Proc. Natl. Acad. Sci. USAVol. 84, pp. 4645-4649, July 1987Medical Sciences

Transposition of bacteriophage Mu in the legionnairesdisease bacterium

(Legionella pneumophila/intracellular parasite/lac gene fusions)

CLIFFORD S. MINTZ AND HOWARD A. SHUMAN*Department of Microbiology, The College of Physicians and Surgeons, Columbia University, 701 West 168th Street, New York, NY 10032

Communicated by Harold S. Ginsberg, March 25, 1987

ABSTRACT Legionnaires disease is an acute respiratorydisease that is often fatal for immunocompromised patients.The causative agent of this disease, Legionella pneumophila, isa Gram-negative bacterium that is present in a variety ofaquatic environments. L. pneumophila is a facultative intra-cellular parasite; it grows within human phagocytic cells andeventually causes their destruction. In contrast to many otherintracellular parasites, L. pneumophila is a Gram-negativebacterium that can be grown in standard microbiologicalculture medium. To determine the factors that enable thisorganism to enter, survive, and multiply within human mono-nuclear phagocytes, we chose bacteriophage Mu, a powerfulgenetic tool that transposes within the host cell genome, togenerate insertion mutations and gene fusions in the Legionellagenome. Certain derivatives ofMu are able to generate fusionsbetween target genes and the lac operon from Escherichia coli.We have determined that although Mu is unable to attach to L.pneumophila or complete its life cycle within Legionella, it doestranspose within the Legionella genome. Transposition wasdetected with a mini-Mu phage that carries the lac operon ofE. coli.

The legionnaires disease bacterium Legionella pneumophilais a facultative intracellular parasite (1). This organism is ofparticular interest because it shares many characteristicswith other intracellular parasites such as Leishmania,Toxoplasma, and Mycobacteria. It is phagocytosed efficient-ly by human mononuclear leukocytes in the absence ofopsonizing antibody by a coiling mechanism (2). Legionellaevade the antimicrobial defenses of monocytes and macro-phages by inhibiting phagosome-lysosome fusion (3). Final-ly, Legionella multiply exponentially within a specializedvacuole and eventually kill the monocyte (1). In contrast tothe intracellular parasites mentioned above, L. pneumophilais a Gram-negative bacterium that can be cultivated in thelaboratory.To determine which Legionella functions are required for

killing monocytes, we have chosen a genetic approach. Sincebacteriophage Mu has been a powerful tool in other Gram-negative bacterial systems, we selected Mu and Mu deriva-tives that contain the lac operon genes to study Legionella.Bacteriophage Mu grows vegetatively in many Gram-nega-tive bacteria (4, 5). Early in replication, the Mu genometransposes essentially at random to many sites in the host cellDNA (6), thus generating insertion mutations and rearranginggenes. Mini-Mu phages that carry the lac genes of Esche-richia coli can result in fusions of the lac operon to genes ofthe host cell genome (7).We present evidence that Mu is able to transpose within L.

pneumophila. We detected transposition of mini-Mu lacphages from a broad host range RK2-related plasmid to

various sites in the Legionella genome, using Southern blots.Although transposition was identified easily, Mu phageparticles could not be detected after induction. Our resultssuggest that Mu can be used to generate insertion mutationsand gene fusions within the Legionella genome. TheseLegionella mutants could then be used to determine whichfunctions are required for Legionella to survive and grow inphagocytic cells.

Despite considerable interest in the pathogenesis and cellbiology of the interaction between L. pneumophila andphagocytic cells, little progress has been made to geneticallyanalyze this organism. Genetic studies have been limited todemonstrating transfer of broad host range plasmids and Tn5into Legionella and to cloning Legionella DNA sequencesthat encode antigens (8-10). Thus Mu and mini-Mu phageswill provide additional tools for the genetic analysis of thisorganism.

