INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

At the 1976 American Legion convention in Philadelphia, an outbreak of pneumonia, termed Legionnaire's disease, led to the identification of the genus Legionella (31, 52). Although more than 40 species of Legionella are now known (7), surveillance studies have attributed 80 to 90% of Legionnaire's disease cases to Legionella pneumophila (14, 50), and it has been studied in the greatest detail. L. pneumophila is an opportunistic pathogen: disease is typically restricted to people with underlying health conditions, including the elderly, cigarette smokers, individuals receiving immunosuppressive therapy, or organ transplant recipients (14, 50). In contrast, healthy individuals associated with outbreaks, recognized by their low titers of Legionella-specific antibodies, are often asymptomatic (34).

In nature, legionellae are found in fresh water and soil as parasites of amoebae (11, 28, 65). L. pneumophila may adapt to its distinct intracellular and aquatic environments by alternating between a "replicative" and a "virulent" form in response to growth conditions (13). When amino acids and other conditions are favorable, L. pneumophila replicates within a host cell vacuole. When amino acids are limiting, the bacteria become cytotoxic, resistant to osmotic stress, motile, sensitive to sodium, and competent to evade phagosome-lysosome fusion (13), traits which likely enable the progeny to survive and to disperse in the environment, then reestablish a protected intracellular replication niche. Many of the traits required by L. pneumophila to parasitize amoebae also contribute to its survival and growth in macrophages (18, 33, 72).

Although the molecular mechanisms of L. pneumophila pathogenesis remain largely undefined, a number of traits which are correlated with virulence have been described. When aerosols containing L. pneumophila are inhaled, alveolar macrophages ingest the bacterium by coiling or conventional phagocytosis (39, 62). Phagosomal bacteria can evade phagosome-lysosome fusion, associate sequentially with mitochondria and the rough endoplasmic reticulum (ER), then replicate to high numbers (37, 38, 77). To establish this replication niche, L. pneumophila may secrete virulence factors by a specialized transport apparatus. A number of dot and icm genes are required by L. pneumophila for both intracellular growth in macrophages and conjugal DNA transfer (70, 80), and some of the predicted Dot and Icm proteins resemble components of the Agrobacterium tumafaciens vir complex (17), the Bordetella pertussis ptl system (84), and the Helicobacter pylori cag complex (16), three secretion systems important for virulence. In addition, it has been shown that a type II secretion system contributes to the virulence of L. pneumophila (49).

The second most common etiologic agent of Legionnaire's disease is Legionella micdadei (27, 50, 63), which infects immunocompromised hosts primarily (27, 47, 55, 56, 64, 68). Consistent with its epidemiology, L. micdadei is less virulent than L. pneumophila in guinea pig and tissue culture models of infection (29, 83). The genetic basis for the differential virulence of L. micdadei and L. pneumophila has not been determined. Indeed, very few studies have directly compared the pathogenesis of L. pneumophila with that of L. micdadei, and the variety of L. micdadei clinical isolates examined further complicates interpretation of this literature.

L. micdadei and L. pneumophila do share some features associated with virulence but not others. For example, like L. pneumophila, L. micdadei encodes flagella (6, 35), acid phosphatase (26), Mip (19, 57), and common antigen, or GroEL (5, 61), and both species inhibit superoxide anion generation by macrophages (25, 26). However, L. micdadei lacks phospholipase C and zinc metalloprotease, two factors thought to play a role in L. pneumophila virulence (26). And unlike L. pneumophila, macrophage internalization of L. micdadei via coiling phagocytosis (62) and association with the host ER (83) has not been observed. Indeed, whether L. micdadei belongs in the genus Legionella is controversial. As judged by the 16S rRNA similarity (30, 32) and DNA relatedness profiles (10), it can be argued that L. micdadei and L. pneumophila are distinct taxonomically. In spite of their genetic heterogeneity, some contend that the legionellae form a natural and practical phenotypic genus (7, 10).

