Deletions at stalled replication forks occur by two different ... - NCBI

The EMBO Journal Vol.16 No.11 pp.3332–3340, 1997

Deletions at stalled replication forks occur by two different pathways

He´le`ne Bierne1, S.Dusko Ehrlich and Beńe´dicte Michel2 Laboratoire de Geńe´tique Microbienne, Institut National de la Recherche Agronomique, Domaine de Vilvert, F-78352 Jouy en Josas Cedex, France 1Present

address: Unite´ des Interactions Bacte´ries-Cellules, Institut Pasteur, 25 rue du Dr Roux, F-75724 Paris Cedex 15, France 2Corresponding

author

Replication blockage induces non-homologous deletions in Escherichia coli. The mechanism of the formation of these deletions was investigated. A pBR322– mini-oriC hybrid plasmid carrying two E.coli replication terminators (Ter sites) in opposite orientations was used. Deletions which remove at least the pBR322 blocking site (named Ter1) occurred at a frequency of 2310–6 per generation. They fall into two equally large classes: deletions that join sequences with no homology, and others that join sequences of 3–10 bp of homology. Some 95% of the deletions in the former class resulted from the fusion of sequences immediately preceding the two Ter sites, indicating a direct role for blocked replication forks in their formation. These deletions were not found in a topA10 mutant, suggesting a topoisomerase I-mediated process. In contrast, deletions joining short homologous sequences were not affected by the topA10 mutation. However, the incidence of this second class of deletions increased 10-fold in a recD mutant, devoid of exonuclease V activity. This indicates that linear molecules are intermediates in their formation. In addition, ~50% of these deletions were clustered in the region flanking the Ter1 site. We propose that they are produced by repair of molecules broken at the blocked replication forks. Keywords: double-strand break/illegitimate recombination/RecBCD/replication terminators/ topoisomerase I

Introduction Recombination between sequences of little or no homology, termed illegitimate, is one of the major causes of rearrangements in higher eukaryotes (reviewed in Roth and Wilson, 1988; Meuth, 1989). Among these rearrangements, deletions are probably the most dramatic alterations, since they lead to the irreversible loss of genetic information. Illegitimate recombination is a multi-pathway phenomenon (reviewed in Ehrlich, 1989). Sequencing of recombination junctions allowed their classification in two groups, depending on the presence of short-homologous sequences, at least 3 bp, at the new junction. Two major mechanisms were proposed for rearrangements joining short-homo3332

logous sequences. The slipped mispairing, or copy-choice mechanism, results from template switching by DNA polymerases. It was proposed in Escherichia coli by Albertini et al. (1982) and is supported by in vivo and in vitro experiments in E.coli (D’Alençon et al., 1994; Canceill and Ehrlich, 1996) and in vivo in yeast (Gordenin et al., 1992). A second mechanism is conceptually related to the single-strand annealing model, originally proposed to account for homologous recombination events in yeast (Lin et al., 1984). The key event is the occurrence of a DNA double-strand break (DSB) between the repeats, followed by exonucleolytic erosion and pairing of exposed complementary sequences. Such rearrangements using short direct repeats for the pairing step have been described in E.coli (Conley et al., 1986; Kong and Masker, 1994) and in Saccharomyces cerevisiae (Schiestl and Petes, 1991; Moore and Haber, 1996). In many spontaneous rearrangements, short homologies are not found at the new junction, suggesting the existence of strictly non-homologous processes. In higher eukaryotes, end-joining reactions can be catalysed by polymerases and ligases without the help of short homologies (Roth and Wilson, 1988; Thode et al., 1990; King et al., 1994). The fusion of non-homologous sequences can also be mediated by enzymes whose physiological function involves the generation and joining of DNA ends (Ehrlich, 1989). Topoisomerases, which control the topological state of chromosomes through transient breakage and rejoining of DNA strands, are the main enzymes involved in nonhomologous rearrangements (for review, see Champoux and Bullock, 1988). Other nicking–closing enzymes, such as site-specific recombinases, also accidentally mediate illegitimate recombination (Ehrlich, 1989). Studies on DNA structural and functional parameters which stimulate deletion formation indicated a correlation between deletion hotspots and replication pause sites. This correlation was noticed for recombination events between short homologous as well as strictly non-homologous sequences (Bierne and Michel, 1994). A more direct evidence was obtained with the E.coli replication terminator TerB (Hill, 1992), which proved to be a strong deletion hotspot (Bierne et al., 1991). This hotspot was dependent on the presence of the Tus protein, which induces replication arrest upon binding at Ter sites. This suggested that stalled replication forks could be deletionprone structures. Tus–Ter complexes were also shown to stimulate homologous recombination (Horiuchi et al., 1994). A possible common origin for this hyper-recombination activity in different systems may be the occurrence of DNA DSBs at replication arrest sites (Michel et al., 1997). In the present work, we investigated the mechanism of deletions induced by replication arrest in E.coli. A pBR322–mini-oriC hybrid plasmid carrying two Ter sites in opposite orientations was used. Two different types of © Oxford University Press

