Development of a new CARD-FISH protocol for quantification of Legionella pneumophila and its application in two hospital cooling towers


Alexander K.T. Kirschner, Medical University Vienna, Institute for Hygiene and Applied Immunology, Water Hygiene, Kinderspitalgasse 15, 1095 Vienna, Austria. E-mail:


Aims:  Open cooling towers are frequent sources of infections with Legionella pneumophila. The gold standard for the detection of Leg. pneumophila is based on cultivation lasting up to 10 days and detecting only culturable cells. Alternative fluorescence in situ hybridization (FISH) protocols have been proposed, but they result in faint fluorescence signals and lack specificity because of cross-hybridization with other Legionella species. Our aim was thus to develop a new FISH protocol for rapid and specific detection of Leg. pneumophila in water samples.

Methods and Results:  A novel catalysed reporter deposition FISH (CARD-FISH) protocol for the detection of Leg. pneumophila was developed, which significantly enhanced signal intensity as well as specificity of the probe through the use of a novel competitor probe. The developed protocol was compared with the culture method for monitoring the seasonal development of culturable and nonculturable Leg. pneumophila in two hospital cooling tower systems. Seasonal fluctuations of Leg. pneumophila concentrations detected via CARD-FISH were related to the development of the total bacterial community in both cooling towers, with temperature and biocide as the main factors controlling this development.

Conclusions:  Our results clearly showed that the majority of the Leg. pneumophila cells were in a nonculturable state. Thus, detection of Leg. pneumophila with culture methods may underestimate the total numbers of Leg. pneumophila present.

Significance and Impact of the Study:  Rapid, sensitive and specific detection and quantification of Leg. pneumophila in water systems is prerequisite for reliable risk estimation. The new protocol significantly improves current methodology and can be used to monitor and screen for Leg. pneumophila concentrations in cooling towers or other water systems.


Legionellae are a frequent cause of nosocomial infections (Stout et al. 2007). However, most Legionella infections are community-acquired without identification of the source (Pedro-Botet and Sabria 2005). The overwhelming majority of cases of Legionnaires’ disease are caused by Legionella pneumophila (Joseph and Ricketts 2010). Legionellae are nearly exclusively conveyed via contaminated aerosols as they are produced by showers, whirl pools, air conditioners or humidifiers. In recent years, open cooling towers have been increasingly identified as sources of aerosols contaminated with legionellae (Ferre et al. 2009; Mouchtouri et al. 2010). Cooling towers are mostly situated on top of large buildings including hospitals, from where the bacteria can be transported over several kilometres.

The gold standard method for detecting and quantifying legionellae is still based on the cultivation on selective agar (preferably buffered charcoal yeast extract agar = BCYE with supplements = GVPC) after different treatments to reduce background flora (ISO 11731:1998). The main drawbacks of this method are its long duration of up to 10 days and the fact that a high proportion of the legionellae may persist in the environment in a viable but nonculturable (VBNC) state (Alleron et al. 2008).

Alternatively, for the quantification of nonculturable cells, real-time PCR (qPCR; Yamamoto et al. 1996; Wellinghausen et al. 2001), direct fluorescent antibody counts (DFA, Aurell et al. 2004) or fluorescence in situ hybridization (FISH; Manz et al. 1995; Grimm et al. 1998; Declerk and Ollevier 2006) have been in use. With qPCR, the limit of detection of 102–103 genomic units (GU) per sample volume (Morio et al. 2008; Wery et al. 2008) is high in comparison to the cultivation technique with a detection limit of three colony forming units (CFU) per sample volume (according to Poisson distribution of randomly dispersed particles; ISO, 2000). However, qPCR is a perfect tool to rapidly detect and quantify concentrations higher than 103 GU per sample volume with a high potential of automation. This is useful, as a concentration of 103 CFU per litre is usually the international action level above which measures have to be taken to reduce legionellae concentrations.

The microscopy-based techniques (DFA and FISH) also have a high detection limit of c. 103 cells per sample volume. With automated detection techniques like Laser Scanning Cytometry (LSC, LeBaron et al. 2004), however, this drawback can be overcome indicating that microscopy-based technique is the methodology of choice when the detection of low numbers of target organisms is required. When using monoclonal antibodies to label Legionella, the problem arises that only specific serogroups can be detected (e.g. Leg. pneumophila SG 1, LeBaron et al. 2004), and polyclonal antibodies may be too unspecific for a larger taxonomic unit (species, genus, family). Cross-reactions with other nontarget organisms and differences in cell surface epitopes between naturally grown and stressed environmental cells are currently the major problems (Füchslin et al. 2010).

With FISH, this problem can be circumvented, when specific probes targeting specific segments of the 16S rRNA are designed. For Legionella spp. and Leg. pneumophila, such specific probes have been designed (Manz et al. 1995; Grimm et al. 1998) and used in several studies (e.g. Declerk and Ollevier 2006; Deloge-Abarkan et al. 2007). However, especially in samples of low nutrient concentrations, the target bacteria are often in a state of starvation with very low numbers of ribosomes. FISH signals are extremely weak and indistinguishable from background fluorescence (Wilhartitz et al. 2007). For this reason, catalysed reporter deposition FISH (CARD–FISH) was developed (Pernthaler et al. 2002) which dramatically increases signal intensity. This is achieved by an enzymatic reaction, catalysed by the enzyme horseradish peroxidase (HRP) bound to the specific probe. This enzyme catalyses a colour reaction, where highly reactive phenolic substances (tyramides) are covalently bound to electron-rich residues of adjacent cellular proteins via the intermediate formation of free radicals. Unfortunately, the current FISH probe for Leg. pneumophila is unspecific because of cross-hybridization with other Legionella species.

