Instability of repetitive DNA sequences: The role of ... - Semantic Scholar

Colloquium

Instability of repetitive DNA sequences: The role of replication in multiple mechanisms Malgorzata Bzymek* and Susan T. Lovett† Department of Biology and Rosenstiel Basic Medical Sciences Research, Center MS029, Brandeis University, 415 South Street, Waltham, MA 02454-9110

I

n bacteria, systematic study of repetitive sequence instability has provided some insights into the molecular mechanisms of repetitive sequence rearrangement. In this paper, we will review the genetic properties of tandem repeat rearrangements in Escherichia coli that are informative about the mechanisms of these processes. Although repetitive sequences can rearrange to either increase or decrease the number of repetitive elements, the process of deletion of repeated DNA sequences has been more widely characterized and will dominate our discussion. Nonetheless, many of the properties of repeat amplification (or expansion) are similar to those defined for repeat deletion (1). We will propose and contrast several molecular mechanisms by which tandem repeats rearrange: homologous recombination, simple slipped misalignment, sister-chromosome slipped misalignment, and single-strand annealing. We present new data in support of a single-strand annealing mechanism. The homologous recombination pathways have been well studied in E. coli (2) and mediate interactions of repeated DNA in many contexts. This article will focus on the other more genetically elusive molecular mechanisms. These mechanisms contribute to rearrangements between repeats found in direct orientation and constitute the RecA-independent recombination pathways of E. coli. Discussion of RecA-independent recombination can be found subsumed under the term ‘‘illegitimate’’ recombination—we favor the more descriptive and specific ‘‘RecAindependent’’ recombination to denote these pathways. Rearrangements between repetitive sequence elements underlie many examples of genomic instability in both prokaryotes and eukaryotes. A large subset of mutations that inactivate genes are deletion events between two short regions of sequence homology, both in bacteria and in humans (3–5). In addition to rearrangements between dispersed repeated DNA sequences, instability of repeated sequences juxtaposed in tandem is also observed. Rearrangements causing either addition or deletion of one or more of the sequence repeats can arise. In humans, deletion or duplication between repeated DNA sequences contributes to human genetic disease, both of nuclear genes and in the mitochondrial genome (4, 6, 7). Several genetically inherited www.pnas.org兾cgi兾doi兾10.1073兾pnas.111008398

neuromuscular disorders, such as Fragile X, Huntington’s disease, and myotonic muscular dystrophy, are associated with expansion of a trinucleotide repeat array (8). Given that local rearrangements between dispersed or tandem repetitive sequences contribute significantly to genetic instability in prokaryotes and eukaryotes, an important question is whether there are underlying common mechanisms for these rearrangements. Materials and Methods Bacterial Strains and Growth. All strains used are derived from the

E. coli K-12 strain AB1157 [F⫺ thi-1 hisG4 ⌬(gpt-proA)62 argE3 thr-1 leuB6 kdgK51 rfbD1 ara-14 lacY1 galK2 xyl-5 mtl-1 tsx-33 supE44 rpsL31 rac⫺ ␭⫺; ref. 9] and carry the additional mutant alleles indicated. Experimental strains were constructed by P1 virA transduction (10). Details of the constructions are available on request from the authors and will be published elsewhere. LB medium was supplemented with 100 ␮g兾ml ampicillin or 15 ␮g兾ml tetracycline. Deletion Assays. Plasmid pMB301 was constructed by ligation of

synthetic oligonucleotides of the sequence 5⬘ GATCTTGGG AGCTTGTTCT TGAGCATTCA AACTCCTAGA GGAAGAAGAA CGTAGC and 5⬘ GATCGCTACG TTCTTCTTCC TCTAGGAGTT TGAATGCTC AAGAACAAGC TCCCAA in the BglII site of pSTL57 (11). The appropriate construction was confirmed by DNA sequence analysis. Deletion between the 101-bp direct repeats in tetA on plasmids pSTL57 and pMB301 was assayed as described (11) by determination of the number of tetracycline-resistant colony-forming units (cfu) in the ampicillin-resistant population for a total of 8–64 independent isolates. Deletion rates were calculated by the method of the median (12) by using the following formula: deletion rate ⫽ M兾N, where M is the calculated number of deletion events and N is the final average number of Apr cells in the 1-ml cultures. M is solved by interpolation from experimental determination of r0, the median number of Tcr cells, by using the formula r0 ⫽ M(1.24 ⫹ lnM). A 95% confidence interval was determined as described (13). Linear Transformation Assays. Purified pSTL57 plasmid DNA (prepared with miniprep kits from Qiagen, Chatsworth, CA) was subjected to digestion by BglII restriction endonuclease, which cleaves between the two 101-bp direct repeats in this plasmid. The linear fragment was isolated by agarose gel electorophoresis and purified (QIAquick gel purification kit from Qiagen). Electroporation (ref. 14; Bio-Rad Gene Pulser) was used to

This paper results from the National Academy of Sciences colloquium, ‘‘Links Between Recombination and Replication: Vital Roles of Recombination,’’ held November 10 –12, 2000, in Irvine, CA. Abbreviations: ssDNA, single-strand DNA; DSB, double-strand break; SCE, sister-chromosome exchange. *Present address: Department of Molecular and Cellular Biology, Harvard University, 7 Divinity Avenue, Cambridge, MA 02138. †To

whom reprint requests should be addressed. E-mail: [email protected].

