Detection of culturable and nonculturable Legionella species from hot water systems of public buildings in Japan

Authors


Akio Kimura, Department of Virology, Osaka Prefectural Institute of Public Health, 1-3-69 Nakamichi, Higashinari-ku, Osaka 537-0025, Japan. E-mail: akkimura@iph.pref.osaka.jp

Abstract

Aims:  To investigate the prevalence of culturable and nonculturable Legionella species in hot water systems of public buildings in Japan and assess the risk factors associated with Legionella contamination in hot water systems.

Methods and Results: Legionella species were detected by conventional culture and molecular methods in 130 water samples collected from 40 buildings. A total of 26 (20·0%) water samples from 17 (42·5%) buildings were positive by culture, qualitative PCR or both methods: Legionella pneumophila and Leg. anisa were detected in four samples by a culture method, whereas 23 samples were positive by qualitative PCR, with the presence of various Legionella species confirmed by sequencing. Of these 23 samples, bacterial counts were quantifiable in 21 by real-time PCR (from 1·7 × 105 to 2·6 × 1011 cells per litre). Phylogenetic analysis of amplified partial 16S rRNA gene showed close relations to various species of Legionella, including Leg. anisa and Leg. micdadei, all of which have been associated with respiratory diseases or increased antibody titres in human sera. Assessment of risk factors showed that turbidity, free chlorine concentration, iron concentration and heterotrophic plate count (HPC) were significantly associated with Legionella contamination (P < 0·05).

Conclusions:  Contamination of hot water systems of public buildings with culturable and nonculturable Legionella species may be a potential risk factor for Legionella infection in Japan. Adequate levels of chlorine, low levels of iron and HPC are important maintenance measures in the reduction of Legionella contamination in hot water systems.

Significance and Impact of the Study:  More than 40% of hot water systems in the Japanese public buildings examined were contaminated by not only culturable Leg. pneumophila and Leg. anisa but also by nonculturable pathogenic species. To our knowledge, this is the first report of both culturable and nonculturable Legionella contamination in hot water systems of public buildings in Japan.

Introduction

Legionellae, ubiquitous Gram-negative bacteria present in both natural and artificial aquatic environments (Brooks et al. 2004; Greig et al. 2004), are responsible for legionellosis, a group of related symptoms which include severe pneumonia and nonpneumonic Pontiac fever. While Legionella pneumophila is the major causative agent of these diseases, other Legionella species are also potential pathogens in humans (Fields et al. 2002). Numerous outbreaks of legionellosis have been reported worldwide (Fields et al. 2002). In Japan, more than 900 cases of infection by Legionella species from contaminated artificial whirlpool spas or natural hot springs have emerged in the past decade (Anonymous 2000; Nakamura et al. 2003; Kura et al. 2006); in particular, a major outbreak in 2002 originating from a newly opened hot spring spa involved 295 people, including seven deaths (Okada et al. 2005). Following this, Legionella infection became an important public health concern in Japan.

Infection with Legionella species occurs through inhalation of an aerosol generated from a contaminated aqueous environment or by aspiration of contaminated environmental water (Fields et al. 2002). Legionella species are commonly detected in various types of artificial water environments such as cooling towers, evaporative condensers, whirlpool spas, showers and hot water tanks (Bornstein et al. 1986; Brown et al. 1999; Fields et al. 2001; Rota et al. 2005; Mouchtouri et al. 2007). Contamination of hot water systems in public buildings has been found to be an important risk factor for Legionella infection (Hutchinson et al. 1996; Anaissie et al. 2002). Several studies in European countries have identified a high prevalence of Legionella species in hot water systems in hotels, hospitals and apartments (Zacheus and Martikainen 1994; Zietz et al. 2001; Borella et al. 2005; Leoni et al. 2005; Mouchtouri et al. 2007). While these reports suggest the importance of survey and control of Legionella contamination in hot water systems, little is known about contamination of hot water systems in public buildings in Japan.

