MICROBIOLOGICAL REVIEWS, Mar. 1992, p. 32-600146-0749/92/010032-29$02.00/0Copyright X) 1992, American Society for Microbiology

Virulence Factors of the Family LegionellaceaeJOHN N. DOWLING,`* ASISH K. SAHA,2 AND ROBERT H. GLEW2

Department of Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania 15261,1 and Departmentof Biochemistry, University ofNew Mexico School of Medicine, Albuquerque, New Mexico 871312

INTRODUCTION .................................................... 32

Background.................................................... 32

Purpose and Scope.................................................... 33

Approach.................................................... 34

INTRACELLULAR BIOLOGY OF LEGIONELLA SPP....................................................34

Pathology.................................................... 34

Polymorphonuclear Leukocytes.................................................... 35

Mononuclear Phagocytes .................................................... 36

Alveolar macrophages.................................................... 36

Peripheral blood monocytes .................................................... 36

Amoebae............................. 37

Multiplication in Phagocytes Required for Infection and Disease....................................................37

Phagocytosis and Phagosome-Lysosome Fusion .................................................... 38

BIOCHEMICAL OVERVIEW .................................................... 39

PROPERTIES OF POTENTIAL VIRULENCE FACTORS .................................................... 39

Peptide Toxin.....................................................39

Enzymes .................................................... 41

Phosphatases .................................................... 41

Phospholipase C ................................................... 42

Proteases .................................................... 43

(i) General .................................................... 43

(ii) Zinc metalloprotease .................................................... 43

Enzymes that scavenge reduced-oxygen metabolites.................................................... 46

Protein kinases.................................................... 47

Legionella Cell Envelope .................................................... 48

Cell Surface Legionella Proteins............................... 48

MOMP................................48

Mip protein................................49

Heat shock protein............................... 50

AVIRULENT LEGIONELLA MUTANTS ............................... 50

CONCLUSIONS AND FUTURE RESEARCH ................................... 52

ACKNOWLEDGMENTS ................................... 54

ADDENDUM................................... 54

REFERENCES ................................... 54

INTRODUCTION

Background

During the year following the epidemic of pneumonia at a

Legionnaires' convention in Philadelphia in 1976 (77), thecausative bacterium was isolated (132) and named Legion-ella pneumophila (32). Within 2 years, a bacterium respon-sible for causing pneumonia in immunocompromised pa-

tients in Pittsburgh, Pa. (140), and Charlottesville, Va. (168),was isolated and proved to be a second Legionella species,

L. micdadei (88). Since that time numerous additional spe-

cies have been identified, and it has become apparent thatsome species contain more than one serogroup based on

agglutinating surface antigens. Although proposals havebeen made to add additional genera, by common usage atpresent Legionella is the only designated genus in the family,Legionellaceae. As of mid-1991, a total of 32 Legionella

* Corresponding author.

species which contained 51 serogroups had been named (16),and this list is sure to grow further in the future. Sixteen ofthe Legionella spp., comprising 34 serogroups, have beenreported as pathogenic for humans (16), whereas the othershave as yet been obtained only from environmental reser-

voirs. However, it is possible that any species could cause

human disease under the appropriate conditions.The predominant clinical manifestation of Legionella in-

fection is pneumonia. Approximately 85% of cases of le-gionellosis are due to L. pneumophila, with about 50% of alldisease due to L. pneumophila serogroup 1 and 10% toserogroup 6 (164). The remainder of cases are due to otherserogroups of L. pneumophila, L. micdadei, and a number ofother species. It is not clear whether this distribution is dueto greater inherent virulence of the more frequently isolatedstrains or, more likely, the prevalence of the various strainsin the environment. The legionellae also cause an acute,febrile, nonpneumonic illness referred to as Pontiac fever.Although Pontiac fever follows the inhalation of environ-mental legionellae, as does pneumonia, there is no tissue

32

Vol. 56, No. 1

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 33

invasion and the disease is self-limited. The pathogenesis ofPontiac fever is obscure, and viable legionellae may not berequired for its production. Alternatively, it has been pro-posed that Pontiac fever is produced by Legionella strainsthat are unable to multiply in human cells (73). Since thepathogenesis of Pontiac fever and Legionella pneumonia areso obviously different, the former will not be consideredfurther in this review.

Although the source of the legionellae that infect humansis contaminated environmental water, the precise nutri-tional, pH, and temperature requirements that are necessaryto cultivate Legionella strains in vitro suggest that it is not afree-living aquatic bacterium. It has been found that legionel-lae infect free-living amoebae and ciliated protozoa in vitro(1, 74, 170, 189) and, probably, in the natural environment(6, 31). Legionellae contained in amoebae, and especially inamoebal cysts, could survive environmental temperatureextremes, chlorination, and other adverse conditions. Undercertain conditions in vitro, ingested legionellae multiplywithin the vacuoles of free-living amoebae, so that amoebaemay serve to amplify the numbers of legionellae in theenvironment. Furthermore, infected amoebae and amoebalvesicles containing legionellae would be present in the driftfrom contaminated aquatic environments and provide thevehicles whereby concentrated infectious particles could bedelivered to humans.

Purpose and Scope

Whereas bacteria of the genus Legionella have emerged asrelatively frequent causes of pneumonia, the mechanismsunderlying their pathogenicity are poorly understood. Thelegionellae are facultative intracellular pathogens which mul-tiply within the phagosome of monocytes (107, 199) andalveolar macrophages (120, 142). Furthermore, the legionel-lae are not killed efficiently after being phagocytized bypolymorphonuclear leukocytes (106, 198). Since the le-gionellae have clearly been shown to be intracellular patho-gens, their pathogenicity for humans may largely depend onthe outcome of the interaction between the bacterium andthe professional phagocytes it encounters in the host. Be-cause of the intracellular location of Legionella spp., atten-tion has naturally been focused on the cell-mediated immuneresponses to infection (103). The influence of specific cell-mediated immune mechanisms on the interaction of Legion-ella spp. with alveolar macrophages and mononuclear phago-cytes no doubt is important in the clearance of the bacterialater in primary infection and in the immune host. However,the initial interactions with host phagocytic cells, especiallyneutrophils, may be particularly important in infection owingto facultative intracellular bacteria with relatively shortgeneration times such as Legionella spp. Unless such organ-isms are controlled, they would reach overwhelming num-bers before specific immunity has time to develop.The aim of this review is to relate the basic biology,

particularly the biochemistry, of Legionella spp. to its patho-genesis at the cellular level. Although much is known aboutvirulence factors of extracellular pathogens, which generallyserve to prevent phagocytosis mediated by host humoralfactors, the mechanisms of pathogenicity used by intracel-lular parasites have not yet been fully elucidated (33). Avariety of mechanisms have been implicated in the intracel-lular survival of different bacteria, including extraphagoso-mal location, resistance to oxidative and nonoxidative killingmechanisms, inhibition of phagosome-lysosome fusion, andinterruption of phagocyte activation and the subsequent

production of bactericidal oxygen metabolites (56). Thefunctional defects that might permit the intracellular survivalof the legionellae have remained an enigma until recently.The initial investigations of phagocytes that ingest Legion-ella spp. did not detect any of the variety of mechanismsused by other organisms to evade intracellular destruction. Ithas been demonstrated that both L. pneumophila and L.micdadei are phagocytized and remain within the phago-some. Furthermore, both species are susceptible to killing invitro by H202 and other bactericidal oxygen metabolitesproduced by phagocytic cells (58, 122, 123). It had beenthought that intracellular survival in phagocytes might beexplained by the fact that phagosome-lysosome fusion isinhibited following the ingestion of Legionella spp. It nowappears, however, that fusion may be inhibited by only asingle strain (Philadelphia 1) of L. pneumophila serogroup 1(99); no defect in phagosome-lysosome fusion has beenfound in monocytes or neutrophils which have phagocytizedother species, other L. pneumophila serogroups, or evenother strains of L. pneumophila serogroup 1 (159). Uponinitial contact L. micdadei was a potent stimulus of theneutrophil and monocyte metabolic burst (60, 61). However,it was recently found that within 30 min following theingestion of L. micdadei, the activation of neutrophils andmonocytes in response to both soluble and particulate stim-uli is profoundly impaired and the bactericidal activity ofthese cells for Staphylococcus aureus and Escherichia coli isattenuated (57, 59). These data suggest that one or moreLegionella bacterial cell-associated factors have a strikinginhibitory effect on phagocyte activation and, consequently,on subsequent antibacterial functions.One strategy to elucidate the pathogenic potential of the

legionellae at the cellular level is to attempt to discover thebacterial factors which are responsible for blocking theactivation of phagocytes. Two factors that are elaborated bythe legionellae and that inhibit phagocyte activation havebeen described. We found that L. micdadei bacterial cellscontain a phosphatase which blocks superoxide anion (02-)production by stimulated neutrophils (172), thus reproducingthe refractoriness to stimulated activation which is seenfollowing ingestion of L. micdadei. We have shown that theLegionella phosphatase disrupts the formation of criticalintracellular second messengers in neutrophils (175), provid-ing a logical mechanism by which the phosphatase blocksphagocyte activation. The second moiety known to blockneutrophil oxidative metabolism in response to variousagonists is the Legionella (cyto)toxin (79, 121). There isconsiderable evidence that the Legionella toxin is importantin the pathogenesis of cellular infection. However, no grouphas yet succeeded in purifying the toxin to homogeneity, aprerequisite for defining its mechanism of action and estab-lishing its role as a virulence factor.

In addition to the toxin and phosphatase, several othermoieties elaborated by the legionellae may serve as viru-lence factors by promoting cell invasion or intracellularsurvival and multiplication. Genetic studies show that a cellsurface protein named Mip is necessary for the efficientinvasion of monocytes (45). The mechanism by which Mippromotes bacterial uptake by phagocytes is unknown. Apossible role for a Legionella phospholipase C as a virulencefactor is still largely theoretical. L. micdadei contains aprotein kinase which catalyzes the phosphorylation of eu-karyotic substrates, including phosphatidylinositol (PI) andtubulin (174). Although we have demonstrated that theenzyme phosphorylates P1 in the plasma membrane of intactneutrophils, the function and significance of the L. micdadei

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

34 DOWLING ET AL.

kinase are obscure. However, since the phosphorylation ofeither PI or tubulin might compromise phagocyte activationand bactericidal functions, the enzyme may well be a viru-lence factor.When a purified Legionella exoprotease was aerosolized

into the lungs of guinea pigs, it induced lesions resemblingthose of Legionella pneumonia and caused the death ofmany of the animals (49). Both the cytotoxic nature of theexoprotease and the detection of antibodies in human con-valescent-phase serum reactive with the exoprotease (114,157) suggested that this protein plays a role in the pathogen-esis of legionellosis. However, recent work with a geneti-cally engineered strain has convincingly shown that theprotease is not necessary for intracellular survival or viru-lence (29, 184). Other Legionella moieties, including lipo-polysaccharide (LPS), the major outer membrane protein(MOMP), the heat shock protein, oxygen-scavenging en-zymes, and a hemolysin termed legolysin, which couldconceivably be involved in virulence, will be mentioned onlybriefly since there is no evidence that any of these promoteinvasiveness or intracellular survival.

Approach

For the reader who is not as familiar with Legionella spp.as facultative intracellular pathogens, the vast amount ofdata concerning the interactions of the legionellae with hostphagocytic cells will be briefly reviewed. Since it is onlyrecently that functional lesions in phagocytes infected withLegionella spp. have been identified which explain theability of the bacterium to survive in these cells, this is anopportune time to summarize these studies to provide thebackground for discussing the bacterial factors which mightcause these impairments. Individual putative virulence fac-tors will then be reviewed. The various factors have beencharacterized to a greater or lesser extent by quite differentapproaches. For example, the evidence suggesting that thephosphatase might be a virulence factor for intracellularbacterial survival is primarily biochemical, whereas for theMip protein it is genetic and the mechanism of action isunknown. In contrast, the exoprotease, which was consid-ered a possible virulence factor on pathological and bio-chemical grounds, does not appear to function in thatcapacity now that genetic analyses are complete. These datawill be critically reviewed, for it is important to point outwhere the present evidence falls short of showing that aputative virulence factor contributes to cellular parasitism.With the multiplicity of Legionella species and serogroups

that have been described, it is possible that progress inidentifying the determinants of a complex process such aspathogenicity will be made by comparing and contrasting theintracellular biology of different species and serogroups.When relevant biological differences are established, thebiochemical basis for the difference can be sought. Forexample, in contrast to L. pneiumophila, L. mnicdadei showsa predilection for causing disease almost exclusively inimmunosuppressed patients (63, 140, 168). This suggestseither that L. micdadei possesses virulence factors differentfrom those of L. pneurnophila or that the two species differin their interaction with host defenses. Most of the researchhas utilized one of the serogroups of L. pneirnophila or L.micdadei. One of the goals of this review is to integrate thework on the intracellular biology which has been accom-plished with various species so that the similarities anddifferences may serve as the basis for an examination of thevirulence factors of the genus. Unfortunately, however,

each research group involved in studying putative virulencefactors has tended to use a single Legionella strain. There-fore, in most instances it can be said only that a particularvirulence determinant appears to be pertinent for that strainand that the relevance to other Legionella spp. and sero-groups has not been addressed.

It is relatively easy to derive attenuated Legionella mu-tants by passing the wild-type bacteria on substandardmedia. The properties of these avirulent mutants have beencompared with those of their virulent parental strains in anumber of studies in an attempt to identify the attributeassociated with virulence. A review of these investigationsindicates that the mechanism leading to avirulence is dif-ferent for the various mutants which arise during agarpassage. These findings may support the possibility thatvirulence in the legionellae is multifactorial. However, it isnot clear whether the various mechanisms identified by thisapproach are the ones which contribute to the virulentphenotype of the naturally occurring bacteria.

INTRACELLULAR BIOLOGY OF LEGIONELLA SPP.

Pathology

Legionella spp. produce an acute purulent pneumonia inwhich the alveoli are filled with polymorphonuclear neutro-phils, macrophages, fibrin, and erythrocytes (203). Thenumbers of neutrophils and macrophages are usually ap-proximately equal, but one or the other cell type maypredominate in individual cases. Whether monocytic phago-cytes or neutrophils predominate does not seem to be relatedto the duration of the pneumonia before tissue is obtained forexamination (203). A dramatic microscopic feature of theexudate in many cases is the lytic destruction of the inflam-matory cells, a process which has been termed leukocyto-clastic. There is a close association of the bacteria, orbacterial antigen, with the cellular component of the inflam-matory infiltrate. The majority of bacteria are seen withininflammatory phagocytic cells; only a few are extracellular(84, 203). A careful immunofluorescence study by Hicklin etal. (91) documented the close association of L. pneiumophilaantigen with the cellular component of the inflammatoryinfiltrate and the rarity of bacilli in alveoli which were merelyedematous or congested. Most of the intracellular bacteriaare found within membrane-bound cytoplasmic vacuoles(phagosomes) or, later in infection following extensive intra-cellular bacterial proliferation, within the disrupted cytosolof the cells (40).

Similar histologic changes were observed in the lungs ofguinea pigs inoculated with Legionella spp. intranasally(113), intratracheally (139, 150, 201), by aerosol (8, 9, 55), oreven intraperitoneally (41). In the latter case the pneumoniapresumably eventuates following bacteremia. Although oc-casional intracellular bacteria could be seen to be dividingwithin inflammatory cells when infected pulmonary tissuewas examined by electron microscopy (41, 113), it could notbe determined whether the bacteria multiplied intracellularlyor were phagocytized after extracellular multiplication.Therefore, the initial approach of a number of investigatorsto understanding the pathogenesis of Legionella infection atthe cellular level was to study in vitro the interactions of thebacterium with the three types of professional phagocytes itencounters in the host during natural infection (54), namely,polymorphonuclear neutrophils, alveolar macrophages, andperipheral blood monocyte-macrophages.

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 35

Polymorphonuclear Leukocytes

Antibody is required to fix the third component of com-plement (C3) to L. pneumophila, and only complement-coated bacteria are efficiently bound to or ingested by humanneutrophils in vitro (106, 198). Even when the bacteria werepreincubated with a source of specific antibody and comple-ment, human neutrophils killed only about 0.5 log of aninoculum of virulent L. pneumophila (106). Under the sameconditions neutrophils reduced the number of a serum-resistant, encapsulated strain of E. coli by 2.5 logs. Separa-tion of the neutrophil-associated and unassociated bacteriarevealed that the majority of surviving legionellae wereneutrophil associated, suggesting that the neutrophils fail tokill legionellae after they are bound and presumably inter-nalized.

In contrast, human neutrophils phagocytized virulent L.micdadei with only normal human serum (complement) asthe opsonin; specific antibody was not required (198). How-ever, under these conditions significantly fewer L. inicdadeicells (75%) were phagocytized than S. aureus cells (98%).Use of heat-inactivated serum abolished phagocytosis.There was essentially no killing of ingested L. micdadei cellsby neutrophils, compared with the killing of 97% of S.aureus cells that had been phagocytized. In contrast, 89% ofL. micdadei cells rendered avirulent by multiple passages onagar were killed by neutrophils.

Fitzgeorge et al. (76) examined the effects of neutrophildepletion engendered by the administration of anti-poly-morph serum on Legionella infection in guinea pigs. Elimi-nation of polymorphonuclear leukocytes lowered the dose ofaerosolized L. pneuimophila necessary to cause pneumonia,increased the number of bacteria in the lungs, and producedmuch higher mortality. Neutrophil depletion did not changethe extent or nature of the pulmonary lesions, except thatneutrophils were absent from the infiltrate. These resultsindicate the importance of neutrophils in the defense of thelungs against L. pneiumophila and also suggest that neutro-phils and their enzymes are not responsible for the pulmo-nary lesions in legionellosis.Working with L. inicdadei, we have developed an in vitro

system which begins to elucidate how the legionellae escapeintracellular destruction by neutrophils and monocytes. Ini-tial experiments revealed that neutrophils phagocytizing L.micdadei appear to function normally. Neutrophil oxidativemetabolism, as measured by reduction of Nitro Blue Tetra-zolium dye, luminol-enhanced chemiluminescence, oxygenconsumption, and hexose monophosphate shunt (HMPS)activity, was stimulated by L. rnicdadei ingestion to thesame extent as when opsonized zymosan (OPZ), S. aiueuis,or E. coli served as the phagocytic stimulus (61). L. ,nicdadeiactually stimulated significantly more O,- production thandid either S. aureuis or E. coli at bacterium-to-neutrophilratios of 10:1, 100:1, and 1,000:1. Fusion of phagosomes withboth primary (azurophilic) and secondary (specific) granulesduring phagocytosis of L. micdadei was equivalent to thatobtained with S. alrells, which was used as the positivecontrol. Furthermore, quantitation of extracellular my-eloperoxidase and lysozyme indicated equivalent release ofgranule enzymes by neutrophils ingesting L. mnicdadei and S.aureus.However, within 30 min after having phagocytized L.

micdadei, the activation of neutrophils in response to solubleand particulate stimuli is markedly depressed when com-pared with either nonphagocytizing control neutrophils orneutrophils which have ingested S. aurelus or E. (oli (59).

Once ingested, L. inicdadei inhibited neutrophil chemotaxisin response to the chemotactic peptide, N-formyl-methionyl-leucyl-phenylalanine (fMLP). The quantity of 02 producedin response to fMLP by neutrophils which had phagocytizedL. inicdadei 30 min or more previously was markedlydepressed compared with either nonphagocytizing neutro-phils or neutrophils ingesting S. aureus. The attenuation ofstimulated 02 production varied directly with the numberof L. inicdadei phagocytized and with the time elapsedfollowing ingestion of the bacteria. The inhibitory effect of L.micdadei on neutrophil activation did not depend on theprior phagocytosis of viable bacteria, as neutrophils ingest-ing heat-killed L. inicdadei had depressed fMLP-stimulated02 production compared with either nonphagocytizing neu-trophils or neutrophils which had phagocytized heat-inacti-vated S. au-reuis. However, phagocytosis of either viable ordead bacteria was required for the inhibition of neutrophilactivation. When neutrophils were exposed to L. micdadeiat 4°C or in the presence of complement-depleted serum,conditions which preclude phagocytosis (180, 198), the sub-sequent production of fMLP-stimulated 02- was normal.

Neutrophil phagocytic and bactericidal capacities weredepressed concomitantly with the inhibition of activationfollowing L. ,nicdadei ingestion (59). The percentage ofneutrophils which phagocytized viable S. aureus was signif-icantly lower following ingestion of heat-inactivated L.micdadei when compared with either neutrophils which hadpreviously ingested heat-killed S. alureuis or nonphagocytiz-ing neutrophils. Killing of staphylococci was likewisereduced, from virtually 100% killed by nonphagocytizingneutrophils and neutrophils which had previously ingestedheat-killed S. aiurelus to 37% killed by neutrophils which hadpreviously ingested killed L. micdadei. These results may becontrasted with the results of an experiment done by Hor-witz and Silverstein (106) in which neutrophils simulta-neously phagocytizing L. pneuimophila and E. coli killed thelatter as effectively as neutrophils ingesting only E. coli. Thisresult again demonstrates that some time must elapse fol-lowing the phagocytosis of Legionella cells before the inhi-bition of neutrophil function is evident.Thus, we have demonstrated that L. micdadei, once

ingested, significantly inhibits a variety of neutrophil func-tions including chemotaxis, phagocytosis, stimulated 2-production, and bactericidal activity (59). Several character-istics of this inhibitory effect are noteworthy. First, ingestionof the organism is required; decreasing phagocytosis exper-imentally markedly reduced the degree of inhibition ofneutrophil function. This suggests that the factor(s) respon-sible for the inhibition is released from the bacteria under theconditions existing in the phagolysosome. Second, viablebacteria are not required for inhibition to take place, sincebacteria killed by heating are effective inhibitors of neutro-phil function. Thus, a phagocyte may be able to kill the firstLegionella cell ingested and yet could still become impairedand subsequently be incapable of handling additional phago-cytized bacteria. On the other hand, the ingestion of viableLegioiell/a cells impairs neutrophil function to a somewhatgreater extent than does ingestion of heat-killed bacteria(59). This implies that the inhibitory effect is multifactorial,involving both a heat-stable and a heat-labile factor(s).Third, inhibition was not stimulus specific; significant inhi-bition of neutrophil activation occurred following stimula-tion with fMLP, OPZ, or phorbol myristate acetate (PMA).However, the inhibition of infected neutrophils stimulatedwith fMLP was greater than that produced in response toPMA. We conclude that the legionellae must possess the

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

36 DOWLING ET AL.

capability to actively inhibit the phagocyte activation se-quence leading to the respiratory burst, perhaps at multiplesites in the sequence.A recent, preliminary report indicates that infection of

neutrophils with L. pneumophila inhibited the rise in intra-cellular Ca2" in response to fMLP (34). Both a virulent andan avirulent strain caused marked suppression of the Ca2"response at early time points, but the virulent bacteriumproduced significantly greater suppression at later times. E.coli caused no inhibition of the fMLP-stimulated rise inintracellular Ca2+. Thus, it appears that neutrophils whichhave phagocytized L. pneumophila and L. micdadei demon-strate similar blockade of the intracellular activation path-ways.

Mononuclear Phagocytes

Alveolar macrophages. Kishimoto et al. (115) studied theinteraction between L. pneumophila and cynomolgus mon-key alveolar macrophages that were obtained by bronchoal-veolar lavage and cultured in vitro. Electron micrographsobtained 3 h after infection showed that about 5% of thealveolar macrophages contained intracellular bacteria. At 24h later, many macrophages contained distended vacuolesfilled with L. pneumophila. Multiplication of the legionellaewas so rapid and extensive that the cytoplasm of somemacrophages became filled with vesicles containing bacteriaand the cells were ultimately destroyed. Jacobs et al. (109)investigated the interaction between L. pneiumophila andpigtail monkey alveolar macrophages by using methods thatdistinguished between viable and dead bacteria. In theabsence of antibody the alveolar macrophages phagocytizedabout 1% of the legionellae in the inoculum. The macro-phages killed the majority (60 to 97%) of the ingestedbacteria within 30 min. Phagocytosis of L. pneutmophila wasassociated with a respiratory burst, as visualized by NitroBlue Tetrazolium reduction around ingested bacteria. Killingof the alveolar macrophage-associated Legionella cells wasinhibited by mannitol and by the combination of superoxidedismutase (SOD) and catalase, but not by either of these twoenzymes alone. These results indicated that the killing ofLegionella cells by alveolar macrophages is mediated by thehydroxyl radical (OH'), which is formed from O- and H,O,in the presence of Fe3". The virulent legionellae whichsurvived after the early killing multiplied more than 2 logs inthe 96 h following infection. In contrast, avirulent L. pneii-mophila replicated more slowly over the same period.