MATERIALS AND METHODS

Bacterial Strains, Phages, and Plasmids. All strains of L.pneumophila, E. coli K-12, phage, and plasmids used arelisted in Table 1. L. pneumophila, Philadelphia-1, grown inhen eggs was obtained from Marcus Horwitz and JoelleGabay (Rockefeller University). Bacteria were cultured ei-ther on solid medium or liquid medium (see below) with nodetectable loss of the ability to kill peripheral blood mono-cytes in vitro (J. Gabay, personal communication).

Bacterial Media. E. coli media have been described in detail(15). L. pneumophila were cultured routinely on Aces [N-(2-acetamido)-2-aminoethanesulfonic acid]-buffered char-coal/yeast extract agar plates (ABCYE) when solid mediumwas used or in albumin/yeast extract broth (AYE) whenliquid medium was used (1). Antibiotics were present at thefollowing concentrations when required: streptomycin at 50,ug/ml, kanamycin at 25 Ag/ml, chloramphenicol at 25 ,ug/mlfor E. coli medium and at 5 ,ug/ml for Legionella medium.

Conjugal Transfer of Plasmids Between E. coli and L.pneumophila. Spot matings between E. coli CM 1000 carryingthe indicated plasmids and L. pneumophila, strain Philadel-phia-1, CS1, were performed on ABCYE plates. Both donorand recipient were grown to early exponential phase prior tomating. The recipient to donor cell ratio was 10:1. After 4 hr,the mating mixture was collected in sterile M63 salts (15), andaliquots were spread on ABCYE plates containing strepto-mycin and kanamycin. For some experiments 0.2 ml of ahigh-titer stock of bacteriophage T4 was spread on the platesto ensure complete counterselection of the E. coli. Thematings and selections were carried out at the indicatedtemperatures. Transfer frequencies are reported per 5 x 108recipient cells. All values represent the average of at leastthree experiments.

Abbreviation: X-Gal, 5-bromo-4-chloro-3-indolyl /3-D-galactoside.*To whom reprint requests should be addressed.

4645

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

4646 Medical Sciences: Mintz and Shuman

Table 1. Bacterial strains, phages, and plasmids

Relevant genotypeStrain or properties Ref. or origin

Bacterial strainsE. coliCM 1000

RR 1L. pneumophila

Philadelphia-1CS 1

CS 24CS 26CS 103CS 122

PhagesMu cts62

MudI 1681

MudI ctsB::Tn9(Cmr),(Ampr,lac)

Mu cts62pf7701-A445-3

PlasmidsRK2pRK24

pRK212pRK24.1pRK24.2

pRK24.3

F- A(lac,pro)XIII,rpsE, (Mu')

F- hsd2O (rm-)

wild typerpsL

rpsL, RK2rpsL, pRK212rpsL, pRK24.3rpsL, pRK24.2

Temperature-sensitive repressor

Mini-Mu that carriesthe lac operon ofE. coli and the neo(Kmr) gene of TnS,Mu cts62

Mini-Mu that carriesthe lac operon andthe bla gene andhas an insertion ofTn9 in the Mu Bgene, Mu cts62

Carries a deletion ofthe kil gene.

Contains the neo(Kmr) gene of TnS.

bla, neo, tetRK2 neo::[HindIII

fragment] (Kms)RK2::Mu cts62pRK24::MudI 1681pRK24: :MudI

cts62B::Tn9 (Cmr),(Ampr lac)