Although the detection of Legionella species from environmental water by culture techniques has been commonly used as an objective standard, it does have a number of limitations. One is that the slow growth rate of Legionella makes a 3- to 10-day incubation essential for culture (Fields et al. 2002). A second is the inhibition of the growth of Legionella species by the presence of other micro-organisms (Hussong et al. 1987; Ng et al. 1997). Finally, a viable nonculturable form of Legionella species which cannot be detected by culture methods has been identified (Hussong et al. 1987). In response, many researchers have adopted molecular methods to detect various Legionella species in environmental samples (Fields et al. 2002). In particular, 16S rRNA, 5S rRNA or macrophage infectivity potentiator gene-based PCR detection methods have markedly higher sensitivity than standard culture methods (Mahbubani et al. 1990; Maiwald et al. 1994; Miyamoto et al. 1997b; Villari et al. 1998) and produce a result in only 1 day. In addition, nucleotide sequencing of the PCR product is even now commonly used to identify species (Cloud et al. 2000; Wullings and van der Kooij 2006). Further, recent development of quantitative real-time PCR for Legionella detection provides further information on the risk of Legionella infection (Joly et al. 2006; Yaradou et al. 2006).

Here, we investigated the prevalence of both culturable and nonculturable Legionella species in hot water systems of public buildings in Japan using culture, and qualitative and quantitative PCR methods, and assessed risk factors associated with Legionella contamination in hot water systems.

Materials and methods

Sample collection

Water samples were collected from the hot water systems of 40 public buildings located in Osaka Prefecture, Japan, including 21 hotels, 7 offices, 5 schools, 4 stores and 3 assembly halls. The 40 buildings studied had three different systems for producing warm water: 9 buildings had a central storage tank, 26 had continuous re-circulation and 5 had local instantaneous production by heat exchanger at the point of use (local instantaneous type). Buildings with a central storage tank have a one-way distribution system for hot water from a storage water heater to the local tap with no return (central storage type), while those with continuous re-circulation have a storage water heater with water pumped around the building in a continuous circulating loop (central circulation type) (Table 3). Initially, three kinds of water sample were collected from the 40 buildings; one from original source water distributed from the water treatment plant to evaluate the quality of the source water, and one sample each from a tap before and after the temperature had stabilized to analyse Legionella contamination in the water distributing pipes. In addition to these 120 samples, water from storage water heaters was collected from ten buildings equipped with a tap drain. These four kinds of water sample were named original source, tap 1, tap 2 and storage water heater samples, respectively. A total of 130 water samples of 200 ml each were collected in sterile bottles containing sodium thiosulfate at a final concentration of 0·01%.

Table 3.   Number of Legionella-positive and -negative cases by hot water system type in 40 public buildings in Japan
System type*Positive cases (n = 17)Negative cases (n = 23)Total
  1. *Details of system type are indicated in the Materials and methods.

Central storage639
Central circulation101626
Local instantaneous145

Physical, chemical and microbiological analysis of hot water

Sample temperature and concentration of free chlorine were measured at the time of sampling. Free chlorine concentration was measured by the diethyl-p-phenylenediamine (DPD) method (DPD test Wako; Wako Pure Chemical Industries Ltd, Osaka, Japan). Water samples were delivered to the laboratory immediately after sampling. Microbiological analyses were performed on the day of collection, and physical and chemical analyses were performed within 24 h after collection. The turbidity and colour unit of each water sample were measured with a Water Analyzer-2000 (Nippon Denshoku Co., Tokyo, Japan) by standard methods. Concentrations of iron, copper, lead and zinc were determined using an Atomic Absorption Flame Emission Spectrophotometer AA-890 (Nippon Jarrel Ash Co. Ltd, Tokyo, Japan). The aerobic heterotrophic plate count (HPC) was determined on R2A agar (BD, Franklin Lake, NJ, USA) for 7 days at 30°C. Additionally, total organic carbon (TOC) was measured with an infrared spectrometer (Shimadzu Co. Ltd, Kyoto, Japan). Information on the hot water distribution type, age of the system and frequency of cleaning was obtained from the respective building administrator.