PNAS 兩 July 17, 2001 兩 vol. 98 兩 no. 15 兩 8319 – 8325

COLLOQUIUM

Rearrangements between tandem sequence homologies of various lengths are a major source of genomic change and can be deleterious to the organism. These rearrangements can result in either deletion or duplication of genetic material flanked by direct sequence repeats. Molecular genetic analysis of repetitive sequence instability in Escherichia coli has provided several clues to the underlying mechanisms of these rearrangements. We present evidence for three mechanisms of RecA-independent sequence rearrangements: simple replication slippage, sister-chromosome exchange-associated slippage, and single-strand annealing. We discuss the constraints of these mechanisms and contrast their properties with RecA-dependent homologous recombination. Replication plays a critical role in the two slipped misalignment mechanisms, and difficulties in replication appear to trigger rearrangements via all these mechanisms.

transform100 ng of cleaved and purified DNA into the appropriate bacterial strains. Uncut and purified pBR322 DNA was transformed in parallel as a control for the efficiency of transformation. After serial dilution in 56兾2 buffer (15), the number of transformants was determined by plating on LB medium containing 100 ␮g兾ml ampicillin. Plating within experiments was performed in duplicate. Results and Discussion Homology-Dependent, RecA-Independent Recombination. Deletion or duplication of tandem repeat sequences in E. coli shows weak or no dependence on RecA DNA pairing and strand exchange protein (11, 16–19). In contrast, homologous genetic recombination scored by conjugation crosses is highly dependent on the RecA, with reductions by greater than five orders of magnitude seen in recA mutant strains (20). Although RecA-independent tandem repeat rearrangements can occur between very short regions of sequence homology, it is nonetheless homologydriven. The homology dependence of repeat deletion is quite striking (16, 17, 21), with large repeats deleting at very high rates in the population. So, although short sequence homologies suffer deletion rarely, in the range of spontaneous point mutations, deletion of large repeats approaches the efficiency of homologous RecA-dependent recombination. RecA-dependent recombination of tandem repeats, on the other hand, appears to require a threshold of minimal homology, although that minimal length may be somewhat context-specific. One study found that for homologies less than about 200 bp in length, there is little or no appreciable contribution of RecA-dependent recombination to the observed deletion events (21). For repeats larger than about 200 bp in length, deletion can occur by RecA-dependent homologous recombination, in addition to the RecAindependent pathway (21). In our hands, there is no detectable RecA-dependent recombination contributing to deletion of 100-bp repeats (11); in contrast, approximately 2兾3 of the deletion events between 787-bp repeats occurs by the RecA homologous recombination pathway (18). The genetic basis of RecA-independent recombination is ill defined. RecA-independent recombination is also independent of the known recombination functions of E. coli, including RecBCD, RecET, RecF, RecG, RecJ, RecN, RecO, RecQ, RecR, RuvAB, and RuvC (M.B. and S.T.L., unpublished results; ref. 18). Our screens for transposon insertion mutants decreasing the rate of RecA-independent deletion of tandem repeats have failed to yield candidates, suggesting that if genes are required to mediate RecA-independent rearrangements, they may be essential. A number of genes whose mutation leads to elevated rates of tandem repeat deletion have been defined and include many of the components of the replication apparatus (see refs. 13 and 22 and below). Mutation of the 3⬘ single-strand exonuclease, Exonuclease I (whose gene is named sbcB or xonA), elevates deletion of both short and long sequence repeats (23, 24). Chromosome topology may also play a role: mutants in the type I topoisomerase, topoisomerase III (topB), have been isolated by their hyperdeletion phenotype (25, 26). The mechanism by which inactivation of topB potentiates deletion formation is not understood but could include increased negative supercoiling or increased formation of a structural intermediate sensitive to topo III unwinding. One potential difficulty in the genetic analysis of RecA-independent recombination is the existence of more than one RecA-independent pathway, each with a distinct genetic basis. Transposon excision is a special case of deletion at short direct repeats, promoted by the large inverted repeats of these elements (27, 28). Mutations that elevate the precise and nearly precise excision of the transposon Tn10 or Tn5 at short direct repeats within or flanking the element (29) have defined the so-called tex mutants (27, 28). Mutations in the uup gene also 8320 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.111008398

