Copyright 0 1997 by the Genetics Society of America

Pathways for Homologous Recombination Between Chromosomal Direct Repeats in Salmonella typhimurium

Timothy Galitski and John R. Roth

Department of Biology, University of Utah, Salt Lake City, Utah 841 I2 Manuscript received June 21, 1996 Accepted for publication March 25, 1997

ABSTRACT Homologous recombination pathways probably evolved primarily to accomplish chromosomal repair and the formation and resolution of duplications by sisterchromosome exchanges. Various DNA lesions initiate these events. Classical recombination assays, involving bacterial sex, focus attention on double- strand ends of DNA. Sexual exchanges, initiated at these ends, depend on the RecBCD pathway. In the absence of RecBCD function, mutation of the sbcB and sbcC genes activates the apparently cryptic RecF pathway. To provide a more general view of recombination, we describe an assay in which endogenous DNA damage initiates recombination between chromosomal direct repeats. The repeats flank markers conferring lactose utilization (Lac+) and ampicillin resistance (ApR); recombination generates Lac- Aps segregants. In this assay, the RecFpathway is not cryptic; it plays a major role without sbcBC mutations. Others have proposed that single-strand gaps are the natural substrate for RecF-dependent recombination. Supporting this view, recombination stimulated by a double-strand break (DSB) in a chromosomal repeat depended on RecB function, not RecF function. Without RecBCD function, sbcBC mutations modified the RecF pathway and allowed it to catalyze DSB-stimulated recombination. Sexual recombination assays overestimate the importance of RecBCD and DSBs, and underestimate the impor- tance of the RecF pathway.

RADITIONAL studies of bacterial homologous re- and RecF pathways are not strict alternatives for sexual T combination exploit the sexuality of these organ- recombination. The RecF pathway has come to be re- isms ( LEDERBERGand TATUM1946) . In typical conjugal garded as a minor cryptic recombination pathway that or generalized transduction crosses, recombination oc- can be activated in the absence of the primary RecBCD curs between an injected linear donor DNA molecule pathway. Consequently, it has received less attention. (with two recombinagenic double-strand ends) and a In spite of the close association of homologous re- homologous recipient chromosome. Recombination- combination and sexuality, it seems likely that the prin- proficient recipient strains integrate the donor DNA cipal purposes for which recombination pathways and giverise to recombinant organisms that can be evolved are theimmediately valuable processes ofchro- detected by positive selection. Classical analysis of sex- mosomal repair and the formation and resolution of ual recombination revealed recombination pathways. gene duplications. The severe viability problems and These were defined by mutants with reduced ability to characteristic DNAdamage sensitivities of various rec serve as a recipient in conjugal crosses (reviewed in mutants support this view. The need for recombina- CLARK1971, 1973). In wild-type bacteria, the RecBCD tional DNA repair and the occurrenceof chromosomal pathway is the major sexual recombination pathway (re- rearrangements seems to be universal, whereas the rela- viewed in SMITH1988) ; recB mutants and recC mutants tive contribution of sexual recombination to genetic are Rec- . In recB or recC mutant strains (abbreviated variation differs from species to species. This spectrum recBC) , the combinedeffects of two extragenic suppres- ranges from primarily asexual species with a clonal pop sor mutations, sbcB and sbcC, restore recombination ulation structure tospecies withan obligate sexual stage ability and DNA-damage resistance (KUSHNERet al. of the life cycle. Among bacterial species, population 1971; LLOYDand BUCKMAN1985). In these suppressed structures range from panmictic, in which frequent sex mutants, a previously crypticalternative to the RecBCD randomizes allele combinations, to clonal, in which sig-

nificant linkage disequilibrium (nonrandom allele asso- pathway operates. In theRec + triple mutants, recBC sbcB sbcC, recombination depends on the functions of the ciations) is observed ( ~~A~NARDSMITH et al. 1993). It is clear that some prokaryotes, including the enterics RecF pathway ( HORIIand CLARK 1973). The RecBCD Escherichia coli ( OCHMAN andSELANDER 1984) and Sal- monella typhimun'um ( BELTRANet al. 1988) , exist in Corresponding author: Timothy Galitski, Whitehead Institute for Bio- medical Research, 9 Cambridge Center, Cambridge, MA 02142. clonal populations that engage in relatively infrequent E-mail: [email protected] sexual episodes. For all organisms, the repair of single-

Genetics 146 751-767 (July, 1997) 752 Galitski T. and J. R. Roth

the population when selection is removed. So, the pres- ence of duplications in a population can facilitate the exploitation of availableresources and enhance thesur- vival of a clone under adverse conditions. The high frequency of gene duplications and their reversibility suggest that thisability is used essentially as a gene regulatory mechanism. If the main use of homologous recombination is to perform sister-chromosome exchanges neededfor DNA repair and duplication formationand segregation, then it seems reasonable to use these events in an assay system. Here, we describe a system for observing recom- or bination between direct repeats of a chromosomal seg- a ment. The properties that distinguish this system from standard sexual recombination assays are thefollowing: FIGURE1 .-Chromosomal rearrangement by unequal re- combination. DNA replication (growing fork at right) pro- All recombining sequences and recombination en- duces sister chromosomes. Striped boxes represent repeated zymes are internal to an intact bacterium. Recombi- sequence elements in the same orientation. A nonreciprocal exchange (one pair of flanks rejoined) between ectopic re- nants arise within bacterial clones without the partic- peats can result in duplication ( -) or deletion (- - -) of the ipation of bacteriophage or plasmid functions that intervening loci (b and c) . A reciprocal exchangegives both contribute to sexual exchanges. products. Initiating DNA structures (single-strand gaps, dou- ble-strand breaks, etc.) form spontaneously by en- and double-strand DNA lesions through sister-chromo- dogenous processes, and they can be induced experi- some exchanges probably represents the most frequent mentally. uses of homologous recombination. Both spontaneous and experimentally induced re- The direct tandem duplication of chromosomal in- combination events are very frequent. Selection for tervals results from unequalsisterchromosome ex- rare recombinant genotypes is not required. changes initiated by spontaneous DNA lesions. Recom- The recombinants scored can arise by a single intra- bination between ectopic homologies in direct orienta- molecular nonreciprocal exchange between direct tion ( e.g., rrn loci, transposable elements, Rhs elements, chromosomal repeats. They also can result from REP sequences, duplicated intervals)results in a dupli- more complex events such as multiple exchanges, cation and/or a deletion of the intervening material intermolecular events (between sister chromo- (Figure 1) . One can observe the very frequent duplica- somes), and reciprocal exchanges. tion, and further amplification or the return to hap- Using this system, we provide evidence that both the loidy, of manylarge ( > 1 min or -47 kb) chromosomal RecBCD and the RecFpathways play major roles in intervals in S. typhimurium (ANDERSON and ROTH recombination. Furthermore, whereas the substrate of 1981). Duplication frequencies fall in the range the RecBCD pathwayis the chromosomaldouble-strand to lop3for most loci. For loci mapping between rm break or the double-strand ends of a donated DNA operons, the duplication frequency is a few percent. fragment, the natural substrate of the RecF pathway is The transposable element IS200 provides another im- neither of these. In the absence of RecBCD function, portant source of ectopic repeats used to form sponta- sbcB sbcC suppressor mutations allow the RecF pathway neous duplications in S. typhimurium ( HAACK and ROTH to catalyze recombinationinitiated by double-strand 1995) . The high frequencies of chromosomal duplica- breaks. These results support the idea thatsingle-strand tions and theconservation of repeated elementssuggest gaps in the bacterial chromosome are the natural sub- an importantrole for chromosomal duplications bac-in stratefor RecFpathway recombinationin wild-type terial survival and evolution. cells. The duplication and amplification of chromosomal intervals may be an important source of genetic varia- MATERIALS AND METHODS tion that can increase fitness by reversibly increasing gene dosage (reviewed in ROTH et al. 1996). Various Growth media and genetic methods: The E medium of laboratories have shownthat specific adverse conditions VOGELand BONNER( 1956) supplementedwith 0.2%glucose select for bacterial variants harboring specific duplica- was used as the defined minimal medium. Growth on alterna- tive carbon sources was in NCE salts, described by BERKOWITZ tions or amplifications that arise by recombination be- et al. (1968), supplemented with 0.2% of the appropriate tween direct repeats. Because these rearrangements are carbon source. The complex mediumwas nutrient broth, NB, readily reversed by recombination, they disappear from (8 g/liter, Difco Laboratories) with added NaCl (5 g/liter) . Pathw ays 753 Recombination Pathways

