Vol. 173, No. 12JOURNAL OF BACTERIOLOGY, June 1991, p. 3630-36340021-9193/91/123630-05$02.00/0Copyright © 1991, American Society for Microbiology

Fine-Structure Analysis of the P1 Plasmid Partition SiteKATHY A. MARTIN,t MICHAEL A. DAVIS, AND STUART AUSTIN*

Laboratory of Chromosome Biology, ABL-Basic Research Program, National Cancer Institute-Frederick CancerResearch and Development Center, Frederick, Maryland 21702-1202

Received 12 December 1990/Accepted 3 April 1991

P1 plasmid partition requires two plasmid-encoded Par proteins and a cis-acting site. The site, parS, lies ina region consisting of a 13-bp palindrome and an adjacent AT-rich sequence. A series of point mutations wereanalyzed for their effects on partition site activity. The results indicated that only the left arm of the palindromeand some adjacent bases were needed. The limits of the functional site were further refined to a maximum of22 bp, which includes binding sites for the P1 ParB protein. Mutations in the 22-bp site cause concomitantdefects in partition and the ability to exert partition-mediated incompatibility. Like the region immediately tothe left of the 22-bp region, the right arm of the palindrome is not essential for partition but does containinformation that affects the specificity of incompatibility.

Bacteriophage P1 lysogenizes Escherichia coli as a stableunit copy plasmid. Its rate of loss is less than one in 105 celldivision events, primarily because of an active partitionapparatus which ensures that each daughter cell receives aplasmid copy at the time of cell division (3). The geneticmaterial of P1 that is required for partition has been localizedto a 2.1-kb segment that is adjacent to, but separable from,the primary replication functions of the plasmid. Two essen-tial partition proteins, ParA and ParB, and a small cis-actingpartition site (parS) are required and sufficient for partition(3, 9, 12). The site lies in a region immediately downstreamof the two par open reading frames (Fig. 1) (12). Models forpartition propose that the parS sequence acts as a functionalequivalent to the eukaryotic centromere (3). According tosuch models, this site binds the plasmid to the bacterial"mitotic" apparatus, guaranteeing accurate segregation.The region containing the parS site also acts as a deter-

minant for plasmid incompatibility (incB) (2). Two differentplasmids with the same type of partition region cannot bestably maintained in the same cell line. One proposedexplanation for this assumes that the site is a recognition sitefor choosing pairs of plasmids for directed movement toopposite halves of the dividing cell. Individual copies of twodifferent plasmids, both carrying cis-acting sites of the samespecificity, would not be distinguishable by the partitionsystem, resulting in the missegregation called incompatibil-ity (13).The P1 partition site can exhibit two different incompati-

bility phenotypes (7, 12). In order for a site to give awild-type incompatibility phenotype (IncB+), it requires 84bp that include multiple binding sites for the ParB proteinand a site which binds the host integration host factor (IHF)protein (7) (Fig. 1). However, much of this information isneeded for specificity but not for the actual mechanics ofpartition. Sequences sufficient for partition site activity havebeen mapped to a 35-bp subregion that includes a perfect13-bp inverted repeat and an adjacent AT-rich sequence (6,12). However, plasmids carrying these truncated sites showan altered incompatibility phenotype. Such plasmids areincompatible with their own type but not with the wild type

* Corresponding author.t Present address: Department of Biological Sciences, Central

Connecticut State University, New Britain, CT 06050-4010.

(12). We refer to this altered incompatibility phenotype asIncBd (7). This change in incompatibility properties can beinduced in the 84-bp wild-type site by various mutations inthe sequences flanking the 35-bp region (7) or by a hostmutation that eliminates IHF activity (himA) (10).

In this paper, we seek to define the minimal informationrequired to constitute a functional partition site with IncBdspecificity. It should contain the bases that interact withwhatever cellular machinery is actually responsible for se-lective movement of plasmid copies to daughter cells.

