Recent findings suggest that DNA nicks stimulate homologous recombination by being converted into double-strand breaks, which are mended by RecA-catalysed recombinational repair and are lethal if not repaired. Hyper-rec mutants, in which DNA nicks become detectable, are synthetic-lethal with recA inactivation, substantiating the idea. Escherichia coli dut mutants are the only known hyper-recs in which presumed nicks in DNA do not cause inviability with recA, suggesting that nicks stimulate homologous recombination directly. Here, we show that dut recA mutants are synthetic-lethal; specifically, dut mutants depend on the RecBC–RuvABC recombinational repair pathway that mends double-strand DNA breaks. Although induced for SOS, dut mutants are not rescued by full SOS induction if RecA is not available, suggesting that recombinational rather than regulatory functions of RecA are needed for their viability. We also detected chromosomal fragmentation in dut rec mutants, indicating double-strand DNA breaks. Both the synthetic lethality and chromosomal fragmentation of dut rec mutants are suppressed by preventing uracil excision via inactivation of uracil DNA-glycosylase or by preventing dUTP production via inactivation of dCTP deaminase. We suggest that nicks become substrates for recombinational repair after being converted into double-strand DNA breaks.
Homologous recombination is known for its role in evolution (reduction in linkage disequilibrium), as well as in the repair of double-strand DNA breaks. If homologous recombination plays such a prominent role in evolution, what would be the advantage (or disadvantage) of having more frequent recombination? To look for mutants with increased recombination in Escherichia coli, Konrad (1977) constructed a genetic system that detects homologous exchange between inverted copies of two lacZ genes carrying non-allelic deletions. The nature of hyper-rec mutants was presumed to be either overproduction of recombination enzymes or increase in recombination-stimulating events in DNA (Konrad, 1977). Remarkably, all hyper-rec mutants isolated have the regulation of recombination enzymes untouched, but have shown various defects in DNA metabolism, suggesting that it is irregularities in their chromosomal DNA that stimulate recombination. Among the hyper-rec mutants that Konrad identified were alleles of polA (DNA polymerase I), lig (DNA ligase), dam (DNA-adenine methylase) and sof[later shown to be dut (dUTPase); Tye et al., 1977] (Konrad, 1977). Other researchers, using the Konrad system, found that mutants in rdgB (ITPase) (Clyman and Cunningham, 1987), uvrD (DNA helicase II) and xthA (abasic site endonuclease) (Zieg et al., 1978) are also hyper-rec.
A single-strand interruption in DNA duplex (nick) was suggested to be a common denominator for the hyper-rec phenotype (Konrad, 1977). Indeed, nicks are detectable in polA (Nakayama and Hanawalt, 1975), lig (Pauling et al., 1976) and dam (Marinus and Morris, 1974) mutants as lowering of molecular weight of the chromosomal DNA in alkaline sucrose gradients. Nicks were also expected in DNA of dut mutants (because of more frequent uracil incorporation into DNA and its subsequent excision) and in uvrD mutants (on account of the inability to complete excision repair). Stimulation of homologous recombination by DNA nicks, whether introduced by the tra functions of F-plasmid (Carter et al., 1992), by the filamentous phage-nicking function (Strathern et al., 1991), by an epigenetic modification during the mating-type switching (Arcangioli, 2000) or induced by a eukaryotic Topo I inhibitor camptothecin (Nitiss and Wang, 1988), is documented. According to the Meselson–Radding model, nicks stimulate homologous recombination by priming strand displacement DNA replication, with the displaced single-strand tails engaging homologous chromosomes in strand exchange (Fig. 1A–E) (Meselson and Radding, 1975). Thus, hyper-recombination is explained by additional, but otherwise innocuous, initiation events in DNA.
Somewhat unexpectedly, four out of the seven hyper-rec mutants –polA (Monk and Kinross, 1972), lig (Gottesman et al., 1973a), dam (Marinus and Morris, 1974) and rdgB (Clyman and Cunningham, 1987) – also turned out to be synthetic-lethal in combination with recA inactivation. In other words, not only were these particular hyper-rec mutants engaged in more frequent recombination, but they were also dependent on it for survival. This ‘addiction’ to homologous exchange suggested that nicks in DNA of hyper-rec mutants are somehow toxic for the cell, and the cell uses recombinational repair to counteract the harmful consequences of these nicks. Single-strand DNA interruptions can poison the cells if they are converted into double-strand breaks by DNA replication (Fig. 1F and G); hence, hyper-rec mutants were proposed to suffer from chromosome fragmentation and therefore rely on recombinational repair (Fig. 1G–J) (Kuzminov, 1995). However, uvrD, xthA and dut mutants, although strongly hyper-rec, are viable in combination with recA mutations (Smirnov et al., 1973; Konrad and Lehman, 1975; Clyman and Cunningham, 1987), suggesting a distinct mechanism for their elevated recombination. Moreover, no DNA nicks are detectable in DNA of uvrD mutants (Sinzinis et al., 1973), while xthA mutants, deficient in the major nicking activity that attacks abasic sites, are expected to have a reduced number of nicks in DNA. Thus, the subdivision of the hyper-rec mutants into either ‘nick/break’ (RecA-dependent ones) or ‘no-nick’ (RecA-independent ones) is almost perfect, with the exception of dut mutants, for which there is a strong indication of elevated nicks in the chromosome, but no associated RecA dependence. An economical explanation is that hyper-recombination in dut mutants proceeds via direct utilization of nicks in homologous strand exchange (Fig. 1A–E), without nicks first being converted to double-strand DNA breaks (Fig. 1F–J).
In E. coli, deoxyuridine triphosphatase (Dut) hydrolyses dUTP to generate dUMP for the de novo synthesis of dTTP and to prevent uracil incorporation into DNA in place of thymine (Neuhard and Kelln, 1996). It has been known that dut mutants of E. coli transiently accumulate nascent DNA fragments four or five times shorter than the regular Okazaki fragments resulting from excision repair of incorporated uracils (Konrad and Lehman, 1975; Tye and Lehman, 1977; Tye et al., 1977). Introduction of the ung (uracil-DNA-glycosylase) or dcd (dCTP deaminase) mutations in dut mutants almost completely suppresses the appearance of short Okazaki fragments, whereas introduction of the non-lethal lig or polA mutations results in persistence of the short Okazaki fragments, cessation of DNA synthesis and conditional lethality (Tye and Lehman, 1977). It was suggested that, in dut polA and dut lig mutants, the observed nicks or gaps in DNA, produced as a consequence of attempted uracil excision by the uracil DNA glycosylase and abasic site endonuclease, remain unsealed because of the defective DNA polymerase I or DNA ligase, accumulate and destroy the replication fork (Tye and Lehman, 1977). The findings that dut xthA strains are conditionally inviable, and suppression of this inviability by the ung mutations, indicated that exonuclease III (gpxthA), the major abasic site endonuclease, works downstream of the Ung protein (Taylor and Weiss, 1982).