Deletions at stalled replication forks

Fig. 1. (A) Schematic representation of plasmid pBRToriC. The position and the orientation of pBR322 and oriC replication origins and of replication terminators Ter1 and Ter2 are indicated. ApR and SpecR denote the ampicillin and the spectinomycin resistance gene, respectively. Also shown are the relative positions of the EcoRI (E), HindIII (H) and PvuI (P) restriction sites, used to analyse replication intermediates. (B) Schematic representation of a Ter1–Ter2 replication intermediate, resulting from replication initiation at oriC and replication arrest at both Ter1 and Ter2, and of a pBR–Ter1 replication intermediate resulting from replication initiation at pBR322 origin and replication arrest at Ter1. Template and newly replicated strands are indicated by thin and bold lines, respectively. The bound Tus proteins are shown as open circles. (C) Southern analysis of replication intermediates. Total DNA of exponentially growing E.coli cells harbouring pBRToriC was restricted by HindIII or EcoRI, run on agarose gels, transferred to nylon menbranes and hybridized to the oriC-containing PvuI fragment of the plasmid. L, linear plasmid DNA; chr, chromosomal DNA fragment; Ter1–Ter2, double-Y structure, generated by EcoRI cleavage of the Ter1–Ter2 intermediate; pBR–Ter1, bubble-shaped molecule generated by EcoRI or HindIII cleavage of the pBR–Ter1 intermediate; oriC–Ter1 and oriC–Ter2, single-Y structures produced by HindIII cleavage of the Ter1–Ter2 intermediate. The intermediates are visible in the tus1 strain (lanes tus1) but not in the tus– strain (lanes tus–), where the Ter sites are not functional. Due to the difference in plasmid copy number, 20-fold more DNA was loaded for the tus1 than for the ∆tus strain, which accounts for the difference in the intensity of the oriC-chromosomal signal.

Ter-associated deletions were obtained, that depend on different genetic backgrounds and were presumably produced by two different pathways. Deletions fusing sequences adjacent to the two Ter sites accounted for 45% of total deletions events. They did not depend on the presence of short direct repeats, and they were not found in a strain mutated for DNA topoisomerase I. We propose that they were produced by this enzyme. This suggests that a topoisomerase-mediated rearrangement can be induced by replication arrest. The vast majority of the other deletions occurred between short-homologous sequences, most often immediately flanking the Ter site that arrested the pBR322 replication fork. Their frequency increased more than 10-fold in the absence of exonuclease V, which destroys linear molecules by erosion of DNA doublestrand ends. We propose that this second class of deletions results from the occurrence of a DSB at Ter sites, followed by improper rejoining of DNA ends.

Results Selection of deletions pBRToriC is a pBR322-derived plasmid carrying the minimal oriC replication origin and two TerB replication terminators, named Ter1 and Ter2, in inverted orientation (Figure 1A). Upon binding of the E.coli Tus protein, the Ter sites inhibit the progression of replication forks in an orientation-specific way: Ter1 blocks the progression of pBR322 and oriC counter-clockwise replication, while Ter2 blocks oriC clockwise replication. pBRToriC was constructed in a ∆tus strain and then introduced in tus1 cells. The presence of the Tus protein led to the accumulation of replication intermediates resulting from replication blockage (Figure 1B and C). In addition, it also induced a decrease in plasmid copy number. When replication was impeded at Ter sites, pBRToriC propagated as a low copy number plasmid. Consequently, cells were sensitive to a high concentration of ampicillin (500 µg/ml). Upon

3333

H.Bierne, S.D.Ehrlich and B.Michel

Fig. 2. (A) Sequence analysis of Ter1–Ter2 deletion endpoints. Shown are the sequences preceding Ter1 (lower line) and Ter2 (upper line). Sequences are from the same strand. The first six nucleotides of the Ter sequences are indicated by large letters and the orientation-arresting replication is shown by arrows. Hatched lines represent the rest of the plasmid, oriC being indicated on the left, with the directions of the clockwise (CW) and counter-clockwise (CCW) replication. The distance between two dots corresponds to 20 nucleotides. The nucleotides joined by a deletion (bold letters) are designated with the same number (i.e. 1 in upper line is joined to 1 in lower line, 4, 5, 6, 7 in block A, upper line are joined respectively to 4, 5, 6, 7 in block A, lower line, and so on). A, B and C designate groups of deletion junctions. All deletions were obtained in the wild-type strain (JJC40), except deletion 19 obtained in the recD strain (JJC273). (B) Sequence of the direct repeat deletion hotspot. Top strand pBR322 nucleotides 905–933, bottom strand pBR322 nucleotides 1033–1062. The sequence in the middle is that of the new junction. The 9 bp of homology are in bold.

inactivation of Ter1, pBR322-initiated replication allowed the plasmid to be maintained at 30–40 copies and cells became resistant to that ampicillin concentration. We took advantage of this property to select for Ter1-deleted plasmids. Sequences adjacent to Ter sites are deletion prone The rate of appearance of cells resistant to high concentrations of ampicillin in pBRToriC-containing cultures was determined. Fluctuation tests were used to ensure that all deletion events were independent (see Materials and methods for details of the procedure). Ap500R cells appeared at a frequency of 2310–6 per generation, and most of them carried plasmids derived from pBRToriC by a single deletion event. Restriction analysis of 64 deletant plasmids from four independent experiments indicated that most of them belonged to two major classes. In total, 31 plasmids had lost a 3.3 kb region encompassing the two terminators and the intervening sequence (designated as Ter1–Ter2 deletants), while 11 had lost only Ter1 with ,400 bp of flanking sequences (designated as ∆-Ter1). The remaining plasmids were of variable size. Deletion endpoints were characterized by sequence analysis. Sequences of 18 Ter1–Ter2 junctions were determined. As shown in Figure 2A, deletion endpoints were clustered within the first 50 nucleotides preceding each