A number of quantitative assays have been developed to evaluate particular traits associated with L. pneumophila virulence (13, 15, 41, 46, 60, 69, 79, 80, 82). To determine whether robust intracellular growth correlates with expression of one, or several, of these traits, four L. micdadei clinical isolates were compared to one well-characterized L. pneumophila strain. In addition, using a genetic gain-of-function strategy, we investigated whether any L. pneumophila loci could stimulate L. micdadei growth in cultured macrophages. Results obtained from both the approaches suggest that L. pneumophila and L. micdadei employ different strategies to replicate within amoebae and macrophages.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cell culture. Bone marrow-derived macrophages were prepared from female A/J mice (Jackson Laboratory) as described previously (77). After a 7-day culture period in L-cell conditioned medium, macrophages were collected by centrifugation, suspended in RPMI 1640 containing 10% fetal bovine serum (RPMI-FBS; Gibco BRL), and plated as described below for intracellular infection and replication, phagosome-lysosome fusion, and cytotoxicity assays. Cells of the human monocyte cell line U937 (American Type Culture Collection) were cultured in RPMI-FBS containing 25 mM HEPES buffer, pH 7.0, and differentiated by treatment with phorbol 12-myristate 13-acetate (Sigma) as described previously (60).

Bacterial strains and media. L. pneumophila Lp02, a virulent thymine auxotroph derived from the serogroup 1 Philadelphia-1 strain, was chosen as the prototype because it has been studied extensively (1, 8, 9, 12, 13, 21, 22, 36-38, 40, 46, 66, 67, 69, 70, 73, 74, 77-79). Four clinical isolates of L. micdadei were examined: 31B (University of Pittsburgh Hospital), Rivera (Stanford University Medical Center), and Camilleri (Stanford University Medical Center), kindly provided by Nicholas P. Cianciotto, and D-2676 (National Center for Infectious Diseases), kindly provided by Barry S. Fields. Bacterial strains were maintained at 70°C as glycerol stocks. Prior to experiments, strains were colony-purified either on N-(2-acetamido)-2-aminoethanesulfonic acid (ACES)-buffered charcoal-yeast extract agar (CYE) or on CYE supplemented with 100 µg of thymidine/ml (CYET) and then cultured in ACES-buffered yeast extract broth (AYE) or in AYE supplemented with 100 µg of thymidine/ml (AYET).

Intracellular bacterial growth. The abilities of Legionella strains to enter and survive in macrophages and to replicate within macrophages were measured as described previously (13, 77). Macrophages and U937 cells plated at a density of 2 × 10⁵ to 3 × 10⁵ cells per well in 24-well tissue culture dishes were infected for 2 h; next, extracellular bacteria were removed by washing the monolayers with RPMI-FBS, then incubating with 10 µg of gentamicin/ml in RPMI-FBS for 30 min. Infection initiation efficiency was calculated for triplicate samples by dividing the number of CFU associated with the monolayer at 2 h by the number of CFU added initially to the monolayer, then multiplying by 100. Bacterial replication in mouse macrophages and U937 cells was measured in triplicate over the subsequent 2- or 3-day period. Macrophages were infected with 31B and Lp02 at an approximate multiplicity of infection (MOI) of 1. For Camilleri, Rivera, and D-2676 infections, the MOI was increased to 5 to 10 to ensure that the number of CFU at 2 h was within the range of detection; neither the efficiency of infection initiation nor the growth rate of the strains was affected by the MOI (data not shown).