L. pneiumophila Philadelphia 1 also multiplied rapidly inhuman alveolar macrophages obtained by bronchoalveolarlavage and cultivated as monolayers in vitro (142). Specificantibody combined with complement promoted the phago-cytosis of L. pneumophila, but alveolar macrophages wereable to kill less than 10% of an inoculum even in the presenceof both opsonins. L. pneiumophila multiplied 2.5 to 5 logsover 3 days, and at the peak of bacterial multiplication themacrophage monolayers were destroyed. Electron micros-copy showed that the bacteria were located intracellularlywithin membrane-bound vacuoles which were studded withribosomes. In the absence of serum, L. pneiinophili wasalso taken up by and multiplied rapidly in guinea pig and ratalveolar macrophages obtained by bronchoalveolar lavage(69). No extracellular multiplication of the bacteria oc-curred. Growth was inhibited when the alveolar macro-phages were pretreated with cytochalasin D. which pre-vented phagocytosis of the legionellae.

In the absence of any opsonins, L. inicdadei cells were

taken up by and multiplied within guinea pig alveolar mac-rophages, so that the cell-associated titer increased morethan 100-fold over 20 h (120). L. micdadei opsonized witheither complement or specific antibody multiplied withinalveolar macrophages to the same extent as did unopsonizedbacteria. Treatment of the macrophages with cytochalasin Bor incubation at 4°C reduced the percentage of alveolarmacrophages containing intracellular bacteria, confirmingthat the uptake of Legionella cells by alveolar macrophagesis dependent on an intact macrophage microfilament systemand occurs by a process compatible with phagocytosis.Guinea pigs infected with L. micdadei by intratrachealinoculation and then treated with antimicrobial agents wereimmune to subsequent challenge with an otherwise lethaldose of L. micdadei. However, the growth curves of L.micdadei in alveolar macrophages obtained from these im-mune animals and infected in vitro were identical to those inalveolar macrophages from naive guinea pigs. On the otherhand, cell-free supernatants from blood mononuclear cellsstimulated with concanavalin A (ConA) inhibited the multi-plication of L. pneumophila in alveolar macrophages (142).This indicates that lymphokines may activate the alveolarmacrophages in the lungs to restrict L. pneumophila multi-plication more effectively.

Peripheral blood monocytes. Horwitz and Silverstein (105)studied the interaction of virulent L. pneumophila Philadel-phia 1 and human peripheral blood monocytes in vitro. L.pneiurnophila cells multiplied several logs when incubatedwith the monocytes. Peak growth was associated with de-struction of the monocyte monolayer. L. pneumophila mul-tiplied only in the adherent cell population, indicating mul-tiplication in monocytes rather than lymphocytes. Byelectron microscopy, L. pneumophila was found in mem-brane-bound cytoplasmic vacuoles studded with host cellribosomes. Peripheral blood monocytes bound more thanthree times as many virulent L. pneumophila in the presenceof both specific antibody and complement than when op-sonized with complement alone (107). Monocytes requiredboth antibody and complement to kill any L. pneumophilacells. However, even in the presence of both opsonins,monocytes killed only 0.25 log of an inoculum of virulent L.pnelunophila. The surviving bacteria multiplied several logsin the monocytes over 4 days following infection, regardlessof whether they were opsonized with specific antibody inaddition to complement.

Similarly, virulent L. micdadei multiplied within humanperipheral blood monocytes following phagocytosis (199).Intracellular bacterial growth was rapid, with a 100-foldincrease to the peak titer occurring over 12 h. Under thesame conditions, L. pneirmophila grew more slowly, reach-ing peak titer in 48 h, but L. pnelimophila multiplied to ahigher final titer than L. rnicdadei did. Electron microscopyafter 18 h showed that the L. micdadei cells were intracel-lular in normal-appearing phagosomes. At the same time,intracellular L. pneio(nophila organisms were located inphagosomes studded with ribosomes. L. inicdadei activatedthe complement system and was opsonized by C3. However,the use of complement-depleted serum as the opsonic sourcehad no effect on the ingestion or growth of L. micdadei inmonocytes.As with neutrophils, oxidative metabolism, as measured

by chemiluminescence, oxygen consumption, or HMPS ac-tivity, in monocytes actively phagocytizing L. micdadei wasequivalent to or greater than in monocytes ingesting S.aiur eius or OPZ (60). Preliminary assays carried out onmonocytes which had previously ingested heat-killed L.

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 37

micdadei revealed a similar attenuation of function as wasobserved with neutrophils, although the degree of inhibitionwas more variable (57). In monocytes which had ingested L.micdadei, chemotaxis was inhibited by 50% compared withmonocytes which had ingested S. aiureius or nonphagocytiz-ing monocytes (P < 0.01). Oxidative activity of monocyteswhich had phagocytized L. micdadei in response to OPZ, asmeasured by HMPS activity, was reduced by 94% comparedwith controls (P = 0.01). However, chemiluminescence,another parameter of oxidative function, was reduced byonly 50% (P < 0.02). Unlike neutrophils, monocytes whichhad ingested L. micdadei demonstrated no inhibition ofphagocytosis of S. aureus and only minimal reduction in thekilling of S. aureus (80% of ingested bacteria killed versus100% killed; P < 0.05). Whether the inhibition of monocytefunction would be greater following phagocytosis of viablerather than heat-killed L. micdadei cells remains to bedetermined.Horwitz and Silverstein (108) showed that soluble prod-

ucts of stimulated lymphocytes were able to at least partiallyactivate mononuclear phagocytes against L. pneirnophila.Peripheral blood monocytes exposed to the products ofConA-treated mononuclear cells (primarily lymphocytes)exhibited decreased phagocytosis of L. pnelonophila. Al-though these cells were still unable to kill ingested L.pneumophila cells, the intracellular multiplication of thebacilli was inhibited. Similarly, peripheral blood monocytesactivated with human recombinant gamma interferon inhib-ited the multiplication of L. pneurmophila (20). In the ab-sence of complement and antibody, neither gamma interfer-on-activated monocytes nor unactivated monocytes killed L.pneumophila. Even in the presence of both opsonins,gamma interferon-activated monocytes killed only 0.5 log ofan inoculum, which was not more than nonactivated mono-cytes did. These in vitro results were extended to show that,when incubated with formalin-killed L. pneirnophila, theperipheral blood mononuclear cells from patients who hadrecovered from Legionnaires' disease also produced cyto-kines which inhibited the growth of L. pnei,mophila infreshly explanted monocytes (100).

Amoebae

Rowbotham (170) was the first to demonstrate that le-gionellae are ingested by free-living amoebae of the generaAcanthamoeba and Naegleria in vitro. There are obviousparallels between Legionella uptake and infection of free-living amoebae and human phagocytic cells. After ingestionthe legionellae within amoebae are confined to vacuoles and,under appropriate conditions, multiply to large numbersuntil the amoebal cell ruptures. Interestingly, the intracellu-lar vacuoles containing a strain of L. pneirnophila serogroup1 (not further defined) demonstrated the same alignment ofmitochondria and ribosomelike structures along the vacuolemembrane (143) as was described for human phagocytephagosomes containing the Philadelphia 1 strain (98, 105,199). However, these morphologic findings and the intracel-lular multiplication of Legionella organisms were related tothe medium in which the amoebae were maintained. The"abnormal" vacuoles were observed for amoebae main-tained in saline, in which intracellular bacterial multiplica-tion was absent, and apparently were not observed inamoeba culture medium, in which intracellular Legionellamultiplication occurred. The interaction may also be influ-enced by the species of the amoebae and the legionellae (171)and, as with human phagocytes, the virulence of the Legion-

ella strain employed (171, 189). For example, electron-microscopic studies have revealed two forms of L. micdadeibacteria: a smooth, thick-walled, banded form containingpoly-3-hydroxybutyrate granules, and thinner, rumpled-wall, unbanded bacteria with little or no poly-p-hydroxybu-tyrate (85). The band appears to represent an unusually thicklayer of peptidoglycan. Since the banded form was prevalentin infected human lungs and among L. micdadei cells prop-agated in cell cultures, whereas the unbanded form waspredominant in agar-grown bacteria, the former may repre-sent a more virulent form of the bacterium. It was found thatthe banded bacilli readily infect and multiply in Acan-thamnoeba cells while the unbanded L. micdadei organismseither do not infect or do not grow in these amoebae (171).The similarity of the interaction of legionellae with amoe-

bae and human phagocytic cells suggests that the formermay be useful as a model to elucidate the latter. However,among amoeboid protozoa, phagocytosis is not only themeans of defense against foreign cells and particles, but alsothe principal mode of ingesting food. Thus, there is acomplex relationship between digestion of ingested bacteriaand bacterial multiplication following ingestion. Low envi-ronmental temperatures favor digestion of ingested legionel-lae (1, 141), whereas higher temperatures favor infection andintracellular multiplication (1, 170). Moreover, virulence foramoebae does not necessarily coincide with virulence forhuman phagocytes. L. anisa and the amoeba Hartmannellav'ermiforrnis were both isolated from an indoor fountainwhich had been implicated as the source of an outbreak ofPontiac fever. The L. anisa strain multiplied in H. vermi-ftrinis in vitro, but failed to infect guinea pigs, cultures of thehuman mononuclear cell line U937, or human peripheralblood monocytes (73). These results suggested that the L.anisa strain could reach high concentrations in the environ-ment as a result of growth in H. vermiformis, but producedPontiac fever rather than pneumonia because of the inabilityto multiply in human phagocytic cells. When a large panel ofLegionella strains of various species were cocultivated withH. veriniforinis at 37°C in a system in which killed Pseudo-monas paiicirnobilis cells served as food for the amoebae,the local environmental strains were more likely to multiplyin the amoebae than were strains from other sources, includ-ing clinical isolates (194). However, it is possible that thedifference in the ability of the legionellae to multiply in thecocultures was related to differences in passage histories ofthe strains. Almost all the environmental strains had beentransferred on charcoal yeast extract agar fewer than fivetimes, while the passage history of most of the other strainswas unknown. More concerning the cell biology of theamoebae must be understood before the various factorsinvolved in the Legionella-amoeba interaction which deter-mine whether the amoebae will inhibit or support the growthof legionellae can be elucidated. From the data available atpresent, it does not appear likely that the virulence determi-nants for infection of, and multiplication in, amoebae andphagocytes are necessarily the same.

Multiplication in Phagocytes Required for Infectionand Disease

The hypothesis that infection, and ultimately disease,depends on intracellular multiplication of the legionellae isbased on three lines of evidence (43). First, in an experimen-tal model of aerosol infection in guinea pigs, Davis et al. (54)found a rapid increase in the number of viable Legionellacells obtained by pulmonary lavage during the 24 h following

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

38 DOWLING ET AL.

infection; by 16 h 86% of the viable bacteria were associatedwith the cell pellet. From 24 to 48 h after infection there wasa rapid influx of neutrophils which resulted in a mixedinflammatory cell population that coincided with a levelingoff of the Legionella titer. Most viable bacteria were found inthe cell fractions containing the largest number of alveolarmacrophages, whereas fewer viable bacteria were found infractions containing predominantly neutrophils. Large num-bers of morphologically intact bacteria were present only inalveolar macrophages. Morphologically intact bacilli werealso found in neutrophils, but the majority of bacteria inthese cells appeared structurally damaged. Similar resultswere obtained by Jepras et al. (112), who compared theeffects of aerosol infection with virulent and avirulentLegionella strains in guinea pigs. The virulent strain multi-plied rapidly in the lungs, reaching a peak titer of over 1011viable bacteria per lung in 6 days, whereas the avirulentbacteria were unable to replicate and were cleared between14 and 21 days following infection. Lung lavages performedon infected animals showed that the virulent legionellae weremainly intracellular, whereas the avirulent bacteria werepredominantly extracellular. There were approximately 10times the number of viable virulent Legionella cells inalveolar macrophages than in neutrophils. In contrast, therewere about equal numbers of viable avirulent bacteria inmacrophages and neutrophils. Therefore, during the earlyperiod after infection the primary site of Legionella multi-plication appears to be the alveolar macrophage. Unlike thefinding in short-term in vitro experiments, neutrophils ap-pear capable of killing at least some Legionella cells in vivoover days, but this relatively inefficient killing is not suffi-cient to eliminate the bacteria.

Second, the susceptibility of a given animal species toinfection is correlated with the ability of L. pneiumophila tomultiply within macrophages from that species (205, 206).However, there were exceptions, in that Legionella cellsgrew in the peritoneal macrophages of golden hamsters andrats, species which are resistant to infection (206). Theseresults indicate that in at least some species other factorsmust be involved in susceptibility to infection. Third, L.pneumophila mutants that are impaired in intracellulargrowth possess reduced virulence for animals (45, 102, 112,154).

Phagocytosis and Phagosome-Lysosome Fusion

As indicated in the preceding sections, the most funda-mental biological difference which has been found betweenvarious Legionella strains is their mode of entry into phago-cytic cells and their intracellular fate. Both L. pneirnophilaand L. micdadei must fix complement in order to be ingestedby neutrophils. However, in vitro opsonization of L. pneu-mophila Philadelphia 1 is dependent on specific antibody-mediated activation of the classical complement pathway;activation of the alternative pathway could not be detected(106, 190). On the other hand, L. micdadei fixes complementby the alternate pathway in the absence of specific antibody,and normal serum is a sufficient opsonin for neutrophilphagocytosis (180, 198). These findings lead to the conclu-sion that L. pneumophila should not be efficiently phagocy-tized by neutrophils early in primary infection before anti-body is produced. Since neutrophils appear to be able to killat least some legionellae in vivo, this could be an effectivevirulence mechanism. However, there is no evidence frompathological studies of human or animal material that L.pneumophila cells are not phagocytized by neutrophils to the

same extent as L. mnicdadei. Perhaps the uptake of L.pneiinophila by neutrophils is explained by later studiesinvolving more sensitive methods which showed that thePhiladelphia 1 strain does fix complement in nonimmuneserum, albeit at 10-fold lower levels than E. coli does (14).Horwitz (101, 103) observed that L. pneumophila Phila-

delphia 1 is ingested by human monocytes, alveolar macro-phages, and polymorphonuclear leukocytes in an uniquemanner termed coiling phagocytosis. In this process, a longphagocyte pseudopod coils around the bacterium as theorganism is internalized. The bacterium thus ends up in thecenter of a large coil and eventually comes to reside in anabnormal, ribosome-studded phagosome (98). The phago-some containing L. pneumophila Philadelphia 1 does notfuse with either primary or secondary monocyte lysosomes,as measured by acid phosphatase cytochemistry or by pre-labeling lysosomes with thorium dioxide, respectively (99).The Philadelphia 1 strain also inhibits phagosome acidifica-tion in monocytes (104). Since live, glutaraldahyde-killed,and heat-killed L. pneuimophila Philadelphia 1 cells are allinternalized by coiling phagocytosis (101), the intracellularfate of the bacteria must be determined by factors other thanthe unusual mode of entry.

In contrast to the situation with Philadelphia 1, severalworkers observed that engulfment of other L. pneumophilaserogroup 1 strains occurred by conventional phagocytosisin which the bacterium is surrounded by extensions ofpseudopods until their tips meet and fuse on the distal side ofthe particle (43, 69, 148). Likewise, we found that L.micdadei is ingested by conventional phagocytosis and thatthe phagosome in which it is contained appears to be normal(198, 199). The apparent contradictions were clarified byRechnitzer and Blom (159). They confirmed that the Phila-delphia 1 strain of L. pneumophila serogroup 1 is internal-ized by coiling phagocytosis. However, the Knoxville 1strain of L. pneirnophila serogroup 1 and L. micdadei werephagocytized in the classical manner. More importantly, theformation of phagolysosomes was seen following phagocy-tosis of cells of the Knoxville 1 strain and L. micdadei, butnot the Philadelphia 1 strain. Since pseudopod coil forma-tion, formation of an abnormal phagosome, and inhibition ofphagosome-lysosome fusion (and probably inhibition ofphagosomal acidification) may be specific for the Philadel-phia 1 strain of L. pneumophila, these phenomena areindependent of bacterial virulence. It is improbable, there-fore, that these processes are necessary for the intracellularsurvival of legionellae (159). On the other hand, thesecharacteristics could promote or enhance the virulence ofPhiladelphia 1 vis-a-vis other strains of L. pneumophila andthe other Legionella spp.One can postulate that there may be one or more funda-

mental virulence factors possessed by all legionellae whichexplain their ability to survive and multiply within phago-cytes. Both those fundamental mechanisms which are rele-vant to the entire genus and the additional or ancillaryfactors which are strain specific may be elucidated mostrapidly by comparing and contrasting the virulence determi-nants of the various strains. The foregoing evidence suggeststhat the outcome of the interaction between Legionella spp.and phagocytes is multifactorial and could depend for eachstrain on at least the following factors: whether the straininhibits phagosome-lysosome fusion and phagosomal acidi-fication; whether the strain inhibits phagocyte activation andthe subsequent generation of bactericidal oxygen metabo-lites by the particular phagocyte; and the susceptibility of thebacterial strain to the toxic oxygen metabolites produced by

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 39

phagocytes. It is also possible that the legionellae can defeatnonoxidative killing mechanisms of phagocytic cells, but thistheoretical virulence mechanism has not yet been examined.

BIOCHEMICAL OVERVIEW

Since they multiply extracellularly on complex laboratorymedia, the members of the Legionellaceae are classified asfacultative intracellular bacteria. They are obligate aerobes.In a comprehensive review of the biochemistry and physiol-ogy of Legionella spp., Miller and Hammel (133) concludedthat although L. pneumophila can hydrolyze starch and canoxidize certain sugars, albeit slowly, its energy metabolismis based on oxidation-dependent rather than fermentativepathways. Furthermore, most of the glucose assimilated bythe organism appears to be used to provide carbon skeletonsfor various biosynthetic pathways. Much of the glucose thatis consumed is metabolized by the pentose phosphate path-way, presumably to provide pentoses for nucleic acid syn-thesis and NADPH to satisfy the need of the cell for reducingpower in various biosynthetic pathways (e.g., fatty acidbiosynthesis). Legionella cells seem to derive most of theirenergy from the oxidation of amino acids and related com-pounds by means of the tricarboxylic acid cycle, which,along with a complex electron transport chain, is completelyexpressed in these organisms. Serine, threonine, and gluta-mate are especially good substrates for the legionellae. Inaddition, the following organic acids stimulate oxygen con-sumption by these bacteria: lactate, pyruvate, acetate,malate, fumarate, and oxaloacetate.

L. pneumophila is capable of synthesizing its own fattyacids. A unique feature of the organism is that it producesmostly branched-chain fatty acids, the predominant onebeing isopalmitate, a saturated, branched, 16-carbon fattyacid (136). The organism also contains small amounts ofhydroxy fatty acids, (,B-hydroxyisomyristic acid, ,-hy-droxyarachidic acid), most of which are confined to cell wallstructures (130). The major phospholipids of L. pnei,nophilahave been determined to be, in order of decreasing abun-dance, phosphatidylcholine, cardiolipin (diphosphatidylglyc-erol), phosphatidylethanolamine, phosphatidylglycerol, andphosphatidyldimethylethanolamine (75).Thorpe and Miller (187) examined 10 strains of L. pnei-

mophila for the production of extracellular enzymes. Allstrains produced detectable levels of extracellular protease,phosphatase, lipase, DNase RNase, and ,B-lactamase activ-ity. Weak starch hydrolysis was also demonstrated for allstrains. Elastase, collagenase, phospholipase C, hyaluroni-dase, chondroitinase, neuraminidase, and coagulase werenot detected. However, all these strains had been passed onagar, so that the production of additional enzymes bywild-type legionellae could not be excluded. Unlike L.pneumophila, L. bozemanii, L. dumoffli, and L. gorrnanii, L.micdadei does not produce a P-lactamase (151).

PROPERTIES OF POTENTIAL VIRULENCE FACTORS

How the legionellae inhibit phagocyte functional activitiesis not clear. Since L. micdadei depress a variety of phago-cyte functions (57, 59), it is likely that the bacteria disrupt aprocess which occurs early in the course of cell activationand is a common precedent to each of the functions weexamined. Signal transduction linking cell surface stimula-tion with intracellular activation is such a process. Duringthe last several years, a variety of signal transductionsystems have been defined and shown to be important in

neutrophil activation (179). These include (i) hydrolysis ofthe membrane phosphoinositide phosphatidylinositol-4,5-bisphosphate (PIP,) by phospholipase C, (ii) arachidonicacid release via phospholipase A2 activation, and (iii) intra-cellular fluxes of calcium (11, 19, 83, 128). Disruption ofsignal transduction by bacterial products such as pertussistoxin has clearly been shown to block a variety of cellfunctions similar to what we observed after ingestion of L.micdadei by neutrophils (12, 13, 30, 119, 191).Two peptide factors elaborated by the legionellae which

inhibit phagocyte activation have now been described. Ma-terial has been partially purified from supernatants andsonicates of L. pneumophila and L. micdadei that hasdemonstrated cytotoxic activity and lethality for animals andinhibits neutrophil oxidative metabolism (78, 79, 89, 90, 121).The second moiety elaborated by the legionellae whichinhibits phagocyte activation is a specific phosphatase thatwe found blocked oxidative metabolism by stimulated neu-trophils (172).

Peptide Toxin

The first Legionella factor known to inhibit neutrophilactivation dates to the observation by Friedman et al. (78)that the culture filtrates in which L. pneumophila was growncontained cytotoxic activity for Chinese hamster ovary(CHO) cells. The cytotoxic activity was found to be heatstable, could pass through dialysis tubing with a molecularweight cutoff of 1,000, and was sensitive to pronase andpapain but insensitive to trypsin. These properties suggestedthat the cytotoxin was a small polypeptide. One of thecritical microbicidal actions of phagocytic cells is the gener-ation of toxic oxidative metabolites including 02 and H202from molecular oxygen. The reduced-oxygen species de-rived from O,- and H202 play an important role in the killingof bacteria and parasites. Quantities of toxin partially puri-fied from L. pnei,nophila serogroup 1 (Knoxville 1) whichhad no effect on neutrophil viability or phagocytosis de-pressed HMPS activity and oxygen consumption duringphagocytosis and inhibited both bacterial iodination and thekilling of E. coli by neutrophils (79). Treatment of neutro-phils with toxin partially purified by a different methoddepressed neutrophil HMPS activity stimulated by latexbeads and the calcium ionophore A23187, but not shuntactivity stimulated by ConA, PMA, or the potassium iono-phore valinomycin (121). Treatment of neutrophils withcrude toxin preparations also blocked membrane depolariza-tion in response to A23187, but not in response to PMA.Independently, Hedlund and Larson (90) partially purifiedtoxin from both L. pnetumophila and L. micdadei bacterialcell sonicates and showed that the toxin preparations of thetwo species shared a common antigen unrelated to Legion-ella LPS. Their toxin preparations inhibited the chemilumi-nescent response of neutrophils (89).