Isolation and characterization of Legionella species by conventional culture

Two hundred millilitres of each water sample was concentrated by filtration through a 0·2-μm pore size polycarbonate filter (Advantec Tokyo Co. Ltd, Tokyo, Japan). The membrane was then immersed in 4 ml of sterile deionized water, shaken vigorously 50 times and vortexed for 2 min. One millilitre of the suspension was stored at −20°C for DNA extraction. The remaining suspension was heated in a water bath at 50°C for 30 min, followed by inoculation of a 200-μl sample onto two Wadowsky-Yee-Okuda agar plates (100 μl per plate) supplemented with alpha-ketoglutarate (WYO-alpha plate) (Eiken Chemical Co. Ltd, Tokyo, Japan). After incubation for 5–7 days at 37°C, from 1 to 50 colonies showing characteristics of Legionella species were selected and cultured on blood agar plates (Nikken Seibutsu Co. Ltd, Tokyo, Japan) and also on plates of buffered charcoal–yeast extract agar with alpha-ketoglutarate (BCYE-alpha; Eiken Chemical Co. Ltd). After 3 days’ culture at 37°C, isolates that did not grow on blood agar plates but did grow on BCYE-alpha agar plates were examined by Gram stain to confirm the presence of Legionella species. The colonies were observed under UV light, and then from 1 to 5 colonies were randomly selected for the latex agglutination test (Kanto Chemical Co., Tokyo, Japan) and immune serum agglutination test (Denka Seiken Co. Ltd., Tokyo, Japan) to identify the serogroup of Leg. pneumophila, Leg. bozemanii, Leg. dumoffii, Leg. gormanii and Leg. micdadei. In addition, a DNA–DNA hybridization test which used ELISA plates coated with 16S rRNA gene of 25 Legionella-type strains was performed according to the manufacturer’s instructions (Kyokuto Pharmaceutical Industrial Co. Ltd, Tokyo, Japan).

DNA extraction

DNA was extracted using the simplified alkaline DNA preparation method (Beige et al. 1995). In brief, 1 ml of a 50-fold concentrated sample was centrifuged at 13 000 g for 10 min at 4°C, and the supernatant was discarded to a volume of 40 μl. A suspension of the pellet was then mixed with 50 μl of 50 mmol l−1 NaOH by vortexing and then boiled for 15 min. After rapid cooling, the material was neutralized with 8 μl of 1 mol l−1 Tris-HCl (pH 7·0), centrifuged at 13 000 g for 10 min at 4°C and the supernatant was stored at −20°C until use.

Qualitative and quantitative PCR

For qualitative PCR, the Legionella species-specific primers P1·2 (5′-AGGGTTGATAGGTTAAGAGC) and cp3·2 (5′-CCAACAGCTAGTTGACATCG) were used to amplify a 386 bp product of a partial 16S rRNA gene according to Jonas et al. (1995). Five microlitres of extracted DNA was used in a 50 μl reaction mixture containing 5 μl of 10 × PCR buffer [100 mmol l−1 KCL, 20 mmol l−1 MgCl2, 20 mmol l−1 Tris-HCl (pH 8·0)], 4 μl of 2·5 mmol l−1 dNTP mixture, 0·5 μl each of 100 mmol l−1 primers and 0·25 μl of 5 U μl−1 Ex Taq DNA polymerase (Takara Bio Co., Shiga, Japan). Thermal cycling was performed with a PCR Thermal Cycler MP (Takara Bio Co.). Cycling conditions began with an initial incubation at 95°C for 5 min, followed by 40 cycles of denaturation at 94°C for 1 min, annealing at 57°C for 1·5 min and extension at 72°C for 1 min. Finally, incomplete PCR products were extended for 10 min at 72°C. Genomic DNA from Leg. pneumophila (ATCC33152) was used as a positive control. Amplified DNA was detected by 1% agarose gel electrophoresis and the DNA was visualized by ethidium bromide staining. Additionally, cell numbers of nonculturable Legionella species were investigated by quantitative real-time PCR with a Cycleave PCR Legionella (5S rRNA) Detection Kit (Takara Bio Co.). For external standards, genomic DNA was isolated from Leg. pneumophila (ATCC33152) as described above and used. Cell numbers of bacteria of the initial purified DNA solution were calculated according to the method of Joly et al. (2006). PCR reactions of duplicate standards, positive control, negative control and samples were performed using an ABI Prism 7000 Real Time PCR System (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s instructions. In brief, 25 μl of reaction mixture containing 12·5 μl of 2 × Cycleave Reaction Mixture, 5 μl of 5S primers/Probe Mix, 2·5 μl of dH2O and 5 μl of each DNA sample was subjected to PCR reaction. Cycling conditions began with an initial incubation at 95°C for 10 s, followed by 45 cycles of denaturation at 95°C for 5 s, annealing at 55°C for 10 s and extension at 72°C for 20 s. The cell numbers of each sample were automatically calculated according to the threshold cycle values using the ABI Prism 7000 sds Software (Applied Biosystems).