MATERIALS AND METHODS

Bacterial strains. N100 (5) is recA13 galK pro Smr. BR825is our laboratory isolate of CM5649 [polA supD(Ts) Tcrtrp(Am)] that has lost the temperature-sensitive supD allele(11). CC241 is BR825 transformed with plasmid pALA283,which supplies the P1 ParA and ParB proteins. CC240 isBR825 containing the control vector pLG338. RW1840 issupE supF metB r- m- galAl82 himA: :TnlO and was kindlysupplied by R. Weisberg.Map coordinates and designation of mutations. Coordinates

from the standard map of the P1 plasmid-maintenance region(1) are referred to as P1-4380, etc. The first bp of the 34-bppartition site region is P1-4357 (Fig. 1). For the purposes ofthis paper and for brevity, bases within the 34-bp parS regionare designated bp 1 through bp 34, so that P1-4357 isredefined as bp 1 (Fig. 1). Mutations are designated by thebase changes and bp numbers changed. Thus, T-*C,4 has aT to C base change at the fourth bp of the 34-bp sequence;AT,2 is deleted for bp 2, and A23-34 is deleted for bases 23through 34.

Plasmid constructions. The pSC101-based plasmids pALA283 and pLG338 and the dual-origin vector pALA136 weredescribed previously (12).

Dual-origin plasmids based on the P1 rep-containing vectorpALA136. Plasmids pALA415 and pALA419 were con-structed by using single-stranded oligonucleotides. The parSregion of these plasmids is inserted between the uniqueEcoRV and AvaI sites of pALA136, as previously described(12). Plasmids pALA455, pALA456, pALA442, pALA453,and pALA449 were made by inserting double-stranded oli-gonucleotides with BamHI- and SphI-complementary endsinto BamHI-SphI-digested pALA136. Variants of pALA455(pALA502, pALA504, etc.) that have randomly generated

3630

on February 7, 2018 by guest

http://jb.asm.org/

Dow

nloaded from

P1 PLASMID PARTITION SITE 3631

P1-4303Taq I

4290 4310- 4330TCGATAAAAAGCCGAAGCCTTAAACT17CGCCATTCAAATTTCA&ATTAACTGACTGTTT

L a I

IHF recognition site

P1-4357 PI-4390Dra I Sty I

* 4350 - ---.- 4370 -

TrAAAGTAAATTACTCTAAAATCMTT S;TGAAAI,CGCCACGkTTTCACITTGG

bp 10 bp 20 bp 30bp bp3

34 bp parS sequence -pFIG. 1. The TaqI to Styl fragment containing the 84-bp P1 incB+

region. Coordinates above the sequence are conventional for the P1plasmid maintenance region. The 84-bp region from P1-4303 toP1-4386 is sufficient to exert IncB+ incompatibility (7). The coordi-nates below (bp 1 through 34) are used in this paper for bases withinthe 34-bp region, including the parS site. Boxes mark heptamersequences thought to be recognition sites for binding of the ParBprotein, and the site resembling the IHF-binding consensus se-quence is marked (7). Half-arrows mark the 13-bp perfect palin-drome within the 34-bp sequence. Black bars above the sequencemark restriction sites.

mutations were constructed in the same way. PlasmidspALA430 and pALA439 were made by removing the SspI-AvaI fragments of pALA425 and pALA438, respectively,that span the mutant P1 bases and inserting them intoEcoRV-AvaI-digested pALA136.Random oligonucleotide mutagenesis. All oligonucleotides

were made on an Applied Biosystems (Foster City, Calif.)380B DNA synthesizer. To generate random mutations inthe 34-bp parS sequence, each base was contaminated with1.66% of each of the other three bases. These mixes wereused to synthesize an oligonucleotide with BamHI- andSphI-complementary ends. A wild-type complementarystrand was synthesized and hybridized to the mutated oligo-nucleotide pool, and the resulting double-stranded oligonu-cleotides were inserted into BamHI-SphI-digested pALA136. These plasmids are thus variants of plasmid pALA455described above. Plasmids pALA504 and pALA502 were

generated by this method, as were the unnumbered mutantsin Fig. 2.High-copy-number plasmids based on the vector pBR322.

Plasmids pALA412, pALA417, pALA425, pALA431, andpALA438 were made by inserting double-stranded syntheticoligonucleotides containing the appropriate P1 sequence intoBamHI-SphI-digested pBR322 DNA. Each oligonucleotidewas designed to have the appropriate complementary ends.Plasmids pALA498 and pALA499 were created by removingthe HindIII-NruI fragments from pALA449 and pALA504,respectively, and reinserting them between the HindIII andNruI sites of pBR322.

Plasmid pALA464, a Kmr Aps derivative of pBR322, wasdescribed by Davis et al. (7). A 174-bp TaqI-HincII fragmentcontaining bp P1-4278 to P1-4451 (1) was inserted betweenEcoRI and Styl of pALA464, generating pALA465.

