Site-specific recombination in the replication

The EMBO Journal vol.14 no.7 pp.1561-1570, 1995

Site-specific recombination in the replicationterminus region of Escherichia coli: functionalreplacement of dif

Nicholas R.Leslie and David J.Sherratt1 2

Department of Genetics, University of Glasgow. Church Street,Glasgow GIl 5JS, UK'Present address: Microbiology Unit, Department of Biochemistry,University of Oxford, South Parks Road, Oxford OXI 3QU, UK

2Corresponding author

Communicated by J.Gross

The replication terminus region of the Escherichia colichromosome encodes a locus, dif, that is required fornormal chromosome segregation at cell division. dif isa substrate for site-specific recombination catalysed bythe related chromosomally encoded recombinases XerCand XerD. It has been proposed that this recombinationconverts chromosome multimers formed by homo-logous recombination back to monomers in order thatthey can be segregated prior to cell division. Strainsmutant in dif, xerC or xerD share a characteristicphenotype, containing a variable fraction of fila-mentous cells with aberrantly positioned and sizednucleoids. We show that the only DNA sequencesrequired for wild-type dif function in the terminusregion of the chromosome are contained within 33 bpknown to bind XerC and XerD and that putative activesite residues of the Xer recombinases are required fornormal chromosome segregation. We have also shownthat recombination by the loxPICre system of bacterio-phage P1 will suppress the phenotype of a dif deletionstrain when loxP is inserted in the terminus region.Suppression of the dif deletion phenotype did not occurwhen either dif/Xer or loxP/Cre recombination actedat other positions in the chromosome close to oriC orwithin lacZ, indicating that site-specific recombinationmust occur within the replication terminus region inorder to allow normal chromosome segregation.Key words: chromosome segregation/dif/site-specificrecombination

IntroductionLinear chromosomes, for example those of eukaryotes,have evolved mechanisms to ensure that chromosomeends are protected from exonucleolytic attack and can

be replicated completely. Circular chromosomes are notsubject to these problems. However, their circularity makeshomologous recombination a threat to their integrity andmaintenance. Any number of homologous recombinationevents between linear homologous chromosomes generateslinear products that have the same overall structure as

their parents. In contrast, any odd number of homologousexchanges between circular chromosomes generates a

fusion of the two circles. Such dimeric molecules might

K Oxford University Press

create difficulties in segregation at cell division, or inpackaging when the circles are viral. Clearly, a unit copyreplicon in dimeric form cannot normally be partitionedinto two daughter cells. Multimerization also interfereswith the stable inheritance of high copy number plasmids(Summers and Sherratt, 1984; Summers et al., 1993). Itis therefore not surprising to find that circular genomeshave evolved mechanisms to ensure that multimers canbe effectively converted to monomers. In Escherichia coliand related bacteria, we believe that both plasmids andthe bacterial chromosome use site-specific recombinationto convert multimers to monomers (Austin et al., 1981;Sherratt et al., 1993).

Both the E.coli chromosome and many high copynumber plasmids contain recombination sites acted uponby the two related recombinases XerC and XerD (Collomset al., 1990; Blakely et al., 1993). These lambda integrasefamily recombinases are encoded by unlinked genes on theEcoli chromosome, xerC and xerD. Their chromosomalsubstrate, dif, is located in the replication terminus regionat min 33.6 of the genetic map, kilobase 1608 of thephysical map, between the innermost terminator sites TerAand TerC (Kuempel et al., 1989, 1991). Strains carryingmutations of xerC or dif display similar morphologicalphenotypes under microscopic analysis, as a significantportion of cells do not undergo normal cell division(Blakely et al., 1991; Kuempel et al., 1991). These arebelieved to be cells in which chromosome multimers haveformed by recombination between sister chromatids; theyhave an aberrant distribution of nucleoids and are fila-mentous. Although the SOS response can become inducedin these strains, it appears not to be the primary cause offilamentation (Kuempel et al., 1991).The dif sequences required in cis for suppression of the

dif deletion phenotype are contained within a 532 bpClal-Sau3A fragment from the terminus (Kuempel et al.,1991). This fragment contains a 33 bp region that canbind XerC and XerD in vitro and which, when present ina plasmid, acts as a substrate in both resolution andmultimerization of plasmids (Blakely et al., 1991, 1993).Here we further investigate the extent of dif sequencesrequired for wild-type function in situ, the effect ofchromosomal location on dif function and whether otherrecombination systems can fulfil the role of dif in normalchromosome segregation.

ResultsChromosomal flanking sequences have nodetectable effect on in vivo recombination at dif ina plasmid substrateA 33 bp DNA fragment containing the 28 bp dif core sitehas been shown to be a substrate for Xer-mediatedmultimerization and resolution in vivo, when cloned into

1561

N.R.Leslie and D.J.Sherratt

high copy number plasmids (Blakely et al., 1991). ThisDNA fragment binds XerC and XerD in vitro and cansupport a partial recombination reaction in vitro (Blakelyet al., 1993; Sherratt et al., 1995). In other site-specificrecombination systems, DNA sequences flanking such acore site can play a role in determining the efficiency andselectivity of recombination (see, for example, Landy,1989; Summers, 1989; Stark et al., 1992). We wished toinvestigate whether sequences adjacent to the dif core sitehad such roles. There is good evidence that a 532 bpchromosomal fragment containing the dif core site carriesall the sequences required for the wild-type functioningof dif in its normal chromosomal location (Kuempel et al.,1991). Therefore, we used an in vivo recombinationassay to compare the Xer-dependent multimerization andresolution of plasmids with and without these extrasequences flanking the core difsite. Each plasmid consistedof a dif fragment inserted into the polylinker of pUC 18(Figure 1A). Plasmid pUC18 alone and a similar plasmidcontaining a full-length cer site were used as controls.Either recF or recA strains deficient in plasmid homologousrecombination were transformed with plasmid monomersor dimers purified from agarose gels. Plasmid DNA fromtransformed strains was prepared after different numbersof generations, from both individual transformants andfrom pooled collections of transformants. Ethidiumbromide staining of electrophoresed samples did not showany distinguishable difference between the fraction ofplasmid DNA present as monomers or multimers for thethree dif plasmids tested (Figure 1B). Thus the extrasequences flanking the core site have no detectable effecton Xer-mediated recombination in vivo on plasmidsubstrates.