Nucleotide sequencing and phylogenetic analysis

The qualitative 16S rRNA PCR products were purified using a QIAamp PCR purification kit (Qiagen K.K., Tokyo, Japan) and directly sequenced in both directions using primers P1·2 and cp3·2 with a BigDye Terminator v1·1 Cycle Sequencing kit (Applied Biosystems). Homology search was performed using the Blast software at the NCBI home page (http://www.ncbi.nlm.nih.gov/). The sequence data obtained in this study were aligned with the sequences of 57 Legionella species or types deposited in the GenBank using the ClustalW software (http://www.nig.ac.jp) and the phylogenetic tree was then displayed by the neighbour-joining (NJ) method using the Njplot programme.

Nucleotide sequence accession numbers

The nucleotide sequences determined in this study have been submitted to the DNA Data Bank Japan (DDBJ) database and assigned the accession numbers AB246278 to AB246297.

Statistical analyses

All calculations were performed using spss 12·0j software (SPSS Japan Inc., Tokyo, Japan). Comparisons of median values, mean values and percentages between two groups were conducted using the Mann–Whitney’s U-test, t-test and Fisher’s exact test, respectively.

Results

Detection of Legionella species by culture and PCR

Legionella species were detected from 26 samples (20·0%) collected from 17 buildings (42·5%) by culture, PCR or both (Table 1). Specifically, 4 of the 26 samples were positive by culture (3·1%), 23 by PCR (17·7%) and 1 by both culture and PCR methods. Despite the limited number of school and store buildings investigated, positive rates were 60·0% and 50·0%, respectively. Among the 130 samples, furthermore, those collected from schools also showed a higher positive rate (46·7%) than those from the other types of building. Nevertheless, no significant differences were observed between either buildings or samples as calculated by Fisher’s exact test (P = 0·947 and P = 0·138, respectively). Positive rates of original source, tap 1, tap 2 and storage water heater samples were 15·0% (6/40), 22·5% (9/40), 17·5% (7/40) and 40·0% (4/10), respectively. Although storage water heater samples were shown to be more positive, no significant differences were observed among the four kinds of sample by Fisher’s exact test (P = 0·342).

Table 1.   Positive samples of Legionella species in hot water systems of 40 public buildings in Japan
 No. of positive samples/total no. (%)
Among buildingsAmong samples
Facility type17/40 (42·5)26/130 (20·0)
 Hotel8/21 (38·1)13/70 (18·6)
 Office3/7 (42·9)3/22 (13·6)
 School3/5 (60·0)7/15 (46·7)
 Store2/4 (50·0)2/13 (15·4)
 Assembly hall1/3 (33·3)1/10 (10·0)