Microscopic assay for intracellular growth. The ability of individual intracellular bacteria to replicate was analyzed by fluorescence microscopy essentially as described previously (79). Macrophages cultured on 12-mm glass coverslips were infected at approximates MOI of 1 for Lp02 and 31B, 10 for Rivera, 50 for Camilleri, and 25 for D-2676, conditions which ensured that only very rarely were macrophages infected by more than one bacterium. After 2 h, extracellular bacteria were killed by incubating the cultures with gentamicin (10 µg/ml) for 30 min. At this time, one set of coverslips was fixed with periodate-lysine-paraformaldehyde (53) containing 4.5% sucrose and prewarmed to 37°C, washed three times with phosphate-buffered saline (PBS), then stored at 4°C until further use. The remaining coverslips were incubated for an additional 16 h at 37°C to allow Lp02 to replicate within, but not escape from, the primary host cell. The 18-h samples were fixed as described above; then all of the samples were methanol extracted and washed three times with PBS before the DNA was stained fluorescently with 0.1 µg of 4',6-diamidino-2-phenylindole (DAPI) per ml of PBS. For each time point, at least 50 infected macrophages were located, and the number of intact bacteria per macrophage was counted.

Phagosome-lysosome fusion assay. The ability of each Legionella strain to evade phagosome-lysosome fusion was analyzed 2 h after infection by using fluorescence microscopy and the soluble endocytic probe Texas red-ovalbumin (TRov) as described previously (77). Prior to methanol extraction of the preparations, extracellular bacteria were stained with rabbit serum specific for L. micdadei (Monoclonal Technologies, Inc., Norcross, Ga.) or L. pneumophila (77) (a kind gift from Ralph R. Isberg) followed by Cascade Blue-conjugated anti-rabbit immunoglobulin G (IgG) antibody (Molecular Probes, Eugene, Oreg.). After methanol extraction of the preparations, both intracellular and extracellular bacteria were stained with L. micdadei- or L. pneumophila-specific antiserum followed by Oregon Green-conjugated anti-rabbit IgG (Molecular Probes). The efficiency of phagosome-lysosome fusion was calculated by dividing the number of intracellular bacteria that colocalized with TRov by the total number of intracellular bacteria, then multiplying by 100.

Cytotoxicity assay. To quantify contact-dependent cytotoxicity, macrophages plated at a density of 3 × 10⁴ to 5 × 10⁴ cells per well of a 96-well tissue culture dish were incubated for 1 h with dilutions of each Legionella strain. Next, the infection medium was aspirated, the monolayers were incubated for 4 h with 0.1 ml of 10% (vol/vol) Alamar Blue (Accumed, Inc., Chicago, Ill.) in RPMI-FBS, and the redox-specific absorbance of the dye was measured as the OD₅₇₀ and OD₆₀₀ with a SpectraMax 250 spectrophotometer (Molecular Devices). The percentage of macrophages that were viable was calculated for triplicate samples from the slope of a plot of the A₅₇₀/A₆₀₀ determined for triplicate samples of six known densities of uninfected macrophages in the range of 10³ to 5 × 10⁴ cells per well. To determine the actual MOI, duplicate samples of the infection medium were plated onto CYE or CYET.

Sodium sensitivity and osmotic sensitivity. The Na^s of exponential- and post-exponential-phase broth cultures was determined as described previously (13) and calculated as (CFU on CYE or CYET agar containing 100 mM NaCl)/(CFU on CYE or CYET) × 100. The osmotic sensitivity of exponential- and post-exponential-phase broth cultures was determined as described previously (13) by using AYE or AYET broth which did or did not contain 0.3 M KCl and was calculated as (CFU of KCl-treated samples)/(CFU of untreated samples) × 100.

Conjugation efficiency assay. How efficiently each Legionella strain donated plasmid DNA by conjugation was quantified by the method of Vogel et al. (80). First, each of the Legionella strains was transformed with pMS8 (1), a derivative of the mobilizable IncQ RSF1010 plasmid, which confers kanamycin resistance. Approximately 10⁹ CFU of exponential-phase donor cells was mixed with 10⁸ to 10⁹ CFU of the recipient strain, Escherichia coli DH5. The bacteria were collected by centrifugation, spread onto CYE or CYET, then incubated for 4 h at 37°C. To select for transconjugants, the mating mixture was collected onto a sterile swab, diluted into water, and plated onto Luria broth (LB)-kanamycin, which does not permit Legionella growth. Conjugation efficiency was calculated as (CFU of transconjugants)/(CFU of donor) × 100.