AChromosomal dif fragments clioned into pUC18

CIaA Styl Sau3A

pN532bp__ ___\_532bpdif \ V\

pN30247 bp dif

pMIN3333bpdif

B

teWtrainwersc trifler -

rUi..( (-l iIn"',l c,rreldL-xed chiller -

SC diTiler 4

SC i1YoTVnomr -

Z~r" -II ,l C,

Deletion of the chromosomal dif siteOne of our primary aims was to determine what sequencesneed to be present in the chromosome terminus region toconfer dif function as determined by normal cell divisionand nucleoid localization. We also wished to investigatethe ability of dif to function in different chromosomallocations and the ability of other site-specific recombina-tion systems to substitute for dif and its recombinases. Aninitial prerequisite was to delete the chromosomal dif site.

In E.coli the dif site is located in the replication terminusregion at min 33.6 of the genetic map, kilobase 1608 ofthe physical map (Kuempel et al., 1991). Plasmid pBS 12,containing 28.5 kbp of the terminus region (Bejar andBouche, 1983) was used as a source of DNA. A range ofbi-directional plasmid deletions were produced, extendingoutwards from the two Styl sites adjacent to the dif coresite, and marked by a kanamycin resistance gene (Figure2). DNA from these plasmids was then used to replacethe normal dif chromosomal locus using homologousrecombination (Jasin and Schimmel, 1984; Winans et al.,1985). Due to the filamentous nature of the recBC sbcBCstrains used in this method, the marked deleted lociwere then moved to other strains by bacteriophage P1transduction (Miller, 1972). Replacement of the normalDNA between the two StyI sites by the KmR cassetteyielded two strains retaining dif core sites, designatedNLAIN and NLAOUT, differing only in the orientation ofinsertion of the kanamycin resistance gene (Figure 2). Thestrain (NL40) containing the largest deletion of -1500 bp

Fig. 1. Chromosomal flanking sequences have no detectable effect onin vivo recombination at dif in a plasmid substrate. (A) The three dif-containing fragments of sizes 532, 247 and 33 bp cloned into thepolylinker of pUC18 are shown. pKS492 is of similar construction butcontains a full cer site (Summers and Sherratt, 1984). Boxes labelledC or D correspond to the nucleotide sequences bound by XerC andXerD respectively, with the central region between. (B) An example ofthe in vivo recombination assays performed. A recF strain (DS941)was transformed with purified monomers of each of the five plasmids,pUC18, pMIN33, pN30, pN532 and pKS492. DNA was prepared after40 h of growth and analysed by 1.0% agarose gel electrophoresis.Different plasmid forms on the gel, including higher multimeric forms,are labelled, supercoiled being abbreviated to sc.

was used in all further experiments. All strains in whichthe dif core site was absent were found to show the samephenotype previously described for dif or xerC mutants(Figure 3; Blakely et al., 1991; Kuempel et al., 1991).Strains with mutations in xerD (STLl 16 and DS9008,Blakely et al., 1993) also showed this phenotype (Figure3). Direct phenotypic comparisons of these three typesof mutant by microscopy and analysis of cell lengthdistributions and of growth and viable cell counts in batchculture have failed to demonstrate any difference betweenthem (Figure 3, Table I and data not shown). The phenotypecaused by these mutations is the same in recF orrecombination-proficient strains (data not shown). Thefact that mutations in xerC, xerD or dif all give the samecharacteristic phenotype is consistent with the view thatthey all function in the same biological process.

1562

Site-specific recombination at dif

Wild typeH St St SaI I L I

I LC dif

core-site

MutantsInsertion of Kmr gene

A) Without exonuclease digestionNIUIN

H

E

1 bl

Kmr gene Sa._________________ ~~~~~~~~I

H

ESt;B Transcripton-I

NIAOUTHI

sa

H

H

ESIB Transcription B

B) With exonuclease digestionNL4O difmutant

Transcription --

H

E

Fig. 2. Mutant strain constructions. These are diagrammaticrepresentations of the chromosomal sequences around the dif locus inthe wild-type and the mutant strains described, showing restrictionsites and other features. The core site represented by a crosscorresponds to the 28 bp XerC and XerD binding region known to bea sufficient substrate for plasmid recombination. The hatched boxesrepresent the entire kanamycin resistance cassette, includingpromoters. Restriction enzyme sites are labelled: E, EcoRI; H, HindIll;C, ClaI; St, Stvl; Sa, Sau3A; B, BamHI. Only 7 bp of DNA separatethe closest Styl site from the XerC binding site of dif. Deletions wereconstructed in plasmids before crossing into the chromosome. A5.5 kbp BamHI restriction fragment from pBS12 with the dif sitecentrally located was cloned into the BamHI site in pUC18.Restriction of the resultant plasmid (pNI) with Stv' removed 367 bpadjacent to the dif core site. Deletion was then extended by digestionwith exonuclease III for times ranging from 0 to 6 min. Ligation to akanamycin resistance gene (Pharmacia) produced a series of plasmidswith deletions ranging in size from 367 bp (with the dif core siteintact) to -1500 bp, all marked by a kanamycin resistance gene. Theseconstructs were then used to replace chromosomal dif loci. The NL40dif construct shown corresponds to the largest deletion produced.

A 33 bp DNA fragment containing the dif core sitecan suppress the dif deletion phenotype wheninserted into the replication terminus regionKuempel and co-workers (1991) showed that a 532 bpfragment containing dif can phenotypically suppress a12 kbp dif deletion when introduced into the terminusregion of the chromosome by plasmid integration. Wewanted to determine whether smaller sequences wouldfully suppress the phenotype, particularly the 33 bpsequence that contains a dif site functional in plasmidrecombination assays in vivo. We also wished to usemethods of integration that did not involve the introductionof plasmid sequences into the chromosome. Therefore weintroduced 532 bp and 33 bp dif+ fragments (Figure 1A)into strain NL40 bearing the 1500 bp deletion of the difregion. These fragments were re-introduced by homo-logous recombination into their natural chromosomal loca-tion in the replication terminus, in both orientations relativeto the terminus and in a number of contexts relative tothe antibiotic resistance markers (Figure 4). These novelterminus region constructions were introduced into DS941

(dif ) and the dif-deleted strain NL40. Details are shownin Figure 4. Strain construction was verified by Southernhybridization and the recombinational activity of theintroduced sites and chromosomal recombinases was testedby plasmid integration assays (described below).

