Moiecuiar Microbiology (1992) 6(13). 1829-1839

PhoA gene fusions in Legioneiia pneumophiiagenerated in vivo using a new transposon, MudphoA

Mark A. Albano,' Juan Arroyo,' Barry I. Elsensteln''^and N. Cary Engleberg^ ^*'Department of Microbiology and tmmunotogy. and^Department of tnternat Medicine. University of MichiganMedical School. Ann Arbor, Michigan 48109, USA.

Summary

To enable effective use of phoA gene fusions inLegioneiia pneumophUa, we constructed \Au6phoA, aderivative of the mini-Mu phage Mu dii4041, which iscapable of generating gene fusions to the Escheri-chia coli alkaline phosphatase gene (EC 3.1.3.1).Although an existing fusion-generating transposon,JnphoA, has been a useful tool for studying secretedproteins in other bacteria, this transposon and otherTn5 derivatives transpose inefficiently in Legioneiiapneumophiia, necessitating the construction of amore effective vector for use in this pathogen. UsingMudphoA we generated fusions to an E. coii geneencoding a periplasmic protein and to an L. pneumo-phiia gene encoding an outer membrane protein;both sets of fusions resulted in alkaline phosphataseactivity. We have begun to use MudphoA to mutatesecreted proteins of L. pneumophiia specifically,since this subset of bacterial proteins is most likelyto be involved in host-bacterial interactions. Thismodified transposon may be useful for studies ofother bacteria that support transposition of Mu, butnot Tn5, derivatives.

Introduction

Bacteriai proteins that mediate the virulence of pathogensare most likely to be expressed at the bacterial surface orexported from the bacterial cell. In the case of Legioneiiapneumophiia. several secreted (or exported) proteins withpotential roles in pathogenesis have been identified.These include Mip, a 24 kDa surface protein (Cianciottoet at.. 1989). MSP. a 38 kDa protease (Dreyfus andIglewski, 1986), and legiolysin. a 39 kDa protein that con-fers haemolysis, pigment production and fluorescenceactivities (Williams etat., 1991)- Several genetic methods

Received 13 August. 1991: revised 17 February, 1992; accepted 18March. 1992. 'For correspondence. Tei. (313) 936 0879; Fax (313) 7643562.

for mutationa! analysis in L. pneumophiia have beenreported. Among these methods are allelic exchange ofmutant for wild-type genes mediated by tri-parental conju-gational matings, introduction of Tn5 on the broad-host-range piasmid pRK340. and introduction of bacterio-phage Mu on the broad-host-range plasmid pRK2 (Cian-ciotto et ai, 1988: Dreyfus and Iglewski. 1985; Keen etai. 1985; Mintz and Shuman. 1987).

One way of identifying genes encoding proteins thatare localized specifically to the bacterial envelope orperiplasm is by using transposons that generate fusionsbetween host proteins and enzymes that are active onlywhen secreted. For example, TnphoA generates fusionsto alkaline phosphatase and JnbtaMlo p-lactamase, bothnormally periplasmic proteins in Escherichia coti (Manoiland Beckwith. 1985; Broome-Smith et ai, 1990). Wefound that TnphoA functions poorly as an insertionalmutagen in L. pneumophiia. Alternatively, Mu trans-posons are known to be functional in this bacterium(Mintz and Shuman, 1987). We therefore decided to mod-ify Mu dtl4041 by ligating a phoA gene fragment near itsright end, thereby constructing Mu dWphoA (herein calledMudphoA). Mu dll404i was chosen because it contains aSamHI site 116 base pairs (bp) from the Mu right end andligations into this site are capable of maintaining an openreading frame from the Mu right end through the tigatedfragment (Castilho et at., 1984; Kahmann and Kamp,1979). Similar mini-Mu (Ap,/ac) transposons have beenconstructed for in vivo generation of transcriptional andtransiational gene fusions by ligating an E. coti lac frag-ment into the fiamHI site (Casadaban and Chou, 1984).

Results

Constnjction of MudphoA

We inserted the E co/; alkaline phosphatase gene {phoA)near the right end of mini-Mu transposon Mudn4041 (Castilho et al., 1984) (Fig 1). To facilitate thisconstruction, Mudll4041 was first transposed into plas-mid pUC18A, using strain BAC101 (Table 1). A DNA frag-ment encoding the phoA gene, missing amino-terminalsequences, was ligated into the SamHI and Xho\ sites onpUCi8A::Mudl!4041. The Tn5 kanamycin resistancegene {neo) (Beck et ai, 1982) was inserted adjacent tothe phoA fragment. The resulting transposon, Mu dWphoA

1830 M. A. Albano, J. Arroyo, B. I. Eisenstein and N. C. Engleberg

pUCi8A::Mu dl 14041

pUCISA:;Mu 6\\phoA

phoA

Fig. 1. Construction of lAu6phoA. Mu CIII4041 was transposed ontopUCt8A, a derivative of pUC18 that lacks the small Pi'ult fragment con-taining the multiple cloning sequence (Yanisch-Perron era/,. 1985). TheBcl\-Xho\ fragment containing the E co/ialkaline phosphatase [phoA)gene was isolated from pSMJt t ,4"Tnp/io^ (see Table 1) and ligated topUC18A;:Mudll4041, replacing the BamHI-Xho I fragment in Mudll4041, which contained Ihe neogene{Km"'and most o(thelS50sequences (Manoil and Beckwith. 1985). An Xho\ fragment frompSMJl 1 ."...inphoA containing the Km" gene {neo) was iigated to theunique Xho\ site on pUC18A::Mudp^j4. All further work was done usingthe clone with the Km" gene transcribed in the same direction as thephoA gene. Mu right end and Mu left end refer to ihe right and left Mutemiinal repeat ('attachment') sequences, respectively. A and B are theessential Mu transpOEition-replication genes and cts is the temperature-sensitive Mu repressor gene, cts62(Casadaban andChou. t984).

), possesses only those Mu sequences neces-sary for transposition, it is defective for making phage par-ticles, and it has a temperature-sensitive repressor gene(cts62).

Characterization of the junction between the inside Muright end and phoA

For UudphoA to be effective in generating proteinfusions, its phoA fragment must be capable of expressingalkaline phosphatase activity, and the junction betvi/eenthe right end of Mu and phoA must maintain a transla-tional reading frame through the Mu right end. TheBci\-Xho\ fragment from JwphoA lacks the phoA pro-moter, ribosome-binding site, and signal peptidesequences, as well as the coding sequence for 28 aminoacids beyond the signal sequence cleavage point, toAsp-29 (Chang et ai, 1986). Since Jn5-phoA junctionsup to the 40th amino acid beyond the signal cleavagepoint are still capable of generating normal levels of phos-phatase activity (Manoil and Beckwith, 1985), weexpected that the phoA fragment in MutiphoA would alsoexpress the activity.