exhibit a Tex phenotype (30–32). The function of Uup is not known but it appears to be a general suppressor of RecAindependent rearrangements because RecA-independent tandem repeat deletion (unassociated with transposons) is also enhanced in uup mutants (32). Some tex mutants (including those inactivating the dam mutHLS uvrD- dependent mismatch repair pathway) appear to be somewhat specific to transposon deletion, by stabilizing a mismatched stem-loop structure formed by the long inverted repeats of the transposon (32), which is stimulatory to the excision events. Other tex genes may have more general effects on deletion, including ssb [singlestrand DNA (ssDNA) binding protein], topA (topoisomerase I), polA (DNA polymerase I), and recBC (subunits of Exonuclease V). Increased persistence of ssDNA during replication (as in ssb or polA mutants) may encourage slipped mispairing (described below). Mutations in topoisomerase I may increase the probability of secondary structure formation of the inverted repeats by elevating supercoiling density. Inactivation of the potent RecBCD exonuclease may encourage rearrangements by stabilizing broken DNA molecules (see below). The role of special sequences in promoting high-efficiency RecA-independent recombination is largely unexplored. Because many of the models for RecA-independent recombination invoke strand displacement during DNA replication, replication stall sites might be imagined to be hotspots for deletion. In polymerase I mutants, CTGG and GTGG sequences are found to flank sites of spontaneous deletion (33). The Tus兾Ter system that arrests replication helicase movement evokes nearby RecAindependent deletion events at short homologies (34). Inverted repeat sequences with the potential for secondary structure formation promote deletion at flanking short sequence direct repeats (29, 35–39). (Deletion associated with inverted repeats will be discussed specifically in a section below.) RecAindependent deletion of tandem repeats has been primarily studied on ColE1-based plasmids for experimental ease, but has also been observed for tandem repeats on the E. coli chromosome (18, 22). Chromosome context effects on deletion rates have been noted with the repeats present on a ColE1-derived plasmid, an F factor, as a lambda prophage or integrated into lacZ of E. coli (40), but the molecular basis of these effects is elusive. Constraints on High-Efficiency RecA-Independent Recombination. Be-

cause a mutation in RecA blocks most other homologous recombination events, what special features allow tandem repeats to rearrange at high frequency in the absence of RecA? The homology dependence suggests that strand pairing is required for RecA-independent recombination, as it is for RecAdependent recombination. However, high-frequency RecAindependent rearrangements always involve interactions between homologies that can take place intramolecularly—in contrast, recombination between homologies on separate DNA molecules, which must occur intermolecularly, exhibits a strong dependence on RecA. Moreover, proximity of the repeat sequences clearly plays an essential role in RecA-independent recombination. There is an exponential decrease in deletion rate as the distance between the homologies increases (11, 21, 41). This contrasts to RecA-dependent recombination of tandem repeats that shows little proximity effect (21). Such high dependence on sequence proximity has supported the idea that many of these RecA-independent rearrangements occur in the context of the replication fork. The requirement of RecA as a strandpairing protein may be obviated by the proximity of the interacting sequences as well as the single-stranded nature of DNA in the fork. In E. coli, the average Okazaki fragment is 1–2 kb in length (42), which might indicate that several kb should be the upper limit for the distance between sequences that can undergo efficient RecA-independent recombination. Bzymek and Lovett

Support for a Replicative Mechanism. Several genetic properties of

The Replication ‘‘Slippage’’ Model. An attractive model for RecAindependent recombination is slipped misalignment of the nascent DNA strand during replication (Fig. 1). Originally proposed to explain frameshift mutations in nucleotide sequence repeats (47), this model has been invoked for larger repeat rearrangements (48, 49). However, for these longer-range interactions over hundreds or thousands of bases, it is difficult to imagine that Bzymek and Lovett

Fig. 1. Replication misalignment (‘‘slippage’’) model for genetic rearrangements. A slipped alignment of the nascent strain with respect to template can generate deletion or expansion of a directly repeated sequence and any intervening DNA.

a simple polymerase ‘‘skip’’ accounts for the rearrangement. Rather, it seems likely that the polymerase must dissociate from its template, allowing the nascent strand to dissociate and translocate to a new pairing position. Resumption of DNA synthesis at the mispaired position then accomplishes the deletion or expansion event. Replication slippage in vitro during replication of hairpin structures is correlated with the inability of various polymerases to mediate strand displacement and therefore their tendency to be blocked by the hairpin (50). In vivo, active displacement of the nascent strand by helicase action may aid the slipped realignment process, although no specific helicase in E. coli has yet been assigned this role. The replication slippage model explains the homology and proximity dependence of RecA-independent rearrangements and why most repeat rearrangements are not reciprocal—that is, deletions are not formed concurrent with expansions. Its intimate association with the replication process explains the hyperrecombinogenic nature of many DNA replication mutants, for which polymerization may be less processive and ssDNA may tend to accumulate. It is likely that many misalignments involve the 3⬘ nascent strand end because mutations in the 3⬘ exonucleases, ExoI and DnaQ, substantially elevate RecA-independent deletion (13, 23, 24). However, it is also possible that some slipped misalignments may involve the 5⬘ end of the Okazaki fragments, as has been proposed in eukaryotes to account for the high frequency of duplication mutations seen in yeast rad27 mutants that fail to process Okazaki 5⬘ ends (51). The comparable Okazaki processing function in E. coli is the 5⬘ exonuclease activity of DNA polymerase I (42). Sister-Chromosome Exchange (SCE)-Associated Slipped Misalignment Model. The simple slipped misalignment model, despite its

pleasing features, cannot account for all of the observed RecAindependent deletion events. A subset of RecA-independent rearrangements, selected by a decreased or increased number of tandem repeats, are associated with replicon dimerization (1, 16–18, 26, 52). These dimeric products are not consistent with simple slippage of the nascent strand on its template. However, crossing over between these circular plasmid replicons can generate circular dimers (Fig. 2). Experiments show that RecAindependent dimeric products do not result from intermolecular crossing-over, but rather from intramolecular sister-chromosome interactions (18). In our experiments selecting gain or loss of a 787-bp tandem repeat, 21% of selected expansion products and 27% of deletion products are dimeric in structure: among these dimers, 90% of expansion product dimers and 4% of deletion product dimers represent reciprocal exchanges (1, 18),

