoI site is within gene 1.3, which encodes T7 ligase. This mutation does not alter the amino acid sequence of the product of gene 1.3. For some experiments inserts of DNA (Fig. 1) were place in the XhoI site so as to inactivate gene 1.3. A T7 mutant deficient in ligase (T7 1.3) is able to grow on lawns of wild-type E. coli but cannot form plaques, even at the permissive temperature on a lig-7(Ts) host (strain N2668), which is temperature sensitive for E. coli ligase (25). Wild-type T7 is able to form plaques on either wild-type or lig-7(Ts) E. coli (25).

View larger version (26K): [in this window] [in a new window]   FIG. 1.   Inserts placed in gene 1.3 (ligase) of bacteriophage T7. DNA sequences of inserts X76/17, X72/5, and X76/171.3-1.7 are shown. The top line shows the sequence near the XhoI site at position 6663, into which the inserts were placed. The sequences of the inserts are shown in uppercase letters, and the surrounding T7 genome sequences are shown in lowercase letters. Direct repeats are double underlined. The restriction sites are marked with a line above the sequence, and the arrows indicate the positions where the restriction enzymes cut. Stop codons are marked with an asterisk over the middle base in the codon.

Construction of T7X76/171.3-1.7. A T7 genome with the nonessential genes 1.3, 1.4, 1.5, 1.6, and 1.7 totally removed or inactivated was constructed. The T7X genome was cut with XhoI, and a 76-bp insert of double-stranded DNA (X76/17) was ligated into the XhoI ends (Fig. 1). This insert has a unique BamHI site in the midst of the insert and a unique SacI site near the right end. When it is placed in the T7 genome, a pair of 17-bp direct repeats are formed. The left repeat is contained within the insert, while the right repeat consists of DNA sequence present in the original T7 genome. This genome is designated T7X76/17. With the X76/17 insert in place, the T7 genome is ligase deficient. Deletion between the direct repeats will restore the original T7X genome, which is wild type for ligase. T7X76/17 DNA was cut with SacI. An adapter oligonucleotide with the sequence 5' GATCAGCT 3' was joined to the SacI ends in order to create an end compatible with that generated by a BclI restriction cut. T7 wild-type DNA was digested with BclI, which cuts at a unique site at position 8311. The X76/17 SacI-digested DNA with the BclI-compatible adapter was joined to the BclI fragment formed from wild-type T7. The resulting DNA molecules were incubated in the T7 in vitro packaging system to select for genomes capable of producing viable phages. Resulting phages were tested for gene 1.3 inactivation by screening for inability to grow on an E. coli lig-7(Ts) host (N2668). One of the resulting phages was selected, and its DNA was examined by restriction digest analysis and by dideoxy sequencing performed on double-stranded DNA with T7 gene 6 exonuclease and T7 Sequenase (both from U.S. Biochemicals) as previously described (55). The resulting phage was designated T7X76/171.3-1.7. The relevant portion of its genome is shown in Fig. 1. T7X76/171.3-1.7 is 1648 nt shorter than wild-type T7 but grows normally on a wild-type E. coli host. It is ligase deficient due to part of the X76/17 insert remaining at the XhoI site and deletion of the rest of gene 1.3. The T7X76/171.3-1.7 phage will not form plaques on an E. coli lig-7(Ts) host.

Growth conditions. Bacteria were grown in L broth (30) at either 32 or 37°C with rapid shaking. Phages were grown on agar plates made from T broth (30) and incubated at 32°C.

Preparation of DNA. DNA was prepared as described by Richardson (39). DNA concentrations are given in nucleotide phosphorous equivalents. Thus, in the tables below, 1.5 nmol of nucleotide phosphorous represents 1.1 × 10¹⁰ T7 genomes. When 1.5 nmol of a restriction digest of donor DNA is added to the genomes, there is one molecule of the relevant restriction fragment for each genome.

