Stabilization of Diverged Tandem Repeats by Mismatch Repair: Evidence for Deletion Formation Via a Misaligned Replication Intermediate Susan T

Proc. Natl. Acad. Sci. USA Vol. 93, pp. 7120-7124, July 1996 Genetics

Stabilization of diverged tandem repeats by mismatch repair: Evidence for deletion formation via a misaligned replication intermediate SusAN T. LovErr* AND VLADIMIR V. FESCHENKO Department of Biology and Rosenstiel Basic Medical Sciences Center, Brandeis University, Waltham, MA 02254-9110 Communicated by Paul Modrich, Duke University Medical Center, Durham, NC, April 1, 1996 (received for review January 11, 1996)

ABSTRACT A functional methyl-directed mismatch re- initiated when replication is blocked as part of a recombina- pair pathway in Escherichia coli prevents the formation of tional repair reaction (8). Once displaced, the nascent strand deletions between 101-bp tandem repeats with 4% sequence may choose an incorrect pairing partner if the stall occurs in divergence. Deletions between perfectly homologous repeats the context of a repetitive sequence array. are unaffected. Deletion in both cases occurs independently of Experimental evidence for the role of replication in rear- the homologous recombination gene, recA. Because the meth- rangements between tandem repeats is sparse. Mutations yl-directed mismatch repair pathway detects and excises one affecting DNA polymerase III ofE. coli (Catherine J. Saveson strand of a mispaired duplex, an intermediate for RecA- and S.T.L., unpublished results) and DNA polymerases a and independent deletion of tandem repeats must therefore be a 5 of Saccharomyces cerevisiae (15) elevate deletion formation heteroduplex formed between strands of each repeat. We find between tandem direct repeats. Although this suggests that that MutH abnormal replication stimulates deletion, it is still possible that endonuclease, which in vivo incises specifically the normally, in a strain with intact DNA polymerases, deletion newly replicated strand of DNA, and the Dam methylase, the occurs nonreplicatively. The best evidence for replicational source of this strand-discrimination, are required absolutely deletion formation comes from studies where the efficiency of for the exclusion of "homeologous" (imperfectly homologous) deletion between short direct repeats is dependent on the tandem deletion. This supports the idea that the heteroduplex direction of replication (16, 17). In these studies, a palindromic intermediate for deletion occurs during or shortly after DNA sequence was asymmetrically placed between the direct re- replication in the context of hemi-methylation. Our findings peats such that a slipped alignment could be stabilized by confirm a "replication slippage" model for deletion formation hairpin-formation by only one of the two template strands. whereby the displacement and misalignment of the nascent These studies have suggested that replication-dependent de- strand relative to the repeated sequence in the template strand letion between direct repeats may occur preferentially during accomplishes the deletion. lagging-strand synthesis. However, other similar constructs detecting deletion between short repeats in E. coli have failed Tandemly repeated sequences are genetically unstable and to show effects of leading versus lagging-strand synthesis on subject to expansion or deletion. These rearrangements are of deletion efficiency (18). The universality of the role of DNA interest because they represent an important source of human replication in tandem repeat instability is therefore unclear. In genetic disease (1-3). Although the processes of repeat ex- addition, an effect of replication direction was not seen for pansion and deletion are similar in theory, deletion formation direct repeats that lack an associated inverted-repeat structure has been subjected to more extensive molecular In (16). A replication misalignment mechanism could be, in analysis. the theory, specific to deletions associated with inverted repeats bacterium Escherichia coli, many of the features that govern because more distant sequences may require hairpin formation deletion efficiency have been elucidated. In general, deletion to juxtapose the repeats for