Approaches To-Half-Tetrad Analysis in Bacteria

Approaches to-Half-TetradAnalysis in Bacteria: Recombination Between Repeated, Inverseorder Chromosomal Sequences

Anca M. %gall’ and JohnR. Roth Department of Biology, University of Utah, Salt Lake City, Utah 841 12 Manuscript received March 26, 1993 Accepted for publication October 5, 1993

ABSTRACT In standard bacterial recombination assays, a linear fragment of DNA is transferred to a recipient cell and, at most, a single selected recombinant typeis recovered from each merozygote. This contrasts with fungal systems, for which tetrads allow recovery of all meiotic products, including both ultimate recombinant products of an apparent single act of recombination. We have developed a bacterial recombination system in which two recombining sequences are placed in inverse order at widely separated sites in the circular chromosome of Salmonella typhimurium. Recombination can reassort markers between these repeated sequences (double recombination and apparent gene conversion),or can exchange flanking sequences, leading to inversion of the chromosome segment between the recombining sequences. Sincetwo recombinant products remainin the chromosomeof a recombinant with an inversion, one can, in principle, approachthe capability of tetrad analysis. Using this system, the following observations have been made. (a) When long sequences(40 kb) recombine, conversion frequently accompanies exchange of flanking sequences.(b) When short sequences(5 kb) recombine, conversion rarely accompanies exchange of flanks.(c) Both recA and recB mutations eliminate inversion formation. (d) The frequency of exchanges between short repeats is more sensitive to the distance separating the recombiningsequences in thechromosome. The results are presented with the assumptionthat inversions occur by simpleinteraction oftwo sequences in the samecircular chromosome. In an appendix we discuss mechanistically more complex possibilities, some of which 11 could also apply to standard fungalsystems.

HE study of homologous recombination in fungi mer following cutting at cos. Thus, in these standard T has benefited from the recovery and character- recombination systems, substrates are provided with ization of both products of a single meiotic recombi- features that are unlikely to be present in “normal” nation event associated as haploid spores in an ascus uninfected (or unmated) cells but are likely to influ- (reviewed by STAHL1979; PETES, MALONEand SYM- ence the recombination process. INGTON 1991). In bacteria, the haploid nature of the Transfer of chromosome fragments (mating, trans- genome makesit difficult to recover both productsof duction, transformation) appears to rare be for enteric an exchange. In most bacterial recombination systems, bacteria in nature (OCHMANand SELANDER, 1984; a single selected recombinant type is recovered; the MILKMAN and BRIDGES,1993). Inview of this,it seems second, unselected, product of the same exchange is likely that therecombination systems of these bacteria lost. Exceptions to this ruleare phage crosses, in which evolved, not for mating, but to perform the internal both recombinationproducts may berepresented sister-strand exchanges needed for repair and dupli- among the progeny of a single burst (SARTHYand cation formation, processes for which the initiating MESELSON1976). However, since multiple rounds of substrates (nicks or breaks) must be generated endog- phage recombination occur,the origin of complemen- enously. This endogenous recombination is like that tary recombinant types remains uncertain. Much in- in eukaryotic cells, which also uses internal substrates formation on bacterial recombination has been gath- and does not normally deal with injected fragments. ered from conjugation and transduction experiments, The sexual recombination systems used to study bac- both of which involve interaction between a linear terial recombination may focus on events that differ donated molecule (with recombinogenicdouble- in important respects from the events for which the stranded ends) and the uninterrupted circular bacte- recombination system of bacteria evolved. rial chromosome. Double-stranded endsare also pres- If the bacterialrecombination system normally ent in the infecting lambda phage genome prior to acts primarily on sequences within a single chromo- circularization and in the replicated lambda concata- some, it has the potential to cause chromosome re- ’ Present address: National Institute of Mental Health, Laboratory of arrangements. Selective forces acting to conserve the Molecular Biology, Building 36, Room 1B-08. Bethesda, Maryland 20892. genetic mapmay have secondarily shaped the bacterial

Genetics 136 27-39 (January, 1994) A. M. Segall and J. R. Roth

/ FIGURE1 .-Recombination events be- Full Half tween inversely oriented homologous se- exchange recombinationDouble Conversion exchange quences in the same circularbacterial chromosome. Thick lines represent homologous sequences. Triangles represent insertions of drug resistance elements within the homologous sequences. Note that inversion requires a full exchange, 1 C defined as an event in which both pairs of flanking sequences are rejoined. y lac'+ \ lac+ + Inv c C

Inversion No Inversion No Inversion Lethal recombination mechanisms so as to minimize inver- reflect a paucity of sites for initiating or resolving the sion formation (which destabilizes the map), while required full exchange (SECALL,MAHAN and ROTH permitting the selectively valuable repair and dupli- 1988; L. MIESEL, A.M. SECALLand J. R. ROTH, cation processes (ANDERSONand ROTH 1977; TLSTY, manuscript submitted). ALBERTINIand MILLER 1984; SONTIand ROTH1989). Direct order repeats can recombine to form dupli- We have previously described a recombination de- cations regardless of the chromosomal position of the tection system in which all the recombining sequences recombining sequences (ANDERSONand ROTH 198 1). are located within the circular bacterial chromosome These internal recombinantscan arise by sister strand (SECALL,MAHAN and ROTH 1988).In this system, exchanges inwhich either one pair of flanking se- initiatingsubstrates for recombination (nicks or quences is rejoined (half exchange) or by exchange of breaks) must begenerated endogenously. Two ho- both flanks (MAHANand ROTH 1988, 1989) (see Fig- mologous sequences, containing distinct genetic mark- ure 4). The unrestrictedformation of duplications ers, areplaced in inverse orientation at separatedsites and the fact that they can form by a half exchange in thechromosome (adapted from KONRAD 1969, suggests that they might form by a different mecha- 1977). Recombination betweenthese inverse-order nism from that leading to inversion. sequences can have either of two outcomes (see Figure We have tested this prediction by observing the 1). Some recombinants carry an inversion of the chro- dependence of various internal recombination events mosomal segment between the sites of the recombin- on the RecBCD enzyme, which is required for the ing sequences. These could be caused by a single full major, reciprocal recombination pathway in Esche- crossover event. Other recombinants show reassort- richia coli (CLARK 1973; THALERand STAHL1988; ment of genetic markers between the repeated se- KINGand RICHARDSON1986; TAKAHASHIet al. 1992). quences but have not acquired an inversion. Some of We show here that mutations in the recB gene elimi- these retain all genetic markers and may have been nate inversion formation, but have little effect on the caused by double crossovers; others have lost one of frequency of duplications. We suggest that this differ- the genetic markers and appearto have resulted from ence reflectsthe fact that inversion formation formally gene conversion events. requires a full exchange (both flanks rejoined), which Only a subset of the many chromosomal intervals can beprovided only by the RecBCD pathway. In tested are permissive for inversion. Inverse repeats contrast, duplications can be generated either by a placed at nonpermissive sites can exchange informa- full exchange (using RecBCD) or by a half-exchange tion(apparent double exchanges and conversions), event (RecBCD-independent), in which only one pair but do not generate a recoverable inversion (KON- of flanking sequences is exchanged and two broken RAD 1969, 1977; ZIEG and KUSHNER1978; REBOLLO, ends are generated. FRANCOISand LOUARN 1988; SEGALL, MAHANand Furthermore, we examine the products of recom- ROTH 1988; SECALLand ROTH 1989). The nonper- bination between inversely oriented sequences located missivity of recombination at some sites is not due to at sites that permit inversion of the intervening chro- lethality of the final inversion product,but could mosomal segment (permissive sites). We have assessed Half-Tetrad Analysis in Bacteria 29