Further characterization of the Legionella toxin has beenhampered by difficulty in purifying this peptide factor tohomogeneity. As was true for previous workers, we havebeen able only to partially purify the L. micdadei toxin frombacterial cell sonicates by conventional chromnatographicmethods. We noted that when neutrophils were treated withtoxin-containing preparations, subsequent 02 productionin response to fMLP was greatly inhibited, and we used thisas an assay for biological activity. Gel filtration of toxin-containing preparations on a Bio-Gel P-2 column eluted with10 mM phosphate buffer yielded numerous biologically ac-tive, low-molecular-weight peaks which had similar amino

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

40 DOWLING ET AL.

02

G protein1+

Pi - PIP 0P''2p Phospholipase C

Al DAGIP3 1n OxNADPH

I Protein Kinase C---- Oxidase

Ca2+

InatracellularCa2+ stores

schematic diagram of the cellular pathways leading to NADPH oxidase activation. Abbreviation: DAG, sn-1,2-

acid compositions (62). This suggested that spontaneousaggregation or oligomerization of the toxin molecules occursat low salt concentration, which may be one of the impedi-ments in attempts to purify the toxin. The predominant,biologically active A206 peaks obtained from Bio-Gel P-2 gelfiltration were examined for their effect on O2 productionby neutrophils in response to various stimuli. The inhibitionof the 02 response to fMLP was not abolished by heatingthe partially purified toxin to 100°C for 60 min. Preincubationof neutrophils with toxin for 10 min markedly inhibitedsubsequent 02 production in response to both ConA andOPZ to about the same extent as the inhibition of fMLP-stimulated production measured simultaneously; treatmentwith toxin had no effect on 02 produced in response toPMA. These results are not completely congruent with thoseof Lochner et al. (121), who found that partially purified L.pneumophila toxin did not inhibit neutrophil HMPS activitystimulated by either PMA or ConA. The difference found inthe inhibition of ConA-stimulated neutrophils by toxin couldrepresent true dissimilarities between the toxins derivedfrom L. micdadei and L. pneiumophila or could merely by afunction of the specific neutrophil response (HMPS activityor 02 generation) being examined. However, this apparentinconsistency is more likely to be due to extraneous mate-rials in the relatively crude toxin preparations which havebeen examined.

Since preincubation of neutrophils with toxin inhibitedneutrophil 02 production in response to fMLP and OPZ,but had no effect on the 02 produced in response to PMA,treatment of neutrophils with partially purified toxin alsodoes not exactly reproduce the refractoriness to stimulationwhich is seen following infection of neutrophils with L.micdadei (59). In the latter case, PMA-stimulated 0,-production was attenuated, albeit to a lesser extent thanfMLP- or OPZ-stimulated oxygen metabolism. However,since the toxin is the only heat-stable putative virulencefactor presently known, it must be considered the primecandidate responsible for the inhibition of neutrophil activa-

tion which occurs following the ingestion of heat-killed L.miedadei.The mechanism of toxin action is unknown. The pattern of

stimulus specificity indicates that the Legionella toxin acts atan early step in the neutrophil activation sequence. Thesteps in signal transduction between stimulus binding and0- production are best delineated for fMLP (179). Asshown in Fig. 1, occupation of the neutrophil fMLP-receptoris coupled to the activation of a polyphosphoinositide-specific phosphodiesterase (phospholipase C) through a per-tussis toxin-inhibitable guanine nucleotide-binding (G) pro-tein (83). Phospholipase C catalyzes the hydrolysis of PIP2 toyield the intracellular second messengers, rnyo-inositol-1,4,5-trisphosphate (UP3) and sn-1,2-diacylglycerol. IP3 is

involved in the mobilization of CA2+ from intracellularstores in the endoplasmic reticulum and calciosomes. Theincreases in both intracellular Ca2+ concentration and dia-cylglycerol lead to the activation of protein kinase C (110).The phorbol esters can substitute for cellular diacylglycerolby binding directly to and activating protein kinase C. It is

not known whether the neutrophil NADPH oxidase, whichcatalyzes the reduction of molecular oxygen to 02-7 iSactivated directly by protein kinase C or through additionalintermediaries; it is likely that at least one component of theNADPH oxidase complex is a substrate for protein kinase C(2). That 02 production by PMA-stimulated neutrophilspretreated with partially purified toxin is not inhibited wouldindicate that the toxin acts prior to the activation of proteinkinase C. If ConA- as well as fMLP-stimulated 02 produc-tion is inhibited by the toxin, it implies that the toxinprevents the activation of phospholipase C without affectingfMLP binding or the pertussis toxin-inhibitable G protein.ConA binds to a neutrophil receptor which is distinct fromthe fMLP-receptor, since ConA binding, and the subsequentproduction of 0,- can be reversibly inhibited by a-methylmannoside (47). ConA stimulation also bypasses the Gprotein, since 0- generation in response to ConA is notblocked by pertussis toxin (191). Furthermore, ConA trig-

FIG 1. Simplifieddiacylglycerol.

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 41

gers the breakdown of PI, but not PIP2, and produces nosignificant increase in 1P3 (117), indicating that ConA doesnot activate the polyphosphoinositide-specific phospholi-pase C. However, both ConA and fMLP stimulation pro-duces a rapid increase in levels of diacylglycerol and phos-phatidic acid. Although the source of the phosphatidic acid isnot known, it appears that the Legionella toxin may affectother activation pathways in addition to blocking the poly-phosphatidylinositol-specific phospholipase C.The present evidence that the Legionella toxin is of

importance in the pathogenesis of infection at the cellularlevel is compelling. First, toxin is produced by all fiveLegionella spp. which have been examined (89). Second,legionellae secrete the toxin into the medium in which theyare grown, and exogenously applied, partially purified toxinblocks neutrophil activation, indicating that the toxin pro-duced by both extracellular and intraphagosomal bacteriacould adversely affect phagocytes. Third, the toxin is stableat pH 3.5, so that it would not be inactivated in thephagosome even if phagosomal acidification takes place.Fourth, with the possible exception of PMA stimulation, thetreatment of neutrophils with toxin reproduces the refracto-riness to stimulation and decrement in bactericidal activitywhich is seen following the phagocytosis of L. ,nicdadei.Fifth, the toxin is heat stable and can therefore explain theinhibition of neutrophil functions which follow the ingestionof heat-killed L. micdadei. Finally, immunization of micewith toxin-containing material protects against a lethal chal-lenge of Legionella cells (89), indicating that toxin produc-tion by legionellae is an important pathogenetic mechanism.This latter finding raises the possibility that a toxoid vaccinewould be efficacious for protection against all species in thegenus. However, it must be borne in mind that all of thesestudies were accomplished with crude preparations thatmust have contained considerable extraneous material. Fur-ther progress in elucidating the place of the toxin in Legion-ella pathogenicity at the cellular level and its mechanism ofaction awaits purification of this moiety to homogeneity.

EnzymesExtracellular enzymes, such as proteases and hemolysins,

have been implicated as factors that might be responsible forsome of the pulmonary and extrapulmonary manifestationsof legionellosis. Thus, the production of a cytolytic enzymecould explain the lysis of the pulmonary inflammatory infil-trate seen in many cases or the infarctlike necrosis observedin some cases (202). An exoproduct could also explain whyprominent extrapulmonary manifestations, such as obtunda-tion, diarrhea, abnormal liver function tests, hematuria, andazotemia, may occur in the absence of remarkable lesionsoutside the lungs (22, 139). Since Miller and Hammel (133)wrote their comprehensive review of the biochemistry andphysiology of Legionella spp., a number of useful reportshave been published on the enzymes and proteins that arelocalized to the surface of the various Legionella species.

Phosphatases. Few studies have been carried out on theLegionella lysosomal hydrolases. In 1981 Muller (138)showed that L. pneumophila produces acid phosphatase,and in 1982 Nolte et al. (146) documented the production ofacid phosphatase by a total of nine Legionella strains includ-ing strains of L. micdadei. Using more traditional plate-substrate assays, as well as tube assays with culture super-natants, Thorpe and Miller (187) detected acid phosphatasein 10 strains of L. pnelimophila representing six serogroups.During acid phosphatase cytochemistry studies of phago-

some-lysosome fusion in peripheral blood monocytes in-fected with L. pneirmophila, Horwitz (99) noted a thin layerof the lead phosphate reaction product between the innerand outer bacterial membranes of the majority of the intra-phagosomal bacteria.The high-speed supernatant obtained after centrifuging a

suspension of L. micdadei that had been freeze-thawed andsonicated contained two acid phosphatases (172). The twoacid phosphatases can be separated by chromatography onhydroxylapatite. Both phosphatases are resistant to inhibi-tion by L-(+)-sodium tartrate. The predominant acid phos-phatase, designated ACP2, was purified to homogeneity bychromatography on QAE-Sephadex, phenyl-Sepharose, andSephadex G-200 (172). ACP2 runs as a single band andexhibits a molecular weight of 68,000 on sodium dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Itsisoelectric point is 4.5, and it has a pH optimum between 5.8and 6.0 when assayed with 4-methylumbelliferylphosphateas the substrate. Some of the preferred natural phospho-monoester substrates for ACP2 are fructose 1,6-diphosphate,AMP, and phosphotyrosine.

Preincubation of human neutrophils with a pure prepara-tion of the L. micdadei acid phosphatase (ACP2) inhibitedthe subsequent production of 02 in response to stimulationwith fMLP (172). The second Legionella acid phosphatasethat we purified (ACP1) has no effect on the ability ofneutrophils to produce O2 when tested at levels 100-foldabove those at which ACP2 is inhibitory. Pretreatment ofneutrophils with the Legionella phosphatase also inhibitedO2 generation in response to ConA, but had no effect on0,- production stimulated by PMA (64).The ability of the purified phosphatase to modulate the

respiratory burst of phagocytes raised the question of themechanism of action of the enzyme. The fact that thereceptor-mediated hydrolysis of phosphoinositides to yieldthe intracellular second messengers diacylglycerol and inosi-tol phosphates is involved in the regulation of phagocyteactivation (19) suggested that dephosphorylation of one ormore of these compounds might be responsible for thebiological effects of the Legionella phosphatase on neutro-phils. Furthermore, this approach had been successful inelucidating the mechanism of action of a cell surface acidphosphatase isolated from Leishmania donovani which sim-ilarly blocked 02 production by activated neutrophils (53).Indeed, we observed that PIP2 and 1P3 are both excellentsubstrates for the Legionella phosphatase in vitro (175). Incontrast to the nonphysiological substrate 4-methylumbel-liferylphosphate, which is hydrolyzed most rapidly at pH6.0, PIP2 is hydrolyzed optimally by the Legionella phos-phatase at pH 7.0. Because of this finding with the possiblephysiological substrate, we now refer to the enzyme as aphosphatase rather than an acid phosphatase.

Neutrophils which had been labeled with 32p; were treatedwith the L. micdadei phosphatase at pH 7.0, and the quantityof PlP, in the cells was determined at various times (175). By30 min, 20% of the radiolabeled [32P]PIP2 had been dephos-phorylated to phosphatidylinositol-4-phosphate (PIP). Noneof the PIP which was formed was further dephosphorylatedto PI. The production of inositol phosphates in neutrophilsfollowing fMLP stimulation was measured in [3HJinositol-labeled cells. When neutrophils were treated with theLegionella phosphatase prior to the addition of fMLP, theaccumulation of 1P3 30 s following stimulation was reduced44% when compared with that in control cells that were notincubated with the enzyme (175). A 30-min preincubation ofneutrophils with the phosphatase prior to addition to fMLP

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

42 DOWLING ET AL.

also decreased sn-1,2-diacylglycerol production by 45% at 2min following stimulation (175).Thus, the active phosphatase appears to reduce the

amount of the second messengers IP, and diacylglycerolproduced following receptor-mediated stimulation by twomechanisms (175). First, the phosphatase catalyzes thedephosphorylation of PIP, both in vitro and in intact neutro-phils. As shown in Fig. 1, this leaves less PIP2 available as asubstrate for phospholipase C following receptor-mediatedstimulation, so that smaller amounts of both 1P3 and diacyl-glycerol are generated. Second, the bacterial phosphatasemay directly dephosphorylate some of the 1P3 formed fromthe hydrolysis of PIP,. Although the majority of the PIP, ineukaryotic cells is believed to be in the inner cytoplasmicmembrane, the fact that PIP2 in intact neutrophils is asubstrate for exogenously added phosphatase implies thatphosphatase from extracellular or intraphagosomal legionel-lae can reach this substrate. Although treatment of neutro-phils with the phosphatase reproduces the refractoriness tofMLP-stimulated activation which is seen following inges-tion of L. micdadei (59), the phosphatase cannot be solelyresponsible for the antiphagocyte effects of ingested le-gionellae. The phosphatase does not prevent 0,- productionin response to PMA, as occurs following phagocytosis of L.micdadei (59). Also, the Legionella phosphatase is heatsensitive and loses both phosphatase activity and its inhibi-tory effect on stimulated neutrophil O- production afterheating (172), whereas ingestion of heated L. rnicdadeiinhibits subsequent neutrophil (59) and monocyte (57) acti-vation. Thus, it is doubtful that the enzymatic activities ofthe Legionella phosphatase fully account for the capacity ofthe bacterium to block the respiratory burst of phagocyticcells following stimulation. More probably, the phosphataseacts in concert with the toxin to produce blocks at multiplesites of the activation pathway following the phagocytosis ofviable legionellae.

Phospholipase C. With the exception of L. inicdadei, all ofthe various Legionella spp. that have been analyzed for theirability to produce phospholipase activity have been found topossess this capability. Baine (3) recognized that the histo-pathology of pneumonia caused by L. pneiumophila wascharacterized by a cytoclastic picture with necrosis of mono-nuclear and polymorphonuclear leukocytes and acknowl-edged that the pathogenesis might involve an extracellularphospholipase. Baine et al. (5) had shown previously thatcultures of L. pneumophila lyse erythrocytes from guineapigs, horses, sheep, rabbits, and humans. Interestingly, itwas the guinea pig and dog erythrocytes, which have thehighest content of phosphatidylcholine, that were hemo-lyzed most rapidly by L. pneumophila (3). The same Legion-ella strains which lysed erythrocytes also caused clouding ofegg yolk agar, suggesting that the bacteria were producingsome sort of lecithinase. In the same study, it was demon-strated that washed cells of seven different Legionella spp.,including L. micdadei, would all catalyze the hydrolysis ofthe nonphysiological sphingomyelinase-phospholipase Csubstrate p-nitrophenylphosphorylcholine. Five of six Le-gionella spp. that hydrolyzed p-nitrophenylphosphorylcho-line would also cleave [3H]phosphorylcholine from[3H]choline-labeled phosphatidylcholine, confirming the hy-pothesis that the lecithin-cleaving factor is a C-type phos-pholipase. The only species which exhibited minimal hemo-lysis, did not digest egg yolk, and was incapable of splittingthe tritiated lecithin substrate was L. micdadei. This obser-vation illustrates the uncertainty inherent in drawing conclu-sions about the nature of enzyme reactions responsible for

cleaving artificial substrates such as p-nitrophenylphospho-rylcholine.The hemolytic factor in the culture filtrates of a strain of L.

bozeinanii appeared to be a macromolecule, since it wouldnot pass a filter with a 10,000-Da cutoff and was voided whenchromatographed on a Sephadex G-100 column (80). Thesupernatant from sonicated L. bozemanii demonstrated farless hemolytic activity than the culture supernatant, indicat-ing that the hemolysin was an extracellular product of thebacteria.

In 1988 Baine (4) reported on the purification and charac-terization of the L. pneumophila phospholipase C. High-yield (85%) purification was achieved after only a 29-foldincrease in the specific activity of the enzyme. Purificationrequired only chromatography on DEAE-Sephadex, manga-nese chloride precipitation, and ammonium sulfate fraction-ation. Enzyme activity was assayed by the hydrolysis ofp-nitrophenylphosphorylcholine and confirmed by the re-lease of radioactivity from L-a-dipalmitoylphosphatidylcho-line labeled on choline. The molecular mass of the enzyme is50 to 54 kDa, and maximum activity on p-nitrophenylphos-phorylcholine substrate is observed at pH 8 to 9. Activitywas stimulated three- to fourfold above the basal level in thepresence of EDTA and by Ba2+, Ca2+, and Mg2+, but wasinhibited by Zn2+, Cu2+, and Fe2+. Neutral detergents (e.g.,Tween 20 and 80, Triton X-100) also stimulated activity two-to threefold. Interestingly, the purified phospholipase C wasnot hemolytic. Thus, although it was originally thought thathemolysis was a marker for phospholipase C activity, iteventually turned out that the hemolytic activity is due toanother moiety, probably a metalloprotease (see below)and/or a recently described 29-kDa nonproteolytic hemoly-sin termed legiolysin (158).Thus, it has been shown convincingly that, with the

exception of L. micdadei, the Legionella spp. release con-siderable quantities of phospholipase C activity into themedium in which they are grown; the enzyme acts onphosphatidylcholine to produce diacylglycerol and phospho-rylcholine. Baine (4) suggested three possible pathophys-iological implications of the phospholipase C that is secretedby Legionella spp. First, hydrolysis of phosphatidylcholine,which is a major component of pulmonary surfactant, couldimpair pulmonary gas exchange. Support for this idea existsin that alveolar hyaline membranes have been described inLegionella pneumonia (203). With the availability of pureLegionella phospholipase C (4), it should now be possible toanalyze the effects of the enzyme on isolated surfactant.Second, since phosphatidylcholine is an important constitu-ent of eukaryotic membranes, the cytolytic action of thephospholipase C might injure both inflammatory cells andlung tissue. It should be noted, however, that there is nodirect evidence that the Legionella phospholipase C iscytolytic or affects eukaryotic cell membranes. Third, theperturbation of phagosome membranes by hydrolysis ofphospholipid might disrupt phagosome-lysosome fusion.However, the evidence that Legionella strains which pro-duce phospholipase C fail to inhibit fusion following inges-tion (69, 148, 159) makes this possibility unlikely.An additional mechanism by which bacterial phospholi-

pase C might influence the interaction of the legionellae withhost phagocytes is by mimicking the activation of cellularphospholipase C and catalyzing the hydrolysis of PIP2 to theintracellular messengers 'P3 and diacylglycerol. Exogenousbacterial phospholipase C has been shown to modify neutro-phil activation. Phospholipase C from both Clostridiumperfringens and Bacilliis cereus elicited neutrophil 02 pro-

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 43

duction and enhanced the oxidative response to NaF whenboth stimuli were presented simultaneously (181). However,when the stimuli were added sequentially, bacterial phos-pholipase C abrogated the subsequent oxygen consumptionresponse of neutrophils to latex beads. The authors sug-

gested that bacterial phospholipase exoenzymes could acti-vate neutrophils before they make contact with bacteria,leaving them unable to respond adequately to the microor-ganism. Whether the legionellae might avoid the bactericidalcapacity of phagocytic cells by triggering a premature respi-ratory burst in this fashion remains to be examined. We (61)found that when L. micdadei cells are added to neutrophils,the amount of 02- produced (8.9 ± 2.0 nmol/106 cells per 15min) was not significantly greater than with S. aiur ells (5.0 ±

0.8 nmol/106 cells per 15 min; P = 0.07). Since L. inicdadeiis the only species that does not produce phospholipase C,this would be the expected result, whereas the other phos-pholipase C-producing species should trigger an early, largerelease of 02- Indeed, a preliminary report suggests thatpurified L. pneumnophila phospholipase may produce a dose-related inhibition of neutrophil function(s) (176).

Proteases. (i) General. The legionellae elaborate a numberof proteolytic enzymes and aminopeptidases, some of whichremain bound to the organism and some of which are

secreted into the extracellular culture medium. One of theearliest studies demonstrating the proteolytic capability ofthe Legionella spp. was that of Muller (137); he showed thatfour strains of L. pneumophila could degrade a number ofhuman serum proteins. Muller incubated suspensions of thebacteria with serum and then used immunoelectrophoresis toanalyze for proteolytic degeneration of 23 different proteins.Five proteins were degraded: al-acid glycoprotein, otl-anti-chymotrypsin, ot-lipoprotein, PlE-globulin, and 32-glycopro-tein I. Muller speculated that inactivation of (xl-antichymot-rypsin by L. pneumophila may contribute to theemphysemalike syndrome seen in legionellosis, which iscomparable to that observed in some patients with (x-antitrypsin deficiency. Thompson et al. (186) extended thiswork by demonstrating that cell-free culture filtrates of eightstrains of L. pneumophila representing six serogroups con-

tained proteolytic activity capable of digesting casein, gela-tin, and hide powder azure, but not elastin.

In addition to showing that L. pneumophila producesproteases, Muller (138) demonstrated the presence of a

variety of aminopeptidases. Again, washed, resuspendedorganisms were incubated with potential substrates, in thiscase synthetic, colorimetric naphthylamine, or nitroanilidederivatives of amino acids; collectively, these substratesconstitute the API ZYM system. For example, L-alanylami-nopeptidase activity is determined by using L-alanyl-2-naph-thylamine or L-alanyl-4-nitroanilide as substrates. All four ofthe L. pneumophila strains that were tested were positive for14 aminopeptidase activities; activities against L-hydrox-yproline and L-proline were absent, as were activities againsttrypsin and chymotrypsin substrates (N-benzoyl-DL-arginineand N-benzoly-DL-phenylalanine-2-naphthylamine). In a

parallel study, Nolte et al. (146) confirmed Muller's obser-vations with L. pneumophila and extended them to sixadditional Legionella spp. As will be described below, use ofthese single-amino-acid derivatives as protease substratescaused these investigators to fail to detect the prominentchymotrypsinlike activity produced by Legionella spp.

The first Legionella aminopeptidase to be isolated in pure

form was phenylalanine-aminopeptidase (86). The enzymewas purified 400-fold from the culture supernatant of L.pneumophila by affinity chromatography on 0-tert-butyl-

Thr-Phe-Pro-Gly-aminosilochrom and by gel filtration andion-exchange chromatography. It has the following proper-ties: MW, 35,000; pl, 5.8; pH optimum, 8.0 to 9.5 withL-phenylalanine p-nitroanilide as substrate. Although theenzyme could be inactivated by being incubated with EDTA,indicating that it is a metalloprotein, activity could not bereconstituted with Zn2+, Ca2 , Mg2+, Mn2+, or Cu2+. Onthe basis of a limited substrate specificity study, Gul'nik etal. (86) concluded that the role of this aminopeptidaseappears to be the liberation of hydrophilic free amino acidsfrom peptides present in the growth medium. These authorsalso speculated that the enzyme might hydrolyze intracellu-lar host cell peptides or the products of host cell proteolysiseffected by Legionella proteases.Whereas the studies of Muller (138) and Nolte et al. (146)

analyzed Legionella strains for exo-aminopeptidases, Berdalet al. (17) used para-nitroanilide (pNA) peptides to test forthe presence of endopeptidase activity in Legionella spp.Specifically, they used substrates in which the pNA moietyis attached to the carboxy-terminal end of tri- and tetrapep-tides. When the amide bond of the pNA peptide is cleaved,the free pNA that is liberated can be estimated spectropho-tometrically. The specificity of the endopeptidases is deter-mined by the sequence of amino acids at the amino-terminalend of these chromogenic substrates. Using chromogenicpeptides of this type which possessed a variety of N-terminalamino acids attached to a prolyl-proline-para-nitroanilidechain (Pro-Pro pNA), Berdal et al. (17) demonstrated thepresence of a proline-specific endopeptidase in three Legion-ella species; the enzyme could cleave pNA from the fol-lowing peptides: O-benzoyl-Ser-Thr-Pro-Pro pNA, 0-ben-zoyl-Ser-Pro-Pro pNA, and benzoyl-Ser-Pro-Pro pNA.The source of protease in these experiments was the con-centrate obtained after subjecting the cell-free growth mediato 100-fold concentration by pressure dialysis.

Using a similar approach, Berdal et al. (18) demonstratedthat L. pneirmophila, L. bozemanii, L. dumoffli, and L.gorm7anii, but not L. micdadei, produced soluble, extracel-lular chymotrpysinlike activity. Two compounds in particu-lar were especially effective substrates: Suc-0-Met-Arg-ProTyr pNA and Suc-Ala-Pro-Tyr pNA. Furthermore, theinhibitor pattern they obtained indicated that the chymo-trypsinlike protease is a serine active-site enzyme. Interest-ingly, when sonicated bacterial cells were analyzed forchymotrypsinlike activity, little activity was found to beassociated with the cells. Thus, these early studies withcrude bacterial products demonstrated that the legionellaesecrete proteases of the endopeptidase and exopeptidasevarieties.

(ii) Zinc metalloprotease. By the mid-1980s, investigatorsbegan fractionating the proteases produced by Legionellaspp. For example, in 1986 Conlan et al. (49) chromato-graphed the culture broth of L. pneumophila on a DEAE-cellulose ion-exchange column and a gel filtration columnand demonstrated the presence of at least six differentproteases, one of which digested collagen, casein, andgelatin and produced lung lesions in guinea pigs. This wasthe first evidence that a Legionella protease could producelesions closely resembling those of Legionella pneumonia.The tissue-destructive protease eluted from a SephadexG-100 column as though its native molecular weight was38,000, and it cleaved collagen and the chromogenic pep-tides Suc-0-Met-Arg-Pro-Tyr pNA and Suc-Ala-Pro-TyrpNA. This protein was devoid of elastase activity.