The aim of the present study was to develop a new CARD-FISH protocol for rapid and specific detection of Leg. pneumophila. This protocol was used to follow the seasonal development of Leg. pneumophila concentrations in two differently operated hospital cooling tower systems in comparison to the conventional standard cultivation method. By relating these data to a variety of environmental (microbial and operational) variables, identification of the driving factors responsible for the seasonal development of culturable and nonculturable Leg. pneumophila in the cooling towers was sought.

Materials and methods

CARD-FISH development

For the development of the CARD-FISH protocol, pure cultures of Leg. pneumophila were spiked into filter-sterilized, autoclaved tap water from the municipal water supply or sterile 1× phosphate-buffered saline (PBS). Serial dilutions were filtered on white 0·2-μm polycarbonate filters (Cyclopore, 25 mm diameter; Whatman, Dassel, Germany). In general, the CARD-FISH protocol is based on the following steps: fixation, permeabilization of the cell envelope, hybridization and signal amplification. For each of these steps, several protocols were applied to come out with the best results. At least, duplicate samples were prepared for each test.


Different cooling tower waters were used to check whether the fixative co-precipitates with potential biocides in the samples. Fixing the bacteria on the filters in a sterile Petri dish with fixative was superior to fixing the whole sample volume prior to filtration because of two reasons. First, the amount of fixative needed is much less when high volumes of sample water have to be investigated, and second, co-precipitation of the fixative with different types of biocides was indeed observed in a variety of cooling tower samples. Fixing the filtered samples directly in 1%p-formaldehyde at 25°C for 1 h, followed by a dehydration series in 50, 80 and 96% ethanol yielded the best results in terms of stability of the cells and stability of the CARD-FISH signal. For each of these steps, filters were placed on top of the fixative and gently submerged with sterile forceps to minimize potential cell loss during the fixation procedure. Filtration of the samples was performed at a pressure difference of maximally −300 mbar. After fixation, the filters were dried and dipped into 0·1% warm (37°C) agarose and dried in glass Petri dishes for 15 min (37°C). Filters were transferred to 1·5-ml Eppendorf tubes for storage at −20°C or immediately processed.


The cells on the filters were incubated in permeabilization mix (100 mg lysozyme, 1 ml 1 mol l−1 Tris–HCl, 1 ml 0·5 mol l−1 EDTA, 8 ml ddH2O) at 37°C for 1 h. Filters were washed in ddH2O and stored for 20 min in 0·01 mol l−1 HCl at room temperature (20–25°C). After a further washing step with ddH2O, filters were dipped in 96% ethanol and dried.


For specific detection of Leg. pneumophila via CARD-FISH, the published probe sequence LEGPNE1 5′-ATC TGA CCG TCC CAG GTT-3′ (Grimm et al. 1998; Declerk and Ollevier 2006) was applied, purchased from Thermo-Electron (Ulm, Germany). At the 5′ end of the probe, HRP was attached. In comparison with normal FISH, lower temperatures and higher stringencies are used for CARD-FISH (Ishii et al. 2004). A formamide concentration of 55% was used for the hybridization buffer consisting of 1·8 ml 5 mol l−1 NaCl, 200 μl 1 mol l−1 Tris–HCl, 5 μl 100% triton X100, 1 g dextran sulphate, 5·5 ml formamide and 1 ml 10% blocking reagent filled up to 10 ml with 1·5 ml of ddH2O. LEGPNE1-HRP-probes were mixed with hybridization buffer to a final concentration of 2·5 ng μl−1. Hybridization took place in an Eppendorf tube, coated with aluminium foil, at 35°C for 12–15 h (Wilhartitz et al. 2007). Thereafter, filters were incubated in prewarmed (37°C) washing buffer (30 μl 5 mol l−1 NaCl, 1 ml 1 mol l−1 Tris–HCl, 500 μl 0·5 mol l−1 EDTA, 50 μl 10% SDS, filled up to 50 ml with ddH2O) for 15 min in a water bath.

Probe specificity

In separate experiments, the specificity of the LEGPNE1-HRP probe was preliminarily tested by hybridizing a variety of Leg. pneumophila serogroups and strains as well as different Legionella spp. All strains had been grown and harvested in the logarithmic phase (OD = 0·7–0·8) from a phosphate-buffered liquid broth (1·5% proteose peptone, 1% yeast extract amended with 0·25 g l−1 ferric phosphate and 0·4 g l−1 l-cysteine; Saito et al. 1981) and suspended in 1× PBS. Logarithmic phase cells were taken because they contain plenty of ribosomes, ideal for specificity testing. The signal intensity of the hybridized cells was analysed from photographs with the free software daime (Digital image analysis in microbial ecology, ver. 1.2; Daims et al. 2006). Photographs were captured with a black and white camera (Digital Sight QilMc; Nikon, Tokyo, Japan) and processed with the software NIS Elements BR 2.3 (Nikon). The same exposure time and digital settings were applied for all strains. For each Legionella strain, ten photographs with at least 200 cells were transferred into daime and analysed. Mean and median intensities were used for assessment of hybridization efficiency.