TABLE 1 Strain list

TT12262 trp(ED)2490::MudA(lacZ478::Tn5)hisD9953::MudA(lacZ95O::TnlO) srl-201 (inverse order MudA insertions) TT12265 trp(E,D)2484::MudA(lacZ478::Tn5)hisD9953::MudA(lacZ950::TnlO) srl-201 (direct order MudA insertions) TT13420 thr-469::MudA(lacZ478::Tn5)pyrB2691::MudA(lacZ95O::TnlO) srZ-201 (inverse order MudA insertions) TT13422 pyrC2688::MudA(lacZ478::Tn5)hisD9953MudA(lacZ95O::TnlO) sd-201 (inverse order MudA insertions) TT13425 pncB220::MudA(lacZ478::Tn5)hisD9953::MudA(lacZ950::TnlO) srl-201 (inverse order MudA insertions) TT13427 trp(ED)2490::MudA hisD9953::MudJ(lacZ951:TnlO) (inverse order MudA and MudJ insertions) TTl3428 trp(EDj2490::MudA hisD9953MudJ(lacZ950::TnlO)(inverse order MudA and MudJ insertions) TT13430 araC661 ::MudJ(lacZ95O::TnlO) thr-458::MudA (inverse order MudA and MudJ insertions) TT13435 urd661::MudJ(lacZ951::TnlO) his(BH)l26l::MudA trpB164 (inverse order MudA and MudJ insertions) TTI 3437 araC661::MudJ(lacZ95O::TnlO)nadA216::MudA (inverse order MudA and MudJ insertions) TTl3477 urd661::MudJ(lacZ950::TnlO) thr-458::MudA leuA414(am) recB556(del)rbg-53::TnlOdCm (inverse order MudA and MudJ insertions) TT 13478 araC661 ::MudJ(lacZ950::TnlO) thr-458::MudA leuA414(am) recB+ zbg-53::TnlOdCm (inverse order MudA and MudJ insertions) TT13481 araC661::MudJ(lacZ950::TnlO)thr-458::MudA leuA414(am) recAl srl-203::TnlOdCm (inverse order MudA and MudJ insertions) TT13482 araC661 ::MudJ (lacZ950::TnlO) thr-458::MudA leuA414(am) rec+ srl-203::TnlOdCm (inverse order MudA and MudJ insertions) TTl7402 trp(E,D)2484::MudA(lacZ478::TnZ)hisD9953::MudA(lacZ950::TnlO)srl-203::TnlOdCm, recA1 (direct order MudA insertions) TT17403 trp(E,D)2484::MudA(lacZ478::Tn5)hisD9953::MudA(lacZ950::TnlO)srl-203::TnlOdCm, recA+(direct order MudA insertions) TTl7404 trp(E,D)2484::MudA(lucZ478::Tn5)hisD9953::MudA(lacZ950::TnJO) rbg-53::TnlOdCm, recB556(del) (direct order MudA insertions) TTI 7405 trp(E,D)2484::MudA(lacZ478::Tn5)hisD9953::MudA(lacZ950::TnlO) zbg-53::TnlOdCm, recB+ (direct order MudA insertions) TT17432 araC661::Mud~(lacZ95l::TnlO)trp(ED)2490::MudA srZ-201 (inverse order MudA and Mudl insertions) the effect of chromosomal position and length of selectively on green indicator plates (CHANet al. 1972), to recombining sequenceson thefrequency of inversions distinguish phage-free colonies from phage-infected colonies. Phage sensitivity was tested by cross-streaking phage- and on the association of gene conversion with ex- free clones with P22 H5, a clear plaque mutant. P22 lysates change of flanking sequences. were made as described (DAVIS,BOTSTEIN and ROTH1980). The results are described as if inversions were gen- MudAand MudJ phage: The MudA phage (originally erated by single full exchanges between inverse-order Mu dl-8(Ap, lac); HUGHESand ROTH 1984) is a transposi- sequences in the same circular chromosome. While tion-defective, full-sized (39 kb) derivative of Mu dl (Ap, we think it is likely that inversions arise in this way, lac) (CASADABANand COHEN1979). This phage transposes only in hosts carrying an amber suppressor; in Suw' (non- we wish to make it clear that alternative explanations suppressor) hosts the prophage can be inherited in crosses exist. Thesemore complicated alternatives are de- only by recombination and, once acquired, is not subject to scribed and discussed in the APPENDIX. transposition. The MudA variant encodes ampicillin resistance (ApR)and carries the promoter-less lacZYA operon with trp translational start signals. The smaller MudJ element (1 1 MATERIALS AND METHODS kb) is the transposition defective Mud11 1683(Kn, lac) mini- Mud element of CASTILHO,OLFSON and CASADABAN(1984). Bacterial strains: All the strains used in this study are In MudJ, a KnR determinant (derived from Tn5) replaces derived from Salmonellatyphimurium strain LT2.Their most of the Mu genes. The MudA and MudJ phages share genotypes are listed in Table 1. Both Mud elements used are transposition-defective and are not subject to movement a promoterless lac operon at oneend and about 1 kb of Mu in the course of the experiments presented. Genotypes of sequence at the opposite end. When inserted within a gene parental strains were tested fully before each experiment, in the proper orientation, both Mud transposons form op since the strains are complicated and interpretation of re- eron fusions that permit lac operon expression using the sults depends on the correctness of these strains. promoter of the target gene. Such fusion strains form blue Media: We used standard media for Salmonella (VOGEL colonies on Xgal indicator plates, and may or may not grow and BONNER1956; DAVIS, BOTSTEINand ROTH 1980). The on plates containing lactose as a sole carbon source, depend- indicator dye 5-bromo-4-chloro-3-indolyl-gDgalactopyran- ing on the strength of the promoter in question. Transduc- oside (Xgal; Bachem Fine Chemicals) was added at a final tional inheritance of the ca. 40-kb MudA element requires concentration of 25 mg/ml. E medium lacking citrate (BER- two transduced fragments (HUGHES and ROTH 1985). KOWITZ et al. 1968) was supplemented with lactose or sor- Therefore, a high m.0.i. (>lo) was used in all P22 transduc- bitol at0.2% w/v when testing growth on these carbon tion crossesin which we selected for inheritance of the sources. All nutritional supplements were obtained from entire insertion by recombination through flanking homol- Sigma. ogies; a low m.0.i. (<1) was used to transduce MudJ inser- Transduction methods:We used a high frequency trans- tions and to transfer markers (e.g., lacZ::TnlO) internal to ducing derivative of phage P22 (HT105/1 int-201) (SAND- either of the two Mud elements. ERSON and ROTH 1983). Transductants were purified non- Tn5 and TnZO insertions: Insertions of Tnl0 in different 30 A. M. Segall and J. R. Roth deletion intervals of lac2 (FOSTER1977) were kindly pro- inversion endpoints, often an unlinked TnlOdCm element, vided by T. FOSTER.The lac2478::Tn5 insertion in lac2 was to show that such markers were inherited normally. generously donated by D. BERG.Its location in lac2 was Construction of recombination-deficient strains:A recB roughly mapped by cotransduction tests with the above deletion mutation (recB556) was constructed by excision of TnlO insertions to a region between the TnlO insertion in a recB::TnlO insertion (K. HUGHES,unpublished results). deletion interval 6 (lacZ; lac295l::TnlO) and the TnlO in- This recB