Rapid Detection and Enumeration of Legionella pneumophila in Hot Water Systems by Solid-Phase Cytometry

Bacterial strains used for development of an IF staining protocol.The Legionella type strains L. pneumophila serogroup (sg) 1 Philadelphia (ATCC 33152) and L. pneumophila sg 3 Bloomington (ATCC 33155) were used for the development of the staining protocol. All Legionella strains were grown on BCYEα (buffered activated charcoal and yeast extract in 2-[(2-amino-2-oxoethyl)amino]ethanesulfonic acid; Oxoid, Dardilly, France) agar for 48 h at 35 ± 1°C with 2.5% carbon dioxide.

Fifteen non-Legionella strains were used for the specificity control (Table 1). These strains were selected because they are frequently found in water systems or have been reported to be cross-reactive with anti-Legionella antibodies.

Artificially contaminated water samples.Tap water samples negative for L. pneumophila were inoculated with cultured L. pneumophila sg 1 cells (for L. pneumophila sg 1 antibodies) or sg 3 cells (for L. pneumophila non-sg 1 antibodies) (48-h cultures). Cells were added at different concentrations, ranging from 1 to 10³ cells per membrane, to evaluate the sensitivity of the assay. In addition, naturally contaminated water samples of different volumes (from 10 to 100 ml) were filtered to assess the potential influence of the volume on cell counts.

Natural water samples.A total of 26 natural water samples were collected from the hot water systems of four hospitals in Lyon, France, between September 2002 and January 2003. Samples were collected from tap water and from showers in sterile bottles (CML, Nemours, France) according to international standard ISO 11731, modified as described below. The reproducibility of counts obtained by the two methods (solid-phase cytometry and culture) was determined from the analysis of seven different sources located in two hospitals. For each sampling point, 4 liters were collected in separate bottles; 1 liter was analyzed in at least three replicates by the cytometry method, and the remaining 3 liters were analyzed by the culture method. Nineteen other water samples received at the French national reference center for Legionella (NRCL) in the course of routine hospital water system surveillance were analyzed in one single replicate by both cytometry and culture.

Enumeration of legionellae by culture. Legionella species and serogroups were enumerated and identified according to the ISO-11731 standard, modified as follows. Before concentration, 200 μl of each sample was directly plated onto selective GVPC medium (BCYE supplemented with 3 g of glycine, 100,000 U of polymykin B, 80 mg of cychloheximide, and 1 ng of vancomycin per liter; Oxoid). Then 1-liter samples were filtered through 0.2-μm-pore-size polycarbonate filters (Millipore, Billerica, Mass.), placed in 5 ml of sterile water, and treated with ultrasonic energy for 2 min with a Bioblock Scientific sonicator (Fisher Bioblock Scientific, Illkirch, France) operated at 35 kHz. One hundred microliters of the concentrate was spread by plating onto selective GVPC medium. Samples with high background growth were subjected to standard heat and acid treatments to eliminate non-Legionella organisms. Plates were incubated at 35 ± 2°C with 2.5% carbon dioxide, and colonies were counted after 3, 5, and 10 days. Colonies were examined for fluorescence under a Wood lamp. Colonies exhibiting Legionella morphology were transferred to BCYEα medium (Oxoid) and blood agar (bioMérieux, Marcy l'Etoile, France) as a control. At least five colonies per sample were identified by Legionella-specific latex reagents (Oxoid).

Labeling reagents for ChemScanRDI detection.Various polyclonal and monoclonal primary antibodies were evaluated for Legionella labeling (Table 2). Two polyclonal antibodies were used: an anti-L. pneumophila sg 1 antibody produced in our laboratory (prepared in rabbits by using L. pneumophila sg 1 ATCC strains Philadelphia, Bellingham, Pontiac, and Knoxville) and labeled with different conjugates and a commercialized anti-L. pneumophila antibody labeled with fluorescein isothiocyanate (FITC) (Accurate Chemical & Scientific Corp., Westbury, N.Y.). Furthermore, the following monoclonal antibodies (MAbs) were tested: three commercially available MAbs specific for L. pneumophila sg 1 to 14 (from Bio-Rad [Hercules, Calif.], Maine Biotechnology Services [Portland, Maine], and Research Diagnostics Inc. [Flanders, N.J.]) and three MAbs specific for L. pneumophila sg 1; L. pneumophila sg 2 to 6, 8 to 10, and 12 to 15; and all Legionella spp., including L. pneumophila. (10).