PNAS 兩 July 17, 2001 兩 vol. 98 兩 no. 15 兩 8321

COLLOQUIUM

RecA-independent recombination support the idea that these rearrangements occur during the process of chromosomal replication. In E. coli, mutations in many of the components of the DNA PolIII holoenzyme result in elevated levels of tandem repeat rearrangements (13, 22). This includes the polymerase subunit (␣, dnaE), the proofreading exonuclease (␧, dnaQ), components of the clamp-loading complex (␶ and ␥, dnaX; ␹, holC), and the processivity clamp (␤, dnaN). Mutations in the replicative helicase (dnaB) and single-strand binding protein (ssb-113) that are reported to fail to interact correctly with the polymerase (43) also cause a hyperdeletion phenotype. For dnaE, dnaQ, and holC, genetic analysis has shown that the stimulation of tandem repeat rearrangements is largely RecAindependent (ref. 13; C. J. Saveson and S.T.L., unpublished results). These observations are consistent with the notion that conditions that stall polymerization or increase the amount of ssDNA in the fork increase the probability of tandem sequence rearrangements. A switch from normal theta replication to ssDNA synthesis greatly enhances RecA-independent ‘‘nearly precise’’ transposon excision, which did not involve transfer of parental DNA to the recombinants, again supporting a replicative mechanism (44). Our study of ‘‘homeologous’’ (slightly divergent) repeat deletion also supports the idea that the majority of RecAindependent tandem repeat deletion events occur during or shortly after DNA replication (45). The mismatch repair system aborts deletion event between repeats with slight sequence dissimilarities; there is no effect of mismatch repair on deletion between perfectly homologous repeats. In the former case, the heteroduplex intermediate formed between the strands of the two repeats presents sequence mispairs that elicit mismatch repair excision, leading to destruction of the potential for deletion. This action of mismatch repair on homeologous repeats requires E. coli’s system to discriminate between nascent and parental DNA—that is, the Dam methylase that marks the parental strand and MutH endonuclease that cleaves unmethylated DNA to target excision to the nascent strand. Therefore, most deletion intermediates must be found in hemimethylated DNA, a state that exists only transiently after DNA replication. These results are concordant with the replication slippage model (see below) where the deletion arises by misalignment of the nascent DNA strand during replication. The presence or absence of the DnaB replicative helicase complex may channel subsequent processing of stalled or blocked replication forks. RecA-independent deletion of tandem repeats does not require reassembly of DnaB onto the fork via the PriA primosomal protein (C. J. Saveson and S.T.L., unpublished results); in fact, priA mutants show hyper-deletion (13). This may indicate that RecA-independent strand rearrangements occur with the DnaB complex still positioned on the fork. On the other hand, mutations that impair the DnaB fork helicase itself stimulate homologous RecA-dependent recombination disproportionately relative to the RecA-independent pathways (13). Polymerase dissociation with the fork helicase complex intact may preferentially lead to RecA-independent strand rearrangements; dissociation of the helicase complex from the replication fork may facilitate fork regression and breakage and subsequent RecA-dependent recombinational repair (46).

of its relative single-strandedness. Several studies in both prokaryotes and eukaryotes support the notion that deletions at short sequence homologies associated with inverted repeats do occur preferentially on the lagging strand (56–58). Perfect palindromes can also adopt cruciform structures in duplex DNA. One study correlated the frequency of excision with the cruciform-forming potential of several inverted repeats (37). By what mechanism do palindromic sequences stimulate RecA-independent deletion? For dispersed homologies, formation of secondary structure at the inverted repeat may serve to increase the local proximity of the direct repeats to facilitate misalignment. In addition, secondary structure can block DNA polymerization (59), increasing the probability of replication slippage events. Secondary structures may be recognized and processed by specific enzymes, creating double-strand breaks (DSBs) or single-strand nicks or gaps, which may be deletionprone substrates. Two Mechanisms for Palindrome-Stimulated Deletion: SbcCD-Dependent and SbcCD-Independent. To study the mechanisms by which

Fig. 2. Unequal crossing-over between circular molecules. Unequal crossingover between direct repeats borne on plasmids generates a circular dimer, with one triplicated and deleted allele.

with concomitant loss and gain of repeats in the two alleles carried on the dimeric product (Fig. 2). Reciprocality is considered a hallmark of ‘‘break-join,’’ conservative recombination and, again, is not consistent with simple replication slippage. Mutations in the RuvAB complex that mediate branch migration of Holliday junctions increase the proportion of RecAindependent deletion events that are recovered as reciprocal crossover dimeric products 13-fold (18), implicating a branched intermediate susceptible to RuvAB action as an intermediate in RecA-independent SCE events. A slipped misalignment model involving mispairing of the two nascent strands of the sister chromosomes (Fig. 3) has been proposed to explain these results (1, 18, 53). It has been proposed that these strand exchange events may reflect a normal fork repair reaction, a template switch gone awry because of mispairing at tandemly repeated sequences. The contribution of this pathway to normal repair of replication blocks has not been established, however, because no mutants that block this pathway have yet been isolated. Although this mechanism shares many features of the simple slipped misalignment model, namely strand displacement and mispairing in the context of a blocked replication fork, several mutations stimulate this dimer-producing pathway disproportionately, such as sbcB (Exonuclease I) and dnaE (the polymerase subunit of polIII) (13, 24). Secondary Structure Stimulation of RecA-Independent Tandem Repeat Rearrangements: A Special Case. Many spontaneous deletions