These strains were analysed microscopically to deter-mine whether they displayed the dif phenotype. All strainscontaining the 532 bp dif fragment were found to bemorphologically wild-type. Strains NL246, NL286 andNL296 contain the 28 bp dif core site within the 33 bpfragment. NL286 and NL296 were indistinguishable fromwild-type cultures when tested (compare Figures 3 and5). Thus the 33 bp fragment appears to contain all of thedif sequences required for suppression of the deletionphenotype in either chromosomal orientation. However,NL246 displays a dif phenotype similar to that of dif orxer mutants (Figure 5), despite containing exactly thesame sequences as NL286, although arranged differently.A similar phenomenon was evident in phenotypic analysisof NLAIN and NLAOUT (Figure 2). Although NLAOUTwas phenotypically wild-type, exponential phase culturesof NLAIN consistently display a phenotype intermediatebetween that of wild-type and a dif deletion mutant.Transformation of these strains with a high copy numberplasmid expressing both XerC and XerD (pMAY5, G.Mayand D.Sherratt, unpublished data) had no effect uponthe phenotype observed, with or without induction ofexpression. Possible explanations for the filamentation ofthese strains are discussed below.

The effect of chromosomal position on dif functionIs the positioning of dif in the replication terminus closeto TerC coincidental or is this positioning required for itsfunction? To investigate this we wished to determinewhether a dif site is able to function if it is positionedelsewhere in the chromosome. Therefore we constructedstrains in which dif had been deleted from its normalposition and an ectopic dif site had been positioned eithernear oriC, the chromosomal origin of replication, or atmin 8, near the mid-point between origin and terminus(Figure 6). Close to the origin, dif was inserted at min 84in a deletion of pstC, pstA, pstB and phoU, genes thatcontrol phosphate metabolism. At min 8, dif was insertedinto a deletion of lacZ, encoding P-galactosidase. Thesepositions were chosen as both are well characterizedand display easily detectable mutant phenotypes withoutaffecting cell morphology. pSN518 (Amemura et al.,1982) and p357 (M.Burke and D.Sherratt, unpublisheddata) were used as sources of DNA from the pst and lacZloci respectively. Using the same strategy as in earlierconstructions, a gentamicin resistance gene, with andwithout a 532 bp dif fragment, was introduced into thesechromosomal locations. Strain construction was verifiedby tests for expression of alkaline phosphatase (AP) andP-galactosidase and Southern hybridization. Strains mutantfor pstCAB and phoU were shown to constitutively expressAP, as previously documented (Wanner and Latterell,1980). Ectopic chromosomal dif sites were shown to beactive for recombination by plasmid integration assays(see below).

Strains with a single dif site positioned in either thepst genes (NL350) or lacZ (NL250) were found to befilamentous and indistinguishable from a xer or dif mutant

1563


A

.. -

^,.

foW

-

C D

Fig. 3. Cell and nucleoid morphology of dif and xer mutants. Exponential phase cultures growing in LB broth at 37°C were prepared forphotography by condensation and staining of cell nucleoids with chloramphenicol and DAPI respectively (Materials and nmethods). (A) Wild-type(DS941). (B) Adif (NL40). (C) xerC (DS984). (D) xerD (DS9008). The long axis of each photograph represents -70 tm. Condensation of thenucleoids does not interfere significantly with the phenotype demonstrated.

(Figures 3 and 5). Strains retaining a wild-type dif site inthe terminus in addition to an ectopic copy appeared wild-type. Insertion of a gentamicin resistance fragment withoutdif in either location had no detectable effect on cellmorphology.

Can other recombination systems suppress the difphenotype?It is our hypothesis that dif exists to resolve chromosomedimers to monomers by site-specific recombination, priorto cell division. If this is true, it might be expected thatother recombination systems with similar properties couldplay the same role. We replaced dif with other recombina-tion systems to see whether these would suppress themutant phenotype. This was done by constructing strainswith other recombination sites in the terminus in place ofdif and expressing any required recombinase enzymesfrom plasmids. The sites used were res of transposon Tn3(Stark et al., 1989), cer of ColEl (Summers and Sherratt,1984), a variant, cer6 (the type II hybrid of Summers,1989) and loxP of bacteriophage P1 (Hoess and Abremski,1990). Recombination at res and cer is exclusively intra-molecular, while recombination at loxP, like that at dif,occurs inter- and intramolecularly. Recombination at cer6occurs both inter- and intramolecularly when the Xerrecombination accessory proteins PepA and/or ArgR areabsent, but when they are both present recombination ispreferentially intramolecular (Summers, 1989). Thereforestrains with this site in the terminus region were made in

Table I. Analysis of cell length distribution

Strain Number of cell Mean cell length + SElengths counted (gm)

DS941 (wt) 167 2.71 + 0.07DS984 (xerC) 233 3.96 + 0.39NL40 (Adif) 167 3.74 + 0.24

The difference in cell length distribution between wild-type andmutant exponential phase cultures was analysed. These data wereobtained by measuring 167 cell lengths each from untreated cultures ofDS941 (wild-type) and NL40 (dij) and 233 from DS984 (xerC). Themean (x), and strandard error of the mean (Xan - I IN ) werecalculated. These findings agree well with previous work on individualmutants (Blakely et al., 1991; Kuempel et al., 1991) and show a verysignificant difference between wild-type and both mutant cultures, butno significant difference between the xerC and the dif mutant culturesthemselves.

argR, pepA and argR+ pepA+ backgrounds (NL279,NL259 and NL289 respectively).

In each case the recombination site was introduced,together with a gentamicin resistance marker, into thekanamycin resistance gene used in the deletion of thechromosomal dif locus (Figure 2). Sites were clonedupstream of the gentamicin resistance gene, in a contextin which a minimal difsite had been shown to be functionalfor suppression of the dif phenotype. In contrast to ourother experiments, loxP site constructions were introducedinto the chromosome by the method of Hamilton et al.

1564


NL 4() difinutart

II Kiil- gene i

E

Recomiibilatioin site anild celentaIIciiresiStanct1 ceeInsertions

an (At traLiinLoeal organisation

3 sbp Wr.t V

hp,, ir;Aincnt > = ==..4..V1

Ni >4 Vr F.~rlli<i

NI._24 __

: r'l"ri,.Jt-r tI1' T1 5 - - =

Fig. 4. Effects of re-introducing dif into the replication terminus.These are diagrammatic representations of the chromosomal difregions of various constructed strains. Various dif fragments wereintroduced alongside a gentamicin resistance gene (abreviated to Gmr)into the terminus of a strain deleted of dif (NL4O). Construction detailsand the phenotype of resultant strains are shown (phenotypes of strainsare abreviated: WT, morphologically wild-type; F, filamentous). Straingenotypes are described in Table IV. Long thin arrows indicate thedirection of transcription of resistance genes. Short fat arrowsrepresent the dif fragments inserted, the XerC binding site beingclosest to the blunt end. The dif sites used were cloned in thepolylinker of pMTL23 next to a gentamicin resistance gene, makingvarious site/marker cassettes. These were then ligated into the codingsequence of the kanamycin resistance gene in pNA6, a plasmidcontaining DNA from the dif region with a 1500 bp dif deletion(described earlier and used in the construction of NL40). Thisproduced a range of plasmids containing dif fragments and gentamicinresistance markers, with flanking homology to the replication terminuson both sides. These constructs were then crossed into thechromosome by homologous recombination.

