Cytopathogenicity and molecular subtyping of Legionella pneumophila environmental isolates from 17 hospitals

Summary

The cytopathogenicity of 22 Legionella pneumophila isolates from 17 hospitals was determined by assessing the dose of bacteria necessary to produce 50% cytopathic effect (CPED₅₀) in U937 human-derived macrophages. All isolates were able to infect and grow in macrophage-like cells (range log₁₀ CPED₅₀: 2·67–6·73 c.f.u./ml). Five groups were established and related to the serogroup, the number of PFGE patterns coexisting in the same hospital water distribution system, and the possible reporting of hospital-acquired Legionnaires' disease cases. L. pneumophila serogroup 1 isolates had the highest cytopathogenicity (P=0·003). Moreover, a trend to more cytopathogenic groups (groups 1–3) in hospitals with more than one PFGE pattern of L. pneumophila in the water distribution system (60% vs. 17%) and in hospitals reporting cases of hospital-acquired Legionnaires' disease (36·3% vs. 16·6%) was observed. We conclude that the cytopathogenicty of environmental L. pneumophila should be taken into account in evaluating the risk of a contaminated water reservoir in a hospital and hospital acquisition of Legionnaires' disease.

INTRODUCTION

Hospital-acquired Legionnaires' disease (LD) has been reported in many hospitals since the first outbreak in 1976 [1]. Although cooling towers were linked to many of the cases of LD in the years after its discovery, potable water has been the environmental source for almost all reported hospital outbreaks [2–4]. Legionella spp. are common commensals of large building water-supply systems. In a previous study we demonstrated that L. pneumophila was present in 17 out of 20 hospital water systems [5] and prospective surveillance showed that cases of hospital-acquired LD appeared in 11 of these hospitals [6]. The remainder did not report any cases of LD, although appropriate techniques were available for the diagnosis of Legionella spp. and most (70%) reported >30% positive peripheral sampling points.

There are some clinical data indicating that the virulence of Legionella spp. may vary, although it unclear whether this is a consequence of differences in bacterial virulence or in host immunity. Examples are: the survival of patients with LD despite not receiving adequate antibiotic treatment [7], the variability in the mortality observed in different community outbreaks [8, 9], or the fact that most cases are caused by L. pneumophila serogroup (sg) 1 [10]. In experimental studies, isolates from amoebas exhibit a variety of phenotypes that appear to increase the incidence and complications of LD in humans [11–13]. Furthermore, some authors have demonstrated that Legionella strains causing pulmonary disease express more virulence traits than those that do not cause human disease [14]. Virulence traits show differences between species [14, 15] and strains of L. pneumophila [16–19].

As the demographic characteristics of the patients in the 17 hospitals reported earlier [5, 6] were similar, we questioned why hospitals with L. pneumophila in their water distribution systems did not experience cases of hospital-acquired LD and hypothesized that the virulence of the L. pneumophila strains may have influenced the appearance of clinical cases. We therefore assessed the cytopathogenicity of L. pneumophila environmental strains isolated from each hospital and determined their relationship with sg1 and the number of different strain genotypes coexisting in the same water distribution system to gain insight into the contribution of cytopathogenicity and potential to cause disease in each hospital.

MATERIAL AND METHODS

Bacteria and culture

A total of 22 environmental L. pneumophila strains from 17 hospital water-supply systems previously characterized [5] by serogroup, and chromosomal DNA subtype by pulsed-field gel electrophoresis (PFGE) were included in the present study (Table 1). The bacteria had a ⩽2 passage history and were stored in brain heart infusion broth (Oxoid, Wesel, Germany) supplemented with 10% glycerol at −80°C. All strains were susceptible to 100 mm NaCl [20], a marker of laboratory attenuation (data not shown).

Table 1. DNA patterns, Legionella pneumophila serogroup, cytopathogenicity and number of cases of hospital-acquired Legionnaires' disease reported in each hospital during the surveillance period

CPED₅₀, The CPED₅₀ was defined as the minimum number of bacteria necessary to produce a cytopathic effect in 50% of the U937 cell infected monolayers after 72 h incubation; Non-sg1, L. pneumophila non-serogroup 1.