The Legionella species detected from 17 buildings by culture, PCR or both are shown in Table 2. Five Legionella isolates were detected in four samples from three buildings (hotels 5 and 8 and office 1) by culture. A high concentration (8·3 × 103 CFU l−1) of mixed Legionella species of Leg. pneumophila serogroup (SG) 1 and Leg. anisa were detected from an original source sample of one hotel (hotel 5). From the same hotel, 1·4 × 103 CFU l−1 of Leg. pneumophila SG 4 was detected from a storage water heater sample, while 3·3 × 102 CFU l−1 of Leg. pneumophila SG 1 was detected from a storage water heater sample of another hotel (hotel 8) and 1·8 × 102 CFU l−1 of Leg. anisa was detected from a tap 1 sample of an office (office 1). Legionella species were detected in 23 samples from 16 buildings by qualitative PCR. Twenty PCR products from 23 PCR-positive samples were successfully sequenced from both directions and designated with an OIPH isolation number. Although complete sequences were not obtained from the remaining three products, they were determined as Legionella species by BLAST homology search. Using quantitative real-time PCR, Legionella species ranging from 1·7 × 105 to 2·6 × 1011 cells per litre were detected from 21 of 23 samples. In two samples (tap 1 from hotel 6 and storage water heater from hotel 7), PCR reactions were not strong enough to quantify the number of cells.

Table 2. Legionella sp. detected by culture and molecular methods in 17 positive buildings
BuildingWater sourceLegionella spp. detected by cultureLegionella sp. detected by PCR
CFU l−1Species and serogroupDirect sequencingCells per litre
  1. *Not sequenced completely from both sides, but confirmed as Legionella species by homology search.

  2. †Accession numbers deposited in GenBank are given in parentheses.

  3. ‡SG, serogroup.

Assembly hall 1Strorage water heaterLegionella sp.*7·1 × 106
Hotel 1Tap 2Legionella sp.*2·6 × 1011
Hotel 2Tap 1OIPH-6b (AB246278)†8·3 × 108
Hotel 3Tap 1OIPH-16b (AB246281)2·7 × 108
Hotel 4Tap 1OIPH-23b (AB246284)7·3 × 107
Tap 2OIPH-23c (AB246285)3·6 × 108
Hotel 5Original source8·3 × 103Leg. pneumophila SG1 and Leg. anisaOIPH-25a (AB246286)9·3 × 109
Storage water heater1·4 × 103Leg. pneumophila SG4
Tap 1OIPH-25b (AB246287)3·2 × 107
Hotel 6Tap 1OIPH-26a (AB246288)Not quantified
Hotel 7Storage water heaterLegionella sp.*Not quantified
Hotel 8Tap 1OIPH-30b (AB246289)1·9 × 106
Tap 2OIPH-30c (AB246290)3·3 × 106
Storage water heater3·3 × 102Leg. pneumophila SG1
Office 1Tap 11·8 × 102Leg. anisa
Office 2Storage water heaterOIPH-31a (AB246291)1·6 × 107
Office 3Original sourceOIPH-34a (AB246292)5·3 × 106
School 1Tap 1OIPH-18b (AB246282)1·5 × 108
Tap 2OIPH-18c (AB246283)3·5 × 109
School 2Original sourceOIPH-38a (AB246293)5·1 × 108
Tap 1OIPH-38b (AB246294)7·8 × 108
Tap 2OIPH-38c (AB246295)8·4 × 108
School 3Tap 1OIPH-39b (AB246296)3·7 × 107
Tap 2OIPH-39c (AB246297)3·5 × 107
Store 1Tap 2OIPH-7c (AB246279)1·7 × 105
Store 2Original sourceOIPH-8a (AB246280)2·5 × 106

Phylogenetic analysis

Genetic relationships among the 20 partial sequences of 16S rRNA gene obtained in this study and the 57 sequences of Legionella in the database were investigated using a phylogenetic tree constructed by the NJ method (Fig. 1). Results showed that the 20 partial sequences were dispersed over several clusters of Legionella species or Legionella-like amoebal pathogen (LLAP) types. Four sequences were identical to Leg. anisa as confirmed by the homology search described above. Six (OIPH-8a, 30b, 30c, 38a, 38b and 38c) of the 16 remaining sequences were found to be included in the clusters of Legionella species known to have associations with respiratory diseases or increased antibody titres in human sera (Fields et al. 2002). The sequence of one isolate, OIPH-26a, was located in a distinct branch from clusters of other Legionella species.