Signal amplification

After hybridization, the filters were transferred to glass Petri dishes with PBS-mix (25 μl Triton X100, filled up with 50 ml 1× PBS). The tyramide substrate mix was freshly prepared on ice, consisting of 493 μl amplification buffer (4 g dextran sulphate, 16 ml 5 mol l−1 NaCl, 400 μl blocking reagens 10%, 23·6 ml of 1× PBS), 5 μl 0·0015% H2O2 and 5 μl Alexa488-tyramide. For amplification, the dry filters were incubated at 37°C for 30 min in sterile 1·5 ml tubes with 50 μl substrate mix. The filters were dried, washed with ddH2O, dehydrated in 96% ethanol, and embedded on microscopic slides in DAPI-Mix, consisting of 5.5 parts Citifluor (Citifluor Ltd, Leicester, UK), 1 part Vectashield (Vector Laboratories, Peterborough, UK), and 0·5 parts PBS with DAPI at a final concentration of 1 mg ml−1. Signal amplification resulted in excellent signal intensity and the cells were easily distinguishable from background.

Detection limit

The detection limit is dependent on the effective filter area as well as on the size and number of the microscopic fields. With an effective filter area of 346·5 mm2, a field size of 10 000 μm2 and a field number of 20 (which is practicable for routine investigations), a theoretical detection limit of 1·7 × 103 cells per filter is achieved. To decrease this limit, 100 microscopic fields can be screened for the presence of Leg. pneumophila resulting in a value of 3·5 × 102 cells per filter. With a filtration volume of usually 100 ml, the detection limit of the method corresponded to 3·5 × 102 cells per 100 ml or 3·5 × 103 cells per litre, respectively.

Quantification limit

For the determination of the quantification limit, several sources of bias have to be considered, like the inhomogeneous distribution of the cells in the sample and on the filter or the loss of cells during sample preparation. In addition, the extrapolation of the number of identified cells on the filter to the total number of cells on the filter leads to an overestimation of low cell numbers. To determine the quantification limit, a pure culture of Leg. pneumophila suspended in 1× PBS was diluted in a series of consecutive 1 : 2 and 1 : 5 dilution steps from 107 to 101 cells per filter. Cell numbers were determined via LEGPNE1-HRP and DAPI counts and compared with the calculated values along the dilution series. The lowest concentration, where LEGPNE1-HRP and DAPI counts were not significantly different from the calculated cell number on the filter was defined as the limit of quantification.

Cooling tower sampling

To apply the newly developed CARD-FISH protocol for assessing Leg. pneumophila concentrations in cooling towers, samples from two hospital cooling tower systems of different size and with contrasting disinfection strategies were collected over several months. These data were compared with culture-based enumerations of Leg. pneumophila and Legionella spp. concentrations and to a variety of microbial, chemophysical and operational variables.

Hospital A

The investigated cooling system of hospital A consists of four cooling tanks with a total volume of c. 300 m3 (Fig. 1a). In summer, all cooling tanks are in continuous operation. In winter, the summer cooling tanks are emptied and the winter cooling tanks are operated if required. All cooling tanks are emptied and cleaned once a year. The water which is lost by evaporation is replaced by tap water, which is amended with an anticorrosive (Trasar3DT250; Nalco, Vienna, Austria). The residence times at the sampling dates varied between 0·4 and 0·7 days, depending on the cooling demand of the hospital. In this cooling tower system, biocides (sodium hypochlorite: sodium bromide 10 : 1) are added once a day (usually at 7 am) at a dosing point situated shortly after the cooling tanks to keep Redox potential at 550 mV. Total chlorine and bromide concentrations, measured with the DPD-method (Lovibond, Dortmund, Germany) are in the range from 5 to 6 mg l−1. TOC concentrations ranged from 2·9 to 13·4 mg l−1 (median: 6·3 mg l−1). Samples from hospital A were taken from May to August 2009 on seven occasions. In total, 29 water samples were taken from 3 to 7 sites within the cooling tower system (Fig. 1a). Sampling site 1 was situated shortly after the cooling tower tanks on the roof. Sampling sites 2 and 3 were positioned before and after the cooling units at the lowest point of the system. On two occasions (July 1 and 14), additional samples were taken directly from the four cooling tower tanks on the roof. At all of these sampling sites, water was taken from taps which were flamed beforehand.

Figure 1.

 General scheme of the investigated cooling system of (a) hospital A with the regular sampling points 1 to 3 and the 4 occasional sampling points (small symbols) at the cooling tanks and (b) hospital B with the regular sampling points 1 to 4 and the 2 occasional sampling points (small symbols) at the cooling tanks (CT).