pRK24::Mu cts62pf7701-A445-3

This laboratory

Ref. 11

Spontaneous Smrderivative ofPhiladelphia-1

M. Howe

Ref. 7

Ref. 12

M. Howe

Ref. 13Ref. 12

Ref. 14

DNA Hybridization on Nitrocellulose Filters. Total genomicDNA was prepared from 50 ml of AYE for Legionella or LB(Luria broth) forE. coli cultures. DNA was cleaved with BglII using the buffer conditions specified by the supplier(Boehringer Mannheim). Cleaved DNA (3 pig) was electro-phoresed on 0.6% agarose gels and then blotted to nitrocel-lulose filters (Schleicher & Schuell). DNA sequences that arehomologous to phage MudI 1681 were detected by hybrid-ization to 32P-labeled plasmid DNA that contains a copy ofphage MudII 1681 inserted in pBR322 (7). This phage issimilar to MudI 1681 and differs only in sequences on the 5'side of the lacZ gene (7). The radioactive probe was preparedby nick-translation with DNA polymerase I, two nucleotide[a-32P]triphosphates and two nonradioactive nucleotide tri-phosphates. The DNA fragments that hybridized to theradioactive probe were detected by autoradiography withKodak X-Omat film.

f3-Galactosidase Assays. B-Galactosidase was measured inwhole cells as described by Miller (15). E. coli strains were

grown in LB, and Legionella strains were grown in AYEbroth. Cells were centrifuged and resuspended in M63 saltsprior to assays. Values are the average value from threeseparate assays (SEM l10%). f3-Galactosidase activity ofLegionella colonies was visualized by replica plating thecolonies to Whatman No. 1 filter paper, incubating the filtersin a solution containing 50 mM Tris HCl (pH 8.0), 5 mMNa2EDTA, and lysozyme at 100 ,ug/ml for 5 min. The filterswere then transferred to a freezer for at least 1 hr. At the endof this time the filters were immersed in M63 salts containingX-Gal (5-bromo-4-chloro-3-indolyl B-D-galactoside at 50pug/ml). The filters were removed from the X-Gal solutionafter 10 min to 1 hr.

RESULTSInsertion ofMu DNA into Broad Host Range Plasmid RK2.

Because L. pneumophila was not killed by Mu cts62 infectioneven when high multiplicities of infection (moi = 100) wereused, we examined whether or not Mu adsorbed to thesebacteria. By measuring the ability of a concentrated suspen-sion of Legionella to decrease the titer of a Mu cts62 stock,we confirmed that phage Mu did not adsorb to Legionella.We incubated a suspension of Legionella (109 cells per ml)with a lysate of Mu cts62. At various times, samples of themixture were centrifuged to remove the bacteria and anyadsorbed Mu particles. The Mu phage remaining in thesupernatant were titrated on a standard E. coli K-12 host. Wedid not detect any decrease in the titer of the unadsorbed Muwith Legionella, even though control experiments with E.coli K-12 showed 90-99% decreases in the titer of unad-sorbed phage after 10 min (data not shown).To introduce the Mu genome into L. pneumophila, we used

RK2, a member of the IncP group of broad host rangeplasmids that are transferred into and stably maintained inLegionella (ref. 8 and C.S.M., unpublished observations).We mated a streptomycin-resistant mutant of L. pneumophi-la, strain Philadelphia-1, with E. coli donor strains thatcontain the IncP plasmid RK2 (13) or the RK2::Mucts62derivative pRK212 (14). The frequencies with which weobtained transfer of the plasmids were 2 x 10-6 for RK2 and8 x 10-7 for pRK212 at 300C. At 370C there was a dramaticdecrease in the frequency with which pRK212 was trans-ferred to Legionella. Exconjugants were obtained at a fre-quency <2 x 10-8 with pRK212, while the frequency oftransfer for RK2 was the same at 30°C and 370C (2.2 x 10-6).These results are consistent with the hypothesis that, afterintroduction into Legionella, the Mu genome reduces theviability of the bacteria. At the higher temperature, the muchlower frequencies observed with pRK212 carrying the Mucts62 mutation resulted from the lack of active Mu repressor.At 30°C, the slight decrease in the frequency of transfer isprobably due to zygotic induction.