The ‘hyper-rec’ nature of the dut mutants was explained as a probable consequence of transient nicks resulting from uracil excision in the newly synthesized DNA (Hochhauser and Weiss, 1978). As outlined above, the original idea of nick-stimulated hyper-recombination posits that nicks (i) stimulate homologous recombination directly, for example via the Meselson–Radding model (Fig. 1A–E), without being converted into double-strand breaks; (ii) yet do not need recombination for repair. In agreement with this idea, dut mutants were found to be viable in combination with recA inactivation (Konrad and Lehman, 1975; El-Hajj et al., 1992). However, direct physical evidence is now available indicating that DNA replication converts nicks into double-strand breaks (Strumberg et al., 2000; Kuzminov, 2001), which are lethal if not repaired. Therefore, hyper-recombination in polA, lig or dam mutants, synthetic-lethal with recA, could be explained by occasional conversion of DNA nicks into double-strand DNA breaks with subsequent repair of these double-strand breaks by homologous recombination. To see which of the two ideas better accounts for the situation in dut mutants, we sought answers to two questions: (i) do dut mutants depend on recombinational repair? and (ii) are there detectable double-strand breaks in the chromosomal DNA of dut mutants?
The nature of dut-1 and dut-11 alleles
As inactivation of the dut gene kills E. coli cells (El-Hajj et al., 1988), we chose to work with the dut-1 and dut-11 mutations as showing the strongest defect in dUTPase activity (Hochhauser and Weiss, 1978). To learn about the nature of the two dut alleles, we have sequenced them and found that dut-1 carries a C-to-T transition at nucleotide 74 of the dut open reading frame (ORF), which results in the threonine-to-isoleucine replacement at position 25 of the 152-amino-acid Dut protein (Fig. 2A). This position is occupied by either threonine or serine in 90% of the known 102 dut genes (aligned with clustalw) and is only four amino acids away from the first of the five conserved dUTPase motifs (Larsson et al., 1996). Sequencing of dut-11 (originally known as dnaS1 or sof-1; Konrad and Lehman, 1975; Hochhauser and Weiss, 1978) revealed a G-to-A transition at nucleotide 440 of the dut ORF, which results in replacement of a conserved glycine by aspartic acid at position 147 in the middle of motif 5 (Fig. 2A). Motif 5 is a portion of the Gly-rich P-loop, disordered in the crystal structure of the protein and suggested to be involved in alpha–beta phosphodiester bond cleavage of dUTP, forming a ‘cover’ for the active site (Fig. 2B) (Cedergren-Zeppezauer et al., 1992), hence the deficiency in dUTPase activity of the dut-11 mutation.
We measured the dUTPase activity in crude extracts of Dut+, dut-1 and dut-11 strains at 25°C and 42°C and estimated the dUTPase activity at 2.2–3.3% of the wild type in dut-1 mutants and at 4.8–11.1% in dut-11 mutants (Fig. 2C), generally corroborating earlier data from others (Tye et al., 1977; Hochhauser and Weiss, 1978); however, we did not see the heat sensitivity of the Dut-1 protein. Another way to quantify the dUTPase defect is to prevent uracil excision from DNA of dut mutants by inactivating uracil-DNA-glycosylase Ung and to compare the plating efficiency of phage isolated from such dut ung double mutants on wild-type (WT) (uracil-excising) versus ung (uracil-permissive) cells (Fig. 2D). As a negative control, we show that λ phage grown on WT or ung-deficient (dut+) cells plates equally well on both WT or ung-deficient hosts independently of the temperature used to grow the phage (Fig. 2D). In contrast, λ grown on dut ung mutants plates much worse on the uracil non-permissive host: the decrease in plating efficiency is about 47 for the dut-11 allele and about 457 for the dut-1 allele if the phage was grown at 42°C (Fig. 2D). Our measurements confirm that: (i) dut mutants accumulate uracil in DNA in the absence of uracil DNA-glycosylase; and (ii) the dut-1 mutant has less dUTPase activity than the dut-11 mutant. It should be noted that, despite dut being an essential gene, both dut-1 and dut-11 mutants are fully viable, although they grow somewhat more slowly than their wild-type counterparts (Fig. 3).
dut-1 and dut-11 mutants are synthetically lethal with recA mutants
To see whether dut recA double mutants are synthetically lethal, we constructed a dut-1 recA200(Ts) strain and compared its viability at 27°C, 37°C and 42°C with the viability of the dut-1 or recA200(Ts) single mutant strains. We found that the dut-1 recA200(Ts) strain did not grow at 42°C (Fig. 3A, row 10). As we were unsuccessful at the introduction of a ΔrecA mutation into the dut-1 mutant strain, we constructed dut-1ΔrecA combinations in strains that carry plasmids with a temperature-sensitive replicon, expressing either recA+ or dut+ genes (Table 1, strains L-11, L-48 and L-49). In these strains, loss of the plasmids at 42°C correlated with the inability of the cells to form colonies (not shown). Thus, in both BW13711 and DH5αF′ backgrounds, combinations of the dut-1 mutation with recA mutations were also synthetically lethal (Table 1, strains L-48 and L-49; data not shown), excluding the possibility that some unspecified mutation in our principal AB1157 background contributes to the synthetic lethality of the dut recA combination.
Table 1. . The list of viable versus inviable combinations with dut mutants found in this work, as well as previously reported inviable combinations.
Known inviable combinations
The tested strains in first column for the dut-1 allele are: L-5, L-6 and L-15, L-7, L-16, L-17; for the dut-11 allele: L-89, L-94, L-100, L-97; the second column for the dut-1 allele: AK106, L-11, L-48 and L-49, AK107, L-47 and L-50, L-10, L-58, and L-14; for the dut-11 allele: L-95.
, band c. The plasmids pEAK2, pJBS2 and pEAK12-4 express wild-type recA, ruvABC or dut genes respectively. These plasmids are lost at 42°C (0.1% of cells keep the plasmid), revealing synthetic lethality of the strains.
