INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Double-strand breaks in DNA confront the cell with potentially disastrous consequences, in the form of permanent loss of genetic information. Double-strand breaks can result from DNA-damaging agents (7), aberrant interactions between topoisomerases and DNA (11, 43), or from collapsed replication forks (2, 19). To counteract the deleterious effects of double-strand breaks, most organisms maintain elaborate repair mechanisms directed against these lesions (3, 7). Recombination with undamaged portions of homologous genomes offers an economical scheme for rescue of partial genomes formed by double-strand breaks. A connection between double-strand breaks and recombination (both homologous and illegitimate) has been well established in a number of biological systems, including yeasts, bacteria, and bacteriophages (9, 12, 32, 44, 46, 53, 54). Our laboratory has been examining the repair of double-strand breaks by using an in vitro system based on extracts made from Escherichia coli infected with bacteriophage T7 (13, 21, 25, 55). In this system, DNA replication closely mimics the in vivo replication of T7 DNA (4, 28). Moreover, the in vitro system is able to carry out at least some steps of homologous recombination (22, 23, 27, 38, 39). To study double-strand break repair, breaks are experimentally introduced with a restriction endonuclease at a predetermined site in the T7 genome. The broken genomes are then incubated in the in vitro system before the DNA is recovered and packaged into infective T7 phage. The yield of viable phage reflects the number of intact genomes and, therefore, the efficiency of double-strand break repair. Double-strand breaks are repaired efficiently in this system, and repair of the breaks is often accompanied by acquisition of genetic information from other DNA molecules present in the same reactions (25). When a double-strand break occurs between a pair of direct repeats, the break can increase the frequency of deletion of the region between the repeats by 2 or more orders of magnitude (13, 55).

Although double-strand breaks in the T7 genome can be repaired by direct ligation of the two partial genomes formed by the break, repair is more than 2 orders of magnitude higher if the reactions contain intact DNA molecules that bracket the break site (21). The repaired DNA acquires genetic information from the donor DNA (25). The relationship between repair efficiency and the relative positions of the double-strand break and the ends of the donor DNA molecules suggests that the double-strand break is widened to a gap prior to its repair (21). The exact function of the donor DNA is unclear, but clues to its role in the repair process can be gleaned from what is known from other biological systems (1, 6, 9, 20, 24, 31, 36, 41, 51, 52). Figure 1 outlines some possible mechanisms by which donor DNA could participate in repair of a double-strand break. A form of the double-strand break repair model, successfully applied to yeast (36, 46, 51), might account for both the repair of the double-strand break and the accompanying transfer of genetic information from donor to T7 genome. In this model, the double-strand break is widened to a gap before the 3' ends of the broken genome each invade a homologous donor DNA molecule. Both strands of the donor DNA serve as a template for DNA synthesis that copies information from the donor to the repaired genome in a nonreciprocal fashion. Variations of this model also use the donor DNA as a template but allow for independent synthesis of single-strand tails on each partial genome. Annealing of these tails repairs the break, again with transfer of information to the repaired genome (36). A second mechanism, shown in the center of Fig. 1, is double-strand break-induced replication fork formation. This mechanism involves invasion of a single-stranded end of the broken DNA into the unbroken duplex region of the donor so as to form a three-stranded D-loop. Nicking of the displaced strand of the D-loop followed by ligation between the invading strand and the nicked strand of the D-loop establishes a replication fork. The replication fork advances until it reaches the end of the donor DNA. If necessary, the process is repeated with the ends of the newly synthesized DNA, forming new replication forks until the genome is completely replicated. This mechanism is characterized by canonical DNA replication fork progression rather than repair-like DNA synthesis that generates single-stranded tails. A third possibility, shown on the right side of Fig. 1, is physical incorporation of the donor DNA into the recipient repaired genomes. It is known that in T7, as in other systems (25, 33, 53), a double-strand break encourages recombination. The molecular nature of the double-strand break-induced recombination events in T7 is not understood, and the enzymes involved are not known. A pair of crossovers between the donor and broken genomes, one on either side of the break, could seal the break while carrying genetic information from donor to recipient. Or, digestion of one DNA strand at the ends created by the double-strand break could produce complementary single-strand tails on the recombining DNA molecules which could allow recombination via single-strand annealing.