Genetic screen for genes that confer increased growth and survival in macrophages on L. micdadei. To identify L. pneumophila genetic loci which enhance the intracellular growth of L. micdadei, we chose strain D-2676 as a cloning vehicle to minimize the background of false-positives. An L. pneumophila genomic library (1) consisting of 5- to 10-kb Sau3A Lp02 genomic fragments cloned into vector pMS8, which confers kanamycin resistance, was transferred by conjugation to D-2676. Transconjugants were collected as three separate pools, each containing approximately 10⁴ members and representing more than 15 genome equivalents. To enrich for L. micdadei strains that replicated efficiently intracellularly, each pool of transconjugants was added at an approximate MOI of 1 to cultures of 2 × 10⁶ macrophages in a six-well tissue culture dish. After a 4.5-h incubation, extracellular bacteria were removed by washing the monolayers with RPMI-FBS, then incubating the infected cultures with 10 µg of gentamicin/ml for 4 h. After 3 days, monolayers were lysed in PBS, and the surviving bacteria were recovered on CYE-kanamycin. Each pool of surviving bacteria was then subjected to a second enrichment cycle.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Intracellular replication of Legionella. A hallmark of Legionella pathogenesis is the capacity to survive and to replicate within macrophages (48, 49, 57). Therefore, entry and survival in primary bone marrow macrophages by L. micdadei 31B, Rivera, Camilleri, and 31B and L. pneumophila Lp02 were compared. Macrophages were infected for 2 h; then extracellular bacteria were killed with gentamicin and cell-associated CFU were enumerated. As observed previously (13), approximately 8% of the inoculum of post-exponential-phase L. pneumophila cells initiated an infection of macrophages (Fig. 1). In comparison, only 1 to 3% of the L. micdadei strains did so, indicating less-efficient binding, entry, and/or survival in macrophages.

View larger version (16K): [in this window] [in a new window]   FIG. 1.   Efficiency of initiation of infection. Macrophages were incubated with each Legionella strain for 2 h; then the numbers of viable and cell-associated bacteria were determined by a standard gentamicin resistance assay (see Materials and Methods). Shown are the means and standard errors of the means calculated in one experiment performed in triplicate; similar results were obtained in six additional experiments.

View larger version (13K): [in this window] [in a new window]   FIG. 2.   Intracellular growth of Legionella strains. (A) U937 cells were infected with the Legionella strains shown; then, at the times indicated, the numbers of viable bacteria were determined. The relative number of CFU was calculated by dividing the CFU at each time point by the CFU at the first time point; the actual initial CFU was 3.5 × 104 for Lp02 (open circles), 1.4 × 104 for 31B (solid circles), 2.7 × 104 for Rivera (open squares), 1.0 × 104 for Camilleri (open triangles), and 1.1 × 104 for D-2676 (solid triangles). The mean CFU was determined from triplicate samples; the standard error of the mean is indicated by error bars. Similar results were obtained in two other experiments. (B) Macrophages were infected with the Legionella strains shown; then intracellular growth was determined as described above. The actual initial CFU was 2.9 × 104 for Lp02 (open circles), 8.6 × 103 for 31B (solid circles), 2.1 × 104 for Rivera (open squares), 8.5 × 103 for Camilleri (open triangles), and 2.4 × 104 for D-2676 (solid triangles). Similar results were obtained in six other experiments.

