CURRENT MICROBIOLOGYVol. 25 (1992), pp. 19-23
Current Microbiology 9 Springer-Verlag New York Inc, 1992
Comparison of Culture Methods and an Immunofluorescence Assay for the Detection of Legionella pneumophila in Domestic Hot Water Devices Michel Alary 1 and Jean R. Joly 2 IDepartments of Social and Preventive Medicine and 2Microbiology, Facult6 de M6decine, Universit6 Laval, Qu6bec, Canada
Abstract. The objective of this study was to compare an indirect immunofluorescence assay
with culture methods for the identification of Legionella pneumophila serogroups 1 to 6 in hot water samples taken from domestic environments. Hot water samples were obtained from the water heater, the shower heads, and the most frequently used faucet of 211 private houses. Concentrated water samples were inoculated on buffered charcoal yeast extract agar (BCYE) and on a semi-selective culture medium (GPV). Colonies with a morphology similar to that of Legionellaceae were subcultured on BCYE and on blood agar plates; those that grew on the former but not the latter were further characterized and identified by direct immunofluorescence techniques. The concentrated samples were also smeared on multiple-well microscope slides and tested by indirect immunofluorescence with monoclonal antibodies against L. pneumophila, serogroups 1 to 6. Of the houses studied, 30% were found to contain culturable L. pneumophila in at least one water sample, whereas 63% were positive by indirect immunofluorescence. The sensitivity of this assay compared with culture varied from 16.7-21.1%, and its specificity was between 76.7% and 88.3% depending on the sample source (water heater, shower heads, or faucet). In the 38 houses with at least one positive sample found by both immunofluorescence and culture, total or partial agreement between serogroups identified by both techniques was only 34%. The results obtained in this study strongly suggest that indirect immunofluorescence is not an adequate alternative for the identification of L. pneumophila in hot water systems.
Initial studies examining water samples for the presence of Legionellaceae relied heavily on animal inoculation (guinea pig) for isolation [9, 17]. These techniques were meticulous, expensive, and time consuming. They were, however, essential, since semi-selective culture media were not available and Legionellaceae were readily overgrown by other environmental bacteria [5]. These early procedures appeared to lack sensitivity [15]. Although a few authors proposed methods to improve the sensitivity of animal inoculation [15, 20], these procedures became obsolete with the development of semi-selective media. Fitzgeorge and Dennis (1983) [8] showed that direct plating of water samples on semi-selective media was the method of choice for the isolation of Legionella pneumophila from water samples and that the recovery of these bacteria was independent of the water source from which the sample was
taken. The two semi-selective media used in that study were described byEdelstein (1981) [4] and by Wadowsky and Yee (1981) [22]. These media were based on the charcoal yeast extract medium described by Feeley et al. (1980) [7] and subsequently improved by Pasculle et al. (1980) [18]. Results obtained by Roberts et al. (1987) [19] suggest that the use of different culture media might be necessary to recover maximum numbers of Legionellaceae. Direct immunofluorescence techniques have been used to identify the presence of Legionellaceae in water samples obtained from aquatic environments [9]. However, this technique has not been largely adopted by investigators because of potential specificity problems. Indeed, cross-reactions with other bacteria and nonspecific fluorescence may occur with the direct fluorescence assay [2, 6, 10]. Recently, an indirect immunofluorescence assay (IFA) has been proposed to overcome this potential
Address requests to: Dr. Jean R. Joly, Groupe de recherche en ~pid6miologie, H6pital du Saint-Sacrement, 1050, chemin Ste-Foy, Qu6bec, Qc, Canada, G1S 4L8.