View larger version (11K): [in this window] [in a new window]   FIG. 1.   Possible roles for donor DNA in repair of a double-strand break. In all three models the double-strand break is first widened to a gap. In scheme A, both strands of the donor DNA serve as a template for new DNA synthesis. The 3' ends on each partial genome invade the donor DNA and synthesize long single-strand tails which can then anneal with each other to close the break while transferring information from donor to repaired genome in a nonreciprocal fashion (36, 51). Alternatively, the DNA molecules could form a pair of Holliday junctions which, after resolution, would repair the genome and transfer markers from the donor DNA (17, 53). In scheme B, ends formed by a double-strand break invade a homologous segment of DNA and form a replication fork (1, 8, 20, 32). In our experimental design, ends on either the genome with one double-strand break or the donor DNA could initiate formation of a new replication fork. DNA synthesis elongates the partially repaired genome. This process is repeated until an intact, fully functional genome is completed. In scheme C, recombinational crossovers between the genome with one double-strand break and the donor DNA physically incorporate the donor DNA into the gap in the repaired genome. The nature of the crossovers is poorly understood and might involve elements of scheme A. The third mechanism (C) is imagined to proceed with very limited DNA synthesis.

Extensive DNA synthesis accompanies recovery from double-strand breaks in many biological systems, including E. coli and its phages lambda and T4. Recombination-dependent replication, the formation of new replication forks via recombination with intact DNA molecules (as in Fig. 1B), offers an efficient and economical means of rescuing partial genomes generated by double-strand breaks (9, 31, 32). Reconstruction of DNA replication forks from recombination between partial and intact genomes represents a major part of bacteriophage T4's infective cycle (8, 32). In E. coli, collapsed replication forks are reestablished by recombination with intact regions of the bacterial chromosome (1, 19, 20). A dnaE mutation in the major E. coli DNA replicase, DNA polymerase III, blocks formation of recombination intermediates that accompany repair of a double-strand break (20). Given the importance of recombinational reconstruction of DNA replication forks in other systems and the similarity between the basic DNA replication process in E. coli, lambda, T4, and T7, we considered whether in T7 formation of new replication forks at a break site (Fig. 1B) provides a major route to double-strand break repair. Inhibition of DNA replication should put an end to any repair pathway involving extensive DNA replication via establishment of new replication forks after invasion of ends created by double-strand breaks into intact homologues. In particular, recombination-dependent replication beginning at a recombination event at the break site and extending to the end of the genome would be precluded by inhibition of normal DNA replication. This would either force the double-strand ends into alternative repair pathways or permanently inactivate partial genomes that invade homologous DNA molecules by freezing the newly formed replication forks. The data presented below show that in T7, simple double-strand breaks and gaps of 1,600 nucleotides (nt) are repaired with about the same efficiency whether or not T7 DNA polymerase is present at its normal level. These data suggest either that establishment of new replication forks is not a primary means for double-strand break repair in the in vitro system employed in this study or that alternative pathways for this mode of repair are robust enough to handle a double-strand break in every DNA molecule with no apparent loss in efficiency when the phage DNA polymerase is inactivated.

An intrinsic limitation of using packaging to assay double-strand break repair is that for some applications it is impossible to separate the steps taking place during the in vitro repair reactions from those taking place in either the packaging reactions or the in vivo growth step needed to convert the single phage particle created by in vitro packaging into a visible plaque. To circumvent this concern, we repaired broken T7 genomes in vitro, recovered the DNA product from those reactions, and cut it with KpnI to separate the segment that had the double-strand break from the rest of the genome. After electrophoresis, an intact restriction fragment that covered the region where the double-strand break had been placed was recovered under conditions in which donor DNA was present and repair could proceed. This KpnI "D" band was not found in the product of reactions that did not contain donor DNA, thereby indicating that the intact KpnI D band arose as a result of repair of the double-strand break. Double-strand break repair, as monitored by direct visualization, was seen whether or not T7 DNA polymerase was present. These data corroborate the experiments employing in vitro packaging and demonstrate repair without normal levels of T7 DNA polymerase.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacteria and bacteriophage. E. coli strains included W3110 (wild type, sup⁰) O11' (supE), and N2668 [lig-7(Ts)]. T7 phage were from the collection of F. W. Studier (47, 49). Amber mutants used in the extract preparations included am29 in gene 3 (endonuclease I), am28 in gene 5 (DNA polymerase), and am147 in gene 6 (5'  3' exonuclease). The A mutation is a deletion from the promoter of gene 1.3 (T7 ligase) to gene 1.5 (35). (The functions of the nonessential genes 1.4 and 1.5 have not been determined.) The sequence of the entire T7 genome has been reported by Dunn and Studier (5). In some experiments, inserts of DNA were placed in an XhoI site that had been engineered at position 6663 in the T7 ligase gene (gene 1.3). The construction of this phage, designated T7X, has been described previously (35). The construction and sequences of the specific T7 inserts used in this study have been described in detail previously (21).