the slipped misalignment to occur. frequencies increase with the length of homology within the Hairpin formation may be more efficient on the lagging strand repeats (4-7). Despite this dependence on homology, the because of its single-stranded nature. homologous recombination gene, recA, is not required for We have examined in E. coli the deletion of tandem 101-bp deletion to occur (5-8). For a given length repeat, closely repeats that lack inverted repeat structures, with the repeats juxtaposed repeats delete more frequently than those more either perfectly homologous or with 4% heterology. The distant (7, 9-11). deletion of the imperfectly homologous repeats is greatly Streisinger et al. (12) proposed a model of "replication inhibited by a functional methyl-directed mismatch repair slippage" to explain addition or loss of very short repetitive system. This suggests that virtually all detected deletions occur sequences during DNA replication which lead to frameshift via a heteroduplex intermediate composed of one strand from mutations; this model has been extended to more distant each repeat. The impact of the mismatch repair pathway in this repeat rearrangements in both E. coli (13) and humans (14). system also allows us to conclude that the majority of detected The nascent DNA strand realigns relative to its template at a deletion events occur during or shortly after DNA replication. second repeat (see Fig. 1), producing a loop in either the Our observations are in support of replication realignment template (leading to deletion) or the nascent strand (leading models for tandem repeat deletion. to expansion). Although small loops that would accompany frameshift mutation realignments may be readily accommo- MATERIALS AND METHODS dated by the replication complex, it is more difficult to imagine spontaneous dislocations of the nascent strand over many Bacterial Strains and Media. The strains used in this study hundreds of bases that would be required to explain deletions are derived from AB1157 [F- thi-1 hisG4 A(gpt-proA)62 argE3 between more distant sequences. We have proposed that displacement of the nascent strand may be an active process, Abbreviations: Tc, tetracycline; Tcr, Tc resistant; Ap, ampicillin; Apr, Ap resistant; LB, Luria-Bertani. *To whom reprint requests should be addressed at: Department of The publication costs of this article were defrayed in part by page charge Biology and Rosenstiel Basic Medical Sciences Center, 415 South payment. This article must therefore be hereby marked "advertisement" in Street, Brandeis University, Waltham, MA 02254-9110. e-mail: accordance with 18 U.S.C. §1734 solely to indicate this fact. [email protected]. 7120 Downloaded by guest on September 24, 2021 Genetics: Lovett and Feschenko Proc. Natl. Acad. Sci. USA 93 (1996) 7121 thr-1 leuB6 kdgK51 rfbDl ara-14 lacYI galK2 xyl-5 mtl-1 tsx-33 cttgcat g, and (ii) 5'-caaggagatg gcgccCaaca gtcccccggc supE44 rpsL31 rac- A- (19)]. Strains ES1582 (mutH34), cacgggGcct gccaccatac ccacgccGaa acaagcgctc atgagcccGa ES1590 (mutL25), and GM3819 (dam-16::kan) were obtained agtggcgagc cagatctcat g. The structure of the tetA repeat for from M. Marinus (University of Massachusetts Medical pSTL57, pSTL113 and pSTL115 was confirmed by DNA School, Worcester, MA); JC10287 [srl-recAA304, (20)] and sequence analysis. Oligonucleotides were purified by ethanol RDK1517 (mutS201::Tn5) were provided by R. Kolodner precipitation. Plasmid DNA was purified by the alkaline/SDS (Dana-Farber Cancer Institute, Boston). Strain STL1526 method (24) and transformed into E. coli by the Mg+2- [uvrD254::Tn5, (21)] and STL2172 (mutS201::Tn5 srl- polyethylene glycol-dimethyl sulfoxide treatment method (25) recAA304) are derived from our collection. STL2172 was or by electroporation (26). DNA sequence determination was constructed by PlvirA transduction (22) from donor RDK1517 performed by using dideoxy terminators and Sequenase 2.0 to recipient JC10287, selecting kanamycin-resistance, UV- (Amersham). sensitivity, and mutator phenotypes. Strains were grown rou- Deletion Assays. Deletion frequencies were measured by tinely in Luria-Bertani (LB) medium (23), with plate media determining the number of Tcr cells in the ampicillin-resistant