Figure 1.

 NJ phylogenetic relationship between the sequences of partial 16S rRNA gene in 20 isolates obtained in this study and 52 species or strains of Legionella deposited in GenBank. Escherichia coli (J01859) was used as an outgroup and the root was removed from the figure. The GenBank accession numbers are given in parentheses. Reference species with an underlined strain name are type strains of Legionella species. References with an asterisk have been reported to be associated with human diseases. Branch length is proportional to the calculated genetic distance (scale shown).

Characteristics associated with Legionella contamination

Regarding the three different systems of producing hot water, buildings with a central storage system were shown to be more positive for Legionella (66·7%; 6/9) than those with central circulation system (38·5%; 10/26) or those that had a local instantaneous production system of hot water at the point of use (20·0%; 1/5) (Table 3). However, these differences were not statistically significant by Fisher’s exact test (P = 0·245). The fact that the positive buildings studied were cleaned at a similar frequency to the negative buildings (positive: 1·3 ± 2·9 times per year and negative: 1·1 ± 2·6 times per year) and were of more or less the same building age (positive: 14·4 ± 12·7 years old and negative: 12·9 ± 8·7 years old) did not allow the drawing of any conclusions about the true influence of these factors on the presence of Legionella. Among the physical, chemical and microbiological characteristics analysed, in contrast, the t-test identified significant associations between contamination and turbidity, free chlorine concentration, iron concentration and HPC (P = 0·012, P = 0·026, P = 0·015 and P = 0·036, respectively), with Legionella-positive samples showing significantly lower free chlorine concentrations and significantly higher turbidity, iron concentration and HPC (Table 4).

Table 4.   Physical, chemical and microbiologic characteristics associated with Legionella contamination in hot water samples
CharacteristicValue*Significant†
Legionella positive (n = 26)Legionella negative (n = 104)Positive vs Negative
  1. *Mean ± SD.

  2. †Significance calculated by the t test, P < 0·05.

Temperature (°C)43·7 ± 14·744·2 ± 17·1No
pH7·46 ± 0·377·35 ± 0·40No
Colour (CU)6·8 ± 25·01·8 ± 7·8No
Turbidity (NTU)2·2 ± 6·40·4 ± 1·4Yes
Free chlorine (mg l−1)0·09 ± 0·130·19 ± 0·22Yes
Lead (mg l−1)0·01 ± 0·040·003 ± 0·01No
Zinc (mg l−1)0·05 ± 0·090·16 ± 1·19No
Copper (mg l−1)0·27 ± 0·520·13 ± 0·23No
Iron (mg l−1)1·43 ± 5·370·09 ± 0·41Yes
TOC (mg l−1)1·06 ± 1·082·95 ± 20·76No
HPC (CFU ml−1)2·8 ± 8·8x1040·6 ± 2·9 × 104Yes

Hotel 5, in which a higher number of culturable Leg. pneumophila and Leg. anisa were detected in samples collected during September, as well as DNA of Legionella species, was only open for use during the summer period (from June to September). Free chlorine concentration in all four samples from this hotel was 0·00 mg l−1. Moreover, the heater tank water from this hotel had a higher iron concentration (8·93 mg l−1) than samples from other buildings.

Discussion

Reports by European groups have revealed strikingly high detection rates of Leg. pneumophila (20–60%) by bacterial culture of water samples from hot water systems of buildings in Finland, Greek, Germany and Italy (Zacheus and Martikainen 1994; Zietz et al. 2001; Borella et al. 2005; Leoni et al. 2005; Mouchtouri et al. 2007). Although we used a sample volume of 200 ml, five times smaller than the European standard method (ISO11731), which requires analysis in one litre of water, the detection rate of Leg. pneumophila and Leg. anisa by culture in our study was only 3·1% (4/130). This result is comparable to that of Furuhata et al. (8·8%) for Leg. pneumophila in water samples from hot water systems of office buildings in Japan (Furuhata et al. 1994). However, the mean count of cultured Legionella species in our study was 2·55 × 103 CFU l−1, similar to the number reported by the Italian group (mean count of Legionella species: 1·9 × 103 CFU l−1) (Borella et al. 2005). Furthermore, Legionella species obtained by bacterial culture in our study were Leg. pneumophila SG1, SG4 and Leg. anisa; among these, Leg. pneumophila SG1 is the predominant species detected from legionellosis patients worldwide while Leg. anisa is recognized as one of the pathogens responsible for Pontiac fever (Fenstersheib et al. 1990; Fields et al. 2002).