It was reported that dut-11 recA mutant is viable (Konrad and Lehman, 1975). We tested dut-11 mutation in combination with recA200(Ts) allele and found that the strain was indeed viable at 42°C (data not shown). As recA200(Ts) allele is not completely deficient in recombinational repair, we constructed dut-11ΔrecA strains in the presence of plasmids carrying either dut + or recA+ genes (Fig. 3B; data not shown). Unlike the dut-1ΔrecA strains, these strains continue to form colonies even when the plasmids are lost, implying that dut-11ΔrecA combination is viable (Fig. 3B, row 4 at 42°C). However, subsequent restreaking of the dut-11ΔrecA colonies that have lost the plasmids reveals only dead cells (Fig. 3B, right). We conclude that (i) dut-11ΔrecA double mutants are synthetically lethal although, unlike dut-1 recA combinations, they show delayed lethality; and (ii) the dut-11 allele is less defective in dUTPase activity compared with the dut-1 allele, which agrees with other assays (Fig. 2C and D). In most of our subsequent experiments, we used dut-1 allele as the one with the tightest dUTPase defect.
dut-1 mutants do not depend on the SOS response
RecA protein has two major functions: (i) it catalyses homologous strand exchange; and (ii) it induces the SOS response, an increased expression of some 30 genes (among them recA itself) in reaction to DNA damage, through the cleavage of the SOS repressor, the LexA protein (Friedberg et al., 1995). Within the SOS response, RecA has yet another function, as it helps directly in translesion DNA synthesis (Tang et al., 1998). As dut mutants were reported to be ‘hyper-rec’ in the recombination assay (Konrad, 1977) and were slow to close ssDNA interruptions in newly replicated DNA (Tye and Lehman, 1977), we tested the dut-1 strain for SOS induction. To this end, we transformed wild-type and dut-1 strains with the plasmid pEAK7 carrying the tet gene under the SOS-inducible umuDC promoter. pEAK7 confers resistance to tetracycline only to SOS-induced cells (Fig. 4A, rows 5 and 6 versus 1 and 4 on the 42°C × 22 h tetracycline plate). We found that, when grown at 42°C, the dut-1 pEAK7 strain is resistant to 12 µg ml−1 tetracycline, whereas its isogenic control strain is not (Fig. 4A, row 2 versus 1 on the 42°C × 22 h tetracycline plate), revealing a degree of SOS induction in the dut-1 mutant.
As dut-1 cells are SOS induced, the synthetic lethality of dut recA strains could result from their inability to induce the SOS response, their recombinational repair deficiency or a defect in translesion DNA synthesis. To distinguish between the three possibilities, we built a recA200(Ts) dut-1 derivative, in which the SOS response is constitutively expressed on account of a lexA deletion. To introduce a deletion of lexA, we had to delete the sulA gene first, which encodes the SOS-inducible inhibitor of cell division (Lutkenhaus, 1983). We verified the induction of the SOS response in the resulting ΔlexA strain as the tetracycline resistance conferred by pEAK7. When grown on LB plates, the ΔlexA strain is resistant to 12 µg ml−1 tetracycline, whereas an isogenic control strain is inhibited by 3 µg ml−1 tetracycline (Fig. 4A, rows 4 versus 5 on the 42°C × 54 h tetracycline plate; data not shown). We found that, although the SOS genes must be constitutively expressed in the ΔlexA dut-1 recA(Ts) strain, the strain shows the same growth defect at 42°C as an isogenic dut-1 recA(Ts) strain (Fig. 4B, rows 3 and 4). We conclude that RecA itself is required for the viability of dut mutants; it may either act as a recombinational repair protein or help with translesion DNA synthesis. To see whether the SOS-elevated expression of the recA+ gene is needed for viability of dut mutants, we attempted to construct a dut-1 lexA3 mutant (the uncleavable LexA3 protein is unable to increase recA expression via SOS) and found it to be viable (Table 1). Therefore, neither SOS-induced levels of RecA, nor some other SOS-induced proteins, are required for the viability of dut mutants. The viability of the dut-1 lexA3 mutant also argues against translesion DNA synthesis (unavailable in the absence of SOS induction in lexA3 cells) as the reason for the RecA requirement in dut mutants. From this genetic analysis, we conclude that the recA dependence of dut-1 mutants results solely from their requirement for recombinational repair.
dut-1 mutants depend on the RecBC pathway of recombinational repair
In wild-type E. coli, there are two pathways of recombinational repair, both dependent on the recA gene. The RecBC pathway deals with the double-strand breaks, whereas the RecF pathway repairs daughter-strand gaps (Kuzminov, 1999). To determine which of the two kinds of two-strand DNA lesions are produced in dut mutants, we introduced the dut-1 mutation into various mutants defective in homologous recombination. We found that the dut-1 recBC(Ts) strain is inviable at 37°C (Fig. 3A, left, row 6 versus 12), whereas the ΔrecBCD dut-1 p(ori-Ts)Dut+ strain is inviable at 42°C upon plasmid loss (Table 1). However, we were able to recover the resulting barely viable ΔrecBCD dut-1 mutant at 27°C, its viability confirming the temperature sensitivity of the dut-1 allele (Hochhauser and Weiss, 1978). RecBCD protein works by processing double-strand DNA ends and by loading RecA on the processed ends. RecBCD is thus critical for the recombinational repair of double-strand breaks through its exonuclease (ExoV) and helicase activities. The lethality of the dut-1 recBC(Ts) strain is not the result of inactivation of ExoV activity, as dut-1 recD strains are viable (Table 1). RecJ nuclease is known to be important for recombinational repair of chromosomal double-strand breaks in the absence of ExoV (Lloyd et al., 1988; Lovett et al., 1988); in agreement, we found that the dut-1 recD recJ strain could only be built in the presence of a plasmid carrying the wild-type dut gene and became barely viable upon plasmid loss (Fig. 5A, row 4).
We could build the dut-1ΔruvABC strain only in the presence of plasmid-borne ruvABC+ or dut+ genes. Upon plasmid loss at 42°C, the dut-1ΔruvABC strain became inviable (Fig. 5A, row 2). Another inviable combination was dut-1 polA4113 (Table 1). We also tested the viability of the dut-11 mutant in combination with the ΔrecBCD and ΔruvABC alleles. We found that, unlike the dut-1 allele, the dut-11 allele appears to be fully viable in combination with both ΔrecBCD and ΔruvABC after plasmid loss (Fig. 5B, rows 2 and 4 at 42°C), but then the double mutant strains grow slowly upon subsequent propagation at 30°C (Fig. 5B, right). We attribute the observed difference in behaviour between the dut-1 and dut-11 alleles to the difference in levels of dUTPase activity in the two mutants (Fig. 2C and D).