Solid media contained BBL agar at 1.5%. Auxotrophic re- Inheritance of the purA:MudJ element causes purine auxotro- quirements on minimal media weresatisfied with supple- phy; these transductants do not grow on the minimal me- ments at final concentrations recommended by DAVIS et al. dium. Recombination between the donor andrecipient Mud ( 1980). Phage plates used to score plaque morphologies con- elements replaces the MudAjoin pointwith a MudJjoin point, tained 1%Bacto tryptone (Difco) ,0.8%NaC1, and 1.2% agar. and gives rise to prototrophic kanamycin-resistant transduc- Phage top agar was the same except it contained 0.7% agar. tants. A Lac+ KmR (unstable) Aps Pur+ transductant was Final concentrations of antibiotics were as follows: tetracycline savedas strain TT18947 { DUP1732[(pyrB2694) *MudJ* hydrochloride, Tc, at 20 pg/ml; kanamycin sulfate, Km, at ( thr-469) ] } . 50 pg/ml; chloramphenicol, Cm, at 20 pg/ml; sodium ampi- Construction of rec derivatives of duplication strains: Vari- cillin, Ap, at 30 pg/ml for single-copy elements ( MudA) or ous rec mutations were introduced into wild-type strain LT2, at 100 pg/ml for multicopy elements (plasmids). The chro- and into rec+ duplication strains IT18931 (DUP1731) and mogenic @galactosidase substrate 5bromo-4chloro-3indolyl- TT18939 (DUP1732). Insertion mutations with a drug resis- P-n-galactopyranoside (Xgal) was obtained from Diagnostic tance, such as recB503:TnlO ( MAHAN and ROTH 1989a), Chemicals Ltd. and used at afinal concentration of 25 ,ug/ml. recD54l::TnlWCm ( MIESEL and ROTH 1994), recF521::Tn5 The nonmetabolizable inducer of the Lac1 repressor protein, (Runnand MENZEL1987), recJ504:MudJ (MAHAN et al. isopropyl-f+D-thiogalactopyranoside (IPTG) , was used at a 1992),and recN555:TnlWCm (T. GALITSKI,unpublished final concentration of 1 mM.All incubations were at 37”. data), were transduced by selecting on NB plates with the Sterile 0.85% NaCl (saline) was used to dilute cultures except appropriate antibiotic. To construct derivatives carrying the as noted. recA1 mutation (WINGet al. 1968) , the srl-203:Tn IWCm mu- Transductional crosses, mediated by the phage mutant P22 tation of strain TT11183 (srl-203:TnlWCm) was introduced HT105/ 1 int-201 (SCHMIEGER 1971),were performed as de- by transduction selecting CmR. Then, the srP-linked recAl scribed by ROTH (1970). mutation was cotransduced into these derivatives by selecting Constructionand verification of duplicationstrains: a Srl’ phenotype on NCE sorbitol plates. Transductants car- Strains carrying direct-order duplications of predetermined ryingrecAl were identified by their UV sensitivity. All the chromosomal intervals were constructed and verified as de- rec alleles used here are recessive to the wild-type allele and scribed by HUGHESand ROTH ( 1985). These merodiploids presumably confer null phenotypes. Note that recJ504:MudJ are the progeny of triparental P22-mediated crosses involving is a Lac- insertion of the lac operon fusion element MudJ in two donors, each with a simple MudA (HUGHES andROTH the reggene; the lac genes are wild type but they are presum- 1984) insertion that will define one of the duplication end- ably in the wrong orientation for transcription. BlueXgal points, and a haploid recipient. In this method, recombina- staining of duplication strains carrying this insertion is due tion between MudA elements (during transduction) results to the lac fusion at the duplication join point, not the MudJ in the duplication of the intervening loci. The duplicated element at the recJlocus. sequences flank a MudA element located at the duplication The S. typhimurium recNmutation used was isolated starting join point. The MudA element of these strains carries an with a DNAdamage-inducible MudJ insertion that was shown ampicillin-resistance determinant and the lacZYA genes tran- to map at the recN locus closely linked to the pheA gene at scriptionallyfused to an adjacent chromosomal operon. Thus, 58‘ (T. GALITSKIand D. THALER,unpublished results). A duplication-bearing strains have Lac+ and ApR phenotypes recMTnlWCm insertion was identified based on close linkage that are lost when an exchange occurs between the flanking to the original MudJ insertion, on its mitomycin-C sensitivity, repeats. and on its Rec- phenotype when placed in a recB sbcB sbcC The duplications used here, DUPl731[ ( leuAll79) *MudA* strain. ( nadC220) 3 and DUP1732[ (pyrB2694)*MudA*( thr-469) 1, Tests of UV sensitivity and phage-plaque morphology: To include the region from Zed to nadC (3’ to 4’) and the verify strain genotypes and to investigate the responses of region from pyrB to thr (97‘ to 100 ’ ) , respectively (Table 1) . previously untested genotypes, W sensitivity and phage- For each duplication, we confirmed the diploidy of a locus plaquing phenotypes of bacterial strains were assessed. Sensi- within the duplication and the haploidy of two loci mapping tivity to ultraviolet light (W)was scored by spotting l@pl outside the duplication. These three test lociare guaC (inside aliquots of serial 100-fold dilutions of overnight cultures on DUP1731, outside DUP1732), serB (inside DUP1732, out- two nutrient agar plates. One of the plates was exposed to a side DUP1731), and nu& (outsideboth DUP1731 and UV dose that reduced survival of wild-type strains to about DUP1732). These tests weredone by transduction of an auxo- one-third of unirradiated controls. Ratios of UV survivors to trophic TnlO (tetracycline-resistant, TcR) insertion for each unirradiated controls were determined. Phage-plaquing phe- test locus. If the recipient was diploid for a test locus, allTcR notypes of bacterial strains were scored by spotting serial 100- transductants were prototrophic. If the recipient was haploid fold dilutions of wild-type P22, P22 erfam (H1173), andP22 for atest locus, all TcRtransductants were auxotrophic. When c2-5 h21 A327 ( abcl abc2) lysates on freshly poured bacterial haploid (Ap’ Lac-) segregants were isolated from prototro- lawns in soft agar on phage plates. When the spots dried, the phic tetracycline-resistant duplication strains, the segregants plates were incubated overnight and scored. Bacteriophage fell into two classes, tetracycline-resistantauxotrophs and tet- strains were supplied by N. BENSON. racycline-sensitive prototrophs (data not shown). Thus, these Colony-sectoringrecombination screen: Recombination merodiploid heterozygous transductant strains produce hap- between chromosomal direct repeats was scored qualitatively loid segregants carrying one oftwo alleles present in the in colonies by plating duplication strains with lac fusions at parent. the join point on NB Xgal plates and incubating these for 4 The MudA join point of a strain carrying DUP1732 was days. Haploid (Lac- Aps) recombinant clones form white converted to MudJ ( CASTILHO et al. 1984) . The MudJ element sectors within the colonies. Photographs of such colonies were is the same as the MudA element except it carries a kanamy- taken using 400 ASA color print film in a Canon AE-1 camera cin-resistance determinant instead of the ampicillin-resistance mounted on anOlympus SZH-ILLD microscope with indirect gene. Using strain TT12319 (purA2158::MudJ) as a donor, transillumination and 15X magnification. strain TT18939 (DUP1732 with a MudAjoin point) was trans- Quantitative recombination assay: Rates of recombination duced to kanamycin resistance on E glucose kanamycinplates. between chromosomal direct repeats were determined by ob 754 T. Galitski and J. R. Roth serving the frequencies of Lac- segregants in broth cultures strain SA2337 (9urC7) to form merodiploid strain IT18948 of duplication strains. For each strain assayed, an overnight (purC7 DUP1732[ (pYrB2694) *MudJ* ( thr-469)] 1. The dupli- NB Ap culture, started from a single colony, was diluted 10'- cation DUP1732 spans the region from pyrB (97') to thr fold in fresh NBAp and grown with aeration to incipient ( 100'). This interval includes the serB locus, mutations of turbidity. Immediately, the culture was diluted lo4-fold in whichcause serine auxotrophy. The serB gene is -125kb fresh NB and multiple 1 ml NB cultures (without antibiotic from pyrB (unlinked by P22 transduction) and -25 kb from to allow the growth of Aps Lac segregants) were started with thr (15% linked).The serB1466::TnlWTcmutation was trans- a 10 ~1 inoculum (typically fewerthan 10 cells) from the lo4- duced into duplication strain TT18948 by selecting TcR.Since fold dilution. These NB cultures were incubated for 24 hr this strain has two copies of SUB, these transductants were TcR (except as noted). The inoculum size was determined by Ser+. The TnlWTcinsertions of these transductants could be measuring the cfu/ml in the lo4-fold dilution from which in either copyof serB. Because serBis linked to thr, it was inocula were derived. These platings were on NBXgal to possible to determine which copy carried the TnlWTc inser- verify simultaneously the Lac+ phenotypeof cells in the inoc- tion. An insertion in the counterclockwisecopy would be ula (data not shown) ; thus, segregants found in themultiple linked to the duplication join point ( hR).An insertion in 1 ml NB cultures arose independently. For each strain assayed, the clockwise copy would be linked to thr+. Strain 33'18949 20 independent nonselective cultures were prepared. These hadthe serB1466::TnlWTc insertion (DBS site) inthe cultures were diluted lO'-fold in saline; an appropriate vol- counterclockwise copy (data not shown); it was used for this ume of this dilution (0.1 mi for rec' and recFstrains, 0.2 ml study. A control strain was constructed also. This strain, for recA strains, 0.3 ml for recB and recB recFstrains) was spread TT18950, carries the nadA379:TnlWTc insertion, not the to each of four NB Xgal plates (40 plates for recA and recB serB1466::Tn 1WTc insertion. The nadA locus maps at 17 ' (far recF strains). The plates were incubated for 2-4 days. Cole from DUP1732). nies were scored as either Lac+ at plating (some blue stain- In the construction of mutant Rec- derivatives of strain ing), or Lac- at plating (no blue staining) and the total TU8949 (pure7 DUP1732[(pyrB2694 serBl466:TnladTc) population size was noted. The Lac- segregant frequency of *MudJ*( thr-469)3 }, we avoided insertion mutations in rec each culture is the ratio of Lac- colonies to total colonies genes, especially TnlO insertions, both because of a lack of (segregants/cell) . The segregation rate was calculated using available drug resistances and because TnlO insertions would the median Lac- segregant frequency. The number of gener- provide additional sites of double-strand breakage. All rec al- ations was calculated from the final and initial population leles wereintroduced by the same general strategy (described size. Segregation rate is the final segregant frequency divided above) used to introducethe recAl allele to duplication by elapsed generations (segregants/ cell / generation ) . strains; in each case a mutant phenotype generated by an Plasmid constructions: Plasmids allowing the regulated ex- insertion mutation was corrected in a transduction cross in pression of various alleles of the IS10 transposase gene were which the rec allele was coinherited. The recBl0 allele ( EISEN- constructed. The 1300-bp EcoRI-Hind111 fragment of plasmid STARK et al. 1969) and the recD561 allele ( MIESELand ROTH pMJR1560 (ST- 1987) contains the lac19 gene, expressing 1994) were introduced by linkage to the argA locus. The elevated levels ofthe Lac1 repressor. This fragment was ligated recF558allele (T. GALITSKI,unpublished data) was introduced to the 2700-bp EcoRI-Hind111fragment of plasmid pGC2 (MY- by repair of a dgo mutation. The deletion DEL1792 (phssbcB) ERS et al. 1985). Theresulting plasmid, pZT379, has a polyli- ( BENSONand ROTH 1994) was introduced by linkage to the nker upstream of a complete kcZq gene, followed by an M13 his operon. The sbcCD8 allele ( BENSONand ROTH 1994) was origin, a pBR322derived origin, and an ampicillin-resistance introduced by linkage to poC. All of these rec and sbc alleles determinant. Between the ClaI and BamHI polylinkersites are recessive to the correponding wild-type alleles. Table 1 of vector plasmid pZT379, fragments ( 1300-bp ClaI-BamHI) lists the complete strain genotypes. Note that insertions of containing IS10 transposase alleles and the LacI-repressor- TnlWCm were commonly used as linked markers for the sensitive Ptac promoter were introduced. These fragments introduction of rec and sbc alleles. These insertions were re- were derived from plasmids, kindly provided by D. HANIFORD moved in the course of the constructions as described above and N. KLECKNER, including pDHlO, pDH12, pDH10-GD163, for srl-203:TnlWCm and recA1. Since TnIWCm is a defective and pDH10-PS167 ( HANIFORD et al. 1989). Theresulting plas- element (ELLIOTTand ROTH 1988), andsince no active TnlO mids are pZT380 (wild-type transposase, Trip+ ) , pZT381 (a elements were used during these constructions, the final transposase null allele with an internal in-frame deletion be- strains harbor no ISlOinsertions providing unwanted double- tween two NcoI sites, TnpA) , pZT382 and pZT383 (express- strand break sites. ing transposase mutants that catalyze the excision, but not The recF558 allele was isolated (after hydroxylamine-in- the transposition, of TnlOderived elements). For the experi- duced local mutagenesis described by HONCand AMES 1971) ments presented, we used plasmid pZT382 expressing uans- and verified during the course of this work. The phenotype posase mutant GD163, designated Tnp*. This plasmid allows and map position of the recF558 allele is identical to that of the introduction of a double-strand break at a defined time the recF521::Tn5 insertion. Both causeUV sensitivity in a wild- ( the addition of IPTG) and at a single defined site (a TnlO type background and a Rec- phenotype in a recBC sbcB sbcC insertion) in the chromosomes of a bacterial population. background. Both are transductionally linked to the dgo locus Construction of double-strand-breakassay strains: These (data not shown; see RUDDand MENZEL1987) . These recF558 strains differ from previously described duplication strains in phenotypes are fully complemented (data not shown) by plas- three respects. They possessa single TnlOelement (providing mid pMAJ34, a minimal E. coli recP clone (kindly provided a DSB site), and they carry a plasmid providing expression by R. MYERS;BIANAR et al. 1984) . of amutant IS10transposase (catalyzing excision of the The introduction of transposaseexpressing plasmids was TnlO) . In addition, a MudJ duplication join point with KmR the final step in all constructions. For each strain, two plas- (not the ApR of MudA) was used to allow selective mainte- mids were introduced by transductions selecting ApR. One nance of the transposase plasmids that confer ampicillin resis- of the plasmids, pZT382, expresses Tnp*, the mutant IS10 tance. The MudJ duplication join point (and consequently transposase that catalyzes double-strand breakage but not the entire duplication) of strain 'IT18947 (DUP1732[ (pyrB2- transposition. The other plasmid, pZT381, expresses TnpA, 694) *MudJ*( thr-469)1; described above] was transduced to a deletion mutant transposase that catalyzes neither double- Pathways 755 Recombination Pathways strand breakage nor transposition.Strains withplasmid pZT381 served as negative controls that show the spontaneous frequencies of haploid Lac- segregants. These spontaneous frequencies were subtracted from those observedwith the strainscarrying plasmid pZT382, providing double-strand breaks at the TnlOdTc insertion site. This difference is the frequency of haploid Lac- segregants induced by a double- strand break as presented in Figure 7. In Figure 6, the uncor- rected numbers are presented to illustrate the nature of the assay system. Assays of recombinationstimulated by double-strand breaks The stimulatory effect of transposase-induced DSBs was tested in various rec backgrounds. The frequency of Lac- recombinants in cultures of DSB-assay strains was measured. Independent (started from separate single colonies) starter cultures of assayed strains were grown to saturation in NB Km Ap. Each culture was diluted 10-fold in 4.5 ml saline. From each of these dilutions, a 0.01-ml aliquot was transferred to each of three 2-ml tubes of NB Apwith IPTG. The IPTG FIGURE2.-A screenfor recombination between direct induces expression of the mutant transposase that catalyzes chromosomalrepeats. The merodiploidstrain IT18939 excision of the TnlOdTc, leaving a double-strand break in 1DUP1732 [ (pyrB2694)*MudA*(thr469)l) carries a 150-kb the chromosome. Some control cultures hadno IPTG. These duplication of loci from pyrB at 97’ to thr at 100 ’. These direct experimental cultures were incubated overnight. Both the repeats flanka MudAjoin-pointelement conferring Lac’ and starter cultures and the experimental cultures were diluted ApR phenotypes. An exchange between the repeats leads to appropriately in saline to provide 0.1-ml aliquots containing the segregation of haploid Lac- Aps recombinants. at least 50 colony-forming units. These 0.1-ml aliquots were spread on NB Xgal plates (four for each culture), incubated for 2 days, and scored. Colonies were scored aseither all white repeats results in the segregationof a haploid recombi-