containing 1.5% agar. Tetracycline (Tc), kanamycin, and (Apr) population for several independent cultures, as de- ampicillin (Ap) were used at concentrations of 15, 30 and, 100 scribed (8). Transformants were streaked on LB + Ap plates /Lg/ml, respectively. Mutator phenotypes were assayed by to obtain single colonies. Entire fresh overnight colonies were mutation to rifampin-resistance (at 100 ,ug/ml). then resuspended and grown in 1 ml LB + Ap broth for 2 h. Plasmids. The deletion assay plasmid pSTL57 is a derivative Serial dilution in 56/2 buffer (23) were spotted on LB + Ap of plasmid pBR322 which confers ampicillin-resistance and and LB + Ap + Tc plates and incubated overnight at 37°C. carries a perfect 101-bp duplication within the tetA (tetracy- Deletion rates were calculated by the method of the median cline-resistance) gene (10). Two additional deletion assay (27) using the formula: deletion rate = M/N, where M is the plasmids, pSTL113 and pSTL115, were constructed to carry calculated number of deletion events and N is the final average imperfect repeats with silent base substitutions at four posi- number of Apr cells in the 1 ml cultures. M is solved by tions within one of the 101-bp repeats. In these plasmids, the interpolation from experimental determination of ro, the me- repeats are immediately juxtaposed with no separating non- dian number ofTcr cells determined among the cultures, by the homologous DNA. Plasmid pSTL113 carries the mutations in formula ro= M(1.24 + lnM). the tetA downstream repeat; plasmid pSTL115 carries the mutations in the upstream repeat. Plasmid pSTL113 was RESULTS constructed by the ligation of two annealed synthetic 101-bp oligonucleotides into the SphI site of pBR322: (i) 5'-agatctggct Previously, we had created a assay plasmid for deletion, cgccacttTg ggctcatgag cgcttgtttT ggcgtgggta tggtggcagg Tcccgt- pSTL57, by exactly duplicating a 101-bp internal segment of ggcc gggggactgt tAggcgccat ctccttgcat g, and (ii) 5'-caaggagatg tetA on plasmid pBR322 (10). Deletion of the repeat in this gcgccTaaca gtcccccggc cacgggAcct gccaccatac ccacgccAaa plasmid can be scored by restoration of Tc resistance to the acaagcgctc atgagcccAa agtggcgagc cagatctcat g. (The muta- cell. Deletion detected in this system is frequent and occurs tional changes relative to tetA + are shown above in uppercase independently of the homologous recombination gene, recA. letters.) Plasmid pSTL114 was a selected tetracycline-resistant We made a similar construct, pSTL113, by introducing four (Tcr) deletion product of pSTL113 which has inherited in its silent transition mutations in one of the two 101-bp repeats. single tetA locus the four silent base-substitution mutations (as These changes are evenly spaced, at 21 bp apart, and create 4% confirmed by DNA sequence analysis). Plasmid pSTL115 was heterology between the tandem repeats. If during the deletion constructed by the ligation of two annealed synthetic 101-bp process heteroduplex DNA is formed between DNA strands oligonucleotides with the wild-type tetA sequence into the SphI derived from each repeat, the pSTL113 heteroduplex will site of pSTL114. The sequences of these oligonucleotides were contain mismatched base pairs. Such heteroduplex DNA is as follows: (i) 5'-agatctggct cgccacttCg ggctcatgag cgcttgtttC predicted by mechanisms that propose displacement and pair- ggcgtgggta tggtggcagg Ccccgtggcc gggggactgt tGggcgccat ctc- ing of the two repeats, as illustrated in the slipped misalign- Replication realignment leading to: A. Deletion of tandem repeats B. Expansion of tandem repeats ,Nascent strand Nascent strand - - :.: . :::::::-.:::'.- :" ::::: ::::: : .:....s: .:::. k-Template strand -Template strand

Slipped alignment I,..i :. 4 # Slipped alignment