Ligation of the 5'-end of the phoA fragment into the rightend of Mu dll4041 maintains not only the reading frame ofthe phoA gene but also the Sau3A restriction site at theMu right end-phoA junction (Fig. 1). To confirm that thissite was preserved, we analysed a deletion derivative ofpUC18A::Mudpho>4, since it was easier to map the Sau3Asite on this smaller plasmid (Fig. 2). The presence of thisjunctional Sau3A site was taken as evidence that the liga-tion was precise and that the junction maintained the cor-rect translational reading frame.

Transposition ofMudphofi^ and construction of a doublefi/lu lysogen. EA17

To test whether MudphoA would transpose as expected,strain M8820Mucts was transformed with pUC18A::Mud-phoA and one of the transformants was purified. Plasmid

a/W7,..CATTACATGGGJGATCAGACTGCCCCTCTGCGT,. . .L*u His Gly Asp Gin Thr Ala Ata L«u Arg.

I

700bp650bp

Fig. 2. Analysis of the Mu dil4041 phoA ligation junction. The inferredsequence of the junction is shown at the top of the figure with the junc-tional SaU3A site marked by a solid circle, pUCt8A::Mudp/io/1 wasdigested with PvuM and religation of the smaller fragmeni resulted inrecovery of a deletion plasmid that contained Ihe junctional Sau3A slle(•) as well as several other 5au3A sites (O). The 700 bp and 650 bpfragments indicated on the circular map are shown in the restrictionenzyme analysis of the pUCt 8A::Mudpho^ deletion plasmid (lower right).As indicated on the map, digestion of the 650 bp Sau3A fragment withAatW results in two smaller fragments of 400 bp and 250 bp. Note thedisappearance of the 650 bp fragment on the gel in the double digest.Double digestion with Sspl and Sau3A also eliminated the 650 bp frag-ment and produced fragments of Ihe expected size (data not shown).

PhoA gene fusions in L. pneumophila using MudphoA 1831

Table 1. Bacterial strains and plasmkJs.Strain/Plasmid Description or Genotype Source or Reference

E. coli

M8820

M8820MU

M8820fUludSMG1655JC 10240

BAC101

EA4EA17

X2981

F , araD139, A[ara-leu)7697,A.iproAB-argFtaclPOZYA)X\U, rpsL

F", araD139, A(ara~leu) 7697, A{proAB-argF-laclPOZYA)X\i\, rpsL. Mu c"

M8820 with Mu cts62Prolofroph, F ,̂ X~Hfr, thr300. recA56, srf300;;Tn/0(Tc").

reiAI. riv318, spoT, thit,rpsE2300. X"M8820 recA56 sri::Tn rO(Tc") Mu cts62 with

Mu dl 14041M8820MUCIS. recA5e. sri300:-Jn tO (Tc")EA4 with Mudpho/1 prophageE. CC'//K-12. F"' A41[proe-/acV7].>."

T3r. Aasd-A4, zhf'2:-Jr\10. cycAlZ2981 with pTLP7::Mudp/)0/»

X2981(pRK212.1)CC118

L. aneumoohila

AAIOOAA103

AA110AA111AA112AA113AA114AA118

^2981 withpRK212.1araDt39. Mara. leu)7696. S tacX74.

phoA.A20. galE. gatK. thi, rpsE, rpoB.srgEa,,, recA t

L pneumophilaSG^ 130bL. pneumof^ilaSG^ 130b high-frequency

conjugation recipieniAA103 with Mudprto,^ fusionAA103 with MudphoA fusionAA103 with MudphoA fusionAA103 with Mudp/io>1 fusionAAIOO with pNC31.5::MudptioAAA103with MudphoA iusian

Casadaban and Chou (1984)

Casadaban and Chou (1984)Guyerera/. (1981)Csonka and Clark (1979)

Casadaban and Chou (1984)

This wofkThis wofkR. Curtiss

This work

J. E. RogersManoii and Beckwith (1985)

Engleberg eiai. (1984a)Cianciotto er a/. (1988)

This workThis workThis workThis workThis workThis work

pUCi8A pUC18,Pi'uH deletion removing the polylinkerregion, Ap"

pUC18A::Mu dll4041 pUC18A with Mu dl!4041 Insertion. Ap". Km"pUC18Awith MudphoA. Ap" Km"Ap", pa/;;Tnpfto>l

PNC31.5 Ap", Cm". oriT. mippTLP7 Cm". saoB, oriTpJLP7::MudphoA Cm". Km", sacB. oriTpRK212.1 Ap", Tc^.Tra-

This work

This workThis wor1<Engleberg era/. (1991)Cianciotto er a/. (1989)This workThis workFigurski efa/. (1976)

DNA isolated from this strain was compared topUC18A::Mudpho-A by restriction enzyme digestions toensure that pUC18A::Mudpfto/^ had been stably intro-duced and had not undergone any rearrangements ordeletions. The restriction mapping indicated that the twoplasmids were identical. M8820Mucts harbours a com-piete Mu prophage that is capable of making phage parti-cles. Since it also possesses the temperature-sensitiverepressor gene (cts62), phage growth and transpositionare inducible at the permissive temperature of 42°C(Castilho efa/.. 1984).

A culture of M8820Mucts (pUCi8A::Mudp/io>A) wasgrown at SO'̂ C and then shifted to 42'"'C to initiate transpo-sition and lysis. A lysate prepared from this culture wasused to transduce E. coli strain MG1855 to kanamycinresistance (Km"). The M8820Mucts {p\JC^8^:Mu^phoA)strain produced lysate titres of 10^-10^ plaque-forming

units (pfu) per ml when titred on MG1655, and the fre-quency of Km" transduction was tO"^-1O"per recipientcell at a multiplicity of infection of 1. Among the Km"transductants, 97% were ampicillin sensitive (Ap^), indi-cating that most Mu particles in the lysate containingMudphoA did not contain pUC18A. We concluded that theseparation of these antibiotic markers was probablycaused by the transposition of MutiphoA at the permis-sive temperature.

We adapted the soheme of Castilho et at. (1984) to testthe ability of MudphoA to generate active phoA fusionproteins. In this scheme a donor strain is constructedcarrying one copy each of MudphoA and the helperprophage Mucts. The target plasmid is then introducedinto the donor strain, transposition is derepressed by rais-ing the temperature, and the lysate generated by this pro-cess is used to transduce a Mu-immune recipient. In

1832 M. A. Albano, J. Arroyo. B. 1. Eisenstein and N. C. Engteberg

some transductants, homologous recombination betweenMu sequences results in recovery of Mu insertions in thetarget plasmid (Castilho etai. 1984).