(1989), to avoid the use of bacteriophage P1 transduction.Tn3 resolvase and Cre recombinases were expressed fromthe plasmids pPAK316 (Kitts et al., 1983) and pRH200(Mack et al., 1992) respectively. cer and its cer6 derivativewere acted upon by chromosomally encoded enzymes.The activity of recombinases in these strains was verifiedby observing in vivo recombination between directlyrepeated copies of appropriate sites on reporter plasmids(pMA2 1 testing for resolvase, Bednarz et al., 1990;pKS455 testing for XerC and XerD, Stirling et al., 1988a).Chromosomal cer6 and loxP sites were shown to befunctional for recombination by plasmid integration assays(see below).

Strains in which di)'had been replaced by cer, cer6 or theTn3 res/resolvase systems demonstrated the filamentousphenotype (data not shown). A strain with a loxP site inplace of di)' (NL208) is filamentous without recombinase.However, transformation of this strain with the Cre expres-sion vector pRH200 suppresses this phenotype (Figure 5).Neither pUC18 nor pSDC105 (a XerC expression vector,

Colloms et al., 1990) cause this suppression. pRH200 hasno effect on the phenotype of a standard difmutant (NL40).

If the loxP/Cre recombination system can suppress thedif phenotype when introduced into the terminus, wouldit be able to effect this suppression if introduced intoother chromosomal locations? To approach this question,we introduced a loxP site into the chromosome either atlacZ or close to the origin of replication, as we previouslyhad with dif (above and Figure 6), producing two strains,NL202 and NL203 respectively. These strains were exactlyanalogous to NL250 and NL350 described above, butwith loxP not dif inserted ectopically. Unlike NL208,the filamentous phenotype of NL202 and NL203 is notsuppressed by transformation of the strains with the Creexpression vector pRH200 (data not shown).

Recombination sites introduced into thechromosome are active for recombinationIt was important to show that sites in the chromosomewere active for site-specific recombination. The plasmidintegration method of Kuempel et al. (1991) was used(Materials and methods). This method uses a chloram-phenicol resistance plasmid that is temperature-sensitivefor replication, pMAK705 (Hamilton et al., 1989).Plasmid-bearing cells raised to the non-permissivetemperature should only form a drug-resistant colony if acopy of the plasmid is replicated by its integration intothe chromosome. The number of colonies formed at 30°Cand 42°C therefore depends upon the fraction of cellscontaining a plasmid copy integrated into the chromosomeat the time of plating or which integrate in the firstgenerations after plating. The results (Table II) show thata 231 bp dif fragment (pLIM701, Kuempel et al., 1991),a loxP site (pN78) or a full cer6 site (pN79) can recombinewith a homologous site in the chromosome when thecognate recombinase is present. The integrated fractionwas highest for difwhen the chromosomal site was in theterminus region (37-62%), and for cer6 in pepA and argRderivatives of NL289 (22-33%). In a pepA+ argR+ strain,when cer6 recombination is directed intramolecularly, theintegrated fraction was reduced almost 100-fold (0.45%).The integrated fraction for loxP recombination was 1%,while that for dif close to oriC or lacZ was just under1%. These results compare with controls in which recom-binases or chromosomal or plasmid recombination sitesare lacking (9X10-4-4.9x10-6).The significance of differences in the integrated fraction

is not clear. For example, the integrated fraction at 30°Cmay depend on the relative frequencies of integration andexcision at that temperature and the relative growth ratesof cells with autonomous and integrated yet replicationallyactive plamsids (Yamaguchi and Tomizawa, 1980). Therelative contribution of these and new integrations afterplasmid replication is inhibited at 42°C is not known.Therefore, we cannot be sure whether differences inintegrated fraction reflect differences in rates of recombina-tion or differences in equilibria that arise as a consequenceof many factors, or both. That rates of recombination arein part being analysed is supported by the results withcer6; the integrated fraction is 50-70 times higher whenrecombination should show no selectivity (argR and pepAstrains) compared with when intramolecular resolutionshould be favoured (NL289). Attempts to assay integration

1565

A

C

E F

Fig. 5. Microscopic analysis of cell and nucleoid morphology. Exponential phase cultures grown in LB broth at 37°C were prepared for photographyby condensation and staining of cell nucleoids with chloramphenicol and DAPI respectively (Materials and methods). (A) NL296 (33 bp dif sitereplacement). (B) NL246 (33 bp dif site replacement). (C) NL250 (532 bp dif site at min 8 of chromosome). (D) NL350 (532 bp dif site at min 84of chromosome). (E) NL208 (loxP site in place of dif). (F) NL208 + pRH200 (Cre expression vector). The long axis of each photograph represents-70 ,um. Condensation of the nucleoids does not interfere significantly with the phenotype demonstrated.

rate more directly, by analysis of the ratio of CmRtransformants at 30°C and 42°C immediately after trans-formation, were not successful because of the low trans-formability of the plasmids at all temperatures.

Putative active site residues of XerC and XerD are

required for normal nucleoid segregationThe recombinases XerC and XerD are closely relatedmembers of the lambda integrase family of recombinasesand have substantial homology over two conserveddomains believed to contribute to the active sites of theseproteins (Colloms et al., 1990; Blakely et al., 1993).

Derivatives of XerC and XerD mutant in either thepresumptive active site tyrosine nucleophile (Pargelliset al., 1988; XerCY275F and XerDY279F) or a conserveddomain II arginine implicated in DNA phosphodiesteractivation (Parsons et al., 1988; XerCR243Q andXerDR247Q) promote recombination at cer in in vivoplasmid resolution assays at undetectable or greatlyreduced levels (Blakely et al., 1993). The mutant proteinsare still able to bind recombination sites. In order to testwhether cells containing these mutations are defective inchromosome segregation/cell division, the xerC mutantstrain DS984 and the XerD mutant DS9008 were

1566


B

D


difand

difandother sites

84.18.033.6

Fig. 6. Chromosomal locations used to study positional effects upon

dif function. This representation of the Ecoli chromosome shows thepositions into which ectopic dif and loxP sites were introduced,relative to the natural location of dif in the replication terminus andthe position 0 min. The xerC and xerD genes and oriC are also shown.Positions are taken from Bachman (1990), Kuempel et al. (1991) andBlakely et al. (1993).

transformed with plasmids producing wild-type XerC andwild-type XerD or with plasmids expressing each of theputative active site mutants. DS984 expressing wild-typeXerC from pSD104 (Colloms et al., 1990) and DS9008expressing wild-type XerD from pRM130 (Blakely et al.,1993) showed wild-type cell morphology. In contrast,derivatives of DS984 and DS9008 expressing the mutantproteins maintained their mutant cell morphology pheno-type (data not shown), indicating that difrecombination inits normal chromosomal location and normal chromosomesegregation require catalytically active recombinases.