In contrast to the recBC or ruv combinations, combinations recG dut-1, recO dut-1 and recF dut-1 are viable (Table 1 and Fig. 5A, row 6). Interestingly, although viable, recO dut-1 and recF dut-1 strains grow more slowly at 30°C than the corresponding single mutants and produce elongated cells, with a significant fraction of very long filaments. Yet, the viability of the recF dut-1 and recO dut-1 strains at 37°C is close to the viability of their dut+recF or recO progenitors (70–90%; Lloyd et al., 1987a; Miranda and Kuzminov, 2003; data not shown). Moreover, the viability of dut-1 recF seems to be worse at 27°C than at 42°C (Fig. 5A, row 6), probably because of the alleviation of the recF defect at high temperatures (Ganesan et al., 1988). Synthetic lethality of dut-1 mutants in combination with recBC and recD recJ mutations and their independence of the recF and recO genes suggests that two-strand DNA lesions in dut mutants are predominantly of the double-strand end type, rather than of the daughter-strand gap type.
Chromosomal fragmentation in dut-1 mutants
As dut-1 mutants depend on the double-strand break-repair enzymes, they must suffer double-strand breaks in their chromosome. A few double-strand breaks in the chromosome will generate subchromosomal fragments that can be separated from the intact circular chromosome by pulsed field gel electrophoresis (PFGE). To detect double-strand breaks in the chromosomal DNA of dut mutants, we grew the dut-1 recBC(Ts) strain at both permissive and non-permissive temperatures, isolated chromosomal DNA in agarose plugs and ran it in a pulsed field gel. In pulsed field gels, circular chromosomal DNA stays at the origin (in the plug) (Fig. 6, ‘–’ lanes); if the chromosome is fragmented by double-strand breaks, linear chromosomal fragments migrate into the gel, forming a smear. Double-strand breaks cannot be repaired in the recBC(Ts) mutant at the non-permissive temperature; the fragmented chromosome accumulates because the main linear DNA degradation activity in E. coli is also inactivated by the recBC defect. The pulsed field gels show that the chromosome of the dut-1 recBC(Ts) strain is intact at 22°C, but becomes fragmented at 37°C: it migrates into the gel as a smear and also accumulates in a compression zone, typical of PFGE, with molecular weight of 1500–2000 kbp (Fig. 6, lanes 8 versus 9 and 15 versus 16). Control single mutant strains dut-1 and recBC(Ts) show little chromosomal fragmentation at either temperature (Fig. 6, lanes 1–4). We conclude that, in dut mutants, chromosomal DNA suffers double-strand breaks, which are revealed if the double-strand break-repair and linear DNA degradation are disabled as a result of the recBC mutation.
ung, dcd or pst inactivation suppresses dut rec lethality via decreasing chromosomal fragmentation
Chromosomal abnormalities in dut mutants could result from thymine-less death, caused by the decreased production of dUMP and, consequently, of the DNA precursor dTTP. In many organisms in conditions of low dTTP, chromosomal replication continues, but the chromosome becomes progressively inactivated; recombinational repair plays an unspecified role in both chromosomal inactivation and chromosomal repair (Ahmad et al., 1998). Published data on the thymidine auxotrophy of dut mutants are contradictory (Hochhauser and Weiss, 1978; Taylor and Weiss, 1982; El-Hajj et al., 1992); in our hands, dut-1 strains in AB1157 or R594 backgrounds are not thymidine auxotrophs, whereas those in a BW13711 background show a temperature-sensitive requirement for thymidine in minimal medium, but not in rich medium. If the dTTP pool depletion is the cause of lethality, supplementing media with thymidine should suppress the synthetic lethality of dut rec mutants. However, adding up to 150 µg ml−1 thymidine to either rich or minimal media did not rescue the dut-1 recA(Ts) or dut-1 recBC(Ts) lethality at 42°C in the AB1157, R594 or BW13711 backgrounds (data not shown). The AB1157 background may be impaired in thymidine uptake because of a tsx mutation (Benz et al., 1988); however, we were able to stimulate growth of AB1157 Δdcd and Δdcd dut-1 strains in M9 media by adding thymidine (data not shown), suggesting that the strains of the AB1157 background can take in some thymidine.
If double-strand DNA breaks in dut mutants are caused by uracil incorporation, then mutations in genes that either participate in the production of dUTP or begin the excision of uracil from DNA should suppress the dut rec lethality. One potential suppressor is dcd, the gene coding for dCTP deaminase, the enzyme that produces 75% of dUTP, the major precursor for dUMP that is used in the dTTP synthesis (Neuhard and Kelln, 1996). Inactivation of uracil DNA-glycosylase (Ung) should also suppress dut rec synthetic lethality via elimination of the source of abasic sites and single-strand interruptions in DNA.
We looked for inactivational suppressors of the dut recA(Ts) or dut recBC(Ts) lethality at non-permissive temperatures after insertional mutagenesis. As the AB1157 background is deleted for the Rac prophage (Kaiser and Murray, 1980), we did not isolate any sbcA suppressors of the recBC defect (Gillen and Clark, 1974). Insertions in three genes, pstS, pstC and dcd, allowed slow growth of dut rec(Ts) mutants at the non-permissive temperatures (Fig. 7A, rows 4, 5, 9 and 10; Table 2). pst mutations, inactivating one of the two phosphorus transporters in E. coli (Wanner, 1996), suppress the dut rec synthetic lethality via yet to be identified mechanism(s). If there is a single mechanism of pst suppression, it is unlikely to result from the expected constitutive activation of the pho regulon (increased expression of several unlinked genes in response to phosphorus starvation) in pst mutants (Wanner, 1996) because, although the pstS337 or pstC602 mutations do activate the pho regulon, the pstS45 suppressor does not (Fig. 7C, sectors 6, 8 and 9 on the ‘high phosphate’ plates). The common mechanism is also unlikely to be via production of an alternative dUTPase activity because, of the three pst suppressors, only pstC602 exhibits such an activity (Fig. 7B), whereas extracts from both pstS suppressors cannot hydrolyse dUTP (data not shown). In contrast to the genuine dUTPase, the alternative dUTPase activity from the pstC602 suppressor also hydrolyses dUDP (Fig. 7B). As the sequencing reveals no changes to the original dut-1 allele in the pstC602 suppressor, the nature of this non-specific dUTPase activity remains to be determined.