To prevent the formation of target plasmid multimers,which would be recombinationally unstable following Muinsertion, it was necessary to construct this donor strain ina recombination-deficient background. Accordingly,M8820Mucts was transduced with a PI lysate made fromstrain JC10240 irecA56 srl::Tn10 (Tc")) (Csonka andClark, 1979). selecting for tetracycline resistance. Theresulting strain, EA4. was purified and tested for the recAmutation by u.v. light sensitivity (Maniatis et ai. 1982).Strain EA4 was then transduced by a M8820Mucts{p\JC\::MudphoA) lysate, followed by selection for Km"iysogens in which MutiphoA had inserted despite Muimmunity. While the frequency of obtaining these doubleMu lysogens was low. one was selected, i.e. strain EA17.Chromosomal DNA from strains EA17 and EA4 andpUC18A::Mudp^o-4 plasmid DNA was cut with H/ndllland probed with radiolabelled pUCi8A::Mudp/7O^, Com-parison of the hybridizing fragments among these DNAsshowed that EA17 contained the same Mu fragments asEA4 as well as fragments identical to pUC18A::Mudp/7o/lthat were internal to MudphoA, but no fragments specificto the vector pUCi8A. In addition, alkaline lysis prepara-tions of EA17 produced no detectable plasmid DNA. Weconcluded from this evidence (data not shown) that EA17contained chromosomal copies of the two Mu genomes.

Isolation and expression of MudphoA fusions in fi-lactamase and the L. pneumophiia mip gene

The p-lactamase gene on pBR322 encodes a periplasmicprotein (Bolivar et ai. 1977). TnphoA insertions into thisgene generate active phoA fusions (Manoil and Beckwith,1985). To test for MudphoA fusions to this gene. pBR322was introduced into EA17 by transformation. One trans-formant was induced for phage growth and transpositionat 42"C. The lysate from this strain was used to transducethe Mu-immune, recA^ strain M8820Muc*, and to recoverpotential MudphoA insertions. The Mu-immune strainM8820Muc' was used as a recipient to reduce the like-lihood of additional transpositions by the incoming Mud-phoA.

Table 2 shows the frequencies of the various pheno-types generated by pBR322::Mudp/7o^ insertion. The fre-quency of insertions into the Ap" and Tc" genes waslower than expected given the size of these regions onpBR322 (Sutclifte, 1979). About 14% of the MudphoAinsertions in the Ap" gene were PhoA*, which is close tothe expected frequency (16.7%) of MudphoA in-frametransiational fusions (two orientations, three readingframes)

We next isolated fusions to an L. pneumophiia gene.

Table 2. Phenotype of Km" transduclants with a Mu lysate fromEA17(pBR322).

Phenotype Number counted

Ap'', Tc"- PhoA-Ap", Tc^, PhoA-Ap^. Tc", PhoA*Ap^. Tc". PhoA*

B952030S

94.52.03.00.6

mip. whose product is an outer membrane protein. Whenmip is cloned on an E. coti replicon. its protein productlocalizes to the E. co//surface (Engieberg etal., 1984a,b).A plasmid carrying a cloned copy of mip. pNC31.5 (Cian-ciotto etai, 1989), was introduced into EA17 by transfor-mation. The resulting strain was induced for phagegrowth and two mip::MudphoA insertions having alkalinephosphatase activity were isolated following transductionof M8820Muc*. We introduced the mipy.MudphoA fusionplasmids into L. pneumophiia strain AAlOO by electro-poration. Transformants expressed PhoA activity on XP-containing agarose plates (see the Experimental proce-dures), confirming that MudphoA fusions can beexpressed in JL. pneumophiia as well as in E. coti

To confirm that the alkaline phosphatase activity ofthese strains resulted from fusions to MudphoA and notfrom other events (such as constitutive phosphatasemutations), strain CC118, an E. co//strain thathasap/7o-4deletion (phoA A20) (Manoil and Beckwith, 1985) wastransformed with each of the presumed PhoA' plasmids.All of the resulting CC118 strains acquired alkaline phos-phatase activity, demonstrating that the PhoA' phenotypewas of plasmid origin.

The insertion points of MudphoA in both the fl-lacta-mase gene and the mip gene were mapped and the sizesof the expected fusion proteins were estimated. Several£)/a::Mudp/70>4 insertions mapped in a region approxi-mately 300 bp from the beginning of the bla gene whileother bla::MudphoA insertions mapped approximately400. 500, 600 and 800 bp into the bta gene. Of the twomipy.MudphoA insertions, one mapped approximately200 bp upstream and the other 100 bp downstream froma unique H/ndlll site located near the C-terminal end ofm/p (Engieberg etal.. 1989).

We also analysed whole cell lysates of the CC118strains containing plasmids with MudphoA fusions byimmunoblotting with anti-alkaline phosphatase antibodyand anti-Mip monoclonal antibody (12F4) (Cianciotto etai, 1989). Anti-alkaline phosphatase antibody producedreactive bands in each of the bla::MudphoA samples andthe moiecuiar weights of these protein species closelyapproximated the predicted sizes based on restrictionmapping. In the case of the m/p::Mudp^o-4 samples, bothanti-alkaline phosphatase and anti-mjp antibody reacted

PhoA gene fusions in L. pneumophila using MudphoA 1833

Fig. 3. Oeteclion of alkaline phosphatase activilyon XP agarose plaies. In panel A, the XP agarosehas been adjusted to a pH of 7.0 and in panel 8 toapHof 11.0. In both panels, disc 1 i5AA103.discs 2-5 are PhoA", MudphoA insertion mutantsAA110. AAl 11 AAn2 . and AAl 13 and disc 6 isAA114. a PhoA' L. pneumophilacarrymgpNC31,5::Uudpho^ {mip:.UudphoA]. At pH 7. allof the strains, including AA103 which lacks alAudphoA insertion, produce a blue colour due tothe nalive L pneumophila phosphatase (Muller.1981). At pH 11.0. only the L. pneumophilastrains with MudphoA insertions that resulled m E.co/i aikaiine phosphatase activity produce a bluecolour.

with the same bands, which also closely approximatedthe predicted molecular weights (data not shown). Weconcluded from these observations that MudphoA inser-tions into the bta and mip genes resulted in the formationof stable fusion proteins.

Construction ofadetivery vector. pTiP7:Mudphok

For delivery of MudphoA into Legionetia. we constructeda vector, pTLP7, that has both selectable and counter-selectable genetic markers. We included the broad-host-range plasmid pRK2 origin of transfer sequence, oriT(Guiney and Yakobson, 1983), to mobilize the vector intoL pneumophila using transfer functions provided in transby a helper conjugational plasmid, pRK212.1 (Figurski efat.. 1976). Counterselection of the delivery vector facili-tates the isolation of chromosomal MudphoA insertionssince any kanamycin-resistant transconjugants remainingafter counterselection must be due to non-vector, or chro-mosomal, copies of MudphoA. We chose the sacB alleleof Bacitlus subtitis as the counterselectable marker (Gayetai. 1985). The sacSgene encodes a sucrose-inducibleIevansucra5e that synthesizes levans. branched poly-mers of fructose. Synthesis of these polymers within theperipiasm of L pneumophita is lethal. Thus, U pneu-mophila carrying this gene will not grow on buffered char-coal yeast extract (BCYE) supplemented with sucrose,and only cells that have lost the allele are able to survive(Cianciotto ef a/., 1988).