In the same year, Dreyfus and Iglewski (67) reported onthe purification and properties of the major chymotrypsinlike

VOL. 56. 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

44 DOWLING ET AL.

extracellular proteolytic enzyme of L. pneiumophila. Theenzyme appears to be identical to the tissue-destructiveprotease reported by Conlan et al. (49). In just four steps andwith an overall enrichment of only 10-fold, the purificationprovided a 50% yield and 10 to 20 mg of pure enzyme proteinfrom 6 liters of growth medium. It is a relatively acidicprotein (pl 4.2 to 4.4), and its pH optimum is broad (pH 5.5to 7.5); it has maximum activity against hide powder azureprotein substrate. The protease is a zinc metalloprotein (1mol of Zn2+ per mol of enzyme); however, it is not inhibitedby phenylmethylsulfonyl fluoride, chymostatin, or trypsininhibitor. Resistance to inhibition by these compounds is inconflict with the results obtained by Berdal et al. (18), whofound that the L. pneumophila protease had the properties ofchymotrypsin. This discrepancy probably occurs becausewhereas Dreyfus and Iglewski (67) characterized a pureenzyme, Berdal et al. (18) were analyzing a relatively crudeprotease preparation. The protease is active as a 40-kDamonomer but is readily autolyzed to a 38-kDa species.Rechnitzer et al. (162) described a method that avoidsenzyme autoproteolysis during purification and results in theisolation of a single 42-kDa protease which is cytotoxic to avariety of cell lines.

Several groups of investigators (15, 114, 162, 169, 185)have demonstrated that culture supernatants containing pro-teolytic activity or purified protease from L. pneiiinophilahave cytolytic effects on tissue culture cells including cyto-toxicity for CHO cells. Purified exoprotease from L. pneiu-mophila has also been shown to elicit hemorrhagic andnecrotic lesions when administered to guinea pigs intrader-mally (15, 49, 169). Conlan et al. (49) have used a guinea piganimal model to demonstrate the cytotoxic properties of theexoprotease. Intranasal or intratracheal administration ofthe partially purified L. pneutmophila protease into the lungsof guinea pigs produced pathological changes which werevery similar to those produced in experimental Legionellapneumonia (7, 49). Conlan et al. (50) also demonstrated thatthe tissue-destructive protease was present in the lung tissueof experimentally infected guinea pigs. On day 3 afterinfection, it was found in amounts equivalent to the lethaldose of the purified protease administered intranasally. Inaddition, Williams et al. (200) have shown by using double-labeling techniques that the protease is found at sites withlarge numbers of L. pneumophila cells, and both enzyme andbacteria are found only in association with pulmonary le-sions. These studies suggested that production of this pro-tease during natural infection may play an important role incausing cytolysis and the destruction of pulmonary tissue.Work in several laboratories (15, 114, 157, 169) has

demonstrated that the 38- to 42-kDa zinc metalloproteasefrom L. pneumophila produces the observed cytotoxic ac-tivities. Furthermore, the exoprotease was also shown tohave hemolytic activity. Preliminary screening of selectedLegionella species for proteolytic and hemolytic phenotypessuggested a correlation between these activities (114). Bothvirulent and avirulent strains of L. pneitmophila and mostother Legionella species expressed both extracellular prote-ase and homolytic activities. The exceptions were strains ofL. micdadei and L. feeleii, which expressed neither proteasenor hemolysin activity. When the L. pneuimophila exopro-tease was purified from culture supernatant, the hemolyticactivity was found to copurify with the proteolytic activity,and analyses by SDS-PAGE and immunoblotting revealed asingle protein species (114). Furthermore, an exoprotease-deficient mutant strain of L. pneirmophila (PRT8) was iso-lated and was found to be nonhemolytic (114). Concentrated

supernatants from strain PRT8 were no longer cytotoxic forCHO cells.Quinn et al. (157) cloned the genetic sequence encoding

the 38-kDa protease from L. pneumophila serogroup 1. Theyshowed by transposon inactivation analysis that a singletrifunctional polypeptide encoded on a 1.2-kb cloned DNAfragment, designated pro, is responsible for proteolytic,hemolytic, and cytotoxic properties. These workers thencompared the proteolytic, hemolytic, and cytotoxic activi-ties of strains from several Legionella species (157). Theconcentrated supernatants from all strains of L. pneumo-phila, L. dumoffii, and L. jordanis and from one (of two)strain of L. micdadei demonstrated protease (caseinase)activity. Only the Legionella strains which demonstratedproteolytic activity showed hemolysis on canine blood agar.Cytotoxicity for CHO cells was apparent with culture super-natants or sonicates containing protease from strains of L.pneuimophila, but the protease activities extracted fromother Legionella species possessed only hemolytic activity.A probe consisting of the DNA sequence encoding the38-kDa metalloprotease from L. pneumophila Philadelphia 1hybridized to the chromosomal DNA of all serogroups of L.pneilmophila, but not to any strains of L. micdadei, L.dirnoffii, L. Jeeleii, or L. jordanis that were examined.Furthermore, Western immunoblots done with antisera tothe L. pneulnophila protease demonstrated cross-reactionsamong 38-kDa proteins from strains of L. pneumophila, butno reactions were observed with proteins from other Legion-ella species. The cloned protease from L. pneumophilareacted with convalescent-phase serum from patients in-fected with L. pneiumophila, but not with antiserum frompatients infected with other Legionella species (114, 157).Thus, despite some similarities among the proteolytic activ-ities of Legionella spp., including proteolytic and hemolyticphenotypes, metal requirements for zinc or iron, sensitivityto EDTA, and temperature and pH optima, there are distinctgenetic, immunological, and cytotoxicity differences amongthe preteolytic activities produced by Legionella spp. Inparticular, only in strains of L. pneumophila is the 38-kDametalloprotease associated with cytotoxicity. The detectionof antibodies to the protease during L. pneumophila infec-tions indicated that the protease/cytotoxin is elaboratedduring natural infection and strengthened the possibility thatthe enzyme is involved in the pathogenesis of that species.These same workers (21) then determined the sequence of

the structural gene encoding the L. pneumophila zinc metal-loprotease and found that it contains a large open readingframe which is preceded by consensus promoter and ribo-some-binding sequences. The deduced polypeptide con-tained a putative signal sequence and a total of 543 aminoacid residues with a computed molecular size of 60,775 Da,substantially larger than the 38,000-Da native and recombi-nant proteins. There was extensive amino acid identity withthe Pseiidomonas aeruginosa elastase, which is also en-coded by an open reading frame larger than that predictedfrom the size of the mature protein. The structural identity ofthe L. pneumophila protease and the P. aeruginosa elastasewas most pronounced in the regions forming the enzymaticactive site of the elastase. Competitive inhibitors of Pseudo-monas elastase were equally potent in inhibiting the L.pneiimophila protease. These findings indicate that the L.pneiiunophila protease and the P. aeruginosa elastase sharea similar molecular mechanism of proteolysis. An invasive-ness-defective mutant of the fish pathogen Vibrio anguil-lariin has been isolated and shown to produce lower levelsof a zinc metalloprotease than the wild-type bacterium (147).

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 45

N-terminal amino acid analysis demonstrated considerablehomology between the Vibrio enzyme and the P. aeruoginosaelastase and small but significant homology with the L.pneumophila protease. The soluble hemagglutinin/proteaseproduced by V. cholerae is also a zinc metalloproteasewhich is structurally, functionally and immunologically re-lated to the P. aerliginosa elastase (87). The family ofbacterial zinc metalloproteases also includes proteases fromseveral Bacillus species and Serratia marcescens, whichhave been shown to be structurally related to P. aerulginosaelastase and to share amino acid homology in the regionsforming the enzymatic active sites and zinc-binding domains(87). The ubiquity and conservation of these similar prote-ases among both pathogenic and nonpathogenic bacterialspecies (and in the case of V. cholerae including bothpathogenic and nonpathogenic serotypes) suggest that theymay provide some common survival advantage which is notnecessarily essential for virulence but which, in the case ofpathogenic species, may promote it (87).

Szeto and Shuman (184) succeeded in isolating the struc-tural gene for the zinc metalloprotease from an L. pneiumo-phila genomic library. They termed the protease the majorsecretory protein (Msp), and they termed the structural genemspA. In E. coli which contained plasmids with the mspAgene, Msp protein and caseinase activity were found in theperiplasmic space and cytoplasm. Transposon mutagenesiswith Tn9 insertions in the cloned structural gene for Msp inE. coli yielded mutants which no longer expressed proteaseactivity. One of these Tn9 insertions was transferred to theL. pneumophila chromosome by allelic exchange. L. pneut-mophila strains that contained the transposon insertionwithin the mspA gene failed to synthesize the exoproteaseand failed to produce any hemolysis. However, the mutantstrain was able to grow within and kill human macrophages(PMA-differentiated HL-60 cells) as well as the wild-type,isogenic strain was. The culture supernatants from HL-60cells infected with the mspA mutant were examined for thepresence of revertants which had regained increased levelsof protease production, and none were detected.

Blander et al. (29) found that when guinea pigs werechallenged with various doses of aerosolized Msp- andMsp+ L. pneumophila, the protease-negative and -positivestrains demonstrated equivalent 50% and 100% lethal doses.The Msp- mutant and Msp+ isogenic parental strains mul-tiplied in the lungs of challenged guinea pigs at comparablerates; both strains multiplied 4 logs over 72 h. They alsoproduced comparable pathological lesions in infected guineapigs. By SDS-PAGE and Western blot analysis, no reversionof the Msp- strain occurred following passage in the ani-mals. On the basis of these observations, the cytotoxicexoprotease is not required for intracellular infection ormultiplication nor for lethality in vivo. Furthermore, sincethere was no difference in the pathological findings of thelungs of guinea pigs infected with the protease-negativemutant strain, the protease does not appear to be necessaryfor the production of acute pneumonitis. The cytotoxicexoprotease might still be considered a persistence factorthat enhances virulence through the destruction of macro-phages (114). It has been suggested that self-limited Legion-ella infections, such as Pontiac fever, may result fromaproteolytic organisms colonizing lung macrophages butfailing to elicit the cytotoxic damage necessary for thedevelopment of the more acute disease (114). However, thework of Szeto and Shuman (184) clearly showed that a1,000-fold decrease in caseinolytic activity does not decreasethe cytopathic effect of L. pneumophila in differentiated

HL-60 cells. Surprisingly, these results make it likely thatthere must be some other moiety which is responsible forany in vivo cytopathic effect; the Legionella toxin would bethe leading candidate for this phenomenon.Even if the zinc metalloprotease is not a virulence factor

for intracellular infection, it could have effects on specifichost proteins of potential relevance to the pathogenesis oflegionellosis. Conlan et al. (51) demonstrated that the tissue-destructive protease produced by L. pneumophila and re-leased into the growth medium was capable of thoroughlyinactivating the important blood proteinase inhibitor o1-antitrypsin. Unlike most other proteases such as trypsin andchymotrypsin, which consume and irreversibly inactivatexl-antitrypsin by forming 1:1 molar complexes with theprotease inhibitor, as little as one molecule of the L. pneu-mophila protease is capable of completely inactivating 500molecules of ol-antitrypsin. The inactivation of oal-antitryp-sin by tissue-destructive protease appears to proceed bycleavage of a 5,000-Da polypeptide from the native,57,500-Da parent atl-antitrypsin molecule. The ability of theL. pneiumophila protease to inactivate many times its ownmass of a1-antitrypsin, one of the major defenses againstproteolytic attack, raises the possibility that this particularenzyme, as well as other proteases released by necrosis fromboth the host and phagocytized bacteria, has the potential tocause more pulmonary damage than if it was inhibited byal-antitrypsin. However, as already noted, there does notappear to be any difference in the lung abnormalities ofguinea pigs infected with Msp+ and Msp- strains (29).The Legionella protease may also directly impair phago-

cytic and natural killer cell function. There is a preliminaryreport indicating that the purified L. pneumophila proteaseat concentrations that were not cytotoxic inhibited humanneutrophil and monocyte chemotaxis toward variouschemoattractants in a dose-dependent manner (160); anyeffects on phagocyte bactericidal functions were not ad-dressed in this study. This suggests that production of thebacterial enzyme could result in impaired recruitment ofphagocytic cells to the site of infection. Sahney et al. (177)prepared 10-fold-concentrated supernatants by ultrafiltrationin a cell with a 10,000-Da cutoff from overnight cultures of avirulent L. pneumophila serogroup 1 strain and a protease-deficient mutant derived from an avirulent L. pneumophilaserogroup 1 strain by ethyl methanesulfonate mutagenesis.The concentrated supernatant from the virulent strain inhib-ited spontaneous neutrophil chemotaxis and chemotaxistoward fMLP and 02 generation in response to zymosan-activated particles, PMA, A23187, and fMLP at concentra-tions that had no effect on cell viability. Heat-treated super-natant from the virulent strain had no inhibitory effect on02 generation in response to any of the four stimuli. Incontrast, the supernatant from the protease-negative mutantfailed to inhibit neutrophil 02 response to zymosan-acti-vated particles and PMA and only partially inhibited neutro-phil response to A23187 and fMLP. Neutrophil spontaneousmigration was unaffected by the culture supernatant from themutant, whereas directed chemotaxis was partially inhib-ited. Since the inhibitory effects on neutrophil function wereabsent or smaller when the concentrated supernatant fromthe protease-deficient mutant strain was used, the conclu-sion that the protease was responsible for the effects of thesupernatant from the virulent strain seems inescapable.However, the protease-deficient strain was derived from anavirulent parent strain, but the supernatant from it wastested in parallel with the supernatant from an unrelatedvirulent L. pnelumophila serogroup 1 strain. This leaves open

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

46 DOWLING ET AL.

the question of whether the avirulent parent, and the pro-tease-deficient mutant derived therefrom, lacked anotherexoproduct(s) which is inhibitory for neutrophils. Culturesupernatants might contain phospholipase, phosphatase, anda number of other heat-labile exoproducts in addition toprotease; the low-molecular-weight toxin should have beeneliminated in these experiments by the ultrafiltration step.As noted above, a Legionella phosphatase blocks neutrophilactivation. Also, preliminary studies by this same groupindicated that purified Legionella phospholipase inhibitedneutrophil function (176). It therefore remains to be deter-mined whether the protease adversely affects the bacteri-cidal mechanisms of phagocytic cells.

Rechnitzer et al. (161) found that purified protease inhib-ited natural killer cell cytolytic activity in a concentration-and time-dependent fashion. The inhibitory effect was notreversed by either gamma interferon or interleukin-2. Theprotease appeared to act by inhibiting the binding of effectorcells to target cells. In vivo, this action might reverse thepreviously described lysis of L. pneumophila-infected mono-nuclear cells by a natural killer subset population (25).

Before there was evidence indicating that the tissue dam-age caused by the legionellae cannot be attributed to theeffects of the zinc metalloprotease, the possibility that anti-body to this enzyme would reduce the severity of the diseaseand thus offer a means of prophylaxis and therapy wasconsidered (7). This approach has been explored for theacute bacterial pneumonias due to the extracellular bacteriaS. marcescens and P. aeruginosa. Guinea pigs and micevaccinated with purified S. marcescens protease developedanti-protease antibodies and were partially protected againsta subsequent lethal challenge of the bacterium (124). Simi-larly, immunization of mink with a protease toxoid vaccineof P. aeruginosa resulted in the production of anti-proteaseantibodies and protection against pneumonia due to a lethalinoculum of bacteria (97). Blander and Horwitz (27) ob-served that guinea pigs infected with L. pneumophila de-velop cell-mediated immunity to the zinc metalloprotease.Guinea pigs immunized with a subcutaneous dose of purifiedLegionella protease developed strong humoral and cell-mediated immune responses to the enzyme (27). When theseanimals were challenged with a lethal aerosol dose L.pneumophila, they exhibited significant protective immu-nity, which was associated with the ability to limit Legion-ella multiplication in their lungs. When combined with theprevious studies indicating that the protease is not a neces-sary virulence factor for intracellular infection or tissuedestruction, these results suggest that immunization with theprotease fosters immunity through the induction of specificcell-mediated immunity, rather than protection mediated byanti-protease antibody.When guinea pigs were immunized with L. pneumophila

serogroup 6 protease, protective immunity developedagainst challenge with L. pneumophila serogroup 1 organ-isms (28). However, as would be expected from the dataindicating that the proteases of various species are notgenetically identical, immunization with the protease fromL. bozemanii provided significant protection against chal-lenge with L. pneumophila serogroup 1, but less than thatprovided by immunization with L. pneumophila serogroup 1protease. Surprisingly, when challenged with L. bozemanii,neither guinea pigs immunized with L. bozemanii nor thoseimmunized with L. pneumophila serogroup 1 protease dem-onstrated protective immunity. Thus, the drawback of thisapproach to prophylaxis is that only partial or no protectionmay be produced against species, other than L. pneumo-

phila, which possess proteases genetically different fromthat of L. pneumophila. Nevertheless, if these results can beextended to humans, the purified protease appears to be apromising vaccine candidate for protection against the L.pneumophila serogroups, which cause the majority of cases.

Enzymes that scavenge reduced-oxygen metabolites. Hostphagocytes produce microbicidal reduced-oxygen metabo-lites such as 02 and H202 when they phagocytize bacteria.Enzymes which eliminate these toxic reduced forms ofoxygen, or which block their production in the first place,might allow intracellular pathogens to escape killing. Pine etal. (155) analyzed 40 Legionella strains for the followingreduced-oxygen scavenging enzymes: SOD, catalase, andperoxidase. SOD inactivates 02 by converting it into 2:

202- + 2H+ -- H202 + 02

Catalase converts H202 into H20 and 02 in the absence ofany other substrate:

2H202 -3 02+ 2H20

A peroxidase uses H202 to oxidize some organic compound[R(OH)21:

HO-R-OH + H202 -3 O=R=O + 2H20One commonly used assay for peroxidase activity employso-dianisidine as the reductant; the peroxidase-catalyzed ox-idation of o-dianisidine by H202 to 02 and H20 yields ayellow product in the absence of an organic compound justas catalase does. In this case the peroxidase is said toexpress catalatic activity.

Pine et al. (155) found basically that whereas SOD wasubiquitous among the Legionella spp., catalase and peroxi-dase activities were not distributed uniformly among the 40strains they analyzed. Two of the Legionella spp., L. pneu-mophila and L. gormanii, had no catalase and containedonly a peroxidase which, in L. pneumophila, exhibited lowcatalytic activity. L. dumoffli appeared to have this sameperoxidase plus a separate catalase. All the remaining spe-cies had a catalase only. Because none of these aforemen-tioned enzymes have been isolated in pure form from any ofthe legionellae, final conclusions about the enzymatic activ-ities responsible for the metabolism of H202 by live Legion-ella spp. or cell extracts derived must therefore be regardedas provisional. When Pine et al. (155) examined theseenzymological data from the standpoint of potential relation-ships of these enzymes to the frequency with which thevarious species cause disease, they concluded that (i) theredoes not appear to be a direct association of the peroxidasewith pathogenicity because the peroxidase appears equallydistributed among all of the serogroups of L. pneumophila,and (ii) the distribution of SOD and catalase among theLegionella spp. does not support the view that these en-zymes play a role in pathogenesis.

L. pneumophila appears to be highly sensitive in vitro toexternally added H202 (96, 122, 123), to in vitro products ofthe myeloperoxidase (MPO)-H202-halide system (122, 123),and to products of the xanthine oxidase system (122, 123).The latter system generates 02 and H202 and, in thepresence of Fe2 , OH' and possibly singlet oxygen. Sincekilling of L. pneumophila by products of the xanthineoxidase system in the presence of Fe2+ was reversed by theaddition of catalase or SOD and by the OH' scavengermannitol, it appears that H202, 02 and OH', or productsderived from them are the bactericidal species (122, 123).Likewise, we (58) found that L. micdadei is as susceptible tokilling by H202 in vitro as E. coli or S. aureus is. However,

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 47

both L. pneiumophila and L. micdadei were significantlymore resistant to killing by hypochlorite, the major bacteri-cidal product of the MPO-H20 -halide system of neutro-phils, than E. coli, Klebsiella pneiinoniae, or S. alureius were(58, 118). Whether the latter property is of biologic signifi-cance remains to be assessed.There is evidence that catalase produced by S. aiureius

protects intraphagocytic bacteria by destroying H20 pro-duced by neutrophils (127). It has also been determined thatthe facultative intracellular bacterial pathogen Norcardiaasteroides, which is susceptible to killing by oxygen metab-olites, is protected from oxidative killing mechanisms in vivoby the production of SOD (10). The nocardial SOD issecreted into the growth medium and is neutralized bytreatment of the whole bacteria with anti-SOD antibody (10).The reason why the legionellae are not protected by eitherSOD or catalase is presumably that both of these enzymesare located in the cytoplasm (96), suggesting that the targetsfor damage by peroxide may be located in or external to thecytoplasmic membrane.

Conversely, there is some evidence that legionellae maypossess quantitatively less oxygen metabolite-scavengingenzymes than other gram-negative bacilli. Phagocytosis ofL. pneumophila by primate alveolar macrophages was asso-ciated with a respiratory burst, as indicated by localizedreduction of Nitro Blue Tetrazolium around the ingestedorganism (109). However, when assayed quantitatively, in-gestion of either virulent or avirulent L. pneiumophila byalveolar macrophages stimulated three- to fourfold less NitroBlue Tetrazolium reduction than did the phagocytosis of E.coli, yet fewer E. coli cells were killed than L. pneiumophilacells. L. pneumophila was more susceptible than the E. colistrain to the toxic effects of H,02 and the oxygen metabo-lites generated by the xanthine oxidase system (123). Acombination of SOD and catalase, but not either scavengerof oxygen metabolites alone, significantly reduced the killingof L. pneumophila by alveolar macrophages. Mannitol alsoinhibited the killing of Legionella cells by alveolar macro-phages, and its effect was significantly greater that that ofSOD and catalase combined. These findings suggested thatOH plays a role in the early killing of Legionella cells byalveolar macrophages as well as in vitro (109). The differencein susceptibility of Legionella spp. and E. coli to early killingby alveolar macrophages and by oxygen metabolites corre-lated with the significantly lower content of the oxygenmetabolite scavengers catalase, glutathione peroxidase, glu-tathione reductase, and glutathione in Legionella spp. thanin E. coli (123).

Protein kinases. Protein kinases catalyze the phosphoryla-tion of specific serine, threonine, or tyrosine residues ofprotein substrates. Evidence demonstrating the presence ofmultiple species of endogenous phosphorylated proteins andprotein kinases in prokaryotes has been reported for Salmo-nella typhimurium (197) and E. coli (70, 126). For E. coli, ithas been shown convincingly that NADP+:isocitrate dehy-drogenase undergoes phosphorylation and dephosphoryla-tion (144, 163). Interestingly, both reactions are catalyzed bya single bifunctional enzyme. Protein kinases have beendescribed in protozoa (125), and we have extensively puri-fied and partially characterized protein kinases that arepresent on the outer surface of L. donovani promastigotes(52). The demonstration of this ectokinase activity, togetherwith the finding of acid phosphatase activity on the outersurface of L. donovani (165, 166), led us to speculate thatthis obligate intracellular parasite may be capable of regu-lating the properties and function of host phagocytic cells by

the phosphorylation and dephosphorylation of critical pro-teins. Therefore, we (173) examined the possibility thatLegionella spp. have protein kinase activity in addition tophosphatase activity.The high-speed supernatant from a sonicate of L. micda-

dei bacterial cells was applied to a QAE-Sephadex column(173). Approximately 70% of the protein kinase activityappeared in the QAE-Sephadex breakthrough fractions andwas designated PK I. When the column was developed withan NaCl gradient, a single peak of protein kinase activity waseluted and designated PK II. Both PK I and PK II werefurther purified on a histone affinity column followed byhigh-performance gel filtration. The K,,1 values with mixedhistones as the substrate were nearly identical for PK 1 (0.27mM) and PK 11 (0.29 mM). Both protein kinases exhibitedmaximum activity at pH 6.8 to 7.0. PK II is cyclic nucleotidedependent, and PK II activity is stimulated by calmodulin, aCa+-calmodulin mixture, mixtures of Ca2+ and phosphati-dylserine, or Ca2+ and PI, and a Ca2+-PI-diolein mixture.None of these agents markedly influenced the activity of PKI. All the experiments which follow concern the cyclicnucleotide-independent PK I, which hereafter will be re-ferred to simply as the L. micdadei protein kinase. Thisbacterial kinase was purified to homogeneity by isoelectricfocusing and chromatography on Sephadex G-150 (174). Thepurified protein kinase electrophoresed as a single band witha molecular weight of approximately 55,000 in SDS-contain-ing gels.To approach the question whether the bacterial protein

kinase might affect host phagocytes, we determined whetherneutrophils possess substrates for the enzyme (174). Neu-trophil homogenates were centrifuged to produce cytosoland membrane fractions. Either the cytosol or solubilizedmembrane fraction was incubated with the purified proteinkinase in the presence of [y-32P]ATP, and the 32P-labeledproteins were resolved by SDS-PAGE and visualized byautoradiography. The bacterial enzyme catalyzed the phos-phorylation of 11.5-, 14-, 19-, 23-, 28-, 34-, and 38-kDaproteins present in the extract of neutrophil membranes and11.5-, 13.5-, 25-, and 38-kDa proteins in the cytosol. Further-more, when purified Legionella protein kinase was incu-bated with calf brain tubulin and [-y-32P]ATP in the presenceof Mn2+, extensive phosphorylation of tubulin occurred.