deletion was introduced intostrains with two Mud sertion in deletion interval 26 (lad;lac295O::TnlO). These elements (direct or inverse order) by cotransduction with a intervals of the lac2 gene were defined by ZIPSER et al. nearby TnlOdCm element (zbg-53::TnlOdCm); the donor (1970). Recently, many TnlO insertion sites withinlac2 have strain in this cross (TT12272) was constructed by Michael been sequenced (BENDERand KLECKNER 1992). The Mahan. Transductants were screened for sensitivity to UV lacZ95O::TnlO is located at the major TnlO hotspot in the to identify those inheriting the linked recB deletion. An lac2 gene, at base pair 3026. The lac295l::TnlO insertion isogenic UVs and UVR pair of transductants was used to is most likely located between base pair 178 and base pair study the effect of a recB mutation on formation of inver- 286, the site of a less prominent TnlO hotspot. sions. Construction of experimental strains:The construction The recAl mutation was moved into appropriate back- of strains used in the recombination screens has been de- grounds by linkage to the srl-203::TnIOdCm element (EL- scribed in detail (SEGALLand ROTH 1989). Briefly, the first LIOTT and ROTH 1988). Transductants (Cm”) were tested MudA prophage was transduced intothe desired back- for UV sensitivity, and isogenic Rec+and Rec- transductants ground selectingAp”. The lac2 gene of this first Mud were compared in the recombination screen. prophage was then inactivated by transducing in a lac2::TnS Linkage disruption tests in rec- strains: The presence insertion. The lac2 gene of the second MudA element was of inversions in the red strain was tested initially by trans- independently inactivated with a lac2::TnlO in a separate ducing the Lac+ clones (derived from strain TT13477) to background, and this element MudA(lac2::TnlO) was trans- Thr+ insmall “patch” transductions as described above. duced as a unit into the strain with the MudA(lacZ::Tn5), (This was possible despite the reduction in total recombi- selecting Tc”. Tetracycline-resistant transductants were nants caused by the rec mutation). Recombinant (Lac+) screened for the presence of Tn5 (kanamycin resistance) clones showing even a slight reduction in transduction fre- and for the auxotrophic requirements conferred by both quency compared to the parent strain (recE) were retested Mud insertions. as donors. P22lysates were grown on the questionable The construction of strains with one MudA element and recombinant Lac+ clones and the MudJ element was trans- one MudJ element was similar. A non-transcribed MudA duced into a wild-type recipient. All the Lac+ clones were insertion (Lac-) was transduced intothe background of efficient donors of Kn”, demonstrating that the strains are choice selecting Ap”. The MudJ(1acZ::TnlO) element was all competent donors and show no linkage disruption. Since transduced into this strain selecting Tc”. Since the MudJ the parental MudJ element canbe inherited withhigh element issmall (1 1 kb), it was frequently inherited by efficiency, without any evidence of linkage disruption, we recombination through flanking homology. Tetracycline conclude that no inversions occurred in the recB strain. resistant transductants were tested for antibiotic resistance P22 lysates were grown on all Lac+ recombinants arising (Ap”, Kn”) and for the presence of auxotrophic require- in the recA strain and used to transduce the MudJ element ments conferred by the twoMud insertions. The final into wild-type LT2 as a test of linkage disruption. For testing product of these constructions was a phenotypically Lac- Lac+ clonesthat arose in a recA background, the initial patch strain with two unexpressed lac regions that could recom- tests (above) could not be done. Clones were usedas donors bine to yield a Lac+ derivative. in transduction crosses to test transmission of the parental Linkage disruption tests: Linkage disruption by an in- Mud insertions to awild-type recipient. Only one of the few version endpoint was diagnosed by a decrease in the effi- Lac+ clonesformed in a racA strain carries an inversion. ciency of repair of the auxotrophic requirements caused by the parental Mud insertions (SCHMIDand ROTH1983). Lac+ clones weretransduced with P22 lysates grown on wild-type RESULTS LT2, and transductants prototrophic for one or the other endpoint were selected. In strains carrying an inversion of The recombination system: By observing recom- the interval between the Mud sequences, prototrophic trans- bination between inverse sequences in the bacterial ductants were recovered >lO-fold less frequently than in chromosome, one can recover both the selected and strains without an inversion. The rare prototrophic trans- unselected products ofa single recombinational event. ductants recovered from inversion strains arise in recipient This systemdetects events in which both pairs of clones inwhich inverted material had re-flipped to the parental orientation. sequencesflanking the recombining site are ex- To test large numbers of independent Lac+ clones, 25 changed (full exchanges, leading to an inversion) and transductions were performed on a single plate spread with events in which no flanking sequences are exchanged approximately lo9 phage particles. Single colonies of the (apparent conversions and double recombinants). If recipients were transferredto individual patches on the only one pair of flanks is exchanged (half exchange), phage plate. The parental noninverted strain was included the chromosome is broken and no recombinant can on the same plate, allowing comparison of the number of Lac+ transductants. If a transduction patch showed evidence be recovered (see Figure 1). of linkage disruption (a reduced number of prototrophic This recombination system has been described in colonies), the Lac+ clone was retested individually on an detail previously (SEGALLand ROTH 1989). It employs entire plate, using an LT2 donor phage at a m.0.i. of 0.1. strainscarrying two Mud(Zac) sequencesinserted To verify that the reduction in the number of prototrophic eitherin inverse or directorientation at separated recombinants was specific for the boundary of there- arrangement (i.e., linkage disruption), putative inversion sites in the bacterial chromosome.Each Mud element clones were also transduced with a marker unlinked to the is inserted within a gene whose disruption causes an Analysis in Bacteria Half-Tetradin Analysis 31 - MudA A B