Plasmid pALA411 is a StyI-Styl deletion of pALA270 (1)that removes all P1 bases to the right of bp P1-4390 (bp 34,Fig. 1), leaving an intact P1 par region including parA, parB,and parS. Plasmid pALA271 is like pALA411 but has theparS site deleted (1). Plasmid pALA463 was made bydigesting pBR322 with HindIlI and BamHI and inserting

TABLE 1. The maintenance stability during unselected growthof unit copy constructs when the P1 Par proteins

are supplied in trans

% Retention during 25 generationsPlasmida 34-bp par

region CC241b (supplies CC240CP1 Par proteins) (control)

pALA136 None <2 <2pALA415 Wild type 48 <2pALA455 Wild type 58 <2pALA449 T- C,4 1 2pALA419 A--T,14 4 <2pALA504 C--G,16 <2 <2

a Plasmid pALA455 contains the 34-bp wild-type parS sequence (Fig. 2) inthe dual replicon vector pALA136. It replicates from a low-copy-number P1replication origin in the polA strain used. Plasmid pALA415 is comparable(Materials and Methods; 12) but has a different linker flanking the 34-bpsequence. This introduces an adenine that matches the P1 sequence at P14356(Fig. 1). However, this base is not required for partition (c.f. pALA415 andpALA455). Plasmid pALA419 is identical to pALA415 except for the single-base mutation. Other plasmids are variants of pALA455 with mutations withinthe 34-bp sequence (see Fig. 2).

b Strain CC241 is polA and contains the pSC101-based plasmid pALA283supplying the two P1 Par proteins.

c Strain CC240 is equivalent to strain CC241 but contains the vector(pLG338) with no Par open reading frames.

together the HindIII-DraI fragment of pALA270 and adouble-stranded oligonucleotide containing P1 sequencefrom Dral to the G at bp P1-4378 (bp 22, Fig. 1) and acomplementary BamHI end.

Partition tests. Tests were as described in Martin et al.(12), and the method is outlined in the Results section below.

Incompatibility tests. All growth was carried out at 30°C.Strain N100 or RW1840 was lysogenized with X-Pl:5RCm, aX-mini-P1 phage-plasmid chimera that lysogenizes its hostonly as a stable extrachromosomal P1-driven plasmid (7).The resulting strain was transformed with the appropriatehigh-copy-number plasmid to be tested, selecting only forthe incoming plasmid on agar containing the appropriateantibiotic. Six transformants were restreaked for singlecolonies on the same medium, and approximately ten of theresulting colonies from each streak were screened for thepresence or absence of chloramphenicol resistance, as pre-viously described (7). The percentage of chloramphenicol-resistant colonies is taken as the proportion of cells retainingthe X-P1:5RCm plasmid in the approximately 25 generationsof growth required to form the initial-transformant colonies(7).

RESULTS

Construction of mutations in the parS sequence. We havedeveloped an assay for partition site activity in a strain thatprovides the two P1 Par proteins in trans and used it to showthat a 35-bp segment of P1 DNA contains sufficient cis-actinginformation to direct the active segregation of plasmids todaughter cells (12). A 34-bp sequence (Fig. 1) lacking the firstadenine base is equally effective (Table 1). The 34-bp seg-ment was introduced into the vector pALA136 which con-tains both a P1 and pBR322 origin of replication. Themulticopy origin facilitated construction and manipulation.Partition activity was measured by determining plasmidstability in a polA strain where the high-copy-numberpBR322 origin is nonfunctional, and a low copy number ismaintained (12).We mutagenized the 34-bp sequence by incorporating

VOL. 173, 1991

on February 7, 2018 by guest

http://jb.asm.org/

Dow

nloaded from

3632 MARTIN ET AL.

parS

PARTITION + T T C

l : :: :bp1 10. *.2 0 .30

ATTTCAAGQTQA#ATPGCCACGATTTCACCTTGG

PARTITION - AC A A GTGGFIG. 2. Mutations in the 34-bp parS region. The 34-bp region

sufficient for parS activity is shown. Half-arrows mark the 13-bppalindrome. Alternative bases above the sequence are single-basemutations that gave no detectable change in tests for partition siteactivity relative to the wild-type 34-bp sequence (>25% retention ofthe test plasmid in 25 generations of unselected growth when ParAand ParB were supplied in trans). The alternative bases below thesequence are single-base-change mutations that blocked partitionsite activity (<2% retention in 25 generations). All of the constructsgave <2% retention in control strains lacking P1 ParA and ParBproteins.