between short dispersed DNA sequence repeats are found associated with nearby inverted repeat sequences (36, 54). Studies of this phenomenon support the idea that inverted repeats greatly stimulate genetic rearrangements at short flanking direct repeats, presumably by formation of a stem-loop secondary structure (35, 37, 39, 55). Transposon excision (29) represents a well studied example of deletion between short direct repeats, stimulated by nearby inverted repeats. Secondary structure formation is expected to be more frequent on the lagging-strand template during replication because 8322 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.111008398

palindromic sequences stimulate deletion, we modified an assay for deletion between 101-bp direct repeats in the tetA gene of pBR322 to include a perfect 57-bp palindrome between the direct repeats, creating pMB301. This can be compared with a similar nonpalindromic construct, pSTL57 (11), with perfectly juxtaposed direct 101-bp repeats. Deletion as scored by this latter construct has been studied extensively in our laboratory with the conclusion that deletions arise predominantly by simple replication slippage with a minor contribution of SCE-associated slippage events (11, 24, 45, 53). In this set of constructs, no increase in the proximity of the direct repeats can be afforded by secondary structure formation between the inverted repeats and therefore any stimulation by the palindrome must be via other means. The size of the palindrome was chosen to be well below that which gives rise to marked loss (approximately 200 bp; ref. 60), and we observed no defect in copy number or maintenance of palindrome-containing constructs relative to controls (M.B. and S.T.L., unpublished results; ref. 61). The presence of the palindrome stimulated deletion formation by two to three orders of magnitude (Table 1). This stimulation was independent of RecA and more dramatic at the lower growth temperature. The natural temperature sensitivity may reflect more stable pairing to stabilize the secondary structure at low temperatures. Transformation efficiency of an incompatible plasmid into strains with resident palindrome or nonpalindrome plasmids verified that the stimulatory effect of the palindrome is not due to its relative inefficiency in competing with a newly arising deletion plasmid (M.B. and S.T.L., unpublished results; ref. 61). The most revealing aspect of our genetic analysis was the effect of the sbcD mutation: part but not all of the palindromestimulation was found to depend on the sbcD gene (Table 1). This differentiates two mechanisms by which palindrome DNA stimulates deletion formation, one SbcD-dependent and one SbcD-independent. Other palindromic constructs show this property; with these, sbcD also dramatically alters the distribution of deletion endpoints within the repeated sequences (61). The sbcD gene encodes the nuclease component of the SbcCD complex of E. coli (62). In vitro, SbcCD possesses ssDNA endonuclease and double-strand DNA (dsDNA) exonuclease activities (63, 64) and can cleave hairpin structures (65). In vivo, SbcCD appears to introduce DSBs at palindromic sequences (66, 67). A Single-Strand Annealing Mechanism for RecA-Independent Deletion Formation Associated with Palindromes. These genetic effects lead

us to propose that the SbcCD nuclease cleaves cruciform or hairpin structures formed by inverted repeat sequences—this

Bzymek and Lovett

break is then repaired by RecA-independent single-strand annealing at the repeats after resection of the ends (Fig. 4). The single-strand annealing mechanism is well documented in eukaryotes (68–70). In E. coli, this mechanism does not account for many of the properties of simple tandem repeat deletion and so most likely does not contribute substantially to deletion that is not associated with palindromes. Experiments show that expansions are produced as efficiently as deletion and share many of the same genetic properties (1). These cannot be via single-strand annealing because, by its nature, single-strand annealing can only efficiently produce deletion and not expansion rearrangements (see Fig. 4). The genetic effects on homeologous repeat deletion are also incompatible with deletion mediated by single-strand annealing but, rather, support a simple replication slippage mechanism as the predominant mode of deletion (45). Deletions Bzymek and Lovett

Table 1. Palindrome stimulation of deletion formation Deletion rate ⫻ 106 (95% confidence interval) Growth temperature

Strain genotype

pSTL57 no palindrome

37°

rec⫹ recA recA sbcD recA recB recA

5.0 (1.1–1.6) 4.7 (3.2–8.4) 3.8 (2.0–5.4) 1.7 (0.4–4.8) 1.7 (0.6–7.2)

30°

pMB301 with palindrome 425 (250–480) 220 (150–300) 40 (12–66) 24 (7.0–54) 2400 (1500–5200)

Strains used were AB1157, JC10287, STL4494, and JC5547. Mutant alleles include ⌬(recA-srl)304 (JC10287, STL4494), recA13 recB21 (JC5547), sbcD300⬋Tn10kan (STL4494). PNAS 兩 July 17, 2001 兩 vol. 98 兩 no. 15 兩 8323

COLLOQUIUM

Fig. 3. Model for SCE associated with replication slipped misalignment, yielding deletion (18). (A) A stalled replication fork with direct repeat sequences shown in black. (B) Displacement of nascent strand 3⬘ ends. (C) Misalignment of nascent ends at repeat sequences. (D) Processing of crossed strand to yield Holliday-junction structure. The structure can be processed in two alternative ways. If no realignment of parental strands occurs: (E) repair replication. (F) Junction cleavage to form a reciprocal triplication兾deletion dimeric product. If realignment of parental strands occurs: (E⬘) repair replication. (F⬘) Junction cleavage by RuvC or other resolving activity to form a nonreciprocal deletion兾duplication dimeric product. For deletion, this outcome is favored over that shown in F. In ruvAB mutants, F⬘ is favored, yielding a higher level of reciprocal products.