Long Double exchange Recombining orgene conversion Sequences u A I B n MudJ Short Recombining Sequences H MudA FIGURE2.-Structure wof chromosomes in which recombination between inverse repeats is scored. Top diagram represents strains with two MudA elements, which share roughly 40 kb of homologous sequence. The lower diagram represents strains with one MudA RecornblnatlonUninvolved product copy element and one MudJ element; these elements share about 5 kb Apparent Gene Conversion Without Inversion of homologous sequence in the lac region and about 1 kb of Mu sequence at the opposite end of the elements. In the figure, filled FIGURE%--Sister strand recombination events between in- boxes represent Mu sequences, hatched boxes represent drug re- versely oriented homologoussequences. Note that a single exchange sistance markers and open boxesrepresent lnc operon sequences. between sisters cannot generate an inversion, but a variety of recombination events generate chromosomes that appear to have auxotrophic phenotype. The Mud(1ac) elements are undergone a gene conversion event. In fact, these recombinant chromosomes retain only one product of recombination, so true of two varieties, called MudA and MudJ. The MudA gene conversion cannot be scored. element is a transposition-defective derivative of CAS- ADABAN and COHEN’Soriginal MudI(Ap, lac) (1979) but the silent lac2gene of MudA can provide se- described by HUGH=and ROTH (1984); the ele- quences for repair of the mutant lac2 gene of MudJ ment is nearly 40 kb long. The MudJ element is the (see Figure 2). MudII1683(Kn, lac) of CASTILHO,OLMN and CAS- In this system, we can score three kinds of recom- ADABAN (1 984). This mini-Mud element is roughly 1 1 bination events: single crossovers with and without kb long and carries the neomycin phosphotransferase conversion (inversions), double crossovers, and “ap gene, conferring resistance to kanamycin (KnR).The parent gene conversions” (Figure 1). A simple single two elements, MudA and MudJ, share about 5 kb of crossover (full exchange) between the sites of the homology, including the lactose operon at one end inserted Tn elements generates a Lac+ recombinant and about 1 kb of Mu sequence at the opposite end. and a doubly mutant second lac2 allele; this single (These elements are diagrammed in Figure 2). exchangeinverts the chromosomal region between A Lac+ phenotype for eitherMud element depends the insertion sites. The same inversion event can be on fusing an intact lac2 gene to afunctional host accompanied by loss of one of the insertion elements; promoter that can provide expression. In the parent these are true conversion events associated with the strains used here for recombination tests, the lac se- full exchange. quences of both elements are inactive either because Two crossovers within the Mud elements can simply of a transposon insertion in the lac2 geneor because exchange the transposons reciprocally between the the Mud element is inserted at a chromosomal site that doesnot provide a promoter. In each parent two lac sequences such that one recombinant allele is strain, the two Mud elements are constructed so that Lac+ and the other carries both insertions. A double recombination between them can restore a functional recombinant of this type has no inversion. The third lac2 sequence fused to an active promoter. class of recombinants (apparent geneconversions) are When two MudA elements were used, one lac2 thosethat have no inversion but acquire one lac+ gene carried a TnZO insertion and the other a Tn5 allele associated with loss of the transposon from the insertion. Recombination in the lac2 gene between repaired site. These recombinants appear to reflect the sites of these transposons can restore a functional intrachromosomal gene conversion events. However, lac2 sequence. To select recombination between a this sort of Lac+ recombinant can also be generated MudA and a MudJ element, strainswere used in which by sister strand exchanges as diagrammed in Figure the lac2 gene of MudJ has a functional promoter but 3; the lac+ allele can form either by gene conversion is inactivated by a TnZO insertion. The lac2 gene of or double recombination between different parental the MudA element is intact but has no chromosomal alleles on sister chromosomes. Since the origin of this promoter. Thus, bothelements are individually Lac-, classis unclear, we have termed it “apparent gene 32 A. M. Segall and J. R. Roth conversion” to emphasize that we cannot be surehow version event (93%of gene conversion products). We the recombinant gene arose. have observed this phenomenon previously (SEGALL In yeast, conversion and exchange of flanking se- and ROTH 1989); the rarer removal of the Tn5 is quences are frequently associated (reviewed by PETES, almost certainly due to theclose proximity of the Tn5 MALONE and SYMINGTON1991). We can score this to the end of the homology shared by the Mud ele- association for the full exchange class (carrying an ments. When a TnlO insertion nearer to the end of inversion), since this class maintains both recombina- the MudA element is used, the opposite bias (prefer- tion products. This association between recombina- ential loss of the Tn5) was seen (SEGALLand ROTH tion and conversion is seen in some but not all of the 1989). bacterial situations described here. In the pncB-hisD interval, Lac+ recombinants were Inversions are identified by scoring recombinants selected by their ability to grow using lactose as the for disrupted association of the sequences that flank sole carbon source. While both of these MudA ele- the insertion sites of the parental Mud elements ments are placed so as to be expressed by a chromo- (SCHMIDand ROTH 1983). In the parent strain, the somal promoter strong enough to permit growth on normal association of flanking sequences is demon- lactose, the pncB promoter is much stronger than the strated by transductional repair of each insertion by a his promoter (A. SEGALL,unpublished data). The donated wild-type allele of the insertion site (k., by majority of Lac+ gene convertants (95%) lost kana- transducing a trp::Mud allele to trp’). In a strain with mycin resistance. This bias for conversion of the ele- an inversion endpointat this locus, one of the se- ment located early in the ZacZ sequence may reflect a quences that previously flanked the Mud element will selective advantage of higher /3-galactosidase expres- be moved away by the inversion. This reassociation of sion when growth on lactose as a carbon source is flanking sequences will occur at both of the Mud selected. Alternatively, the high level of transcription elements, making it impossible to repair either of the from the pncB promoter may influence the recombi- parentalauxotrophies by a single transducedfrag- nation event itself by affecting the structure of the ment. (In practice, the auxotrophy of an inversion recombination substrate. Both negative and positive strain can be repaired at a low frequency due to the effects of transcription on recombination frequency presence of rare recipient cells in which recomblna- have been reported (HERMAN1968; DULand DREX- tion has reversed the inversion and restored the pa- LER 1988a,b; STEWARTand ROEDER1989; THOMAS rental combination of flanking sequences.) and ROTHSTEIN1989; Rocco,DE MASSYand NICOLAS Recombinationbetween long inverse-order se- 1992). Firm predictions for the outcome of a cross- quences: We have tested recombination between pairs over cannot yet be made based on its position with of MudA elements placed in inverse order at the ends respect to a promoter. of four different chromosomal segments. Three of The trp-his interval is nonpermissive for inversions the intervals tested were permissive for inversion (P); and has been described previously (SEGALLand ROTH the fourth is nonpermissive (NP). The Lac+ recombi- 1989). Although the frequency of Lac+ clones, 1.1 X nant products recovered are described in Table 2. is similar to the frequency of Lac’ clones in For all the strains in Table 2, the recombining se- permissive intervals, we found no inversions among quences are MudA elements, about 40 kb in length. the Lac+ recombinants. The Lac+ clones were selected To generate a Lac+ recombinant, an exchange must for growth on lactose as sole carbon source. In this occur within the lacZ gene between the sites of the strain,the selection conditions are met onlyif the two transposon insertions, a 1-2 kb region. Lac+ recombinant gene is fused to the his promoter. It should be noted that for allof the first three In 75% of the recombinants, a Lac+ phenotype was strainsdescribed in Table2, full exchangeevents achieved by apparent geneconversion, with the TnlO (inversions) are frequently associated with gene con- elementbeing lost fromthe his::MudA element in version. The inversion class is the only class for which 98% of the recombinants. true gene conversion can be scored. The non-inver- Recombination between short inverse-order ho- sion recombinant class includes double recombinants mologies: Five chromosomal intervals were tested and “apparent gene conversion events,” whichmay using a MudA element in combination with a MudJ arise in several ways. element as regions of homology; four of the tested In the pyrB-thr interval, Lac+ colonies were detected intervals are permissive for inversion. In all these by their blue color onX-gal indicator plates. Since the strains, only the MudJ elements are fused to active color assay requires very little @-galactosidaseactivity, promoters; thus, a selective biasis imposed for the the difference in the strength of the two promoters repair of the MudJ(1acZ) copy. The MudA and MudJ does not bias which of the two lac alleles is repaired. elementsshare about 5 kb of sequence homology In this interval, the TnlO element was removed more (including the lac region), about 8-fold less than the frequently than the Tn5 by the apparent gene con- homology shared by two MudA elements. Reducing Half-Tetrad Analysis in Bacteria 33