Growth conditions. Bacteria were grown at either 32 or 37°C with rapid aeration in L broth (30). Agar plates made with T broth (30) were used to grow bacteria at 32°C.

DNA. DNA was prepared as previously described (37). DNA concentrations are reported as nucleotide phosphorous equivalents. For reference, 1 nmol nucleotide phosphorous of DNA is equivalent to 7.3 × 10⁹ T7 genomes. When 1.5 nmol of BstXI-digested donor DNA is added to 1.5 nmol of DNA in the form of T7 genomes, there is one molecule of the relevant restriction fragment for every genome. Restriction enzymes were purchased from New England Biolabs and used according to the supplier's instructions. For double-strand break repair reactions, breaks were placed in the T7 ligase gene by digestion with a restriction enzyme (usually BamHI or XhoI). The donor DNA used to repair the breaks was T7 6 ss DNA digested with BstXI. The ss mutation in the donor DNA refers to "suicide in Shigella" and confers upon phage that carry that mutation the ability to grow equally well on E. coli or Shigella sonnei. The presence of the ss mutation is incidental to the experiments reported here. Digestion with BstXI cuts the T7 genome into 12 fragments and effectively precludes reassembly into a functional genome. All restriction digests were checked by agarose gel electrophoresis. Radioactively labeled [³²P]dCTP was purchased from ICN. The radioactive precursor was included in DNA repair reactions, portions of the reaction mix were dried on a Whatman GF/C filter, and the radioactivity on the filter was measured with a liquid scintillation counter to allow a determination of the number of ³²P counts per minute that corresponds to a picomole of nucleotide. Product DNA from the in vitro reactions was precipitated with 10% trichloroacetic acid (TCA) including 0.1 M PP_i. The acid-precipitated DNA was collected on Whatman GF/C filters, and the amount of radioactivity was determined with a liquid scintillation counter.

In vitro repair. Extracts used for in vitro repair reactions were prepared by using phage that had genes 1.3 to 1.5 removed with the A deletion. This means that ligase-positive phage could not be produced via recombination with endogenous DNA in the extracts used for repair or packaging. Moreover, since the ligase gene was the site of double-strand breaks, the presence of the A mutation avoided homology between the break site and any endogenous DNA that might remain in the extracts, thereby reducing concern that endogenous DNA might contribute to repair reactions. The A mutation also reduces, by about 1.4 kb, the length of the 3.6-kb KpnI D fragment which in the wild type extends from position 5617 to 9192. This fragment was used to visualize repair of the double-strand breaks. The difference between A and wild-type KpnI bands allows easy distinction between exogenous DNA and any endogenous DNA that contaminates the reaction mixtures. Extracts for double-strand break repair were prepared with either T7 A 3 (lacking ligase and endonuclease), T7 A 3 5 (lacking ligase, endonuclease, and DNA polymerase), or T7 A 3 5 6 (lacking ligase, endonuclease, DNA polymerase, and 5'  3' exonuclease). The preparation of these extracts has been described in detail previously (10, 26, 29). Unless stated otherwise, 0.05-ml reaction mixtures included 1.5 nmol of DNA, consisting of broken or intact T7 genomes plus 1.5 nmol of BstXI-digested donor DNA. The 0.01-ml extract is the equivalent of protein extracted from 10⁹ phage-infected cells. Thus, a typical in vitro reaction begins with about 10¹⁰ genome equivalents of DNA and the amount of protein present in 10⁹ cells. This ratio of DNA to protein is a reasonable approximation of the early stage of a typical T7 infection. The in vitro reactions were provided with 0.3 mM concentrations of each of the four deoxynucleoside triphosphates. This amount is in excess of what is expected to be contributed endogenously by the extracts used for in vitro DNA repair. Details of the reactions are the same as used in earlier studies (14).