Surprisingly, when the purified L. micdadei kinase wasincubated in the presence of [-y-32P]ATP and PI, a linear,time-dependent incorporation of 32p into PIP was observed;the PIP was not further phosphorylated to PIP2 (174). TheK, of the kinase for ATP was 1.5 mM. The PI kinase activityof the L. micdadei enzyme was optimal at pH 7.0, close tothat observed when histone is used as the substrate (pH 7.2[173]). Neutrophils which had been labeled with myo-[2-3H]inositol were treated with the purified PI-tubulin kinase,and the labeled inositol lipids were extracted and separated.After a 4-h incubation with the bacterial kinase, 19 to 24% ofthe PI in the neutrophil plasma membrane had been phos-phorylated to PIP, resulting in a 73 to 87% increase in thequantity of labeled PIP, while the level of PIP2 did notchange by more than 10% (174).

In summary, L. micdadei contains at least two proteinkinases. One of these kinases has the capability of catalyz-ing, at the expense of ATP, the phosphorylation of not onlytubulin (and histone) but also PI. The function and signifi-cance of the Legionella kinase are obscure. Although wehave demonstrated that the enzyme is capable of phosphor-ylating PI and tubulin in vitro and PI in intact neutrophils, itsphysiologically relevant substrates remain to be established.

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

48 DOWLING ET AL.

Although it seems unlikely, the possibility that there areendogenous substrates for the bacterial enzyme remains tobe excluded. The presence of PI is rare in prokaryotic cells;it has been found in only the mycobacteria (193) and myxo-bacteria (36) and was not detected in Legionella spp. (75).Since prokaryotes do not contain tubulin, it would seemmost likely that host phagocyte PI or tubulin or both arepotential physiological substrates. The phosphorylation ofeither of these eukaryotic substrates might block phagocyteactivation or function. For example, since phosphorylationof tubulin monomers inhibits assembly into microtubules(195, 196), the L. micdadei kinase might affect variousantimicrobial activities mediated by tubulin, including motil-ity, phagocytosis, and granule-membrane fusion. Also, all ofthe eukaryotic PI kinases that are closely associated withjrotein kinases have been found to phosphorylate PI at theD-3 position of inositol, rather than the usual D-4 position(38). If this is the case for the Legionella enzyme, the PI-3-Pformed would not be in the pathway for the formation ofPI-4,5-P2, which is the progenitor for IP3 production uponcell activation.

Legionella Cell Envelope

Hindahl and Iglewski (92) fractionated a whole-cell lysateof L. pneumophila Knoxville 1 into five membrane fractionsby sucrose gradient centrifugation. Membranes were char-acterized by enzymatic, chemical, and SDS-PAGE analyses.Two forms of cytoplasmic membrane (CM-1, CM-2), a bandof intermediate density (IM), and two forms of outer mem-brane (OM-1, OM-2) were detected. The CM-1 fraction wasthe purest form of cytoplasmic membrane, whereas fractionCM-2 was primarily cytoplasmic membrane associated withsmall amounts of peptidoglycan. The IM, CM-1, and CM-2fractions were all enriched in peptidoglycan. The IM, OM-1,and OM-2 fractions were expected to be enriched in totalhexose relative to CM-1 and CM-2 as a result of increasedamounts of LPS associated with these fractions. However,the amount of carbohydrate and 2-keto-3-deoxyoctonic acid(KDO) was not appreciably greater in outer membrane thanin cytoplasmic membrane. The high molar ratio of hexos-amine to KDO in the membrane fractions suggested that thegreat majority of hexosamine detected was derived frompeptidoglycan rather than LPS. Phosphatidylethanolamineand phosphatidylcholine were found to be the major phos-pholipids in the membrane fractions, confirming the unusualphospholipid composition of whole Legionella cells found byFinnerty et al. (75). Phosphatidylcholine is seldom observedin bacterial cells. The major outer membrane proteins hadmolecular sizes of 29,000 and 33,000 Da and were bothmodified by heating. The 29,000-Da protein was tightlyassociated with the peptidoglycan and was equally distrib-uted in the IM, OM-1, and OM-2 fractions.Gabay and Horwitz (82) isolated cell envelopes by treating

whole L. pneumophila Philadelphia 1 cells with lysozymeand EDTA to convert them to spheroplasts and then lysingthe spheroplasts osmotically or sonically. They resolved thecell envelopes into two membrane fractions by isopycniccentrifugation. NADH oxidase was localized to the fractionof buoyant density 1.145, which was designated cytoplasmicmembrane, and LPS was found in the fraction of density1.222, which was designated outer membrane. In contrast tothe results of Hindahl and Iglewski (92), KDO was localizedto the outer membrane fraction. SDS-PAGE revealed thatthe L. pneumophila outer membrane contains a single majorprotein species migrating at 28,000 Da; this is the MOMP of

the bacterium. The cytoplasmic membrane also contains asingle major protein species migrating at 65,000 Da. Surfaceiodination of the bacteria and agglutination and immunoflu-oresence studies with rabbit antibody produced against thepurified MOMP revealed that this protein is exposed at thecell surface. These workers isolated LPS from L. pneumo-phila membranes by SDS-EDTA treatment. The patternobtained by subjecting the LPS to SDS-PAGE and stainingthe gel with silver nitrate indicated that L. pneumophila LPSmight be atypical. Two other lines of evidence suggest thatLegionella LPS is not typical of the LPS of most gram-negative bacilli. First, lipid analysis has shown that Legion-ella LPS has an unusual fatty acid composition consistingpredominantly of branched-chain fatty acids; hydroxy-fattyacids, which are generally associated with classical endotox-ins as structural components of lipid A, are absent (75, 136,204). Second, Legionella LPS lacks the classical endotoxic-ity associated with LPS in that it induces a weak pyrogenic-ity response in rabbits and has relatively low toxicity formice (204).Gabay and Horwitz (82) also studied serologic responses

of patients to cell envelope components of L. pneumophilaPhiladelphia 1. Serum samples from patients with evidenceof infection with L. pneumophila serogroup 1 containedlarge amounts of antibody to this strain. Few of theseantibodies recognized the MOMP of L. pneumophila. Incontrast, more than 98% of the antibodies were directedagainst the LPS. These results and the work of others (46,48, 145, 149) have established that LPS is the dominantserogroup-specific antigen and the major antigen responsiblefor the reactivity of patient serum in the indirect fluorescent-antibody assay. Given the similarity of the patterns of LPSfrom virulent and avirulent strains of the same serogroup onSDS-PAGE, it seems unlikely that LPS is a virulencedeterminant for L. pneumophila (48).

Cell Surface Legionella Proteins

About 10% of the total protein mass of Legionella spp. islocalized to the outer membrane of the organism. Althoughouter membrane proteins are of potential use for purposes ofidentification, they are also of interest because they mayplay a role in pathogenesis and because the outer membraneantigens come into primary contact with the inflammatorycells and immune system of the host.MOMP. In the preliminary study of the outer membrane

proteins of L. pneumophila by Ehret and Ruckdeschel (68),it was reported that all 10 serogroups examined possessed a29-kDa MOMP. The MOMP, which is associated with pep-tidoglycan, has been isolated by several groups of investiga-tors (35, 81, 93). Gabay et al. (81) purified MOMP from L.pneumophila by a three-step procedure that involved extrac-tion of bacterial cells with calcium and detergent followed byion-exchange and molecular-sieve chromatography. Prepa-ration of MOMP monomers required the reducing agent2-mercaptoethanol, indicating that the protein aggregatesthrough the formation of interchain disulfide bridges (35, 81).Several researchers have demonstrated that the MOMP of L.pneumophila has a molecular mass between 24 and 29 kDa(35, 68, 81, 93). It is generally agreed that all Legionellaspecies, with the possible exception of serogroup 1 strains ofL. bozemanii (35), possess a disulfide-cross-linked MOMP ofsimilar molecular weight, although Hindahl and Iglewski (93)did not observe MOMP in species other than L. pneumo-phila. Gabay et al. (81) demonstrated that the MOMP of L.pneuimophila is a cation-selective porin. Porins are a class of

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 49

bacterial proteins that are capable of inserting themselvesinto membranes, including those of host cells, and openingup channels through which ions can pass. Horwitz (103) hasspeculated that one mechanism by which L. pneiimophilaPhiladelphia 1 might inhibit acidification of the monocytephagosome is by the insertion of a proton ionophore,namely, the cation-selective MOMP porin, into the mono-cyte membrane. Although as yet unproven, a 95-kDa com-plex may represent the porin consisting of four disulfide-cross-linked subunits of the MOMP, and this complex isbound to peptidoglycan in situ (35).The finding that human monocyte complement receptors

(CR), CR1 and CR3, and complement component C3 inserum mediate the phagocytosis of L. pneiumophila (152)prompted Horwitz to investigate the identity of the C3acceptor molecule on the bacterial surface. Bellinger-Kawa-hara and Horwitz (14) showed that C3 was fixed to L.pneumophila which were opsonized in fresh, nonimmuneserum, and C3 fixation took place via the alternate pathwayof complement activation. Immunoblot analysis of op-sonized L. pneumophila indicated that C3 was fixed exclu-sively to the MOMP; C3 did not fix to LPS on these blots.Furthermore, when purified MOMP was incorporated intoliposomes, the MOMP-liposomes avidly fixed C3 and wereefficiently phagocytized by monocytes. Since the MOMP ispresent in virulent and avirulent Legionella strains (35), itsrole in pathogenesis is unclear. However, it is a substancedeserving of further study. C3 is the major opsonin for thephagocytosis of the legionellae (152, 180), so that if theMOMP were modified it might lead to decreased internaliza-tion of the bacterium and subsequent failure to multiply. Apreliminary report indicates that a strain of L. pnelimophilarendered avirulent by prolonged agar passage showed agreatly diminished expression of the MOMP (167). When a750-bp DNA fragment from the virulent L. pneuimophilastrain was transformed into E. coli, the resulting clonesexpressed the Legionella MOMP and had increased viru-lence in chicken embryos compared with the parent E. colistrain. Thus, reduced expression of MOMP could be themechanism behind the attenuation of some strains which donot appear to be phagocytized as effectively as their virulentcounterparts (see below).Mip protein. Instead of using biochemical analyses to

identify components of Legionella spp. that may be viru-lence factors, Cianciotto, Eisenstein, Engleberg, and theircoworkers initiated a genetic analysis of intracellular para-sitism. To identify targets for mutagenesis, they clonedgenes encoding surface protein antigens and chose the geneencoding a 24-kDa L. pneumoplhila-specific antigen thatinduced strong antigenic reactivity in rabbits immunizedwith killed bacterial cells (153) for further investigation. Thisprotein is distinct from the MOMP of L. pneumophila. Thegene encoding the 24-kDa protein was deleted by site-specific mutagenesis involving allelic exchange with specificloss of 24-kDa antigen expression (45). Compared with theisogenic parent, the mutant was significantly impaired in itsability to infect transformed human U937 cells and humanalveolar macrophages. U937 cells are monoblastic in contin-uous culture, but lose their proliferative capacity and de-velop features of tissue macrophages after treatment withinducers such as PMA. An 80-fold-greater inoculum of themutant strain than of the parent strain was required to infectU937 cells. The mutant strain regained full infectivity fol-lowing reintroduction of the cloned 24-kDa protein gene.Although the titer of the mutant strain was lower immedi-ately following infection, the growth rates of the mutant and

parent strains in U937 cells were similar over the next 40 h,indicating that the mutant strain was less able to initiateinfection, but remained capable of intracellular multiplica-tion. When opsonized with specific antibody, the mutantstrain still demonstrated reduced infectivity despite equiva-lent cell association, indicating that the mutant did not lacka ligand required for macrophage attachment. Thus, the24-kDa protein appears to be required either for optimalinternalization of L. pneumophila by macrophages or forresistance to the bactericidal mechanisms that are operativeimmediately following phagocytosis. These workers havedesignated the gene encoding the 24-kDa protein as mip (formacrophage infectivity potentiator) and the gene product asMip.To substantiate the importance of Mip in the pathogenesis

of L. pneirnophila infection, the lethality of the parent andmutant strains for guinea pigs inoculated intratracheally wasexamined (44). The mip mutant killed fewer animals andproduced illness later after inoculation than did the isogenicparent strain. Although the mutant strain was capable ofcausing illness and death at a high dose, it was significantlyless virulent than the parent strain. Furthermore, bacteriaderived from the mutant strain into which the mip gene hadbeen reintroduced were as virulent for guinea pigs as theparent strain was. The titer of the mip mutant in the lungs ofanimals 48 h after infection was lower than that of the parentstrain, but the difference was not statistically significant; thetiters of lung homogenates from animals infected with thereconstituted strain were intermediate between those of theother two. Interestingly, the spleen titers of animals infectedwith the mnip mutant were 23- and 20-fold lower, respec-tively, than those of animals receiving either the parent orreconstituted mip+ strain; these are highly significant differ-ences. There was no discernible difference between the mipmutant and parent strains in proteolytic and phosphataseactivities, complement fixation, serum resistance, or LPSstructure. A derivative of the mip+ strain which had beenrendered avirulent by multiple passages on agar still ex-pressed a 24-kDa Mip protein. Therefore, a different mech-anism appears to be operative in legionellae rendered avir-ulent by agar passage, suggesting that multiple virulencefactors may be operative in wild-type bacteria.The DNA sequence encoding the Mip protein has been

established (71). The inferred polypeptide is a potent poly-cation with an estimated pl of 9.8. Although the mechanismby which the Mip protein promotes the invasion of hostphagocytes is unknown, the polycationic character of thissurface molecule raises some logical possibilities (71). Poly-cations are known to induce the phagocytosis of inertparticles, and increased phagocytosis of legionellae mightpromote their multiplication. Alternatively, following bacte-rial uptake, this protein could act as a cationic lysosomotro-pic agent to inhibit phagosome-lysosome fusion and acidifi-cation. However, if inhibition of phagosome-lysosomefusion were the mechanism of Mip action, it is unclear whyonly the Philadelphia 1 strain of L. pneumophila has beenobserved to inhibit fusion, since the mip gene is conservedthroughout the species (44). It is intriguing that the carboxy-terminal 114 residues of the Mip protein share 39% identitywith a protein isolated from Neurospora crassa which bindsthe new immunosuppressive drug, FK 506 (188). The Neu-)'ospoIra protein exhibits peptidyl-prolyl cis-trans isomeraseactivity. Prolyl isomerases are thought to catalyze a step inprotein folding. It is not known whether Mip acts as a prolylisomerase or binds FK 506.A large panel of Legionella strains were examined by

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

50 DOWLING ET AL.

Southern hybridization and immunoblot analyses for thepresence and expression of the mip sequence (42). Strainsrepresenting all 14 serogroups of L. pneiumophila containeda mip gene and expressed a 24-kDa Mip protein. Althoughthe DNAs for 29 other Legionella species did not hybridizewith the mip DNA probe under high-stringency conditions,they did so at reduced stringency. Furthermore, these spe-cies each expressed a 24- to 31-kDa protein that reacted withspecific Mip antisera. It was also shown that the clonedmip-like gene from L. micdadei encoded the cross-reactiveprotein. Thus, although the mip gene is conserved in, and isspecific to, L. pneumophila (44), mip-like genes are presentthroughout the genus. Whether the various Mip proteins arefunctionally similar remains to be determined. If the Mip-related proteins of the other Legionella species are function-ally dissimilar to the Mip protein of L. pneumophila itself, itmay help to explain the apparent greater virulence of thelatter species. The mip gene family appears to be limited tothe genus Legionella. DNAs from a number of gram-nega-tive bacilli, gram-positive cocci, and Mycoplasma pneiumo-niae did not hybridize with probes known to contain mip(42). Likewise, Mip-related proteins were not detected inother gram-negative bacilli or Mycoplasma spp. (72).Heat shock protein. All Legionella species and serogroups

that have been examined express a 58- to 60-kDa proteinwhich contains a genus-specific epitope recognized bymonoclonal antibodies, as well as epitopes which are cross-reactive with many species of gram-negative bacteria (94,178). The gene coding for the L. pneumophila 58- to 60-kDacommon antigen has been cloned in E. coli, and its completenucleotide sequence has been determined (94, 95, 178). Theprotein is preferentially synthesized upon heat shock and isserologically cross-reactive and demonstrates considerablehomology with other heat shock proteins including theGroEL protein of E. coli, the Mycobacterium tuberculosis65-kDa antigen, and Coxiella burnetii HtbB. The protein ishighly immunogenic and is the predominant Legionellaprotein reactive with human convalescent-phase serum frompatients with confirmed cases of legionellosis. It has beenshown that the purified L. pneumophila 60-kDa protein isantigenic for human T lymphocytes (95). Furthermore, indi-rect fluorescent-antibody studies indicated that this proteinmay be located in the periplasmic space or expressed on thesurface of intracellular bacteria. Thus, although no evidencehas accrued to establish the Legionella heat shock protein asa virulence factor, it seems likely that this protein would beinduced by the unfavorable conditions in the phagosome.What effects the release of this protein might have onphagocyte function which could lead to the abrogation ofcellular bactericidal mechanisms remain to be determined.

AVIRULENT LEGIONELLA MUTANTS

Mutants, particularly avirulent mutants, can be powerfultools for analyzing virulence determinants. One strategy forelucidating the virulence determinants of the legionellae is toisolate and characterize mutants that are unable to multiplyin monocytes or monocytic cell lines (71) or to survive inneutrophils infected in vitro. In this manner it might bepossible to identify mutants blocked at each of the hypothet-ical steps required for successful intracellular parasitism:uptake by or entry into the cell, resistance to cellularbactericidal mechanisms, and ability to multiply within thephagosome. Two strategies have been used for the produc-tion of avirulent Legionella mutants. Some investigatorshave examined spontaneously occurring avirulent mutants,

whereas others have genetically engineered the bacteria andthen screened for possible decreased virulence. As justdescribed for definition of the Mip protein, the latter ap-proach has the advantage that a fully virulent, isogenicparent strain is available for comparative purposes.When passaged on suboptimal artifical media, L. pneumo-

phila and L. miedadei spontaneously convert to mutantforms which are avirulent for guinea pigs. The mediumcommonly used for this purpose is supplemented Mueller-Hinton agar. The casein hydrolysate component of supple-mented Mueller-Hinton medium has been shown to beinhibitory to the growth of virulent, but not avirulent,bacterial cells, and the inhibitory component of the hydroly-sate was identified as NaCl (39). The importance of intracel-lular growth to virulence expression by Legionella cells isunderscored by the inability of avirulent strains produced byagar passage to replicate intracellularly. Kishimoto et al.(116) were the first to find that virulent strains of L. pneu-mophila serogroup 1 survived and proliferated in guinea pigperitoneal macrophages, whereas an avirulent strain derivedfrom Philadelphia 1 (passage history unknown) was killed bynormal macrophages. The phagocytosis of the virulent andavirulent strains appeared to be equivalent. Jacobs et al.(109) undertook extensive studies of the interaction of viru-lent and avirulent L. pneumophila serogroup 1 strains withpigtail monkey alveolar macrophages; the avirulent strainwas obtained by repeated passage of originally virulentbacteria on modified Muller-Hinton agar. Primate alveolarmacrophages phagocytized comparable numbers of virulentand avirulent bacteria in the absence of specific antibody.The majority of the bacteria which were associated withalveolar macrophages were killed, and virulent and avirulentLegionella cells were equally susceptible to this early killing.The virulent L. pneumophila cells that survived intracellu-larly increased in number by over 2 logs during the 96 h afterinfection. In contrast, the avirulent bacteria multiplied muchmore slowly over the same period. Neither the mechanism ofthe survival of a fraction of both virulent and avirulentLegionella cells following phagocytosis nor the mechanismfor the subsequent unrestrained multiplication of the virulentbacteria is known. Similarly, virulent L. pneumophila sero-group 1 cells multiplied in the human macrophagelike U937cell line and produced cytopathic effect during intracellulargrowth, whereas the avirulent mutant obtained after morethan 200 passages on suboptimal medium did not (45).Horwitz (102) obtained 44 mutant clones of L. pneumo-

phila Philadelphia 1 by batch passaging wild-type bacterianine times on supplemented Mueller-Hinton agar. Then 44colonies were selected, and each was passaged individuallyan additional three times. None of the 44 mutant clonesmultiplied in human monocytes in experiments in whichwild-type L. pneumophila multiplied 2.5 to 4.5 logs. Like thewild type, the avirulent mutants were resistant to the bacte-ricidal effects of complement in the presence or absence ofhigh-titer antibody. Both mutant and wild-type bacteriabound to and were ingested by monocytes, and both enteredby coiling phagocytosis. The wild type formed the distinctiveribosome-lined phagosome, inhibited phagosome-lysosomefusion, and multiplied intracellularly. The avirulent mutantdid not form the distinctive phagosome or inhibit phago-some-lysosome fusion; it survived intracellularly but did notreplicate in the phagosome. Essentially the same resultswere obtained by using the promyelocytic cell line, HL-60,after its differentiation into macrophagelike cells (129).

All of the preceding studies indicate that there is noquantitative difference in the uptake of virulent and avirulent

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 51

L. pneumophila strains by phagocytic cells. In contrast.Dreyfus (66), using an assay in which the extracellularbacteria are killed with an antibiotic, found that L. pnelino-phila serogroup 1 strains rendered avirulent by passage onagar entered HeLa cells approximately 1,000-fold less effi-ciently than did virulent strains. When radiolabeled bacteriawere used, no difference was found in the numbers ofvirulent and avirulent bacteria which were associated withthe HeLa cell monolayers after 1 h of incubation. Thesefindings may apply only to nonprofessional phagocytes, orthe divergent results may be due to the different methodsused to enumerate intracellular bacteria. Alternatively, thevarious strains derived by prolonged agar passage may differin the particular virulence mechanism which has been al-tered.One possible differentiation between virulent and aviru-

lent legionellae was suggested by prior work indicating thatavirulent strains of Salmonella typhi and S. alureius stimu-lated oxygen consumption by neutrophils whereas a virulentstrain of Salmonella typhi did not (134). Indeed, there was apreliminary report indicating that a virulent strain of L.pneumophila stimulated significantly less 02 consumptionand chemiluminescence by neutrophils than an avirulentstrain did (192). Immune human serum was used as theopsonin in these studies, and no qualitative difference inbacterial uptake by the neutrophils was observed. Unfortu-nately, the method by which the avirulent strain was derivedwas not reported. On the other hand, we (60, 61) found thatvirulent L. micdadei stimulated the metabolic burst of neu-trophils and monocytes during phagocytosis to the sameextent as S. aureius and E. coli.