TABLE 2 Recombination between two inversely oriented MudA sequences

Frequency of various Lac+ products (X 105)b Inversions Non-inversions Double Interval Interval Interval size recombinantsTotal Lac+ gene Non-conversionsApparent (strain no.)" (min) clones/105 (min) no.)" (strain (allConversions conversionsTcR KnR) (all TcR KnR) pyrB-thr (P) 2 34.0 (63) 4.3 (8) 2.7 (5 TcS KnR) 12.4 (21 TcS KnR) 14.6 (27) (TTl3420) (2 TcR KnS) fincB-hisD (P) 11 9.4 (193) 0.2 (5) 1.9 (33 TcR KnS) 6.3(lllTcRKnS) 0.9(18) (TT13425) (7 (2 TcS KnR) TcS KnR) (5 TcS KnS) TcS(1 2 KnS) pyrC-hisD (P) 24 7.1 (1 20) 0.8 (1 3) 2.9 (41 TcS KnR) 1.0TcS (12 KnR) 2.4 (41) (TT13422) (7 TcR KnS) (1 TcR KnS) (1 TcS KnS) (4 TcS KnS) trp-his (NP) 10 11.0 (1078) 0 0 TcS8.2 (785 KnR) 2.7 (269) (TT12262) (15 TcR (15 (TT12262) KnS) (9 TcS KnS) " P designates intervals permissive for inversions; NP designates intervals nonpermissive for inversions. Numbers in bold denote the frequency of clones in the particular recombinant class. In parentheses are the number and phenotype of Lac+ clones represented in this class. the length of shared sequence did not alter the fre- the frequency of Lac+ clones was 3.4 X Mostof quency of inversions; however, it did influence the the recombinants carry inversions, and only two of types of recombination products recovered. 129 inversions were accompanied by gene conversion. Of inversions betweenlong sequences (MudA- This confirms the rarity of conversion among recom- MudArecombination; Table 2), 38-88% showed binants generated between short sequences. conversion.Of inversions generated by exchanges In the RecA- background, the frequency of Lac+ between short homologous sequences, only I-4% clones is reduced 1 05-fold. Only one of 1 1 Lac+ clones showed associated gene conversion (Table 3). The recovered carried an inversion; 7 others were double relative frequency of apparent conversion events seen recombinants. The remaining 3 Lac+ products were among non-inversion recombinants is also reduced. clones in which the MudA element (originally at the Role of RecA and RecBCD functions in recombi- thr locus) acquired an active promoter, eitherby trans- nation between short inverse repeats:Using a MudJ position to another site or by transposition of an IS10 element (at the ara locus) and a MudA element (at sequence such that its promoter expressed the previ- the thr locus), we tested the effects of recA and recB ously non-transcribed MudA. We have described such mutations on the frequency of Lac+ recombinants. rare events previously (SEGALLand ROTH 1989). The data are shown in Table 4. In the Rec+ strain, The overall frequency of Lac+ recombinants in the

TABLE 3 Recombination between inversely oriented MudA and MudJ sequences

Frequency of Lac+ products(X

~~ Inversions Non-inversions

Apparent Double Apparent Interval IntervalInterval size Total Lac+ Non-conversions Conversionsconversionsrecombinants (strain no.)a clones/105 (min) VCR) VCS) (TCS) VCR) thr-aruC (P)(TTl3430) 2 320.0 (221) 243.2 (168) 4.3 (3) 15.9 (1 1) 56.5 (39) araC-nadA (P) (TTl3437) 17 22.0 (398) 16.4 (296) 0.3 (5) 0.8 (I 4) 4.6 (83) urd-trpE (P) (TTI 7432) 32 5.9 (99) 3.6 (60) 0.06 (1) 0.06 (1) 2.2 (37) araC-hisH (P) (TT13435) 42 1.4 (105) 0.3 (26) 0.02 (1) 0.1 (10) 0.9 (68) trp(ED)hisD (NP) (TT13428) 10 0.02 ( 198) 0 0 (0) 0.015 (154) 0.005 (42) trp(ED)hisDC (NP) (TTI 3427) 10 0.02 (218) 0 0 (0) 0.01 1 (1 19) 0.009 (96) P designates intervals permissive for inversions; NP designates intervals nonpermissive for inversions. ' Numbers in bold type denote thefrequency of a particular recombinant class. Numbers in parentheses denote the number ofLac+ clones tested for each class. This TnlOelement is located early in the lacZ gene. 34 A. M. Segall and J. R. Roth

TABLE 4 Effect of Rec- mutations on recombination between inverse repeatsof MudJ (at am) and MudA (at thr)

Frequency of various Lac+ products (X lo6)“ Lac+ clones/l O6 Apparent Double Genotype cells Inversions conversions recombinants Transpositions Wild typeb (TT13482) 3350 (1 56) 2770 (1 29) 64.4 (3) 515.4 (24) 0 (0) (TT13478) recA (TT13481) 0.053 (1 1) 0.005 (1) 0 (0) 0.034 (7) 0.014 (3) red (TT13477) 44 (1 88) 0 (0) 34.9 (I 49) 9.1 (39) 0 (0) a Numbers in parentheses represent the number of Lac+ clones scored in each class. The two strains listed are constructed to be isorcenic with the recA and red strains listed below. Since they gave essentially identical results, data for the two rec+ strains was combined. -

RecB- strain (4.4 X 10-5) is about 100-fold lower than final recombinant was uninvolved in the exchange, in the RecB’ strain. None of 188 independent Lac+ one cannot score gene conversion. Regardless of the recombinants in the recB strain has an inversion; for nature of the event, the recombinant clone has one the isogenic recB+ strain, 80% of Lac’ recombinants Lac’ allele and one unaffected allele; we call these had an inversion. The majority (79%) of the Lac+ apparent gene convertants (Table 5; second event in recombinants in the recB backgroundhad lost the Figure 4). TnlO element and are scored as “apparent genecon- When a Lac+ recombinant arises by intrachromo- versions.” Although the recB mutation eliminated in- soma1 exchange between direct repeats, the double versions, it had very little effect on the frequency of recombinant class can be identified because both of apparent gene conversion, 64 compared to 35 (per the two lac regions are altered; one becomes lac+ and 1O6 cells). Double recombinant frequencywas strongly the otherinherits both of the parental drug resistance reduced by the recB mutation, from 515 to9 (per 1O6 markers (fourth eventin Figure 4). Intrachromosomal cells). gene conversion (the third event in Figure 4) yields Role of RecA and RecBCD functions in recombi- one (Lac’) recombinant allele. This event cannot be nation between direct repeats: When the recombin- distinguished from the non-duplication class gener- ing repeated sequences are in the same orientation in ated by sister strand exchanges (described above); all the chromosome, three distinguishable outcomes can are scored as “apparent gene conversion” events (see be detected. These are duplications, apparent gene Figure 4). conversions anddouble recombinants; all are dia- The various recombinant types can be easily distin- grammed in Figure 4. A duplication is generated by guished. Since duplications generate a Lac+ allele at arecombination event between sister strandsthat the join point between two tandem copies of a large exchanges one or bothpairs of sequences flanking the chromosome segment, the join point element (Lac+) recombining sequences. When a Lac+ recombinant is is frequently lost by recombination between the two formed by this sort of event, the Lac+ recombinant copies. This segregation event returns the chromo- allele is left at theduplication join point and the other some to its parental haploid state (Lac-; one drug product of the exchange is lost. The second product resistance at each locus) and can be easily detected by is segregated to a non-viable daughter cell with a screening for loss of the Lac+ phenotype. The pres- broken chromosome (in the case of a half exchange) ence of the TnlO or Tn5 in the lac sequences is or to a daughter with a large lethal deletion (in the signaled by the antibiotic-resistance phenotype;the case of a full exchange). Because only one recombi- position of the lac+ allele and the remaining drug nation product is recovered, it is impossible to score resistance markers are determinedby observing what gene conversion among recombinants that acquire a aspects of phenotype are lost when the auxotrophies duplication. of the parental Mud insertions are repaired by trans- Non-duplication recombinants (lac+) can arise by duction. sister strand exchange (double recombinationor gene The recombinant types were classified for isogenic conversion). Double recombination or gene conver- Ret+, RecA-, or RecB- derivatives of theparent sion between sister strands generates a Lac+ recom- strain; all have MudA(1ac-) elements in direct orien- binant allele at one site, while the second product of tation at the his and trp loci. In the Rec+ strain, Lac+ the exchange is lost. Since only one recombination recombinants are found at a frequency of 1 x product is recovered, and the other lac region in the (Table 5). About 87% of these clones were non- Half-Tetrad Analysis in Bacteria 35

A B

lntrachromosomal elchange (gene conversion) (double recombination)

FIGURE4.-Recombination events between direct order homologous sequences. Note thatexchanges between sister chromosomes generate products in which only the allele at the joinpoint was involved in the recombination event. For the diagrams of sister strand exchanges, it is assumed that the top sister is recovered as a recombinant. Exchanges between elements in the same chromosome can lead to products in which both alleles were involved in the exchange event, but this can only be identified if information in both copies was altered by the event (double recombination); true gene conversion events can not be scored for exchanges between direct order sequences.