Strain EA17 was transformed with pTLP7 and Mud-phoA insertions into the plasmid were isolated. One ofthese pTLP7::Mudpho-4 plasmids was purified, and theinsertion was mapped to ensure that MudphoA had notinserted into a strategically important sequence, such asthe sacB gene. pRK212.1 and pTLP7::Mudp/7O-^ werethen introduced by transformation into strain x2981, adiaminopimelic acid (dap) auxotroph, since exclusion ofdiaminopimelic acid from selective BCYE medium coun-terselects the £. co//donor strains in conjugal matings.

Isolation of MudphoA insertions with alkalinephosphatase activity in L. pneumophila

We used tri-parental conjugative matings to introducepTLP7;:Mudp/7O>4 into L pneumophila (Engieberg et ai,1988), and screened the resulting Km" exconjugants forphoA activity. Strain AAl 03 was used as the recipientstrain because it generates higher frequencies oftransconjugants in such matings (Cianciotto etai. 1988).We did not subject the mating mixtures to sucrosecounterselection, since simultaneous selection for Km"and sucrose counterselection reduced the total number oftransconjugants by about 10-fold. Plasmid counterselec-tion could be more efficiently applied after isolation ofPhoA' transductants. In a typical experiment, wescreened between 1000-2000 Km" transductants. Thefrequency of PhoA' isolates was about 0.1%. Some posi-tive colonies developed colour more rapidly than others,probably indicating different levels of phosphataseexpression.

Figure 3 demonstrates the PhoA screening assay usingXP-agarose indicator plates. At pH 7, both PhoA*̂ andPhoA" strains exhibited phosphatase activity, seen as anintense blue colour on the XP-agarose indicator medium(Fig. 3A). By raising the pH of the medium, the back-ground activity produced by the native L pneumophilaphosphatase was minimized (Fig. 3B, disc 1), whereasstrains with PhoA" gene fusions continued to produce ablue colour (Fig. 3B, discs 2-7). Strain AAl 14 (m/p::Mud-phoA) was included to show the expression of a Mud-phoA gene fusion to a known L. pneumophila surface pro-tein. PhoA' strains AA110 and AA112 consistentlyproduce a blue colour that is less intense than the colourproduced by AAl 11, AAl 13 or AA114.

The PhoA* L pneumophiia isolates were also analysedby Southern hybridization (Fig. 4). Figure 4. panel Ashows genomic DNA from four PhoA^ strains, digestedwith H/ndlll and probed with pTLP7:: MudphoA. TheMudphoA insertions are different In each strain, since the

1834 M. A. Aibano. J. Arroyo. B. I. Eisenstein and N. C. Engleberg

2 3 4 5 2 3 4 5

5.5kb —4.8kb —

3.5kb — trrf1.2kb — I

9.5M1

• 4 • *sacB onT Cm' prtoA neo

Probe A

Probe B

Fig. 4. Southern hybfidizalron of L pneumophiia having MvdphoAinsertions. Lane 1 (control), pTLP7:;Mijdprto4: (ane 2. AAl 10: iane 3,AAn i : !ane4 , AA112: lane 5. AAf 13. The Hindlll restriction map ofpTLP7::Mudp/io^ is shown below the phoiographs. The Ihinner. dia-gonally striped segment corresponds to pTLP7 veclof sequences and thethicker, darkly shaded segment is MudphoA, inclusive of the Mu right andleft ends. The corresponding sequences used as probes are indicatedbelow Ihe map. In panel A, the filler was probed with pTLP7:: MudphoA(probe A). In panel B . the filter was washed free ol probe A and reprobedwith pTLP7 (probe B) to identify vector-specific fragments. Note that all olthe samples in panel A have a 5.5 kilobase (kb) internal MudphoA frag-ment Note that AAl 10 and AAl 11 have only one Mudp^JO^ insertioneach, and they are free ol vector sequences. AAl 12 and AAl 13 haverelatned vector sequences as well as transpositions. The high molecular-weight band in the AAl 10. AAl 11, AAl 12, and AAl 13 samples (lanes2-5) is presumably pRK2i2.i which has a unique H/ndlll site and sharesthe orirsequence with thepfobe (Rgurski etal.. 1976).

of C0IEI replicons in L. pneumophila (Engleberg et ai.1988). To confirm this interpretation, the same fiiter wasprobed with pTLP7 (Fig. 4B). Vector sequences were pre-sent in strains AA112 and AAl 13 (lanes 4,5) but not inAAl 10 and AAl I t . Moreover, the vector fragments inAAl 12 and AAl 13 were of unexpected sizes, indicatingsome rearrangement of the vector sequences.

Detection of MudphoA fusion proteins

To detect the presence of fusion proteins in the PhoA* Lpneumophila strains, we performed immunoblotting ofwhole cell lysates with anti-alkaline phosphatase antibody(kind gift of Dr David A. Lowe)(Fig. 5). In each case, theantibody detected fusion proteins with apparent molecu-lar weights greater than that of native alkaline phos-phatase (Fig. 5, lane 1). The fusion proteins in strainsAAl 11. AAl 13 and AAl 14 (lanes 4, 6, 7) reacted moreintensely than those in strains AAl 10 and AAl 12 (Fig. 5,lanes 3 and 5). This pattern of reactivity correlated withthe differences in intensity of blue colour detected on XP-agarose indicator plates for these strains (Fig. 3). Thesmaller bands seen with most of the samples presumablycorrespond to alkaline phosphatase degradation pro-ducts. These products also appear to be of different sizesand intensities, except for one fragment common to all thesamples that is slightly smaller than the alkaline phos-phatase monomer (Fig. 5. lane 1). Presumably thissmaller, common band represents the product of the trun-cated phoA fragment used in MudphoA.

sizes of the junctional bands, determined by the positionsof flanking genomic H/ndlll sites, are different. EachMudphoA insertion would therefore produce three bands,one corresponding to the 5.5-kb internal fragment, andtwo variable junction fragments. AAl 11 (Fig. 4, lane 3)has three fragments, indicating a single insertion. AA110(Fig. 4, lane 2) has two fragments that react strongly; onecorresponds to the 5.5-kb internal fragment. The thirdfragment is probably the weaker band at approximately4.0 kb. AAl 12 and AA113 (Fig. 4. lanes 4 and5) havemultiple insertions. By comparing lane 1 (pTLP7::Mud-phoA) with AAl 12 and AAl 13 (Fig. 4, lanes 4 and 5) itappears that these last two strains contain vector-specificfragments. Since we did not subject the matings to coun-terselection to eliminate the vector, the presence of thesebands is not surprising. The spontaneous loss of the vec-tor in AAl 10 and AAl 11 is consistent with the behaviour

Fig. 5. Immunoblot of MutiphoA fusion proteins with anti-PhoA serum.Whole cell lysates were prepared from L. pneumophiia strains asdescribed by Pearlman etai. (1985). eiectrophoresed in a SDS-poly-acrylamide gel (8%), Iransterrad to nitrocellulose and reacted with anti-alkaline phosphatase antibody. Lane 1. denatured alkaline phosphatase:lane 2. AA103; lane 3, AAl 10; lane 4. AAl 11; lane 5, AAl 12; lane 6,AAt13; lane 7. AAl 14. Note the absence of reactive protein in AAl03and the presence of strongly reacting bands in the other lanes which con-tain samples (rom PhoA' L pneumophila strains. In lane 7, the mipr.phoAfusion protein had a predicted size of 56 kDa based on restriction enzymeanalysts of the insertion site on pNC31,5. The other strains show variablysized bands al least one of which is larger than alkaline phosphatasemonomer in each case.