Intracellular growth by individual bacteria. The lower yield of CFU in primary macrophages observed for L. micdadei Rivera, Camilleri, and D-2676 could indicate that the majority of cells in the inoculum may replicate at a lower rate than L. pneumophila. Alternatively, a subset of cells in the inoculum may replicate efficiently while the majority do not multiply at all or are killed. To differentiate between these two possibilities, intracellular replication by individual cells was analyzed microscopically under conditions which ensured that only one bacterium entered each macrophage. Late in the primary infection period, approximately one-third of the macrophages infected with L. pneumophila contained more bacteria than could be counted, an indication of robust intracellular growth (Table 1). As expected, strain 31B showed a similar profile, confirming its efficient replication in macrophages (Table 1; Fig. 2). In contrast, more than 85% of the macrophages infected with Rivera, Camilleri, or D-2676 contained no more than 10 bacteria, indicating that most intracellular bacteria had not replicated efficiently. Interestingly, when infected with Rivera, Camilleri, or D-2676, a small minority of macrophages (<10%) contained bacteria too numerous to count. That a subpopulation of each of these strains established a primary infection in macrophages may reflect heterogeneity in the virulence of the bacterial population; alternatively, macrophages may not respond uniformly to infection.

Evasion of phagosome-lysosome fusion. To determine whether the reduced initiation of infection (Fig. 1) reflected decreased survival of intracellular L. micdadei, the interaction between macrophage lysosomes and phagosomes harboring each of the strains was examined. First, macrophage lysosomes were labeled by endocytosis of the soluble fluorescent marker TRov; 2 h after infection with post-exponential-phase Legionella, colocalization of bacteria and TRov was evaluated by fluorescence microscopy (77). As expected, fewer than 4% of phagosomes harboring virulent L. pneumophila contained the lysosomal marker (Fig. 3A). In contrast, approximately 80% of Rivera, Camilleri, and D-2676 phagosomes had fused with lysosomes (Fig. 3A), a result consistent with the lower numbers of viable intracellular bacteria 2 h after infection (Fig. 1).

View larger version (13K): [in this window] [in a new window]   FIG. 3.   Evasion of phagosome-lysosome fusion by Legionella strains. (A) Macrophages prelabeled with the endocytic probe TRov were infected with the Legionella strain indicated for 2 h; then the percentage of intracellular bacteria that colocalized with TRov was determined by fluorescence microscopy. Intracellular and extracellular bacteria were stained differentially by using Legionella-specific antisera (Materials and Methods). (B) Evasion of phagosome-lysosome fusion was determined essentially as described above, except that macrophages were infected with Lp02, 31B, and D-2676 which had been prelabled with FLUOS. At least 50 intracellular bacteria were scored in each experiment; shown are the means and the standard errors of the means calculated from three experiments.

Cytotoxicity for macrophages. L. pneumophila can kill macrophages by a replication-independent but contact-dependent mechanism (41). Cytotoxicity has been correlated with L. pneumophila virulence, although its role in pathogenesis is not known (46). To measure L. micdadei cytotoxicity, macrophages and bacteria were cocultured for 1 h; then macrophage viability was measured by using the colorimetric dye Alamar Blue (51). As expected, at an MOI greater than 5, L. pneumophila killed approximately 95% of the macrophage monolayer (Fig. 4). In contrast, none of the L. micdadei strains were cytotoxic (Fig. 4). Even at the high MOI of 50, neither exponential- nor post-exponential-phase cultures of L. micdadei affected macrophage viability (data not shown). Surprisingly, even strain 31B, which replicates efficiently in macrophages, lacked contact-dependent cytotoxicity. Therefore, if the cytotoxicity of L. pneumophila is critical for establishment of its unique replication niche (45) and/or escape from amino acid-depleted host cells (13), L. micdadei must use an alternative strategy.

View larger version (20K): [in this window] [in a new window]   FIG. 4.   Cytotoxicity of Legionella strains for macrophages. Macrophages were incubated for 1 h with the bacterial strain indicated; then macrophage viability was assessed by using the redox-sensitive dye Alamar Blue (see Materials and Methods). Shown are the percentages (means and standard errors) of viable macrophages infected at an MOI of 5 with Camilleri, Rivera, and 31B (n = 6), D-2676 (n = 4), and Lp02 (n = 2).