DiscussionIs it site-specific recombination or some other process

that occurs at dif to ensure normal E.coli chromosomalsegregation and cell division? Here we have presentedthree pieces of data that strongly point to it being site-specific recombination. First, a 33 bp fragment containingthe 28 bp dif core site that is functional in site-specificrecombination in in vivo plasmid assays is sufficient fornormal chromosomal segregation and cell division whenplaced in either orientation in its normal position in thereplication termination region. Second, mutations in theputative active sites of XerC or XerD interfere with bothsite-specific recombination and chromosome segregation/cell division. Last, Cre-mediated site-specific recombina-tion at a loxP site inserted into the replication terminationregion can functionally replace Xer recombination at dif,as judged by suppression of the aberrant cell nucleoidmorphology of a difdeletion. Since loxP has little sequencesimilarity to dif and because suppression requires Cre, itseem most reasonable that the process of site-specificrecombination leads to normal chromosome segregation.

It has previously been proposed (Blakely et al., 1991;Kuempel et al., 1991; Sherratt et al., 1993) that the

function of Xer site-specific recombination is to 'undo thedamage' of relatively rare homologous recombinationevents: odd numbers of exchanges between circularchromosomes will generate dimeric molecules that cannotbe segregated effectively. If so, the viability and substantialproportion of normal cells in dif- or Xer- populationsseems to indicate that many cells do not encounter aproblem in segregation and cell division, presumablybecause an odd number of homologous exchanges occursrelatively rarely, perhaps at most every few cell generations(Forro and Wertheimer, 1960; Kuempel et al., 1991). Xerrecombination at natural plasmid sites (e.g. cer in ColE1)is exclusively intramolecular and is therefore ideally suitedto convert multimers to monomers. This selectivity forintramolecular resolution requires about 190 bp ofaccessory sequences adjacent to the core recombinationsite and accessory proteins (Stirling et al., 1988b, 1989;Summers, 1989; Blakely et al., 1993). We have proposed(Blakely et al., 1993; Sherratt, 1993) that the mechanismby which this selectivity is obtained is similar to that usedby the Tn3 res/resolvase and related systems, which showa strong selectivity for either resolution (res/resolvase) orinversion (gix/Gin). In all of these cases, accessory proteinsand accessory sequences appear to be involved in theformation of a recombinational synapse which has thecomplexity to direct the DNA into a precise synapsetopology (see, for example, Stark et al., 1989). We haveno evidence that recombination at dif can show resolutionselectivity; recombination in plasmid substrates occursinter- and intramolecularly and no data support the involve-ment of accessory sequences outside the dif core site oradditional accessory proteins. How then does recombina-tion at dif convert putative chromosomal dimers to mono-mers? Either there is selectivity for intramolecularresolution by an unknown mechanism that does not operatewhen dif is present in multicopy plasmid and does notrequire accessory sequences or recombination at dif in theterminus region of the chromosome leads to chromosomesegregation despite the lack of resolution selectivity. Twoways in which inter- and intramolecular recombination atdif could lead to effective chromosome segregation havebeen proposed (Blakely et al., 1991; Kuempel et al., 1991;Sherratt et al., 1993). In one, rapid recombination betweennewly replicated dif sites would ensure that chromosomeswere monomeric 50% of the time, irrespective of whetherhomologous recombination had acted; they could thereforebe physically separated at or close to the normal time inthe cell cycle by a chromosome partition mechanism. Thesecond model relies on Xer recombination proceeding bya Holliday junction intermediate, as would be expectedfor this class of enzyme (see, for example, Stark et al.,1992). Holliday junction formation by site-specificrecombination at two chromosomal dif sites would produce'figure-of-eight' molecules that are topologically identical,irrespective of whether or not there has been a homologousrecombinational exchange. If the process that leads tochromosome segregation now begins to separate thedaughter chromosomes linked by the Holliday junction, aresolution event that senses the local conformationalchanges induced by separation could lead to monomericdaughter chromosomes. We have no evidence to supportthis latter hypothesis other than the in vivo accumulation

1567

difandloxP

\%~oric

pst

xerD

Positions

psiCAB phoUlacZin terminus

Min.


Table II. Integration of plasmids into the chromosome

Bacterial strain Plasmid recombination site

Name and description Site None dif loxP cer6pMAK705 pLIM701 pN78 pN79

DS941, wild-type dif 1.1 x 10-4 4.6x 10-1 3.9x 10-4 9.5x 10-5NL40, dif deletion None 5.0X10-5 5.3x10-5 5.3x10-4 4.0X10-5DS981, xerC dif 1.7X 10-5 1.3x 10-5DS9008, xerD dif 3.3x 10-5 2.3x 10-4DS902, recA dif 3.3x10-5 4.4xI0-NL246, dif in terminus 33 bp dif 1.3X10-4 4.0X10-NL247, dif in terminus 532 bp dif 4.9x 10-6 3.0X10-1NL287, dif in terminus 532 bp dif 1.6X10-4 5.7X10-1NL296, dif in terminus 33 bp dif 5.3X10-5 3.7XI0-NLAIN, KmR by dif dif 6.9x 10-4 6.3x 10-1NLAOUT,KmR by dif dif 7.5X10-4 3.7x10-1NL250, dif in lacZ 532bp dif 7.5X10-5 7.1X10-3NL350, dif near oriC 532bp dif 1.2x 10-4 8.8x 10-3NL208, loxP in terminus loxP 4.OX 10-4 6.7x 10-5NL208 + Cre loxP 9.0X104 lOXNL289, cer6 in terminus cer6 l.5x l0-5 4.5X 10-3NL259 (NL289 pepA) cer6 1.2x 10-4 2.2X 10-NL279 (NL289 argR) cer6 1.3x 10-5 3.3x10-1

These figures are calculated as number of Cmr colonies at 42°C/number of Cmr colonies at 30°C. They represent the fraction of cells in a populationcontaining a plasmid copy integrated into the chromosome for the strains and plasmids shown. All plasmids used are derived from the Cm' vectorpMAK705 and contain the recombination sites shown. Most results are calculated as the mean fraction from several independent experiments. Allstrains tested were free of other plasmids with the exception of NL208 + Cre, containing the Cre expression vector pRH200. The dif sites in strainsDS941, DS981, DS9008 and DS902 are in the natural wild-type context in the terminus region.