Hospital B

The cooling system of hospital B consists of six cooling tanks on the roof with a total volume of c. 30 m3 (Fig. 1b). The cooling tanks are operated in alternating manner to avoid stagnancy during periods of lower cooling demand. Before winter, cooling tanks are partially emptied to a volume of c. 30% of its capacity. The water which is lost by evaporation is replaced by softened tap water. The residence time usually lies between 1 and 2 days, depending on the cooling demand of the hospital. Starting in April 2009 with an initial biocide shock with bromo-nitropropan-diol (Transhelsa, Leobersdorf, Austria; final conc. c. 35 mg l−1), a biocide mixture of isothiazolinones (CIT/MIT, 1·5%; Transhelsa) was continuously added to the system at various amounts, ranging from 2 to 4 l day−1 (final conc. c. 1–2 mg l−1). TOC concentrations ranged from 4·8 to 15·1 mg l−1 (median: 9·8 mg l−1). Samples from hospital B were taken from April to October 2009 at 13 occasions. A preliminary sampling event took place in September 2008. In total, 55 water samples were taken from 2 to 6 sites within the cooling tower system (Fig. 1b). Sampling sites 1 and 2 were situated at the cooling tower tanks on the roof. As the six cooling towers were operated alternately, one sample was always taken from an active tank and another sample from a tank not in use. Sampling sites 3 and 4 were positioned before and after the cooling units at the lowest point of the system. On three occasions (5 May, 20 May, 10 Jun), additional samples were taken from two active cooling tower tanks on the roof. The samples from the cooling tower tanks were taken at the blowdown sites, where concentrated water above an electrical conductivity of 800 μS cm−1 is discharged from the system. The blowdown ports as well as the taps at sampling sites 3 and 4 were flamed before sampling.


Samples for all microbiological variables were collected in sterile glass flasks. Samples for heterotrophic plate counts and Legionella cultivation were filled into 250-ml glass flasks containing sodium thiosulfate. Data on water temperature, pH and conductivity were determined on site with portable metres (WTW, Weilheim, Germany). Samples were transported at ambient temperature in isolation boxes to the laboratory within 2 h.

Legionella pneumophila CARD-FISH.

Total Leg. pneumophila concentrations were determined via the developed CARD-FISH protocol (see above) by epifluorescence microscopy.

Legionella pneumophila and Legionella spp. cultivation

For quantification of Leg. pneumophila and Legionella spp. via cultivation, the international standard ISO 11731 (1998) was followed. 100 ml subsamples were filtered on polycarbonate filters (0·22 μm, ø 47 mm; Whatman). The filters were transferred into sterile 100-ml glass flasks containing 5 ml of 1 : 40 Ringer solution (Ringer tablet from Oxoid mixed 1 : 40 with distilled water, autoclaved) and placed into an ultrasonic bath for 2 min. at room temperature. The concentrated cell suspension was divided into three portions. One portion was directly streaked onto a BCYE agar plate, supplemented with glycine, vancomycin, polymyxin B and cycloheximide (GVPC; bioMerieux, Vienna, Austria); the second portion was treated before with Leg-10× acid buffer; and the third one was subjected to a heat treatment at 50°C for 30 min in a water bath. Undiluted subsamples and up to two dilutions were streaked on agar plates, which were incubated at 36°C for up to 10 days. Examination of the colonies took place on the 4th (±1), 6th (±1) and 10th day. Legionella species were identified by morphological characteristics like colour and fluorescence. Per treatment, all or at least three colonies with typical or suspected Legionella morphology from each treatment were further streaked onto GVPC-agar as well as blood agar (Columbia agar + 5% sheep blood, bioMerieux) and incubated for 2 days at 36°C. Colonies which grew on Leg-agar without growing on blood agar were assigned as presumptive Legionella spp. For the identification of Leg. pneumophila, an agglutination test (Legionella Latex Test; Oxoid, Basingstoke, UK) was conducted and assigned as serogroups 1 or serogroups 2–14. The limit of detection was 30 CFU l−1, the limit of quantification 350 CFU l−1, according to our validation of the method (data not shown).

Heterotrophic plate counts

Heterotrophic plate counts were determined by pour plate method according to the international standard ISO 6222 (1999). Duplicate 1-ml samples and three 1 : 10 dilutions were mixed with 20 ml of ISO 6222 agar at 45°C. One replicate of each dilution series was incubated at 36 ± 2°C for 44 ± 4 h (HPC 37) and at 22 ± 2°C for 68 ± 4 h (HPC 22).

Bacterial production

Bacterial production was determined by 3H-leucine incorporation into bacterial biomass according to Kirschner and Velimirov (1999). This parameter is a good estimate of total bacterial activity in aquatic ecosystems. Briefly, four subsamples of 1 ml (three samples and one blank killed with 5% final conc. of trichloroacetic acid) were incubated with a 60 nmol l−1 final conc. of 3H-leucine. After 30 min, the samples were killed and the proteins precipitated and purified. The radioactivity in the proteins was determined in a liquid scintillation counter (1900 TR; Perkin Elmer, Brunn, Austria).

Statistical analysis

All statistical analysis was carried out with spss 17.0 (IBM, Armonk, New York). All data were log-transformed for statistical calculations. Spearman Rank correlations were calculated for estimating relationships between variables.