Deletion product 4 Expansion product .-.. -_> -Parental duplicat .:.:.-.-..+:::-:.::.:::. . ::: a d..-- --" ::...'.''''::::::I :::::::::: P::::ar::::ld::::::::n:::..-. Parental duplication Parental duplication FIG. 1. The replication slippage model showing misalignments that lead to deletion (A) or expansion (B) of tandem repeats. (A) Slipped alignment of the nascent strand with its template, producing a loop between the repeats on the template, will result in a deletion after completion of replication. (B) Slipped alignment of the nascent strand with its template, producing a loop between repeats on the nascent strand, will result in an expansion product after completion of replication. Downloaded by guest on September 24, 2021 7122 Genetics: Lovett and Feschenko Proc. Natl. Acad. Sci. USA 93 (1996) ment model in Fig. 1. The pSTL57 heteroduplex, however, would be perfectly paired. If the slipped misalignment model for deletion is correct, our expectation is that the DNA mispairs present in misaligned heteroduplex derived from pSTL113 should initiate a repair reaction leading to excision of the misaligned strand (Fig. 2). We would predict that the operation of the methyl-directed mismatch repair pathway would exclude many deletion events 0 ixio-5- 1 t involving "homeologous" (imperfectly homologous) repeats in pSTL113. As the repeats in pSTL57 are completely homologous, no effect of the mismatch repair system is expected. We examined the deletion efficiency for these two assay plasmids in wild-type E. coli and strains carrying mutations affecting the methyl-directed mismatch repair pathway (Fig. 3). The plasmid with perfectly homologous repeats, pSTL57, mut+ mutS mutL mutH uvrD dam exhibited similar deletion rates in every background tested. Genotype [Deletion rate may be slightly elevated (2-fold) in uvrD mutants.] The homeologous repeat construct with 4% heterology, FIG. 3. Deletion rates in various mismatch repair mutants for pSTL113, showed much lower deletion rates in wild-type plasmid pSTL57 (black bar, homologous tetA repeats) or pSThL13 strains by several orders of magnitude. Mutations in mutS, (gray bar, homeologous tetA repeats). Deletions were scored by the appearance of Tcr cells in the Apr population. Deletion rates were calculated by the method of the median as described for 8-18 Displacement of Homneologous repeats nascent strand independent cultures. M\ >1 .x>-1* S4 m n I mutL, mutH, and dam increased its deletion rate dramatically. + + + X m m m m I The efficiency of deletion of the homeologous repeats, how- 4 CH3 ever, never reached the value of the perfectly homologous + + + repeats and remained from 3- to 8-fold less efficient. This may be because the 4% heterology disrupts the initial pairing of the Slipped alignment misaligned intermediate. A mutation in uvrD showed less of a restorative effect on deletion efficiency of pSTL113 than the

other mutations affecting methyl-directed... ..mismatch. repair. The deletions formed from both the homologous and ho- 4Heteroduplex marker meologous repeats in both mut+ and mutS backgrounds occur + + + m independently of a functional RecA protein (Fig. 4). Switching m the position of the silent mutations to the upstream repeat, in + m / 9 plasmid pSTL115, did not change the results appreciably from that obtained previously for pSThL13, with the mutations in + m the downstream tetA repeat.

MutHLS Recogniton of mismatch UvrD (heioase 11) DISCUSSION and exacsion Single-strand speodic exonuclease The methyl-directed mismatch repair pathway effectively ex- cludes RecA-independent recombination between tandem m m I -CH3

Aborted deletion event i 4.~~~~~~~~~~~~~~~~~~~~~~~~1-1 m m m M 1i~ CH3

FIG. 2. Model for mismatch repair exclusion of homeologous repeat deletion. After partial replication of the repeat structure, we propose that the nascent strand is displaced and pairs with the downstream repeat of the template strand. After this "slippage," the DNA polymerase (Pol) extends the nascent strand as shown. This slipped realignment produces a mispaired heteroduplex marker (de- noted +/m) between strands of the homeologous repeats. Note that the extent of heteroduplex will depend on how much of the upstream recA recA muItS repeat has been replicated before displacement and the extent of this Genotype displacement of the nascent strand. In addition, the loop structure can migrate either to the left, extending the region of heteroduplex, or to FIG. 4. Deletion rates for pSTL57 (black bar, homologous tetA the right, reducing the region of heteroduplex. If a mismatch in the repeats), pSTL113(1ight gray bar, homeologous repeats with mutations heteroduplex region is recognized by the