Table 2. . Suppressors of dut rec lethality.
Allele numbers indicate either the position of the pRL27 insert in the DNA sequence of the corresponding gene or the length of deletion in nucleotides.
. This allele was recovered in the strain carrying an additional nfi mutation.
. The mutations were introduced directly by phage P1-mediated transduction, rather than found in the suppressor screen.
Because we did not find the expected ung suppressors by selection after insertional mutagenesis, we considered the possibility that EndoV endonuclease, which attacks hypoxanthine- and xanthine-containing DNA (Yao et al., 1994; He et al., 2000), but is also known to attack uracil-containing DNA (Gates and Linn, 1977), contributes to the formation of DNA lesions in the absence of the Ung uracil-DNA-glycosylase. We repeated the mutagenesis in the dut-1 rec (Ts) strains carrying additional mutations in the nfi gene, which encodes EndoV, but again failed to isolate ung suppressors, although we did isolate one more dcd suppressor (Table 2). However, when we directly introduced the ung152 allele into ΔrecA dut-1 pEAK12-4 (dut +) and ΔrecBCD dut-1 pEAK12-4 (dut +) strains, the dut ung rec triple mutants easily lost the plasmids and grew normally at 42°C (data not shown). A similar situation was described for dut polA and dut xthA synthetic lethals, which were both found to be suppressible by inactivation of uracil-DNA-glycosylase, yet ung mutants failed to show up in the corresponding suppressor screens (Tye and Lehman, 1977; Taylor and Weiss, 1982). We also built the Δung dut-1 recA(Ts) and Δung dut-1 recBC(Ts) strains; these strains grow at 42°C (Fig. 7A, rows 1 and 6).
To gain insights into the mechanism of suppression of dut rec lethality, we checked the status of chromosomal fragmentation in the suppressor mutants. In dut recBC(Ts) strains carrying these suppressors, there is a substantial decrease in chromosome fragmentation, especially noticeable in the compression zone (Fig. 6, lanes 9 versus 11 and lanes 16 versus 18, 20 and 22). The introduction of the Δung mutation also suppressed the SOS induction phenotype in a dut-1 strain (Fig. 4A, panel 42°C × 22 h, row 3 versus 2). We conclude that the increased uracil incorporation in DNA of dut mutants with subsequent removal by uracil DNA-glycosylase causes chromosomal fragmentation and SOS induction.
dut mutants incorporate uracil in their chromosomal DNA; subsequent uracil excision by uracil-DNA glycosylase was postulated to lead to accumulation of nicks in the DNA of dut mutants, resulting in hyper-recombination. We compared two possible scenarios leading to hyper-recombination: (i) nicks stimulate homologous recombination directly, for example via the Meselson–Radding model (Fig. 1A–E); and (ii) nicks are converted into double-strand breaks, which require recombinational repair (Fig. 1F–J). The ‘direct stimulation’ idea predicts that (i) dut mutants are independent of recA (Konrad and Lehman, 1975; El-Hajj et al., 1992); (ii) there are no double-strand breaks in the DNA of dut mutant cells. In contrast, the ‘double-strand break’ idea predicts (a) recA dependence of dut mutants; and (b) double-strand breaks in DNA of dut mutants. Our results confirm the ‘double-strand break’ scenario for dut mutants and unite them with other hyper-recs (polA, lig, dam and rdgB) that have or are expected to have increased levels of DNA nicks and are dependent on recA and recBC for viability (Kuzminov, 1995). It should be pointed out, however, that our results make unlikely, yet do not rule out, direct stimulation of recombination by nicks, which may still happen in addition to recombinational repair of double-strand breaks.
Besides recA dependence, we found that, although there is an induced SOS response in dut strains, their viability depends on the recombinational function of RecA, rather than on functions of other SOS-induced genes. We also found that the viability of dut mutants depends on the recBC and ruvABC genes. The early stage of recombinational repair is promoted either by the RecBC pathway, acting on double-strand breaks, or by the RecFOR pathway, acting on daughter-strand gaps. The post-synaptic phase of recombinational repair deals with the intermediates of homologous recombination and depends on either RuvABC or RecG. Recently, it was found that, during repair of two-strand DNA lesions after UV irradiation, the RuvABC post-synaptic pathway works with the RecBC presynaptic pathway (double recBC ruv mutants have the UV sensitivity phenotype of single mutants; Bolt and Lloyd, 2002); however, the same is not true for conjugational recombination (Lloyd et al., 1987b). In agreement with this finding, our results with dut mutants implicate the RuvABC resolvasome, but not the RecG helicase, in the repair of two-strand DNA lesions along the RecBC pathway. At the same time, dut mutants in combination with mutations in the RecF recombinational repair pathway are viable. The results of genetic analysis suggest that dut mutants suffer double-strand DNA breaks, and these breaks are then fixed by the RecBC–RecA–RuvABC pathway of recombinational repair. We confirmed chromosomal fragmentation, indicative of double-strand breaks, in dut recBC(Ts) mutants at non-permissive temperatures by pulsed field gels. Inactivation of dcd and ung genes suppresses both the lethality of dut recA and dut recBC strains and the chromosomal fragmentation in dut recBC(Ts) mutants, further supporting the idea that incorporation of uracil, due to the increased level of dUTP, and its subsequent excision by uracil DNA-glycosylase is the primary source of background DNA damage in dut mutants that leads to chromosomal fragmentation and inviability.
Our study indicates that dut mutants have elevated levels of double-strand breaks in their chromosomes, but the nature and the mechanisms of formation of these double-strand breaks are unclear at the moment. Uracil-induced double breaks could result from simultaneous excision of two opposing uracils by uracil DNA-glycosylase and abasic site endonuclease (Dianov et al., 1991). This situation envisages not only accumulation of uracil in new DNA strands, but also uracil presence in old DNA strands, which may be possible only when the uracil excision system is overloaded. A competing model posits that uracil-induced single-strand breaks are converted into double-strand breaks during DNA replication, as a result of replication fork collapse (Fig. 1F and G) (Kuzminov, 1995), requiring uracil presence in one DNA strand only. The two models await to be distinguished experimentally.
Are all hyper-rec mutants synthetic lethal with recA?