The in vitro DNA repair system. The system employed to monitor double-strand-break repair used extracts made from T7-infected E. coli and exogenous T7 genomes with or without double-strand breaks as the substrates (18, 26). DNA products recovered from these reactions were packaged in vitro (21) to produce infective T7 phage. The numbers of phages produced provided measures of the efficiency of double-strand-break repair. The major advantages of the in vitro system are the ability to control the quantity and structure of the DNA molecules added to the reaction mixtures and the high sensitivity provided by the large numbers of phages generated in the experiments. Since in vitro packaging is used as a final step, it must be kept in mind that the DNA molecule generated by the first in vitro reactions might be subject to further modification during the packaging step. This possibility is not viewed as a serious concern since the same processing can presumably take place during normal in vivo packaging. Another cost associated with the high sensitivity provided by packaging is the fact that visualization of the packaged DNA product in the form of plaques on a bacterial host necessitates in vivo growth as a final step, and mismatch repair or recombinational exchanges during this growth step remain a possibility (42). This consideration is of only minor consequence in the work reported here since in most experiments we used fragments of DNA rather than intact DNA substrates to help repair the double-strand breaks. T7 requires unique 160-bp repeats on the ends of its genome (47), and there is no evidence that incomplete T7 DNA molecules without this end sequence can be encapsulated into phage heads and then contribute to repair during the in vivo growth step. In earlier studies that included experiments that could be done both in vivo and in vitro, we found essentially perfect agreement between in vivo and in vitro results (17, 18, 55, 56). This raises confidence that in vitro studies performed with this system are relevant to the in vivo situation.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

What is the efficiency of repair of DNA molecules with a simple double-strand break? T7 X72/5 DNA was digested with BamHI so as to produce one double-strand break in each T7 genome. This broken DNA was incubated, without prior treatment, in the in vitro T7 DNA packaging system, and the number of viable phage was compared with what was found with the same number of intact T7X72/5 DNA molecules (Table 1). The DNA suffered more than a 1,000-fold loss in potential for generating viable phage as the result of the double-strand break at the BamHI site. We performed a similar experiment in which, prior to the in vitro packaging step, the DNA was incubated in the in vitro DNA replication system. The yield of phage from intact genomes increased about 10-fold after incubation in the in vitro DNA replication system. This increase was caused by replication of the intact genomes and by improved potential for packaging, probably by partial maturation of the newly replicated genomes during the first incubation step (4). Incubation of the DNA with the double-strand break in the in vitro system caused an improvement in viability of more than 2 orders of magnitude. Still, Table 1 shows that even after incubation in the in vitro DNA replication system, viability of the DNA with the double-strand break was only about 1% of what was found with the intact genomes. Another reaction mixture included T7 X72/5 DNA with a double-strand break and DNA from T7 genomes that had been digested with BstXI. The BstXI restriction endonuclease produces 11 cuts in the T7 genome, thereby effectively precluding restoration of intact, potentially viable, T7 genomes from the array of restriction fragments. BstXI-treated T7 DNA, by itself, does not produce a significant number of T7 phage in this experimental system (data not shown). A marked improvement in DNA repair was noted when the BstXI restriction fragments were included with broken T7X72/5 genomes (Table 1). Inclusion of the BstXI fragments in the reaction mixtures increased viability of the broken DNA to nearly 40% of what was found with intact genomes. Moreover, essentially all of the repaired DNA had acquired a functional copy of gene 1.3, which allowed the repaired phage to grow on a lig-7(Ts) host as well as on a wild-type host. In contrast, less than 0.1% of the intact genomes became ligase positive as a result of incubation with the BstXI fragments (Table 1). This experimental system was constructed with all endogenous DNA deleted of gene 1.3 so that recombination with contamination DNA could not account for formation of ligase-positive phage. Moreover, the fact that the X72/5 insert is bracketed by repeats only 5 bp long means that this insert cannot be deleted from the genome at a measurable frequency (34). Thus, recombination with the BstXI fragments, which have wild type gene 1.3, provides the only avenue for acquisition of a wild-type ligase gene (Fig. 2). Table 1 shows that the frequency of this recombination is low when the T7 genomes are intact but relatively high when there is a double-strand break in the genome. (Both the frequency and the absolute number of ligase-positive phage increased with the presence of a double-strand break in the genome. Thus, the high frequency of ligase-positive phage obtained after a double-strand break is not simply the result of selection for genomes that have been repaired.) Direct end-to-end joining of the partial genomes formed by the double-strand break would not require the BstXI fragment and would not give rise to ligase-positive phage, as seen in Table 1. The low frequency of recovery from the double-strand break measured in the absence of the BstXI fragments shows that this repair pathway is less important than the repair involving interaction with donor DNA molecules.