methyl-directed mismatch in the tetA downstream repeat) and pSTL115 (dark gray bar, ho- repair system, incision and subsequent excision of the unmethylated meologous repeats with mutations in the tetA upstream repeat). Both nascent strand can dissociate the slipped alignment and thereby abort recA and recA mutS mutant strains were analyzed. Deletions were the deletion event. (We have illustrated incision 3' to the mispair, but scored by the appearance of Tcr cells in the Apr population. Deletion this is arbitrary and could be 5' as well.) Re-replication without rates were calculated by the method of the median as described for slippage will restore the original repeat structure. 10-16 independent cultures. Downloaded by guest on September 24, 2021 Genetics: Lovett and Feschenko Proc. Natl. Acad. Sci. USA 93 (1996) 7123 repeats containing as little as 4% heterology. Therefore, slight strand during replication of the first repeat and subsequent sequence divergence may function to prevent rearrangements mispairing downstream on its template produces a loop on the among tandem repeat arrays. In addition, these results offer template and heteroduplex DNA within the newly paired some insight into the mechanism by which these deletions are region. The extent of the heteroduplex will be determined by formed. several factors: the amount of the repeat replicated before Our observations confirm the existence of a heteroduplex displacement, the amount of DNA displaced and then paired, intermediate for RecA-independent deletion formed between and migration of the loop stricture once the misalignment is DNA strands of the two repeats, as predicted by the slipped made. misalignment models for deletion (Figs. 1 and 2). The fact that Cleavage by MutH endonuclease of the nascent strand, if a mismatch repair inhibits at least 99% of the deletion events mispair is present in the slipped misalignment, makes the between homeologous repeats suggests that virtually all dele- deletion strand vulnerable to excision while preserving the tions in this system occur via a heteroduplex intermediate. If methylated template strand. Therefore, if the slipped misalign- misalignment occurs throughout this region, the heteroduplex ment occurs between homeologous repeats, the process of must span at least 21 bases in length, the distance between mismatch excision and re-replication of the excision gap (as heterologies in pSTL113. Second, because mutations in mutH depicted in Fig. 2) could abort the deletion event, as we indeed endonuclease and dam methylase remove this inhibition to observed. To account for the severe reduction in homeologous homeologous deletion, the slipped heteroduplex must consist deletion rates, excision must be extensive enough to destabilize of hemi-methylated DNA, formed during or shortly after DNA the slipped configuration in virtually every case. (In theory, replication. excision could also remove the mispair but not destabilize the The methyl-directed pathway in E. coli is responsible for slipped alignment.) Although we have illustrated in Fig. 2 removing the bulk of errors that arise during DNA replication excision 3' to the mismatch, this is arbitrary; we do not know and is especially sensitive to transition mispairs (reviewed in which of the GATC sites in the deletion assay plasmids is used ref. 28), a common replication error and the type predicted to for the incision which aborts the deletion event. Note that the be present in our pSTL113 heteroduplex. The events during looped structure formed by the slipped alignment (present in methyl-directed mismatch repair have been elucidated by both the perfect and the homeologous repeat intermediates) numerous genetic and biochemical experiments (reviewed in does not itself elicit the repair reaction. This is consistent with ref. 28 and summarized below). The Dam protein methylates observations that mismatch repair does not recognize loops of adenine residues in GATC sequences postreplicatively. For a 4 bases or greater (33, 34). Our results also imply that mismatch few minutes after DNA replication, there is transient hemi- excision cannot be initiated from the nascent strand 3' termi- methylation (29) which allows the repair system to discriminate