Fig. 4. Single-strand annealing model for deletion associated with palindromes. Palindromic DNA can adopt cruciform structure. SbcCD initiates cleavage of the cruciform. Resection followed by annealing at the flanking direct repeats generates a deletion.

between repeats containing mismatched bases are aborted by the methyl-directed mismatch repair (MDMR) pathway, dependent on strand discrimination via Dam methylase and MutH endonuclease (45). For the slipped mispairing model, MutHdependent incision and targeted excision of the nascent slipped strand by MDMR (the top strands in Fig. 5) explains the loss of potential deletion products. In the absence of methylation for strand discrimination, excision of the parental template strand should occur, causing fixation of the deletion; deletion rates of mismatched repeats are indeed greatly elevated by loss of the Dam methylase (45). For the single-strand annealing model (Fig. 5), it is difficult to reconcile destabilization of deletion intermediates dependent on both methylation and the MutH endonuclease (which cleaves at unmethylated GATC sites). Although mismatch excision may destabilize the annealing intermediate (Fig. 5), mismatch repair excision proteins should be able to access the DNA from the broken ends, without requiring MutH incision. Furthermore, there is no good explanation for the rescue of deletion products by the absence of methylation. In E. coli single-strand annealing is inefficient, most likely because of the rampant DNA degradation by the RecBCD enzyme. For example, healing of linear DNA, either produced in vitro or introduced by transformation, at terminal repeats is inefficient unless cells are RecBCD⫺ (refs. 71 and 72; Table 2). RecB and RecC mutations are among the tex mutations that elevate deletion of transposons from the E. coli chromosome (27, 28). The plasmids used in our study lack Chi sites to attenuate RecBCD nuclease (73) and so should be especially prone to RecBCD-mediated degradation. Although mutation abolishing RecBCD nuclease elevated by 10-fold the healing of transformed broken DNA with terminal direct repeats (Table 2), it did not elevate deletion promoted by SbcCD on palindromic resident plasmids (Table 1). If anything, deletion formation may be reduced in recB mutants, both for palindrome and the nonpalindrome-containing constructs. This may implicate 8324 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.111008398

Fig. 5. Dissolution of homeologous deletion intermediates by mismatch repair. (A) Slipped misalignment model. MutH incision of the unmethylated strand removes the slipped strand, thereby aborting the deletion event. In the absence of Dam methylation, the ability to target excision to the slipped deletion strand is lost, resulting in the observed elevation of homeologous deletion rates (45). (B) Single-strand annealing model. Mismatch excision of either DNA strand destabilizes the intermediate. However, preexisting nicks should obviate the need for MutH. There is likewise no reason to explain the stabilization of the intermediate when Dam methylase is lacking.

RecBCD, either directly or indirectly, in the generation of the deletion event. Part of this reduction may be due to cellular inviability (74), or plasmid loss in this strain because rates are reduced 3-fold for the nonpalindromic control plasmid. Mutations in recBC also increase rolling-circle replication, causing plasmid DNA to accumulate in a linear form (75). Because cruciform extrusion is facilitated by negative supercoiling, cruciform formation should be diminished in this linear state. The lack of any RecBCD inhibitory effect on palindromestimulated deletion may be because SbcCD remains bound to the ends of the DSB and protects it from complete degradation by RecBCD. Potentially dimerizing through the coiled-coil domain Table 2. DSB repair after transformation of linear pSTL57 Strain genotype recA recA sbcD recA recB

Repair efficiency* ⫻ 104

% deleted (tet⫹)†

1.5 1.1 11.0

100 82 88

Strains and genotypes are listed in Table 1. *Repair efficiency is the number of transformants for the linear pSTL57 molecule relative to those for uncleaved pBR322. †Percent of ampicillin-resistant transformants that are tetracycline-resistant, indicating deletion at terminal repeats.