View larger version (15K): [in this window] [in a new window]   FIG. 2.   Models of mechanisms of double-strand-break repair. (A) Direct repair. The direct repair model depicts joining of the two partial genomes formed by the double-strand break and subsequent ligation to form an intact genome. Presumably, such direct joining is facilitated by sticky ends present on each of the partial genomes. (B) Single-strand annealing. In this model the double-strand break forms between a pair of direct repeats. One strand of DNA is digested by an exonuclease to produce single-stranded tails, thereby exposing complementary versions of the repeated regions, which can subsequently anneal with one another. Trimming and some limited DNA synthesis reestablish a repaired genome with the region between the repeats deleted. (C) Recombinational repair. In the recombinational repair model the break is repaired by recombination with a third molecule of DNA essentially homologous to the broken genome. The recombination is assumed to be promoted by the break, and widening of the break to a gap via exonuclease activity may precede the recombinational repair.

Comparison of repair mediated by annealing between repeated sequences or with the help of a donor molecule. When a double-strand break occurs between a pair of direct repeats in the T7 genome, the frequency of the deletion of the region between those repeats is very high and exceeds 1% for repeats of 20 bp (18, 56). Thus, annealing of the complementary sequences provided on each partial genome may serve to assist in the repair of the break, with a consequent loss of all genetic information between the repeats (Fig. 2). Another possibility is that misalignment between repeated sequences during homologous recombination initiated at the double-strand breaks leads to deletion (18, 26, 56). Although repair of double-strand breaks contributes to deletion between those repeats, it was not clear whether the presence of the repeats contributed significantly to repair. To monitor the involvement of a pair of 17-bp direct repeats in repair of a double-strand break placed between those repeats, we used the X76/17 and the X72/5 inserts shown in Fig. 1. T7 genomes with the X76/17 insert were cut with BamHI. The two fragments were separated in a neutral sucrose gradient as previously described (17) so as to isolate left BamHI fragments and to remove intact genomes from which the insert had already been deleted. (Since phage that deleted the insert do not have a BamHI site, their genomes remain intact after exposure to this restriction enzyme and therefore sediment much faster than the 6,691-bp left BamHI fragment from genomes that still have the insert. The left BamHI fragments are essentially free of contamination from genomes from which the insert has already been deleted [17].) In the experiment described in Table 2 the left DNA molecules came from genomes with an X76/17 insert. The right partial genomes came from T7 with an X72/5 insert that had been cut with PstI. The right PstI fragment was separated from the left fragment by sucrose gradient sedimentation. As seen in Fig. 1, the sequence of the X72/5 insert right of the BamHI site matches the X76/17 insert exactly. However, since the direct repeats flanking the X72/5 insert are only 5 bp long, this insert cannot be deleted from the genome at any measurable frequency. (In T7, repeats of 5 bp or less promote deletion at frequencies below 10¹⁰ [34, 35].) Thus, we are not concerned that X72/5 genomes might have been contaminated with T7 from which the insert had already deleted (17). Furthermore, use of a BamHI cut on the left fragment and an incompatible PstI cut on the right fragment helped to ensure that full-length genomes would not be reassembled by ligation of the restriction cuts on the right and left fragments. This means that repair via end-to-end ligation of the left and right partial genomes was not an option in this experiment. However, the experimental design permits repair via annealing of complementary versions of the 17-bp repeat on the left and right partial genomes. This annealing mode of repair deletes the insert and gives rise to a viable, ligase-positive, genome, as depicted in Fig. 2. In the experiments whose results are shown in Table 2, all genomes had double-strand breaks. This fact precluded using the ratio of phage yield from broken and intact DNA substrates as a measure of repair efficiency. To estimate relative repair efficiency, we used the data shown in Table 1 indicating that repair was highest when donor DNA was present. Thus, the number of phage produced on wild-type E. coli with X72/5 left and right DNAs (both with BamHI cuts) and X32/4 donor DNA (with a BstXI cut) in Table 2 is considered 100% relative phage yield and other values in the table provide relative measures of the efficiency of repair. Repair resulting from annealing of direct repeats produces ligase-positive phage, whereas any other mode of repair should produce phages that are viable on the wild-type host but not on the lig-7(Ts) host. Table 2 shows that without donor DNA present, about 0.6% of the viable genomes generated in these reactions had the insert deleted via annealing of the 17-bp sequences repeated in the left and right partial genomes. The number of ligase-positive phage generated was essentially unchanged whether or not the BstXI fragments (which have gene 1.3 inactivated by an X32/4 insert) were present (Table 2), but the number of viable phage, as measured on wild-type E. coli, was over 100-fold higher when the BstXI fragments were included in the reaction mixtures. The 17-bp sequence that is duplicated on the X76/17 and X72/5 fragments is not present on the BstXI digest of X32/4 DNA. Table 2 shows that repair caused by recombination with the BstXI fragment was more than 2 orders of magnitude greater than the amount of repair caused by annealing of the 17-bp repeats on the flanks of the double-strand break. Although, in T7, double-strand breaks increase deletion frequency by about 2 orders of magnitude and, depending upon repeat length, can lead to deletion frequencies greater than 10² (18, 56), the mode of repair that leads to these deletions is only a minor component of the mechanisms available to correct double-strand breaks.