(Lac - at plating) or blue with white sectors(Lac + at plating) . nant thathas lostthe join-pointmarkers; these recombi-

The Lac ~ segregant frequencyof each culture is the ratio of nants are Lac- and ampicillin sensitive (Figure 2). white (Lac-) colonies to total colonies. For each strain, the Though Figure 2 illustrates an intermolecular nonre- Lac- segregant frequency is that of the experimental culture ciprocal exchange, intramolecular and reciprocal ex- minus the low segregant frequency inthe starter culture from which it was derived. changesalso will produce Lac - Apsrecombinants. These segregationevents occur spontaneously; the assay system provides neither DNA damage nor fragment RESULTS endsto initiate recombination. Presumably,endoge- Detection of recombination between extensive direct- nous processes provide the initiating DNA lesions. The order chromosomal repeats: Duplication strains were observed segregation events occur very frequently and used to detect recombination betweenextensive direct- can be observed without selection for recombinants. order chromosomal repeats. In these strains thedupli- Properties of the recgenotypes investigatedTo both cated regions flank a MudA element, conferring ApR verify strain genotypes and investigate the responses of and Lac+ phenotypes. The chromosomal structure of previously untested genotypes, UV sensitivity and one of these duplication strains, TT18939 (DUP1732 phage-P22 plaquing phenotypes of mutant rec strains [ (pyrB2694) *MudA*( thr-469) ] } , is diagrammed in Fig- were assessed. The data of Table 2 summarize observa- ure 2. The full genotypes of merodiploidstrains tions made with a ret+ strain and its rec derivatives. TT18931 (DUP1731)and TT18939 (DUP1732)are These properties aregenerally identical tothe behavior listed in Table 1. The duplication DUP1731[ ( lewill79) of analogous E. coli mutants, tested with phage lambda. *MudA*( nadC220) ] is a one minute ( -40 kb) direct- UV survival, relative to unirradiated controls, was as- order duplication of the region from the leuA locus at sayed at a UV dose that reduces survival of ret+ strains 3 ‘ to the nadC locus at 4’ with a transcriptional fusion to about one-third. As expected, a recD mutation did of the MudA lac operon to led at the joinpoint. Simi- not confer UV sensitivity ( MIESELand ROTH 1994). larly, DUP1732 [ (pyrB2694)*MudA* ( thr-469) ] is a Mutant recA strainsshowed a UV hypersensitivity. three minute ( -150 kb) direct-order duplication from Strains with a mutant allele of either recB, recF, or recJ the pyrB locus at 97 ’ to the thr locus at 100 ’ with a showed a less extreme UV sensitivity. Of particular note transcriptional lac fusion to thr at the joinpoint. Note is the severe UV sensitivity of recB recF double mutant that DUP1731 and DUPl732 do not overlap; observa- strains, and therecA-like UV sensitivity of recB recJdouble tions made with both strains are not dueto peculiarities mutants previously reported by MAHAN et al. ( 1992) . of a shared interval. The construction and verification Moreover, recB recJdouble mutants were difficult to con- of theseduplications is describedin MATERIALS AND struct, maintain, and study due to viability problems METHODS. and very slow growth. An exchange between the direct-order chromosomal Sincegrowth of somemutant bacteriophage P22 756 T. Galitski and J. R. Roth strains requireshost recombination functions, these tors in recA derivatives. These rare sectors arise by rec- phage strains were used to verify or phenotypically char- independent deletion events. Some of these rare sectors acterize rec genotypes. Wild-type P22 was used as a con- are homogeneous populations of bacteria with an un- trol. Normally, the formation of lysogens by wild-type usual phenotype (e.g.,Lac- ApR;data not shown). The P22 produces turbid plaques; most of the host strains best explanation for these observations is that each sec- investigated gavethis result. However,wild-type P22 tor contains descendants of a single recombinant cell. produced clear plaques on recB recJ double mutants. Effects of rec mutations: To investigate the recombi- This observation, combined with the extremely poor nation-pathway dependence of the events scored in the viabilityof these doublemutants, might indicate an colony-sectoring recombination assay, various rec deriva- SOS-constitutive phenotype due to a general defect in tives of duplication strains TT18931 and TT18939 were the repair of potentially lethal chromosomal lesions: tested. For each genotype, three independent isolates SOS constitutivity precludes lysogen formation. The were observed in the colony-sectoring assay described phage P22 erfamber mutant lacks the essential recombi- in MATERIALS AND METHODS. Essentially, duplication nationfunction needed for growth on recA strains strains arestreaked for single colonies on NB Xgal ( BOTSTEINand MATZ 1970). The Erf function is in- plates. After 4 days colonies were scored for the pres- volved in recircularization by recombination at the ter- ence of white (Lac-) sectors. Figure 3 shows rep- minal redundancies of linear injected P22 genomes. In resentative colonies of strain TT18939 (DUPl732 we+) these tests, recA mutants, of course, and recB recJdouble (A) , and derived rec mutants: recAl ( B ) , recBSO3:TnIO mutants, curiously, were nonpermissive hosts for P22 (C), recF521::Tn5 (D), recJ504::MudJ (E), and erjam. Also, this phage produced only small plaques at recBSO3:TnlO recF52I::Tn5(F) . The same results were a reduced efficiency on recJ mutants; this defect was obtained with strain TT18931 (DUP1731 we+) and its reported by "IAN et al. ( 1993). Phage P22 abc c2-5 is rec derivatives (Table 1; data not shown). The recFmu- a clear-plaque phage lacking an Znti-RecKD function tants and regmutants could not bedistinguished from that inhibits the exonuclease V activity ( POTEETEet al. ret' strains in this assay. The same was true of a recD 1988) of the host RecBCD enzyme. Exonuclease V de- mutant (data not shown).Typical recA colonies showed grades rolling-circle replication intermediates. The P22 no sectors. Strains with a recB mutation showed only a abc mutant forms pinpoint plaques on recBCD+ strains, slight defect. The proficiency of recB strains in duplica- but it can form large plaques on recB, recC, or recD mu- tion segregation has been observed in E. coli by SCLA- tant hosts. The total nonpermissivity of regmutants for FANI and WECHSLER(1981) and in Salmonella by P22 abc, and the suppression of this effect by recB muta- SECALL(1987). However, recB recF double mutants ex- tions, has been reported by MAHAN et al. (1992). hibited an obvious defect in recombination between The data summarized by Table 2 confirm the rec ge- chromosomal direct repeats. Sickly recB regdouble mu- notypes of the duplication strains used in this study. tants are notshown in Figure 3; after extended incuba- Some novel aspects of these observations contribute to tion, they formed irregular colonies with many white theinterpretation of recombinationdata presented sectors suggesting that they are ableto support duplica- below. tion segregation. A recN mutation had no effect in the The colony-sectoring recombination assay: The seg- colony-sectoring assay either as a single mutant, in dou- regation of Lac- recombinant clones within colonies ble mutant combinations with recB or recE; or as part of forms the basis of a screen for recombination defects. a triple mutant combination with recB and recF (data Homologous recombination between the direct repeats not shown). of duplication strains like TT18931 and TT18939 results The colony-sectoring assay of recombination is a con- in the segregation of haploid Lac- clones (Figure 2 ) . venient qualitative screen for recombination defects.