* Hospital that reported hospital-acquired Legionnaires' disease (1996–2001).

For cytopathogenicity and intracellular growth assays, strains were cultured on BCYE-α agar for 72 h and resuspended in antibiotic-free tissue culture medium containing 5% fetal calf serum (FCS).

Cytopathogenicity assay

The human monocyte cell line U937 was cultured in RPMI 1640, 2 mml-glutamine and 10% FCS. Prior to infection, the cells were differentiated for 72 h with 10⁻⁸m phorbol 12-myristate 13-acetate (PMA) (Sigma Chemical Co., St Louis, MO, USA) as described previously [21]. The relative cytopathogenicity of L. pneumophila strains was determined by the 50% cytopathic effect (CPE₅₀). CPED₅₀ was defined as the minimum number of bacteria necessary to produce a cytopathic effect in 50% of the U937 cell infected monolayers after 72 h incubation. Briefly, eight replicate monolayers containing 2×10⁵ macrophage-like cells in 96-well microtitre plates were infected with serial dilutions of each strain (range of inocula, 10¹–10⁹ bacteria per monolayer) for 90 min. The monolayers were washed and treated with 80 μg/ml of gentamicin and then incubated with RPMI containing 10% FCS at 37°C for 72 h. The monolayers were stained with tetrazolium salt 3-(4,5-dimethylthiazol-2-yl)-2,5,-diphenyltetrazolium bromide (MTT; Sigma) and the viability of the cells was determined measuring the optical density (OD) and expressed as percentage of cell death compared with uninfected cells using the formula

The number of bacteria that reduced the OD by 50% compared with the OD of uninfected cells was defined as the CPED₅₀ value. Each experiment was performed in triplicate.

Intracellular growth

U937 macrophage-like cells plated at a density of 1×10⁶ cells per well in 24 tissue culture dishes were infected at a multiplicity of infection (MOI) of 0·2. Thereafter, extracellular bacteria were removed by washing the monolayers with RPMI followed by incubation with gentamicin in RPMI for 30 min. Intracellular growth was measured in duplicate at 0, 24 and 48 h and expressed as c.f.u./monolayer. Variations between experiments were corrected with the MOI value (dividing the result c.f.u./well by MOI). The relative increase in c.f.u. was calculated by subtracting the c.f.u. at each time-point from the first time-point.

Statistical analysis

Groups of cytopathogenicity were determined with the K means cluster analysis using SPSS software program version 12.0 (SPSS Inc., Chicago IL, USA). The Mann–Whitney U and Kruskal–Wallis tests were used for statistical analysis of quantitative variables and Fisher's exact test was used for qualitative variables. P values <0·05 were considered significant.

RESULTS

All strains were able to infect and grow in macrophage-like cells and produce a significant cytopathic effect. Values ranged from log CPED₅₀ 2·67 c.f.u./ml to log CPED₅₀ 6·73 c.f.u./ml and five groups were determined, ranging from 1 (most) to 5 (least) cytopathogenic strains (Fig. 1). Cytopathogenic groups (1 and 2) had a higher intracellular infection efficiency (T=0 h, P=0·031). At 24 h all isolates showed increased intracellular growth to 2·05±0·74 log units (range 1·04–3·07) and at 48 h to 3·29±0·74 log units (range 1·75–4·93) (Fig. 2).

Fig. 1. Cytopathogenic groups of the different environmental isolates. Data of each isolate are represented by a point and error bars represents standard deviations. Mean data of each cytopathogenic group is represented by a horizontal bar. Groups were established with the K means cluster analysis (SPSS). The CPED₅₀ groups were statistically different (P<0·001, Kruskal–Wallis test).

Fig. 2. Intracellular growth of cytopathogenic groups of L. pneumophila isolates in co-cultures with U937 macrophage-like cell monolayers. •, Cytopathogenic group 1; ■, group 2; ○, group 3; ▲, group 4; ×, group 5. The values are the mean of groups and error bars represents standard deviations.