To overcome the limitations of culture methods, we additionally used PCR to detect Legionella species. A number of reports have described higher detection rates of Legionella species from environmental samples by PCR-based methods than by culture (Catalan et al. 1994; Miyamoto et al. 1997b; Ng et al. 1997; Bates et al. 2000; Buchbinder et al. 2002), as we also observed in the present study. However, they were several incongruent findings between qualitative PCR and culture results. In the samples from hotels 5, 8 and office 1, Leg. pneumophila and Leg. anisa were detected by culture but not by PCR. Theoretically at least, PCR should provide substantially greater sensitivity than culture in the detection of Legionella (Villari et al. 1998). Nevertheless, PCR is occasionally inhibited by the presence of unknown materials in environmental water samples (Maiwald et al. 1994; Miyamoto et al. 1997a). We assume that such inhibition occurred in these three hot water samples (hotels 5, 8, office 1). In contrast, Leg. anisa was detected from hotels 2, 3 and school 1 by PCR only. Samples in the present study were preheated for culture, and we acknowledge the possibility that the isolation of heat-sensitive Legionella species like Leg. anisa and others may have resulted in their failure to grow. These results suggest that the investigation of Legionella contamination of hot water systems by culture should be supplemented by PCR-based methods.

Additional quantitative real-time PCR to investigate cell numbers of nonculturable Legionella species detected by qualitative PCR was successful in 21 of 23 samples. The concentration of Legionella species showed wide variation, from 1·7 × 105 to 2·6 × 1011 l−1. Joly et al. (2006) and Yaradouet al. (2006) reported higher concentrations of bacteria by real-time PCR than culture because real-time PCR detects both living and dead cells. We observed the same phenomenon in our study. Although dead cells were possibly included in our samples, our real-time PCR results suggested that the hot water samples were contaminated by notable numbers of nonculturable Legionella species. It is not presently known how many cells per litre of nonculturable Legionella species would represent a threat to human health. However, two reports have suggested that concentrations above 104 to 105 CFU l−1of culturable Legionella may increase the risk of human infection (Meenhorst et al. 1985; Patterson et al. 1994). Further investigation of the correlation between the number of nonculturable Legionella and human infection appears necessary.

Our Blast search of 20 sequences obtained from 23 PCR products and a phylogenetic analysis identified the presence of various non-Leg. pneumophila species in hot water systems of public buildings. Except for the sequence OIPH 26a, similarities between the 19 sequences and other Legionella species ranged from 96% to 100%. The sequence OIPH-26a formed a distinct branch of the tree and showed only 91% identity to the others, and thus possibly represents an as yet unidentified species of Legionella, or even a species not belonging to the genus Legionella. Four sequences, OIPH-6b, 16b, 18b and 18c, showed 100% identity with Leg. anisa, one of the pathogens of Pontiac fever (Fenstersheib et al. 1990). Among the 15 remaining sequences, 6 showed a phylogenetic relationship with six Legionella species known to be associated with respiratory diseases or increased antibody titres in human sera (McNally et al. 2000; Fields et al. 2002). Sequences OIPH-8a, 38a, 38b and 38c showed a close relation with Leg. drozanskii and Leg. birminghamensis. Of these, Leg. drozanskii (formerly called LLAP-1) was first isolated with free-living amoebae from the tank of a potable water well in 1981 (Adeleke et al. 2001). Several researchers have reported that Leg. drozanskii is a pathogenic agent of community-acquired pneumonia (Adeleke et al. 1996; McNally et al. 2000; Marrie et al. 2001). Further, Leg. birminghamensis was isolated from a lung biopsy specimen from a cardiac transplant recipient with no symptoms of pneumonia (Wilkinson et al. 1987) and was later found in environmental water in Australia (Wilkinson et al. 1990). Sequence OIPH-30b and 30c were included in the same cluster as Leg. micdadei and Leg. maceachernii, and occupied the same clade as Leg. oakridgensis and Leg. lansingensis. These four species are also recognized as etiologic agents of Legionella infections (Tang et al. 1985; Wilkinson et al. 1985; Thacker et al. 1992; Thomas et al. 1992; Yu et al. 2002; Medarov et al. 2004); in particular, Leg. micdadei is the third-most common agent following Leg. pneumophila and Leg. anisa, accounting for approx. 2–5% of reported cases in the United States (Benin et al. 2002; Muder and Yu 2002). At least three cases of pneumonia caused by Leg. micdadei have also been reported in Japan (Miyashita et al. 1996; Matsubara et al. 1998; Takiguchi et al. 1999).