These events aredetected on Xgal plates as white The validity of the colony-sectoring assay was affirmed, (Lac - ) sectors within blue (Lac + ) colonies (Figure and its sensitivitywas calibrated by quantitating duplica- 3A). The white sectors reflect recombination events tion segregation rates during growth in liquid nutrient that cause loss of the duplicated region and the join broth. point. The cells in these white sectors are Aps (nojoin Quantitative recombination assay:A duplication seg- point)and haploid (no duplication)(data not regation rate, segregants per cell per generation, can shown) . Furthermore, the appearance of these sectors be calculated from the segregant frequencyin indepen- requireshomologous recombination functions (de- dent cultures that havegrown for a known number scribed below). Thus, they arose by a recombination of generations (see MATERIALS AND METHODS). Twenty event like that depicted in Figure 2, not a mutational independent segregant frequencies were observed in event affecting LacZ function. broth cultures of strain TT18931 (DUP1731 ret+) and Each sector is assumed to be a recombinant clone rec derivatives thereof. As an example of the results ob- arising from a single event. This assumption is sup- tained, Table 3 lists these frequencies in ascending or- ported by characterization of very rare single Lac- sec- der. The median frequency, the mean of the 10th and Recom bination Pathways Recombination 757

11th values, was chosen to represent each strain to avoid nation pathway dependence of both sexual and asexual disproportionatecontributions by “jackpots.” The recombination(Figure 4). Thismodel is elaborated quantitative data of Table 3 paralleled the results of the more fully in DISCUSSION. colony-sectoring assay. Data from 10 cultures of strain The main feature of this model is the different sub- TT18933 (DUP1731 recAl) arenot tabulated. Of strate specificities of the RecBCD and RecF recombina- 144,904 colonies of this strain, three, fromtwo cultures, tion pathways. It is proposed that the primary initiating were Lac-. This yields an overestimate of lop5for its substrate of the RecF pathway is a single-strand gap, median segregant frequency. Other than recA mutants, whereas a double-strand break is the substrate of the only recB recF double mutants showed a substantial re- RecBCD pathway (Figure 4A). This general idea has combination defect. beenproposed and independentlysupported pre- Cultures ofrecB recF doublemutant strains grew viously (e.g., WANG and SMITH 1983; LLOYDand slowly and, since all cultures in Table 3 were incubated THOMAS1984). Both types of lesions occur spontane- for 24 hr, spentless time in stationary phase. One could ously in the chromosomeand stimulate repair and rear- argue that the relative dearth of segregants in recB recF rangement events. Either pathway can produce an ex- cultures was due to shorter incubation time in station- change between direct repeats. Eliminating one path- ary phase. This possibility was eliminated by the observa- way does not result in a substantial defect. However, tion that 10 cultures of the recB recFdouble mutant gave the elimination of both pathways causes a Rec- pheno- the same distribution of Lac- segregant frequencies type. In classical sexual recombination assays (conjuga- after an additional 24hr period of incubation. The re- tion or transduction; Figure 4B) , the donor DNA has sults of extended stationary phase incubation areconsis- double-strand ends. Thus, the inheritance of a donor tent with the expectation that duplication segregation marker requires RecBCD function but not RecF func- occurs mainly during exponential growth. tion. In the absenceof the RecBCD pathway,suppressor

The median Lac - segregant frequencyand the num- mutationscombined withRecJ exonuclease activity ber of generations are two pieces of information allow RecF pathway enzymes to operate on the ends of needed to calculate a duplication segregation rate. The donor DNA. This might be due to processing of a dou- number of generations was calculated from the initial ble-strand end to generate a single-strand tail that ap- and final numbers of colony-forming units in the cul- proximates the featuresof a single-strand gap, theRecF tures (see MATERIALS AND METHODS). For each strain, pathway’s primary substrate. the Lac- segregation rate, a (segregation events per This model predicts that recombination stimulated cell per generation) , is by a specific type of DNA lesion will depend on a spe- cific pathway. For example, if recombination between a = 2f,/g, direct repeats were induced above spontaneous levels where f is the median segregantfrequency, and g is the by the introduction of a DSB in one of the repeats, number of generations (Table 4). this increase in recombination will depend on RecB Spontaneous recombination between chromosomal function, but not on RecF function. Furthermore, sbc direct repeats The results of the colony-sectoring (Suppressor of rec&BC) mutations will restore DSB-in- assay, the Lac- segregation rates, and the relative re- duced recombination to a recB mutant by allowing the combination proficiencies ( the ratio of the Lac - segre- RecF pathway to substitute. In wild-type strains, DSB- gation rate to that of the reef parent) of various rec induced recombination will be indifferent to a recFmu- genotypes are summarized in Table 4. All of the single tation, whereas in recBsbcB sbcC strains DSB-induced mutant types, except recA, were Ret+. The most striking recombination will depend on RecF function. These observations are the recombination proficiency of recB predictions are tested below. mutants and recFmutants, and the recombination defi- Induction of a specific DNA double-strand break: A ciency of the recB recF double mutant. Whereas sexual system was developed for the regulated induction of a recombination assays show a strong dependence on the single predetermined double-strand break at thesite of RecA and RecBC functions,but no dependence on a TnlO insertion in the bacterial chromosome. This RecF function ( except in arecBC sbcB sbcC mutant back- system exploits an IS10 transposase mutant isolated and ground), recombination between chromosomal direct characterized by HANIFORDet al. ( 1989). This mutant repeats requires RecA plus either RecBC or RecF func- transposase (Tnp”) catalyzes the excision of an ele- tion. This result indicated that some spontaneously ini- ment with IS10 ends from a donor site. Excision leaves tiated events depend on RecBCD-pathway recombina- a DSB and causes a transient induction of the SOS tion, whereas some depend onrecombination by a RecF regulon of DNA-damage-induced genes. However, this pathway that is active in wild-type cells. mutant transposase is defective for target site interac- A model and predictions: The genetics of spontane- tions. Excision produces primarily a flush cut double ous recombination between chromosomaldirect re- strand end anda freeabortive excised fragment bearing peats suggested a model that accounts for therecombi- bound transposase. Both genetic (BENDERet al. 1991 ) 758 T. Galitski and J. R. Roth

TABLE 1

Bacterial strains

Strain" Genotype* Source"

LT2 Wild type Lab collection TT11183 srl-203:Tn 1OdCm Lab collection SB2452 = TR2246 HfrB2 strA rnetA22 recAl J. WYCHE TT14556 recB503:Tn 10 Lab collection TT16812 recD541::TnlOdCm Lab collection KRS2322 = TT11293 recF521::Tn5 K. RUDD TT15278 recJ504:MudJ Lab collection TT18831 recN555:Tn 1OdCm TT12319 purA2158::MudJ Lab collection SA2337 = TR7206 purC7 K. SANDERSON TR5 124 recBl0 hisDl447 Lab collection TT17766 recD56ldel Lab collection TT18923 ara-9 recF558 TT19082 recAl/pMAB4 (pBR322 re&) Lab collection TT18924 fels2- leuA414am DEL1792(phmbcB) Lab collection TT17582 fels2- leuA414am recB497::MudJ sbcBl sbcCD8 Lab collection TT17427 recAl/pZT379 (vector, ApR, lacP) TT17428 recAl/pZT380 (trip+) TT17429 recAl/pZT381 (tnpA) TT17430 recAl/pZT382 (tnfGD163) TT1743 1 recAl/pZT383 (tnpPS167) TI'18925 recAl TT18926 recB503:TnlO TT19079 recD541::TnlOdCm TT18927 recF521::Tn5 TT18928 recJ504:MudJ TT18929 recB503:TnlO recF521::Tn5 TT18930 recB503:TnlO recJ504:MudJ TT18931 DUP173l[(leuAll79)*MudA*(nadC220)] TT18933 DUP1731[(leuA1179)*MudA*(nadC220)]recAl TT18934 DUP173l[(leuAll79)*MudA*(nadC220)]recB503:TnlO TT19080 DUP1731[(leuAll79)*MudA*(nadC220)]recD541::TnlOdCm TT18935 DUP1731[(6uA1179)*MudA*(nadC220)]recF521::Tn5 TT18936 DUP1731[ (leuAll79)*MudA*(nadC220)]recJ504:MudJ TT18937 DUP1731[ (leuAl179)*MudA*(nadC220)]recB503:TnlO recF521::Tn5 TT18938 DUP1731[(leuA1179)*MudA*(nadC220)]recB503:recJ504:MudJ TT18939 DUP1732[(pyrB2694)*MudA*(thr-469)] TT18941 DUPl732 [ (pyrB2694) *Mu&*( thr-469)] recAl TT18942 DUP1732[(pyrB2694)*MudA*(thr-469)]recB503:TnlO TT19081 DUP1732[(pyrB2694)*MudA*(thr-469)]recD541::TnlOdCm TT18943 DUPl732 [ (pyrB2694) *MudA*(thr-469) 3 recF521::Tn5 IT18944 DUP1732[ (pyrB2694)*MudA*( thr-469)] recJ504:MudJ TT18945 DUP1732[ (pyrB2694)*MudA*(thr-469)]recB503:TnlO recF521::Tn5 TT18946 DUPl732 [ (pyrB2694) *MudA*(thr-469)] recB503:Tn 10 recJ504:MudJ TT18947 DUP1732[(pyrB2694)*MudJ*(thr-469)] TT18948 purC7 DUP1732[ (pyrB2694) *MudJ*(thr-469) ] TT18949 purC7 DUPl732[ (pyrB2694 serBl466::TnlWTc)*MudJ*(thr-469)] TT18950 purC7DUP1732[(pyrB2694)*MudJ*(thr-469)]nadA379::TnlOdTc TT18964 purC7DUP1732[(pyrB2694 serB1466::TnlOdTc)*MudJ*(thr-469)]/pZT381 TT18965 purC7 DUPl732[(pyrB2694serB1466::Tn1OdTc)*MudJ*(thr-469)]/pZT382 TT18966 purC7 DUP1732[ (pyrB2694) *MudJ*(thr-469)] /pZT382 TT18967 purC7 DUP1732[ (pyrB2694)*MudJ*( thr-469)] nadA379::Tn1WTc/pZT382 'IT18968 purC7 DUP1732[(pyrB2694 serB1466:TnlOdTc)*MudJ*(thr-469)]recAl/pZT381 TT18969 purC7DUP1732[(pyrB2694 serBl466::TnlWTc)*MudJ*(thr-469)]recAl/pZT382 TT18970 purC7DUP1732[(pyrB2694 serB1466::TnlOdTc)*MudJ*(thr-469)]recBlO/pZT381 'IT18971 purC7DUPl732[(pyrB2694 serBl466::TnlWTc)*MudJ*(thr-469)]recBlO/pZT382 TT18972 purC7 DUP1732[(pyrB2694 serB1466:TnlOdTc)*MudJ*(thr-469)] recD561/ pZT381 Recom bination Pathways 759 Pathways Recombination