Specificity of the CARD-FISH probe

Results from preliminary specificity experiments displayed that all investigated Legionella spp. hybridized with the LEGPNE1-HRP probe, at slightly weaker intensities (Table 1). Because such a high percentage of cross-hybridization was observed, the ribosomal databases RDPII, SILVA and Greengenes were checked for target sequence similarities. Highest similarity with the target sequence was found only for bacteria belonging to the genus Legionella. Two mismatches occurred always at the same positions. From this information, the following competitor probe, named LEGCOMP2, was designed: 5′-ATC TGA CCT GCC CAG GTT-3′. The majority of known Legionella species and serogroups (n = 67) were then examined for cross-hybridization with LEGPNE1-HRP in presence of LEGCOMP2 following the developed protocol. With the exception of Legionella yabuuchiae, no cross-hybridization was observed for any Legionella strain, when the competitor probe was used (Table 2). All Leg. pneumophila serogroups and strains gave an intense signal.

Table 1.   Signal intensities (% of median and mean values) estimated via DAIME from photographs with different Legionella spp. hybridized with LEGPNE1-HRP, without and with competitor probe (LEGCOMP2)
SpeciesWithout competitor probeWith competitor probe
% of mean values% of median values% of mean values% of median values
Legionella pneumophila100100100100
Legionella bozemanii826600
Legionella dumoffi786300
Legionella micdadei806500
Legionella longbeachae806500
Legionella feeleii776300
Legionella sainthelensi735800
Table 2.   Results from specificity testing of the used CARD-FISH probe LEGPNE1-HRP with competitor probe LEGCOMP2
No.SpeciesNC/DSM no.+/−
  1. +, bright signal; −, no signal.

 1Legionella pneumophila SG 1- Allentown 1NC 12024+
 2Leg. pneumophila SG 1- Bellingham 1NC 11404+
 3Leg. pneumophila SG 1- BenidormNC 12006+
 4Leg. pneumophila SG 1- Camperdown 1NC 12098+
 5Leg. pneumophila SG 1- FranceNC 12007+
 6Leg. pneumophila SG 1- Heysham 1NC 12025+
 7Leg. pneumophila SG 1- Knoxville 1NC 11286+
 8Leg. pneumophila SG 1- OldaNC 12008+
 9Leg. pneumophila SG 1- OxfordNC 12009+
10Leg. pneumophila SG 1- Philadelphia 1NC 11192+
11Leg. pneumophila SG 1- Philadelphia 2NC 11193+
12Leg. pneumophila SG 2NC 11230+
13Leg. pneumophila SG 3NC 11232+
14Leg. pneumophila SG 5NC 11417+
15Leg. pneumophila SG 6NC 11287+
16Leg. pneumophila SG 7NC 11984+
17Leg. pneumophila SG 8NC 11985+
18Leg. pneumophila SG 9NC 11986+
19Leg. pneumophila SG 10NC 12000+
20Leg. pneumophila SG 11NC 12179+
21Leg. pneumophila SG 12NC 12180+
22Leg. pneumophila SG 13NC 12181+
23Leg. pneumophila SG 14NC 12174+
24Leg. pneumophila ssp. pneumophilaDSM 7513+
25Leg. pneumophila ssp. fraseriDSM 7514+
26Leg. pneumophila ssp. pasculleiDSM 7515+
27Legionella donaldsoniiNC 13292
28Legionella fairfieldensisNC 12488
29Legionella feeleiiNC 12022
30Legionella quateirensisNC 12376
31Legionella quinlivaniiNC 12433
32Legionella adelaidensisDSM 19888
33Legionella anisaDSM 17627
34Legionella beliardensisDSM 19152
35Legionella birminghamensisDSM 19232
36Legionella brunensisDSM 19236
37Legionella busanensisDSM 22853
38Legionella cherriiDSM 19213
39Legionella cincinnatiensisDSM 19233
40Legionella drozanskiiDSM 19890
41Legionella erythraNC 11977
42Legionella falloniiDSM 19889
43Legionella geestianaNC 12373
44Legionella gratianaDSM 21233
45Legionella gresilensisDSM 21218
46Legionella hackeliaeDSM 19214
47Legionella impletisoliDSM 18493
48Legionella israelensisDSM 19235
49Legionella jamestowniensisDSM 19215
50Legionella jordanisDSM 19212
51Legionella lansingensisDSM 19556
52Legionella londiniensisDSM 21234
53Legionella longbeachae SG1DSM 10572
54Legionella moravicaDSM 19234
55Legionella nautarumDSM 21805
56Legionella oakridgensisDSM 21215
57Legionella parisiensisDSM 19216
58Legionella rubrilucensDSM 11884
59Legionella sainthelensiDSM 19231
60Legionella santicrucisDSM 23075
61Legionella shakespeareiDSM 23087
62Legionella spiritensisDSM 19324
63Legionella steigerswaltiiDSM 23076
64Legionella taurinensisDSM 21897
65Legionella tucsonensisDSM 19246
66Legionella worsleiensisDSM 21907
67Legionella yabuuchiaeDSM 18492+

Quantification limit

The quantification limit which was defined as the lowest concentration, where LEGPNE1-HRP and DAPI counts were not significantly different from the calculated cell number on the filter amounted to 3·5 × 103 cells per filter (data not shown). With a filtration volume of usually 100 ml, the quantification limit of the method corresponded to 3·5 × 103 cells per 100 ml or 3·5 × 104 cells l−1, respectively.