Induction of Mu DNA Replication Is Lethal. The survival ofvarious Legionella strains that contained either RK2 or RK2derivatives at 300C and 42°C is shown in Table 2. At 420CLegionella strain CS 26, which contains Mu cts62, has a muchlower plating efficiency than the control strain (CS 24)confirming that the induction of Mu functions is lethal toLegionella.There are two distinct Mu functions that can result in the

death of the host cell: one is MuB-dependent DNA replica-tion and transposition (16), and the other is the kil genefunction (17). To determine which of these two functions wasresponsible for the killing of Legionella following thermoin-duction, we measured the viability of Legionella strainscarrying plasmids that contain Mu genomes defective in thesetwo functions. One of these, MudI B::Tn9(Cm9)(Amp,lac)(pRK24.2) (12) is defective in DNA replication and transpo-sition due to a Tn9 insertion in the Mu B gene. This phage

Proc. Natl. Acad. Sci. USA 84 (1987)

Proc. Natl. Acad. Sci. USA 84 (1987) 4647

Table 2. Survival of Mu-containing Legionella

Legionella survival,cfu/ml Relative

Strain Plasmid 300C 420C efficiencyCS 24 RK2 3.5 x 107 2.0 x 107 0.57CS 26 pRK212 5.0 x 108 1.3 x 104 2.6 x 10-5CS 103 pRK24.3 1.3 x 108 8.3 x 104 6.4 x 10-4CS 122 pRK24.2 3.2 x 108 4.3 x 108 1.34

The indicated plasmids were introduced into L. pneumophila,Philadelphia-1, at 30'C. The plasmid-containing strains were purifiedtwice on selective medium and grown overnight at 30'C in AYEbroth. Dilutions of this culture were spread on ABCYE platescontaining kanamycin or chloramphenicol and incubated either at30'C or 420C. Colonies were counted, and the number of colony-forming units (cfu) are reported per ml. The relative efficiency is thenumber of cfu at 420C divided by the number of cfu at 30'C.

genome probably expresses genes downstream from Mu B,including the kil gene at a low level due to polarity of the Tn9insertion. The other phage genome Mu cts62pf7701-A445-3(pRK24.3) contains an active Mu B gene but contains adeletion that removes all ofthe kil gene. The bacteria (CS 103)that contain the Mu kil- phage genome show the samesurvival as those (CS 26) that contain the Mu kil+ phagegenome, while bacteria (CS 122) that contain the Mud B: :Tn9phage genome show no loss of viability (Table 2). Weconclude that the lethal events associated with Mu DNAreplication and transposition are responsible for the decreasein viability at 420C.

Detection of Mu Transposition with Iac Gene Fusions. Wenext investigated whether Mu transposition could be detect-ed. Transposition occurs very early in the Mu life cycle andrequires only the products of the Mu A and Mu B genes aswell as the two ends of the Mu genome (18, 19). To determineif transposition of the Mu genome occurred in Legionella, weused a mini-Mu phage, MudI 1681, that carries a selectabledrug marker (neo-Km9 and a copy of the E. coli lacZYAgenes without their normal promoter. Transcription of the lacgenes depends on upstream transcription signals (7).

First, we inserted this phage into a derivative of RK2,pRK24, that does not confer resistance to kanamycin (20).We intentionally selected a plasmid, pRK24.1, with a lowlevel of lac operon expression and transferred it from an E.coli donor into a Legionella recipient at 30'C. The exconju-gants were selected on ABCYE agar plates that containedkanamycin and streptomycin. To detect transposition, welooked for colonies of Legionella that expressed higher levelsof /B-galactosidase than the original parent E. coli strain.These would presumably result from transposition of theMudI 1681 to a different site that contained a strongertranscriptional signal than the original site of insertion in thepRK24.1 plasmid.

Since the commonly used chromogenic /3-galactoside,X-Gal, does not produce a blue color in the presence of thehigh concentrations of cysteine used in ABCYE media, weexamined the Lac phenotype of the exconjugants by trans-ferring the colonies from the ABCYE/kanamycin/strepto-mycin agar plates to Whatman filter paper, lysing the bacteriawith lysozyme at 100 Ag/ml and 5 mM EDTA, and incubatingthe filters in a X-Gal solution. Most of the colonies were verypale blue. However, we found dark blue colonies at afrequency of about 10% among the exconjugants, and manyof the pale blue colonies contained darker blue sectors (Fig.1). Since the bacteria with increased lac gene expressionwere found among the exconjugants, we believe that trans-position could have occurred prior to the establishment ofMulysogeny in Legionella.