Summersgill et al. (182) examined four strains of L.pneumophila serogroup 1 and their avirulent variants (cre-ated by multiple passages on agar) for their effect on neutro-phil function, as determined by measurement of peakchemiluminescence and 02 production during phagocyto-sis. When neutrophils were exposed to either virulent oravirulent strains of L. pneiumophila in the presence ofnormal serum, their 2- production and chemiluminescencewere markedly lower than those observed with E. coli. Thisobservation may be explained by the fact that both specificantibody and complement are required for efficient phago-cytosis of L. pneumophila by neutrophils (106); a reduceduptake in the absence of specific antibody would be expectedto result in less stimulation of oxidative metabolism. How-ever, the reduction in neutrophil oxidative function wasmore pronounced with the virulent member of each pair ofstrains in both assays. To determine whether the differencein reduced neutrophil function was due to C3 binding, thefour pairs of bacteria were incubated in normal human serumand examined by quantitative immunofluoresence. Bothvirulent and avirulent legionellae bound less C3 than E. colidid, but the relative quantities of C3 bound did not match therelative reduction in oxidative function produced by eachstrain. Three of the four virulent bacteria bound less C3 thanthe avirulent member of the pair did. These workers con-cluded that some of the depressed neutrophil function fol-lowing exposure to virulent L. pnelumophila may be relatedto reduced C3 binding, although other virulence-associatedfactors may also be involved. The relationship of virulenceand opsonization was supported by Plouffe et al. (156), whostudied two subtypes of L. pneumophila serogroup 1 isolatedat one medical center. The attack rate for one strain was10-fold higher than for the other. The clinically less virulentstrain was killed by fresh serum and fixed qualitatively more

complement, whereas little killing of the more virulent strainoccurred in fresh serum and it fixed less complement.

Summersgill et al. (183) then examined the interactions ofthese same four virulent-avirulent pairs of L. pneumophilawith human peripheral blood monocytes. All of the L.pneumnophila strains elicited less oxidative response bymonocytes in the presence of normal serum, as measured byboth O,- and H,O, production, than did E. coli. Also, theavirulent member of each pair of bacteria evoked moreoxidative response than did the virulent strain. All of thevirulent strains were capable of multiplying in monocytes toa high titer over a period of 3 days. The avirulent strainsfailed to multiply, and the titers fell to undetectable levelsafter 24 h of incubation. To assess the effects of complementfixation by each strain, phagocytic indices were determinedunder various conditions. In the presence of autologousdonor serum, all L. pneiumophila strains had phagocyticindices markedly lower than that of E. coli. There were nodifferences between the phagocytic index of the virulent andavirulent member of each pair. When heat-inactivated serumwas used, all L. pneiumophila phagocytic indices fell mark-edly, as did that for E. coli, and the indices were restored bythe addition of human complement to heat-inactivated se-rum. Therefore, although the importance of complement inthe adherence and uptake of L. pneumophila was confirmed,the binding of complement did not provide an explanation ofthe difference between virulent and avirulent bacteria ineliciting an oxidative response or multiplying in monocytes.There is conflicting evidence regarding whether avirulent

Legionella strains are more susceptible to killing by oxygenmetabolites than are virulent strains (111, 122, 123). Locks-ley et al. (123) and Lochner et al. (122) reported that virulentand avirulent strains of L. pneiumophila were equally sus-ceptible to the antimicrobial species generated by the xan-thine oxidase reaction and the MPO-H202-halide system. Incontrast, Jepras and Fitzgeorge (111) found that two virulentL. pneiumophila serogroup 1 strains resisted the bactericidalactivity generated by the xanthine oxidase system while twoavirulent strains did not. In addition, the catalase activity ofthe strains was directly correlated with their virulence andresistance to H202 and the toxic oxygen metabolites gener-ated by the xanthine oxidase system. All four strains of L.pnelimophila were equally susceptible to MPO-mediatedkilling. These results would help explain why the legionellaecan survive and multiply in mononuclear phagocytes but arekilled, albeit inefficiently, by neutrophils. It is not clear whydifferent results were obtained in the three studies. Theearlier workers (122, 123) used strains which had theirvirulence defined by intraperitoneal infection of guinea pigsand lethality for embryonated eggs, whereas Jepras andFitzgeorge (111) defined virulence by the dose required tokill guinea pigs inoculated by aerosol. Although the latter isa more relevant definition of virulence, it may still be thatavirulent strains derived in the laboratory by agar passagediffer in the particular virulence mechanism(s) which isaltered.

Mintz et al. (135) isolated several independent thymidine-requiring auxotrophs of L. pneumophila serogroup 1. Thethymidine auxotrophs exhibited a marked decrease in via-bility when they were deprived of thymidine, and they wereincapable of intracellular survival or multiplication in phe-ripheral blood monocytes. In contrast, both the wild-typestrain and prototrophic revertants were capable of multipli-cation in monocytes. It was not clear whether the thymidineauxotrophs died intracellularly directly as a result of thymi-dine deprivation or because they were unable to produce a

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

52 DOWLING ET AL.

substance necessary for intracellular survival and growth, orboth.

Virulent strains of L. pneirnophila that are preincubatedwith specific antiserum are cytotoxic for guinea pig alveolarmacrophages when added at a multiplicity of infection of 100bacteria per macrophage (37). Incubation of virulent L.pneiumophila (Philadelphia 2) with complement and specificantibody resulted in an increase in the cytopathic effect. Incontrast, an avirulent strain, obtained by passage of thevirulent strain on supplemented Mueller-Hinton agar, killedfew of the macrophages even when preincubated with spe-cific antiserum or antiserum and complement. The cytotoxiceffect required the phagocytosis of viable bacteria, suggest-ing that a bacterial toxin was involved. However, attemptsto demonstrate toxic activity in concentrated culture fluidsor in sonicated cell suspensions were unsuccessful. Thevirulent strain was also resistant to the bactericidal activityof normal human serum and failed to fix complement com-ponents in the presence of normal human serum. In contrastto the preceding avirulent L. pneiinophila serogroup 1strains derived from Philadelphia 1 (102), this avirulentbacterium was serum sensitive and both C3 and C9 weredeposited on its surface.These results are not easy to interpret because it is not

clear which bacterial moiety causes the cytotoxicity pro-duced by the legionellae. The assay originally used byFriedman et al. (78) to define the presence of toxin in thesupernatants of broth-grown L. pniei,nophila was cytotox-icity for CHO cells, as measured by a change in pH of themedium after prolonged incubation. As noted, the zincmetalloprotease is also cytotoxic for CHO (and other) cells.McCusker et al. (131) found that after 3 to 4 h of incubationwith L. pneiumophila, protein synthesis by CHO cells wasmarkedly inhibited, as shown by the reduction of incorpora-tion of radiolabeled amino acids into proteins. L. pnelino-phila did not inhibit the transport of amino acids or causedegradation of newly synthesized proteins in CHO cells.Cytochalasin D blocked the entry of Legionella organismsinto CHO cells, but did not reverse the inhibition of proteinsynthesis, indicating that an extracellular substance, such asthe toxin or protease, was responsible. However, the eluentfrom washed L. pneurnophila cells did not inhibit CHO cellprotein synthesis. Furthermore, inhibition of protein synthe-sis was not observed after L. pneiumophila had been boiledfor 2 min, suggesting that the toxic factor was neitherendotoxin nor the low-molecular-weight cytotoxin, both ofwhich are heat stable (62, 78). This finding also indicates thatdifferent moieties are responsible for the inhibition of proteinsynthesis and the inhibition of neutrophil and monocyteactivation by L. inicdadei, since the latter effect was ob-served when heat-killed bacteria were used (57, 59). It isintriguing that L. pneirnophila cells that had been killed withantibiotics prior to incubation with CHO cells still inhibitedprotein synthesis, indicating that inhibition of CHO cellprotein synthesis occurred in the absence of de novo proteinsynthesis by the bacteria. This result also implies thatbacteria which are inhibited or killed by antibiotic therapymay still have adverse effects on host cells. McCusker et al.(131) state that preliminary experiments show that L. pneiu-mophila also inhibits protein synthesis in U937 cells andhuman monocytes, cells which are more relevant to ourunderstanding of the pathogenesis of Legionella infection atthe cellular level.

In summary, avirulent mutants of L. pneirnopliia derivedby different means have been shown to be unable to replicateintracellularly. It is unlikely that strains derived by agar

passage in different laboratories necessarily have a defect inthe same virulence mechanism. The phagocytosis of theseagar-passed mutants is decreased in some cases, whereas forother attenuated strains uptake is normal. The mechanism ofreduced uptake may be decreased complement fixation,perhaps owing to decreased expression of the MOMP (167),in some, but not necessarily all, instances. Avirulent mu-tants which are ingested normally presumably have a defectin their ability to withstand or nullify cellular bactericidalmechanisms. However, the particular mechanism has notbeen defined for any avirulent mutant, so it is not possible tolink the work done on these mutants with the effects of anydefined antiphagocyte moiety. Interpretation of these resultsis further complicated by the fact that mutants derived bypassing a virulent strain on artifical medium may containmultiple mutations (45).

CONCLUSIONS AND FUTURE RESEARCH

The systematic search for Legionella virulence factorswhich promote survival within phagocytes can proceedalong one of two routes. In the first approach individualvirulence factors can be isolated and purified, utilizing somemodel system to demonstrate virulence. Once purified, thebiochemical or molecular mechanism of the action of thefactor can be pursued; if this is consistent with the functionalactivity of the factor, it adds credence to the importance ofthe moiety in pathogenesis. Virulence factors identified bythis route can be classified as sufficient to promote virulencein the particular model system used. The second systematicroute to the identification of virulence factors is to usemolecular biological methods to create mutants which areisogenic with the parent strain in all respects except theproduction of a single gene product and to show that themutant is avirulent in some model system in which theparent is virulent. It would appear that this plan would haveto be used to identify moieties which are structural compo-nents of the bacterium rather than exoproducts. This schemewould seem the only one whereby necessary, as opposed tosufficient, virulence factors may be identified. However, thedata obtained may not always be clear-cut. The secondapproach was the one used in identifying the Mip protein asa virulence factor, but the Mip- mutant, although attenu-ated. was not totally avirulent in the model systems studied.The limitation of both schemes is that the acceptance of themoiety as a virulence factor depends entirely on the rele-vance of the model system for virulence which was used andhow faithfully the system represents the events which takeplace in nature.Although all investigations of Legionella virulence have

not been as systematic as the schema just described, theevidence that a particular moiety is important in the ability ofthe legionellae to invade and persist in host phagocytic cellstakes one of three forms: functional, biochemical, or genetic.The evidence that the Legionella toxin is a virulence factoris entirely functional in that partially purified toxin adverselyaffects neutrophil activation. The biochemical basis for toxinaction cannot be addressed until this moiety is purified tohomogeneity and identified. In addition to functional effectson neutrophils, a biochemical mechanism of action has beenestablished for the phosphatase. However, as yet, otherlegionellae have not been examined to determine whetherthey contain the same active phosphatase which is present inL. inicdatdei. It has not been rigorously determined whetherthe phosphatase is secreted by Legionella spp. or is found inthe periplasmic space, nor is it known how the phosphatase

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 53

reaches its substrates in the phagocytic cell. Also, the basicwork required to engineer a phosphatase-negative mutantand determine whether it is avirulent has not been done.At the other end of this spectrum is the Mip protein. The

gene encoding this Legioniella cell surface 24-kDa proteinhas been cloned, and an isogenic mutant lacking a functionalgene has been constructed. Compared with the parent strain,the mutant was impaired in its ability to infect macrophagecell lines and produce disease in guinea pigs. The function ofthe Mip protein is unknown, but it appears to be involvedwith the initiation of intracellular infection; multiplication ofthe mutant lacking Mip after being internalized was appar-ently normal. The possibility remains that these propertiesare not functions of Mip itself. Mip is probably a prevalentprotein in the outer cell membrane. The absence of Mipcould so alter the structure of the outer membrane that someother component required for invasion or intracellular sur-vival is altered or the overall surface is so distorted thatingestion is impaired. It remains to be determined whetherthe Mip protein is a virulence factor only for L. pieitinophilaor is operative throughout the genus. Either the function ofMip must be elucidated and the Mip-like proteins of theother species shown to have similar actions, or geneticallyengineered legionellae of other species lacking their Mip-likeproteins would have to be shown to be attenuated. It shouldalso be noted that, although Mip appears to be necessary foroptimal cellular invasion, the inip mutant did multiply inmacrophages and produced illness in guinea pigs. That L.pneiiinophila inip mutants are not totally avirulent indicatesthat additional factors are involved in virulence. In addition,all putative virulence factors other than the toxin are heatlabile. Since it has been shown that following the phagocy-tosis of heat-killed L. nicdaldei the activation of neutrophilsand monocytes is severely impaired, heat-labile factors mayplay only secondary or ancillary roles to that of the toxin inregard to defeating phagocyte bactericidal functions.

In addition to the toxin, phosphatase, and Mip protein,there are other putative virulence factors which may in-volved in the intracellular survival and multiplication oflegionellae. The demonstration of protein kinase activity inL. micdadei together with phosphatase activity indicatesthat the bacterium may possess the capability of regulatingthe properties and function of host cell proteins and lipidsthrough phosphorylation and dephosphorylation reactions.Although it was shown that the L. nic-dadei enzyme phos-phorylates PI in the plasma membrane of intact neutrophils,there is no evidence that the purified protein kinase ad-versely affects phagocyte function. It also has not beendetermined whether other Legionella species possess aPI-protein kinase, where the kinase is located in the bacterialcell or if the bacterial kinase phosphorylates PI or othercellular substrates during infection. The possibility that theLegionella phospholipase C is a virulence factor is entirelytheoretical. Although a number of mechanisms whereby aphospholipase C might contribute to intracellular parasitismcan be put forward, there are no functional or biochemicaldata in this regard for the purified protein.The zinc metalloprotease secreted by Legionelli spp. is

cytotoxic and hemolytic and induces pulmonary lesionswhich resemble those found in infection. However, despitethese indications that the extracellular protease might be avirulence factor, there is now strong evidence that theprotease is not necessary for cellular invasion and growth ofthe legionellae. Not only is the protease produced by aviru-lent legionellae, but also a genetically engineered. protease-negative strain is fully virulent. On the other hand, guinea

pigs immunized with the purified L. piieirnopliila proteaseare resistant to challenge with an otherwise lethal quantity ofL. piweuml7opllil(i cells. Although the protease turns out not tobe a virulence factor, the research effort expended on thismoiety has yielded a very promising immunogenic agent. Infact. one generalization that can be made is that there may belittle relationship between the role of a particular moiety asa virulence factor, defined as promoting invasiveness andintracellular survival or multiplication, and the utility of themoiety in producing protective immunity. Not only does thepurified protease engender protection, but also guinea pigsimmunized by aerosol exposure to L. pneuinopliila renderedavirulent by multiple agar passage are protected against anotherwise lethal aerosol challenge of wild-type bacteria (26).Apparently, the common denominator for the protectiveefficacy is that immunization with either the protease oravirulent bacteria invokes strong cell-mediated immune re-actions (26. 27). Cell-mediated immune responses which areeffective in eliminating intracellular bacteria or killing Le-gioilell(a-infected cells are then recalled upon challenge withvirulent bacteria. This scheme does not exclude the possi-bility that a necessary virulence factor will be found, immu-nization with which would protect by preventing the infec-tion of host phagocytes in the first place. For example,Hedlund (89) demonstrated that inoculation of AKR/J micewith crude toxin preparations from either L. pneitmophila orL. Inindadei protected against a lethal dose of either crudetoxin or viable bacteria from the same or the other species.These promising experiments have not been followed upbecause it has not yet been possible to purify and character-ize the Legiol7ella toxin.The foregoing considerations indicate that the capacity of

the legionellae to survive and multiply intracellularly ismultifactorial. This is not surprising since there are at leastfour attributes which Legionella spp. must possess to be asuccessful intracellular pathogen. First, the bacteria mustadhere or bind to the cell and, second, they must bephagocytized in order to gain access to the favored site formultiplication. In very few studies has any attempt beenmade to distinguish attachment from ingestion. However, itappears that some avirulent strains obtained by prolongedagar passage fix complement like wild-type Legionellastrains and are ingested normally by phagocytes, whereasother similarly derived avirulent mutants are not taken up byphagocytic cells. Third, the legionellae must either inhibitphagosome-lysosome fusion and/or oxidative metabolism orresist killing by bactericidal oxygen species and phagosomalcontents, in order to survive immediately after ingestion.Finally, a virulent bacterium must be capable of multiplica-tion within the cell, including, for example, having thecapacity to obtain the appropriate nutrients within the pha-gosome. The inability of thymidine-requiring mutants togrow intracellularly may be an example of avirulence basedon inability to acquire a necessary nutrient. Loss of theability to carry out any one of these steps may cause thebacterium to be attenuated to some extent, but perhaps nottotally. Therefore, it is probable that more than one factor isnecessary for the virulent phenotype and unlikely that asingle entity will be found which is, of itself, sufficient toexplain the intracellular multiplication of Legionella cells.This conceptualization may explain why multiple putativevirulence factors are being discovered, some of which ap-pear to promote entry into the cell and others which blockvarious antibacterial properties of phagocytes. If there isredundancy on the part of the legionellae, there may be aneven greater number of virulence factors. This does appear

VOL. 56. 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

54 DOWLING ET AL.

to be the case since at least two Legionella moieties (toxinand phosphatase) have similar inhibiting effects on neutro-phil activation.As is obvious from the data on the cytotoxicity and protein

synthesis inhibition engendered by the legionellae, as well asthe similar effects of various moieties on phagocyte func-tions, a primary future task is to determine which bacterialfactor(s) is responsible for each of the defects in phagocyteactivation and bactericidal activity produced by Legionellaspp. As the number of possible virulence factors grows,determining the role of each in pathogenesis at the cellularlevel will depend on the development of mechanistic data. Atfirst, this data may be descriptive; e.g., a particular factorinhibits phagocyte oxygen metabolism or phagosome-lyso-some fusion. Eventually, the molecular or biochemicalmechanism of action should be determined; to date this hasbeen accomplished only for the Legionella phosphatase. Asthe ultimate test, the techniques of molecular biology shouldbe applied to each putative virulence factor. It should bedemonstrated that a genetically engineered bacterium whichlacks the virulence factor is less pathogenic in a relevantcellular or animal model than the isogenic parent strain.The elucidation of the pathogenetic mechanisms by which

the legionellae evade host phagocyte bactericidal propertiesduring the early stages of Legionella infection may beexpected to yield two more immediate practical benefits inthe therapy and prevention of legionellosis. First, the limitsof conventional antimicrobial therapy may have beenreached now that it is understood that inhibitory or bacteri-cidal antimicrobial agents which penetrate phagocytes mustbe used (65, 150). Further therapeutic success may bepossible only by using immunotherapeutic modalities basedon a knowledge of pathogenesis at the molecular level. Forexample, if the basis of the failure of neutrophils to killlegionellae is due to a Legionella-induced block in cellactivation, therapy with a specific cytokine, such as tumornecrosis factor, might reverse the inhibition (23, 24). Sec-ond, it may be possible to develop an efficacious vaccine thatfunctions by counteracting one or more of the intracellularvirulence mechanisms rather than by stimulating cell-medi-ated immunity. The more that is known about pathogenesisat the cellular and molecular level, the more likely the abilityto specifically attack the weakest link in the chain of viru-lence determinants.

ACKNOWLEDGMENTS

Work in our laboratory was supported in part by Public HealthService grant R01 A117047 from the National Institutes of Health.We thank our coinvestigator, Gerald R. Donowitz, for reviewing

the manuscript and for his generous permission to include unpub-lished data from his laboratory.

ADDENDUM

After it was demonstrated that other Legionella speciespossess analogs of the L. pneumophila Mip protein (42), theL. micdadei mip gene was cloned and expressed in E. coli(5a). DNA sequence analysis of the L. micdadei mip genedisclosed 71% homology with the L. pneumophila gene. Thepredicted secondary structures of the two Mip proteins arequite similar. An extended alpha helix from residues 55 to120 of L. micdadei Mip protein could represent an elongatedstructure projecting from the surface of the bacterium, aspreviously suggested (71). A similar, extended alpha-helicalstructure was identified in the corresponding proteins of L.pneumophila and Chiamydia trachomatis (123a). The most

obvious difference between the L. pneumophila and L.micdadei Mip proteins is in the initial 22 carboxy-terminalamino acids. These residues are hydrophobic and have thepredicted structure of an alpha helix. This probably repre-sents a secretory signal, but in L. micdadei no typical signalcleavage site can be identified. In contrast, the L. pneumo-phila mip gene has an easily distinguishable leader sequencewith a typical cleavage site. Southern hybridization experi-ments indicated that the mip gene of L. micdadei hasextensive homology with the mip-like genes of severalLegionella species. It was suggested that the mip gene familyof various Legionella strains can be divided into threehomology groups: (i) the mip gene of L. pneumophila, whichhas moderate homology to its analogs in all other Legionellaspecies (42); (ii) the mip-like genes of L. micdadei, L. feeleii,L. jamestow,niensis, L. oakridgensis, L. quinlivanii, L.sainthelensi, L. spiritensis, and L. wadsworthii, which havemoderate homology to the L. pneumophila gene but exten-sive (.90%) homology with L. micdadei mip; and (iii) themip-like genes of the remaining Legionella species, whichhave moderate homology with both the L. pneumophila andL. ,nicdadei genes. Whether this genetic heterogeneity hasimplications regarding the virulence of the various speciesremains to be determined. Also awaiting elucidation is thestriking sequence similarity between the carboxy-terminalend of the two Legionella mip genes and the N. crassa (188)and human FK 506 binding proteins, as well as a crypticNeisseriia meningitidis sequence (179a). The extended alphahelix in the N-terminal part of the Mip proteins may form arodlike spacer arm extending from the bacterial cell whichmay bring the enzymatically active carboxy-terminal endclose to a target membrane structure. Whether the Mipproteins act enzymatically as prolyl-isomerases, as do theFK 506-binding proteins, remains to be determined. Prelim-inary experiments showed that a recombinant E. coli ex-pressing the L. inicdadei Mip protein was taken up better byhuman monocytes than was E. coli carrying only the vector(5b), indicating that the L. micdadei Mip protein, like its L.pnelumophila counterpart, is necessary for optimum uptakeby phagocytic cells.A specific DNA probe for the legiolysin gene (lly) cloned

from L. pneirmophila Philadelphia 1 has been used in South-ern hybridizations to detect lly-specific DNA in the genomesof legionellae and other gram-negative pathogenic bacteria(1Sa). Under conditions of high stringency, the ily reactedonly with DNA from L. pneumophila isolates. However,under low-strigency conditions, hybridization was observedfor all the Legionella strains tested. No hybridization oc-curred with DNAs from bacteria of other genera. All but oneL. pneiumophila strain also produced Lly proteins that weredetected in Western blots by using anti-Lly antibodies. Theprotein was not detected in non-L. pneumophila strains or inother gram-negative bacteria, so that legiolysin appears to bespecific to L. pneumophila. Since the zinc metalloprotease isalso specific to L. pneumophila (157), the hemolysis ob-served for numerous Legionella species appears to be due toa moiety other than legiolysin or the protease. As yet, thereis no evidence the legiolysin is a virulence factor or plays arole in pathogenesis at the cellular level. In fact, lly wasdetected in an avirulent strain of L. pneumophila Philadel-phia 1.

REFERENCES1. Anand, C. M., A. R. Skinner, A. Malic, and J. B. Kurtz. 1983.

Interaction of L. pneumnophila and a free living amoeba (Acan-thamoeba palestinensis). J. Hyg. Camb. 91:167-178.

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 55

2. Badwey, J. A., and M. L. Karnovsky. 1986. Production ofsuperoxide by phagocytic leukocytes: a paradigm for stimulus-response phenomena. Curr. Top. Cell. Regul. 28:183-208.

3. Baine, W. B. 1985. Cytolytic and phospholipase C activity inLegionella species. J. Gen. Microbiol. 131:1383-1391.

4. Baine, W. B. 1988. A phospholipase C from the Dallas 1Estrain of Legionella pneiumophila serogroup 5: purification andcharacterization of conditions for optimal activity with anartifical substrate. J. Gen. Microbiol. 134:489-498.

5. Baine, W. B., J. K. Rasheed, D. C. Mackel, C. A. Bopp, J. G.Wells, and A. F. Kaufmann. 1979. Exotoxin activity associatedwith the Legionnaires disease bacterium. J. Clin. Microbiol.9:453-456.

Sa.Bangsborg, J. M., N. P. Cianciotto, and P. Hindersson. 1991.Nucleotide sequence analysis of the Legionella micdadei mipgene, encoding a 30-kilodaltron analog of the Legionella pneiu-mophila Mip protein. Infect. Immun. 59:3836-3840.