Dynamics of Legionella pneumophila in cooling towers

Hospital A

Samples taken from the different sampling sites were pooled for data presentation, as no significant differences were observed for all bacterial variables between sites (anova; < 2·1; P > 0·1). With the exception of Leg. pneumophila concentrations determined via cultivation, all microbiological variables followed the same seasonal trend (Fig. 2a,b). Until the end of June, bacterial concentrations and activity increased, thereafter a decrease until the end of July was recorded. By the end of August, values increased again. Legionella pneumophila concentrations determined via CARD-FISH showed significant correlation to Legionella spp. CFUs, HPC 22°C and HPC 37°C as well as to bacterial production and temperature (Table 3). Legionella pneumophila CFUs showed no correlation with CARD-FISH results and with any other variable, while Legionella spp. CFUs correlated with most microbiological variables (Table 3). Only a small proportion between zero and 20% of Legionella spp. was identified as Leg. pneumophila. Besides Leg. pneumophila, Leg. anisa and Leg. bozemanii were identified. Basic information on pH, water temperature and electrical conductivity can be found in Table 4.

Figure 2.

 Development of bacterial variables in the cooling tower system of hospital A during the period from May to August 2009. Values represent the mean of samples taken from 2 to 6 sites. (a) Legionella pneumophila and Legionella spp. concentrations determined via cultivation and L. pneumophila via CARD-FISH. The upper line depicts the limit of quantification and the lower line the limit of detection for the CARD-FISH method. The limits of detection (30 CFU l−1) and quantification (350 CFU l−1) for the culture method are not shown. Single measurements resulting in values below the quantification limit 350 CFU l−1 were replaced by an average value of 175 CFU l−1. (b) Bacterial production (BP) and heterotrophic plate counts (HPC) incubated at 22°C and 37°C. a: (inline image) Leg. pneumophila (CFU l−1); (inline image) Legionella sp. (CFU l−1) and (inline image) Leg. pneumophila-CARD (cells l−1). b: (inline image) HPC37 (CFU l−1); (inline image) HPC22 (CFU l−1) and (inline image) BP (ng C l−1 h−1).

Table 3.   Spearman′s rho correlation matrix for microbiological variables in the cooling tower system of hospital A
VariableLegionella pneumophila CultLegionella spp. CultHPC 37°CHPC 22°CBPTemp
  1. HPC, heterotrophic plate count; BP, bacterial production; CF, CARD-FISH; Cult, cultivation; Temp, temperature.

  2. Conductivity and pH were not included because no correlation was observed with these variables.

  3. *< 0·05; **< 0·01; ***< 0·001; = 22–25.

Leg. pneumophila CF−0·160·40*0·45*0·51*0·58**0·46*
Leg. pneumophila Cult0·050·01−0·033−0·380·15
Legionella spp. Cult 0·49*0·380·42*0·59**
HPC 37°C  0·45*0·66***0·14
HPC 22°C   0·70***0·48*
BP    0·15
Table 4.   Basic chemophysical data of the cooling towers in hospital A and B during the investigation period
 Hospital AHospital B
pHEC (μS cm−1)Temp (°C)pHEC (μS cm−1)Temp (°C)
  1. EC, electrical conductivity at 25°C; Temp, water temperature.


With CARD-FISH, two to three orders of magnitude higher concentrations of Leg. pneumophila in comparison with cultivation were observed (Fig. 2a). Legionella pneumophila CFUs varied between 102 and 103 CFU l−1, while cell numbers ranged from 104 up to 5 × 105 cells l−1. In the middle of June, no Leg. pneumophila could be detected by cultivation, while 2·5 × 105 cells l−1 were detected with CARD-FISH. The addition of biocide was obviously ineffective in eradicating culturable and nonculturable legionellae in this cooling system.

Hospital B

For the cooling tower system of hospital B also no significant differences were found between sites (anova; < 1·1; P > 0·1). In contrast to hospital A, the seasonal development of the bacteria in the system was mainly influenced by the addition of biocides. No correlations were found with any of the physicochemical parameters listed in Table 4. With the exception of Leg. pneumophila determined via CARD-FISH, all microbiological variables showed significant negative correlation with the biocide (Table 5). After high values for all measured bacterial variables in September 2008, the addition of biocide during restarting the system after winter break in April 2009 led to a massive decrease of culturability and activity of the bacterial population, including legionellae (Fig. 3). After the beginning of the treatment, HPC 22°C and HPC 37°C as well as BP dropped partly to undetectable levels, and Legionella spp. and Leg. pneumophila were not detectable by cultivation anymore until the end of the observation period. In contrast, although not correlated with the biocide concentrations, Leg. pneumophila detected via CARD-FISH followed the biocide pattern with a time lag. In the beginning CARD-FISH Leg. pneumophila concentrations remained at high levels, when biocide was added to the system at a concentration of c. 1 mg l−1 day−1. Not before the upshift of biocide concentration to c. 2 mg l−1 day−1Leg. pneumophila CARD-FISH concentrations dropped below the limit of quantification, still remaining at detectable level. After stopping biocide addition in mid-July, Leg. pneumophila CARD-FISH concentrations started to increase again and values above the quantification limit were observed. When biocide addition was restarted in September at 2 mg l−1 day−1, Leg. pneumophila CARD-FISH concentrations dropped again in the beginning but re-increased quickly, especially after the end of biocide addition (Fig. 3).