Initially, four dark blue and three light blue colonies from asingle mating were purified and used as donors to transfer the

A ..B

FIG. 1. Detection of transposition with mini-Mu lac phages:Plasmids pRK24.1 and pRK24.2 were introduced into L. pneumo-phila strain CS 1 from E. coli CM 1000 by conjugation. Theexconjugants were transferred to Whatman filter paper, and the3-galactosidase activity of the colonies was detected. (A) PlasmidpRK24.1. (B) Plasmid pRK24.2.

pRK24.1 plasmid back to E. coli. We found that among these,all transferred bla, tet, and neo to E. coli CM 1000. The Lacphenotype of the E. coli strains that had acquired the plasmidsfrom Legionella was light blue and indistinguishable from theoriginal donor E. coli strain. We performed /3-galactosidaseassays to confirm the observations (Table 3). These resultsshow that the Lac phenotype of the darker blue colonies ofLegionella was not due to an alteration that increases theexpression of the lac genes on the pRK24.1 plasmid in E. coli.The plasmids were then retransferred back to Legionella asecond time, and the Lac phenotype of the exconjugants wasevaluated with X-Gal. All of the plasmids yielded the sameproportion of light blue colonies and dark blue colonies orsectors as in the original mating experiment. This shows that theexconjugants that were originally dark blue in Legionella werenot the result of mutations that increased the level of lac geneexpression from the original pRK24.1::MudI 1681 insertion

Table 3. f3-Galactosidase activity of MudI 1681-containing strains

p8-Galase,Strain Plasmid units/ml

CM 1000 pRK24.1 6.0CS 1 pRK24.1

Light blue colony 7 4.3Light blue colony 25 4.2Light blue colony 26 8.3Dark blue colony 4 62.5Dark blue colony 6 62.5Dark blue colony 17 70.5Dark blue colony 18 34.5

CM 1000 pRK24.1From light blue colony 7 4.8From light blue colony 25 4.5From light blue colony 26 4.7From dark blue colony 4 5.0From dark blue colony 6 4.2From dark blue colony 17 3.1From dark blue colony 18 4.6

Enzyme assays were performed as described (15). Strain CM 1000contains a deletion of the lac operon, and Legionella strain CS 1produces no endogenous 3-galactosidase (,3-Galase) and has no DNAthat hybridizes to a lac operon probe (data not shown). The dark blueand light blue colonies of Legionella were restreaked twice onABCYE plates containing kanamycin and streptomycin and werethen grown in AYE broth for the assays. The plasmids from thesewere mated back to CM 1000, and these strains were also assayed.The values are the average of three determinations (SEM s10%).

Medical Sciences: Mintz and Shuman

4648 Medical Sciences: Mintz and Shuman

either in E. coli or in L. pneumophila. When these matingexperiments were repeated with a plasmid that carried the MudIB::Tn9(Cm9(Amp,lac) phage genome that is defective in trans-position, dark blue colonies occurred at a much lowerfrequency(-10-s) (see Fig. 1). We conclude that the dark blue colonies arethe result of transposition of the MudI 1681 phage to theLegionella genome.