TABLE 1 Continued

Strain" Genotype' Source" TT18973 purC7 DUP1732[ (pyrB2694 serB1466::TnIWTc) *MudJ* (thr-469)]recD561/ pZT382 TT18974 purC7 DUP1732[ (pyrB2694serBl466::TnIWTc)*MudJ*( thr-469)] recF558/pZT381 TT18975 purC7 DUP1732 [ (pyrB2694 serB1466::TnIWTc) *MudJ* (thr-469)] recF558/pZT382 TT18976 purC7DUP1732[(pyrB2694 serB1466::TnIWTc)*MudJ*(thr-469)]recBl0 DELl792(phs-sbcB)sb~W8/pZT381 TT18977 purC7 DUP1732[ (pyrB2694serB1466::TnIWTc)*MudJ*( thr-469)] recBl0 DEL1 792(phs-sbcB) sbcCD8/pZT382 TT18978 purC7 DUP1732[ (pyrB2694serBl466::TnlWTc)*MudJ*( thr-469)] recBl0 DELl792(phs-sbcB)sbcW8 recF558/pZT381 TT18979 purC7 DUP1732[ (pyrB2694serB1466::TnIWTc)*MudJ*( thr-469)] recBI0 DELl792(phs-sbcB)sbcW8 recF558/pZT382

~~ "All strains are derivatives of S. typhimun'um LT2. bThe chromosomal structure of strain 'IT18939 is diagrammed in Figure 2. Except where noted, all strains were constructed during the course of this work. and biochemical evidence ( HANIF~RDet al. 1991) argue in the presence of IPTG. The Tnp* mutant enzyme strongly against rejoining of the donor site following catalyzes the excision of the TnlO from the chromo- cleavage. Thus, the fate of the donor site (the DSB) is somal site and leaves a double-strand break. either degradation or repair by homologous recombi- Stimulation of recombination by a specific DSB: In nation. The source of the template homology for this this assay, recombination stimulation requires expres- repair could be either an intact sister chromosome or sion of Tnp* (thetransposase that makes breaks) and an ectopic repeat of the sequence flanking the break a Tn IOdTc ( the break site ) in one of the repeats (see site. Figure 6) . Cultures of DSB assay strains (described in The DSB system requires only a plasmid, pZT382, MATERIALS AND METHODS) were diluted from selective to provide regulated expression of the Tnp* protein medium (counterselecting Lac- Aps segregants) into ( excision-proficient transpositiondefective transpo- nonselective medium containing IPTG to induce trans- sase) anda transpositiondefective TnlOelement (with posase expression. The cultures were incubated and the endsof ISIO) at any chromosomal site. The plasmid their Lac - segregant frequencies were determined as carries the transposase allele under the expression of described above. The uninduced Lac- segregant fre- the Ptac promoter. It also carries the ZucIY gene (see quency was -0.08. The stimulation of Lac- segregation MATERIALSAND METHODS and Figure 5 ) . The expression above this level required a TnlWTc (Figure 6, no of Tnp * protein is repressed without IPTG and induced TnlO) , the plasmid expressing Tnp * (Figure 6,

TABLE 2 W-survival and phage-plaquing phenotypes of rec mutants

P22 plaque morphologies" Strain"Genotype w survivalb typewild P22 P22 em P22 abc c2-5 LT2 rec" 0.3 +++ turbid +++ turbid + clear TT18925 recA 10-6 +++ turbid 0 + clear TT1892610-2 recB +++ turbid +++ turbid +++ clear TT19079 recD 0.3 +++ turbid +++ turbid +++ clear TT18927 rec F 2 x 10-2 +++ turbid +++ turbid + clear 'IT18928 recJ 10-2 +++ turbid ++ turbid 0 TT18929 recB recFTT18929recB 5 X 10-~ +++ turbid +++ turbid +++ clear TT18930 recB recJ 10-~ + + + clear 0 +++ clear "The data tabulated here compare isogenic haploid strains. Data for the two ret+ merodiploid strains TT18931 and lT18939 and their rec derivatives (Table 1) were similar. In all cases, identical results were obtained for eachof three independent isolates from the crosses performed to construct these strains. Relative survival of UV-irradiated bacteria to unirradiated controls. See MATERIALS AND METHODS. Plaque morphology isindicated as follows: + + +, large; + +, small; +, pinpoint; 0, no plaques. SeeMATERIALS AND METHODS for the full genotypes of the P22 strains. 760 T. Galitski and J.R. Roth

RGURE3.-Colony-sec- toringrecombination assay.Recombination be- tween the direct chromo- somalrepeats of strain TT18939(DUP1732 [ (BrB2694)*MudA* ( thr- 469) ] 1 generates Lac- clones within colonies. Theserecombinants ap pear as whitesectors on NB Xgal plates. (A) rec+ (TT18939). (B) redl (TT18942).(C) rec- B503::TnlO (TT18943). (D) re cF5 2 1 :: T n5 (TT18944). (E) rec- J504:MudJ (TT18944). (F) recB503::TnlO rec- F521::Tn5 (TT18945).

TnpA ) , and IPTG to induce Tnp * expression (Figure at serB) increases the segregant fryquency from "0.08 6, no IPTG) . Also, the stimulation required a TnlWTc to "0.4; one-third of the cells (0.4 minus 0.08) are in one of the repeats; a TnladTc far from the repeats DSB-stimulated Lac-recombinants. This increase (Figure 6, TnlO atnadA) had no effect. Strains lacking above backgrounddoes not depend on the duration of any essentialcomponent show no stimulation of recom- exponential growth. This is because transposase does bination above background spontaneous frequencies. not work continuously during growth; it appears to in- Strains with all these elementsand a ret+ genotype show duce DSBs in bursts, one upon initial exposure to IPTG, a substantial increasein Lac- segregant frequencydue and one near the end of exponential phase growth to double-strand breakageof one of the repeats (Figure (data not shown) . The efficiency of double-strand 6, TnlO at serB) . breakage must be at least as high as the frequency of The absolute increase in the Lac- recombinant fre- DSBstimulated Lac- recombinants (one-third) ; some quency above the background level (seen in Figure 6) doublestrand breaks will be repaired either using the is of interest; the relative increase is arbitrary. Under same repeat on a sister chromosome (not an ectopic the specified conditions (MATERIALS AND METHODS) the repeat), or using an ectopic repeat but without the background Lac- segregant frequency is -0.08. These exchange of flanking markers. spontaneous segregants (background) accumulate dur- Induction of transposase activity in the strains tested ing exponential growth (see above) ; the background shows very little effect on the course of exponential Lac - segregant frequency canbe controlled arbitrarily growth; there is at most a brief cessation of growth by varying the number of generations of exponential (data not shown). This probably reflects the fact that growth (data not shown) . The induction of a transpo- transposase cutting is not 100% efficient and many in- sase-catalyzed DSBin one of the repeats (Figure 6, TnlO troduced breaks can be efficiently repaired. Recombination Pathways 761

TABLE 3 spontaneousLac- recombinants. An isogenic strain Lac- segregant frequencies in cultures of a ret' duplication with the pZT382 plasmid (Tnp*) showed a Lac - re- strain and rec derivatives" combinant frequency comprised of this background fre- quency of spontaneous events plus the frequency of CulturerecB rec" recF recB recF events stimulated by a double-strand break. Thus, for 1 0.058 0.016 0.035 0.0053 each rec genotype, the DSB-stimulated Lac- recombi- 2 0.060 0.022 0.041 0.0054 nant frequency was the Lac- frequency of the strain 3 0.062 3 0.026 0.044 0.0062 with Tnp* minus the Lac - frequency of the strain with 4 0.065 4 0.030 0.053 0.0067 TnpA (Figure 7) . 5 0.067 0.033 0.053 0.0067 The effects of various rec mutations on DSB-stimu- 6 0.067 6 0.036 0.053 0.0068 lated recombination between direct repeats support the 0.0069 7 0.068 7 0.037 0.054 recombination model. Figure 7 shows that recombina- 8 0.069 0.037 0.056 0.0073 9 0.071 0.043 0.060 0.0074 tion initiated by a double-strand break depends primar- 10 0.071 0.047 0.061 0.0080 ily on RecB function. A recF mutation had a twofold Medianb 0.071 0.048 0.061 0.0082 effect on the frequency ofDSB-stimulated events, 11 0.071 0.048 0.061 0.0084 whereas a recB mutation nearly eliminated this stimula- 12 0.073 0.048 0.068 0.0088 tion. As expected, sbc suppressor mutations restored 13 0.073 0.048 0.069 0.0088 DSB-stimulated recombination to arecB mutant to levels 14 0.083 0.049 0.076 0.0092 approaching the wild type. As predicted by the model, 15 0.083 0.059 0.077 0.0096 16 0.084 0.060 0.078 0.0120 this restored DSB-stimulated recombination activity in 17 0.084 0.063 0.080 0.0140 a recB sbcB sbcC strain depended on RecF function. As 18 0.085 0.063 0.091 0.0140 was observed for spontaneous recombination events, 19 0.087 0.220 0.096 0.0180 DSB-stimulated recombination depends completely on 20 0.100 0.350 0.120 0.7200 RecA function. Essentially no Lac- segregants are ob- The full genotypesof strain TT18931 (DUP1731ret+) and served in a recA mutant under any tested conditions. its rec derivatives, TT18934(DUP1731 recB), TT18935 The effect of a redl mutation, eliminating the exo- (DUP1731 recF), andTT18937 (DUP1731 recBrecF), are listed nuclease V activity (double-stranded DNA exo- in Table 1. nuclease) of the RecBCD enzyme, was a twofold de- The median Lac- segregant frequency from 20 indepen- crease in DSB-stimulated recombination. These find- dent cultures sorted in ascending order (see MATERIALS AND METHODS). ings are discussed below.