Repair of a large gap in the genome. We considered whether a large gap in the T7 genome could be effectively repaired by the in vitro system. This experiment used genomes from T7X76/171.3-1.7 (Fig. 1). Double-strand breaks were placed in these genomes by cutting the unique BamHI restriction site. Intact genomes able to generate viable phages can be formed by rejoining the sticky ends formed by the BamHI cut. This type of repair gives rise to ligase-negative phage. Because of the large deletion in T7X76/171.3-1.7, restoration of the ability to grow on a ligase-deficient bacterial host can be achieved only by filling in the deleted region between genes 1.3 and 1.7. Thus, in this phage restoration of ligase function as part of the repair of a double-strand break is equivalent to filling in a 1,648-nt gap. Table 3 shows that even without added BstXI fragments, the in vitro system was able to markedly improve the viability of the broken DNA (with numbers of phage on the wild-type host from 3.0 × 10² to 6.9 × 10⁴). This level of repair brings the viability to 0.7% of what was found without a double-strand break in the genomes (Table 3). With the BstXI fragments present, repair was considerably higher. Experiments with intact and cut DNAs incubated in an extract of E. coli infected with T7 3 phage and with a BstXI digest of mutant gene 6 DNA (Table 3) showed repair equivalent to 15% of what was measured without a double-strand break. It is noteworthy that essentially all of the DNA molecules that were repaired in the presence of the BstXI donor molecules became ligase positive and were able to grow on a lig-7(Ts) host. This result contrasts with other results (Table 3) which showed that in the absence of the double-strand break, the intact genomes did recombine with the BstXI-digested DNA but that acquisition of functional gene 1.3 was only 0.03%. The high recovery of ligase-positive phage during repair of the double-strand break indicated that genetic information on the BstXI fragment was incorporated into the repaired genome. The functional gene 1.3 could not come from endogenous DNA contamination in either the extract used for DNA replication or the extract used for in vitro packaging, since both of those extracts were made with T7 phage that had gene 1.3 completely deleted. Instead, genetic information in the BstXI fragment was carried into the repaired genomes either by physical incorporation of the donor DNA molecule into the repaired genome or by new DNA synthesis with the donor DNA as the template.