Sodium sensitivity. The ability of L. pneumophila to survive and to replicate in macrophages has been correlated with Na^s (15, 69, 82), a trait expressed in the post-exponential phase (13). Yet two broad surveys of Legionella indicate that this correlation is not complete (2, 58). To assess the association between the virulence and Na^s of L. micdadei, bacteria collected from exponential- and post-exponential-phase cultures were plated onto CYE which did or did not contain 100 mM NaCl; then colony formation was quantified. Consistent with the differences observed between virulent and avirulent L. pneumophila, the severely attenuated L. micdadei strain D-2676 was Na^r, while the most virulent strain, 31B, was relatively Na^s in both the exponential and the post-exponential phase (Fig. 5). Rivera and Camilleri, which appear to be partially attenuated, were slightly Na^s during the exponential phase and more so in the post-exponential period, consistent with a previous report (58). In general, two rules held: relative Na^s correlated with virulence, and the Na^s of each of the strains increased in the post-exponential phase.

View larger version (17K): [in this window] [in a new window]   FIG. 5.   Sodium sensitivities of Legionella strains. Bacteria cultured to the exponential (E) and post-exponential (PE) phase were plated in duplicate onto CYE or CYET to quantify total CFU and onto CYE or CYET containing 100 mM NaCl to enumerate sodium-resistant CFU. Shown is the percent plating efficiency on sodium-containing medium determined in three experiments, each represented by a unique symbol (open circles, open squares, or solid triangles).

Conjugal transfer of DNA. It has been proposed that L. pneumophila secretes virulence factors via the putative Dot-Icm complex, which also mediates the conjugal transfer of plasmid DNA (73, 80). To determine whether conjugation activity correlates with the virulence of L. micdadei, we compared the conjugation efficiencies of the four strains of L. micdadei with that of L. pneumophila. As predicted by the Dot-Icm secretion model (73, 80), the virulent L. micdadei strain 31B transferred DNA to the recipient more efficiently than L. pneumophila, whereas the three attenuated L. micdadei strains donated DNA at a rate approximately 10-fold lower than that of L. pneumophila (Fig. 6). In contrast, L. pneumophila dotB mutants donate plasmid DNA 100- to 1,000-fold less efficiently than the wild type (80) (data not shown). Whether transfer of plasmid DNA by L. micdadei is mediated by a Dot-Icm complex remains to be tested.

View larger version (19K): [in this window] [in a new window]   FIG. 6.   Conjugal DNA transfer. Legionella organisms cultured to exponential phase were mixed with an approximately equal number of CFU of the recipient strain DH5, then incubated for 4 h. Appropriate dilutions of the mating mix were plated in duplicate onto LB-kanamycin plates to quantify conjugation efficiency. Shown is the number (mean and standard error) of conjugants obtained per donor in one experiment; similar results were obtained in at least one other experiment. The difference in conjugal efficiency between L. pneumophila and the L. micdadei strains was significant (P < 0.001 by the 2 test of independence).

Osmotic resistance. Like E. coli (43), L. pneumophila becomes osmotically resistant as it exits the exponential phase (13). A similar phenotypic switch was observed for all four strains of L. micdadei. As cultures of each of the strains entered the post-exponential phase, their osmotic resistance increased approximately 10-fold (Fig. 7). Thus, osmotic resistance correlated with the growth phase, but not the virulence, of L. micdadei.

View larger version (24K): [in this window] [in a new window]   FIG. 7.   Osmotic sensitivity of Legionella. Exponential- and post-exponential-phase cultures of Legionella were incubated for 1 h in broth with or without 0.3 M KCl, diluted into distilled water, then plated in duplicate onto CYE or CYET to quantify CFU. Shown are the means and standard errors of the means obtained in six experiments.