nus, presumably because it is occluded by the replication the newly replicated DNA strand. The MutH endonuclease apparatus, and that cleavage by MutH is required to incise the delivers a nick to the unmethylated strand of the hemi- nascent strand and initiate the repair reaction. methylated GATC sites and the MutS protein detects and When the DNA is unmethylated, MutH incision occurs binds mispaired bases. The exact role of MutL is not known but randomly on either unmethylated DNA strand (35). According may form a complex with both MutH and MutS to coordinate to the proposed slipped misalignment mechanism, the opera- their action. The UvrD DNA helicase unwinds the DNA from tion of mismatch repair in a Dam methylase mutant would the site of MutH-delivered incision to the site of the mispair. lead, 50% of the time, to the excision of the parental template Degradation of this displaced strand is accomplished by one of strand causing fixation of the slipped (and deleted) configu- several single-strand DNA specific exonucleases (30). ration. In agreement with this idea, dam mutants exhibited the The methyl-directed mismatch repair pathway has been highest homeologous deletion rate of any strain we tested. shown previously to exclude RecA-dependent recombination Examination of the deletion of homeologous tandem re- between imperfectly homologous (homeologous) sequences in peats has allowed us to confirm several predictions of repli- bacteria: in interspecies crosses between Salmonella and E. coli cation-misalignment models of deletion. A heteroduplex in- (80% homology) (31), and between diverged repetitive se- termediate formed by pairing one strand of each repeat is quences on the E. coli chromosome (98% homology) (32). As experimentally confirmed. Moreover, the influence of MutH we observe, the exclusion to homeologous recombination is endonuclease and Dam methylase suggest that the slipped reversed by mutations in mutS, mutL, and partially by uvrD. misalignment occurs during or shortly after DNA replication, The consistently weak effect of uvrD in all reports may mean with the deleted product formed on the newly replicated that its unwinding function is not required for mismatch repair strand. We do not believe that our conclusions are restricted of homeologous sequences, either because other helicases to plasmid replication events. RecA-independent deletion of substitute for its role or because the heterologies facilitate tandem repeats occurs on the E. coli chromosome as well as on unwinding of DNA by MutS and MutL unaided by DNA plasmids (8). The ColEl-derived plasmids used in our exper- helicases. iments employs the E. coli replication machinery (36). Al- The results reported here, however, differ in one important though these results do not address which DNA polymerase is aspect. Whereas we observe a dramatic effect of mutH, responsible for deletion events, our unpublished studies have equivalent to mutL and mutS, in the exclusion of homeologous implicated the replicative DNA polymerase III. Several mu- RecA-independent deletion, both of the previously cited in- tants impairing the polymerase III complex lead to elevated stances of homeologous RecA-dependent recombination re- rates of tandem repeat deletion on the chromosome as well as port little or no effect by mutH. This has been explained by the on plasmids (Catherine J. Saveson and S.T.L., unpublished notion that excision may be directed from the ends provided results). by the recombining molecules rather than an incision pro- It has been reported that MutS and MutL proteins block moted by MutH endonuclease (31). Our observations of a RecA-mediated branch migration in vitro and that this may be MutH and Dam requirement for exclusion of homeologous responsible for inhibition of homeologous recombination (37). deletion, comparable to that seen for spontaneously occurring However, RecA-mediated strand exchange is not required for replication errors, support the idea that much of the deletion our observed deletion events, and therefore the MutLS inhi- detected by this system occurs during DNA replication. bition of RecA-mediated strand exchange cannot be solely These results support a model for deletion involving a