TABLE 5 reflect reversion of the recB mutation.) The major Effect of Rec mutations on recombination between direct effect of the recB mutation was to reduce the fre- repeats of MudA elements at the his and trp loci quency of the “apparent gene conversion” class and the “double recombinant”class. Remarkably, the recB Frequency of various Lac+ products mutation did not significantly affect the frequency of (X 106)“ duplications, confirming earlierobservations (MAHAN Lac+ clones/ Apparent Double Genotype IO6 cells Duplications conversions recombinants and ROTH 1988, 1989). Since detection of a duplication depends on scoring Wild typeb it (TTI7403) 130 (71) 1.8 (I) 114 (62) 14.2 (8) segregation, is important that this event can occur (TT17405) in theparent strain used. Since the recA mutation recA prevents duplication segregation, recombinants aris- (TTl7402) 0.028 (45) 0.0006 (1) 0 (0) 0.027 (44) ing in a recA background were made RecA+ before recB being tested for duplication segregation. The segre- (TT17404) 12 (104) 8.9 (77) 2.4 (2 I) 0.7 (6) gation of duplications is independent of the RecBCD a Numbers in parentheses represent the actual number of Lac+ protein. We have tested this by introducinga clones scored in each class. The two strains listed are constructed to be isogenic with the recB::TnZO insertion into a strain bearing apre-exist- recA and recB strains listed below. Since the two rec+ strains gave ing duplication of the trp-his segment with a essentially identical results, data for thetwo strains were combined. Mud(Lac+) element at the join pointbetween the two copies. We detected recombination between the du- duplication,“apparent gene conversion” types (see plicated material by scoring the formation of white, above) and 1 1% were double recombinants. Onlyone Lac- colonies on Xgal medium. The Mud element duplication was seen among71 independent Lac+ was lost with equal frequency inRecB+ and RecB- clones. In other experiments, up to 10% of the Lac+ backgrounds (between 45 and 60% of the cells in an clones carried duplications (SEGALLand ROTH 1989). overnight culture started from a single colony). Simi- The frequency of Lac+ recombinants in the RecA- lar data have been reported by SCLAFANIand WECHS- background was reduced 104-fold. Most (92%) of the LER (1 98 1) and MAHANand ROTH(1 989). In contrast, recovered Lac+ clones were double recombinants, 6% introducing a recA mutation into a pre-existing dupli- were apparent gene conversions, and one clone seg- cation reduces the frequency of segregants to less than regated Lac- colonies (when made recA+) as expected 2%. for a duplication. (True gene conversion cannot be scored among recombinants between direct repeats.) DISCUSSION The Lac+ clones arising in the recB deletion background grew slowly (4 days) but theoverall frequency We have assayed recombination using a system that of Lac+ clones was reduced only 10-fold. (All Lac+ permits recovery of both products arising from an clones maintained a UV-sensitive phenotype and a individual exchange event. This was made possible lower frequency of transduction, and thus did not through use of strains in which the recombining se- 36 A.and M. Segall J. R. Roth plex DNA. In the case of short repeats, the distance

300 - Short recombining sequences between the scored markers is a large fraction of the (MudJ and MudA) total recombining sequence. Under these conditions 250 - 4 conversion tracts are likely to fall entirely between the 200 - markers and affect neither of the markers.

150- A recB mutation eliminates inversion formation: Long recombining sequences The absence of inversions in a recB mutant strongly / (MudA x MudA) 100- supports amodel in which the full exchange (rejoining