DISCUSSION Effects ofmc mutations on DSB-stimdated recombi- nation: For each rec genotype studied, two strains were Recombinationbetween chromosomal direct re- tested in assaysof DSB-stimulated recombination.A peats We have studied homologous recombination strain with the pZT381 plasmid ( TnpA ) served as a events between intact extensive chromosomal direct re- negative control to show the background frequency of peats flanking join-point markers. The detected events

TABLE 4 Effect of rec mutations on spontaneous recombination between chromosomal direct repeats

Colony-sectoringsegregationLac- Relative recombination Strain rec genotype assay" proficiency' rate'

TT18931 rec" 4.9 X 10-~ 1 TT18933 recA <4.6 X <0.00094 TT18934 recB 3.4 X 10-3 0.69 TT19080 redl ND ND TT18935 rec F 4.4 X 10-3 0.90 TT18936 recJ + ND ND TT18937 recB recF 0 5.9 X 10-4 0.12 Strains were scored for the presence of white Lac- segregant sectors in colonies on NB Xgal plates as illustrated in Figure 3. Lac- segregation rate, segregants per cell per generation (2 X median segregant frequency X genera- tions"). The number of culture generations was calculated fromdeterminations of the final and initial colony- forming units in cultures. ND, no data. "Relative recombination proficiency is the ratio of the Lac- segregation rate to that of strain "I8931 (DUP1731 rec"). 762 T. Galitski and J. R. Roth

A homologous duplexes would givethe Lac- recombinants. For the reasons out- 1-1 1-1 lined below, we think this possibility is unlikely. 1-1 1-1 1. This alternative model might predict that thehomol- double-strand ogous strand invasioncatalyst, the RecA protein, would be superfluous. However, both spontaneous and DSB-stimulated events observed here depended absolutely on RecA activity. 2. The resection and single-strand annealing model predicts that recombination should depend on exo- nuclease activities. The exonuclease V activity of the RecBCD enzyme wouldbe the best candidate since it is the major exonuclease of E. coli and S. typhimurium et k \ (reviewed in KOWALCZYKOWSKI al. 1994). This is donor DNA especially true for recombination events stimulated li=l) recipient chromosome by a DSB, the RecBCD substrate. Loss of the RecD subunit of RecBCD eliminates exonuclease V activ- ity, but notRecBGmediated recombination ( AMUND SEN et a[. 1986; LOVETT et al. 1988) . Contrary to the predictions of the single-strand annealing model, a 1-1 1-1 I recD mutation eliminated neither spontaneous nor I DSB-stimulated recombination between chromo- t RecFt somal direct repeats. 3. Recombinant frequencies are independent of the ‘X\ ‘X,1- length of material between the repeats; this was tested by comparing segregant frequencies of identi- FIGURE 4.-A model for bacterial homologous recombina- cal duplications, one with a MudJ (10 kb) and one tion pathways. DNA strand polarities are indicated with a half arrow at the 3‘ end. Strands that are degraded or lost are with a MudA (38 kb) at the join point (data not represented by dashed lines. x sites and their orientations shown ) . The single-strand annealing model predicts are indicated. (A) Chromosomal repair and rearrangement a decrease with increasing distance between repeats. through homologousrecombination initiated by double- strand breaks or single-strand gaps. The RecBCD pathway and We conclude that the recombination events scored the RecF pathway are alternatives that operate on different in this system occur through homologous strand inva- initiating DNA lesions. (B) Sexual recombination by integra- sion and exchange of flanking markers. tion of donor DNA at thehomologous region of the recipient Two major recombination pathways: Classical sexual chromosome. Events depicted are at one endof the donated fragment. Because the donor DNA has double-strand ends, recombination assays (in which double-strand ends are these events depend primarily on RecBCD function. In the provided as substrates) revealed a major RecBCD path- absence of RecBCD function, RecJ exonuclease produces a way in wild-type strains and a minor or cryptic RecF suitable substrate for RecF pathway recombination; this sub- pathway that includes most of the known recombina- strate can be stabilized by loss of SbcB exonuclease activity. tion functions. This curious state of affairs suggeststhat recombination pathways can be studied more profitably using an assay system that more closely represents their eliminate the join-point markers located between the natural roles. The recombination functions of E. coli repeats. A DNA lesion (either spontaneous or experi- and S. typhimurium probably evolved primarily to accom- mentally induced) in one of the repeats provides a plish DNA repair, and the formation and reversion of substrate that initiates homologous strand invasion at gene duplications. The assay presented here closely ap- the other repeat followed by resolution to exchange proximates this inferred natural role of homologous flanking sequences. recombination. An alternative model for the recombination events The effects of rec mutations on recombination be- studied here involves resection and single-strand an- tween chromosomal direct repeats suggest that both nealing ( SYMINGTONet al. 1985) , For example, a dou- the RecBCD pathway and the RecF pathway are major

ble-strand break might allow rather extensive RecBCD recombination pathways in wild-typebacteria. Both recB catalyzed degradation such that the duplication join mutants and recF mutants are Rec + by this assay; recB point (Lac+) is lost and complementary single-strand recF double mutants are Rec-. Thus, recombination tails derived from the two repeats are available to pair. between direct repeats can occur bytwo alternative Simple single-strand annealing (not homologous equally significant routes, one RecB-dependent and strand invasion) followed by gap filling and ligation one RecF-dependent. Furthermore, in this assay, the Recombination Pathways 763

IPTG induces Tnp* expression

double-strand break

FIGURE5.-A system for the regulated induction of a single predetermined double-strand break in the bacterial chromosome. Plasmid pZT382 carries a mutant TnlO transposase gene, tnp *. This gene is transcribed from the Ptac promoter that is repressed by LacI repressor protein encoded by the ladqgene (which overproduces LacI protein). The addition of IPTG inactivates LacI and thereby induces Tnp* protein expression. This mutant Tnp* transposase catalyzes the excision of a TnlOdTc element from the bacterial chromosome and leaves a double-strand break. Because Tnp* protein is defective for transposition, excision also produces an abortive excised transposon fragment (not shown).

RecF-dependent alternative does not depend on sup RecJ and RecN functions may be needed to allow useof pressors ( sbcB and sbcC) needed to activate this pathway double-strand end substrates in a recBC sbcB sbcC strain. for sexual recombination. One should note, also, that Substrate specificities: The biochemistry and genet- the recombination defectof recB recFdouble mutants is ics of the RecBCD enzyme and the RecF protein suggest not as severe asthat of a recA mutant. Some homologous that the primary substrate of the RecBCD pathway is a recombination activity remains in recB recF double mu- double-strand break or the double-strand ends of donor tants. The source of this could be either distincta minor DNA, whereas the primary substrate of the RecF path- recombination pathway or partial activity of one of the way is a chromosomal single-strand gap or single-strand major pathways. Partial activity could be due to func- tails of donated DNA. An extensive literature (reviewed tional overlap with a related enzyme. in KOWALCZYKOWSKJet al. 1994; MYERS and STAHL1994) The RecF pathway redefined: The RecF pathway as documents the activity of the RecBCD enzyme at DNA observed in this work not only has an elevated status, double-strand ends. Recently, the biochemical interac- it depends on a differentset of enzymes. The classically tions ofRecF proteinand DNAhave beenstudied defined RecF pathwayfunctions include theRecJ single- (GRIFFINand KOLODNER1990; MADIRAJU and CLARK strand exonuclease and the RecN protein of unknown 1991; HECDEet al. 1996). TheRecF protein binds pref- function. Mutations of either recErecJ (HORIIand erentially to double-stranded DNA with a single-strand CLARK1973) or recN (LLOYDet al. 1983) knock out gap. This affinity exceeds that for either double- sexual recombination in recBCsbcB sbcC strains. How- stranded DNA (dsDNA) or single-stranded DNA ever, in the direct-repeat assay, neither a recJmutation (ssDNA). This implies that RecF binds to dsDNA- nor a recN mutation has the same effect as a recFmuta- ssDNA junctions. Thus, RecF is poised to catalyze re- tion. Whereas a recF mutation combined with a recB combination initiated at DNA gaps, whereas the mutation causes a Rec- phenotype in the direct-repeat RecBCD enzyme is known to produce ssDNA tails after assay, recBrecJ double mutants and recBrecN double loading at adouble-strand end anddegrading the DNA mutants are Rec +. Thus, RecF-dependent sexual re- until encountering ax site ( DIXONand KOWALCZYKOW- combination and RecF-dependent recombination be- SKI 1993; KUZMINOVet al. 1994). tween chromosomal direct repeats are not equivalent. The functional dependencies of double-strand-break- We propose that the defining difference is the nature induced recombination between chromosomal direct of the initiating DNA substrate. In the sexual assay, the repeats support the proposed substrate specificities of 764 T. Galitski and J. R. Roth

