Community-acquired Legionnaires' disease caused by Legionella pneumophila serogroup 10 linked to the private home

Legionnaires' disease, the pneumonic form of legionellosis, usually is acquired by inhalation or aspiration of legionellae from contaminated environmental sources. Potable water is an important source of both nosocomial and community-acquired Legionella infections (Fields et al., 2002). Since 2001, legionellosis is a notifiable disease in Germany. Around 500 cases are reported through the public health system each year, which is considered to be the tip of the iceberg (Robert Koch Institut, 2006). Based on an ongoing study on the aetiology and outcome of community-acquired pneumonia in Germany (), the frequency of Legionella pneumonia is estimated to be 15 000 to 30 000 cases per annum (Lück et al., 2006). The majority of community-acquired cases are caused by strains belonging to Legionella pneumophila serogroup 1 (Helbig et al., 2002, Yu et al., 2002). Among the 1300 clinical isolates from Europe, serogroup 10 is the fourth most common cause of Legionella pneumonia (Helbig et al., 2002).

Due to the ubiquitous prevalence of legionellae in water supply systems, strains isolated from patients and environmental sources must be compared by molecular typing techniques to confirm or exclude a suspected environmental reservoir as the source of the infection. These techniques include mAb typing, various DNA restriction patterns [amplified fragment length polymorphism (AFLP) or PFGE] (Fry et al., 1999; Lück et al., 1994; Sax et al., 2002) and more recently sequence-based typing (SBT) (Gaia et al., 2005). This method is in fact a variation of multilocus sequence typing and uses sequence variations in five virulence-associated L. pneumophila genes (flaA, pilE, mip, mompS, proA) and one house-keeping gene (asd). It has been successfully used to type L. pneumophila serogroup 1 strains, and a few strains belonging to serogroup 6, 8 and 10 (Gaia et al., 2005). It is now established as the standard subtyping technique within the European Working Group for Legionella Infections (EWGLI).

In this study the SBT method was used to compare the allelic profiles of L. pneumophila serogroup 4, 5, 8 and 10 isolates from patients and environmental sources. These strains were selected because they are cross-reacting and might be considered a group of antigenically related strains.

A 66-year-old man was hospitalized at the University Hospital of Berlin from August 23rd to September 8th 2004. Eighteen days later, he was readmitted to the intensive care unit of the same hospital with a 3 day history of fever. The patient has received a cadaveric kidney transplant 4 months earlier and was now on cortisone therapy. On admission, his vital signs were as follows: core body temperature 32.8 °C, pulse 92 beats min^–1, blood pressure 90/65 mmHg, respiratory rate 28 breaths min^–1 and his chest radiograph showed bilateral consolidation. He was intubated, given ventilatory assistance and empiric antibiotic therapy with macrolides. His clinical status improved slowly and the X-ray showed a resolution of the pneumonia. The further course was complicated by invasive aspergillosis, cytomegalovirus pneumonia, failure of the transplanted kidney and septic anaemia. Four months after developing Legionella pneumonia he died of multi-organ failure. At the time of death no radiological signs of pneumonia remained. However, it must be emphasized that the Legionella infection was the onset of the chain of events leading to the fatal outcome.

A bronchoalveolar lavage fluid collected on day 2 after admission was positive for L. pneumophila-species antigen by direct immunofluorescence mAb staining (Bio-Rad). Subsequently, Legionella colonies grow on buffered-charcoal yeast extract agar. By using mAbs the strain was typed as serogroup 5/10, since both serogroup-specific mAbs reacted brightly with this strain (Helbig et al., 2002). A urine sample taken on the same day gave a positive result using the Biotest Legionella urinary antigen test.

Environmental samples were taken from hospital and home water supplies according to standard protocols (ISO, 2004). Water samples from the hospital where the patient stayed yielded L. pneumophila serogroups 1 and 9. The concentrations ranged from 10 000 to 20 000 c.f.u. l^–1. L. pneumophila serogroup 5/10 was not detected. However, water samples from the private home of the patient collected 3 and 5 weeks after the onset of the clinical illness yielded 30 and 280 c.f.u. l^–1 of L. pneumophila serogroup 5/10, respectively. An environmental isolate from the hospital in the patient's town was also serogroup 10. All isolates were then referred to the German Legionella Reference Laboratory in Dresden and typed by using the Dresden panel of mAbs (Table 1). Epidemiologically unrelated strains of antigenically related serogroups 4, 5, 8 and 10 were taken from the strain collection in Dresden.

Table 1. Allelic profiles of L. pneumophila isolates belonging to serogroups 4, 5, 8 and 10 from the patient, his private home, and nine unrelated strains

To obtain an initial indication of the genetic relatedness of the L. pneumophila isolates from the patient, the hospital water system and from tap water at his home, the isolates were genotyped by the AFLP method as described by Fry et al. (1999). Subsequently, SBT was used to genotype all strains listed in Table 1 as described previously (Gaia et al., 2005). The allele numbers of the isolates were determined by using the EWGLI software tool (available online at ). The results of AFLP and SBT typing confirmed that the clinical strain and environmental isolates from the patient's home were indistinguishable. This is a strong argument for the transmission from the patient's home water supply to the patient. All unrelated strains were of different SBT patterns (Fig. 1, Table 1). To date, the SBT profile of the patient's isolate has only been recorded once before from an isolate from Tokyo, Japan ().