Several detection systems were evaluated as shown in Table 2. The polyclonal rabbit antibodies could be detected by anti-rabbit secondary antibodies, and the MAbs could be detected by anti-mouse secondary antibodies. The biotin-labeled antibodies were detected with streptavidin conjugates, and horseradish peroxidase (HRP)-labeled antibodies were detected with tyramide-FITC. For detection in the red channel, the polyclonal anti-L. pneumophila sg 1 antibody (NRCL) was revealed by an Alexa Fluor 647-RPE (Molecular Probes, Leiden, The Netherlands) or RPE-Cy5 (Dako, Glostrup, Denmark) conjugate.

The dilution buffers and concentrations of antibodies were chosen for each type of staining after a preliminary assay (data not shown). Various membrane filters were evaluated for each staining (data not shown). Black polyester membranes (pore size, 0.26 μm; diameter, 25 mm; Chemunex) provided optimal characteristics for FITC labeling. For the red staining, white polycarbonate membranes (pore size, 0.2 μm; diameter, 25 mm; Whatman, Maidstone, United Kingdom) were selected, since the other membranes tested presented excessive fluorescence baselines or nonspecific antibody fixation.

To reduce the background due to nonspecific fixation of the staining reagents, we used (i) various blocking reagents including bovine serum albumin (BSA) (Sigma, St. Louis, Mo.) and different normal sera and (ii) the counterstains Evans blue (Réactifs RAL, Martillac, France), CSE/2, CSA, and CSB (Chemunex). Blocking reagents were either applied prior to application of the primary antibody or mixed with the primary antibody, and the counterstain was mixed with the detection reagent.

Labeling technique.To optimize the labeling protocol, bacteria from a 48-h culture were filtered at a concentration of about 100 bacteria per membrane. For water samples, different volumes ranging from 1 to 20 ml were filtered, depending on the concentration of Legionella cells in the sample. Samples were filtered under a maximum vacuum of 100 mm of mercury through the labeling membranes. For labeling of bacteria, the membrane was incubated for 60 min at 37°C with 100 μl of labeling solution containing the antibody in phosphate-buffered saline (PBS) (bioMérieux)-0.01% Tween 20 (Sigma)-2% BSA (PBS-T-BSA). The membrane was rinsed by transfer to an absorbent pad soaked with 500 μl of PBS-T-BSA and a 3-min incubation to eliminate excess antibody, followed by transfer to a second absorbent pad to eliminate the remaining washing buffer. The secondary antibody was diluted in PBS-T-BSA with 0.01% Evans blue and applied for 30 min as described for the primary antibody. The EnVision antibody was used according to the supplier's suggestions. Streptavidin conjugates were diluted in PBS-T-BSA and applied in the same manner as the secondary antibody, except that incubation was performed for 30 min at room temperature. Tyramide-FITC was diluted with the dilution buffer supplied with the tyramide and was incubated for 15 min at room-temperature.

Solid-phase cytometer enumeration.After labeling, the membrane was transferred to the sample holder of the ChemScanRDI solid-phase cytometer (Chemunex). The holder had previously been overlaid with a support pad (black membrane; Chemunex) soaked with 100 μl of PBS-20% glycerol. The entire filter surface is scanned within 3 min. The ChemScanRDI laser scanning device has been described previously (18, 27). Briefly, light is provided by an air-cooled argon laser emitting at 488 nm, and the fluorescence emission is collected in the green channel (500 to 540 nm) for FITC and in the red channel (655 to 705 nm) for Alexa Fluor 647-RPE and RPE-Cy5.