In vitro packaging. Extracts for in vitro packaging were prepared with T7 A 3 5 6, as previously described (18). DNA recovered from the in vitro repair reactions was diluted in an appropriate reaction buffer (14) and incubated at 32°C for 60 min with packaging extracts. The dilutions were such that 7.4% of the in vitro repair reaction volume was added to each packaging reaction. For a repair reaction mixture including 1.5 nmol of T7 DNA, the equivalent of 110 pmol of DNA was present in each packaging reaction. Repair reactions were carried out in triplicate. Values shown in the tables are averages of these determinations. This averaging procedure gives reproducible results and avoids errors that might be caused by an occasional anomalous packaging reaction.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

We compared the relative abilities of extracts made with T7 with amber mutations in genes 3, 5, and 6 and extracts made with T7 deficient in only gene 3 to repair double-strand breaks in a T7 genome. Gene 3 encodes an endonuclease and gene 6 encodes an exonuclease, both of which are essential for breakdown of host DNA to be used as a precursor for T7 DNA replication (47). Because deoxynucleoside triphosphates are provided as precursors during in vitro DNA reactions, neither of these proteins are essential for in vitro DNA replication. Because only very limited DNA synthesis goes on during in vivo preparation of phage-infected cells to be used for extracts in the in vitro system, the amount of endogenous DNA present in extracts from gene 3 mutants is low. This reduces contamination of the in vitro reactions with endogenous DNA and increases the dependence on exogenous DNA to be used as a substrate. Moreover, reducing gene 3 protein helps avoid spurious nuclease activity against the exogenous DNA in the in vitro reactions. The amount of in vitro DNA replication carried out by this system is essentially the same irrespective of whether wild-type, 3, or 3 6 phage are used for extract preparation (28). Our earlier studies (14, 21, 25, 55) had shown that the gene 3 protein is not necessary for efficient repair of double-strand breaks, and all experiments in the present study were performed by using T7 with gene 3 inactivated. The extracts made with T7 3 5 6 phage are similar to the ones used to package T7 DNA and, in addition to the endonuclease deficiency, are deficient in the T7 DNA polymerase encoded by gene 5 and the 5'  3' exonuclease encoded by gene 6 (50). Table 1 shows that with normal levels of DNA polymerase and exonuclease present, double-strand breaks in the T7 genome were repaired with nearly 20% efficiency. This repair depends on extracts from T7-infected E. coli and on donor DNA molecules that overlap the site of the break. Extracts made with T7 missing the phage DNA polymerase and the exonuclease encoded by gene 6 as well as the gene 3 endonuclease were more than 60-fold-less capable of dealing with the double-strand break (Table 1). The limited amount of double-strand break repair that was achieved without normal levels of the gene 5 or 6 product depends heavily upon the availability of donor DNA (Table 1). The lower yield of phage found by using intact DNA and extracts from T7 3 5 6 rather than from T7 3 was not unexpected. Previous studies have shown that the absence of gene 6 reduces the ability of the packaging system to produce viable phage (4). This may be due to a role for the gene 6 product in maturation of the T7 genomes prior to packaging (28). The lack of carryover of gene 6 product from the DNA repair reactions to the packaging reactions can account for the lower phage yield found when T7 3 5 6 were used as the source of extract.

To increase the sensitivity of repair reactions performed by using extracts made with T7 3 5 6, we supplemented these reactions with a small amount of extract made with T7 3 5. These reactions do not contain T7 DNA polymerase but, because of the low levels of gene 6 exonuclease added to the repair reactions, produce DNA that can be packaged very efficiently so as to generate large amounts of T7 phage. Table 2 shows that with intact T7 DNA, nearly the same phage yield was achieved whether an extract from T7 3 or combined extracts from T7 3 5 6 and T7 3 5 were used. The combined extracts were also able to repair double-strand breaks with an efficiency more than half that achieved with T7 3 extracts (Table 2). This result shows that double-strand breaks can be repaired without normal levels of T7 DNA polymerase but, because of the need for the extract from T7 3 5, also shows the need for the gene 6 exonuclease in double-strand break repair. The data in Table 2 also show that, in the absence of normal levels of T7 DNA polymerase, double-strand break repair requires donor DNA molecules and that information from these donor molecules is recombined into the repaired genomes.