PhoA gene fusions in L. pneumophiia using MudphoA 1835

a>Or-enatDo

ab

>

9080

70

60

50

40

30

20

10

n

Exp«rim9nt I

AA103 -PhoA* MudphoA atrains--

100,9080706050 J403020100

Experiment I

AA103 — PhoA+

Bg. 6. Reproducible cylopatriic efieci of PhoA' MudphoA mutants of L.pneumophiia in a U937 cell assay. The iwo sxperiments shown were per-(ormed on separate days with separate cultures of bacteria and mono-cytlc cetls with inocula of 10^ bacteria. After 72 h at 37''C, the U937 cellviability was detefmined by staining with MTT (% viable U937 cells =mean OD^io for each strain ^ mean ODSM O' uninoculated. conlrol U937cell wells X 100), AA103 is the virulent, parent stram and contains noMudpfto^ insertions. PhoA* mutants studied were (from left to right):AA110,AA111.AA112.AAl13.andAAn8

Retative Infectivities of PhoA* Wt/dphoA L. pneumophiiamutants in U937cetl assays

MudphoA contains the temperature-sensitive repressorallele cts{62) that regulates transposition in a tempera-ture-dependent manner. The utility of MudphoA in identi-fying virulence factors in L. pneumophita (or any otherpathogen) depends upon the effect, if any. on bacterialcell viability of growth at 37-C. That is. if Mu transpositionis sufficiently derepressed at 37''C to kill Mudp/7o-4-con-taining bacteria, all strains having functional transposonsvi/ould appear to be attenuated in virulence assays con-ducted at this temperature. Since we intend to screenPhoA' MudphoA L. pneumophita mutants for an in vitroU937 cell cytopathicity effect (CPE) at 37"C (Pearlman etai., 1988), it is important to insure that any observedreduction in bacterial virulence is caused by insertionalinactivation of a virulence-related gene rather than bybacterial death resulting from derepressed Mu transposi-tion.

To address this concern, we tested the plating efficien-cies of PhoA' MudphoA L. pneumophita mutant strains at

30"C and 37''C. In all strains examined, there was no sig-nificant difference in the cfu numbers of individual Mud-phoA L. pneumophiia mutant strains grown at either tem-perature (data not shown). Next, we conducted severalindependent experiments to compare PhoA' MudphoAL. pneumophiia mutants and the parent strain, AA103. inthe cytopathicity assay using U937 cetls. Figure 6 showsthe results of two of these assays. Clearly, this screeningassay detects reproducible differences between thePhoA' MudphoA mutants; some are as virulent as theparent strain, whereas two mutants were significantlyattenuated. These experiments suggest that any sec-ondary transposition induced by growth at 37 C is not suf-ficient to affect the infectivity of these strains, as mea-sured by the U937 cetl assay.

Discussion

We described construction of MudphoA. a derivative ofMu dll404i (Castilho et ai.. 1984) that can be used togenerate alkaline phosphatase gene fusions in vivo. Togenerate a PhoA* fusion, MudphoA must insert into agene with an amino-terminus encoding a signal peptide,since alkaline phosphatase is normally active only when itis exported from the cytoplasm {Michaelis et ai, 1983;Hoffman and Wright. 1985; Manoil and Beckwith, 1985).Insertions of MudphoA in several different positions in thegene for the periplasmic p-lactamase protein gave fusionproteins with alkaline phosphatase activity. Based onrestriction mapping, many of these insertions were in thesame region of the [Mactamase gene, possibly indicatingnon-random Mu-insertion activity or lethality of certaininsertions (Castilho etai. 1984; Hall and Silhavy, 1981),Two different insertions in the gene for the membrane-localized Mip protein of L. pneumophiia similarly alsoresulted in enzyme activity, both in E. coli and in L. pneu-mophiia.

It is reasonable to assume that detection of PhoA'MudphoA fusions in L. pneumophita will identify genesencoding signal peptides. a hallmark of proteins that aresecreted from the cytoplasm. Secreted proteins are thosemost likely to be involved in mediating L. pneumophitapathogenesis; therefore we can use MudphoA to identifyand to mutate the genes encoding these proteins. Anadvantage of using MudphoA is that chromosomal inser-tions can be selected in vivo under different environmen-tal conditions, thereby allowing identification of selectivelyexpressed genes that may be important for virulence, butare otherwise difficult to identify using in vitro methods.For example, one can select for PhoA* MudphoA inser-tions at lower temperatures, under iron limitation, oxygenstress or low/high osmolarity — conditions known toaffect expression of virulence determinants in otherpathogenic bacteria.

1836 M. A. Albano, J. Arroyo, B. I. Eisenstein and N. C. Engleberg

Mintz and Shuman reported the ability of Mu to trans-pose to many sites within the Legionella genome, by gen-erating insertion mutations and lac gene fusions (Mintzand Shuman. 1987). We have obtained similar resultswith f^\J6phoA. All of the fusion strains isolated to datestably maintain the PhoA' phenotype after several pas-sages on BCYE/Km media.

Fusions generated by Uu6phoA insertion contain 38extra amino acids resulting from translation of the Muright end sequence (Fig. 2). The presence of this extrasequence had no effect on iac fusion protein expression(Castilho etai. 1984); however, its effect en the secretionand dimerization of PhoA*̂ fusion proteins was unpre-dictable. Our results indicate that this linking peptide doesnot prevent expression of the enzyme, although pro-teolytic cleavage in this region seems to cccur frequently,judging by the products seen in Fig. 5.

To screen L. pneumophita strains for the presence of E.co//alkaline phosphatase activity, we developed an in situassay for PhoA activity. This assay helped us overcomeseveral technical difficulties associated with screening forPhoA activity in L. pneumophila. For example, the blackcolour of BCYE medium interferes with blue colour detec-tion of PhoA activity, and incorporation of the colour indi-cator substrate, XP, into BCYE is toxic to L pneumophila.In addition, the native phosphatase activity of L. pneu-mophiia masks the E. co//phosphatase activity at pH lev-els that permit bacterial growth; however, this backgroundactivity becomes insignificant at a pH of about 11.0.