3334

Ter site. Hence, these deletions were likely induced by replication arrest at the two terminators. A more detailed analysis of the endpoints revealed that there was only 1 bp or no homology between the recombining sequences. Sequence analysis of the 11 plasmids specifically deleted of the region flanking Ter1 (∆-Ter1) showed that seven resulted from recombination between the same 9 bp direct repeats lying 50 and 80 bp from Ter1 (Figure 2B). In the other four, two perfect 8 bp direct repeats and two nearly perfect repeats (8/9 bp and 6/8 bp) were present at the junction. The 9 bp sequence most often found at the deletion junction is the longest direct repeat in the region flanking Ter1. The presence of short direct repeats at the junction of all the Ter1-flanking deletions suggested that they occurred via a mechanism different from the one which governs the formation of Ter1–Ter2 non-homologous deletions. The sequence of recombination sites was also determined for the 22 remaining plasmids. In 16 cases, between 3 bp and 10 bp of short homologies were present at the junction, while in six cases there was only 1 or no bp of homology. Interestingly, three of these last deletants were not random, as they resulted from the joining of the pBR322 primosome assembly site (pas BL; Zipursky and Marians, 1980) to a nucleotide located 20 bp (found twice) or 200 bp before Ter2. The locations of all deletion endpoints are shown


classes of deletions, the mutation which affected those flanking the Ter1 site increased 20-fold, while the frequency of Ter1–Ter2 deletions was essentially unchanged (Table I). Among the three deletions flanking Ter1 that were sequenced, two occurred at the 9 bp major hotspot and one had only 1 bp of overlap at the junction. The frequency of deletions which do not belong to the two major classes was also increased (Table I); five were sequenced. They all occurred between short direct repeats, 5–9 bp long. These results show that, in the absence of exonuclease V activity, at least 90% of deletions occurred between direct repeats, and that the major effect of the recD mutation is an increase in the frequency of this type of deletion. This indicates that most of the deletions joining short direct repeats involve a linear intermediate. As the deletion endpoints were preferentially clustered in the region flanking Ter1, we propose that the linear intermediates result from DSBs at Ter1. Deletions flanking the Ter2 site may occur by the same pathway, but they cannot be detected here, as they do not confer resistance to a high concentration of ampicillin. Fig. 3. Distribution of deletion endpoints in the wild-type strain JJC40. (A) Deletion between non-homologous sequences (0 or 1 bp homology). (B) Deletions between short direct repeats (3 to 9 bp homology). (C) Linear representation of pBRToriC. The plasmid is open arbitrarily at nt 1600 of the SpecR. The position and the orientation of the replication origins and the Ter sites are indicated. The position of the pas-BL sequence is shown. Bars show the number of deletion endpoints within 100 bp intervals. Solid bars are used for deletions in which both endpoint are close to a Ter site, hatched bars for deletions in which one endpoint is close to a Ter site. Open bars are used for deletions in which both endpoints are at least 200 bp distant from each Ter site.

in Figure 3. Among deletions occurring between nonhomologous sequences (0 or 1 bp homology; Figure 3A), 95% were Ter1–Ter2 deletants. Among the deletions between short direct repeats (3 to 10 bp; Figure 3B), 70% had at least one endpoint in the Ter1-flanking region, and most of them had both endpoints in this region. The frequency of deletions between short direct repeats increases in a recD mutant We previously proposed that illegitimate recombination induced by replication pauses at protein–DNA complexes could result from DNA breakage at arrested forks (Bierne et al., 1991). Such breakage has recently been observed in the E.coli chromosome (Michel et al., 1997). A model for the repair of broken replication forks by the RecBCD recombination pathway was proposed (Horiuchi and Fujimara, 1995; Kuzminov, 1995; Uzest et al., 1995). In this model, the DNA double-strand ends formed at broken replication forks are substrates for the exonuclease activity encoded by the recBCD genes (exo V). It was tempting to speculate that if deletions were induced by breakage of arrested replication forks, they would be stimulated in a recD strain, since the life-time of the linear intermediate molecules would increase. We therefore measured deletion frequencies in a recD::Tn10 strain. The deletion frequency increased 10-fold in the recD strain (Table I), while replication arrest at Ter sites was not affected by this mutation as shown by a constant number of Ter-induced replication intermediates (Table II). Restriction analysis showed that, among the two major

Ter1–Ter2 deletions were not found in a topA10 mutant In the search for an enzyme that might both act at replication forks and have a DNA breaking and joining activity, we tested the effects of a topoisomerase I mutation on deletion formation. The topA10 mutant RS2 was used. Replication arrest at Ter sites was similar in this strain to that in the wild-type strain JJC40, as judged by the amount of Ter1–Ter2 and pBR322–Ter1 replication intermediates detected by Southern hybridization (Table II). Deletion frequency was decreased 2-fold in RS2 compared with JJC40 (Table I), which may not be significant. More interestingly, the distribution of deletant molecules was dramatically affected. Electrophoretic analysis of intact and restricted deletant molecules revealed that none out of 34 plasmids obtained in six independent experiments was a Ter1–Ter2 deletant; indeed, half of them were ∆-Ter1 plasmids (Table I). Junctions were sequenced in two plasmids and deletion had occurred in both cases at the 9 bp hotspot. The rest of the plasmids were of random sizes (Table I, class ‘Others’) and the three sequenced deletions occurred between direct repeats of 3, 4 and 6 bp. This shows that, in the absence of topoisomerase I, Ter1–Ter2 deletion formation is specifically abolished, while deletions joining short direct repeats are not affected. JJC40 and RS2 are not isogenic. The wild-type strain isogenic to RS2, JTT1, could not be directly tested for deletion formation, as it maintained pBRToriC at a slightly higher copy number (1.3 copy per chromosomal oriC; compare with Table II), allowing residual growth of parental cells and thus precluding the selection of deletants. To diminish pBRToriC copy number, the Tus protein was overproduced from the plasmid pACmTus. In JTT1 (top1) containing both pACmTus and pBRToriC, deletions occurred at a frequency of 4310–6. The distribution of deletant plasmids was as follows: 44% of Ter1–Ter2 deletants, 29% of ∆-Ter1, and 27% of plasmids of random sizes. This is not significantly different from that in the other wild-type strain (Table I). In a control experiment, we checked that overproduction of Tus did not modify the results in RS2 (not shown); this confirms that the