Homology between the broken DNA and the donor DNA near the break site is not essential to efficient repair. The experiment described in Table 1 was performed under conditions where the DNA sequence in the immediate vicinity of the ends produced by the double-strand break did not match any sequence present on the BstXI fragment that helped to repair the break. An experiment was performed to test repair of a double-strand break in the presence of donor DNA molecules that are fully homologous to the T7 genomes with a double-strand break. Genomes with an X72/5 insert were left intact or cut with BamHI. Repair was attempted without donor DNA or with BstXI fragments derived from either T7 genomes with wild-type gene 1.3 or from T7 DNA with the same X72/5 insert that was present in the BamHI-digested genomes. The sequence near the ends of the BamHI-induced double-strand break exactly matched the sequence in the BstXI fragments from T7X72/5. But, as in Table 1, the ends of the double-strand break did not match BstXI fragments from T7 DNA with wild-type gene 1.3 T7. Either type of BstXI fragment proved very effective in facilitating the repair of the double-strand breaks (Table 4). Phage yield from the BamHI-cut DNA was over 3 orders of magnitude higher when the BstXI fragments were present. In the presence of homologous donor DNA, about 50% of the genomes were repaired. About half of that number were repaired when the break site was nonhomologous with the donor sequence. Considering the experimental uncertainties inherent in the in vitro system, it is unlikely that this twofold difference is of much significance. All of the repaired genomes arose through recombination. Thus, perfect homology between the ends formed by the double-strand break and the donor DNA that contributes to the repair is not necessary, although it may make the repair a little more efficient. We assume that homology is required for single-strand annealing or D-loop formation associated with repair of the double-strand break. Thus, elimination of the nonhomologous region before or during the repair process may be important to efficient repair. Our data may mean that there is considerable digestion of the DNA near the site of the double-strand break so that the nonhomologous region is essentially eliminated, or it may mean that recombination originates at the ends of the relevant BstXI fragment. The latter explanation seems less likely because the ends on the BstXI fragment do not contribute nearly as much to recombination when the recipient genome is fully intact. In fact, it is noteworthy that the number of ligase-positive phage (not just the percentage) is 2 orders of magnitude higher when there is a cut in the genomic DNA.

Repair of double-strand breaks requires a minimal length of donor DNA. If the double-strand breaks are normally widened to a gap prior to their repair, there should be some minimal length of donor DNA below which the donor molecules are too short to fill the gap. The relevant BstXI fragment which rescues the broken genome is 5,736 bp long and extends from positions 3863 to 9623. As diagrammed in Figure 3, the cut site at position 6663 is approximately in the middle of this BstXI fragment. For comparison, we tried donor DNA made with two different restriction enzymes. Figure 3 shows the endpoints of the donor DNA relative to the position of the double-strand break. BsrGI cuts T7 DNA into 14 pieces, including a fragment that covers the double-strand break and extends from positions 5516 to 10714. This 5,198-bp fragment is nearly the same length as the relevant BstXI fragment. However, the left end of the BsrGI fragment is only 1,147 bp from the double-strand break. The BsrGI fragments proved very effective at repairing double-strand breaks (Table 5). With 1.5 nmol of the BsrGI fragments in the reaction mixtures (the same amount of donor DNA that was used when the BstXI restriction enzyme cut the donor DNA), more than 50% of the broken genomes were repaired. Moreover, all of the viable genomes were ligase positive, indicating that they had lost the insert that had been present on the broken genomes by recombining with the 5.2-kb BsrGI fragment. Thus, a fragment with an end 1.1 kb from the break site is adequate for repair. We also digested the T7 genome with DrdI, which cuts it into 12 pieces. The 755-bp DrdI fragment that covers the region of the double-strand break extends from positions 6209 to 6964. Table 5 shows that 1.5 nmol of DrdI-digested DNA was unable to improve repair of the double-strand break beyond what was seen without donor DNA present. Broken DNA was restored to only about 1% of the viability of intact DNA, and about 19% of the resulting phage had lost the X72/5 insert from gene 1.3. Thus, the 755-bp DrdI fragment was apparently able to make some contribution to the survival of broken genomes. It seems likely that the remainder (80%) of the phage arose from direct ligation of the partial genomes left by the double-strand break. With 1.5 nmol of the DrdI digest added to the reaction mixture, the genomes with a double-strand break and the relevant donor DNA molecules were present at a 1:1 ratio. When the concentration of DrdI fragments was increased by a factor of 10, the repair efficiency improved to 3.2%. Most of the repaired phage arose from recombination, as judged from the observation that over 80% of the viable phage were now ligase positive. Thus, donor DNA from the DrdI digestion, which has its right end only about 300 bp from the double-strand break, can assist in the repair of the break but, even at high concentration, the efficiency of this repair is limited.