Gain-of-function genetic screen for L. pneumophila virulence factors. Two observations indicated that the L. micdadei strains are equipped for intracellular growth: all of the strains replicated in the U937 cell line (Fig. 2A), and a subpopulation of each exhibited robust growth in primary macrophages (Table 1). Yet, in nearly every phenotypic test, L. micdadei appeared less virulent than L. pneumophila. In an attempt to identify the genetic basis for this difference, we designed a gain-of-function, interspecies complementation strategy applied previously for a variety of pathogenesis model systems (4, 20, 23, 42, 54, 59, 75, 85). In particular, we screened an L. pneumophila genomic library for loci which conferred a robust intracellular growth phenotype on L. micdadei. To maximize the selective pressure exerted by the macrophages, we chose as the parent strain D-2676, a clinical isolate which is killed in macrophage lysosomes (Fig. 2B and 3).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

We took a comparative approach toward understanding the pathogenesis of the two causative agents of Legionnaire's disease, L. pneumophila and L. micdadei. Using a series of quantitative phenotypic assays, we sought to identify those traits most likely to be required by Legionella for robust growth in macrophages. In nearly every test, L. micdadei strains appeared less virulent than L. pneumophila (Table 3). One hypothesis consistent with these results is that L. pneumophila and L. micdadei encode the same battery of virulence factors, but L. micdadei lacks a factor critical for high-level expression of one or more traits. Attempts to augment the intracellular growth of L. micdadei by providing L. pneumophila genomic fragments in trans were not successful (Table 2). Therefore, simplistic models in which a single virulence determinant or locus differentiates these pathogenic Legionella spp. were not supported. Instead, results of this study and others (62, 83) suggest that L. pneumophila and L. micdadei, though similar in some respects, use different strategies to parasitize host cells.

L. pneumophila expresses a contact-dependent cytotoxicity (41) concomitant with exit from the exponential phase of growth (13). Recently, a pore-forming activity that correlated with L. pneumophila virulence was characterized. When added to either erythrocytes or macrophages at a high MOI, L. pneumophila organisms insert a pore which leads to osmotic lysis and rapid cell death (46). Under normal infection conditions, L. pneumophila may insert pores into its nascent phagosomal membrane to block, by some undefined mechanism, subsequent fusion with the lysosomal network (45). Alternatively, or additionally, the cytotoxin may facilitate the lysis of host membranes by starved bacteria, which must escape to locate a new supply of nutrients (13). Surprisingly, none of the L. micdadei strains were cytotoxic for macrophages (Fig. 4), yet three of the strains replicated efficiently in U937 cells, and strain 31B also exhibited robust growth in the more-stringent environment of primary mouse macrophage cultures (Fig. 2). Moreover, strain 31B efficiently evaded phagosome-lysosome fusion in macrophages (Fig. 3B). Therefore, L. micdadei and L. pneumophila may establish and/or escape from replication vacuoles by different mechanisms. Alternatively, cytotoxicity may be dispensable for L. pneumophila growth in macrophages, despite a genetic or regulatory linkage to virulence (45, 46).

The L. pneumophila dot and icm genes are required for intracellular growth and for efficient conjugal transfer of plasmid DNA (71, 81). Thus, like A. tumefaciens (17) and B. pertussis (84), L. pneumophila may use a type IV secretion system to export virulence factors. A functional Dot-Icm apparatus has been correlated genetically with two other L. pneumophila virulence traits: Na^s and contact-dependent cytotoxicity (69, 81, 82). By one model, export of a cytotoxin or other virulence factors by the Dot-Icm transport apparatus permits the accumulation of inhibitory levels of NaCl in the cytosol; mutational inactivation of this machinery confers salt resistance, loss of cytotoxicity, and avirulence (82). In our studies of L. micdadei, a strict correlation between conjugation efficiency, Na^s, and cytotoxicity was not observed. For example, although the Rivera, Camilleri, and D-2676 strains of L. micdadei appeared similarly competent to transfer plasmid DNA by conjugation (Fig. 6), their degrees of Na^s differed (Fig. 5), and none were cytotoxic (Fig. 4). Apparently, conjugal transport systems of Legionella do not necessarily confer Na^s.