0.5. 0.35 - T T 0.3 0.4 - h C e, 2 0.25 0.3 2s &E ar 2 & 0.2 . iM 3Y g 0.2 2 g&- 0.15 a2 3 0.1 ' 0.1 . 4Y 0.05 0 noTnlO TnpA noIPTG TnlOat TnlOat nadA serB 0. - recB recD recF recB recB FIGURE6.-Characterization of the double-strand break sbcB sbcB system. In a strain carrying direct-order chromosomal repeats sbcCsbcC flanking a Lac+ join-point element,a DSB in one of the re- recF peats (TnlO at SUB, TT18965; see Figure 5) stimulates the segregation of Lac- recombinants. Background spontaneous FIGURE7.-Effects of rec mutations on recombination stimu- levels of recombination are observed in strains ( 1) without a lated by a double-strand break. Mutant derivativesof the strain DSB site (no Tn10, TT18966),(2) expressing a null transpo- depicted in Figure 5 were constructed. The frequencies of sase allele (TnpA, TT18964), (3) in the absence of Tnp* Lac ~ segregation events stimulated by a double-strand break expression (no IPTG, TT18965), and(4) with a DSB site far (see MATERIALS AND METHODS) are shown for the rec+ parent, from the duplicated interval (TnlO at nu&, TT18967). TT18965 {purC7 DUP1732[(pyrB2694serB1466::TnIOdTc) *Mu&* (thr-469)] / pZT382 1, and derivatives IT18969(redl), n18971 (recBI0), TT18973 (recD561),TT18975 (recF558), the RecBCD and RecF pathways. A specific DSB in one TT18977 [recBIO DELl792(phs-sbcB) sbcCD81, and TT18979 of the repeats stimulates recombination that depends [ recBl0 DEL1792 ( phs-sbcB) sbcCD8 recF5581. primarily on RecB function, not RecF function. The RecB dependence of DSB-stimulated recombination is recD defect in double-strand-break-induced duplication due to the recombinase activity of the RecBC enzyme, segregation (Figure 7) could reflect an increase in not the exonuclease activity of the RecBCD enzyme. equal sister-chromosome exchanges. This in indicated by thegreater severityof the recB Control experiments demonstrated that the stimula- mutant defect compared to that of a recD mutant. A tion of recombination by a DSB was due to the initiation recFmutation caused a minor defectin DSB-stimulated of recombination at the site of the DSB. One might recombination; this may reflect conversion of some have predicted that the stimulation of recombination flush double-strand ends to ends with a single-strand is due to secondary effects of a DSB such as the SOS overhang, which can serveas a RecF substrate. This induction caused by Tn 10/ IS 10excision ( ROBERTSand conversion may be more efficient, or the single-strand KLECKNER 1988).However, testsshowed that thestimu- tail may be more stable in a recB sbcB sbcC strain. Thus, lation was a cis effect (at the site of the DSB) , not atrans a recB sbcB sbcC strain showed a restoration ofDSB- effect (like SOS induction) ; there was no stimulation of stimulated recombination; this recombination activity recombination if the DSB site was far from the recom- depended on RecF function. bining repeats. The same conclusion was reached by Throughout this article we have discussed mutations HACEMANNand CRAIG(1993) studying the effects of that cause a reduced frequencyof duplication segrega- Tn 7 transposition ( through a similar mechanism leav- tion and we have assumed that the defect is due to a ing a double-strand break) on recombination. Further- failure of recombinational repair. However, it is possi- more, the fact that induced DSBs increase the propor- ble that a mutation could reducethe frequency of dupli- tion of recombinants indicates that the frequency of cation segregation by increasing the fidelity of recombi- initiating DNA lesions limits the frequency of recombi- nation,that is, reducingunequal sister strand ex- nation events. changes in favor of equal sister strand exchanges. We A recombination model: The key feature of the re- think this alternative is unlikely to explain the defect combination model presentedin Figure 4 is that differ- seen in recB, recF single or double mutants, because ent pathways operate ondifferent initiating DNA struc- these mutations generally cause a defect in recombina- tures. The experiments and model described here do tion regardless of the assay used. However a recD muta- not address the final processing of these substrates to tion stimulates sexual recombination; thus the minor mature recombinant products. Figure 4A depicts spon- Recom bination Pathways Recombination 765 taneous recombination between generic homologous ultimately can repair the damage. Failure to repair a sequences. These sequences could be intermolecular double-strand break may be amore life-threatening or intramolecular;they could be ectopic or at the same problem. The poor viability of recB mutants (data not position on sister chromosomes. Figure 4B shows the shown)supports this. However, theoccurrence of a sexual recombination of a linear homologous donor double-strand break is not always a terminal condition DNA and an intact recipient chromosome. in the absence of RecBCD activity. This is indicated by DNA double-strand ends may be generated directly, the genetics of sexual recombination and DSB-stimu- or indirectly through a single-strand gap intermediate. lated recombination. In recB mutants, residual recombi- The mechanisms of rolling circle replication, endonu- nation initiated at double-strand ends can be restored clease cleavage, transposition, transduction, and conju- to high efficiency by the addition of suppressor muta- gation produce DNA molecules with double-strand tions, sbcB and sbcC ( KUSHNERet nl. 1971, 1972; Figure ends. Indirectly, single-strand gaps may be converted to 7). Conventional thought on this is that exonucleases double-strand breaks through DNA replication, like RecJ can degrade one strand leaving a recombina- nuclease activity, or recombination. genic 3’-ended single-strand tail, allowing RecF-medi- It has been pointed out previously that a replication ated recombination. The elimination of SbcB’s 3 ’ exo- fork encountering a single-strand gap in the template nuclease activitystabilizes this end. Alternatively, we molecule will collapse; this leaves one intact daughter propose that the effect is not strictly the production chromosome and one terminating at a double-strand and stabilization of a single-strand tail; rather, it is the end (reviewed in AsAI et al. 1994; KUZMINOV1995). generation and stabilization of a dsDNA-ssDNA junc- The RecBCD enzyme can load at the broken end and tion, a suitable RecF substrate. The role of the RecF degrade the molecule until encounteringx sitea where protein may be either to catalyze strand separationor to it will start producing single-stranded DNA through its facilitate RecA filament formation on a single-stranded helicase activity ( DIXONand KOWALCZYKOWSKI1993). region. This single-stranded DNA is a substrate for RecA-cata- The phenotypes of recB recJdouble mutants further lyzed homologous strand invasion of the intact mole- support the existence of alternatives to RecBCD for cule. The resulting D-loop structure with an invading 3’ repair of double-strand breaks. RecJmay be a RecF end may serve as a reinitiation site for DNA replication. pathway functionin sexual assays only because it is Nuclease activities could convert a single-strand gap needed to convert a flush double-strand end to an end to a double-strand break with flush ends. BENSONand with a recessed single-strand (a dsDNA-ssDNA junc- ROTH ( 1994) have proposed that the SbcB protein has tion) for RecF binding. This idea predicts that repair two nuclease activities, its known 3 ’ to 5‘ exonuclease of DSBs depends primarily on RecBCD and secondarily activity and a putative endonuclease activity. The puta- (in the absence of RecBCD) on RecJ, whereas recombi- tive endonuclease activity cuts the continuous strand nation initiated at single-strand gaps depends on nei- opposite a gap at thejunction with the 3’ endthereby ther of these. This prediction is supported by the follow- generating one flush double-strand end and one with ing. ( 1) The recB recJdouble mutants show SOS-consti- a 3 ‘ single-strand tail. The exonuclease I activity (3’ to tutivity and extremely low viability, lower than that of 5 ’ single-strand exonuclease; KUSHNERet al. 1972) of either recB mutants or recJ mutants. (2) The recB recJ SbcB degrades this tail to produce a second flush dou- doublemutants are nonpermissive for P22 erf. The ble-strand end. Thus, SbcB might have two activities injected linear P22 genome of this mutant phage re- that serve to prepare flush double-strand ends, known quires recircularization by host-mediated recombina- substrates for the RecBCD enzyme. tion at double-strand-ended terminal redundancies; Nonreciprocal recombination by the RecF pathway this occurs in single recB or recJ mutants but not in (or any pathway) would produce onerecombinant mol- the recB recJ double mutant. ( 3) The recB recJ double ecule and one broken molecule. If the ends of this mutants, although sick, are still recombination profi- molecule are not flush, single-strand exonucleases like cientin the colony-sectoring assay.We suggest that RecJ ( 5 ’ to 3 ’ ; LOVETTand KOLODNER1989 ) or SbcB these exchanges are stimulated by single-strand gaps ( 3 ’ to 5 ’ ; KUSHNERet al. 1972) can make them flush. that continueto be effective in initiating recombination The broken molecule will then be a substrate for a in a recB recJdouble mutant. subsequent recombination event mediated by RecBCD. The assay system described here provides a way to The first recombinant molecule could serve as a target score recombination events that are initiated by struc- for the second exchange. This scenario resembles the tures other than double-strand breaks. This is unlike two-step model for recombination proposedby MAHAN sexual recombination assays that all provide substrates and ROTH ( 198913). with double-strand ends. The single-strand gap is likely Gaps may be converted easily to chromosome breaks. to be the predominant, recombination-initiating alter- Thus, recf’mutants, which fail to repair gaps, have good native to a double-strand break. Single-strand gaps are viability ( data not shown)because the RecBCD pathway formed when a replication fork traverses a region con- 766 T. Galitski and J. R. Roth taining damaged (non-pairing) basesin one strand. mosomal IS200 elements supports frequent duplication forma- tion in Salmonella typhimurium. Genetics 141: 1245-1252. Single-strand excision tracts formed during the repair HAGEMANN,A. T., and N. L. CRAIG, 1993 Tn 7 transposition creates of damaged DNA bases alsomay serve as recombination a hotspot for homologous recombination at the transposon do- substrates. We have used the colony-sectoring assay as nor site. Genetics 133: 9-16. HANIFORD, D.B., A. R. CHELOUCHEand N. KLECKNER, 1989 A spe- a screen for mutants with low rates of recombination. cific class of IS10 transposase mutants areblocked for target site Preliminary characterization of such mutants indicates interactions and promote formation of an excised transposon that oxidative DNA damage may be the major source fragment. Cell 59 385-394. HANIFORD, D.B., H. W. BENJAMINand N. KLECKNER,1991 Kinetic of endogenous recombination-initiating lesions. and structural analysis of a cleaved donor intermediate and a strand transfer intermediate in TnlOtransposition. Cell 64 171- We thank N. BENSON,R. DAWSON,K. HAACK, D. HANIFORD,N. 179. KLECKNER, A. KUZMINOV,R.MYERS, L. MIESEI.,G. SMITH,F. STAHL, HEGDE,S. P., M. RA~AGOPALANand M. V. V. S. MADIRAJU,1996 Prefer- and D. THALERfor sharing strains, plasmids, and influential discus ential binding of Escherichia coli RecF protein to gapped DNA in sions. This work was supported by grant GM-27068 (J.R.R.) and pre- the presence of adenosine ( y-thio) triphosphate. J. Bacteriol. doctoral training grant 5 T32 GM07464-18 (T.G.) from the National 178: 184-190. Institutes of Health. HONG,J-S., and B. N. AMES, 1971 Localized mutagenesis of any small region of the bacterial chromosome. Proc. Natl. Acad. Sci. USA 68: 3158-3162. LITERATURE CITED HORII,Z. I., and A. J. CLARK,1973 Genetic analysis of the RecF path- way of genetic recombination in Escherichia coli K12: isolation AMUNDSEN,S. K., A. F. TAYLOR, A. M. CHAUDHURYand G. R. SMITH, and characterization of mutants. J. Mol. Biol. 80: 327-344. 1986 recD: a gene for anessential third subunit of exonuclease HUGHES,K. T., and J. R. ROTH, 1984 Conditionally transposition- V. Proc. Natl. Acad. Sci. USA 83: 5558-5562. defective derivative of Mud1 (Amp Lac).J. Bacteriol. 159: 130- ANDERSON,P., and J. R. ROTH, 1981 Spontaneous tandem genetic 137. duplications in Salmonella typhimurium arise by unequal recombi- HUGHES,K. T., and J. R. ROTH, 1985 Directed formation of dele- nation between rRNA (rm)cistrons. Proc. Natl. Acad. Sci. USA tions and duplications using Mud(Ap, lac). Genetics 109: 263- 78: 3113-3117. 282. ASAI,T., D. B.BATES and T. KOGOMA,1994 DNA replication trig- KOWALCZYKOWSKI,S. C., D. A. DIXON,A. K. EGGLESTON,S. D. LAUDER gered by double-strand breaks in E. coli: dependence onhomolo- and W.M. REHRAUER, 1994 Biochemistry of homologous re- gous recombination functions. Cell 78 1051-1061. combination in Escherichia coli. Microbiol. Rev. 58: 401-465. BELTRAN,P., J. M. MUSSER,R. HELMUTH,J.J. FARMER,W. M. FRERICHS KUSHNER,S. R., H. NAGAISHI,A. TEMPLINand A. J. CLARK,1971 Ge- rt al., 1988 Toward a population genetic analysis of Salmonella: netic recombination in Escherichia coli: the role of exonuclease genetic diversity and relationships among strains of serotypes S. I. Proc. Natl. Acad. Sci. USA 68: 824-827. choleraesuis, S. derby, S. dublin, S. enteritidis, S. heidelberg, S. infantis, KUSHNER,S. R., H. NAGAISHIand A. J. CLARK,1972 Indirect suppres- S. naupod, typhimurium. 85: and S. Proc. Natl. Acad. Sci. USA sion of recB and recC mutations by exonuclease I deficiency. Proc. 7753-7757. Natl. Acad. Sci. USA 69: 1366-1370. BENDER,J., J. Kuo and N. KLECKNER, 1991 Genetic evidence against KUZMINOV, 1995A,, Collapse and repair of replication forks in Esche- intramolecular rejoining of the donor DNA molecule following richia coli. Mol. Microbiol. 16: 373-384. IS10 transposition. Genetics 128: 687-694. KUZMINOV, E.A,, SCHABTACH and F.W. STAHL,1994 x sites in combi- BENSON,N. R., and J. R. ROTH, 1994 Suppressors of recB mutations nation with RecA protein increase the sunrival of linear DNA in in Salmonella typhimurium. Genetics 138 11-29. Escherichia coli by inactivating exoV activity of RecBCD nuclease. BERKOWITZ,D., J. M. HUSHON,H. J.J. WHITFIELD,J. ROTH and B. N. EMBO J. 13: 2764-2776. AMES,1968 Procedurefor identifymg nonsensemutations. J. LEDERBERG,J., and E. L. TATUM,1946 Gene recombination in Esche- Bacteriol. 96: 215-220. richia coli. Nature 158 558. BIANAR,M. A,, S. J. SANDLER,M. E. ARMENGOD,L. W. REAMand A. J. LLOYD,R. G., and C. BUCKMAN,1985 Identification and genetic CIARK,1984 Molecular analysis of the recF gene of Escherichia analysis of sbcC mutations in commonly used recBC sbcB strains coli. Proc. Natl. Acad. Sci. USA 81: 4622-4626. of Escherichia coli K12. Bacteriol. 164: 836-844. BOTSTKIN,D., and M. J. MATZ,1970 A recombination function es- J. LLOYD,R. G., and A. THOMAS,1984 A molecular model for conjuga- sential to the growth of bacteriophage P22. J. Mol. Biol. 54: 417- tional recombination in E. coli K12. Mol. Gen. Genet. 197: 328- 440. CASTILHO,B. A,, P. OLFSONand M. J. CASADABAN,1984 Plasmid in- 336. sertion mutagenesis and lac gene fusion with mini-Mu bacterio- LLOYD, R. G., S. M. PICKSLEYand C. PRESCOTT,1983 Inducible ex- phage transposons. J. Bacteriol. 158: 488-495. pression of a gene specific to the RecF pathway for recombina- CIARK,A. J., 1971 Toward a metabolic interpretation of genetic re- tion in Escherichia coli K12. Mol. Gen. Genet. 190: 162-167. combination. Annu. Rev. Microbiol. 25: 437-464. LOVETT,S. T., and R. D. KOLODNER,1989 Identification and purifi- CIARK,A. J., 1973 Recombination deficient mutants of E. coli and cation of a single-stranded-DNA-specific exonuclease encoded other bacteria. Annu. Rev. Genet. 7: 67-86. by the redgene of Eschm'chia coli. Proc. Natl. Acad. Sci. USA 86: DAVIS,R. W., D. BOTSTEINand J. R. ROTH, 1980 Advanced Bacterial 2627-2631. Cmrtics. Cold Spring Harbor Laboratory, Cold Spring Harbor, LOVEIT, S. T., C. LUISI-DELUCAand R. D. KOLODNER,1988 The ge- netic dependence of recombination inrecDmutants of Escherichia NY. DIXON,D. A., and S. C. KOWALCZYKOWSKI,1993 The recombination coli. Genetics 120 37-45. hotspot x is a regulatory sequence that acts by attenuating the WIRAJU, M. V. V. S., and A. J. CLARK,1991 Effect of RecF protein nuclease activity of the E. coli RecBCD enzyme. Cell 73: 87-96. on reactions catalyzed by RecA protein. Nucleic Acids Res. 19: EISENSTARK,A,, R. EISENSTARK,J. VAN DILLEWJN and A. RORSH,1969 6295-6300. Radiation-sensitive and recombinationless mutants of Salmonella MAHAN, M. J., and J. R. ROTH, 1989a recB and recC genes of Salmo typhimurium. Mutat. Res. 8: 497-504. nella typhimurium. J. Bacteriol. 171: 612-615. ELLIO~,T., and J. R. ROTH, 1988 Characterization of TnlOd-Cam: MAHAN,M. J., and J. R. ROTH, 1989b Role ofRecBC function in a transpositiondefective TnlO specifying chloramphenicol resis- formation of chromosomal rearrangements: a two-step model tance. Mol. Gen. Genet. 213: 332-338. for recombination. Genetics 121: 433-443. GRIFFIN,T. J., and R. D. KOLODNER,1990 Purification and prelimi- MAHAN,M. J., J. CA~ADESUSand J. R. ROTH, 1992 The Salmonella nary characterization of the Esche&hia coli K12 RecF protein. J. typhimurium RecJ function permits growth of P22 abc phage on Bacteriol. 172: 6291-6299. recBCD' hosts. Mol. Gen. Genet. 232: 470-478. HAACK,K. R., and J. R.ROTH, 1995 Recombination between chro- MAHAN, M. J., A. GARZONand J. CASADESUS,1993 Host RecJ is re- R ecom bination Pathways Recombination 767