To determine the extent to which the amber mutation in gene 5 had blocked DNA replication in this system, we measured the amount of DNA synthesis during in vitro reactions using both intact T7 genomes and genomes that had a double-strand break at the BamHI site. DNA repair reactions including [³²P]dCTP were performed. Samples were removed from the reactions at timed intervals, and the amount of acid-precipitable radioactive DNA was measured. Figure 2A shows that under the reaction conditions used to repair double-strand breaks, the amount of DNA synthesis carried out using extracts made with T7 3 5 6 was about 10% of what was measured under identical conditions using extracts made with T7 3. With extracts made from T7 3 5, DNA synthesis was higher but still much lower than what was found with the T7 3 extract. Given the small amount of synthesis measured, the small difference is of questionable significance. A determination of DNA synthesis was made both with T7 genomes with a double-strand break and with donor DNA fragments present in the reactions (Fig. 2B). The level of synthesis achieved by extracts made from T7 3 5 6 was much lower than that carried out with extract from T7 3-infected cells even when broken T7 genomes and donor DNA were both present in the reactions. The mutation in gene 5 causes a major reduction in DNA synthesis. The level of residual DNA synthesis was about the same irrespective of whether double-strand break repair-proficient extracts from T7 3 5-infected cells or double-strand break repair-deficient extracts from T7 3 5 6-infected cells were used. Also, little difference in the amount of DNA synthesis could be seen when broken and intact T7 genomes were compared, irrespective of a functional gene 5 product. Thus, when replication of the T7 genomes is blocked, some residual DNA synthesis persists whether or not repair takes place. Taken at face value, the data (Fig. 2) show approximately 40 pmol of difference between synthesis using T7 3 5 and that using T7 3 5 6 extracts. This amounts to about 1,000 bp of new synthesis per 40,000-bp genome. The source of this DNA synthesis is not known but may represent repair-like DNA synthesis needed to maintain the integrity of T7 genomes. There is no reason to believe that the residual synthesis seen with the gene 5 mutants represents repair-like DNA synthesis associated with the repair of double-strand breaks.

View larger version (16K): [in this window] [in a new window]   FIG. 2.   In vitro DNA synthesis. (A) In vitro reaction mixtures of 0.250 ml included a 0.050-ml extract from T7-infected E. coli, 7.5 nmol of DNA in the form of intact T7 genomes, and 7.5 nmol of BstXI-digested T7 6 ss DNA as the donor. Reaction mixtures, which included [32P]dCTP, were incubated at 37°C. At intervals, 0.050 ml of the reaction mixture was removed and 3 ml of 10% (wt/vol) TCA with 0.1 M PPi was added. The DNA was collected on Whatman GF/C filters, and the radioactivity was determined. The plot shows picomoles of total DNA present in a 0.050-ml sample. Extracts were prepared with T7 A 3 (), T7 A 3 5 (), or T7 A 3 5 6 (). (B) Identical experiment except that the wild-type T7X genomes were cut at position 6663 with XhoI. Symbols in panel B are the same as in panel A.

T7 genomes with a >1,600-nt gap were prepared with the substrate shown in Fig. 3. This DNA substrate has a 76-bp insert placed in the XhoI site of gene 1.3. However, because of a deletion between genes 1.3 and 1.7, the insert ends with a BclI site in gene 1.7. Effectively, a cut at one of the restriction sites within the inserts generates a >1,600-nt gap in the T7 genome. Because there are no repeats at its ends, the insert will not delete from the genome at detectable frequency and cannot grow on a lig-7(Ts) host (34, 35). Genes 1.3, 1.4, 1.5, 1.6, and 1.7 are non-essential (48), so T7 with this insert and deletion can grow normally on a wild-type E. coli host. The insert has five unique restriction sites which can be used to introduce double-strand breaks. Break sites in genomes cut with these restriction enzymes have no homology with donor DNA from wild-type T7 in the region between genes 1.3 and 1.7, so intact ligase-positive genomes can be formed only by filling in the gapped region between gene 1.3 and 1.7. This could be accomplished by a recombinational crossover between broken genome and the BstXI fragment left of position 6663 and a second crossover right of position 8311 in gene 1.7, or the donor DNA could be used as template for at least 1,600 nt of DNA synthesis. Previous experiments in which ligase-positive revertants of this gapped genome were sequenced showed that in all cases where intact genomes were formed after interaction with wild-type donor DNA, they took on the wild-type sequence of the donor (21). T7 genomes with the insert and deletion shown in Fig. 3 were cut with BamHI and then added to the in vitro reactions in place of the genomes with a simple double-strand break to see if they could be repaired sufficiently to generate viable T7 phage. Table 3 shows that gaps in a T7 genome are repaired efficiently in the absence of normal levels of T7 DNA polymerase. Under normal reaction conditions, with T7 DNA polymerase present, about 14% of the double-strand breaks were repaired and nearly 80% of these were derived from the donor DNA, as evidenced by their ligase-positive genotype (Table 3). When extracts missing T7 DNA polymerase were used, the amount of DNA repair was essentially unchanged from what was measured with normal levels of DNA synthesis. Again, in 75% of the cases, the repaired genomes became wild type for the T7 ligase gene.