of Xer-mediated Holliday junction-containing moleculesunder some conditions (McCulloch et al., 1994).The inability of the dif sites in chromosomal locations

outside the replication terminus to suppress the mutantcell morphology phenotype could have several explana-tions. For example, dif may be able to recombine, butunable to fulfil its biological function because of itsunnatural positioning away from the replication terminus.Alternatively, the strains containing ectopic dif sites mayexhibit reduced recombination or may be influencedadversely by local context. The plasmid integration assaysshow that the ectopic difsites are functional for recombina-tion, but give few integrants as compared with the normallylocated dif sites. We cannot be sure whether this reflectsreduced recombination and cannot tell if it relates to thelack of normal chromosomal segregation. One explanationof why dif may only be able to fulfil its biological role iflocated inside the replication terminus region could relateto physical separation of the two copies of the site afterreplication. Once copied, non-terminus sequences mightbe expected to be physically separated rapidly by anactive partition mechanism acting upon a replicatingchromosome. This separation would prevent recombina-tion between two ectopic difcopies on sister chromosomes.Sequences in the terminus region may not only be the lastto be replicated, but may remain in physical proximityafter replication for longer periods than sequences locatedelsewhere, thus maximizing their recombination potential.

It appears that though a 33 bp fragment containing difis sufficient to allow normal chromosome segregationwhen it is located in the terminus region, the context ofthe sequence can influence its function. For example,NLAIN and NL246, though they contain what should bea functional dif site in the terminus region, show aberrantcell and nucleoid morphology. Ability to support difrecombination, as judged by the integration assay, was no

different from that of the related morphologically wild-type strains NLAOUT and NL296. It is possible that theinability of these dif sites to suppress the mutant celldivision phenotype is related to the transcription of adja-cent antibiotic resistance genes. The manner in which thiseffect is mediated is the subject of further work. Incontrast, dif can function in chromosome segregation ineither orientation, demonstrating that it does not need tobe correctly oriented with respect to other chromosomalsequences.

Neither the Tn3 res/resolvase recombination system northe ColEl cer/Xer system suppressed the filamentousmutant phenotype when cer and res were substituted atthe position of the normal dif site. Recombination at eachof these sites shows a strong resolution selectivity, whichwe believe is mediated as a consequence of a requirementfor a recombinational synapse in which a precise smallnumber of supercoils are trapped (see, for example, Starket al., 1989; S.Colloms and D.Sherratt, unpublished data).Although two sites on a chromosome dimer would bedirectly repeated, they would be 4.7 Mbp apart and randominteractions between such distantly spaced sites would beunlikely to result in synaptic complexes of a definedtopology. Because Cre recombination at loxP, like that ofXerC/D at dif, occurs without apparent resolutionselectivity (Hoess and Abremski, 1990), it was not un-expected to find that recombination at loxP in the terminusregion could substitute for that at dif. However, it wassurprising to find that a cer6 variant site inserted into theterminus region did not suppress the difmutant phenotypein wild-type, argR or pepA backgrounds. In a wild-typebackground, recombination at this site shows intra-molecular selectivity, but mutations in the accessoryproteins argR or pepA abolish this constraint (Summers,1989). This behaviour is supported by our assays ofplasmid integration. We would have predicted, therefore,

1568


Table III. Plamids

Plasmids Description Reference/source

pUC18 Cloning vector Yanisch-Perron et al. (1985)pMTL23 Cloning vector Chambers et al. (1988)pMAK705 Temperature-sensitive pSC 101 replicon vector Hamilton et al. (1989)pLIM701 pMAK705 + dif for plasmid integration assay Kuempel et al. (1991)pUC71K Source of 1.4 kb Km fragment Vieira and Messing (1982)pGM 160 Source of 1.5 kb GmR fragment Muth et al. (1989)pMIN33 33 bp dif minimal oligonucleotides in pUC18 Blakely et al. (1991)pMA1441 282 bp Tn3 res site in pUC18 M.BoocockpTypell Full cer6 site in pUC9 Summers (1989)pNI pUC18 + 5.5 kb BamHI dif fragment from pBS12 This workpN20 pMTL23 + Ndel-HindIII cer6 site fragment from pTypelI This workpN40 pMTL23 cut with HindlIl + loxP site oligonucleotides 01 and 02 This workpN78 pMAK705 + PstI-XbaI loxP site from pN40 This workpN79 pMAK705 + BamHI-SstI cer6 site fragment from pN20 This work

Table IV. Ecoli K12 strains and oligonucleotides

Strain Genotype Reference/construction

AB1 157 (see text) Bachmann (1972)DS941 ABI 157 recFJ43, lacZAM15, lacIq Summers and Sherratt (1988)DS902 AB 1157 recA 13 Bednarz et al. (1990)JC7623 AB 1157 recB21C22 sbcBI5C201 Kushner et al. (1971; Lloyd and Buckman, 1985)DS956 DS941 xerA9 (argR::Tpr) Colloms et al. (1990)DS957 DS941 xerBI (pepA::TnS) Stirling et al. (1989)DS984 DS941 xerC::YJ7 (::mini Mu, Cmr) Colloms et al. (1990)DS981 DS941 xerC2(::Kmr) Colloms et al. (1990)DS9008 DS941 xerD2(::mini TnJO-9, Kmr) Blakely et al. (1993)NL40 DS941 difA6 Replacement of dift of DS941 with difA6::Kmr from pNA6NLAIN DS941 difAOIN Replacement of dift of DS941 with difAOIN::Kmr from pNAOINNLAOUT DS941 difAOOUT Replacement of dijf of DS941 with difAOOUT::Kmr frompNAOOUTNL244 NL40 difA6::resGmr Replacement of difA6::Kmr of NL40 with difA6::resGmrfrom pN24NL245 NL40 difA6::cerGmrl Replacement of difA6::Kmr of NL40 with difA6::cerGmrjfrom pN25NL246 NL40 difA6::33bpdiyGmrl Replacement of difA6::Kmr of NL40 with difA6::33bpdi]Gmrl from pN26NL247 NL40 dlfA6::S32bpdifl3mrl Replacement of difA6::Kmr of NL40 with difA6::532bpdiftmrl from pN27NL285 NL40 difA6::cerGmr2 Replacement of difA6::Kmr of NL40 with difA6::cerGmr2 from pN65NL287 NL40 difA6::532bpdiiGmr2 Replacement of difA6::Kmr of NL40 with difA6::S32bpdifmr2 from pN67NL286 NL4O difA6::33bpdiffjmr2 Replacement of difA6::Kmr of NL40 with difA6::33bpdiffimr2 from pN66NL296 NL40 difA6::33bpdiiGmr3 Replacement of difA6::Kmr of NL40 with difA6::33bpdiiGmr3 from pN76NL259 DS941 pepA difA6::cer6Gmr Replacement of dif+ of DS957 with difA6::cer6Gmr from pN69NL279 DS941 argR difA6::cer6Gmr Replacement of dif+ of DS956 with difA6::cer6Gmr from pN69NL289 NL40 difA6::cer6Gmr Replacement of difA6::Kmr of NL40 with difA6::cer6Gmr from pN69NL233 DS941 lacZ::Gmr Replacement of lacZ+ of DS941 with lacZ::Gmr from pN32NL238 DS941 IacZ::532bpdiyGmr Replacement of lacZ+ of DS941 with lacZ::532bpdiftGmr from pN37NL250 NL40 IacZ::532bpdifjmr Replacement of lacZ+ of NL40 with IacZ::532bpdijGmr from pN37NL343 DS941 A(pstCAB phoU)::Gmr Replacement of pst+ of DS941 with A(pstCAB phoU)::Gmr from pN42NL348 DS941 A(pstCAB phoU)::532bpdifGmr Replacement of pst+ of DS941 with A(pstCAB phoU)::532bpdiGmr from pN47NL350 NL40 A(pstCAB phoU)::532bpdiflGmr Replacement of pst+ of NL40 with A(pstCAB phoU)::532bpdiiGmr from pN47NL208 NL40 difA6::loxPGmr Replacement of difA6::Kmr of NL40 with difA6::loxPGmr from pN58NL202 NL40 lacZ::loxPGmr Replacement of lacZ+ of NL40 with lacZ::loxPGmr from pN83NL203 NL40 A(pstCAB phoU)::1oxPGmr Replacement of pst+ of NL40 with A(pstCAB phoU)::IoxPGmr from pN84