Physical Evidence for Transposition. To confirm that thephysical location of the MudI 1681 phage is different in thedarker blue Legionella exconjugants than in the parental E.coli strain, we performed Southern blot experiments. Totalgenomic DNA was isolated from a variety of dark blueexconjugants that were obtained from independent matings.The DNA was cleaved with restriction endonuclease Bgl II.The cleaved DNA was then resolved on agarose gels andexamined by Southern blotting and hybridization with 32P-labeled plasmid DNA that includes MudII 1681 lac se-quences. As seen in Fig. 2, in all cases the pattern offragments obtained from the dark blue colonies (lanes b-g,i-m, o-r, and t-y) is different than in the parental E. colidonor strain (lanes a, h, n, and s). In most cases two newjunction fragments are seen in addition to the fragmentsderived from pRK24. 1. In one case (lane d) no new junctionfragments were observed but the smaller plasmid junctionfragment has a reduced size. This may be due to a deletionthat resulted in increased expression of the lacZ gene.Southern blots done on several Legionella exconjugants thathad a light blue phenotype were identical to those of the E.coli donor strain (data not shown). These results confirm thatin the strains that produce larger amounts of P-galactosidase,there are new DNA sequences adjacent to the MudI 1681phage in addition to those in the original pR24::MudI 1681insertion. In addition there is considerable variety in the sizesof the junction fragments that were detected in the dark bluecolonies. This indicates that Mu transposed to different siteswithout detectable preference for a specific location in thedifferent exconjugants. The intensity of the new junctionfragments seems to be less than the intensity of the originaljunction fragments from pRK24. 1. This is consistent with thenew fragments having a lower copy number relative to the

plasmid-derived fragments as would be expected for chro-mosomal insertions.To determine if any rearrangement of plasmid sequences

had occurred, the blots shown in Fig. 2 were treated toremove the original radioactive probe and rehybridized witha probe that contains RK2 sequences and no Mu or lacsequences. Only the two larger junction fragments charac-teristic of pRK24.1 were observed in 19 out of the 20 strains.This indicates that in these strains we were unable to detectany alteration of plasmid sequences. In the one other strainthe 9-kilobase (kb) fragment was not present (correspondingto lane d of Fig. 2). In most cases, however, there was noevidence either for rearrangement of plasmid sequences ortheir integration into the Legionella genome. These resultsargue strongly that the Mu transposition system operatesefficiently in L. pneumophila.

Bacteriophage Mu Is Unable to Complete Its Life Cycle inLegionella. We next wanted to determine whether Mu is ableto complete its life cycle in Legionella and produce plaque-forming units. We grew cultures of a L. pneumophila strain,CS 26 that contains RK2::Mu cts62 at 30°C and then shiftedthe cultures to 42°C for various amounts of time between 2and 24 hr. At no time between 2 and 24 hr was there anydetectable lysis or decrease in the turbidity of the culture,although the number of colony-forming units decreased by afactor of more than 1000. At different times after thetemperature shift, we centrifuged the culture, resuspendedthe bacteria in buffer containing lysozyme at 100 ,g/ml and1 mM EDTA, and sonicated the bacteria for three 20-secperiods to release any phage trapped inside the bacteria. Theculture supernatant and the broken cells were cultured on E.coli K-12 RR1 (restriction minus) and E. coli C cells using theoptimum conditions for detecting Mu plaques (21). We werenot able to observe any plaques from the Legionella cultures,even though controls demonstrated that we were able todetect Mu with either the Mu G(+) or Mu G(-) host ranges(21). We conclude, therefore, that although bacteriophageMu is able to carry out part of its life cycle in Legionella,including transposition, it is unable to complete its fulldevelopmental program and assemble active phage particles.

a b c a e f g h i j k m n o p q r s t u v w x y

BgBg BgKmR trpE'lI 'KmR Ap oriVI Tc

r cAB

pRK24.1

IL neo A Y Z'