We also investigated the characteristics of hot water systems and buildings associated with Legionella contamination. Although statistical analysis showed no significant differences between the three systems for producing hot water and presence of Legionella, buildings with a central storage system were shown to be more contaminated than those with a central circulation system or those with a local instantaneous system that had instantaneous production of hot water at the point of use. This finding is consistent with a report that warm water stagnation in water distributing pipes and storage tanks significantly promoted Legionella multiplication (Ciesielski et al. 1984). Indeed, one hotel (hotel 5) using a central storage system was found to be highly contaminated by Leg. pneumophila SG1 and SG4 and Leg. anisa. This hotel was open only during the summer period, an interesting finding that supports a recent report by Mouchtouri et al. that seasonal hotel operation is a risk factor for Legionella contamination of hot water systems (Mouchtouri et al. 2007).

We evaluated 11 physical, chemical and microbiological characteristics of hot water samples for their association with Legionella contamination. Statistical analysis showed a significant association for turbidity and the number of HPC, reflecting the results of Stout et al. (1985), who demonstrated that the combination of sediment and environmental bacteria in water act synergistically to increase Legionella contamination in water distribution systems. We also found that free chlorine concentration was a major factor affecting contamination, consistent with previous observations (Fujimura et al. 2006; Mouchtouri et al. 2007). Additionally, among the four metals examined, only iron concentration significantly correlated with Legionella contamination. This promotive activity of iron agrees well with the findings of States et al. (1985), who showed that a high concentration of iron (>1·0 mg l−1) in hot water tanks enhanced Legionella growth 2- to 100-fold. In contrast, we did not observe the inhibitory effect of copper on contamination or promotive activity of zinc described by other groups (Rogers et al. 1994; Borella et al. 2004; Leoni et al. 2005).

Our study found that more than 40% of hot water distribution systems in the Japanese public buildings examined were contaminated not only by culturable Leg. pneumophila and Leg. anisa but also by notable numbers of nonculturable pathogenic species such as Leg. drozanskii. These results suggest that the hot water of public buildings is a potential risk factor for Legionella infection in Japan. Water stagnation in hot water systems might cause a reduction in free chlorine concentration and an increase in iron concentration and HPC, factors which strongly enhance the growth of Legionella. Prevention of water stagnation in hot water systems, especially in public buildings with lower utilization frequency, appears to play an important role in reducing Legionella contamination and preventing legionellosis.

Acknowledgements

The authors gratefully acknowledge Drs I. Watanabe and S. Kumagai of the Osaka Prefectural Institute of Public Health for their helpful discussions. The authors are also grateful to Dr G. Harris of DMC Corporation for his review of the manuscript. This work was supported in part by the New Bioscience Research Project of Osaka Prefectural Institute of Public Health, 2005 to A.E. and Grants-in-Aid for the Program of the Founding Research Center for Emerging and Reemerging Infectious Diseases from the Ministry of Education, Culture, Sports, Science and Technology, Japan, to Y.S.

Ancillary