The broad-host-range plasmid pRK212.1, whioh wasused in this study to provide transfer functions, is itselftransferred at a high frequency (Fig. 4). II is possible toeliminate this plasmid either by passage on non-selectivemedia or by 'curing'. Its use may be avoided by intro-ducing MutiphoA from an E, coli strain carrying the Trafunctions on the chromosome or by placing MudphoA ona conjugative 'suicide' vector (Engleberg et ai, 1988;Miller. 1972; Keen et ai, 1985). !n any case, the presenceof pRK212.1 in L. pneumophila confers no detectablechange on intracellular infeotivity (J. E. Rogers, unpub-lished observations). In some cases (Fig. 4B) the deliveryvector, or derivatives cf it, is present in the fusion strains.The presence of the apparently modified vector frag-ments suggests possible Mu-induced inversions or dele-tions (Pato. 1989). Sucrose counterselection effectivelyeliminates sacS-ccntaining plasmids frcm L pneu-mophita, but it is not known how this counterselection willeffect these unexplained derivatives (Cianciotto et ai,1988).

We have isolated several PhoA* L pneumophilatransoonjugants, and we have found two of these strains(AA110, AA118) to be significantly attenuated in a U937cell-infectivity assay. However, certain biological featuresof the transposon impose limitations on the use of Mud-

phoA for identification of virulence genes. For example,since derepression of Mu cts(62) phage transposition canbe reliably induced at 42"C, it is unlikely that animal chal-lenge experiments could be performed using MudphoAmutants. The febrile temperatures would probably induceenough secondary transposition to limit the growth andsurvival of L pneumophila strains that carry Mu6phoA. Amore immediate concern is that some leakiness of the Mucts(62) might occur at the lower, semi-permissive temper-ature of 37'̂ C and result in a similar bacterial attrition inthe U937 cell-infectivity assay. At the very least, passageof the bacteria at 37"C might be expected to result in aheterogeneous population of bacteria containing multiplecopies of the transposon in diverse chromosomal loca-tions.

To limit secondary transpositions, we routinely maintainand cultivate all L. pneumophila {\Au6phoA) strains at3O''C or store them at -7O'''C. All bacterial inocula for theU937 cell assay are grown at 30°C, and any bacteria iso-lated from cell cultures (at 37"C) are discarded. By exer-cising this precaution, we have shown that most PhoA' L.pneumophiia strains are as cytopathio as the parentalstrain in the U937 cell assay. This normal level of cyto-pathicity has been demonstrated in repeated experimentson different days, using different cultures of the L. pneu-mophiia test strains and different batches of differentiatedU937 cells. In the same experiments, the two PhoA'mutants that were attenuated in our initial screeningassay were comparably attenuated in all subsequentassays. We conclude from this experience that L. pneu-mophila does not become attenuated in the U937 cellassay simply because of the presence of either MudphoAor a PhoA* gene fusion. Furthermore, the attenuatedPhoA* mutants are likely to acquire this phenotype as aresult of transposon-induced mutations in virulence-related genes.

The multiplicity of Uu^phoA transpositions in L pneu-mophila presents a second potential problem in charac-terizing the phenotype of attenuated mutants, since mosttransconjugants contain more than cne copy of the trans-poscn (typically 1—4 copies). Although it is likely that onlyone of these copies is responsible for the PhoA' activity, itis by no means certain that the fused gene is also respon-sible for the attenuated phenotype. For this reason (andbecause of the additional concern about possible sec-ondary transpositions, outlined above), all mutations ofinterest will eventually have to be isolated by molecularcloning and then reintroduced into a wild-type strain byhomologous recombination. This manoeuvre will deter-mine whether the PhoA' gene fusion or some other Mud-phoA insertion is responsible for the attenuated pheno-type. In addition, a reconstructed strain will not have afunctional transposon and could be used for animal chal-lenge studies and extended phenotypic characterization.

PhoA gene fusions in L. pneumophila using MudphoA 1837

Experimental procedures

Bacteriai strains, plasmids and growth conditions

The E. coli and L. pneumophila bacterial strains used In thisstudy are listed in Table 1. Strain BAC101, obtained from M. J.Casadaban, was our source for Mu dll4041 (Casadaban andChou, 1984). Strain CC118 is an E. coti phoA deletion strainobtained from C, Manoil (Manoil and Beckwith, 1985). L. pneu-mophila strain AAIOO is a serogroup 1 clinical isolate, 103b{Los Angeles), and AA103 is a derivative of AA100 that is ahigh-frequency conjugal recipient (Engleberg et ai. 1988;Cianciotto etai. 1988}.

Plasmids used include pUCi8A. which was derived in thislab. from pUCIS by deletion of the small PvuW fragment con-taining multiple ctoning sequences (Yanisch-Perron ef a/.,1985). pRK212,1 is a deletion derivative of pRK2 that is Km^{Figurskiefa/.. 1976).

E. coii strains were routinely cultured from frozen stocksonto Luria-Bertani (LB) medium or in LB broth containingantibiotics as required. Antibiotics were used at the followingconcentrations; kanamycin (Km) 25 |ig ml" ' , tetracycline |Tc)15 tig rn\~\ chloramphenicol (Cm) 20 (ig ml"^ and ampicillin(Ap) 50 ng mrV L. pneumophita strains were also culturedfrom frozen stocks onto BCYE medium, containing 25 ^g ml"'kanamycin as required.

Genetic and DNA manipulations

Chromosomal DNA was prepared from E. coti and L pneu-mophiia as described by Ausubel ef ai. (1987). Plasmid DNApreparation, DNA cloning and Southern hybridizations wereperformed using standard methods (Maniatis et ai. 1982).DNA restriction enzymes were used as suggested by themanufacturers. ^^P-!abelled DNA probes were made with themulti-prime kit (Amersham).

Cloning and isolation of recombinant plasmids were done inE. CO//strain HB101, Transformation of temperature-sensitiveMudp/Jo>^-containing plasmids was performed by electropora-tion, using a Bio-Rad Gene Pulser apparatus (Bio-Rad) or bythe method of Chung etai (1989). Other transformations wereby standard procedures (Maniatis etai. 1982). £ co//and L.pneumophiia cells were prepared for electroporation by themethod of Dower et al. (1988). Methods for handling bacterio-phage Mu and PI mediated transduction of £. coji were asdescribed by Bukhari and Ljungquist (1977) and Sifhavy et ai(1984).

Tri-parental bacterial conjugations for introduction of Mud-phoA into L. pneumophila were performed as described byEngleberg ef al. (1988), with minor modifications. Mating mix-tures of AA103. x2981(pRK212.1) and x2981 (pTLP7;;Mud-phoA) were incubated on nitrocellulose filters on the surface ofBCYE medium for 6 h at 30"C. The mating mixtures were thenresuspended and plated onto BCYE/Km at appropriate dilu-tions to obtain 100-200 colonies per plate. The plates wereincubated for an average of 6 days at 30°C, until the colonieshad reached optimum size for transfer to XP-agarose plates.