L. pneumophila sg1 strains had a significantly higher mean cytopathogenicity (log CPED₅₀ 4·92±1·28 vs. 5·75±0·68, P=0·003) and were distributed in the top three groups more frequently than other strains [50% (5/10) vs. 8·3% (1/12), P=0·06] (Table 1). Strains from hospital water systems harbouring more than one DNA type (PFGE) more often fell into the top three cytopathogenic groups than strains from systems that yielded a single genotype (3/5 vs. 2/12 hospitals respectively, P>0·05). Similarly, cytopathogenicity was greatest among strains from hospitals reporting hospital-acquired LD cases than those without clinical cases (4/11 vs. 1/6 hospitals respectively, P>0·05).

DISCUSSION

It is not clear why some hospitals with environmental colonization by L. pneumophila have nosocomial LD and others do not. This is often attributed to differences in engineering design and maintenance, bacterial concentration at peripheral points (taps and shower faucets) and host susceptibility. The virulence of the bacteria is also thought to play a major role.

We found that all environmental isolates of L. pneumophila tested were able to infect and grow in U937 macrophage-like cells resulting in a significant cytopathic effect. Almost three-quarters of the strains fell into weakly cytopathogenic groups and were invariably non-sg1; half of the 10 strains of sg1 were grouped as strongly cytopathogenic. Moreover, strains from mixed Legionella populations in water systems were more often cytopathogenic than those recovered in single strain culture and importantly, this association held for environmental strains from hospitals with cases of LD compared with those free of cases.

It is well established that Legionella spp. differ in virulence determined by cytopathogenicity and other assays and can be grouped accordingly [14, 19, 22–25]. However, there are few studies evaluating virulence traits of environmental isolates of L. pneumophila [10, 18, 24, 26]. Most cases of disease are caused by sg1 strains and, consequently, it has been suggested that this serogroup is more virulent than other serogroups of L. pneumophila, but this has not been demonstrated experimentally [14, 26, 27].

The higher cytopathogenicity of L. pneumophila from mixed strain populations in water systems has not been previously described but our data are consistent with the finding of the coexistence of different L. pneumophila genotypes in cooling towers where the limitation of nutrients in the non-potable water induces an overgrowth of the most virulent strain and displacement of other strains leading to an outbreak of disease [28]. Indeed, it has been demonstrated by in vitro studies that depleted nutrients enhance the expression of virulence traits [20]. The coexistence of more than one strain population in the hospital water distribution system may provide the microbial complexity to facilitate biofilm formation and intracellular growth in protozoa. L. pneumophila growing within amoeba changes phenotypically and exhibits increased invasion ability [29] compared with cells grown in conventional media. The similarities in the model of intracellular infection between protozoa and alveolar macrophages [30] suggest that the virulence of L. pneumophila for alveolar macrophages is a consequence of its evolution as a parasite of amoeba [31]. Thus, considering that protozoa are important determinants in the ecology of L. pneumophila in aquatic environments, studies to determine the different populations of amoeba in water supplies of hospitals and the ability of different strains to grow in them and alveolar macropages would be informative of the ecology and pathogenicity of the species.

Hospitals harbouring cytopathogenic strains of L. pneumophila strains in water supplies tended to report hospital-acquired LD cases more frequently. Unfortunately, owing to a lack of active surveillance clinical isolates were not available for comparison with the environmental isolates. However, several authors have noted the importance of strain virulence in the appearance of hospital-acquired LD since in hospitals colonized by more than one strain genotype, the clinical strain was often identical to the most virulent environmental strain [16, 17].

Some authors have pointed out that the presence of Legionella spp. in a hospital water distribution system, the use of specific techniques for the diagnosis of Legionella pneumonia and in-patient host factors such as immunosuppression, are the most important prognostic factors for the appearance of hospital-acquired cases [1]. Nevertheless, since the complete eradication of Legionella spp. from a potable water distribution system is impossible without highly cost-effective disinfection methods, knowledge of the cytopathogenicity of an environmental Legionella spp. isolate may be useful in evaluating the risk of the emergence of hospital-acquired LD from a contaminated water source.