random bases into a synthetic oligonucleotide according tothe method of Derbyshire et al. (8). After construction, thesequence of each mutant clone was determined, and theresulting plasmid was assayed for partition. Sixty-two plas-mids were screened: 19 contained single mutations, 23 werewild type, and the other 20 had multiple changes. Anadditional mutant plasmid (pALA419, A-+T,14) was con-structed by using a defined oligonucleotide (12).

Fig. 2 shows the results of assays for partition function ofeach of the different single mutations isolated. Severalmutants showed no significant reduction in stability levelsrelative to the appropriate wild-type control. (It must benoted, however, that the polA stability assay is not sensitiveenough to determine small decreases in stability.) The othermutants were very unstable. Loss rates were comparable tothose obtained with the vector lacking all partition se-quences. Examination of Fig. 2 shows that silent changes

pALA455 AlCACMGGTGAAATCGCC(WT)

pALA456 AMCAAGGTGAAATCGCC(A23-34)

pALA439(CCA-iTGG, 18-20)

rALA430(CCA 4AAT, 18-20)

pALA453(AT ,2)

pALA442rI,114')

DAC

OAC

can be made in the right half of the 13-bp inverted repeat, butmost changes in the left half are not tolerated. The quanti-tative data from representative mutants that were used forfurther study are given in Table 1.

Refinement of the boundaries of parS. The data in Fig. 2suggest that the boundary of the essential sequences forparSlies close to the left boundary of the 34-bp sequence, but thatthe right portion of the sequence might be dispensable forpartition. We tested these predictions directly by deletionanalysis (Fig. 3). A 22-bp oligonucleotide lacking the right-most 12 bp of the sequence (Fig. 3, A23-34) was introducedinto the pALA136 vector. It stabilized the fragment toapproximately the same degree as the 34-bp sequence whenthe Par proteins were supplied in trans (Fig. 3). Deletion intothe left end of the 34-bp sequence abolished partition (mu-tation A1-4, Fig. 3). The mutation AT,2 (Fig. 3) can beregarded as a 2-bp substitution of the left end of the 34-bpsequence (substituting CA for the AT of the first two bases).This is also partition defective (Fig. 3), showing that the leftboundary of the partition site is at bp 1 or 2.Our initial data suggested that the sequences constituting

bp 18-21 might not be required for partition, since twosingle-base changes there have no obvious effect (Fig. 2).However, results obtained with two mutants with multiplechanges in this sequence show that it does contain informa-tion important for partition, even though some changes aretolerated. Both mutants (CCA--TGG,18-20 and CCA-+AA&T,18-20) are partition defective (Fig. 3). Since the basesbeyond bp 22 are dispensable for partition but substitution inbp 18-20 is not tolerated, the right boundary of the minimalparS site must lie in the interval bp 19-22 of the 34-bpsequence.

Partition-mediated incompatibility. The region containingthe parS site acts as a determinant for partition-mediatedplasmid incompatibility. Our standard test for this effect is asfollows: a small fragment including the region is cloned intothe high-copy-number vector pBR322. This is introduced bytransformation into a cell containing a resident P1 miniplas-mid. Loss of the resident plasmid is rapid and can be scored

% plasmid retainedCC241 CC240

(ParA+,ParB+)

:GA1TTCACCTTGG 58 2

:G 43 <2

<2 <2

<2 <2

<2 <2

<2 <2

FIG. 3. The partition site activity of deletions and multiple mutations that define the boundaries of the minimal parS site. Half-arrows markthe 13-bp palindrome or the remaining portions of it. Only the P1 bases are shown, with those in bold representing bases altered from thewild-type sequence. The A symbols indicate deletions. All P1 sequences were made from synthetic oligonucleotides (see Materials andMethods). CC241 is the polA strain that supplies P1 ParA and ParB proteins. CC240 is an isogenic control strain lacking ParA and ParB.