50 both flanking sequences), required to form an inversion, can only be accomplished by the RecBCD path- 0 0 10 20 30 40 50 way (MAHAN and ROTH 1988). Using very short Distance between recombining Mud element (minutes) recombiningsequences, SCHOFIELD,ACBUNAC and FIGURE5.-Effect of distance between recombining inverse-or- MILLER(1 992) have also found that inversion forma- der sequences on the frequency of recombinants. Note that prox- tiondepends heavily on RecBC function. The fre- imity stimulates recombination most for the short recombining quency of the double recombinantclass (not associated sequences. with inversions) is also depressed significantly in the recB mutant,but is noteliminated. We inferthat quences are placed in inverse order in the same chro- noninversion double recombinants can be generated mosome. In this situation, the selected and the unse- both by the RecBCD and also by some alternative lected products of recombination remain in the same recombination pathway (perhaps the recF pathway). chromosome and are available for analysis. As for The frequency of Lac+ recombinants without an half-tetrad analysis, this assay allows recovery of both inversion (apparent conversions) is unaffected by a products of arecombination event. Since we must recB mutation, and themajority of these recombinants select one recombinant phenotype (Lac+), we cannot may occur by a different pathway (perhaps recF). We recover products of exchanges in which both lac se- suspect that many of these apparent conversion re- quences remain inactive. It should be understood that combinant types may arise by sister strand exchange situations thatreduce apparent recombination fre- as diagrammed in Figure 3. This sort of exchange quency could merely bias the nature of the product leads to correction of one lac locus (apparent conver- in favor of undetected events. (This problem is not sion). We cannot score true gene conversions in this peculiar to this recombination system.) situation because the final recombinant (lac+)carries Recombinationproducts were classifiedwith re- only one of the two lac loci known to have been spect to exchange of flanking sequences (inversions) involved in the exchange. and loss of information (gene conversion) associated For direct repeats, a recB mutation reduces total with the crossover event. Gene conversion leads to loss of one of the drug resistance markers placed recombinants but has no effect on duplication fre- within the recombining repeats. quency: The frequency of duplications, in contrast to Below is a summary of our observations: the frequency of inversions, is unaffected by a recB When long repeats (39 kb)recombine, full ex- mutation. We interpret this as evidence that duplica- changes frequently show concomitant conversion: tions, which can form by a half exchange (one pair of All of the expected products arise from exchanges flanks rejoined), can form by an alternative pathway between long homologies. Results generally resemble that does not involve the RecBCD function. Since this those obtained in fungi. Recombinant (Lac+) clones alternative can not accomplish inversions, we interpret arise with and without exchangeof flanking sequences this result as indicating thatthe major alternative and exchanges of flanks are frequently associated with recombination pathway is unable to perform a full gene conversion. Inthe caseof longrepeats, the exchange (rejoining both flanks). majority of initiatingevents may occur (in Muse- Double recombinantsbetween direct repeats are quences) far from the eventual site of resolution (in assumed to reflect exchanges between two sequences lac sequences). Since the heteroduplex region must in the same chromosome since information is moved extend from the initiation site to the resolution site, reciprocally between two lac sites; these products are the intervening marker allele has a high likelihood of reduced by the recB mutation. The “apparent gene being converted. conversion” class of recombinants between direct re- When short repeats recombine,conversion is rare: peats can arise from several different events (see Fig- Gene conversion occurs rarely in exchanges between ure 4). Mud elements that share only 5 kb of homologous Effect of distance betweensites on recombination sequence. Most inversions occur withoutan associated frequency: Inconsidering this andthe nextpoint conversion event. The small number of gene conver- (below), it is important to note that thereis an uncon- sion events may reflect shorter regions of heterodu- trolled variable in these experiments. The chromo- Half-Tetrad Analysis in Bacteria 37 soma1 intervals tested with short homologies are dif- interval (data in Table 3). In both situations the ele- ferent from the onestested with long homologies. ments are separated by about 17% of the chromo- Therefore it is possible that differences between short some. This differencebetween a permissive and a and long homologies could reflect idiosyncrasies of nonpermissive interval is not seen when long homol- the particular sites used in each case. ogous sequences recombine (MudA-MudA). It is pos- Frequency of recombination between shortinverse- sible that longer homologous sequences provide suf- order repeatsappears to be more sensitive to the ficient initiatingsubstrates within the recombining separation between the repeats thanis recombination sequences to support frequent gene conversion. Ini- between long repeats (Figure5; data presentedis from tiation of gene conversion between short sequences Tables 2 and 3). We have varied the distance between may have to rely on nearby sites in the chromosome the repeated sequences from 2 to 44% of the chro- (outside the recombining sequences). For nonpermis- mosome. Proximity ofthe recombining sequences sive intervals, we propose that regions adjacent tothe seems to stimulate recombination between short se- recombining sequences are deficient in providing ini- quences (MudA-MudJ) more than that between long tiating structures; this results in a much lower fre- sequences (MudA-MudA). This suggests that the long quency of recombination between sequences at these sequences provide sufficient initiating substrate (DNA sites. strand ends) to saturate the recombination process Summary: We describe a recombinationsystem that regardless of the chromosomal distance separatingthe is useful for studying the nature of intrachromosomal interacting sequences; for short homologies, the num- exchange events and for testing the effect of known ber of initiating ends may be limiting and therefore recombinationfunctions onthe various product recombination is stimulated by bringingthe rarer classes. We consider two of our observations particu- structures closer together. larly significant: (1) Inversion formation dependscom- Recombination is very frequent between the most pletely on the recBCD function while duplications oc- closely placed Mud elements, which are separated cur in the absence of the recBCD function. (2) Gene by only 2% of the chromosome (am-thr, 3.2 X lo-’; conversion is often seen among recombinants with a thr-pyri3, 3.4 X Despite the highfrequency of full exchange between long sequences; conversion is recombination events, these intervals show a distri- rarely associated with recombination between short bution of productssimilar to that seen for morewidely sequences. separated Mud elements;thus, recombination does Overall, these results suggest that exchanges initi- not appear to bequalitatively affected by the distance ated by different DNA substrates may lead to distinct separating the interactingsequences. The high recom- outcomes. The frequency with which various initiat- bination frequency may reflect a nearby hotspot for ing substrates (nicks, breaks) and various critical sec- initiation of recombination, or an underlying feature ondary sequences (chi sites) arise in particular se- of chromosome structure thatfavors thejuxtaposition quences may determine the distribution of products of homologies separated by short stretches of DNA. seen in each situation. It should be possible to use the Effect of sequence length on recombination fre- system described here to learn how the several bacte- quency: For all butthe two shortestintervals, the rial recombination functions act on various chromo- length of total homology shared between the two somal substrates and how initiating substratesare gen- repeats has a minimal effect on recombination fre- erated. quency. Repeated elements that share either 5 or 40 This work was supported by National Institutes of Health grant kb of identical sequence yield lac+ recombinants at GM 27068 to J.R.R. We would like to thank MICHAELLICHTEN for about the same frequency. Excluding the am-thr and critical reading of the manuscript. pyrB-thr intervals, the average frequency of recombination between short repeats (MudA-MudJ) is 9.8 X LITERATURE CITED 10-5 asopposed to 8.2 x betweenlong repeats ANDERSON,P., and J. ROTH, 1981 Spontaneous tandem genetic (MudA-MudA). This suggests that 5 kb of homology duplications in Salmonella typhimurium arise by unequal re- is longer than the minimal length required for effi- combination between rRNA (rrn) cistrons. Proc. Natl. Acad. cientpairing. As mentionedabove, the two short Sci. USA78 3113-3117. intervals recombine more frequently than expected ANDERSON,R. P., and J. R. ROTH, 1977 Tandem genetic dupli- 5). cations in phage and fungi. Ann. Rev. Microbiol. 31: 473-505. (Figure BENDER,J., and N. KLECKNER,1992 IS10 transposase mutations A more striking exception to the general observa- that specifically alter target site recognition. EMBO J. 11: 741- tion is the large reduction in the recombination fre- 750. quency seen for short sequences(MudA-MudJ) flank- BERKOWITZ,D., J. HUSHON,H. WHITFIELD,J. R. ROTH and B. N. ing an interval that is nonpermissive for inversion. AMES, 1968 Procedure for identifying nonsense mutations.J. Bacteriol. 96 215-220. The nonpermissive trpA-hisD interval showed a 1000- CASADABAN,M. J., and S. N. COHEN,1979 Lactose genes fused fold lower recombinant frequency than thearaC-nadA to exogenous promoters in one step using a Mu-lac bacterio- 38 A. M. Segall and J. R. Roth