Table 5.   Spearman’s rho correlation matrix for microbiological variables in the cooling tower system of hospital B
VariableLegionella pneumophila CultLegionella spp. CultHPC 37°CHPC 22°CBPBiocide
  1. HPC, heterotrophic plate count; BP, bacterial production; CF, CARD-FISH; Cult, cultivation.

  2. Temperature, conductivity and pH were not included because no correlation was observed with these variables.

  3. *< 0·05; **< 0·01; ***< 0·001; = 52–59.

Leg. pneumophila CF0·000·000·160·29*0·08−0·03
Leg. pneumophila Cult1·00***0·38**−0·121−0·31*−0·29*
Legionella spp. Cult  –0·38**−0·121−0·31*−0·29*
HPC 37°C  0·79***0·50***−0·78***
HPC 22°C   0·66***−0·56***
BP    −0·57***
Figure 3.

 Development of bacterial variables in the cooling tower system of hospital B during the period from Sep 2008 to October 2009. Values represent the mean of samples taken from 2 to 6 sites. (a) Legionella pneumophila and Legionella spp. concentrations determined via cultivation and L. pneumophila via CARD-FISH. The upper line depicts the limit of quantification and the lower line the limit of detection for the CARD-FISH method. The limits of detection (30 CFU l−1) and quantification (350 CFU l−1) for the culture method are not shown. (b) Bacterial production (BP) and heterotrophic plate counts (HPC) incubated at 22°C and 37°C. a: (inline image) Leg. pneumophila (CFU l−1); (inline image) Legionella spp. (CFU l−1) and (inline image) Leg. pneumophila-CARD (cells l−1). b: (inline image) HPC37 (CFU l−1); (inline image) HPC22 (CFU l−1) and (inline image) BP (ng C l−1 h−1).

Before the first use of the biocide, Leg. pneumophila concentrations detected by CARD-FISH and by cultivation were of similar magnitude (105–106 cells per CFU l−1). However, after biocide application, no legionellae could be detected by cultivation anymore, while Leg. pneumophila remained detectable with CARD-FISH until the end of the investigation period with concentrations up to 1 × 106 cells l−1 (May).


The use of the published FISH probe for direct detection of Leg. pneumophila (Grimm et al. 1998) led to unsatisfactory results in terms of signal intensity and background fluorescence in our cooling tower samples. We therefore developed a CARD-FISH protocol which significantly improved detectability of the cells. The cells were much brighter and larger and could be easily differentiated from background fluorescence. However, when applying this protocol to a series of related Legionella species, we observed a high degree of cross-hybridization. Actually, all tested Legionella species showed a rather strong fluorescence signal. A sequence comparison using the ribosomal databases RDPII, SILVA and Greengenes revealed that almost all legionellae showed the same two mismatches in comparison with the Leg. pneumophila target sequence. By applying the newly designed competitor probe, cross-hybridization could be prevented completely. For specific detection of Leg. pneumophila in environmental samples like cooling towers, either with FISH or CARD-FISH, it is thus necessary to use this new competitor probe.

In comparison with the results from cultivation, Leg. pneumophila concentrations determined via CARD-FISH are available within 24 h, instead of 10 days. Most importantly, they were also significantly higher in both investigated cooling tower systems. In hospital A, the difference was on average two to three orders of magnitude. In hospital B, the difference was up to six orders of magnitude, as after the first application of the biocide in hospital B, no culturable legionellae could be detected until the end of the investigation period. There are several reasons for these discrepancies. First, the culture method may lead to underestimation of viable cells. When abundant competing bacterial flora is present, overgrowth and inhibition of slowly growing legionellae is possible, even when the samples are pretreated with acid and heat to reduce background flora (Kimiran-Erdem and Yazici 2008). Second, a large proportion of the Leg. pneumophila cells detected with CARD_FISH may have been in a viable but nonculturable (VBNC) state (Alleron et al. 2008), where they are not detectable on traditional culture media, but where detection is possible with alternative molecular biological methods like PCR (Dusserre et al. 2008) or FISH (Moritz et al. 2010). Whether such VBNC cells are of importance to human health is still under debate (Rowan 2004). It has been shown that in the presence of amoebae, legionellae are able to resuscitate from VBNC state and regain culturability (Steinert et al. 1997; Alleron et al. 2008). Third, the CARD-FISH method may also detect inactive or starved cells, which may not be relevant for human health. In comparison with normal FISH, bright signals are also achieved, when the ribosomal content of the cells is low (Pernthaler et al. 2002), as it is the case with starving or nonviable cells.