v_

A1TTCAAGGTGAAATCGTGGCGA1TTCACCTTGG

%, A AATTTCAAGGTGAAATCGAATCGATTTCACC1TGG

_______TGAMTG leAATCAAGGTGAAATCGCCACGA1TTCACCTTGG

CAAGGTGAAATCGCCACGATTTCACCTTGG

J. BACTERIOL.

on February 7, 2018 by guest

http://jb.asm.org/

Dow

nloaded from

P1 PLASMID PARTITION SITE 3633

TABLE 2. Incompatibility exerted by incB region constructs'

% Loss of mini-Pl in 25Plasmid incB region generations from strain: Phenotype

N100 (IHF+) RW1840 (IHF-)

pALA464 None 0.4 2 IncBpALA412 34-bp wild type 0.3 99 IncBdpALA465 84-bp wild type 98 96 IncB+

pALA411 Intact >98 >98pALA463 A22-34 9 >98pALA271 A1-34 <2 <2

a Incompatibility is expressed as the loss of the resident X-P1:5RCmplasmid induced by transformation of the cells with pBR322 derivativescarrying par region fragments. Upper section: the results in the upper panelparallel those obtained by Funnell (10), with plasmids carrying different butcomparable fragments. The sequences of the 84- and 34-bp regions are shownin Fig. 1. Plasmids pALA465 and pALA412 are incB+ and incBd respectively.They serve to illustrate the IncB+ and IncBd phenotypes: in IHF+ strains,incoming incB+plasmids compete with the incB+ resident plasmid but incBdplasmids do not (7); whereas in IHF- strains, all plasmids with active partitionsites (including the resident plasmid) behave as though they were incBd andtherefore compete with each other (7, 10). Lower section: the plasmidpALA411 is derived from pBR322 and carries the intact P1 par region,including the open reading frames for the Par proteins and the incB+ regionending at the Styl site at P1-4390 (bp 34 of the 34-bp parS sequence). PlasmidpALA463 is identical to pALA411 except for the deletion of bp 22-34. PlasmidpALA271 is comparable to pALA411 but lacks all bases to the right of theDraI site at P1-4338, including all of the 34-bp sequence (P1-4357 to P1-4390,Fig. 1).

by analysis of the population after a number of generationsof unselected growth. In these tests, it is always the residentP1 miniplasmid that is lost, because the incoming plasmid isselected for and is maintained at high copy number.The upper panel of Table 2 shows typical results obtained

for wild-type fragments in IHF-proficient and -deficientstrains containing a resident X-P1:5R plasmid. When the84-bp incB+ region (Fig. 1) was tested, the resident P1plasmid was displaced from both strains (pALA465, Table2). However, when the incoming test plasmid contained the34-bp partition site (Fig. 1), displacement was seen only inthe IHF-deficient strain (pALA412, Table 2). This testserves to define the IncB+ and IncBd phenotypes for thepurposes of this paper (see footnote a to Table 2) and wasused to screen point mutations.We tested the incompatibility properties of three of the

point mutations in the 34-bp fragment (Table 3). The 34-bp

TABLE 3. Incompatibility exerted by mutants inthe 34-bp context

% Loss of resident mini-Pl inPlasmida incB region 25 generations from strain:

N100 (IHF+) RW1840 (IHF-)

pALA412 34-bp wild-type <2 >98pALA498 34-bp, T-*C,4 <2 31pALA417 34-bp, A- T,14 <2 37pALA499 34-bp, C-*G,16 <2 16pALA425 34-bp, CCA-*AAT,18-20 <2 <2pALA438 34-bp, CCA-*TGG,18-20 <2 <2pALA431 34-bp, AT,2 <2 <2

a The mutant plasmids tested are derived from pBR322 and are comparableto pALA412 but have the mutations shown.

b The resident plasmid was the P1 miniplasmid X-P1:5RCm. Its loss wasmonitored after 25 generations of unselected growth after transformation withthe plasmids tested.

mutant fragments were recloned into pBR322, eliminatingthe P1 replication sequences. The wild-type 34-bp sequencein these constructs scored as IncBd, inducing displacementof the mini-Pl from the IHF-defective but not the wild-typestrain. However, all three single-base-change mutationscaused significant defects in IncBd, with that of the mutationC->G,16 being the most severe (Table 3). Although someincompatibility is still seen in the tests with the mutants, weregard the phenotypic changes of all three mutants assignificant. The competition is only partially effective at thehigh-copy-number ratio (ca. 20:1) of the competing mutantsequence to the wild-type mini-Pl, whereas a wild-typeincBd sequence is an effective competitor at a copy numberratio of 1:1 to the resident target plasmid (12). A single-base-deletion mutation, AT,2, eliminated the IncBd phenotypealtogether (Table 3). Because all of the four mutations testedaffect the IncBd phenotype and also eliminate partition siteactivity (Table 2), we conclude that the two phenotypes areintimately linked effects.The role of the sequences to the right of the 22-bp parS site.