3335


Table I. Deletion frequencies in wild-type and mutant strains Strain

Relevant mutation

Ter1–Ter2

∆-Ter1

Others

Total deletions

JJC40 JJC273 RS2

wild-type recD topA10

9.5310–7 1.5310–6 (1.6)a ,8.4310–8 (,0.08)b

4.4310–7 9.1310–6 (20) 4.5310–7 (1)

7.1310–7 9.6310–6 (13) 4.6310–7 (0.64)

2.1310–6 2.0310–5 (10) 9.1310–7 (0.4)

Results are the average of four (JJC40) or six (recD and topA strains) independent experiments. For each experiment, the total rate of deletions was calculated (–log [non-grown wells/total wells]/average number of cells per well; see Materials and methods), and the frequency of each class of deletion was then deduced from the proportion of deletant plasmids belonging to each class; as determined by restriction mapping and/or sequence analysis. Standard deviations were between 0.13% and 0.7% of the values. The values between parentheses were obtained by dividing the deletion frequencies in the mutant strains by those in JJC40. aOne Ter1–Ter2 junction was sequenced, and, as expected, no homology was found between the recombining sequences (Figure 2, deletant 19). bThe limit value ranged from ,8.4310–8 to ,3.7310–7 in the different experiments. Table II. Quantification of replication intermediates in tus1 cells Strain

JJC40 JJC273 RS2

Genotype

wild-type recD topA10

Plasmid forms/chromosomal oriC

Copy number/ chromosomal oriC

CCCa

Ter1–Ter2b

pBR–Ter1c

0.2360.11 0.5060.03 0.7160.24

0.0460.01 0.0460.03 0.0560.02

0.0860.03 0.0760.04 0.0760.03

0.3 0.6 0.8

Values are the average of 4–12 independent determinations. Signals were quantified with a phosphoimager:SI apparatus (Molecular Dynamics). aCovalently closed circular DNA, deduced from the amount of linear DNA. bReplication intermediates resulting from initiation at oriC and replication arrest at both Ter sites, Y-shaped after restriction with EcoRI or HindIII. cReplication intermediates resulting from initiation at the pBR322 origin and arrest at Ter1, bubble-shaped after restriction with EcoRI or HindIII.

absence of Ter1–Ter2 deletants in RS2 results specifically from the inactivation of topoisomerase I.

Discussion In this work, we showed that sequences adjacent to replication terminators are hotspots for illegitimate recombination. Two major classes of deletions were found, which differ by their localization on the plasmid, their requirement for short homologous sequences at the junction, and their genetic dependence. All these deletions are the consequence of replication arrest, but they occur by two different mechanisms. The most frequent rearrangement, representing 45% of the deletion events, results from the fusion of sequences immediately adjacent to the Ter1 and Ter2 sites. Although these two sites are 3.3 kb apart, deletions joining them occur at a frequency of 10–6 per plasmid copy. The substrate for these deletions is likely to be the replication intermediate which accumulates upon Ter-induced replication arrest of oriC-initiated bi-directional replication. Since these replication intermediates represent 10% of total plasmid copies, Ter1–Ter2 deletions may occur at a frequency of 10–5 per molecule of substrate, which is a high frequency for strictly non-homologous deletion events. Topoisomerase I mediates Ter–Ter deletions Ter1–Ter2 deletions were not found in a topA mutant, indicating that the topoisomerase I of E.coli may be involved in their formation. As the expression of several genes is altered in topA mutants of E.coli (Steck et al., 1993), we cannot exclude an effect of the topA mutation on the synthesis of proteins required specifically for the formation of Ter1–Ter2 deletions. It is also conceivable that in a topA mutant the linking number of the molecule is modified so that the probability of contacts between the

3336

two Ter sites is strongly decreased. However, several arguments support a direct role for the topoisomerase in the deletion process. Evidence has accumulated that topoisomerases can mediate illegitimate recombination in prokaryotes as in eukaryotes. In E.coli, DNA gyrase can mediate the joining of non-related molecules (Ikeda, 1990). The eukaryotic topoisomerase I has also been proposed to promote illegitimate recombination (Champoux and Bullock, 1988). In vitro, the cutting and ligating action of eukaryotic topoisomerase I can be uncoupled, the topoisomerase I covalently linked to the 39 end of a ‘donor’ molecule can then catalyse a ligation with a heterologous ‘acceptor’ DNA (Shuman, 1992; Christiansen et al., 1993; Henningfeld and Hecht, 1995). In mammalian cells and in yeast, illegitimate integration occurs preferentially at sites resembling the eukaryotic consensus sequence for TopI action. Furthermore, these illegitimate integration events increase in yeast strains overproducing topoisomerase I and disappear in a TOPI mutant strain (Zhu and Schiestl, 1996). The E.coli topoisomerase I is linked covalently to the 59 end of the DNA (Tse-Dinh et al., 1983), and exhibits a marked structural preference, since in vitro cleavage is always located near a junction between single-stranded and double-stranded DNA (Kirkegaard and Wang, 1985). The location of Ter1–Ter2 deletions at arrested replication forks is consistent with deletions being mediated by enzymes that act at single-strand–double-strand junctions. Although E.coli topoisomerase I does not act preferentially during replication termination, (Hiasa et al., 1994), it could possibly act on forks accidentally stalled during elongation. Thus, at each Ter site, repeated attempts to restart replication could be accompanied by a local activity of topoisomerase I at increased frequency. In case of blockage of the replication fork from the opposite direc-