phage: in vivoprobe for transcriptional control sequences. Proc. cereuisiae ARG4 initiator of meiotic gene conversion and its Natl. Acad. Sci. USA 76 4530-4533. associated double-strand DNA breaks can be initiated by tran- CASTILHO,B. A., P. OLFSONand M. J. CASADABAN,1984 Plasmid scriptional interference. Proc. Natl. Acad. Sci. USA 89 12068- insertion mutagenesis and lac gene fusion with mini-Mu bac- 12072. teriophage transposons. J. Bacteriol. 158 488-495. SANDERSON,K. E., and J. R. ROTH, 1983 Linkage map of Salmo- CHAN, R. K.,D. BOTSTEIN, T. WATANABEand Y. OGATA, nella typhimurium. Edition VI. Microbiol. Rev. 47: 410-453. 1972 Specialized transduction of tetracycline by phage P22 SARTHY,V., and M. MESELSON, 1976 Single burst study of rec- in Salmonella typhimurium. 11. Properties of a high frequency and red-mediated recombination in bacteriophage lambda. transducing lysate. Virology 50 883-898. Proc. Natl. Acad. Sci. USA 73: 46 13-46 17. CLARK,A. J., 1973 Recombination deficient mutants of E. coli SCHMID,M. B., and J. R. ROTH,1983 Genetic methods for analysis and other bacteria. Annu. Rev. Genet. 7: 67-86. and manipulation of inversion mutations in bacteria. Genetics DAVIS,R. W., D. BOTSTEIN and J. R. Roth, 1980 AdvancedBac- 105: 517-537. terialGenetics. Cold Spring Harbor Laboratory, Cold Spring SCHOFIELD,M. A.,R. AGBUNAGand J. H. MILLER, 1992 DNA Harbor, N.Y. inversions between short inverted repeats in Escherichia coli. DUL,J. L., and H. DREXLER,1988 Transcription stimulates re- Genetics 132: 295-302. combination. I. Specialized transduction of Escherichia coli by SCLAFANI,R. A., and J. A. WECHSLER,1981 High frequency of htrp phages. Virology 162: 466-470. genetic duplications in the dnaB region of the Escherichia coli DUL,J. L., and H. DREXLER,1988 Transcription stimulates re- K12 chromosome. Genetics 98: 677-689. combination. 11. Generalized transduction of Escherichia coli by SEGALL,A. M., and J. R.ROTH, 1989 Recombination between phages T1 and T4. Virology 162: 471-477. homologies in direct and inverse orientation in the chromo- ELLIOTT, T., and J. R. ROTH, 1988 Characterization of some of Salmonella: intervals which are nonpermissive for TnlOdCm: A transposition-defective TnlO specifying chlor- inversion formation. Genetics 122: 737-747. amphenicol resistance. Mol. Gen. Genet. 213: 332-338. SEGALL,A. M., M. J. MAHAN and J. R. ROTH, 1988 FOSTER,T. J., 1977 Insertion of the tetracycline resistance trans- Rearrangement of the bacterial chromosome: forbidden inver- location unit TnlO in the lac operon of Escherichia coli K12. sions. Science 241: 13 14-1 3 18. Mol. Gen. Genet. 154 305-309. SONTI,R. V., and J. R. ROTH, 1989 Role of gene duplications in HERMAN,R. K., 1968 Effect of gene induction on frequency of the adaptation of Salmonella typhimurium to growth on limiting intragenic recombination of chromosome and F-merogenote carbon sources. Genetics 123: 19-28. in Escherichia coli K-12. Genetics 58: 55-67. STAHL,F. W., 1979 GeneticRecombination: Thinking About It In HILL,C. W., and B. W. HARNISH,1981 Inversions between ribo- Phage and Fungr. W. H. Freeman, San Francisco. somalRNA genes of Escherichia coli. Proc. Natl. Acad. Sci. STEWART,S. E., and G. S. ROEDER,1989 Transcription by RNA USA 78 7069-7072. polymerase I stimulates mitotic recombination in Saccharomyces HUGHES,K. T., andJ. R. ROTH,1984 Conditionally transposition- cereuisiae. Mol. Cell. Biol. 9 3464-3472. defective derivative of Mu dl(Amp Lac). J. Bacteriol. 159 TAKAHASHI,N.K., K. YAMAMOTO,Y. KITAMURA, S. Q. Luo, H. 130-137. YOSHIKURAand I. KOBAYASHI,1992 Nonconservative recom- HUGHES,K. T., and J. R. ROTH, 1985 Directed formation of bination in Escherichia coli. Proc. Natl. Acad. Sci.USA 89: deletions and duplications using Mu d(Ap, lac). Genetics 109 5912-5916. 263-282. THALER,D. S., and F. W. STAHL,1988 DNA double-chain breaks KING, S. R., and J. P. RICHARDSON,1986 Role of homology and in recombination of phage lambda and of yeast. Annu. Rev. pathway specificity for recombination between plasmids and Genet. 22: 169-197. bacteriophage lambda. Mol. Gen. Genet. 204: 141-147. THOMAS,B. J., and R. ROTHSTEIN,1989 Elevated recombination KONRAD,E. B., 1969 The genetics of chromosomal duplications. rates in transcriptionally active DNA. Cell 56 619-630. Ph.D. Thesis, Harvard University, Cambridge, Mass. TLSTY,D. T., A. M. ALBERTINIand J. H. MILLER,1984 Gene KONRAD,E. B., 1977 Method for the isolation of Escherichia coli amplification in the lac region of E. coli. Cell 37: 2 17-224. mutants with enhanced recombination between chromosomal VOGEL,H., andD. BONNER,1956 Acetylornithinase of Escherichia duplications. J. Bacteriol. 130 167-172. coli: partial purification and some properties. J. Biol. Chem. MAHAN,M. J., and J. R.ROTH, 1988 Reciprocality of recombi- 218: 97-106. nation events that rearrange the chromosome. Genetics 120 ZIEG,J., and S. R. KUSHNER,1977 Analysisof genetic recombi- 23-35. nation between two partially deleted lactose operons of Esche- MAHAN,M. J., and J. R. ROTH, 1989 Role of recBC function in richia coli K12. J. Bacteriol. 131: 123-132. formation of chromosome rearrangements: a two-step model ZIPSER, D., S. ZABELL,J. ROTHMAN,T. GRODZICKERand M. WENK, for recombination. Genetics 121: 433-443. 1970 Fine structure of the gradient of polarity in the z gene MILKMAN,R., and M. M. BRIDGES,1993 Molecular evolution of of the lac operon of Escherichia coli. J. Mol. Biol. 49 251-254. the Escherichia coli chromosome. IV. Sequence comparisons. Communicating editor: E. W. JONES Genetics 133: 455-468. OCHMAN,H., and R. K. SELANDER,1984 Evidence for clonal population structure in Escherichia coli. Proc. Natl. Acad. Sci. USA 81: 198-201. APPENDIX PETES, T. D.,R. E. MALONE and L. S. SYMINGTON,1991 Recombination inyeast, pp. 407-521 in TheMolecular and Recombinational exchanges of flanking sequences Cellular Biology of the Yeast Saccharomyces, edited by J. R. and gene conversion (gain and loss of information in BROACH,J. R. PRINGLEand E. W. JONES. Cold Spring Harbor an exchange) can be observed without regard to the Laboratory, Cold Spring Harbor, N.Y. exact mechanism of the recombinational process. We REBOLLO,J.-E., V. FRANCOISand J.". LOUARN,1988 Detection have described these events with the assumption that and possible role of two large nondivisible zones on the Esche- inversions occur by a single reciprocal exchange be- richia coli chromosome. Proc. Natl. Acad. Sci. USA 85: 9391- 9395. tween two sequences within one circular chromosome Rocco, V., B. DE MASSY and A. NICOLAS,1992 TheSaccharomyces as diagrammed at the left side of Figure 1 above. We Half-Tetrad Analysis id Bacteria 39 which multiple events are morelikely. We have tested thesecounter-arguments by constructing cells with two independent duplications and observing the seg- a a regationfrequency of eachduplication (L. MIESEL and J. R. ROTH, unpublished results); the two dupli- I! I! cations segregate independentlyand show no evidence b b of recombinationally “hot” cells. We think it highly C C unlikely that inversions form by two coincident, un- d d related sister strand exchanges. More likely are scen- IL 1L arios in which the “hot” end generated by the first half exchange is involved in a cascade of events that e e ultimately results in inversion. Half-reciprocal exchanges followed bydouble- strand break repair: Formation of an inversion can FIGURE6.--Inversion formation by coinciding half-exchanges. be accomplished without a reciprocal exchange. The The sister strands of a replicating chromosome are presented. If process could be initiated by a half-exchange between the inverse-order repeats line up properly, two independent half- inverse-order sequences in the same chromosome (see exchanges can generate an inversion. Figure 7). This leads to aninversion with an associated double-strand break that must be repaired. If the two ends are restricted to the region of shared homology think it likely that the events described here occurin (top right), the break can be repaired using any copy this way. However other mechanisms can also gener- of the repeated sequence as template.If one such end ate aninversion. Below we present two possibilities. is degraded beyond the region of inverse homology, Coincidental half-exchanges betweensister chro- the gap can be repaired only by using the indicated mosomes: This mechanism (Figure 6) for inversion copyof the repeat in the sister chromosome. This formation has been discussed previously (SEGALLand mechanism involves two copies of the inverse-order ROTH 1989). By two half-exchanges with a sister chro- sequence in one chromosome and a copy of the re- mosome, an inversion is made in one sister chromo- peated sequence in a sister chromosome. If inversions some and the other sister chromosome is destroyed. occur by a process of this sort, they could arise without A single half-exchange of this type can occur between a standard reciprocal exchange, and the interpreta- direct repeats to generate a duplication. Monitoredin tion of gene conversion is made more complicated. this way, the frequency of each event is about The problem ofhow apparently reciprocal ex- We have arguedthat the two coincidentalevents changes are formed is not unique to the recombina- needed to generate an inversion would be expected tion system described here; even in tetrads for which at a frequency of lo-’ and could not be detected in all ultimaterecombination products are recovered, our system. Counter-argumentsare: (a) The first the observed reciprocality could result from a half- event mightcause a general increasein recombination exchange followed by repair of the produced double rate, making the second more likely and (b) All such strand breaks, perhaps using a sister chromosome as events might occur in recombinationally “hot”cells in template.

bhReparable at A INV c any L fr-4b FIGURE7.-Inversion formation by ahalf-ex- / change followed by double strand break repair. An initial half-exchange can generate an inversion and an associated double strand break; the ends may be Reparable at b highly recombinogenic and stimulate the subsequent events needed to repair the break.This general proc- of sister ess can also be initiated by a half exchange between d sisters. e -1 Reparable at e b of sister