5b.Bangsborg, J. M., N. P. Cianciotto, J. B. Wright, and P.Hindersson. 1991. The Legionella micdadei mip analog: DNAsequence analysis of the gene and an evaluation of its role inmonocyte infection. Program Abstr. 31st Intersci. Conf. Anti-microb. Agents Chemother., p. 171.

6. Barbaree, J. M., B. S. Fields, J. C. Feeley, G. W. Gorman, andW. T. Martin. 1986. Isolation of protozoa from water associ-ated with a legionellosis outbreak and demonstration of intra-cellular multiplication of Legionella pneumophila. AppI. Envi-ron. Microbiol. 51:422-424.

7. Baskerville, A., J. W. Conlan, L. A. E. Ashworth, and A. B.Dowsett. 1986. Pulmonary damage caused by a protease fromLegionella pneumophila. Br. J. Exp. Pathol. 67:527-536.

8. Baskerville, A., A. B. Dowsett, R. B. Fitzgeorge, P. Hambleton,and M. Broster. 1983. Ultrastructure of pulmonary alveoli andmacrophages in experimental legionnaires' disease. J Pathol.140:77-90.

9. Baskerville, A., R. B. Fitzgeorge, M. Broster, and P. Hamble-ton. 1983. Histopathology of experimental legionnaires' dis-ease in guinea pigs, rhesus monkeys and marmosets. J. Pathol.139:349-362.

10. Beaman, L., and B. L. Beaman. 1990. Monoclonal antibodiesdemonstrate that superoxide dismutase contributes to protec-tion of Nocardia asteroides within the intact host. Infect.Immun. 58:3122-3128.

11. Becker, E. L. 1986. Leukocyte stimulation: receptor, mem-brane and metabolic events. Fed. Proc. 45:2148-2150.

12. Becker, E. L., J. C. Kermode, P. H. Naccache, R. Yassin, M. L.Marsh, J. J. Munoz, and R. 1. Sha'afi. 1985. The inhibition ofneutrophil granule enzyme secretion and chemotaxis by per-tussis toxin. J. Cell Biol. 100:1641-1646.

13. Becker, E. L., J. C. Kermode, P. H. Naccache, R. Yassin, J. J.Munoz, M. L. Marsh, C. K. Huang, and R. I. Sha'afi. 1985.Pertussis toxin as a probe of neutrophil activation. Fed. Proc.45:2151-2155.

14. Bellinger-Kawahara, C., and M. A. Horwitz. 1990. Comple-ment component C3 fixes selectively to the major outer mem-brane protein (MOMP) of Legionella pneumophila and medi-ates phagocytosis of liposome-MOMP complexes by humanmonocytes. J. Exp. Med. 172:1201-1210.

15. Belyi, I. F., I. V. Vertiev, I. S. Tartakovskii, I. V. Ezepchuk,and S. V. Prozorovskii. 1988. Characteristics of the cytolysin ofLegionella pneumophila. Zh. Mikrobiol. Epidemiol. Immuno-biol. Feb.(2):4-7.

15a.Bender, L., M. Ott, A. Debes, U. Rdest, J. Heesemann, and J.Hacker. 1991. Distribution, expression, and long-range map-ping of legiolysin gere (lv)-specific DNA sequences in legion-ella. Infect. Immun. 59:3333-3336.

16. Benson, R. F., W. L. Thacker, J. A. Lanser, N. Sangster, W. R.Mayberry, and D. J. Brenner. 1991. Legionella adelaidensis, anew species isolated from cooling tower water. J. Clin. Micro-biol. 29:1004-1006.

17. Berdal, B. P., K. B0vre, 0. Olsvik, and T. Omland. 1983.Patterns of extracellular proline-specific endopeptidases inLegionella and Flat'obacteriun spp. demonstrated by use ofchromogenic peptides. J. Clin. Microbiol. 17:970-974.

18. Berdal, B. P., 0. Olsvik, S. Myhre, and T. Omland. 1982.Demonstration of extracellular chymotrypsin-like activity fromvarious Legionella species. J. Clin. Microbiol. 16:452-457.

19. Berridge, M. J. 1984. Inositol trisphosphate and diacylglycerolas second messengers. Biochem. J. 220:345-360.

20. Bhardwaj, N., T. W. Nash, and M. A. Horwitz. 1986. Interfer-on-y-activated human monocytes inhibit the intracellular mul-tiplication of Legionella pneumophila. J. Immunol. 137:2662-2669.

21. Black, W. J., F. D. Quinn, and L. S. Tompkins. 1990. Legion-ella pneuinophila zinc metalloprotease is structurally and func-tionally homologous to Pseudomonas aeruginosa elastase. J.Bacteriol. 172:2608-2613.

22. Blackmon, J. A., F. W. Chandler, W. B. Cherry, A. C. EnglandIII, J. C. Feeley, M. D. Hicklin, R. M. McKinney, and H. W.Wilkinson. 1981. Legionellosis. Am. J. Pathol. 103:429-465.

23. Blanchard, D. K., J. Y. Djeu, T. W. Klein, H. Friedman, andW. E. Stewart 2d. 1988. Protective effects of tumor necrosisfactor in experimental Legionella pneumophila infections ofmice via activation of PMN function. J. Leukocyte Biol.43:429-435.

24. Blanchard, D. K., H. Friedman, T. W. Klein, and J. Y. Djeu.1989. Induction of interferon-gamma and tumor necrosis factorby Legionella pneumophila: augmentation of human neutro-phil bactericidal activity. J. Leukocyte Biol. 45:538-545.

25. Blanchard, D. K., W. E. Stewart 2d, T. W. Klein, H. Friedman,and J. Y. Djeu. 1987. Cytolytic activity of human peripheralblood leukocytes against Legionella pneumophila-infectedmonocytes: characterization of the effector cell and augmen-tation by interleukin 2. J. Immunol. 139:551-556.

26. Blander, S. J., R. F. Breiman, and M. A. Horwitz. 1989. A liveavirulent mutant Legionella vaccine induces protective immu-nity against lethal aerosol challenge. J. Clin. Invest. 83:810-815.

27. Blander, S. J., and M. A. Horwitz. 1989. Vaccination with themajor secretory protein of Legionella pneumophila inducescell-mediated and protective immunity in a guinea pig model ofLegionnaires' disease. J. Exp. Med. 169:691-705.

28. Blander, S. J., and M. A. Horwitz. 1991. Vaccination with themajor secretory protein of Legionella induces humoral andcell-mediated immune responses and protective immunityacross different serogroups of Legionella pneumophila anddifferent species of Legionella. J. Immunol. 147:285-291.

29. Blander, S. J., L. Szeto, H. A. Shuman, and M. A. Horwitz.1990. An immunoprotective molecule, the major secretoryprotein of Legionella pneumophila, is not a virulence factor ina guinea pig model of Legionnaires' disease. J. Clin. Invest.86:817-824.

30. Bokoch, G. M., and A. G. Gilman. 1984. Inhibition of receptormediated release of arachidonic acid by pertussis toxin. Cell39:301-308.

31. Breiman, R. F., B. S. Fields, G. N. Sanden, L. Volmer, A.Meier, and J. S. Spika. 1990. Association of shower use withlegionnaires' disease. Possible role of amoebae. JAMA 263:2924-2926.

32. Brenner, D. J., A. G. Steigerwalt, and J. E. McDade. 1979.Classification of the legionnaires' disease bacterium: Legion-ella pneumophila, genus novum, species nova, of the familyLegionellaceae, familia nova. Ann. Intern. Med. 90:656-658.

33. Brubaker, R. R. 1985. Mechanisms of bacterial virulence.Annu. Rev. Microbiol. 39:21-50.

34. Buster, B., J. Summersgill, and R. Miller. 1991. Intracellularfree calcium responses in human neutrophils infected withLegionella pneumophila, p. 37. Abstr. 91st Gen. Meet. Am.Soc. Microbiol. 1991. American Society for Microbiology,Washington, D.C.

35. Butler, C. A., E. D. Street, T. P. Hatch, and P. S. Hoffman.1985. Disulfide-bonded outer membrane proteins in the genusLegionella. Infect. Immun. 48:14-18.

36. Caillon, E., B. Lubochinsky, and D. Rigomier. 1983. Occur-rence of diakyl ether phospholipids in Stigmatella aurantiaca.J. Bacteriol. 153:1348-1351.

37. Caparon, M., and W. Johnson. 1988. Macrophage toxicity and

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

56 DOWLING ET AL.

complement sensitivity of virulent and avirulent strains ofLegionella pneumophila. J. Infect. Dis. 10:S377-S381.

38. Carpenter, C. L., and L. C. Cantley. 1990. Phosphoinositidekinases. Biochemistry 29:11148-11156.

39. Catrenich, C. E., and W. Johnson. 1989. Characterization ofthe selective inhibition of growth of virulent Legionella pnell-mophila by supplemented Mueller-Hinton medium. Infect.Immun. 57:1862-1864.

40. Chandler, F. W., J. A. Blackmon, M. D. Hicklin, R. M. Cole,and C. S. Callaway. 1979. Ultrastructure of the agent ofLegionnaires' disease in the human lung. Am. J. Clin. Pathol.71:43-50.

41. Chandler, F. W., J. E. McDade, M. D. Hicklin, J. A. Blackmon,B. M. Thomason, and E. P. Ewing, Jr. 1979. Pathologic findingsin guinea pigs inoculated with the legionnaires' disease bacte-rium. Ann. Intern. Med. 90:671-675.

42. Cianciotto, N. P., J. M. Bangsborg, B. I. Eisenstein, and N. C.Engleberg. 1990. Identification of mip-like genes in the genusLegionella. Infect. Immun. 58:2912-2918.

43. Cianciotto, N., B. I. Eisenstein, N. C. Engleberg, and H.Shuman. 1989. Genetics and molecular pathogenesis ofLegionella pneumnophila, an intracellular parasite of macro-phages. Mol. Biol. Med. 6:409-424.

44. Cianciotto, N. P., B. I. Eisenstein, C. H. Mody, and N. C.Engleberg. 1990. A mutation in the mip gene results in anattenuation of Legionella pneumophila virulence. J. Infect.Dis. 162:121-126.

45. Cianciotto, N. P., B. I. Eisenstein, C. H. Mody, G. B. Toews,and N. C. Engleberg. 1989. A Legionella pnewimophila geneencoding a species-specific surface protein potentiates initia-tion of intracellular infection. Infect. Immun. 57:1255-1262.

46. Ciesielski, C. A., M. J. Blaser, and W. L. Wang. 1986.Serogroup specificity of Legionella pneumophila is related tolipopolysaccharide characteristics. Infect. Immun. 51:397-404.

47. Cohen, H. J., M. E. Chovaniec, M. K. Wilson, and P. E.Newburger. 1982. Con-A-stimulated superoxide production bygranulocytes: reversible activation of NADPH oxidase. Blood60:1188-1194.

48. Conlan, J. W., and L. A. E. Ashworth. 1986. The relationshipbetween the serogroup antigen and lipopolysaccharide ofLegionella pneumophila. J. Hyg. Camb. 96:39-48.

49. Conlan, J. W., A. Baskerville, and L. A. E. Ashworth. 1986.Separation of Legionella pneumophila proteases and purifica-tion of a protease which produces lesions like those of Legion-naires' disease in guinea pig lung. J. Gen. Microbiol. 132:1565-1574.

50. Conlan, J. W., A. Williams, and L. A. E. Ashworth. 1988. Invivo production of a tissue-destructive protease by Legionellapneumophila. J. Gen. Microbiol. 134:143-149.

51. Conlan, J. W., A. Williams, and L. A. E. Ashworth. 1988.Inactivation of human a-1-antitrypsin by a tissue-destructiveprotease of Legionella pneumophila. J. Gen. Microbiol. 134:481-487.

52. Das, S., A. K. Saha, N. K. Mukhopadhyay, and R. H. Glew.1986. A cyclic nucleotide-independent protein kinase in Leish-mania donovani. Biochem. J. 240:641-649.

53. Das, S., A. K. Saha, A. T. Remaley, R. H. Glew, J. N. Dowling,M. Kajiyoshi, and M. Gottlieb. 1986. Hydrolysis of phos-phoproteins and inositol phosphates by cell surface phos-phatase of Leishmania donovani. Mol. Biochem. Parasitol.20:143-153.

54. Davis, G. S., W. C. Winn, Jr., D. W. Gump, and H. N. Beaty.1983. The kinetics of early inflammatory events during exper-imental pneumonia due to Legionella pneumophila in guineapigs. J. Infect. Dis. 148:823-835.

55. Davis, G. S., W. C. Winn, Jr., D. W. Gump, J. E. Craighead,and H. N. Beaty. 1982. Legionnaires' pneumonia after aerosolexposure in guinea pigs and rats. Am. Rev. Respir. Dis.126:1050-1057.

56. Densen, P., and G. L. Mandell. 1980. Phagocyte strategy vs.microbial tactics. Rev. Infect. Dis. 2:817-838.

57. Donowitz, G., J. Dowling, and D. Focht. 1989. Legionellamicdadei inhibits monocyte function. Program Abstr. 29th

lntersci. Conf. Antimicrob. Agents Chemother., abstr. 329.58. Donowitz, G. R., and D. Focht. Unpublished data.59. Donowitz, G. R., I. Reardon, J. Dowling, L. Rubin, and D.

Focht. 1990. Ingestion of Legionella micdadei inhibits humanneutrophil function. Infect. Immun. 58:3307-3311.

60. Donowitz, G. R., I. Reardon, and D. Focht. Unpublished data.61. Donowitz, G. R., I. Reardon, D. Focht, and J. N. Dowling.

Unpublished data.62. Dowling, J. N., J. D. Hempel, and R. H. Glew. Unpublished

data.63. Dowling, J. N., A. W. Pasculle, F. N. Frola, M. K. Zaphyr, and

R. B. Yee. 1984. Infections caused by Legionella micdadei andLegionella pneuimophila among renal transplant recipients. J.Infect. Dis. 149:703-713.

64. Dowling, J. N., A. K. Saha, and R. H. Glew. Unpublished data.65. Dowling, J. N., R. S. Weyant, and A. W. Pasculle. 1982.

Bactericidal activity of antibiotics against Legionella micdadei(Pittsburgh pneumonia agent). Antimicrob. Agents Chemo-ther. 22:272-276.

66. Dreyfus, L. A. 1987. Virulence associated ingestion of Legion-ella pneumnophila by HeLa cells. Microb. Pathog. 3:45-52.

67. Dreyfus, L. A., and B. H. Iglewski. 1986. Purification andcharacterization of an extracellular protease of Legionellapneuimophila. Infect. lmmun. 51:736-743.

68. Ehret, W., and G. Ruckdeschel. 1985. Membrane protein ofLegionellaceae. I. Membrane proteins of different strains andserogroups of Legionella pneumophila. Zentralbl. Bakteriol.Mikrobiol. Hyg. Ser. A 259:433-435.

69. Elliott, J. A., and W. C. Winn, Jr. 1986. Treatment of alveolarmacrophages with cytochalasin D inhibits uptake and subse-quent growth of Legionella pneumophila. Infect. Immun.51:31-36.

70. Enami, M., and A. Ishihama. 1984. Protein phosphorylation inEscherchia coli and purification of a protein kinase. J. Biol.Chem. 259:526-533.

71. Engleberg, N. C., C. Carter, D. R. Weber, N. P. Cianciotto, andB. I. Eisenstein. 1989. DNA sequence of mip, a Legionellapnetinophila gene associated with macrophage infectivity.Infect. Immun. 57:1263-1270.

72. Engleberg, N. C., E. Pearlman, D. Dixon, and B. I. Eisenstein.1986. Antibodies isolated by using cloned surface antigensrecognize antigenically related components of Legionellapneumophila and other Legionella species. J. lmmunol. 136:1415-1417.

73. Fields, B. S., J. M. Barbaree, G. N. Sanden, and W. E. Morrill.1990. Virulence of a Legionella anisa strain associated withPontiac fever: an evaluation using protozoan, cell culture, andguinea pig models. Infect. Immun. 58:3139-3142.

74. Fields, B. S., E. B. Shotts, Jr., J. C. Feeley, G. W. Gorman, andW. T. Martin. 1984. Proliferation of Legionella pneumophilaas an intracellular parasite of the ciliated protozoan Tetrahy-mena pvriformis. Appl. Environ. Microbiol. 47:467-471.

75. Finnerty, W. R., R. A. Makula, and J. C. Feeley. 1979. Cellularlipids of the Legionnaires' disease bacterium. Ann. Intern.Med. 90:631-634.

76. Fitzgeorge, R. B., A. S. R. Featherstone, and A. Baskerville.1988. Effects of polymorphonuclear leucocyte depletion on thepathogenesis of experimental Legionnaires' disease. Br. J.Exp. Pathol. 69:105-112.

77. Fraser, D. W., T. R. Tsai, W. Orenstein, W. E. Pardin, H. J.Beecham, R. G. Sharrar, J. Harris, G. F. Mallison, S. M.Martin, J. E. McDade, C. C. Shepard, P. S. Brachman, and theField Investigation Team. 1977. Legionnaires' disease: descrip-tion of an epidemic of pneumonia. N. Engl. J. Med. 297:1189-1197.

78. Friedman, R. L., B. H. Iglewski, and R. D. Miller. 1980.ldentification of a cytotoxin produced by Legionella pneumo-phila. Infect. Immun. 29:271-274.

79. Friedman, R. L., J. E. Lochner, R. H. Bigley, and B. H.Iglewski. 1982. The effects of Legionella pneumophila toxin onoxidative processes and bacterial killing of human polymor-phonuclear leukocytes. J. Infect. Dis. 146:328-334.

80. Frisch, C. F., and W. B. Baine. 1984. Production of extracel-

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 57

lular hemolytic cytotoxin by Legionella bozemanii. Microbiol.Immunol. 28:1267-1272.

81. Gabay, J. E., M. Blake, W. D. Niles, and M. A. Horwitz. 1985.Purification of Legionella pneumophila major outer membraneprotein and demonstration that it is a porin. J. Bacteriol.162:85-91.

82. Gabay, J. E., and M. A. Horwitz. 1985. Isolation and charac-terization of the cytoplasmic and outer membranes of theLegionnaires' disease bacterium (Legionella pneumophila). J.Exp. Med. 161:409-422.

83. Gilman, A. G. 1987. G proteins: transducers of receptor-generated signals. Annu. Rev. Biochem. 56:615-649.

84. Glavin, F. L., W. C. Winn, Jr., and J. E. Craighead. 1979.Ultrastructure of lung in Legionnaires' disease. Observationsof three biopsies done during the Vermont epidemic. Ann.Intern. Med. 90:555-559.

85. Gress, F. M., R. L. Myerowitz, A. W. Pasculle, C. R. Rinaldo,Jr., and J. N. Dowling. 1980. The ultrastructural morphologicfeatures of Pittsburgh pneumonia agent. Am. J. Pathol. 101:63-78.

86. Gul'nik, S. V., M. P. Yusupova, G. I. Lavrenova, I. S.Tartakovsky, S. V. Prozorovsky, and V. M. Stepanov. 1986.Proteinases of Legionella: phenylalanineaminopeptidase of L.pneumophila. J. Gen. Microbiol. 132:387-392.

87. Hase, C. C., and R. A. Finkelstein. 1990. Comparison of theVibrio cholerae hemagglutinin/protease and the Pseudomonasaeruginosa elastase. Infect. Immun. 58:4011-4015.

88. Hebert, G. A., A. G. Steigerwalt, and D. J. Brenner. 1980.Legionella micdadei species nova: classification of a thirdspecies of Legionella associated with human pneumonia. Curr.Microbiol. 3:252-257.

89. Hedlund, K. W. 1981. Legionella toxin. Pharmacol. Ther.15:123-130.

90. Hedlund, K. W., and R. Larson. 1981. Legionella pneutnophilatoxin, isolation and purification, p. 81-89. In F. M. Everaerts(ed.), Analytical isotachophoresis. Elsevier Scientific Publish-ing Co., Amsterdam.

91. Hicklin, M. D., B. M. Thomason, F. W. Chandler, and J. A.Blackmon. 1980. Pathogenesis of acute Legionnaires' diseasepneumonia. Immunofluorescent microscopic study. Am. J.Clin. Pathol. 73:480-487.

92. Hindahl, M. S., and B. H. Iglewski. 1984. Isolation andcharacterization of the Legionella pneumophila outer mem-brane. J. Bacteriol. 159:107-113.

93. Hindahl, M. S., and B. H. Iglewski. 1986. Outer membraneproteins from Legionella pneumophila serogroups and otherLegionella species. Infect. Immun. 51:94-101.

94. Hoffman, P. S., C. A. Butler, and F. D. Quinn. 1989. Cloningand temperature-dependent expression in Eseherichia coli of aLegionella pneumophila gene coding for a genus-common60-kilodalton antigen. Infect. Immun. 57:1731-1739.

95. Hoffman, P. S., L. Houston, and C. A. Butler. 1990. Legionellapneumophila htpAB heat shock operon: nucleotide sequenceand expression of the 60-kilodalton antigen in L. pneumophila-infected HeLa cells. Infect. Immun. 58:3380-3387.

96. Hoffman, P. S., L. Pine, and S. Bell. 1983. Production ofsuperoxide and hydrogen peroxide in medium used to cultureLegionella pneumophila: catalytic decomposition by charcoal.Appl. Environ. Microbiol. 45:784-791.

97. Homma, J. Y., C. Abe, K. Tanamoto, Y. Hirao, K. Morihara,H. Tsuzuki, R. Yanagawa, E. Honda, Y. Aoi, Y. Fujimoto, M.Goryo, N. Imazeki, H. Noda, A. Goda, S. Takeuchi, and T.Ishihara. 1978. Effectiveness of immunization with single andmulticomponent vaccines prepared from a common antigen(OEP), protease and elastase toxoids of Pseudoinonas aerug-inosa on protection against hemorrhagic pneumonia in minkdue to P. aeruginosa. Jpn. J. Exp. Med. 48:111-133.

98. Horwitz, M. A. 1983. Formation of a novel phagosome by theLegionnaires' disease bacterium (Legionella pneumophila) inhuman monocytes. J. Exp. Med. 158:1319-1331.

99. Horwitz, M. A. 1983. The Legionnaires' disease bacterium(Legionella pneumophila) inhibits phagosome-lysosome fusionin human monocytes. J. Exp. Med. 158:2108-2126.

100. Horwitz, M. A. 1983. Cell-mediated immunity in Legionnaires'disease. J. Clin. Invest. 71:1686-1697.

101. Horwitz, M. A. 1984. Phagocytosis of the Legionnaires' dis-ease bacterium (Legionella pneumophila) occurs by a novelmechanism: engulfment within a psuedopod coil. Cell 36:27-33.

102. Horwitz, M. A. 1987. Characterization of avirulent mutantLegionella pneumophila that survive but do not multiplywithin human monocytes. J. Exp. Med. 166:1310-1328.

103. Horwitz, M. A. 1988. Phagocytosis and intracellular biology ofLegionella pneumophila, p. 283-302. In M. A. Horwitz (ed.),Bacteria-host cell interaction. Alan R. Liss, Inc., New York.

104. Horwitz, M. A., and F. R. Maxfield. 1984. Legionella pneumo-phila inhibits acidification of its phagosome in human mono-cytes. J. Cell Biol. 99:1936-1943.

105. Horwitz, M. A., and S. C. Silverstein. 1980. Legionnaires'disease bacterium (Legionella pneumophila) multiplies intra-cellularly in human monocytes. J. Clin. Invest. 66:441-450.

106. Horwitz, M. A., and S. C. Silverstein. 1981. Interaction of thelegionnaires' disease bacterium (Legionella pneumophila) withhuman phagocytes. I. L. pneumophila resists killing by poly-morphonuclear leukocytes, antibody, and complement. J.Exp. Med. 153:386-397.

107. Horwitz, M. A., and S. C. Silverstein. 1981. Interaction of thelegionnaires' disease bacterium (Legionella pneumophila) withhuman phagocytes. 11. Antibody promotes binding of L. pneu-mnophila to monocytes but does not inhibit intracellular multi-plication. J. Exp. Med. 153:398-406.