Fig. 4. Models for TopA-mediated deletion formation between divergent replication forks blocked at Ter sites. The Ter1–Ter2 replication intermediate is partially represented, template strands are shown as thin lines, newly synthesized leading and lagging strand as continuous and interrupted thick lines respectively. O and T stand for oriC and replication terminators respectively. The 59 and 39 ends generated by topoisomerase cleavage are represented by a point and a thin arrow respectively. Deletion can result either from the junction of the template strands (model A) or from the junction of the newly synthesized strands (B). In step A1, topoisomerase-mediated cleavages occur at each of the two replication forks, in the vicinity of Ter sites. A2: a topoisomerase molecule (a) covalently linked to the 59 end generated on the leading strand template catalyses by error a ligation with the 39 end created by another topoisomerase molecule (b) at the other replication fork. A3: this leads to the excision of a gapcontaining molecule, which is converted into a circular double-strand plasmid by continuation of the leading strand synthesis. In step B1, a topoisomerase molecule cleaves the lagging strand in the vicinity of a Ter site. B2: this molecule, bound to the 59 end, catalyses the joining to the 39 end of the leading strand at the other blocked replication fork. B3: a circular double-strand deletant plasmid is generated by another round of replication.

tion, recurrent action of topoisomerase I on partly singlestranded molecules could favour the uncoupling of the cutting and ligating reactions and hence the joining to an ectopic acceptor site (Figure 4). This acceptor site is preferentially the other blocked replication fork; hence topoisomerase I could act at both replication forks (Figure 4A). The double-strand DNA ends would then be protected from exo V degradation by bound topoisomerase molecules. This would account for the lack of effect of the recD mutation on this class of deletions. Alternatively, the acceptor nucleotide could be the 39 end of the stalled replication fork (Figure 4B).

DNA breakage at Ter sites induces deletions between direct repeats In the wild-type strain, 45% of total deletions resulted from recombination between 3 to 10 bp direct repeats. The longest direct repeat flanking Ter1 (9 bp) was used preferentially. We also observed a bias in favour of direct repeats with a high GC content (70% GC in the direct repeats present at the deletion junctions versus 50% GC in the plasmid). Such a bias was previously observed in other systems (Singer and Westyle, 1988) and may reflect the need to stabilize the pairing of short complementary sequences. The presence of short homologies at the recombining junctions suggests that deletions occurred either by replication slippage or during double-strand break repair. Direct repeats effectively favour these two processes which both involve an intermediate step stabilized by the pairing of short complementary sequences. In E.coli, linear plasmid molecules are degraded by RecBCD until the encounter with a CHI site protects them from the exo V action (Dabert et al., 1992). Our plasmid is likely to be entirely degraded by RecBCD upon linearization as it does not carry a CHI site. In order to distinguish between the replication slippage and the DSB repair models, we used a recD mutant, inactivated for exonuclease V. In the absence of exo V (recD mutants), the incidence of deletions between short homologous sequences increased 15-fold. This indicates that linear molecules are intermediates in deletion formation, and are present in a limiting amount in wild-type cells, due to the degradation by RecBCD. Three lines of evidence support the hypothesis that the blockage of replication at Ter sites is responsible for the appearance of these linear molecules: (i) in wild-type cells, most of the deletions between short homologous sequences have at least one end in the vicinity of the Ter1 site (Figure 3B); (ii) the proportion of these Ter1-associated deletions is increased in the recD mutant (Table I); and (iii) an inverted-Ter structure similar to the one used here induces DSBs in the chromosome of E.coli, and the double-strand ends are accessible to the RecBCD proteins (Michel et al., 1997). We propose that deletions between short-homologous sequences result from the repair of Ter1-induced DSBs. This repair may proceed by the single-strand annealing pathway (Figure 5). Although deletions occurred here between short-homologies, similar reactions would be favoured by the presence of longer homologous sequences and the same model may apply to homologous recombination events. Replication pause sites and deletion formation Ter sites were used here as a model system to block replication progression. It is tempting to speculate that other nucleoprotein complexes, known to cause replication pauses, could also induce deletion formation by erroneous action of topoisomerases and/or improper repair of DSBs. For example, transcribing RNA polymerases have been shown to cause transient arrest of replication forks (Liu and Alberts, 1995; Deshpande and Newlon, 1996) and transcribed regions are hotspots for illegitimate recombination, provided that they interfere with replication (Vilette et al., 1995). The LacI–lac operator complex is similarly a replication-dependent deletion hotspot (Vilette et al., 1992). Finally, in the E.coli chromosome, ionizing radiation stimulates the formation of deletions which were

3337


replicative DNA helicases by a mutation leads to replication pauses that cause DSBs (Michel et al., 1997) and could therefore lead to rearrangements. In higher eukaryotes, deletions occur more frequently via illegitimate than via homologous recombination. Some of these illegitimate recombination events have been proposed to occur at replication arrest sites (Krawczak and Cooper, 1991; Stary and Sarrasin, 1992). In eukaryotic genomes, the replication complexes are fixed to the nuclear matrix with the replicating strands extruding as loops. In the resulting structure, opposite-growing replication points are in close proximity, which could facilitate recombination (Painter and Kapp, 1991). Replication arrest could favour interactions between replication forks and thus lead to deletion formation as described here for E.coli. Some of the excised molecules could be a potential source of polydisperse circles commonly found in eukaryotic cells (Roth and Wilson, 1988). Illegitimate recombination would thus be the ultimate way to rescue blocked replication forks, at the expense of genetic integrity.