Piate assays for detection of PhoA fusions

For detection of alkaline phosphatase activity in E. co//strainsresulting from MutiphoA insertions, we incorporated 5-bromo-

4-chloro-3-indolyl phosphate (XP) at a concentration of 40 ngml"' into the selective LB media. The high phosphate concen-tration in LB medium effectively suppressed expression of thenative £. co//alkaline phosphatase, allowing easy detection ofPhoA activity resulting from Mu6phoA fusion proleins-

This strategy fails to detect E. coli alkaline phosphataseactivity in L. pneumophiia because the black backgroundcolour of BCYE prevents visualization, XP inhibits L. pneu-mophila growth, and L- pneumophila produces a broad pHrange (pH 4-8) phosphatase activity that is not inhibited byhigh phosphate concentration (Muller, 1981). To circumventthese technical problems, we developed XP-containingagarose plates for detection of PhoA' L. pneumophilastrains.The XP-agarose plates were made as follows- t.5 g ofelectrophoresis grade agarose was mixed with 100 ml of a 0.1molar solution of 3-(cyclohexyiamino)-1 -propanesulphonic acid(CAPS) buffer (Sigma). The suspension was then brought topH 11.0 with 10 M NaOH and heated lo dissolve the agarosecompletely. After the solution cooled to approximately 40'^C.XP (40 mg ml"' in dimethyiformamide) was added at a concen-tration of 40 |ig m l ' and 10 ml of the XP-agarose solution wasallowed to solidify in Petri plates.

L. pneumophita Km^ transconjugant colonies were liftedonto nitrocellulose filters, and the filters were applied colony-side-down onto the surface of XP-agarose plates. PhoA*colonies were generally apparent after 1-2 h incubation atroom temperature. In each experiment PhoA* and PhoA" con-trol strains were included on the filters to assess the develop-ment of blue colour because of £. coti alkaline phosphataseactivity. When PhoA* colonies were detected, master plateswere re-incubated at 30°C until good colony growth wasachieved (about 3 days). The PhoA* colonies were identrfied,picked, and streaked for purity on BCYE/Km media. After re-isolation, strains were always re-tested on fresh XP-agaroseplates. The XP-agarose plates were efficient at detecting alka-line phosphatase activity as long as the plates were used whenfreshly made. After about 48 h (he plates began to lose speci-ficity, probably owing to a spontaneous reduction in pH.

Antibodies and immunoassays

To detect alkaline phosphatase fusion proteins, whole celllysates of strains were prepared and reacted with anti-alkalinephosphatase antibody. L. pneumophita strains were grown onplates for 48-72 h at fhe appropriate temperature. Cells weresuspended in water to an optical density at 550 nm (OD550) of1.0 and samples were boiled for 5 min in the presence of SDSand 2-mercaptoethanol (Pearlman ef ai. 1985). The lysateswere centrifuged briefly to pellet debris, and equal amountswere elecfrophoresed in an 8% discontinuous SDS-polyacryl-amide gel. Alkaline phosphatase (Sigma) was treated as aboveand included in the immunoassays. The separated proteinswere transferred to nitrocellulose, and immunoblots were pre-pared as previously described (Cianciotto etai., 1989). Horse-radish peroxidase-conjugated goat anti-rabbit antibody (Cap-pel Industries) was used as the secondary antibody anddiaminobenzidlne (Sigma) was used as the colour indicatorsubstrate.

U937 celi cytopathicity

Determination of the cytopathic effects of L pneumophila

1838 M. A. Atbano. J. Arroyo. B. t. Eisenstein and N. C. Engieberg

strains was performed as described by Pearlman et al. (1988)with slight modifications. Bacteria grown on either BCYE orBCYE/Km at 30"C were added to adherent U937 cells in 96-well plates (10^ cells per well), and the infection was allowed toproceed at 37"C for 72 h. The infected monolayers wereneither washed nor treated with antibiotics during this incuba-tion. At the end of the incubation, the U937 cell-cytopathiceffect was determined as previously described. The vital staintetrazolium salt 3-(4.5-dimethyl thiazol-2-yl)-2,5-diphenyl tetra-zolium bromide (MTT; Sigma) was used to distinguish live fromdead U937 cells. In alt experiments, a set of eight replicateinfections for each bacterial strain was examined. The lowestand highest OD^so readings were eliminated from the set ofvalues used to calculate an average and standard deviation.

Acknowledgements

We thank Dr David A. Lowe for his kind gift of anti-alkalinephosphatase antibody, and M. J. Casadaban, C. Manoil, B.Bachmann, R. Curtiss III, and J. E. Rogers for strains. Thiswork was supported by Public Health Service grants R01AI26232 (to N.C.E.). RO1 AI24731 (to B.I.E.). M. A. Albano issupported by NIH training grant 5T32 AI07360.

References

Ausubel, F-M.. Brent. R.. Kingston, R.E-, Moore. D.D., Seid-man, J.G., and Struhl, K. (1987) Current Protocols in Molec-ular Biology. New York: John Wiley and Sons.

Beck, E-, Ludwig, G., Auerswald, E.A.. Reiss, B., and Schaller,H. (1982) Nucleotide sequence and exact localization of theneomycin phosphotransferase gene from transposon Tn5.Gene 19: 327-336.

Bolivar, F., Rodriguez, R.L., Greene, P.J., Betlach, M.C.,Heyneker, H.L., and Boyer. H.W. (1977) Construction andcharacterization of new cloning vehicles. II. A multipurposecloning system. Gene 2: 95-113.

Broome-Smith. J.K., Tadayyon, M., and Zhang, Y. (1990) Beta-lactamase as a probe of membrane protein assembly andprotein export. Moll\^icrobiol4: 1637-1644.

Bukhari, A.l.. and Ljungquist, E. (1977) Bacteriophage Mu;methods (or cultivation and use. In DNA Insertion Elements,Ptasmids and Episomes. Bukhari, A.I., Shapiro, J.A., andAdhya, S.L. (eds). Cold Spring Harbor, New York: ColdSpring Harbor Laboratory Press, pp. 749-756.

Casadaban, M.J , and Chou, J. (1984) In vivo formation ofgene fusions encoding hybrid beta-galactosidase proteins inone step with a transposable Mu-/ac transducing phage.Proc NatI Acad Set USA 81 : 535-539.

Castilho. B.A.. Olfson, P., and Casadaban. M.J. (1984) Plas-mid insertion mutagenesis and /ac gene fusion with mini-mubacteriophage transposons. J Bacterion58: 488-495.

Chang, C.N., Kuang, W.J., and Chen, E.Y. (1986) Nucteotidesequence of the alkaline phosphatase gene of Escherichiacoii. Gene 44; 121-125.

Chung, CT., Niemela, S.L., and Miller. R.H. (1989) One-steppreparation of competent Escherichia coli: transformationand storage of bacterial cells in the same solution. Proc NatIAcad Sci USA 86: 2172-2175.