View larger version (20K): [in this window] [in a new window]   FIG. 3.   T7 genome with a 1.6-kb gap. An insert of double-stranded DNA with the sequence shown in the upper part of the figure was placed between the XhoI site in gene 1.3 and the BclI site in gene 1.7. The sequence of the insert is shown in uppercase letters, while the surrounding T7 sequence is shown in lowercase letters. The insert has a unique BamHI site which, when cut effectively, produces a >1,600-nt gap in the T7 genome. The figure shows the generation of this gap next to the relevant BstXI fragment of donor DNA. The homology needed for recombinational repair of the gapped genome by any of the models outlined in Fig. 1 is available only in genes 1.3 and 1.7. The bottom line of the figure shows a repaired genome with either a portion of the donor DNA or the information from that donor DNA inserted into the gap.

The data presented thus far demonstrate that, as determined by production of genomes that can be packaged so as to produce infective phage, either a double-strand break or a 1,600-nt gap in the T7 genome can be repaired efficiently without the normal amount of T7 DNA polymerase. The high efficiency of repair prompted us to attempt visualization of double-strand break repair without resorting to in vitro packaging of the repaired genomes as a means of detecting successful repair events (Fig. 4). Genomes from T7X, wild-type T7 with a single XhoI site created at position 6663, were treated with XhoI to introduce double-strand breaks in the genomes. The broken DNA molecules were incubated in the in vitro repair reactions together with extracts made from E. coli infected with T7 3 5 6 supplemented with extracts made with T7 3 5. In one reaction, donor DNA, in the form of a BstXI digest of T7 genomes, was included, while the other reaction served as a control in which double-strand break repair was limited by the absence of donor DNA, as in Table 2. DNA from these reactions was extracted with phenol to remove proteins and then treated with RNase. The DNA was digested with KpnI and subjected to electrophoresis. Figure 4 shows a genetic map of bacteriophage T7 with the XhoI and KpnI sites marked. The KpnI D fragment extends from position 5617 to 9192 and covers the double-strand break site near 6663. Thus, a double-strand break obliterates this restriction fragment, while the restoration of a 3.6-kb fragment is diagnostic of repair of the break. Use of the KpnI digest also removes concern that endogenous DNA in the extracts might yield products which appear as repaired genomes. Since all extracts used in this study carry the A mutation, any endogenous DNA remaining in the reactions would not be able to produce a normal-sized KpnI D fragment. Furthermore, digesting the product DNA with KpnI avoided complications due to formation of concatemers of T7 genomes which form end to end during normal T7 maturation either in vivo (42) or in this in vitro system (28). The presence of concatemers frustrates detection of single genome-sized DNA molecules, so internal restriction fragments such as KpnI D become a much more reliable indicator of the physical integrity of the region near the double-strand break. Figure 4 shows that the KpnI D fragment was clearly detectable if donor DNA molecules were present. In numerous repeats of this experiment, the 3.6-kb band was detected after repair involving donor DNA molecules even though the 1.0- or 2.5-kb fragments expected after KpnI digestion of the unrepaired XhoI-digested DNA were barely detectable. The poor recovery of the smaller fragments and the BstXI fragments of donor DNA may be due to digestion of DNA near double-strand ends, as previously suggested (21). DNA not repaired early after introduction into the in vitro system appears to be at least partially degraded during the 30-min duration of the in vitro DNA repair reactions.

View larger version (35K): [in this window] [in a new window]   FIG. 4.   Visualization of repaired T7 genomes. The map of the T7 genome shows all KpnI recognition sites and the single XhoI site. In vitro repair reaction mixtures were either complete or, as a control, missing donor DNA. Both reactions included a 9:1 mixture of extracts from E. coli infected with T7 A 3 5 6 or T7 A 3 5. Products from the in vitro reactions were subjected to electrophoresis at 46 V for 2 h in a 0.4% agarose gel in 0.5× Tris-borate-EDTA buffer (40). The gel was stained with ethidium bromide to give the pattern shown. Lane M consists of a HindIII digest of lambda DNA as molecular weight markers. Lane 1 is missing donor DNA, and lane 2 shows the product of the complete reaction.