Materials and methods Bacterial strains and plasmids The bacterial strains used in this work are E.coli JJC40 (ABII57, hsdR; Bierne et al., 1991), JJC256 (as JJC40, but ∆tus::kanR; P1 transduced from TH205; Hill et al., 1988), JJC273 (as JJC40 but recD::Tn10; Michel et al., 1997), JTT1 and RS2 (wild-type and topA10, respectively; Sternglanz et al., 1981). The 7 kb-long pBRToriC plasmid is composed of: (i) the pBR322 plasmid sequence deleted of the AvaI–PvuII region; (ii) a non-transcribed fragment of the EmR gene of plasmid pE194 (NT 2141, 2900; Horinouchi and Weisblum, 1982) cloned into the ClaI site of pBR322; (iii) the SpecR Ω fragment of plasmid pHP45Ω (Prentki and Krisch, 1984) inserted at the BamHI site of pBR322; (iv) a 460 bp EcoRI fragment of plasmid pOC84 containing the minimal replication origin oriC (Messer et al., 1985). In addition, two identical pairs of complementary oligonucleotides containing the TerB sequence were cloned in inverted orientation, one into the NruI site of pBR322 (Ter1) and the other one into the BclI site of the EmR sequence (Ter2). The oligonucleotides used are: 59-CTGCAGAATAAGTATGTTGTAACTAAAGTAGTACT-39 39-GACGTCTTATTCATACAACATTGATTTCATCATGA-59 Fig. 5. Models for deletion formation between short direct repeats. A replication fork arrested at a Ter (T) site is represented. Short repeated sequences (1 and 2) are shown as bold lines. (a) Breakage of the lagging strand template in the vicinity of Ter, generating a DSB. (b) Nucleolytic degradation of the exposed 59 ends, generating a 39-tailed single-strand region. (c) Annealing of complementary sequences 1 and 2. (d) The intermediate is repaired by removal of the 39 tail, gap filling and ligation. A round of replication produces the deletant plasmid molecule. Alternatively, the leading strand template is also broken and repair of both strands leads directly to the deletant plasmid.

The sequences of Ter1 and Ter2 were verified. Plasmid pACmTus was constructed by replacing the TetR gene of plasmid pACYC184 with a fragment carrying the araC repressor gene and the tus coding sequence under the control of an arabinose-inducible promoter. This fragment was cloned from the pBAD–Tus plasmid (Sharma and Hill, 1995). Cells were grown on standard media (Luria broth). For the Tus induction experiments, the medium was supplied with 0.15% arabinose. Plasmid preparation and analysis, preparation and transformation of E.coli competent cells have been described previously (Bierne et al., 1991). Restriction enzymes, ligase and polymerase were from commercial sources and were used according to the suppliers’ recommendations.

proposed to result from the illegitimate joining of replication forks blocked by lesions (Hutchinson, 1993). In the present work, three of the deletions occurring between nonhomologous sequences occurred at the pBR322 primosome assembly site (pas-BL; Zipursky and Marians, 1980) (Figure 3A). The pas-BL sequence has the potential to form a stem and loop structure, and is the binding site for the PriA protein, and hence for primosome assembly (Zipursky and Marians, 1980). It is conceivable that deletions between pas-BL and Ter2 result from replication arrest at both endpoints, the pas-BL acting as a weaker replication pause site. In addition, the inactivation of

Analysis of replication intermediates and copy number determination To detect accumulation of replication intermediates by agarose gel electrophoresis, total DNA of the pBRToriC-carrying strains was prepared as described by te Riele et al. (1986) and treated with EcoRI or HindIII. The migration of replication intermediates is delayed on agarose gel compared with linear fragments (Sharma and Hill, 1995). DNA–DNA hybridization experiments were carried out by a method derived from Southern with transfer performed on nylon membranes by vacuum blotting. The probes used were fragments or all of pBRToriC DNA. The use of different restriction fragments as probes allowed the unambiguous identification of all bands. To quantify the amount of replication intermediates and to measure the plasmid copy number, blots were scanned with a phosphorimager and submitted to a gel analyser program (Image quant, Molecular Dynamics). The EcoRI or HindIII oriCchromosomal DNA fragments, which also hybridized with the probe,

3338

Deletions at stalled replication forks were used as a reference. The plasmid copy number was calculated by summing the amount of the different molecular plasmid species present in restricted DNA. For each strain, 4–12 independent experiments were performed.

Deletion frequency determination Deletion frequencies were measured using a fluctuation test derived from the Luria and Delbruck test (Chedin et al., 1994). The principle of the test is as follows. A large number of identical cultures are inoculated with a number of bacteria small enough not to contain mutant cells. Cells are grown until, on average, there is one mutant cell per culture. Selection is then applied to block the growth of parental cells while mutant cells will grow. The distribution of the events leading to the mutant phenotype in the different cultures follows a Poisson distribution. The frequency is given by the ratio m/N, where m is the average number of mutation events per culture, deduced from the relation P0 5 e–m, calculated from the proportion of cultures not-grown, and N the average number of viable cells. The mutant phenotype tested here was the ability to grow in presence of 500 µg/ml Ap. This concentration was determined to be the one which blocked growth of parental pBRToriC-containing cells, if ,106 cells were inoculated, while allowing growth of a single cell containing a Ter1-lacking (deletant) plasmid. Experiments were performed at 37°C. pBRToriC was transformed into the tested strain selecting clones on plates containing 10 µg/ml of ampicillin. Fresh colonies (1–4) were resuspended into 25 ml of LBT containing 20 µg/ml Ap and incubated with agitation for 1 h. 100 µl aliquots were then distributed into two microwell plates, each of 48 wells. One microwell plate was used to measure recombination frequencies as follows. Cells were grown for 5–7 h without shaking (4–8 generations), the number of viable cells per culture was determined, and each culture was resuspended into 1 ml of LBT containing a final concentration of 500 µg/ml Ap and grown overnight with agitation. This allowed a single Ap500R cell to grow to saturation. To analyse the plasmid DNA content of the Ap500R cells, 10 µl of each growing cultures was inoculated into 5 ml of LBT containing 500 µg/ml Ap, culture was grown until saturation and plasmid DNA was isolated and analysed. At the onset of the experiment, we checked the absence of pre-existent recombinants by applying immediately the selective pressure (500 µg/ml Ap) to the 48 cultures of the control microwell plate. Pre-existent deletant plasmids (rarely obtained) were analysed and identical plasmids obtained in the recombination experiment were not taken into account. Analysis of deletant plasmids To map deletion endpoints, plasmid DNA isolated from Ap500R cells was analysed by electrophoresis on agarose gel, intact and treated with different restriction enzymes. The nucleotide sequences of deletion junctions were determined for 49 deletant plasmids from JJC40, nine deletants from JJC273, and five deletants from RS2. Plasmid DNA for sequencing was prepared as described by Sorokin et al. (1995). PCR sequencing was performed by use of Applied Biosystems PRISM dye terminators sequencing kit on the Perkin Elmer 9600 thermal cycler, and analysed in an Applied Biosystems 373 DNA sequencer.