Bg Bg

FIG. 2. Physical evidence for transposition of MudI 1681. Plasmid pRK24.1 is a derivative of pRK24 that was made by isolating an insertionof MudI 1681 in pRK24 that was Tra+, Apr, and Tcr, as well as pale blue on X-Gal solid medium. Plasmid pRK24 contains a 5-kb insertion ofDNA that includes the trpE gene of E. coli at the HindIl site of RK2 (20). This insertion inactivates the kanamycin resistance gene of RK2.We do not know the location of the MudI 1681 on pRK24.1. Therefore, we did not know the predicted sizes of the junction fragments betweenMu and RK2-derived sequences. RK2 contains a single Bgl II site near the oriV and pRK24 acquired two additional Bgl II sites within the trpEgene. MudI 1681 has two internal Bgl II sites that are 1.5 kb apart. The map of MudI 1681 is adapted from ref. 7. Brackets indicate the endsof Mu. This phage contains the ISSOL and neo gene from TnS, but the ISSOR has been deleted. The lac operon genes lacZYA are present butwithout the lac promoter. Transcription of the lac genes depends on initiation at upstream sequences. Fragments indicated by a dot have sizesof 13 kb, 9 kb, or 1.5 kb. The two larger fragments arejunction fragments between RK2-derived sequences and MudI 1681 sequences. The 1.5-kbfragment is derived from within MudI 1681 phage DNA. Lanes a, h, n, and s contain DNA from the E. coli donor strain that contains pRK24.1.The remaining lanes contain DNA from dark blue Legionella exconjugants. The fourth faint band that is visible in the E. coli DNA is probablyderived from the Mu prophage in the chromosome of the donor strain.

Proc. Natl. Acad. Sci. USA 84 (1987)

4NOW--~ 4wb.!4:.,:, "', " I-W:~ MW* 0 ...... 40AWMI.Oft WAp..

A

-.0- - 1406W.40040

f::-U: 0 !"lp"

4. ...W

Proc. Natl. Acad. Sci. USA 84 (1987) 4649

DISCUSSIONOur results indicate that bacteriophage Mu is able to trans-pose to many sites within the Legionella genome. Wedetected transposition with mini-Mu phages that carry the lacoperon of E. coli. Thus it is possible to use Mu and relatedmini-Mu phages to generate insertion mutations and genefusions within the Legionella genome.

Although we were able to detect transposition easily, thebehavior of Mu and of mini-Mu in Legionella differs consid-erably from their behavior in other bacteria such as E. coliand Salmonella typhimurium. First, Mu does not adsorb toLegionella. Second, we were unable to detect Mu phageparticle development in Legionella. This is not true for otherspecies of bacteria; in all other cases where Mu cts has beenintroduced via infection or conjugation, phage particle pro-duction has been detected. The number of particles producedvaried among different types of bacteria from 103 plaque-forming units/ml to 1010 plaque-forming units/ml (22). Theconditions that we used to detect Mu particles after inductionshould have detected any phage produced above 10 plaque-forming units/ml. The mini-Mu lac fusions arise frequently inLegionella after conjugal transfer of the mini-Mu; Fig. 1shows that a large proportion of the exconjugants are eitherentirely dark blue or have dark blue sectors. We haveexamined the stability of the Lac phenotype in these darkblue colonies by restreaking single colonies twice and retest-ing the Lac phenotype of 20 resulting colonies. The Lacphenotype appears to be relatively stable; all 20 resultingcolonies were phenotypically identical to the original isolate.A potential disadvantage of the method that we describe

for isolating gene fusions in Legionella is the persistence ofthe pRK24.1 plasmid in the strains that contain the fusion.This plasmid could be eliminated either by introduction of anincompatible plasmid and screening for loss ofpRK24.1 or byusing a "suicide"-type vector that cannot replicate in Legion-ella for delivering the mini-Mu.We have developed a semi-defined solid medium for

culturing L. pneumophila that results in a high platingefficiency for wild-type Legionella. We have used thismedium to look for auxotrophs among kanamycin-resistantexconjugants that contain the MudI 1681 phage. Among 3084exconjugants we have detected 9 auxotrophs. This frequency(0.3%) is much higher than the frequency of spontaneousmutations and is approximately that observed in other orga-nisms containing the MudI 1681 phage. At the present timewe are not certain that the auxotrophic phenotypes are duedirectly to the insertion of the MudI 1681 phage genome intoparticular genes.