In E. coli, six mutants are known to be dependent on RecA and recBC for survival: polA, lig, dam, rdgB, fur and sodAB (reviewed by Kuzminov, 1999). fur and sodAB mutants experience elevated levels of oxidative DNA damage, which results in the production of single-strand breaks in DNA (Keyer et al., 1995; Touati et al., 1995). In dam mutants, nicks are detected in chromosomal DNA (Marinus and Morris, 1974); the reason for dam recA lethality is proposed to be the formation of double-strand breaks at these nicks (Wang and Smith, 1986). In polA and lig mutants, a defect in maturation of Okazaki fragments was reported (Kuempel and Veomett, 1970; Okazaki et al., 1971; Gottesman et al., 1973b; Konrad et al., 1973); therefore, the lethal defect of polA recA and lig recA mutants is linked to the high level of single-strand interruptions in the newly synthesized DNA. dut mutants, like polA and lig mutants, produce short Okazaki fragments, although they are not grossly defective in Okazaki fragment maturation (Konrad and Lehman, 1975; Tye et al., 1977). Our findings of the dut-1 recA inviability encouraged us to take a closer look at the viability of dut-11 recA. Eventually, we found that the dut-11 recA combination is synthetic-lethal, but the lethality is observed only with the ΔrecA allele and is significantly delayed, so that the dut-11ΔrecA cells can form regular-looking colonies despite the loss of the recA+ plasmid, but these colonies are composed of dead cells. Thus, dut recA becomes the seventh known synthetic-lethal with recA in E. coli.
Of the seven RecA-dependent mutants, five (polA, lig, dam, rdgB and dut) are also known hyper-recs. Elevated homologous recombination in all five cases is attributed to the formation of double-strand breaks in the chromosome with subsequent recombinational repair (Kuzminov, 1995). However, there are two more hyper-rec mutants, uvrD and xthA, which are known to be RecA independent (Smirnov et al., 1973; Mattern and Houtman, 1974; Clyman and Cunningham, 1987; Wang and Chang, 1991), suggesting different mechanisms. UvrD protein is a copious DNA helicase II, the protein involved in late stages of several DNA repair reactions in E. coli (Friedberg et al., 1995). XthA protein is the exonuclease III, the main abasic site endonuclease in E. coli (Friedberg et al., 1995). One possibility is that hyper-recombination in uvrD and xthA mutants is initiated by nicks and follows the Meselson–Radding model (Fig. 1A–E). Nicks could be suspected in DNA of uvrD mutants on account of the inability of the uvrD mutant cells to finish excision repair or methyl-directed mismatch repair; however, no single-strand interruptions (nicks) are detectable in DNA of uvrD mutants (Sinzinis et al., 1973). xthA mutants are deficient in the major nicking activity that attacks abasic sites; therefore, if anything, an xthA defect should reduce the number of nicks in DNA. The regular, or maybe even lower, number of nicks in DNA of uvrD and xthA mutants makes it likely that hyper-recombination in these mutants has a different nature. As UvrD helicase can disrupt RecA-promoted pairing between short homologies in vitro (Morel et al., 1993), the hyper-rec phenotype of uvrD mutants could result from the lower fidelity of recombinational repair, rather than the actual increase in DNA lesions (Kuzminov, 1999). In xthA mutants, the primary DNA lesions may be abasic sites, which, when encountered by replication forks, cause the formation of daughter-strand gaps, repaired by the RecFOR recombinational pathway (Kuzminov, 1999), which will also cause hyper-recombination, but why this should make xthA recA double mutants viable is not clear.
Interception of improper/modified DNA precursors
All the known dut synthetic lethalities, as well as their suppressors, can be combined in a metabolic network around ‘regular’ DNA (Fig. 8). In this scheme, dUTP is processed by Dut to dUMP, which is used to synthesize dTTP. dTTP, together with dCTP, is used in the synthesis of regular DNA. In dut mutants, the dUTP pool expands while the level of dTTP drops, leading to a more frequent uracil incorporation into DNA (DNA-U). Uracil excision from DNA is initiated by Ung, followed by ExoIII (XthA), PolA and finished by DNA ligase. With still unknown frequency, the DNA nick intermediate of the uracil excision gives rise to double-strand breaks (the mechanism of this breakage is also unclear, as discussed above), which are repaired by the RecBC–RecA–RuvABC pathway of homologous recombination. Inactivation of the ‘yellow’ activities, xthA, polA, lig, recA, recBC and ruvABC, is lethal in dut mutants (magenta), while inactivation of the ‘blue’ activities, dcd and ung, suppresses the synthetic lethality.
The scheme in Fig. 8 emphasizes the strategy of double-strand DNA break avoidance by dUTPase via intercepting an improper DNA precursor, dUTP, thus preventing incorporation of a wrong base, uracil, into DNA and avoiding the chromosomal consequences of its subsequent excision. Until recently, the only known protein with an analogous function was MutT, hypothesized to intercept 8-oxo-dGTP, the oxidized form of dGTP. 8-Oxo-guanine incorporation causes transversional mutagenesis; therefore, the organism removes 8-oxo-guanine from DNA by MutM DNA glycosylase (Friedberg et al., 1995). mutT recA mutants seem to be viable (E. Rotman and A. Kuzminov, unpublished), suggesting that either incorporation of 8-oxo-guanine into DNA is not a frequent event [there is neither detectable 8-oxo-dGTP in the nucleotide pools of mutT mutant (Tassotto and Mathews, 2002) nor 8-oxo-G in the DNA of mutT mutM mutants (Wójcik et al., 1996)] or the excision of 8-oxo-guanine from DNA is inefficient, which would be consistent with the high mutation rates characteristic of mutT mutants (Vidmar and Cupples, 1993).
Recently, enzymes hydrolysing inosine/xanthosine triphosphates were characterized from Methanococcus jannaschii, Homo sapiens and E. coli (Chung et al., 2001; Lin et al., 2001). Inactivation of the rdgB gene, coding for this enzyme in E. coli, makes cells dependent on recombinational repair (Clyman and Cunningham, 1987; Bradshaw and Kuzminov, 2003), leads to chromosomal DNA degradation in recA mutants (Clyman and Cunningham, 1987) and chromosomal fragmentation in recBC mutants (Bradshaw and Kuzminov, 2003). Synthetic lethality of rdgB recA and rdgB recBC mutants and chromosomal fragmentation in rdgB recBC mutants is suppressed by inactivation of EndoV (Bradshaw and Kuzminov, 2003), which initiates excision of hypoxanthines and xanthines from DNA (Yao et al., 1994; He et al., 2000). It is argued that rdgB mutants are likely to have an increased incorporation of hypoxanthine/xanthine in DNA (because of expanded dITP/dXTP pools) with their subsequent excision by EndoV. This excision leads to chromosomal fragmentation repaired via the RecBC pathway (Bradshaw and Kuzminov, 2003). The overall rdgB phenomenon is similar to the dut phenomenon, the only major difference being that, unlike ΔrdgB, Δdut mutations are inviable in both E. coli (El-Hajj et al., 1988) and yeast (Gadsden et al., 1993). In the light of our findings, the mechanism of Δdut mutant inviability is likely to be massive chromosomal fragmentation exceeding the cell's capacity for recombinational repair.