We have shown that at least 12 bases at the right boundaryof the 34-bp sequence are not needed for parS function.However, the fact that they bind the ParB protein (6, 7)suggests that they have some role in partition. We con-structed a variant of the wild-type par region that lacks thesebases and tested it for IncB+ incompatibility. Unlike thewild-type sequence (pALA411) that scores as IncB+, thevariant (pALA463) fails to exert incompatibility in a wild-type strain. It does exert incompatibility in the IHF- strain,though, and is therefore IncBd (Table 2). Further deletionfrom the right end that removes the 22-bp parS site elimi-nates all incompatibility (pALA271, Table 2). We concludethat, like the sequences that lie to the left of the 22-bpminimal partition site (7), the 12 additional bases to its rightare not essential for partition or for the IncBd phenotype butare part of the information required for the wild-type IncB+specificity.The limits of the parS and IncBd determinants. The bound-

aries of the determinant for the IncBd phenotype weremapped. Bases to the right of bp 22 are not needed for IncBd(pALA463, Table 2), and the two multiple mutations(CCA-*TGG,18-20 and CCA-*AAT,18-20) that block parti-tion by modifying the right boundary of the 22-bp sequenceabolish the IncBd phenotype (Table 3). The partition-defec-tive mutation AT,2, which effectively replaces the first twobases of the 22-bp sequence, also blocks IncBd (Table 3).Thus, the boundaries for both the partition site and IncBd arethe same and lie in the intervals bp 1-2 and bp 18-22.

DISCUSSION

We have defined a 22-bp P1 sequence that is required andsufficient to direct the segregation of plasmids to daughtercells when the P1 Par proteins are supplied. It contains onlythe left half of a 13-bp perfect palindrome and, because basesflanking this half are essential, the right half cannot substi-tute. No readily measurable differences in partition are seenbetween this 22-bp sequence and larger (up to 174 bp) P1fragments (12) that contain the 22-bp sequence. When the P1Par proteins are supplied in trans, the maintenance of theseartificial constructs is about 10-fold better than expected forrandom distribution (12). This underestimates partition effi-ciency because some of the loss events are due to theformation of multimers in the rec+ strain used in the assayrather than to failure of the partition mechanism itself (4).Although partition promoted by the 22-bp fragment is less

VOL. 173, 1991

on February 7, 2018 by guest

http://jb.asm.org/

Dow

nloaded from

3634 MARTIN ET AL.

efficient than observed with the wild-type par region in itsnatural context (12), this small site must nevertheless con-tain sufficient information for recognition and interactionwith the putative host machinery responsible for selectivemovement of the copies.What do the 22 bp do? They bind ParB (6, 7). The leftmost

15 bases (bp 1-15) consist of an inverted repeat of theproposed ParB-binding consensus sequence, 5'ATTTCAC(the first repeat has an A in the last base). Both elements ofthis repeat motif are strongly protected by ParB binding inDNase protection (footprinting) assays (7). All but one of thesingle base changes in these 15 bp cause the site to bepartition defective. This suggests that the formation oractivity of a specific ParB-DNA complex involving bothParB-binding sites is essential for partition. The remaining 7bp (bp 16-22) contain some information important for parti-tion, since changes there can eliminate partition site activity.We are currently investigating whether ParA or some hostprotein might bind there. These bases overlap the sequenceTCGCCA (bp 15-20). It might be significant that this motif isalso present at the extreme left boundary of the incB+sequences (P1-4303-4308, Fig. 1), where it is required for theIncB+ phenotype (7).