OligonucleotidesTop and bottom strand loxP site oligonucleotides for production of pN40

01 5'AGCTCGAGATAACTTCGTATAATGTATGCTATACGAAGTTAT 3'02 5'AGCTATAACTTCGTATAGCATACATTATACGAAGTTATCTCG 3'

that cer6 could have functionally replaced dif in argR andpepA strains. We have no explanation for this result.

Though we have not characterized homologues of difin other bacteria, we suspect that they exist and functionin normal chromosomal segregation. This is supported bythe observation that relatives of E.coli contain sequenceshomologous to XerC and XerD (G.Blakely and D.Sherratt,unpublished data).

Materials and methodsBacterial strains, plasmids and growth conditionsBacteria used were derived from the Ecoli K12 strain AB 1 157 (Bachmann,1972). ABI 157 is thrl leuB6hisG4 thil aral4 A(gpt-proA)62 argE3 galK2supE44 xylS mtll tsx33 lacYl rpsL31. Principle plasmids, strains andoligonucleotides are detailed in Tables III and IV. Plasmids used in theconstruction of strains are not included on the grounds of brevity. Detailsof these plasmids are available on request. Strains constructed as part ofthis study were verified by hybridization of radiolabelled DNA fragments

1569


to Southern blots of restriction enzyme-digested genomic DNA. Bacteriawere grown in LB media (Miller, 1972) unless otherwise stated, andsupplemented with antibiotics as required (ampicillin 50 ,ug/ml, kanamycin50 ,ug/ml, gentamicin 5 ,ug/ml, chloramphenicol 25 gg/ml, streptomycin100 jig/ml, tetracycline 10 gg/ml). Tests for expression of alkaline phos-phatase and f-galactosidase were performed using colonies growing onsolid media, following the methods of Wanner and Latterell (1980) andMiller (1972) respectively.

Genetic techniquesGeneralized transduction with bacteriophage PIk,. was carried out asdescribed by Miller (1972). Plasmid-borne mutations were generally intro-duced into the chromosome by transformation of strains deficient in exo-nuclease V with linearized DNA (Jasin and Schimmel, 1984; Winans et al.,1985) before transduction into other genetic backgrounds. However, themethod of Hamilton et al. (1989) was used for constructs containingloxP sites.

Phenotypic analysis/microscopyThe filamentous phenotype evident in microscopic analysis of xer or difmutant cultures depends greatly upon the conditions under which the cellshave been grown. Therefore assays were performed upon cultures producedin one of two ways. Either liquid cultures originating from a 5000-folddilution of a stationary phase culture into fresh medium and grown intolate exponential phase or cells taken from colonies grown on solid mediafor -20 h and suspended in liquid medium were used.

Microscopic analysis of nucleoid morphology followed the methods ofHiraga et al. (1989) and Eliasson et al. (1992). Nucleoids were generallycondensed with chloramphenicol (250 ,ug/ml) for 20 min before stainingwith DAPI (4',6-diamidino-2-phenylindole). These preparations wereexamined using combined phase contrast and fluorescence microscopy.Assays of cell length were performed with untreated cultures using eitherphase contrast or Nomarski optics. Measurements of cell length (Table I)were calculated from projected images of photographic slides. One wayanalysis of variance was used to compare sets of cell length data.

In vitro DNA manipulationsRoutine techniques used for construction and preparation of plasmid DNA,agarose gel electrophoresis, Southern hybridization and radioactive label-ling of synthetic oligonucleotides were as described by Sambrook et al.(1989). Radiolabelling of plasmid DNA restriction fragments was per-formed using a Random Primed DNA Labelling Kit (BoehringerMannheim). Restriction fragments were purified from agarose gels usingSPIN-X filtration columns (Costar, Cambridge, MA). Preparation ofchromosomal DNA followed the method of Neumann et al. (1992).

Plasmid integration assaysThis method was adapted from Kuempel et al. (1991). Strains to be testedwere transformed with pMAK705 or derivative plasmids, then grown inliquid culture with chloramphenicol into stationary phase. Cells werediluted and plated onto pre-warmed plates at 30°C or 42°C. Colonies werecounted after 24 h of growth. The number of colonies formed at 42°Cdivided by the number at 30°C was taken to be the fraction of transformedcells containing a plasmid copy integrated into the chromosome. Pilotexperiments showed that in all of the experiments performed, this fractiondid not vary significantly with the time from transformation to plating,growth phase of the culture at plating or transformation frequency.

AcknowledgementsWe would like to thank Garry Blakely and Marshall Stark for criticalreading of this manuscript, ideas and other help, Lidia Arciszewska forplasmids and advice on expressing mutant Xer proteins and Peter Kuempel,Martin Boocock, Gerhard May, lain Hunter and Brian Sauer for kindlyproviding plasmids and strains. We also thank our colleagues in the Genet-ics Department for their many contributions, ideas and discussions, particu-larly Kevin O'Dell, Richard McCulloch, Jennifer Roberts and Karen Grant.This work was supported by the Medical Research Council.