Media, growth conditions and general methods
Cells were grown routinely in LB broth (10 g of tryptone/5 g of yeast extract/5 g of NaCl per 1 l) or on LB plates (15 g of agar per 1 l of LB broth). Tryptone broth (TB; 10 g of tryptone/5 g of NaCl per 1 l, pH 7.2) was used for phage λ plating and growth. M9 minimal medium (Miller, 1972) was used to test for thymidine requirements. For growth of bacteria resistant to antibiotics or for plasmid-harbouring bacteria, the media were supplemented with the required antibiotic in the following concentrations: 100 µg ml−1 ampicillin, 50 µg ml−1 kanamycin, 12.5 µg ml−1 chloramphenicol or 10 µg ml−1 tetracycline.
For viability testing, strains were grown overnight in LB at either 27°C or 30°C, diluted in the morning 100-fold into fresh broth, grown to 2 × 108 cells ml−1, and 10 µl of several dilutions of a strain was spotted to detect the effect of cell density, if any. NaCl (1%) was used for dilutions. Plates were incubated at the temperatures indicated in the figure legends. P1-mediated transduction was performed as described previously (Miranda and Kuzminov, 2003). Small-scale and preparative plasmid DNA isolation was exactly as described previously (Birnboim, 1983). Insertional mutagenesis with a kanamycin resistance cassette of pRL27 was carried out as described previously (Bradshaw and Kuzminov, 2003).
The E. coli K-12 strains used in this study can be found in Supplementary material, TableS1. dut mutations were moved around using their linkage to zic-4901::Tn10 and were scored on solid media for sensitivity to uracil (Hochhauser and Weiss, 1978) (Fig. 3A, rows 2, 4 and 6), 10 mM for dut-1 allele or 2 mM for dut-11 allele. Both dut-1 and dut-11 mutations were verified by measuring dUTPase activity in cell extracts (see below). To lose the zic-4901::Tn10 marker in the strain L-33, the strain was treated with fusaric acid and scored for sensitivity to tetracycline as described previously (Maloy and Nunn, 1981). ung mutations were verified by sensitivity of the mutants to the uracil-containing λ phage grown on the L-36 dut-1 ung152::Tn10 host (the phage plating was up to 700-fold higher on the ung mutants compared with the isogenic ung+ host) (Duncan, 1985). recA, lexA3, recBC, recF, recG, ruvABC, recJ, recO and recD recJ mutants were confirmed by their characteristic UV sensitivities. Deletion mutations Δung689::cat, Δdcd581::cat, ΔlexA609::cat, ΔsulA510::kan and ΔmltB1085::cat were created by replacing genes in the chromosome with either chloramphenicol or kanamycin resistance markers according to the method of Datsenko and Wanner (2000). Numbers indicate the size of the deleted sequence starting with the A in the first ATG of the corresponding open reading frame (ORF). The sequences of primers used to construct and verify deletion replacement alleles are available upon request.
pho regulon induction in pst mutants was estimated by expression of the bacterial alkaline phosphatase (BAP) on the indicator plates. The strains were streaked on the low-phosphate minimal medium plates (Metcalf and Wanner, 1991) containing 0.04% glucose, 60 µg ml−1 5-bromo-4-chloro-3-indoylyl-phosphate, supplemented with either 0.1 mM phosphate (low-phosphate conditions) or 2 mM phosphate (high-phosphate conditions). BAP constitutive mutants produced dark blue colonies on both 0.1 mM and 2.0 mM phosphate plates.
The following plasmids have been described: pJBS2 (Bradshaw and Kuzminov, 2003), pSU19 (Martinez et al., 1988), pMTL22 (Chambers et al., 1988), pRL27 (Larsen et al., 2002) and pK134 (Miranda and Kuzminov, 2003). pEAK1 is pHSG415 [pSC101-ori(Ts) derivative] (Hashimoto-Gotoh et al., 1981), in which the ≈ 1.75 kbp EcoRI–HindIII fragment was replaced by the 85 bp EcoRI–HindIII fragment of pMTL21 (Chambers et al., 1988) carrying the multiple cloning site. pEAK2 is pEAK1, in which the ≈ 550 bp BamHI–BamHI fragment has been replaced by the 3274 bp BamHI fragment from pBEU14 (Uhlin and Clark, 1981) carrying the wild-type recA gene of E. coli. pEAK5 is pEAK1, in which the ≈ 550 bp BamHI–BamHI fragment has been replaced by the ≈ 3.8 kbp BamHI–BglII fragment from pWB300 (Lehming et al., 1987) carrying a modified functional lacZ gene in the orientation that puts lacZ co-directional with the bla gene. pEAK7 was constructed by cloning an HindIII–EcoRI 300 bp fragment (containing the upstream region of the umuD ORF) amplified from E. coli DNA (strain AB1157) by polymerase chain reaction (PCR) with the Taq polymerase using primers ACCCAAGCTTAATAATCTGCCTGAAGTTATAC and CGGA ATTCTAATTCATAGTTAGCCGG into pBR322 digested with HindIII and EcoRI. pEAK11 was constructed by cloning the 645 bp EcoRI–BamHI fragment (containing the dut+ gene of E. coli) amplified by PCR with primers dutEcoRI (GGAATTC CCAGCCAACTCAAGG) and dutBamHI (CGGGATCCACT TGCCAGGCGGCACC) into pMTL22 digested with EcoRI and BamHI. pEAK12-4 was constructed by cloning the 650 bp BglII–BamHI fragment from pEAK11, containing the dut gene, into the unique BamHI site of pEAK5. Expression of the recA+ and dut + genes from the corresponding plasmids was verified by complementation tests.
Overnight cultures of AB1157, L-80, L-84 and L-101 strains were grown in LB. Cultures were diluted 100 times in the morning and grown in 2 ml of TB to the density of about 5 × 108 cells ml−1. The cultures were diluted with 2 ml of TM buffer (10 mM Tris-HCl, pH 8.0, 10 mM MgSO4), stored at 4°C and used within 1–3 days.