Experiments similar to those represented by Fig. 4 were performed by using extracts with normal levels of T7 DNA polymerase (T7 3). Figure 5 shows a comparison of a repair reaction performed under conditions of normal DNA replication (i.e., with an extract prepared with T7 A 3) with and without donor DNA. Again, a double-strand break was placed in the XhoI site of T7X DNA, and the DNA was incubated with (lane 2) or without (lane 1) BstXI fragments as donor DNA and then treated with KpnI and RNase prior to electrophoresis on an agarose gel. Figure 5 shows recovery of the normal-sized KpnI D fragment only when donor DNA was present. As an additional control, an experiment similar to that in Fig. 4 was carried out by using an EcoRII fragment as an indicator of double-strand break repair. EcoRII cuts the 39,937-bp T7 genome at two sites, 2365 and 8187, to produce fragments of 2,365, 5,822, and 31,750 bp long (5). A double-strand break at position 6663 cleaves the 5.8-kb fragment into 4.3- and 1.5-kb pieces. Thus, the presence of a normal-sized 5.8-kb fragment after EcoRII digestion is indicative of repair of the double-strand break. Figure 6 shows an in vitro repair experiment carried out with combined extracts made by using T7 A 3 5 6 and T7 A 3 5 and DNA that had a double-strand break at the XhoI site. The reaction shown in lane 2 included donor DNA in the form of BstXI fragments of the T7 genome and shows a pronounced band at the 5.8-kb position. This 5.8-kb band was not found in the reaction without donor DNA present (lane 1).

View larger version (66K): [in this window] [in a new window]   FIG. 5.   Visualization of double-strand break repair during normal DNA replication. In vitro repair reactions were performed as described in the legend to Fig. 4 except that an extract made with T7 A 3 was used instead of the extract made with a T7 DNA polymerase mutant. DNA recovered from the reactions was digested with KpnI and subjected to electrophoresis as in Fig. 4. Lane M consists of a HindIII digest of lambda DNA as molecular weight markers. Lane 1 is missing donor DNA, and lane 2 shows the product of the complete reaction.

View larger version (23K): [in this window] [in a new window]   FIG. 6.   Visualization of double-strand break repair with DNA cut by EcoRII. In vitro repair reactions were performed with a 9:1 mixture of extracts from E. coli infected with T7 A 3 5 6 or T7 A 3 5. DNA recovered from the reactions was digested with EcoRII and subjected to electrophoresis as in Fig. 4. Lane M consists of a HindIII digest of lambda DNA as molecular weight markers. Lane 1 shows a reaction without donor DNA, and lane 2 shows the product of a complete reaction.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

We have previously shown that recombination with a homologous DNA molecule represents the major avenue for repair of double-strand breaks in an in vitro system based on T7-infected E. coli (21). In this study, we consider the extent to which the donor DNA is replicated during double-strand break repair in T7. There are at least three possible mechanisms by which a segment of donor DNA could rescue a double-strand break or gap in the T7 genome. In one model, after the break is widened to a gap, 3' ends on each fragment of the broken genome invade the duplex donor DNA molecule. These ends then become the primers for DNA synthesis that copies information from the donor to the repaired recipient. The two invasion steps can occur in a coordinate fashion or could be independent. The latter case would produce single-strand tails extending from each of the broken genomes. Annealing between these complementary single-stranded tails, perhaps mediated by the T7 gene 2.5 product (15, 16), would rejoin the broken genome and transfer genetic information from the donor DNA to the repaired genome in a nonreciprocal fashion. This type of repair is essentially the same as the original double-strand break model of recombination (36, 51). It requires only a limited level of DNA synthesis to copy the template provided by the donor DNA. A second possible mechanism involves invasion of the broken strand ends into a homologous segment, followed by establishment of a replication fork. This mechanism provides a very attractive means for rescuing collapsed replication forks and would provide T7 with an economical way of recouping its investment in synthesis of partial genomes which would otherwise be lost (19). Establishment of new replication forks via recombination is a major factor in bacteriophage T4's infective cycle (32). Repair of double-strand breaks in T4 requires extensive DNA synthesis (8). There is compelling evidence that in E. coli broken genomes use the RecBC pathway of homologous recombination to establish new DNA replication forks and that this repair pathway is blocked in the absence of a major subunit of E. coli DNA polymerase III (20). In the experimental scheme employed in the present work, the donor DNA is a BstXI restriction fragment extending only from position 3870 to 9633. Therefore, rescue by this second mechanism would require two steps and establishment of a new replication fork at each end of the donor DNA molecule. This model requires substantial DNA replication. The amount of DNA replication depends on the order and position of invasion events and the direction of replication fork progression. Synthesis from the break site to the end of the T7 genome would require, on average, replication of half of the 40,000-bp genome. The third possibility is physical incorporation of the donor DNA into a gap on the genome. Presumably this would require crossovers between the broken genome and the donor DNA on either side of the gap. This model does not require extensive DNA synthesis, although a limited amount of new DNA may need to be manufactured at the ends where the crossovers take place. An earlier study is in accord with double-strand exchange between recombining T7 DNA segments (22). That study is not strictly compatible, since it was based on plasmid-genome recombination. A different study, also dealing with plasmid-genome recombination, favored incorporation of only a single strand of DNA from the plasmid to the genome and new synthesis of a complementary strand (45). Thus, the possibility of double-stranded crossovers between donor and recipient DNA as a means of repairing double-strand breaks appears to be an open question.