The device is able to differentiate between labeled organisms and fluorescent particles present in the sample based on the optical and electronic characteristics of the signals generated (18). When this discrimination procedure is not used, all fluorescent events detected, including the background signal, are counted by the cytometer and can be manually distinguished by using an epifluorescence microscope (Olympus BX41) fitted on a motorized stage driven by the computer of the cytometer, as described elsewhere (5, 18, 20). Data obtained from manual microscope validation of bacteria were used to optimize the setting of the instrument and the automatic discrimination procedure. These optimized discrimination settings were used for the analysis of natural water samples; in addition, the identities of bacteria were confirmed by manual microscope validation.

Specificity control of the IF protocol.The specificity of IF labeling was assessed by application of the IF protocol to a set of 15 different species of non-Legionella bacterial strains obtained from clinical and environmental collections (Table 1). Assays were performed using the protocol described above for Legionella strains. The specificity of the L. pneumophila antibody from Accurate Chemical & Scientific was additionally tested on L. pneumophila strains of sg 1 to 15.

Statistical analysis of L. pneumophila counts in natural waters.The mean numbers of Legionella organisms in seven hot-water samples were determined from at least three replicates and were compared by Student's t test. A Pearson's correlation test was applied to direct counts and culture CFU counts for 26 samples. All analyses were performed with SPSS software (SPSS Inc., Chicago, Ill.).

Development of a staining protocol.One of the challenges in the development of direct methods for the detection of pathogens by combining fluorescent probes and solid-phase cytometry is to provide bright fluorescent signals and to optimize the signal-to-noise ratio so as to reduce the effect of nonspecific signals. Therefore, various types of staining procedures were applied and compared on cultured L. pneumophila cells (Table 2; Fig. 1).

• Open in new tab
• Download powerpoint

FIG. 1.

Micrographs of L. pneumophila sg 1 cultured cells stained by IF. (A) Anti-L. pneumophila sg 1 antibody (MAb from Dresden) detected by an FITC-conjugated secondary antibody. (B) Anti-L. pneumophila sg 1 antibody (NRCL) detected by an Alexa Fluor-647-RPE-conjugated secondary antibody.

Solid-phase cytometry allows the detection of at least two different fluorescence emission signals. Therefore, we tested and compared different staining protocols providing green or red fluorescence signals. When a red-emitting dye was used for staining, the baseline due to membrane fluorescence emission was generally too high (about 2,500 fluorescence units) to enable the discrimination of labeled cells. Various other membranes were tested, and although some had a correct baseline, they unfortunately gave rise to excessive background staining (data not shown). The brightest staining intensity (700) was obtained with Alexa Fluor-RPE 647 nm (Table 2; Fig. 1), but the background fluorescence due to nonspecific fixation was too prominent.

In contrast, FITC emission was better adapted to the use of membranes, because it results in less nonspecific fixation. We therefore continued the optimization by evaluating polyclonal antibodies and MAbs detected in the green channel. The simplest labeling procedure, using an FITC-conjugated primary antibody (Accurate Chemical & Scientific) (Table 2), resulted in weak fluorescence intensity. However, the fluorescence signal was significantly increased when amplified with an FITC-labeled secondary antibody. The unconjugated anti-L. pneumophila sg 1 antibody also provided bright staining when detected by an FITC-labeled secondary antibody (Fig. 1). Other amplification systems (biotin-streptavidin or HRP-tyramide-FITC) were evaluated, but the fluorescence was lower than that obtained with the FITC-labeled secondary antibody. Three commercially available MAbs, obtained from Maine Biotechnology Services, Research Diagnostics Inc., and the Bio-Rad Monofluo kit, and three MAbs from the Dresden panel were used with a secondary antibody. The best staining was obtained with two of the MAbs from Dresden (specific for L. pneumophila sg 1 and for sgs 2 to 6, 8 to 10, and 12 to 15); staining intensities were 400 to 500 and 300 to 400, respectively, and background levels were low (Table 2; Fig. 1). We tried to amplify the staining intensity by using either the EnVision antibody combined with tyramide-FITC or an FITC-conjugated “tertiary” antibody specific for the secondary antibody. The EnVision antibody resulted in no signal. Although the tertiary antibody amplified the staining intensity, the signal-to-noise ratio decreased due to an increase in the fluorescent background. Based on these results, the secondary FITC-conjugated antibodies were selected as the best detection method for both polyclonal antibodies and MAbs.