ReferencesAmemura,M., Shinagawa,H., Makino,K., Otsuji,N. and Nakata,A. (1982)

J. Bacteriol., 152, 692-701.Austin,S., Ziese,M. and Sternberg,N. (1981 ) Cell, 25, 729-736.Bachmann,B.J. (I1972) Bacteriol. Rev., 36, 525-557.Bachmann,B.J. (1990) Microbiol. Rev., 54, 130-197.

Bednarz,A.L., Boocock,M.R. and Sherratt,D.J. (1990) Genes Dev., 4,2366-2375.

Bejar,S. and Bouche,J.P. (1983) J. Bacteriol., 153, 604-609.Blakely,G., Colloms,S.D., May,G., Burke,M. and Sherratt,D.J. (1991) New

Biologist, 3, 789-798.Blakely,G., May,G., McCulloch,R., Arciszewska,L.K., Burke,M.,

Lovett,S.T. and Sherratt,D.J. (1993) Cell, 75, 351-361.Chambers,S.P., Prior,S.E., Barstow,D.A. and Minton,N.P. (1988) Gene,

68, 139-149.Colloms,S.D., Sykora,P., Szatmari,G. and Sherratt,D.J. (1990) J.

Bacteriol., 172, 6973-6980.Eliasson,A., Bernander,R., Dasgupta,S. and Nordstrom,K. (1992) Mol.

Microbiol., 6, 165-170.Forro,F.J. and Wertheimer,S.A. (1960) Biochim. Biophys. Acta, 40, 9-21.Hamilton,C.M., Aldea,M., Washburn,B.K., Babitzke,P. and Kushner,S.R.

(1989) J. Bacteriol., 171, 4617-4622.Hiraga,S., Niki,H., Ogura,T., Ichinose,C., Mori,H., Ezaki,B. and Jaffe,A.

(1989) J. Bacteriol., 171, 1496-1505.Hoess,R.H., and Abremski,K. (1990) In Eckstein,F. and Lilley,D.M.J.

(eds), Nucleic Acids and Molecular Biology. Springer-Verlag, Berlin,Germany, Vol. 4, pp. 99-109.

Jasin,M. and Schimmel,P. (1984) J. Bacteriol., 159, 783-786.Kitts,P.A., Symington,L.S., Dyson,P. and Sherratt,D.J. (1983) EMBO J.,

2, 1055-1060.Kuempel,P.L., Pelletier,A.J. and Hill,T.M. (1989) Cell, 59,581-583.Kuempel,P.L., Henson,J.M., Dircks,L., Tecklenburg,M. and Lim,D.F.

(1991) New Biologist, 3, 799-811.Kushner,S.R., Nagaishi,H., Templin,A. and Clark,A.J. (1971) Proc. Natl

Acad. Sci. USA, 68, 824-827.Landy,A. (1989) Annu. Rev. Biochem., 58, 913-949.Lloyd,R.G. and Buckman,C. (1985) J. Bacteriol., 164, 836-844.Mack,A., Sauer,B., Abremski,K. and Hoess,R. (1992) Nucleic Acids Res.,

20,4451-4455.McCulloch,R., Coggins,L.W., Colloms,S.D. and Sherratt,D.J. (1994)EMBO J., 13, 1844-1855.

Miller,J.H. (1972) Experiments in Molecular Genetics. Cold Spring HarborLaboratory Press, Cold Spring Harbor, NY.

Muth,G., Nussbaumer,B., Wohlleben,W. and Puhler,A. (1989) Mol. Gen.Genet., 219, 341-348.

Neumann,B., Pospiech,A. and Schairer,H.U. (1992) Trends Genet., 8,332-333.

Pargellis,C.A., Nunes-DUby,S.E., Vargas,L. and Landy,A. (1988) J. Mol.Biol., 263, 7678-7685.

Parsons,R.L., Prasad,P.V., Harshey,R.M. and Jayaram,M. (1988) Mol. Cell.Biol., 8, 3303-33 10.

Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning: ALaboratory Manual. Cold Spring Harbor Laboratory Press, Cold SpringHarbor, NY.

Sherratt,D.J. (1993) In Eckstein,F. and Lilley,D.M.J. (eds), Nucleic Acidsand Molecular Biology. Springer-Verlag, Berlin, Germany, Vol. 7, pp.202-216.

Sherratt,D.J., Blakely,G., Burke,M., Colloms,S., Leslie,N., McCulloch,R.,May,G. and Roberts,J. (1993) In Heslop-Harrison,J.S. and Flavell,R.B.(eds), The Chromosome. Bios Scientific Publishers, Oxford, UK,pp. 25-41.

Sherratt,D.J., Arciszewska,L., Blakely,G., Colloms,S., Grant,K.,Leslie,N.R. and McCulloch,R. (1995) Proc. R. Soc. B., 347, 37-42.

Stark,W.M., Boocock,M.R. and Sherratt,D.J. (1989) Trends Genet., 5,304-309.

Stark,W.M., Boocock,M.R. and Sherratt,D.J. (1992) Trends Genet., 8,432-439.

Stirling,C.J., Stewart,G. and Sherratt,D.J. (1988a) Mol. Gen. Genet., 214,80-84.

Stirling,C.J., Szatmari,G., Stewart,G., Smith,M. and Sherratt,D.J. (I 988b)EMBO J., 7,4389-4395.

Stirling,C.J., Colloms,S.D., Collins,J.F., Szatmari,G. and Sherratt,D.J.(1989) EMBO J., 8, 1623-1627.

Summers,D.K. (1989) EMBO J., 8, 309-315.Summers,D.K. and Sherratt,D.J. (1984) Cell, 36, 1097-1103.Summers,D.K. and Sherratt,D.J. (1988) EMBO J., 7, 851-858.Summers,D.K., Beaton,C.W.H. and Withers,H.L. (1993) Mol. Microbiol.,

8, 1031-1038.Viera,J. and Messing,J. (1982) Gene, 19, 259-268.Wanner,B.L. and Latterell,P. (1980) Genetics, 96, 353-366.Winans,S.C., Elledge,S.J., Krueger,J.H. and Walker,G.C. (1985) J.

Bacteriol., 161, 1219-1221.Yamaguchi,K. and Tomizawa,J.I. (1980) Mol. Gen. Genet., 178, 525-533.Yanisch-Perron,C., Viera,J. and Messing,J. (1985) Cene, 33, 103-119.

Received on November 9, 1994,revised on Januarv 12, 1995

1570

Documents

Site-specific recombination in the replication