The data presented in this study show that repair of a double-strand break is very efficient in extracts made from T7-infected E. coli. The high repair efficiency (approximately 20% [Table 1]) is even more remarkable considering that under these experimental conditions essentially every T7 genome had a double-strand break and the protein-to-DNA ratio was chosen to match that of a phage-infected cell a few minutes after infection (10 genome equivalents per cell equivalent of protein). Thus, especially aggressive repair activity is required to handle this level of molecular catastrophe. It should be kept in mind, however, that the data we present here do not prove that rescue of a collapsed replication fork via the model shown in Fig. 1B does not take place in T7-infected E. coli. Also, there is always some concern that the in vitro system does not truly mimic the in vivo situation. We do claim that blocking such a repair pathway by inhibiting replication has only a small effect on the efficiency of double-strand break repair and that other, replication-independent, mechanisms are sufficient to handle double-strand break repair in this in vitro system. Our data are not in accord with establishment of new replication forks at the break site as a primary repair mechanism. Single-strand annealing (Fig. 1A) remains a viable alternative but, if it occurs, must involve only a limited level of DNA synthesis, possibly by one of the E. coli DNA polymerases. The third model, double-stranded crossovers between donor and recipient DNA, is easiest to reconcile with the lack of dependence on T7 DNA polymerase.

Extracts made with T7 having an amber mutation in gene 5 show markedly lower rates of DNA synthesis (Fig. 2). In the absence of normal levels of gene 5 product, the residual DNA synthesis is the same whether broken or intact DNA is used for the substrate. Extracts mutant in genes 3, 5, and 6 are unable to perform double-strand break repair but show about the same rate of DNA synthesis as extracts made from T7 with only genes 3 and 5 inactivated. These observations suggest that the residual DNA synthesis seen with the gene 5 mutants is not due to DNA synthesis that is part of the double-strand break repair. Although the residual levels of DNA replication are low, they have been detected reproducibly. We suggest that perhaps nicks appear in the DNA as the result of DNA damage or as part of normal metabolism. Repair-like DNA synthesis at these nicks, perhaps by E. coli DNA polymerase I, leads to this background level of DNA synthesis. Initial efforts to produce infective T7 genomes in this system showed dependence on host DNA polymerase I in order to generate full-length genomes (29). The repair synthesis level may be exaggerated in the experiments presented here because inactivation of T7 ligase by the A mutation makes the phage DNA metabolism entirely dependent on the host ligase, thereby reducing the capacity to close transient interruptions in the DNA and inviting increased activity of DNA polymerase I.

The high efficiency of double-strand break repair in this system allowed visualization of KpnI restriction fragment D, which is normally found with intact T7 genomes but which breaks into smaller fragments when a double-strand break is placed in the XhoI site (Fig. 4). Visualization of the repair product was achieved both with and without normal levels of T7 DNA polymerase. Visualization of the repaired genome is a less quantitative determinant of DNA repair than packaging of DNA to produce viable phage, as in Tables 1 through 3. While it is clear from the relative intensities of the C and D bands in the right lane of Fig. 5 that substantial repair of the double-strand break took place, this estimate of repair efficiency is crude relative to the quantitative data recovered after packaging. However, the visualization technique has the advantage of decoupling the repair reaction from the packaging reaction and provides evidence that repair of the double-strand breaks (without normal levels of T7 DNA polymerase) takes place during the in vitro reactions rather than either during the packaging step or during the in vivo growth step that follows packaging. Techniques based on detection of recombinant products after electrophoresis and hybridization with appropriate probes, rather than on the phenotype of phage produced after packaging, provide the most convincing evidence for in vitro recombination of T7 DNA (22). Similarly, visualization of a repaired product DNA complements the packaging technique as a way of monitoring double-strand break repair in the T7 system.