Bzymek and Lovett

1. Morag, A. S., Saveson, C. J. & Lovett, S. T. (1999) J. Mol. Biol. 289, 21–27. 2. Kowalczykowski, S. C. (2000) Trends Biochem. Sci. 25, 156–165. 3. Farabaugh, P. J., Schmeissner, U., Hofer, M. & Miller, J. H. (1978) J. Mol. Biol. 126, 847–857. 4. Krawczak, M. & Cooper, D. N. (1991) Hum. Genet. 86, 425–441. 5. Meuth, M. (1989) in Mobile DNA, eds. Berg, D. E. & Howe, M. M. (Am. Soc. Microbiol., Washington, DC), pp. 833–860. 6. Hu, X. & Worton, R. G. (1992) Hum. Mutat. 1, 3–12. 7. Lestienne, P. & Bataille, N. (1994) Biomed. Pharmacother. 48, 199–214. 8. McMurray, C. T. (1995) Chromosoma 104, 2–13. 9. Bachmann, B. J. (1996) in Escherichia coli and Salmonella: Cellular and Molecular Biology, ed. Neidhardt, F. C. (Am. Soc. Microbiol., Washington, DC), Vol. 2, pp. 2460–2488. 10. Miller, J. H. (1992) A Short Course in Bacterial Genetics (Cold Spring Harbor Lab. Press, Plainview, NY). 11. Lovett, S. T., Gluckman, T. J., Simon, P. J., Sutera, V. A., Jr., & Drapkin, P. T. (1994) Mol. Gen. Genet. 245, 294–300. 12. Lea, D. E. & Coulson, C. A. (1949) J. Genet. 49, 264–285. 13. Saveson, C. J. & Lovett, S. T. (1997) Genetics 146, 457–470. 14. Dower, W. J., Miller, J. F. & Ragsdale, C. W. (1988) Nucleic Acids Res. 16, 6127–6145. 15. Willetts, N. S., Clark, A. J. & Low, B. (1969) J. Bacteriol. 97, 244–249. 16. Dianov, G. L., Kuzminov, A. V., Mazin, A. V. & Salganik, R. I. (1991) Mol. Gen. Genet. 228, 153–159. 17. Mazin, A. V., Kuzminov, A. V., Dianov, G. L. & Salganik, R. I. (1991) Mol. Gen. Genet. 228, 209–214. 18. Lovett, S. T., Drapkin, P. T., Sutera, V. A., Jr., & Gluckman-Peskind, T. J. (1993) Genetics 135, 631–642. 19. Matfield, M., Badawi, R. & Brammar, W. J. (1985) Mol. Gen. Genet. 199, 518–523. 20. Low, B. (1968) Proc. Natl. Acad. Sci. USA 60, 160–167. 21. Bi, X. & Liu, L. F. (1994) J. Mol. Biol. 235, 414–423. 22. Bierne, H., Vilette, D., Ehrlich, S. D. & Michel, B. (1997) Mol. Microbiol. 24, 1225–1234. 23. Allgood, N. D. & Silhavy, T. J. (1991) Genetics 127, 671–680. 24. Bzymek, M. & Lovett, S. (1999) J. Bacteriol. 181, 477–482. 25. Whoriskey, S. K., Schofield, M. A. & Miller, J. H. (1991) Genetics 127, 21–30. 26. Yi, T.-M., Stearns, D. & Demple, B. (1988) J. Bacteriol. 170, 2898–2903. 27. Lundblad, V. & Kleckner, N. (1982) in Molecular and Cellular Mechanisms of Mutagenesis, eds. Lemontt, J. F. & Generoso, W. M. (Academic, New York), pp. 245–258. 28. Lundblad, V. & Kleckner, N. (1985) Genetics 109, 3–19. 29. Foster, T. J., Lundblad, V., Hanley-Way, S., Halling, S. M. & Kleckner, N. (1981) Cell 23, 215–227. 30. Hopkins, J. D., Clements, M. & Syvanen, M. (1983) J. Bacteriol. 153, 384–389. 31. Reddy, M. & Gowrishankar, J. (1997) J. Bacteriol. 179, 2892–2899. 32. Reddy, M. & Gowrishankar, J. (2000) J. Bacteriol. 182, 1978–1986. 33. Jankovic, M., Kostic, T. & Savic, D. J. (1990) Mol. Gen. Genet. 223, 481–486. 34. Bierne, H., Ehrlich, S. D. & Michel, B. (1997) EMBO J. 16, 3332–3340. 35. Albertini, A. M., Hofer, M., Calos, M. P., Tlsty, T. D. & Miller, J. H. (1982) Cold Spring Harbor Symp. Quant. Biol. 47, 841–850. 36. Glickman, B. W. & Ripley, S. (1984) Proc. Natl. Acad. Sci. USA 81, 512–516. 37. Sinden, R. R., Zheng, G., Brankamp, R. G. & Allen, K. N. (1991) Genetics 129, 991–1005.

Bzymek and Lovett

Summary. Evidence supports the existence of three RecA-

independent mechanisms distinguished by their genetic properties that contribute to genetic rearrangements in E. coli: simple replication slippage, SCE-associated replication misalignment, and single-strand annealing. For rearrangements between large sequence homologies, these mechanisms can be comparable in their efficiency to RecA-dependent homologous recombination. RecA-independent processes contribute exclusively to rearrangements between short sequence homologies but are constrained to intramolecular rearrangements over fairly short distances. The single-strand annealing pathway, identified for E. coli only recently, appears to contribute primarily to deletions associated with palindromic sequences and depends on the SbcCD nuclease, presumably to introduce a DSB at the hairpin structure formed by the inverted repeats, and possibly to protect the broken ends. Difficulties in DNA replication contribute to sequence rearrangements via all four mechanisms. This work was supported by the General Medical Institute of the National Institutes of Health Grant RO1 GM51753 and by Predoctoral Training Grant T32 GM07122. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75.