108. Horwitz, M. A., and S. C. Silverstein. 1981. Activated humanmonocytes inhibit the intracellular multiplication of Legion-naires' disease bacteria. J. Exp. Med. 154:1618-1635.

109. Jacobs, R. F., R. M. Locksley, C. B. Wilson, J. E. Haas, andS. J. Klebanoff. 1984. Interaction of primate alveolar macro-phages and Legionella pneumophila. J. Clin. Invest. 73:1515-1523.

110. Jaken, S. 1989. Diacylglycerol: the role of stimulated produc-tion in activation of protein kinase C, p. 163-178. In R. H.Michell, A. H. Drummond, and C. P. Downes (ed.), Inositollipids in cell signaling. Academic Press, Inc., San Diego.

111. Jepras, R. I., and R. B. Fitzgeorge. 1986. The effect ofoxygen-dependent antimicrobial systems on strains of Legion-ella pneumophila of different virulence. J. Hyg. Camb. 97:61-69.

112. Jepras, R. I., R. B. Fitzgeorge, and A. Baskerville. 1985. Acomparison of virulence of two strains of Legionella pneumo-phila based on experimental aerosol infection of guinea-pigs. J.Hyg. Camb. 95:29-38.

113. Katz, S. M., and S. Hashemi. 1982. Electron microscopicexamination of the inflammatory response to Legionella pneu-mophila in guinea pigs. Lab. Invest. 46:24-32.

114. Keen, M. G., and P. S. Hoffman. 1989. Characterization of aLegionella pneumophila extracellular protease exhibiting he-molytic and cytotoxic activities. Infect. Immun. 57:732-738.

115. Kishimoto, R. A., M. D. Kastello, J. D. White, F. G. Shirey,V. G. McGann, E. W. Larson, and K. W. Hedlund. 1979. Ihvitro interaction between normal cynolmolgus monkey alveo-lar macrophages and legionnaires disease bacteria. Infect.Immun. 25:761-763.

116. Kishimoto, R. A., J. D. White, F. G. Shirey, V. G. McGann,R. F. Berendt, E. W. Larson, and K. W. Hedlund. 1981. lnvitro response of guinea pig peritoneal rnacrophages to Legion-ella pneurnophila. Infect. Immun. 31:1209-1213.

117. Korchak, H. M., L. B. Vosshall, G. Zagon, P. Ljubich, A. M.Rich, and G. Weissmann. 1988. Activation of the neutrophil bycalcium-mobilizing ligands. I. A chemotactic peptide and thelectin concanavalin A stimulate superoxide anion generationbut elicit different calcium movements and phosphoinositideremodeling. J. Biol. Chem. 263:11090-11097.

118. Kuchta, J. M., S. J. States, A. M. McNamara, R. M. Wad-owsky, and R. B. Yee. 1983. Susceptibility of Legionellapneiinophila to chlorine in tap water. Appl. Environ. Micro-biol. 46:1134-1139.

119. Lad, P. M., C. V. Olson, and P. A. Smiley. 1985. Association of

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

58 DOWLING ETAL.McoBo.RV

the N-formyl-Met-Leu-Phe receptor in human neutrophils with

a GTP-binding protein sensitive to pertussis toxin. Proc. Nati.

Acad. Sci. USA 82:.869-873.

120. Levi, M. H., A. W. Pasculle, and J. N. Dowling. 1987. Role of

the alveolar macrophage in host defense and immunity to

Legionella micdadei pneumonia in the guinea pig. Microb.

Pathog. 2:269-282.

121. Lochner, J. E., R. H. Bigley, and B. H. Iglewski. 1985.

Defective triggering of polymorphonuclear leukocyte oxidative

metabolism by Legionella pneuimophila toxin. J. Infect. Dis.

151:42-46.

122. Lochner, J. E., R. L. Friedman, R. H. Bigley, and B. H.

Iglewski. 1983. Effect of oxygen-dependent antimicrobial sys-

tems on Legionella pneumophila. Infect. Immun. 39:487-489.

123. Locksley, R. M., R. F. Jacobs, C. B. Wilson, W. M. Weaver,

and S. J. Klebanoff. 1982. Susceptibility of Legionella pneu-

mophila to oxygen-dependent microbicidal systems. J. Immu-

nol. 129:2192-2197.

123a.Lundemose, A. G., S. Birkelund, S. J. Fey, P. M. Larsen, and

G. Christiansen. 1991. Chlamvdia trachomatis contains a pro-

tein similar to the Legionella pneumophila mip gene product.

Mol. Microbiol. 5:109-115.

124. Lyerly, D. M., and A. S. Kreger. 1983. Importance of serratia

protease in the pathogenesis of experimental Serriatia marces-

cens pneumonia. Infect. Immun. 40:113-119.

125. Majumder, G. C., E. Shrago, and C. E. Elson. 1975. Changes in

cyclic AMP-dependent protein kinase activity in Tetrrahvmenapyriformis. Biochim. Biophys. Acta 384:399-412.

126. Manai, M., and J. Cozzone. 1982. Endogenous protein phos-

phorylation in Escherichia coli extracts. Biochem. Biophys.

Res. Commun. 107:.981-988.

127. Mandell, G. L. 1975. Catalase, superoxide dismutase, and

virulence of Staphvlococcius aurieius. J. Clin. Invest. 55:561-

566.

128. Maridonneau-Parini, I., S. M. Tringale, and A. Tauber. 1986.

Identification of distinct activation pathways of the human

neutrophil NADPH-oxidase. J. Immunol. 137:2925-2929.

129. Marra, A., M. A. Horwitz, and H. A. Shuman. 1990. The

HL-60 model for the interaction of human macrophages with

the legionnaires' disease bacterium. J. Immunol. 144:2738-

2744.

130. Mayberry, W. R. 1981. Dihydroxy and monohydroxy fatty

acids in Legionella pneutinophila. J. Bacteriol. 147:373-381.

131. McCusker, K. T., B. A. Braaten, M. W. Cho, and D. A. Low.

1991. Legionella pneumophila inhibits protein syntheis in

Chinese hamster ovary cells. Infect. Immun. 59:240-246.

132. McDade, J. E., C. C. Shepard, D. W. Fraser, T. R. Tsai, M. A.

Redus, W. R. Dowdle, and the Laboratory Investigationi Team.

1977. Legionnaires' disease: isolation of a bacterium and

demonstration of its role in other respiratory disease. N. Engl.J. Med. 297:1197-1203.

133. Miller, R. D., and J. M. Hammel. 1985. Biochemistry and

physiology of Legionella, p. 83-107. In S. M. Katz (ed.),

Legionellosis, vol. 1. CRC Press, Boca Raton, Fla.

134. Miller, R. M., J. Garbus, and R. B. Hornick. 1972. Lack of

enhanced oxygen consumption by polymorphonuclear leuko-

cytes on phagocytosis of virulent Salmonella tvplhi. Science

175:1010-1011.

135. Mintz, C. S., J. X. Chen, and H. A. Shuman. 1988. Isolation

and characterization of auxotrophic mutants of Legionella

pneumophila that fail to multiply in human monocytes. Infect.

Immun. 56:1449-1455.

136. Moss, C. W., R. E. Weaver, S. B. Dees, and W. B. Cherry.1977. Cellular fatty acid composition of isolates from Legion-naires disease. J. Clin. Microbiol. 6:140-143.

137. Muller, H. E. 1980. Proteolytic action of Legionella pneii/no-

phila on human serum proteins. Infect. Immun. 27:51-53.

138. Muller, H. E. 1981. Enzymatic profile of Legionella pneu/no-

phila. J. Clin. Microbiol. 13:423-426.

139. Myerowitz, R. L. 1982. An update on legionnaires' disease and

pneumonias caused by "new" Legionella-like bacteria.

Monogr. Pathol. 23:123-137.

140. Myerowitz, R. L., A. W. Pasculle, J. N. Dowling, G. J. Pazin,

M. Puerzer, R. B. Yee, C. R. Rinaldo, Jr., and T. R. Hakala.1979. Opportunistic lung infection due to "Pittsburgh pneumo-nia agent." N. Engl. J. Med. 301:953-958.

141. Nagington, J., and D. J. Smith. 1980. Pontiac fever andamoebae. Lancet ii:1241.

142. Nash, T. W., D. M. Libby, and M. A. Horwitz. 1984. Interac-tion between the legionnaires' disease bacterium (Legionellapneiumophila) and human alveolar macrophages. Influence ofantibody, lymphokines, and hydrocortisone. J. Clin. Invest.74:771-782.

143. Newsome, A. L., R. L. Baker, R. D. Miller, and R. R. Arnold.1985. Interactions between Naegleria fowleri and Legionellapneuttnophila. Infect. Immun. 50:449-452.

144. Nimmo, G. A., A. C. Borthwick, W. H. Holms, and H. G.Nimmo. 1984. Partial purification and properties of isocitratedehydrogenase kinase/phosphatase from Escherichia coliML308. Eur. J. Biochem. 141:401-408.

145. Nolte, F. S., C. C. Conlin, and M. A. Motley. 1986. Electro-phoretic and serological characterization of lipopolysaccha-rides of Legionella pneumophila. Infect. Immun. 52:676-681.

146. Nolte, F. S., G. E. Hollick, and R. G. Robertson. 1982.Enzymatic activities of Legionella pneuimophila and Legionel-la-like organisms. J. Clin. Microbiol. 15:175-177.

147. Norqvist, A., B. Norrman, and H. Wolf-Watz. 1990. Identifica-tion and characterization of a zinc metalloprotease associatedwith invasion by the fish pathogen Vibric anguillarum. Infect.Immun. 58:3731-3736.

148. Oldham, L. J., and F. G. Rodgers. 1985. Adhesion, penetrationand intracellular replication of Legionella pneumophila: an invitro model of pathogenesis. J. Gen. Microbiol. 131:697-706.

149. Otten, S., S. lyer, W. Johnson, and R. Montgomery. 1986.Serospecific antigens of Legionella pneumophila. J. Bacteriol.167:893-904.

150. Pasculle, A. W., J. N. Dowling, F. N. Frola, D. A. McDevitt,and M. A. Levi. 1985. Antimicrobial therapy of experimentalLegionella ,nic-dadei pneumonia in guinea pigs. Antimicrob.Agents Chemother. 28:730-734.

151. Pasculle, A. W., J. N. Dowling, R. S. Weyant, J. M. Sniffen,L. G. Cordes, G. M. Gorman, and J. C. Feeley. 1981. Suscep-tibility of Pittsburgh pneumonia agent (Legionella micdadei)and other newly recognized members of the genus Legionellato nineteen antimicrobial agents. Antimicrob. Agents Chemo-ther. 20:793-799.

152. Payne, N. R., and M. A. Horwitz. 1987. Phagocytosis ofLegionella pneuimophila is mediated by human monocytecomplemnent receptors. J. Exp. Med. 166:1377-1389.

153. Pearlman, E., N. C. Engleberg, and B. I. Eisenstein. 1985.Identification of protein antigens of Legionella pneumophilaserogroup 1. Infect. Immun. 47:74-79.

154. Pearlman, E., A. H. Jiwa, N. C. Engleberg, and B. 1. Eisenstein.1988. Growth of Legionella pneumophila in a human macro-phage-like (U937) cell line. Microb. Pathog. 5:87-95.

155. Pine, L., P. S. Hoffman, G. B. Malcolm, R. F. Benson, andM. G. Keen. 1984. Determination of catalase, peroxidase, andsuperoxide dismutase within the genus Legionella. J. Clin.Microbiol. 20:421-429.

156. Plouffe, J. F., M. F. Para, and K. A. Fuller. 1985. Serumbactericidal activity against Legionella pneumophila. J. Clin.Microbiol. 22:863-864.

157. Quinn, F. D., M. G. Keen, and L. S. Tompkins. 1989. Genetic,immunological, and cytotoxic comparisons of Legionella pro-teolytic activities. Infect. Immun. 57:2719-2725.

158. Rdest, U., E. Wintermeyer, B. Ludwig, and J. Hacker. 1991.Legiolysin, a new hemolysin from L. pneuimophila. Zentralbl.Bakteriol. Parasitenkd. Infektionskr. Hyg. Abt. 1 Orig. ReiheA 274:471-474.

159. Rechnitzer, C., and J. Blom. 1989. Engulfment of the Philadel-phia strain of Legionella pneumophila within pseudopod coilsin human phagocytes. Comparison with other Legionellastrains and species. APMIS 97:105-114.

160. Rechnitzer, C., M. Diamant, and B. K. Pederson. 1989. Inhibi-tion of human natural killer cell activity by Legionella pneu-

MICROBIOL. REV.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

VIRULENCE FACTORS OF THE LEGIONELLACEAE 59

mophila protease. Eur. J. Clin. Microbiol. Infect. Dis. 8:989-992.

161. Rechnitzer, C., A. Kharazmi, G. Doring, and M. Tvede. 1987.Legionella pneumophila exoprotease inhibits human neutro-phil and monocyte chemotaxis, abstr. 380. Eur. J. Clin. Invest.17:A66.

162. Rechnitzer, C., M. Tveda, and G. Doring. 1989. A rapid methodfor purification of homogeneous Legionella pneumophila cyto-toxic protease using fast protein liquid chromatography.FEMS Microbiol. Lett. 59:39-44.

163. Reeves, H. C., and P. J. Malloy. 1982. The phosphorylation ofisocitrate dehydrogenase in Escherichia coli. Biochem. Soc.Trans. 10:321-322.

164. Reingold, A. L., B. M. Thomason, B. J. Brake, L. Thacker,H. W. Wilkinson, and J. N. Kuritsky. 1984. Legionella pneu-monia in the United States: the distribution of serogroups andspecies causing human illness. J. Infect. Dis. 149:819.

165. Remaley, A. T., S. Das, P. I. Campbell, G. M. LaRocca, M. T.Pope, and R. H. Glew. 1985. Characterization of Leishmaniadonovanii acid phosphatase. J. Biol. Chem. 260:880-886.

166. Remaley, A. T., D. B. Kuhns, R. E. Basford, R. H. Glew, andS. S. Kaplan. 1984. Leishmanial phosphatase blocks neutrophil02- production. J. Biol. Chem. 259:11173-11175.

167. Rodgers, F. G., and A. S. High. 1991. Expression of Legionellapneumophila virulence-associated major outer membrane pro-tein in Escherichia coli, abstr. B-71, p. 37. Abstr. 91st Gen.Meet. Am. Soc. Microbiol. 1991. American Society for Micro-biology, Washington, D.C.

168. Rogers, B. H., G. R. Donowitz, G. K. Walker, S. A. Harding,and M. A. Sande. 1979. Opportunistic pneumonia. A clinico-pathological study of five cases caused by an unidentifiedacid-fast bacterium. N. Engl. J. Med. 301:959-961.

169. Rosenfeld, J. S., F. Kueppers, T. Newkirk, R. Tamada, J. J.Meissler, Jr., and T. K. Eisenstein. 1986. A protease fromLegionella pneumophila with cytotoxic and dermal ulcerativeactivity. FEMS Microbiol. Lett. 37:51-58.

170. Rowbotham, T. J. 1980. Preliminary report of the pathogenicityof Legionella pneumophila for fresh water and soil amoebae. J.Clin. Pathol. 33:1179-1183.

171. Rowbotham, T. J. 1986. Current views on the relationshipsbetween amoebae, legionellae and man. Isr. J. Med. Sci.22:678-689.

172. Saha, A. K., J. N. Dowling, K. L. LaMarco, S. Das, A. T.Remaley, N. Olomu, M. T. Pope, and R. H. Glew. 1985.Properties of an acid phosphatase from Legionella micdadeiwhich blocks superoxide anion production by human neutro-phils. Arch. Biochem. Biophys. 243:150-160.

173. Saha, A. K., J. N. Dowling, N. K. Mukhopadhyay, and R. H.Glew. 1988. Demonstration of two protein kinases in extractsof Legionella micdadei. J. Gen. Microbiol. 134:1275-1281.

174. Saha, A. K., J. N. Dowling, N. K. Mukhopadhyay, and R. H.Glew. 1989. Legionella micdadei protein kinase catalyzesphosphorylation of tubulin and phosphatidylinositol. J. Bacte-riol. 171:5103-5110.

175. Saha, A. K., J. N. Dowling, A. W. Pasculle, and R. H. Glew.1988. Legionella micdadei phosphatase catalyzes the hydroly-sis of phosphatidylinositol 4,5-bisphosphate in human neutro-phils. Arch. Biochem. Biophys. 265:94-104.

176. Sahney, N., J. Summersgill, M. Raff, and R. Miller. 1989.Inhibition of polymorphonuclear leukocyte (PMN) function byLegionella pneumophila exoproducts, abstr. B-272, p. 76.Abstr. 89th Annu. Meet. Am. Soc. Micribiol. 1989. AmericanSociety for Microbiology, Washington, D.C.

177. Sahney, N. N., B. C. Lambe, J. T. Summersgill, and R. D.Miller. 1990. Inhibition of polymorphonuclear leukocyte func-tion by Legionella pneumophila exoproducts. Microb. Pathog.9:117-125.

178. Sampson, J. S., S. P. O'Connor, B. P. Holloway, B. B. Plikaytis,G. M. Carlone, and L. W. Mayer. 1990. Nucleotide sequenceof htpB, the Legionella pneumophila gene encoding the 58-kilodalton (kDa) common antigen, formerly designated the60-kDa common antigen. Infect. Immun. 58:3154-3157.

179. Sawyer, D. W., G. R. Donowitz, and G. L. Mandell. 1989.

Polymorphonuclear neutrophils: an effective antimicrobialforce. Rev. Infect. Dis. 11:S1532-S1544.

179a.Standaert, R. F., A. Galat, G. L. Verdine, and S. L. Schreiber.1990. Molecular cloning and overexpression of the humanFK506-binding protein FKBP. Nature (London) 346:671-674.

180. Steffensen, D. O., D. L. Weinbaum, and J. N. Dowling. 1985.Opsonic requirements for phagocytosis of Legionella micdadeiby polymorphonuclear leukocytes. Infect. Immun. 49:695-699.

181. Styrt, B., R. D. Walker, and J. C. White. 1989. Neutrophiloxidative metabolism after exposure to bacterial phospholi-pase C. J. Lab. Clin. Med. 114:51-57.

182. Summersgill, J. T., M. J. Raff, and R. D. Miller. 1988.Interactions of virulent and avirulent Legionella pneumophilawith human polymorphonuclear leukocytes. Microb. Pathog.5:41-47.

183. Summersgill, J. T., M. J. Raff, and R. D. Miller. 1990.Interactions of virulent and avirulent Legionella pneumophilawith human monocytes. J. Leukocyte Biol. 47:31-38.

184. Szeto, L., and H. A. Shuman. 1990. The Legionella pneumo-phila major secretory protein, a protease, is not required forintracellular growth or cell killing. Infect. Immun. 58:2585-2592.

185. Tartakovskii, I. S., I. F. Belyi, I. V. Vertiev, I. G. Nagaev, andS. V. Prozorovskii. 1988. Cytolysin as an immunoserologicalmarker of Legionella pneumophila. Zh. Mikrobiol. Epidemiol.Immunobiol. May(5):58-61.

186. Thompson, M. R., R. D. Miller, and B. H. Iglewski. 1981. Invitro production of an extracellular protease by Legionellapneumophila. Infect. Immun. 34:299-302.

187. Thorpe, T. C., and R. D. Miller. 1981. Extracellular enzymesof Legionella pneumophila. Infect. Immun. 33:632-635.

188. Tropschug, M., E. Wachter, S. Mayer, E. R. Schonbrunner,and F. X. Schmid. 1990. Isolation and sequence of an FK506-binding protein from N. crassa which catalyses proteinfolding. Nature (London) 346:674-677.

189. Tyndall, R. L., and E. L. Domingue. 1982. Cocultivation ofLegionella pneumophila and free-living amoebae. Appl. Envi-ron. Microbiol. 44:954-959.

190. Verbrugh, H. A., D. A. Lee, G. R. Elliot, W. F. Keane, J. R.Hoidal, and P. K. Peterson. 1985. Opsonization of Legionellapneumophila in human serum: key roles for specific antibodiesand the classical complement pathway. Immunology 54:643-653.

191. Verghese, M. W., C. D. Smith, and R. Snyderman. 1985.Potential role for a guanine nucleotide regulatory protein inchemoattractant receptor mediated polyphosphoinositide me-tabolism, Ca"+ mobilization and cellular responses by leuko-cytes. Biochem. Biophys. Res. Commun. 127:450-457.

192. Vilde, J. L., P. Rajagopalan, and E. Dournon. 1983. VirulentLegionella pneumophila does not provoke metabolic burst ofneutrophils, unlike a nonvirulent strain. Program Abstr. 22ndIntersci. Conf. Antimicrob. Agents Chemother., abstr. 65.

193. Wadee, A. A., D. Mendelsohn, and A. R. Rabson. 1983.Characterization of a suppressor cell-activating factor (SCAF)released by adherent cells treated with M. tuberculosis. J.Immunol. 130:2266-2270.

194. Wadowski, R. M., T. M. Wilson, N. J. Kapp, A. J. West, J. M.Kuchta, S. J. States, J. N. Dowling, and R. B. Yee. 1991.Multiplication of Legionella spp. in tap water containing Hart-mannella vermiformis. Appl. Environ. Microbiol. 57:1950-1955.

195. Wandosell, F., L. Serrano, and J. Avila. 1987. Phosphorylationof a-tubulin carboxyl-terminal tyrosine prevents its incorpora-tion into microtubules. J. Biol. Chem. 262:8268-8273.

196. Wandosell, F., L. Serrano, M. A. Hernandez, and J. Aliva.1986. Phosphorylation of tubulin by a calmodulin-dependentprotein kinase. J. Biol. Chem. 261:10332-10339.

197. Wang, J. Y. J., and D. E. Koshland, Jr. 1981. The identificationof distinct protein kinases and phosphatases in the prokaryoteSalmonella typhimurium. J. Biol. Chem. 256:4640-4648.

198. Weinbaum, D. L., J. Bailey, R. R. Benner, A. W. Pasculle, andJ. N. Dowling. 1983. The contribution of human neutrophilsand serum to host defense against Legionella micdadei. J.

VOL. 56, 1992

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from

60 DOWLING ET AL. MICROBIOL. REV.

Infect. Dis. 148:510-517.199. Weinbaum, D. L., R. R. Benner, J. N. Dowling, A. Alpern,

A. W. Pasculle, and G. R. Donowitz. 1984. Interaction ofLegionella micdadei with human monocytes. Infect. Immun.46:68-73.

200. Williams, A., A. Baskerville, A. B. Dowsett, and J. W. Conlan.1987. Immunocytochemical demonstration of the associationbetween Legionella pneumophila, its tissue-destructive prote-ase, and pulmonary lesions in experimental Legionnaires'disease. J. Pathol. 153:257-264.

201. Winn, W. C., G. S. Davis, D. W. Gump, J. E. Craighead, andH. N. Beaty. 1982. Legionnaires' pneumonia after intratrachealinoculation of guinea pigs and rats. Lab. Invest. 47:568-578.

202. Winn, W. C., Jr., F. L. Glavin, D. P. Perl, J. L. Keller, T. L.Andres, T. M. Brown, C. M. Coffin, J. E. Sensecqua, L. N.Roman, and J. E. Craighead. 1978. The pathology of Legion-

naires' disease. Fourteen fatal cases from the 1977 outbreak inVermont. Arch. Pathol. Lab. Med. 102:344-350.

203. Winn, W. C., Jr., and R. L. Myerowitz. 1981. The pathology ofthe Legionella pneumonias. Hum. Pathol. 12:401-422.

204. Wong, K. H., C. W. Moss, D. H. Hochstein, R. J. Arko, andW. 0. Schalla. 1979. "Endotoxicity" of the Legionnaires'disease bacterium. Ann. Intern. Med. 90:624-627.

205. Yamamoto, Y., T. W. Klein, C. A. Newton, R. Widen, and H.Friedman. 1988. Growth of Legionella pneumophila in thiogly-colate-elicited peritioneal macrophages from A/J mice. Infect.Immun. 56:370-375.

206. Yoshida, S., and Y. Mizuguchi. 1986. Multiplication of Legion-ella pneumophila Philadelphia-1 in cultured peritoneal macro-phages and its correlation to susceptibility of animals. Can. J.Microbiol. 32:438-442.

on August 23, 2020 by guest

http://mm

br.asm.org/

Dow

nloaded from