Plating phage to obtain individual plaques.
λ MMS5 (λWT form Frank Stahl's collection) was diluted in TM buffer. A sample of 10 µl of a 10−7 dilution of the phage was mixed with 100 µl of a plating culture and incubated at 37°C for 10 min. Top TB agar (1 ml) was added to the adsorption mixture, and the whole content of the tube was plated on a small TB plate. The plates were incubated overnight at 34°C. An individual plaque was recovered from a plate and put in 1 ml of TM for 2 h.
Preparing phage lysates.
Plating cultures (300 µl) were mixed with 30–60 µl of a single plaque phage eluate and incubated at 37°C for 10 min. Top TB agar (3.2 ml) was added to the adsorption mixture, and the whole content of the tube was plated on a regular TB plate. Plates were incubated at either 42°C or 30°C for 5 h, then overlaid with 5 ml of TM buffer and incubated for 16 more hours at the same temperature. The overlaying TM was collected, briefly vortexed with 100 µl of chloroform and centrifuged at 15 000 g for 20 min to remove cells and agar.
Determining the phage plating factor.
Plating cultures (400 µl) from either AB1157 or L-80 were mixed with 5 ml of top TB agar and plated on square plates. After the plate had been dried open for 20 min, 10 µl of serial phage dilutions from each phage lysate were spotted on the plate. The control phages were λ MMS5 isolated from AB1157 (λT) or from L-36 dut-1 ung152 (λU). Plates were incubated at 34°C overnight, and plaques were counted in those spots where the numbers were between 10 and 40. The plating factor was calculated as the ratio of phage titre on AB1157 (Ung+) to the phage titre on L-80 (Δung689::cat).
DNA sequence analysis
To sequence the dut-1 and dut-11 alleles, a 645 bp fragment containing the dut ORF was amplified from purified chromosomal DNA of AK105, LK-105 or BW209 in five independent 20 µl PCRs containing 1.5 mM MgCl2, 200 µM each dATP, dCTP, dGTP and TTP, 0.4 µM each primer, 20 ng of genomic DNA and 0.5 U of Taq polymerase (Invitrogen). The primer dutEcoRI corresponds to the 5′ end of the gene, and the primer dutBamHI corresponds to the 3′ end of the gene. PCR was performed in a Hybaid PCR Sprint thermocycler under the following conditions: 4 min of initial denaturation at 94°C followed by three cycles of 4 s at 94°C, 10 s at 52°C and 30 s at 72°C each, then 28 cycles of 4 s at 94°C, 10 s at 60°C and 30 s at 72°C each, then one cycle of 5 min at 72°C. The five independent reactions were then pooled, and the PCR product was fractionated in 1% agarose and gel-purified with a QIAquick gel extraction kit (Qiagen). Both strands of DNA were sequenced with primers TTCCCAGCCAACTCAAGG and CCGCAAACGAAATGTTTG to verify the mutation. Plasmid DNA retrieved after insertional mutagenesis was purified with QIAprep spin miniprep kit (Qiagen) and sequenced as described previously (Bradshaw and Kuzminov, 2003).
Overnight cultures grown at 25°C were diluted 50–100 times and grown in nutrient broth until OD600 of 0.5. Cells from 0.5 ml of bacterial cultures were collected by centrifugation, washed once with 500 µl of 50 mM MOPS, pH 6.9, and frozen as a pellet at −80°C. Thawed cell pellets were vortexed for 1 min with 100 µl of the CellLytic BII bacterial lysis extraction reagent (Sigma). The suspension was then centrifuged at 16 000 g for 3 min to pellet the cell debris. A sample of 1 µl containing 0.01 mg ml−1 soluble protein fraction in 50 mM MOPS (pH 6.9) buffer was added to 10 µl of reaction containing 30 mM MOPS buffer (pH 6.9)/500 µM MgCl2/2 µM [3H]-dUTP (2 × 104 c.p.m. total) (Moravek Biochemicals)/3 µM dUTP (ICN). Incubation at 25°C was 30 min for wild-type cell extracts and 1 h for the mutants; at 42°C, the incubation was for 20 min. The reaction was stopped by transferring tubes to ice and adding 4 µl of a carrier solution containing 5 mM each unlabelled dUMP and dUTP. Reaction mixture (6 µl) was applied to a polyethyleneimine cellulose thin-layer (TLC) plate and developed for 10 cm with 0.75 M LiCl. After the TLC plate had dried, the dUMP and dUTP spots were visualized with UV light and, together with the dUDP-containing portion of the plate between them, were cut out to determine radioactivity by liquid scintillation counting without prior elution. dUTPase activity is defined as the increase in the relative amount of dUMP over the background, established in parallel blank reactions without cell extracts. Values of dUTPase activity were proportional to the amount of the cell extracts in the reaction conditions used. dUDPase activity was calculated as a decrease in the relative levels of dUDP; our preparation of [3H]-dUTP contained ≈ 25% dUDP.
Protein concentration was determined spectrophotometrically at wavelength 280 nm for the dut mutant strain screening or by the Bio-Rad protein assay, based on the Bradford dye-binding procedure for the dUTPase activity measurements to compare activities in AK105, L-93, AB1157 and L-107 strains.
Pulsed field gel electrophoresis
Growth of cultures and plug preparation were as described previously (Bradshaw and Kuzminov, 2003; Miranda and Kuzminov, 2003). Overnight cultures were grown at 22°C in LB, diluted to 1–2 × 107 cells ml−1 and grown at 22°C or 37°C until late log phase (≈ 5 × 108 cell ml−1). OD600 of all cultures was adjusted to 0.5 to equalize the DNA loading. Plugs were inserted into 1.2% agarose gel on 0.5× TBE buffer and run in a Gene Navigator unit (Amersham-Pharmacia) with hexagonal electrodes. Running conditions: 165 V, 10°C for 46 h, switch time of 90 s for 10 h, 105 s for 18 h and 125 s for 18 h.
We are grateful to Ichizo Kobayashi, Richard Kolodner, Sidney Kushner, Sue Lovett, Gerry Smith and Frank Stahl for bacterial strains and plasmids, and to Gerry Smith for helpful suggestions on the manuscript. Raven Huang (Department of Biochemistry, UIUC) kindly provided co-ordinates for the dUTPase trimer, used in molecular modelling. This work was supported by grant MCB-0196020 from the National Science Foundation.