Canceill, D. & Ehrlich, S. D. (1996) Proc. Natl. Acad. Sci. USA 93, 6647–6652. Weston-Hafer, D. & Berg, D. E. (1989) Genetics 121, 651–658. Kazic, T. & Berg, D. E. (1990) Genetics 126, 17–24. Chedin, F., Dervyn, E., Dervyn, R., Ehrlich, S. D. & Noirot, P. (1994) Mol. Microbiol. 12, 561–570. Kornberg, A. & Baker, T. A. (1992) DNA Replication (Freeman, New York). Chase, J., L’Italien, J., Murphy, J., Spicer, E. & Williams, K. (1984) J. Biol. Chem. 259, 805–814. d’Alenc¸on, E., Petranovic, M., Michel, B., Noirot, P., Aucouturier, A., Uzest, M. & Ehrlich, S. D. (1994) EMBO J. 13, 2725–2734. Lovett, S. T. & Feschenko, V. V. (1996) Proc. Natl. Acad. Sci. USA 93, 7120–7124. Michel, B. (2000) Trends Biochem. Sci. 25, 173–178. Streisinger, G., Okada, Y., Emrich, J., Newton, J., Tsugita, A., Terzaghi, E. & Inouye, M. (1967) Cold Spring Harbor Symp. Quant. Biol. 31, 77–84. Efstratiadis, A., Psalony, J. W., Maniatis, T., Lawn, R. M., O’Connell, C., Spritz, R. A., DeRiel, J. K., Forget, B. G., Weissman, S. M., Slightom, J. L., et al. (1980) Cell 21, 653–668. Albertini, A. M., Hofer, M., Calos, M. P. & Miller, J. H. (1982) Cell 29, 319–328. Canceill, D., Viguera, E. & Ehrlich, S. D. (1999) J. Biol. Chem. 274, 27481–24790. Tishkoff, D. X., Filosi, N., Gaida, G. M. & Kolodner, R. D. (1997) Cell 88, 253–263. Bi, X., Lyu, Y. L. & Liu, L. F. (1995) J. Mol. Biol. 247, 890–902. Feschenko, V. V. & Lovett, S. T. (1998) J. Mol. Biol. 276, 559–569. Galas, D. J. (1978) J. Mol. Biol. 126, 858–863. Pierce, J. C., Kong, D. & Masker, W. (1991) Nucleic Acids Res. 14, 3901–3905. Trinh, T. Q. & Sinden, R. R. (1991) Nature (London) 352, 544–547. Rosche, W. A., Trinh, T. Q. & Sinden, R. R. (1995) J. Bacteriol. 177, 4385–4391. Gordenin, D. A., Malkova, A. L., Peterzen, A., Kulikov, U. N., Paviov, Y. I., Perkins, E. & Resnick, M. A. (1992) Proc. Natl. Acad. Sci USA 89, 3785–3789. LaDuca, R. J., Fay, P. J., Chuang, C., McHenry, C. S. & Bambara, R. A. (1983) Biochemistry 22, 5177–5188. Warren, G. J. & Green, R. L. (1985) J. Bacteriol. 161, 1103–1111. Bzymek, M. & Lovett, S. T. (2001) Genetics 158, in press. Sharples, G. J. & Leach, D. R. (1995) Mol. Microbiol. 17, 1215–1217. Connelly, J. C. & Leach, D. R. (1996) Genes Cells 1, 285–291. Connelly, J. C., de Leau, E. S., Okely, E. A. & Leach, D. R. (1997) J. Biol. Chem. 272, 19819–19826. Connelly, J. C., Kirkham, L. A. & Leach, D. R. (1998) Proc. Nat. Acad. Sci. USA 95, 7969–7974. Leach, D. R., Okely, E. A. & Pinder, D. J. (1997) Mol. Microbiol. 26, 597–606. Cromie, G. A., Millar, C. B., Schmidt, K. H. & Leach, D. R. (2000) Genetics 154, 513–522. Fishman-Lobell, J., Rudin, N. & Haber, J. E. (1992) Mol. Cell. Biol. 12, 1291–1303. Lin, F.-L. M., Sperle, K. & Sternberg, N. (1990) Mol. Cell. Biol. 10, 113–119. Maryon, E. & Carroll, D. (1991) Mol. Cell. Biol. 11, 3278–3287. Lovett, S. T., Luisi-DeLuca, C. A. & Kolodner, R. D. (1988) Genetics 120, 37–45. Luisi-DeLuca, C., Lovett, S. T. & Kolodner, R. D. (1989) Genetics 122, 269–278. Dixon, D. A. & Kowalczykowski, S. C. (1993) Cell 73, 87–97. Capaldo-Kimball, F. & Barbour, S. D. (1971) J. Bacteriol. 106, 204–212. Cohen, A. & Clark, A. J. (1986) J. Bacteriol. 167, 327–335.

PNAS 兩 July 17, 2001 兩 vol. 98 兩 no. 15 兩 8325

COLLOQUIUM

of the SbcC protein (62), the SbcCD complex may also bind both broken ends simultaneously, facilitating their annealing interaction. A mutation in sbcD did not affect healing of transformed linear DNA with terminal repeats (Table 2), suggesting that SbcCD’s role in palindrome-stimulated deletion must include the initiation of the break. SbcCD-initiated DSBs may be special in that they can heal by RecA-independent single-strand annealing if there are homologies flanking the break, thus resulting in a deletion mutation. The homology and proximity-dependence of this SbcCD-dependent single-strand annealing mechanism in E. coli remains to be determined. If the RecA-dependent recombination pathways are available, SbcCD-initiated DSBs can also be repaired by nonmutagenic recombination with the sister-chromosome (66, 67). The SbcCD-independent form of palindrome-stimulated deletion, we believe, represents simple replication slippage promoted by the hairpin’s block to replication. Analysis of deletion endpoints with respect to replication direction is consistent with these blocks occurring predominantly on the lagging strand (61).