We thank David Figurski for introducing us to RK2 and manyhelpful discussions and criticisms. We are grateful to Martha Howe

for supplying us with Mu kil pf7701-A445-3, MudIB::Tn9,(Cmr)(Ampr,lac), and E. coli C as well as helpful suggestions on Mubiology. We thank Eduardo Groisman and Malcolm Casadaban forthe MudI lac 1681 and MudIl lac 1681 phages. We also acknowledgeMarcus Horwitz and Joelle Gabay for their help in initiating theLegionella work in our laboratory. C.S.M. acknowledges N. Rosen-berg for an introduction to the mysteries of the Southern blot. C.S.M.was supported by Public Health Service Training Grant AI 07161-08from the National Institutes of Health. H.A.S. is a Career Scientistof the Irma T. Hirschl Charitable Trust. This research was supportedby Biomedical Research Support Grant 02 S07RR05395 and Grant Al23549 from the National Institutes of Health.

1. Horwitz, M. A. & Silverstein, S. C. (1980) J. Clin. Invest. 66,441-450.

2. Horwitz, M. A. (1984) Cell 36, 27-32.3. Horwitz, M. A. (1983) J. Exp. Med. 158, 2108-2126.4. Schroeder, W. & van de Putte, P. (1974) Mol. Gen. Genet. 130,

99-104.5. van Vliet, F., Silva, B., van Montagu, M. & Schell, J. (1978)

Plasmid 1, 446-455.6. Razzaki, T. & Bukhari, A. I. (1975) J. Bacteriol. 122, 437-442.7. Castilho, B. A., Olfson, P. & Casadaban, M. J. (1984) J.

Bacteriol. 158, 488-495.8. Dreyfuss, L. A. & Iglewski, B. J. (1985) J. Bacteriol. 161,

80-84.9. Keen, M. G., Street, E. D. & Hoffman, P. S. (1985) J. Bacte-

riol. 162, 1332-1335.10. Engleberg, N. C., Pearlman, E. & Eisenstein, B. I. (1984) J.

Bacteriol. 160, 199-203.11. Maniatis, T., Fritsch, E. F. & Sambrook, J, (1982) Molecular

Cloning: A Laboratory Manual (Cold Spring Harbor Labora-tory, Cold Spring Harbor, NY).

12. Baker, T. A., Howe, M. M. & Gross, C. A. (1983) J. Bacte-riol. 156, 970-974.

13. Ingram, L., Richmond, M. H. & Sykes, R. B. (1973) Antimi-crob. Agents Chemother. 3, 279-288.

14. Figurski, D., Meyer, R., Miller, D. S. & Helinski, D. R. (1976)Gene 1, 107-119.

15. Miller, J. H. (1972) Experiments in Molecular Genetics (ColdSpring Harbor Laboratory, Cold Spring Harbor, NY).

16. Bukhari, A. I. (1975) J. Mol. Biol. 96, 87-99.17. van de Putte, P., Westmaas, G., Giphart-Gassler, M. &

Wijffelman, C. (1978) in DNA Insertion Elements, Plasmidsand Episomes, eds. Bukhari, A. I., Shapiro, J. A. & Adhya, S.(Cold Spring Harbor Laboratory, Cold Spring Harbor, NY),pp. 287-294.

18. Casadaban, M. J. & Chou, J. (1984) Proc. Natl. Acad. Sci.USA 81, 535-539.

19. Desmet, R. L., Faelen, M., Lefbbvre, N., Rdsibois, A., Tous-saint, A. & van Gijsegan, F. (1980) Cold Spring Harbor Symp.Quant. Biol. 45, 355-363.

20. Meyer, R., Figurski, D. & Helinski, D. R. (1977) Mol. Gen.Genet. 152, 129-135.

21. Grundy, F. & Howe, M. M. (1984) Virology 134, 296-317.22. Murooka, Y., Takizawa, N. & Harada, T. (1981) J. Bacteriol.

145, 358-368.

Medical Sciences: Mintz and Shuman