Cianciotto, N.P.. Long, R., Eisenstein, B.I., and Engieberg,N.C. (1988) Site-specific mutagenesis in Legionella pneu-mophiia by allelic exchange using counterselectable ColEIvectors. FEh/iS Microbiot LettSS: 203-208.

Cianciotto, N.P.. Eisenstein, B.L, Mody, C,H,, Toews, G.B..and Engieberg, N.C. (1989) A Legiortella pneumophila geneencoding a species-specific surface protein potentiates theinitiation of intracellular infection. Infect Immun 57: 1255-1262.

Csonka, L.N., and Clark. A.J. (1979) Deletions generated bythe transposon Tn 70 in the srt recA region of the Escherichiaco//K-12 chromosome. Gene/Zcs 93: 321-343.

Dower, W.J.. Miller, J.F., and Ragsdale, C.W, (1988} High effi-ciency transformation of E. coli by high voltage electropora-tion. NucI Acids Res 16: 6127-6145.

Dreyfus. L.A.. and Iglewski, B.H. (1985) Conjugation-mediatedgenetic exchange in Legionella pneumophila. J Bacterioi161:80-84.

Dreyfus, L.A. and Iglewski, B.H. (1986) Purification and char-acterization of an extracellular protease of Legioneila pneu-mophiia. Infect Immun 5^: 736-743.

Engieberg, N.C., Drutz, D.J.. and Eisenstein, B.I. (1984a)Cloning and expression of Legionella pneumophiia antigensin Escherichia coli Intect Immun 44: 222-221.

Engieberg, N.C., Pearlman, E., and Eisenstein, B.I. (19B4b)Legionetia pneumophila surface antigens cloned andexpressed in Escherichia coti are translocated to the hostcell surface and interact with specific ant\-Legioneiia anti-bodies. J Bacteriol "iBO: 199-203.

Engieberg, N.C, Cianciotto, N.P., Smith, J., and Eisenstein,B.I. (1988) Transfer and maintenance of small, mobilizableplasmids with ColEI reptication origins in Legioneiia pneu-mophila. Plasmid 20: 83-91.

Engteberg, N.C, Carter C , Weber D.R., Cianciotto N.P.,Eisenstein B.I. (1989) DNA Sequence of mip. a Legioneiiapneumophiia gene associated with macrophage infectivity.tnfect Immun 57:1263-1270

Engieberg, N.C., Howe, D.C. Rogers, J.E., Arroyo, J., andEisenstein, B.I. (1991) Characterization of a Legionellapneumophila gene encoding a lipoprotein antigen. MolMicrobio!5: 2Q2^-2029

Figurski, D., Meyer, R., Mitler, D.S., and Hetinski, D.R. (1976)Generation in vitro of deletions in the broad host range ptas-mid RK2 using phage Mu insertions and a restrictionendonuctease. Gene i : 107-119.

Gay, P., Le Coq. D., Steinmetz. M., Berkelman, T., and Kado,C.I. (1985) Positive selection procedure for entrapment ofinsertion sequence elements in Gram-negative bacteria. J8acfer/o/164: 918-921.

Guiney, D.G. and Yakobson, E, (1983) Location and nucteotidesequence of the transfer origin of the broad host range pias-mid RK2. Proc NatI Acad Sci USA 80: 3595-3598.

Guyer, M.S.. Reed, R.R., Steitz, J.A., and Low, K.B. (1981)Identification of a sex-factor affinity site in E coli as gammadelta. Cold Spring Harb Symp Quant Biol 45 (R 1):135-140.

Hall, M.N., and Silhavy, T.J. (1981) Genetic analysis of themajor outer membrane proteins of Escherichia coli. AnnuRei/Gene/15: 91-142.

Hoffman, CS., and Wright, A. (1985) Fusions of secreted pro-teins to alkaline phosphatase: an approacfi for studying pro-tein secretion. Proc NatI Acad Sci USA 82: 5107-5111.

Kahmann. R., and Kamp, D. (1979) Nucleotide sequences ofthe attachment sites of bacteriophage Mu DNA. Nature 280:247-250.

Keen, M.G., Street, E.D., and Hoffman, P.S. (1985) Broad-host-range plasmid pRK340 delivers Tn5into the Legionellapneumophila chromosome. J Bacleriol ^62:1332-1335.

Maniatis, T., Fritsch, E.F., and Sambrook, J. (1982) Molecular

PhoA gene fusions in L. pneumophiia using Mi/dphoA 1839

Cloning. A Laboratory Manuai Cold Spring Harbor. NewYork: Cold Spring Harbor Laboratory Press.

Manoil, C , and Beckwith, J. (1985) TnphoA: a transposonprobe for protein export signals. Proc Natl Acad Sci USA 82;8129-8133.

Michaelis, S.. Inouye, H., Oliver, D., and Beckwith, J. (1983)Mutations that alter the signal sequence of alkaline phos-phatase In Esctierichia coli. J Bacteriol 154: 366-374.

Miller, J.H. (1972} Experiments in Molecular Genetics. ColdSpring Harbor, New York: Coid Spring Harbor LaboratoryPress.

Mintz, C.S. and Shuman. H.A. (1987) Transposition of bacte-riophage Mu in the Legionnaires' disease bacterium. ProcNatl Acad Sci USA 84: 4645-^649.

Multer, H.E. (1981) Enzymatic profile of iegionetia pneu-mophitia. J Clin Microbioi 13: 423-426.

Pato, M.L. (1969) Bacteriophage Mu. In fvJobile DNA. Berg.D.E.. and Howe. M.M. (eds). Washington, DC: AmericanSociety for Microbiology, pp. 23-51.

Pearlman, E., Engieberg. N.C.. and Eisenstein. B.I. (1985)

Identification of protein antigens of Legionetta pneumophitaserogroup 1. tnfect immun 47: 74-79.

Peariman. E.. Jiwa. A.H., Engieberg. N.C., and Eisenstein. B.I.(1988) Growth of Legionetta pneumophita in a humanmacrophage-like (U937) cell line. Microbiai Pathog 5: 87-95.

Silhavy, T.J., Berman, M.L.. and Enquist, L.W. (1984) Experi-ments with Gene Fusions- Coid Spring Harbor. New York:Cold Spring Harbor Laboratory Press.

Sutcliffe, J.G. (1979) Complete nucleotide sequence of theEscherichia coli plasmid pBR322. Cold Spring Harb SympQuant Biot43 (P\^): 77-90.

Wiiliams, A.E., Maskell. D.J.. and Moxon, E.R. (1991) Relation-ship between intracellular survival in macrophages and viru-lence of Haemophilus inftuenzae type b. J Inf Dis 163:1366-1369.

Yanisch-Perron. C . Vieira, J., and Messing. J. (1985)Improved Ml 3 phage cloning vectors and host strains:nucleotide sequences of the M13mp18 and pUCi9 vectors.Gene 33:103-119.