Acknowledgements We thank J.-Y.Bouet for the gift of pOC84 and M.Uzest for helpful technical assistance in sequencing deletion endpoints. We are very grateful to C.Bruand and E.Cassuto for helpful comments on the manuscript. B.M. is a member of the CNRS staff.

References Albertini,A.M., Hofer,M., Calos,M.P. and Miller,J. (1982) On the formation of spontaneous deletions: the importance of short sequences of homologies in the generation of large deletions. Cell, 29, 319–328. Bierne,H. and Michel,B. (1994) When replication forks stop. Mol. Microbiol., 13, 17–23. Bierne,H., Ehrlich,S.D. and Michel,B. (1991) The replication termination signal TerB of the Escherichia coli chromosome is a deletion hotspot. EMBO J., 10, 2699–2705. Canceill,D. and Ehrlich,S.D. (1996) Copy-choice recombination mediated by DNA polymerase III holoenzyme from Escherichia coli. Proc. Natl Acad. Sci. USA, 93, 6647–6652. Champoux,J.J. and Bullock,P.A. (1988) Possible role for the eucaryotic type I topoisomerase in illegitimate recombination. In Kucherlapati,R.

and Smith,R. (eds), Genetic Recombination. American Society for Microbiology, Washington, DC, pp. 655–666. Chedin,F., Dervyn,E., Dervyn,R., Ehrlich,S.D. and Noirot,P. (1994) Frequency of deletion formation decreases exponentially with distance between short direct repeats. Mol. Microbiol., 12, 561–569. Christiansen,K., Svejstrup,A.B.D., Andersen,A.H. and Westergaard,O. (1993) Eukaryotic topoisomerase I-mediated cleavage requires bipartite DNA interaction. J. Biol. Chem., 268, 9690–9701. Conley,E.C., Saunders,V.A., Jackson,V. and Saunders,J.R. (1986) Mechanism of intramolecular recyclization and deletion formation following transformation of Escherichia coli with linearized plasmid DNA. Nucleic Acids Res., 14, 8919–8932. Dabert,P., Ehrlich,S.D. and Gruss,S. (1992) χ sequence protects against RecBCD degradation of DNA in vivo. Proc. Natl Acad. Sci. USA, 89, 12073–12077. D’Alençon,E., Petranovic,M., Michel,B., Noirot,P, Aucouturier,A., Uzest,M. and Ehrlich,S.D. (1994) Copy-choice illegitimate recombination revisited. EMBO J., 13, 2725–2734. Deshpande,A.M. and Newlon,C.S. (1996) DNA replication fork pause sites dependent on transcription. Science, 272, 1030–1033. Ehrlich,S.D. (1989) Illegitimate recombination in bacteria. In Berg,D.E. and Howe,M.M. (eds), Mobile DNA. American Society for Microbiology, Washington, DC, pp. 799–832. Gordenin,D.A., Malkova,A.L., Peterzen,A., Kulikov,V.N., Pavlov,Y.I., Perkins,E. and Resnick,M.A. (1992) Transposon Tn5 excision in yeast: influence of DNA polymerase α, δ, and ε and repair genes. Proc. Natl Acad. Sci. USA, 89, 3785–3789. Henningfel,K.A. and Hecht,S.M. (1995) A model for topoisomerase Imediated insertions and deletions with duplex DNA substrates containing branches, nicks and gaps. Biochemistry, 34, 6120–6129. Hiasa,H., Digates,R.J. and Marians,K.J. (1994) Decatening activity of Escherichia coli DNA gyrase and topoisomerases I and III during oriC and pBR322 DNA replication in vitro. J. Biol. Chem., 269, 2093–2099. Hill,T.M. (1992) Arrest of bacterial DNA replication. Annu. Rev. Microbiol., 46, 603–633. Hill,T.M., Pelletier,A.J., Tecklenburg,M.L. and Kuempel,P.L. (1988) Identification of the DNA sequence from the E. coli terminus region that halts replication forks. Cell, 55, 459–466. Horinouchi,S. and Weisblum,B. (1982) Nucleotide sequence and functional analysis of pE194. J. Bacteriol., 150, 804–814. Horiuchi,T. and Fujimura,Y. (1995) Recombinational rescue of the stalled DNA replication fork: a model based on analysis of an Escherichia coli strain with a chromosome region difficult to replicate. J. Bacteriol., 177, 783–791. Horiuchi,T., Fujimara,Y., Nishitani,H., Kobayashi,T. and Hidaka,M. (1994) The DNA replication fork blocked at the Ter site may be an entrance for the RecBCD enzyme into duplex DNA. J. Bacteriol., 176, 4656–4663. Hutchinson,F. (1993) Induction of large DNA deletion by persistent nicks: a new hypothesis. Mutat. Res., 299, 211–218. Ikeda,H. (1990) In Cozarelli,N.R. and Wang,J.C. (eds), DNA Topology and its Biological Effects. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp. 341–359. King,J.S., Fairley,C.F. and Morgan,W.F. (1994) Bridging the gap. J. Biol. Chem., 269, 13061–13064. Kirkegaard,K. and Wang,J.C. (1985) Bacterial DNA topoisomerases I can relax positively supercoiled DNA containing a single-stranded loop. J. Mol. Biol., 185, 625–637. Kong,D. and Masker,W. (1994) Deletion between direct repeats in T7 DNA stimulated by double-strand breaks. J. Bacteriol., 176, 5904–5911. Krawczack,M. and Cooper,D.N. (1991) Gene deletions causing human genetic disease: mechanisms of mutagenesis and the role of the local DNA sequence environment. Hum. Genet., 86, 425–441. Kuzminov,A. (1995) Collapse and repair of replication forks in Escherichia coli. Mol. Microbiol., 16, 373–384. Lin,F.L., Sperle,K. and Sternberg,N. (1984) Model for homologous recombination during transfer of DNA into mouse L cells: role for DNA ends in the recombination process. Mol. Cell. Biol., 4, 1020–1034. Liu,B. and Alberts,B.M. (1995) Head-on collision between a DNA replication apparatus and RNA polymerase transcription complex. Science, 267, 1131–1137. Messer,W., Bellekes,U. and Lother,H. (1985) Effect of dam methylation on the activity of the E.coli replication origin, oriC. EMBO J., 4, 1327–1332.