View larger version (15K): [in this window] [in a new window]   FIG. 3.   Positions of the ends of the donor DNA relative to the position of the double-strand break on the T7 genome. The top line shows an abbreviated genetic map of bacteriophage T7. The region around gene 1.3 is shown in greater detail, with the unique XhoI site at position 6663 marked. The insert in gene 1.3 is shown as a white segment within the dark region, which indicates the gene 1.3 sequence. The shaded segments indicate relevant restriction fragments used as donor DNA. The white segments on the ends of the pJP6 PvuII segment indicate regions on the ends of the PvuII fragment that come from plasmid pJP6 (34) and are not homologous with the T7 genome. Donor DNA molecules are drawn to scale. The nucleotide positions correspond to the DNA sequence determined by Dunn and Studier (6).

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The experiments reported here demonstrated a very high efficiency of double-strand break repair in bacteriophage T7. The repair of these breaks derived primarily from recombination with donor DNA molecules that bracket the region of the double-strand break. Direct joining of broken ends and annealing of single-stranded versions of repeated sequences on partial genomes made much smaller contributions to double-strand break repair. Also of interest is the finding that it is not critical that there be good homology between the ends of the double-strand break and other DNA molecules involved in the repair. However, efficient double-strand break repair does depend upon the distance between the break site and the end of the donor DNA molecule. These last two observations suggest exonuclease digestion of ends created by double-strand breaks prior to their repair.

The motivation for the experiment in Table 2 derived from our earlier study (18) that showed increased deletion frequency between a pair of direct repeats when a double-strand break formed between the repeats. That observation immediately suggested that the observed deletions were by-products of a repair mechanism, such as the one described in the legend to Fig. 2B, dedicated to rescuing partial genomes. Table 2 shows that when a break was placed between direct repeats, about 5 × 10⁴ phage were generated. Less than 1% of these had deleted the DNA between the direct repeats, presumably by annealing the complementary sequences provided by the repeats on each partial genome. Some portion of the other 99% of the rescued genomes may have arisen by direct ligation of the partial genomes, but since the BamHI and PstI ends on the left and right fragments are incompatible, direct ligation of these fragments seems unlikely. The presence of fully compatible BamHI ends on both fragments did not improve repair efficiency. A more attractive alternative is that endogenous DNA in the extracts used in the in vitro system may have contributed to repair. The phage yield from the same DNA was more than 2 orders of magnitude greater when donor DNA molecules were present. Thus, although annealing between direct repeats can account for a substantial number of deletions (18), this mechanism represents only a minor route to repair of double-strand breaks relative to what is accomplished via recombination with donor DNA.

The absence of homology between the break sites in the genomes and the donor DNA molecule may interfere with invasion of the ends of the double-strand break into the donor molecule. A comparison of repair with and without perfect homology between the break site and the donor DNA showed repair efficiency that was higher by only about a factor of 2 when the ends of the double-strand break exactly matched DNA sequence within the donor molecule (Table 4). Over 25% of the broken DNA molecules were repaired even when the ends of the double-strand break did not match the donor DNA. This observation might mean that the recombination that salvages the broken genome originates at the ends of the donor DNA, thereby bringing the donor DNA into the repaired genome as a patch. However, Table 6 shows that inhomogeneities at the ends of the double-strand break and at the ends of the donor DNA had essentially no effect on repair efficiency. Thus, a more attractive explanation is that the ends on the DNA molecules may have been digested considerably before the broken genome invaded the donor DNA. It is likely that the fragments of the insert on the ends of the broken genome were removed by exonuclease digestion prior to recombination.