Specificity control.The NRCL anti-L. pneumophila sg 1 polyclonal antibody does not cross-react with other serogroups (data not shown). The Accurate Chemical & Scientific anti-L. pneumophila polyclonal antibody was found to be specific for sgs 1, 6, 7, and 12, as demonstrated by tests on L. pneumophila type strains (data not shown). The specificity of the MAbs from Dresden has already been reported for the different serogroups (10).

Specificity tests performed on 15 non-Legionella strains isolated from both clinical samples and environmental waters (Table 1) showed that the MAbs from the Dresden panel were not reactive with the strains tested. In contrast, the two polyclonal antibodies cross-reacted with several of the strains tested. Therefore, we selected the MAb protocol for further development of the assay.

Artificially contaminated water samples.The MAb labeling protocol was applied to inoculated water samples in order to further optimize the protocol and determine the sensitivity of the method. The background fluorescence was more important than that observed with cultures, due to the nonspecific fixation of antibodies to particles naturally present in waters. Various blocking reagents, such as BSA and different normal sera, and the counterstains Evans blue (Réactifs RAL), CSE/2, CSA, and CSB (Chemunex) were evaluated.

A substantial part of the nonspecific fluorescence could be removed when the following combination of blocking reagents and counterstains was applied. BSA and normal swine serum (Dako) were applied 15 min before staining. The primary antibody dilution buffer contained BSA and normal swine serum, and the secondary antibody dilution buffer contained BSA and Evans blue.

Different volumes of water (10 to 100 ml) containing 1 to 100 bacterial cells were analyzed. In all cases, a single cell could be detected independently of the volume analyzed. Two criteria defined the upper limit of the volume that could be filtered and thus the detection limit of the method: (i) the number of particles present in the volume analyzed, not only because they contribute to the background but also because they contribute to membrane warping, and (ii) the number of targeted cells present on the membrane; when this number exceeded 300, the microscope validation was difficult and time-consuming. However, this issue did not affect the efficiency of the method, since it can be solved by filtering a lower volume.

Detection of L. pneumophila in natural water samples.A total of 26 samples were used to compare cytometry counts with the standard culture method. Seven water samples were collected from the hot water systems of two hospitals in Lyon, France, and analyzed by both cytometry and culture in three replicates, in order to determine the reproducibility of each method. Nineteen other samples were collected from four hospitals as part of the water surveillance scheme at the NRCL. These samples were analyzed in a single analysis by the two detection methods.

Nineteen samples (73.1%) were positive with ChemScanRDI, and 16 (61.5%) were positive in culture (Table 3). Three samples were positive in ChemScanRDI but negative in culture, whereas the opposite trend was never found. The concentration of L. pneumophila organisms enumerated by colony count ranged from 1.2 × 10² to 1.1 × 10⁴L. pneumophila CFU/liter, and the numbers of L. pneumophila organisms quantified by cytometry ranged from 1.4 × 10³ to 4.1 × 10⁶ per liter (Table 3). The concentration reported in the three positive samples found only by cytometry ranged from 1.9 × 10³ to 4.1 × 10⁶L. pneumophila organisms per liter. The ratios of direct counts to CFU counts varied in the range of 1.4 to 38, although two very high ratios were found (148 and 325).

The results from the samples analyzed in triplicate were included in a Pearson's correlation test. When all samples were included, no correlation was observed. However, we eliminated a sample (sample 7) containing a very high level of nonculturable L. pneumophila organisms (10⁶ per liter), and we further omitted one of the three replicates of sample 1 from the analysis, since it was quite different from the other two replicates. With these modified data (37 points, including single samples and replicates), a significant correlation (coefficient of correlation, 0.46; P < 0.05) was found between the counts obtained by the two methods.