quired for growth of P22 erf bacteriophage. J. Bacteriol. 175: RUDD,K. E., and R. MENZEL,1987 his operons of Escherichia coli and 288-290. Salmonella typhimuriumare regulated by DNA supercoiling. Proc. MAYNARDSMITH, J., N. H. SMITH,M. O'ROURKE and B. G. SPRA?T, Natl. Acad. Sci. USA 84 517-521. 1993 How clonal are bacteria? Proc. Natl. Acad. Sci. USA 90 SCHMIEGER,H., 1971 A method for detectionof phage mutantswith 4384-4388. altered transducing ability. Mol. Gen. Genet. 110 378-381. MIESEL,L., and J. R. ROTH, 1994 Salmcmella recD mutations increase SCLAFANI,R. A., and J. A. WECHSLER,1981 High frequency of ge- recombination in a short sequence transduction assay. J. Bacte- netic duplications in the dnaB region of the Escherichia coli K12 ri01. 176: 4092-4103. chromosome. Genetics 98: 677-690. MYERS,R. S., and F. W. STAHL, 1994 x and the RecBCD enzyme of SEGALL,A,, 1987 Studies on genome rearrangements and intrachrc- Escherichia coli. Annu. Rev. Genet. 28: 49-70. mosomal recombination in Salmonella typhimunum. PhD. Thesis, MYERS,R. M., L. S. LERMANand T. MANIATIS, 1985 A general University of Utah, Salt Lake City, UT. method for saturation mutagenesis of cloned DNA fragments. SMITH,G. R., 1988 Homologous recombination in procaryotes. Mi- Science 229: 242-247. crobiol. Rev. 52: 1-28. OCHMAN,H., and R. K. SELANDER,1984 Evidence for clonal popula- STARK, M.J. R., 1987 Multicopy expression vectors carrying the lac tion structure in Escherichia coli. Proc. Natl. Acad. Sci. USA 81: repressor gene for regulated high-level expression of genes in 198-201. Eschm'chia coli. Gene 51: 255-267. POTEETE,A. R., A. C. FENTONand K. C. MURPHY, 1988 Modulation SYMINGTON, L., P. MORRISONand R. KOLODNER,1985 Intramolecu- of Escherichia coli RecBCD activity by the bacteriophage X Gam lar recombination of linear DNA catalyzed by the Escherichia coli and P22 Abc functions. J. Bacteriol. 170: 2012-2021. RecE recombination system. J. Mol. Biol. 186 515-525. ROBERTS, D.,and N. KLECKNER,1988 TnIU transposition promotes VOGEL,H. J., and D. M. BONNER,1956 Acetylornithase of Escherichia coli: partial purification and some properties. Biol. Chem. RecA-dependent induction of a X prophage. Proc. Natl. Acad. J. 218: 97- 106. Sci. USA 85 6037-6041. WANG, T-C. V., and K. C. SMITH, 1983 Mechanisms for recmepen- ROTH,J. R., 1970 Genetic techniques in studies of bacterial metabo- dent and recBdependent pathways of postreplication repair in lism. Methods Enzymol. 3-35. 17A: W-irradiated Escherichia coli UWB.J. Bacteriol. 156: 1093-1098. ROTH,J. R., N. BENSON,T. GALITSKI,K. HAACK, J. LAWRENCE et al., WING,J. P., M. LEVINE and H. 0. SMITH,1968 Recombinationdefi- 1996 Rearrangements of the bacterialchromosome-forma- cient mutant of SalmoneUu t#&murium. J. Bacteriol. 95 1828- tion and applications, pp. 2256-2276 in Escherichia coli and Salmo 1834. nella typhimunum: Molecular and Cellular Biology, edited by F. C. NEIDHARDT.American Society of Microbiology, Washington, DC. Communicating editor: N. L. CRAIG