Legionellosis is most often attributed to inhalation of contaminated aerosols from manmade water systems or, in the case of Legionella longbeachae, to inhalation of contaminated potting mixes or composts [1–3]. From 2003 to 2004, active population-based surveillance for atypical bacterial respiratory pathogens was performed in the province of Sa Kaeo, Thailand, to establish incidence [4]. Immunologic testing performed on sera of suspect legionellosis cases found 20/397 (5%) adult cases exhibited a fourfold rise in titer to L. longbeachae by the indirect immunofluorescence assay. L. longbeachae infection followed the typical legionellosis demographic, with incidences highest among adults over 34 years of age, increasing steadily with age, and infection peaking through September and October. As rice farming is quite prevalent in this area of Thailand, the extensive exposure of farm workers to moist soil would be a plausible means for transmitting L. longbeachae [5]. These findings, coupled with limited access to composted or processed soil, led us to investigate the presence of legionellae, specifically L. longbeachae, in indigenous soils and exposure due to agricultural practices in Sa Kaeo.

In 2009, thirty-nine wet soil samples were collected from eight rural sites within Sa Kaeo province where prior laboratory-confirmed cases of legionellosis were identified and were shipped to the Centers for Disease Control and Prevention in Atlanta for culture [4]. Cultures were performed using a modification of the procedure used to culture legionellae from water [6]. Five grams of soil were weighed, and mixed with 50 mL sterile dH₂O. The filtrate was then strained through sterile gauze into a 50 mL conical tube. A 1 : 5 dilution was made of the filtrate, and 500 μL of the diluted filtrate was acid treated for fifteen minutes with an equal part of KCl/HCl acid (pH 2.3). After acid treatment, 50 μL was cultured on one plate of buffered charcoal yeast extract (BCYE) supplemented with 160 mg/L cycloheximide, two plates additionally supplemented with 100,000 U/L polymyxin B and 5 mg/L vancomycin (PCV), and two plates further supplemented with 2 g/L glycine (GPCV). Plates were examined at four, seven, and fourteen days to check for the presence of Legionella. Any samples overgrown with non-Legionella organisms were retreated with acid in fifteen-minute increments and recultured, until nonrelevant organisms were reduced enough to allow for identification of legionellae. Colonies displaying Legionella morphology were checked for cysteine auxotrophy on BCYE biplates with and without L-cysteine. Colonies requiring L-cysteine for growth were considered presumptive Legionella species.

Monoclonal antibody (MAb) testing was used to identify L. pneumophila serogroup 1 (Lp1) as previously described [7, 8]. Other L. pneumophila serogroups were determined using direct fluorescent antibody staining performed on dried, formalin-fixed suspensions using L. pneumophila serogroup-specific fluorescein isothiocyanate labeled antibody [9]. Select non-L. pneumophila isolates were tested by slide agglutination as described [10].

All non-Lp1 isolates were identified by sequencing the macrophage infectivity potentiator (mip) gene [11]. The resulting sequence was then queried against the National Center for Biotechnology Information (NCBI) GenBank nucleotide database using NCBI's alignment tool Basic Local Alignment Search Tool (BLAST) [12, 13]. Sequences with a minimum of 95% identity to a known Legionella species were assigned. Those sequences with less than 95% identity to sequences of submitted Legionella species were considered a potential novel Legionella organism. All L. pneumophila isolates were genotyped using the sequence-based typing (SBT) epidemiological typing scheme established by the European Working Group for Legionella Infections (EWGLI) as previously described [14, 15]. eBURST analysis was performed to observe the relatedness of the soil isolates to isolates from other countries previously submitted to the EWGLI SBT database (http://eburst.mlst.net/).

Twenty-two (56%) of the 39 soil samples received were positive for Legionella. In total, we obtained 115 isolates, 87 of which were known species, 25 potential novel species, and 3 isolates that were not typable by mip sequencing (Table 1). Nine of the species identified have been reported in association with human disease and represented 70% (80/115) of isolates obtained [16]. Isolates of L. birminghamensis, L. lansingensis, L. pneumophila, L. rubrilucens, and L. sainthelensi were found exclusively in soil samples from the personal residence of suspect cases. Isolates of L. bozemanae, L. erythra, L. gormanii, L. quateirensis, L. quinlivanii, and the potentially novel species identified as Legionella sp. ST24644 (NCBI accession number GU083740; isolated from a cooling tower in Thailand) were found only in soil samples from the workplace of suspect cases. No isolates of L. longbeachae were identified. The highest diversity of species (n = 7) was found in soils taken from the outdoor area for washing at the personal residence and within rice fields. L. pneumophila serogroup 1 was identified in the environment; however, none were positive for monoclonal antibody MAb2, the phenotypic subtype responsible for 65–100% of Lp1-caused legionellosis [17].

Five of 112 isolates tested reacted strongly with L. longbeachae serogroup 1 antisera by direct fluorescent antibody testing. Sequence analysis, however, indicated these isolates were L. bozemanae (n = 2), Legionella sp. ST24644 (n = 2), and a novel nonfluorescent Legionella species (n = 1). Although the species of the three mip untypable isolates remain unknown, they were identified as Legionella spp. using a pan-Legionella real-time PCR assay [18]. Slide agglutination testing found these isolates were not L. geestiana, a species in which the mip gene is known to not amplify with the primers used [11].

Eight allelic profiles were identified by SBT analysis, seven of which were novel profiles (as of November 2, 2010). Two of the seven novel sequence types identified were found to be related to isolates from community-acquired and nosocomial cases through eBURST analysis (as of December 14, 2010; data not shown). The eighth allelic profile matched the existing ST260 which has been associated with community-acquired cases.

Although a primary goal in this study was to identify the environmental source(s) of the suspect causative agent L. longbeachae in the 2003/4 suspect pneumonia cases, we were unable to recover this species from these 39 soil samples. Interestingly, the water-saturated soils collected did support the growth of many other Legionella species not previously associated with this indigenous soil type.

These findings suggest that the L. longbeachae seropositive-pneumonia cases documented in 2003/4 were either (i) due to L. longbeachae that were no longer or never present in these environments or were not found in our limited number of samples as mentioned previously; (ii) due to a serologically cross-reactive strain of legionellae; (iii) serological cross-reaction with another pathogen; or (iv) false-positives. Cross-reactivity is an inherent problem in the use of serology for identification, and the findings of this study further highlight the need for a molecular-based method for identification [19–22]. When possible, a clinical isolate is preferred for diagnosis of suspect non-L. pneumophila cases because of insufficient specificity and sensitivity of non-pneumophila legionellae antisera [23].

For future studies, we wish to conduct prospective surveillance in Sa Kaeo and obtain legionellae isolates from patient samples. A clinical isolate would allow for sequence identification of the etiologic agent and detection of any novel species responsible for respiratory disease in tropical countries such as Thailand. Concurrent environmental sampling would allow identification of settings capable of transmitting these pathogens to susceptible hosts. Although L. longbeachae is widely accepted as the predominant pathogenic, soil-dwelling species of Legionella, the presence of legionellae in soils has been limited to composts and manufactured potting mixes. The findings in this study indicate native soils are a likely reservoir of multiple Legionella species in regions with a geography and climate similar to Sa Kaeo and may play a role in human disease.