To begin to address whether L. micdadei encodes a Dot-Icm transport system, we tested whether L. micdadei contained genomic sequences homologous to dotA, dotB, dotE, dotF, or dotG of L. pneumophila. Under low-stringency Southern hybridization conditions which allowed a 30% base pair mismatch (T_m 47°C) (3), full-length probes of dotA, dotB, dotE, and dotFG genes each hybridized with genomic fragments from each of the four L. micdadei strains (44). When the stringency was increased to allow a 20% base pair mismatch (T_m 30°C) (3), only the dotA probe hybridized with L. micdadei genomic DNA. These data indicate that L. micdadei harbors sequences homologous to both dot-icm region I and region II (71). Whether L. micdadei encodes functional copies of the 23 dot-icm genes or assembles a functional Dot-Icm complex important for conjugation or intracellular growth remains to be established.

In macrophages and amoebae, L. pneumophila replicates in a compartment bounded by the ER (33, 37, 77). Based on morphological and kinetic studies of replication vacuole formation, Swanson and Isberg had proposed that L. pneumophila acquires a rich supply of nutrients by stimulating the eukaryotic autophagy pathway (77). In contrast, none of the L. micdadei strains examined to date appear to associate with the ER (83); instead, L. micdadei occupies a dilated phagosome (62). According to the autophagy exploitation model, the inability of L. micdadei to associate with the ER to obtain nutrients could account for its poor intracellular growth. However, L. micdadei 31B, which replicated in macrophages as efficiently as did L. pneumophila, occupied a dilated phagosome which was not associated with the ER, as judged by immunofluorescence microscopic localization of the ER luminal protein Bip (44, 77). Therefore, even though ER association has been correlated with the intracellular growth of L. pneumophila (37, 77, 79), it does not appear to play a role in L. micdadei pathogenesis. Thus, the intriguing question of how intracellular Legionella organisms obtain the amino acids needed for replication remains.

Designation of 31B as L. micdadei was suspect because its intracellular growth characteristics differed from those of the three other L. micdadei strains studied and because it stained poorly with the L. micdadei-specific antiserum. In particular, immunofluorescent staining of broth-grown 31B required fourfold-higher antiserum concentrations to achieve the intensity seen for Rivera, Camilleri, and D-2676. Interestingly, genetic divergence of 31B was also suggested by Southern hybridization analysis, which revealed restriction fragment polymorphisms in the chromosomal loci homologous to L. pneumophila dotA, dotB, dotE, and dotFG for 31B compared to the other three L. micdadei strains. On the other hand, 31B did not react strongly with L. pneumophila-specific antiserum (data not shown). Furthermore, previous molecular studies of the Mip gene and protein of strain 31B established its relatedness to several other L. micdadei clinical isolates (57). Additional molecular genetic studies are required to explain the potent virulence of 31B relative to other clinical isolates of L. micdadei.

Taken together, our data indicate that the differential virulence of L. pneumophila and L. micdadei is not attributable to a single phenotypic or genetic trait. It remains possible that L. micdadei lacks a particular genetic determinant which is critical for robust intracellular growth but which either was not represented in the L. pneumophila library, was poorly expressed by L. micdadei, or was not sufficient to bypass the more extensive genetic differences of D-2676. However, we favor the alternative view that Legionella is a diverse genus whose species appear to use quite different strategies to parasitize host cells. Perhaps the most intriguing questions raised by this comparative study are whether cytotoxicity, sodium sensitivity, the conjugation apparatus, and the host ER contribute to L. pneumophila pathogenesis, and if so, how.