The length of the donor DNA molecule is a determinant of repair efficiency. Table 5 shows that 5.2-kb BsrGI donor DNA molecules with left ends only a little over a kilobase from the site of the genome break repaired double-strand breaks very efficiently. However, 0.75-kb DrdI fragments with one end about 300 bp from the break site were far less effective at facilitating repair of the break even when they were present at 10 times the concentration of the BsrGI fragments. As shown in Table 6, perfect homology either between the break site and the donor DNA molecule or between the ends of the donor molecules and the genome had no effect on the efficiency of repair. These data argue that substantial digestion takes place at the ends of the DNA molecules. Since the BsrGI fragments are very effective in the repair process, the digestion does not, on average, extend about 500 nt from both the site of the double-strand break and the left end of the BsrGI fragment. However, digestion of 150 nt in either direction from the site of the double-strand break and from the right end of the DrdI fragments would explain why these fragments are, as shown in Table 5, ineffective in assisting in repair of the break. Repair may be improved at a higher concentration of DrdI fragment because the distribution of digested donor molecules includes a sufficient number of individual donor DNA molecules with lengths adequate to support the repair mechanism. These data do not answer the interesting question of what is the minimal length of homology needed for efficient pairing.

The experiments reported here were done with extracts deficient in T7 gene 3 endonuclease, an enzyme known to be involved in homologous recombination in T7 (13, 23, 37, 41, 52), presumably because of its well-established role in resolving Holliday junctions (3, 33). The choice of gene 3-deficient phage for the present experiments was prompted by our earlier evidence for recombination being part of the process of repair of double-strand breaks in the same in vitro system (26). In fact, those experiments (26) showed essentially the same recombination frequencies for wild-type and gene 3-deficient extracts (70 versus 78%) when a break 181 nt from the suicide in Shigella marker (ss) was repaired (26). A trivial explanation for these data is that there is probably some gene 3 endonuclease present in the extracts used in our experiments in spite of the amber mutation present in the phage used for extract preparation. Even a lowered level of gene 3 endonuclease might suffice for repair in this in vitro system. A more interesting possibility is that more than one recombinational mechanism operates in T7-infected E. coli (41) and that the mode of recombination responsible for the repair of double-strand breaks does not require the gene 3 product. For example, it is not obvious that the recombination that repairs a double-strand break in T7 involves a Holliday junction or requires the enzyme that resolves those junctions.

Formation of double-strand breaks may be a common occurrence in a typical T7 infection. DNA damage or errors during the replication process may cause fragmentation of the genome, particularly near the advancing replication fork. Maintenance of an efficient double-strand-break repair mechanism may provide a substantial advantage in rescuing partial genomes and thereby maximizing production of T7 phage during the brief period (18 to 20 min) between initial infection and lysis of the host. The high frequency of deletion observed when a double-strand break is placed between direct repeats in a T7 genome may be a consequence of attempting to repair the break via annealing of complementary copies of the repeat, as in the model depicted in Fig. 2B. The data presented here show that, at least in this in vitro system and under these experimental conditions, annealing of complementary copies of the repeats is a relatively minor repair pathway. The data presented above are more compatible with a mechanism where ends formed by a double-strand break are widened into a gap by exonuclease action and another T7 DNA molecule is used to facilitate closure of the gap. This donor DNA may be physically incorporated as a patch into the resurrected genome, or the donor DNA may serve as the template for new DNA synthesis that produces complementary single-strand tails in the partial genomes. The in vitro system we describe here should prove useful in working out the biochemical details of the predominant mechanism of double-strand-break repair in T7 and should help in identification of enzymes involved in this repair pathway.