responsible for exclusion of homeologous recombination slipped misalignment ofthe repeats during the process of DNA events by the mismatch repair system. replication, which forms the deletion on the nascent, unmethylated strand as depicted in Fig. 2. Displacement of the nascent We thank Timothy Strassmaier for construction of pSTL113 and Downloaded by guest on September 24, 2021 7124 Genetics: Lovett and Feschenko Proc. Natl. Acad. Sci. USA 93 (1996) STL2172 during his laboratory rotation, Martin Marinus for strains, 17. Trinh, T. Q. & Sinden, R. R. (1991) Nature (London) 352, and Jim Haber for comments on the manuscript. This work was 544-547. supported by National Institutes of Health Grant GM43889. 18. Weston-Hafer, K. & Berg, D. E. (1991) Genetics 127, 649-655. 19. Bachmann, B. J. (1987) in Derivations and Genotypes of Some 1. Bates, G. & Lehrach, H. (1994) Bioessays 16, 277-284. Mutant Derivatives ofEscherichia coli K-12, ed. Neidhardt, F. D. 2. Krawczak, M. & Cooper, D. N. (1991) Hum. Genet. 86, 425-441. (Am. Soc. for Microbiol., Washington, DC), pp. 1190-1219. 3. Richards, R. I. & Sutherland, G. R. (1992) Cell 70, 709-712. 20. Csonka, L. & Clark, A. J. (1979) Genetics 93, 321-343. 4. Pierce, J. C., Kong, D. & Masker, W. (1991) Nucleic Acids Res. 21. Lovett, S. T. & Sutera, V. A., Jr. (1995) Genetics 140, 27-45. 14, 3901-3905. 22. Miller, J. H. (1992) A Short Course in Bacterial Genetics (Cold 5. Mazin, A. V., Kuzminov, A. V., Dianov, G. L. & Salganik, R. I. Spring Harbor Lab. Press, Plainview, NY). (1991) Mo. Gen. Genet. 228, 209-214. 23. Willetts, N. S., Clark, A. J. & Low, B. (1969) J. Bacteriol. 97, 6. Dianov, G. L., Kuzminov, A. V., Mazin, A. V. & Salganik, R. I. 244-249. 24. Ausubel, F. A., Brent, R., Kingston, R. E., Moore, D. D., Seid- (1991) Mol. Gen. Genet. 228, 153-159. man, J. G., Smith, J. A. & Struhl, K. (1989) Short Protocols in 7. Bi, X. & Liu, L. F. (1994) J. Mol. Biol. 235, 414-423. Molecular Biology (Greene/Wiley-Interscience, New York). 8. Lovett, S. T., Drapkin, P. T., Sutera, V. A., Jr., & Gluckman- 25. Chung, C. T., Niemela, S. L. & Miller, R. H. (1989) Proc. Natl. Peskind, T. J. (1993) Genetics 135, 631-642. Acad. Sci. USA 86, 2172-2175. 9. Albertini, A. M., Hofer, M., Calos, M. P., Tlsty, T. D. & Miller, 26. Dower, W. J., Miller, J. F. & Ragsdale, C. W. (1988) Nucleic J. H. (1982) Cold Spring Harbor Symp. Quant. Bio. 47, 841-850. Acids Res. 16, 6127-6145. 10. Lovett, S. T., Gluckman, T. J., Simon, P. J., Sutera, V. A., Jr., & 27. Lea, D. E. & Coulson, C. A. (1949) J. Genet. 49, 264-285. Drapkin, P. T. (1994) Mol. Gen. Genet. 245, 294-300. 28. Modrich, P. (1991)Annu. Rev. Genet. 25, 229-253. 11. Singer, B. S. & Westlye, J. (1988) J. Mol. Biol. 202, 233-243. 29. Campbell, J. L. & Kleckner, N. (1990) Cell 62, 967-979. 12. Streisinger, G., Okada, Y., Emrich, J., Newton, J., Tsugita, A., 30. Cooper, D. L., Lahue, R. S. & Modrich, P. (1993) J. Biol. Chem. Terzaghi, E. & Inouye, I. (1966) Cold SpringHarbor Symp. Quant. 268, 11823-11829. Biol. 31, 77-86. 31. Rayssiguier, C., Thaler, D. S. & Radman, M. (1989) Nature 13. Albertini, A. M., Hofer, M., Calos, M. P. & Miller, J. H. (1982) (London) 342, 396-401. Cell 29, 319-328. 32. Petit, M.-A., Dimpfl, J., Radman, M. & Echols, H. (1991) 14. Efstratiadis, A., Psalony, J. W., Maniatis, T., Lawn, R. M., Genetics 129, 327-332. O'Connell, C., Spritz, R. A., DeRiel, J. K., Forget, B. G., Weiss- 33. Parker, B. 0. & Marinus, M. G. (1992) Proc. Natl. Acad. Sci. USA man, S. M., Slightom, J. L., Blechl, A. E., Smithies, O., Baralle, 89, 1730-1734. F. E., Shoulders, C. C. & Proudfoot, N. J. (1980) Cell 21, 653- 34. Dohet, C., Dzidic, S., Wagner, R. & Radman, M. (1987) Mo. 668. Gen. Genet. 206, 181-184. 15. Gordenin, D. A., Malkova, A. L., Peterzen, A., Kulikov, U. N., 35. Welsh, K. M., Lu, A. L., Clark, S. & Modrich, P. (1987) J. Biol. Pavlov, Y. I., Perkins, E. & Resnick, M. A. (1992) Proc. Natl. Chem. 262, 15624-15629. Acad. Sci USA 89, 3785-3789. 36. Scott, J. R. (1984) Microbiol. Rev. 48, 1-23. 16. Rosche, W. A., Trinh, T. Q. & Sinden, R. R. (1995) J. Bacteriol. 37. Worth, L., Jr., Clark, S., Radman, M. & Modrich, P. (1994) Proc. 177, 4385-4391. Natl. Acad. Sci. USA 91, 3238-3241. Downloaded by guest on September 24, 2021