We observed specific variation patterns of Leg. pneumophila concentrations determined via CARD-FISH in both cooling towers systems. In hospital A, Leg. pneumophila followed the same seasonal pattern as the total bacterial community in the system and was correlated significantly to HPC22, HPC 37 and total bacterial production as well as to Legionella spp., determined via cultivation. In hospital B, Leg. pneumophila concentrations determined via CARD-FISH clearly followed the biocide treatments with a time lag. We are thus convinced that the CARD-FISH method based on the used probes (inclusive competitor probe) can be effectively applied for monitoring Leg. pneumophila concentrations in cooling tower systems and other water systems with abundant background flora. Without competitor probe, an overestimation of Leg. pneumophila may occur, because of cross-hybridization of the probe with other Legionella species. Only a single nonpneumophila species (L. yabuuchiae) of all investigated species (n = 41) cross-reacted with the Leg. pneumophila probes, but this species was reported to occur only in Japanese soils contaminated with industrial wastes (Kuroki et al. 2007).

One major drawback of the presented CARD-FISH protocol based on the detection of positive cells by epifluorescence microscopy derives from the fact that a rather high detection limit and quantification limit has to be accepted in comparison with the culture method, where in theory only a single culturable cell (CFU) can be detected per sample volume and where the limit of quantification is 35 CFU per sample volume tested. Using higher filtration volumes for microscopic detection may only sparsely solve this problem. Higher background fluorescence and markedly longer filtration times make these options an inconvenient alternative.

However, this problem may be overcome by automated cell counting technology, like flow cytometry (Füchslin et al. 2010) or solid phase cytometry (Joux and Lebaron 2000). Lower volumes for analysis can be used with these technologies, thus reducing interference problems from fluorescing background particles. Actually, only solid phase cytometry is able to detect rare events at a comparable level as cultivation (Mignon-Godefroy et al. 1997); with this method theoretically one cell can be detected on a filter, analogous to one colony on an agar plate. LeBaron et al. (2004) were the first who quantified Leg. pneumophila serogroup 1 in a solid phase cytometer after staining the cells with several monoclonal and polyclonal antibodies. While polyclonal antibodies being too unspecific, monoclonal antibodies restrict this application to specific serogroups which makes antibody staining useful only for certain issues (e.g. to verify whether a specific serogroup is colonizing a water system of interest, when no legionellae can be cultured for identification). FISH is in this respect more promising for a broader applicability of detecting legionellae at the species level in water samples with a solid phase cytometer, most desirably combined with a viability test. FISH probes for Legionella spp. are available (Manz et al. 1995), but so far no FISH probe has been designed providing enough signal intensity to be used in combination with a solid phase cytometer. The use of the newly developed CARD-FISH probe enables combined application with a solid phase cytometer. As an alternative to culture-based or microscopic quantification of Leg. pneumophila, quantitative real-time PCR has been applied, but here one has to cope with a detection limit of c. 102–103 genomic units per sample volume (Morio et al. 2008; Wery et al. 2008; Chen and Chang 2010).

Based on the observations from our study, future research has to focus on the questions, which factors influence the distribution and multiplication of legionellae, especially of Leg. pneumophila, in cooling towers and which risk emanates from nonculturable legionellae to human health. The factors influencing Legionella dynamics in cooling towers may vary significantly. Seasonal patterns of outbreaks have been reported (Bentham and Broadbent 1993), and more recent studies did not find any seasonal cycles of Legionella concentrations in cooling towers (Turetgen et al. 2005; Ragull et al. 2006). In our study, we found that the Leg. pneumophila concentrations in hospital A followed the variation patterns of the total microbial community. This indicates that abiotic factors having impact on microbial growth have been responsible for these patterns, like seasonal changes in temperature or operational–technical parameters. From the short residence times in hospital A (0·4–0·7 days), it can be concluded that mainly biofilms were responsible for a continuous release of legionellae into the water. Mechanical cleaning/biofilm removal could thus have been an additional and effective measure to control legionellae in this system, but more detailed information on the niches of Leg. pneumophila in this system is required. In hospital B, it was obviously the application of a potent biocide regulating Legionella abundance and culturability. Lower concentrations in the beginning only reduced culturable legionellae, while nonculturable Leg. pneumophila concentrations remained high. Only higher biocide concentrations could effectively reduce nonculturable legionellae.

From the comparison of Leg. pneumophila concentrations determined via CARD-FISH and culture method, it became obvious that detecting legionellae with culture methods only may lead to a severe underestimation of the total numbers of Leg. pneumophila present. Cell-based and nucleic acid-based methods are thus important tools to expand our knowledge on Legionella concentrations and proliferation especially in man-made water systems. Our data suggest that there is a high potential for rapid multiplication of Leg. pneumophila in cooling tower systems, even when culturable Leg. pneumophila concentrations are low or undetectable. This may imply a potential health risk. The combination of the new methods with assays for testing viability or even virulence as well as strategies to reduce the detection limits will enable reliable quantitative risk assessment in the future.


The project was financed by the AUVA and ECHEM. Manfred Hinker (AUVA) and Irene Zweimüller (ECHEM) are gratefully acknowledged for their logistic support. Special thanks also to Michael Burger and Günter Brandl, who enabled sampling in the two hospitals, as well as to Sonja Knetsch, Marion Griessler and Elisabeth Holzhammer for the cultivation of legionellae and sampling at hospital A. All Legionella strains were provided by the Institute for Medical Microbiology and Hygiene, AGES, Vienna, Austria.

Conflict of interest

The authors declare that they have no conflict of interest concerning this publication.