Table 1; lane 3, unrelated isolate of L. pneumophila serogroup 1 from a patient of the same hospital suffering from nosocomial Legionella infection; lane 4, unrelated isolate from the water system of the University Hospital of Berlin, where the patient had been treated (L. pneumophila serogroup 1); lanes 5–7, three isolates from tap water outlets of the patient's home (L. pneumophila serogroup 5/10, isolates of lane 5 and 6 correspond to isolates W04-989 and W04-990 of Table 1); lane 8, L. pneumophila serogroup 1, Knoxville-1 (ATCC 33153).

The frequency of community-acquired legionellosis is often under-reported. This is mainly due to the fact that specific laboratory tests are required to diagnose a Legionella infection. The patient reported here had undergone kidney transplantation 4 months previously and therefore belonged to the high risk group for legionellosis. He presented with the clinical signs compatible with Legionella pneumonia. This aetiology was clinically suspected and confirmed by three laboratory methods. In our case the Biotest urine antigen assay detected the infection caused by serogroup 5/10. Currently, the vast majority of legionellosis cases are detected by urinary-antigen detection (DenBoer & Yzerman, 2004). However, the urinary-antigen assay does not reliably detect non-serogroup 1 infections (Benson et al., 2000; Horn 2001). The direct fluorescence antibody test uses a mAb that detects all serogroups of L. pneumophila and thus has a broader spectrum (Fields et al., 2002). Culturing Legionella strains is still the gold standard method but suffers from only moderate sensitivity (DenBoer &Yzerman, 2004; Fields et al., 2002; Lück et al., 2006). However, it is still the prerequisite for epidemiological molecular typing studies to confirm or exclude a given environmental reservoir as the source of infection. According to data from the Robert Koch Institute (2006), in only half of the urinary-antigen-positive cases is an attempt made to cultivate legionellae from clinical samples, which significantly hampers the epidemiological investigation.

Genotyping by PFGE, AFLP or SBT is an important epidemiological tool, and many studies have demonstrated the usefulness of differentiating isolates of L. pneumophila in order to confirm or refute epidemiological associations (Fields et al., 2002; Fry et al., 1999; Gaia et al., 2005). All band-based techniques suffer from reduced inter-laboratory reproducibility. However, they are excellent when all strains are compared in one run as shown here. Recently, EWGLI adopted SBT as a simple, rapid and reproducible method for genotyping L. pneumophila serogroup 1 (Gaia et al., 2005). In the same study the applicability of SBT for epidemiologically related strains belonging to serogroup 6, 8 and 10 was demonstrated. To further test the utility of this method we applied SBT to strains belonging to serogroups 4, 5, 8 and 10. The present study was initiated as L. pneumophila serogroup 5/10 isolates became available from a patient and environmental samples, i.e. from the private home where the patient had spent the time after discharge from the hospital during the suspected incubation period (2 to 10 days). In our study, all strains belonging to serogroups 4, 5, 8 and 10 were typable. Since the number of strains belonging to these serogroups investigated is rather small, a final assessment of the index of discrimination is difficult to perform. Interestingly, strains with the allelic profiles 2, 10, 3, 28, 9 and 4 were found in this study as well as in a previous publication (Gaia et al., 2005). Taking together SBT represents a valuable tool for molecular epidemiological studies on L. pneumophila strains belonging to serogroups 2–15.

The patient had stayed in hospital in Berlin during the summertime. He felt ill 15 days later, thus exceeding the usual incubation period of 2 to 10 days, which is assumed for nosocomially acquired cases. However, it has been reported that Legionella can colonise the respiratory tract for a certain time and cause pneumonia later (Marrie et al., 1992). Occasionally, the incubation time might be longer than 10 days (DenBoer et al., 2002). Thus the hospital could not be ruled out as the source of the infection. Since the private home of the patient was also considered to be a possible source of the infection, it was inspected by the local health department and water samples were collected. The colony counts were 30 c.f.u. l^–1 on October 18th 2004 and 280 c.f.u. l^–1 3 weeks later. The results from serotyping and SBT methods revealed the potable water supply to be the source of infection (Table 1).

An association between the degree of Legionella contamination and the occurrence of legionellosis has been described (Best et al., 1983; Stout et al., 1992). However, fluctuations in the amount of legionellae in a give water supply system, as well as differences in the virulence properties, make it extremely difficult to predict a correlation between the number of legionellae in the environmental samples and the risk for humans. In the present case the number of Legionella colonies detected 3 and 5 weeks after the onset of the illness in the home of the patient were below the level considered to merit immediate action (<1000 c.f.u. l^–1) (EWGLI, 2005). On this basis, no remedial action and disinfection of the water supply was initiated. We do not know the concentration of the causative strain in the water supply at the moment of the infection. It might be speculated that the number of legionellae were higher during the hot summer. We also have no data concerning the virulence of the given serogroup 5/10 strain. Since the patient belonged to the high-risk group of transplant patients, a Legionella contamination <1000 c.f.u. l^–1 might be relevant, whereas such concentrations possess little or no risk for immunocompetent patients. Keeping in mind that patients with T-cell affecting immunosuppressive therapy have an increased risk of legionellosis, one might consider supplying these patients with Legionella-free water, either by decontaminating the home water system or by using terminal tap water filters.

The results of this investigation emphasize the importance of collaboration and exchange of data between local health authorities, clinics and the reference laboratory. Furthermore, it demonstrates the importance of isolating and typing Legionella, for further strengthening the epidemiological link of a specific environment to a patient.

We are grateful to Carolin Dix, Jutta Paasche and Kerstin Seeliger for technical assistance in culturing water samples and typing of legionellae. This study was supported by the Federal Ministry of Education and Research of Germany, Network of Competence in Medicine CAPNetz.