Partition and IncBd incompatibility require the same 22bp, and point mutations within those bases affect both. Thebases bind ParB. Perhaps then, IncBd incompatibility issimply due to titration of available ParB protein by theParB-binding sites of the competing plasmids. However, thishypothesis is not favored by the data. The plasmid pALA463carries a complete parA-parB operon in addition to an IncBdpartition site (Table 2). It exerts a normal IncBd incompati-bility phenotype (Table 2). Since each of the incoming copiesof the plasmid brings its own capacity to synthesize ParB, itis unlikely that displacement of the resident plasmid is due toParB titration. The observations that ParB synthesis isautoregulated (9) and that incBd plasmids do not competewith the wild-type P1 plasmids, even when in large copynumber excess (12), also provide difficulties for a simpletitration model. Rather, it seems likely that incompatibilityreflects the mechanism by which pairs of plasmids arechosen for partition (3, 13). Introduction of competing plas-mids carrying the same partition site disrupts the choice of"correct" pairs by participating directly in the process. Inthis case, mutations that affect both partition and incompat-ibility do so because they disrupt information required forpair recognition.We show that, like the sequences to the left of the 22-bp

region, the region to its right is part of the informationrequired for the IncB+ phenotype. Thus, the core partitionsite is sandwiched between two blocks of information thatare not essential for partition but are required for wild-typeincompatibility. The left block contains a ParB-binding siteand a binding site for the host IHF protein (7, 10). The newly

defined 12-bp right block also binds ParB (7) and contains theParB-binding consensus sequence 5'ATTTCAC (Fig. 1). Theinvolvement of these separated ParB-binding regions inIncB' incompatibility is consistent with our previouslyproposed model for the region, in which IncB+ is determinedby a folded structure in which the distal ParB-bindingregions are held together by IHF and ParB protein (7).Disruption by mutation (or the absence of IHF) would givean unfolded structure that results in the IncBd phenotype.

ACKNOWLEDGMENT

This research was sponsored by the National Cancer Institute,under contract no. NO1-CO-74101 with ABL.

REFERENCES1. Abeles, A. L., S. A. Friedman, and S. J. Austin. 1985. Partition

of unit-copy miniplasmids to daughter cells. III. The DNAsequence and functional organization of the P1 partition region.J. Mol. Biol. 185:261-272.

2. Austin, S., and A. Abeles. 1983. Partition of unit-copy miniplas-mids to daughter cells. I. P1 and F miniplasmids containdiscrete, interchangeable sequences sufficient to promote equi-partition. J. Mol. Biol. 169:353-372.

3. Austin, S., and A. Abeles. 1983. Partition of unit-copy miniplas-mids to daughter cells. II. The partition region of miniplasmidP1 encodes an essential protein and a centromere-like site atwhich it acts. J. Mol. Biol. 169:373-387.

4. Austin, S., M. Ziese, and N. Sternberg. 1981. A novel role forsite-specific recombination in maintenance of bacterial repli-cons. Cell 25:729-736.

5. Das, A., D. Court, and S. Adhya. 1976. Isolation and character-ization of conditional lethal mutants of Escherichia coli defec-tive in transcription termination factor rho. Proc. Natl. Acad.Sci. USA 73:1959-1963.

6. Davis, M. A., and S. J. Austin. 1988. Recognition of the P1plasmid centromere analog involves binding of the ParB protein.EMBO J. 7:1881-1888.

7. Davis, M. A., K. A. Martin, and S. J. Austin. 1990. Specificityswitching of the P1 plasmid centromere-like site. EMBO J.9:991-998.

8. Derbyshire, K. M., J. J. Salvo, and N. D. Grindley. 1986. Asimple and efficient procedure for saturation mutagenesis usingmixed oligonucleotides. Gene 46:145-152.

9. Friedman, S. A., and S. J. Austin. 1988. The P1 plasmid-partition system synthesizes two essential proteins from anautoregulated operon. Plasmid 19:103-112.

10. Funnell, B. E. 1988. Participation of Escherichia coli integrationhost factor in the P1 plasmid partition system. Proc. Natl. Acad.Sci. USA 85:6657-6661.

11. Kelley, W. S. 1980. Mapping of the polA locus of Escherichiacoli K12: genetic fine structure of the cistron. Genetics 95:15-38.

12. Martin, K. A., S. A. Friedman, and S. J. Austin. 1987. Partitionsite of the P1 plasmid. Proc. Natl. Acad. Sci. USA 84:8544-8547.

13. Novick, R. P., and F. C. Hoppensteadt. 1978. On plasmidincompatibility. Plasmid 1:421-434.

J. BACTERIOL.

on February 7, 2018 by guest

http://jb.asm.org/

Dow

nloaded from