3339

H.Bierne, S.D.Ehrlich and B.Michel Meuth,M. (1989) Illegitimate recombination in mammalian cells. In Berg,D.E. and Howe,M.M. (eds), Mobile DNA. American Society for Microbiology, Washington, DC, pp. 833–860. Michel,B., Ehrlich,S.D. and Uzest,M. (1997) DNA double-strand breaks caused by replication arrests. EMBO J., 16, 430–438. Moore,J.K. and Haber,J.E. (1996) Cell cycle and genetic requirements of two pathways of nonhomologous end-joining repair of doublestrand breaks in Saccharomyces cerevisiae. Mol. Cell. Biol., 16, 2164–2173. Painter,R.B. and Kapp,L.N. (1991) Replication intermediates as substrates for DNA rearrangements. Mutat. Res., 262, 21–23. Prentki,P. and Krisch,H.M. (1984) In vitro insertional mutagenesis with a selectable DNA fragment. Gene, 29, 303–313. Roth,D. and Wilson,J. (1988) Illegitimate recombination in mammalian cells. In Kucherlapati,R. and Smith,G.R. (eds), Genetic Recombination. American Society for Microbiology, Washington, DC, pp. 621–653. Schielst,R.H. and Petes,T.D. (1991) Integration of DNA fragments by illegitimate recombination in Saccharomyces cerevisiae. Proc. Natl Acad. Sci. USA, 88, 7585–7589. Sharma,B. and Hill,T.M. (1995) Insertion of inverted Ter sites into the terminus region of the Escherichia coli chromosome delays completion of DNA replication and disrupts the cell cycle. Mol. Microbiol., 18, 45–61. Shuman,S. (1992) Two classes of DNA end-joining reactions catalyzed by vaccinia topoisomerase I. J. Biol. Chem., 267, 16755–16758. Singer,B.S. and Westlye,J. (1988) Deletion formation in bacteriophage T4. J. Mol. Biol., 202, 233–243. Sorokin,A., Serror,P., Pujic,P., Azevedo,V. and Ehrlich,S.D. (1995) The Bacillus subtilis chromosome region encoding homologues of the Escherichia coli mssA and rpsA gene products. Microbiology, 141, 311–319. Stary,A. and Sarasin,A. (1992) Molecular analysis of DNA junctions produced by illegitimate recombination in human cells. Nucleic Acids Res., 20, 4269–4274. Steck,T.R., Franco,R.J., Wang,J.-Y. and Drlica,K. (1993) Topoisomerase mutations affect the relative abundance of many Escherichia coli proteins. Mol. Microbiol., 10, 473–481. Sternglanz,R., Dinardo,S., Voekel,K.A., Nishimura,Y., Hirota,Y., Becherer,K., Zumstein,L. and Wang,J.C. (1981) Mutations in the gene coding for Escherichia coli DNA topoisomerase I affect transcription and transposition. Proc. Natl Acad. Sci. USA, 78, 2747–2751. Te Riele,H., Michel,B. and Ehrlich,S.D. (1986) Are single-stranded circles intermediates in plasmid DNA replication? EMBO J., 5, 631–637. Thode,S., Scha¨fer,A., Pfeiffer,P. and Vielmetter,W. (1990) A novel pathway of end-to-end joining. Cell, 60, 921–928. Tse-Dinh,Y.C., McCarron,B.G., Arentzen,R. and Chowdhry,V. (1983) Mechanistic study of E. coli DNA topoisomerase I: cleavage of oligonucleotides. Nucleic Acids Res., 11, 8691–8701. Uzest,M., Ehrlich,S.D. and Michel,B. (1995) Lethality of rep recB and rep recC double mutants of Escherichia coli. Mol. Microbiol., 17, 1177–1188. Vilette,D., Uzest,M., Ehrlich,S.D. and Michel,B. (1992) DNA transcription and repressor binding affect deletion formation in Escherichia coli plasmids. EMBO J., 11, 3629–3634. Vilette,D., Ehrlich,S.D. and Michel,B. (1995) Transcription-induced deletions in Escherichia coli plasmids. Mol. Microbiol., 17, 493–504. Zhu,J. and Schiestl,R.H. (1996) Topoisomerase I involvement in illegitimate recombination in Saccharomyces cerevisiae. Mol. Cell. Biol., 16, 1805–1812. Zipursky,S.L. and Marians,K.J. (1980). Identification of two Escherichia coli factor Y effector sites near the origins of replication of plasmids ColE1 and pBR322. Proc. Natl Acad. Sci USA, 77, 6521–6525. Received on December 19, 1996; revised on February 13, 1997

3340