20 specificity problem in man-made environments [2, 3]. According to these authors, nonspecific background fluorescence is reduced with IFA in comparison with the direct assay. Finally, Vickers recently reported that a diagnostic monoclonal antibody failed to detect L. pneumophila in environmental samples [21]. The objective of this study was to compare IFA by use of monoclonal antibodies, to culture for the identification of L. pneumophila in hot water samples taken from the domestic environment. Materials and Methods Sampling within the domestic environment. Between August and December 1989, 211 private houses were selected for this study. Methods of selection have been previously described [1]. In each selected house, water samples were taken from the water heater, the shower heads and the most frequently used faucet. For the water heater, the first 500 ml of water obtained from the drainage valve was used. Similar samples were taken from all shower heads and from the selected faucet. A swab of inner parts of shower heads was also taken. These swabs were transported to the laboratory in 50-ml conical tubes that contained approximately 40 ml of water obtained from the shower that was sampled. Sterile bottles were used throughout the study. Water samples were transported to the laboratory and stored at 4~ for a maximum of 12 h prior to laboratory procedures. Culture and isolation of Legionellapneumophila. The media used in this study were buffered charcoal yeast extract medium supplemented with c~-ketoglutarate (BCYE [18]) and BCYE supplemented with 3 mg of glycine/ml, 5 /xg of vancomycin/ml, and 100 U of polymixin B/ml (GPV [22]). Each lot of culture medium was tested for sterility prior to use and was also evaluated for its ability to support the growth of a wild strain of L. pneumophila. Each water sample was concentrated 100-fold by centrifugation for 30 rain at 5000 g. Thereafter, 100/xl of the concentrated sample and five tenfold dilutions were inoculated on the two culture media. Plates were incubated under normal atmosphere at 35~ -+ 2~ for 7 days and were examined at days 1, 5, and 7. Bacteria were counted on day 7. Colonies with a morphology similar to that of Legionellaceae were subcultured on BCYE and on blood agar plates; those that grew on the former but not the latter medium were further characterized and tested by a direct immunofluorescent antibody technique previously described [ 13]. Swabs from the interior of shower heads were processed as above, except that the final concentration obtained was 16-fold. Bacterial counts were adjusted accordingly. Indirect immunofluorescence assay. Monoclonal antibodies against L. pneumophila serogroups 1-6 were used for indirect immunofluorescence in this study. These have previously been described [12] and shown to be serogroup specific [11]. The analysis was restricted to these serogroups because the specificity of monoclonal antibodies against other serogroups has not been clearly established [11]. Concentrated water samples (5 /~l/well) were smeared on multiple-well microscope slides (40 wells of 2-mm diameter per slide) and allowed to dry. They were heat-fixed and placed in 1% (vol/vol) formalin in water for 10 rain. Staining was performed by
CURRENT MICROBIOLOGY Vol. 25 (1992)
placing 5 ~1 of the appropriate serogroup-specific monoclonal antibody diluted in phosphate-bfiffered saline (PBS) in each well. After 30 rain, slides were washed in PBS and stained with a rabbit anti-mouse IgG antiserum labeled with fluorescein, diluted in PBS containing 0.02% of Evan's blue. Slides were washed in PBS after 30 min, dried, and mounted. Each smear was examined for 5 min (time required for the entire well to be visualized) with an epiillumination fluorescence microscope (x400). Positive controls (L. pneumophila serogroups 1-6) were used throughout all of these experiments. Negative controls employing Pseudomonas aeruginosa were also used in this study. In order to determine whether Legionella might have been inadvertently placed on slides during the manufacturing/cleaning process, slides without any water samples were also stained. All reagents that were used in these assays were prepared with filtersterilized distilled water. Statistical analysis. Indirect immunofluorescence assay (IFA) was compared with culture for its ability (in terms of sensitivity and specificity) to identify the presence ofL. pneumophila serogroups 1-6. In addition, when Legionellae were found by both techniques, the concordance between identified serogroups was examined. Finally, because the volume of water examined by IFA was smaller than that inoculated on culture media, sensitivity of IFA was also examined over the quartile distribution of the highest bacterial counts obtained on culture media. Chi-square was used for statistical testing in this analysis.
Results Culture and immunofluorescence for samples from the houses. In total, 970 samples were taken from 211 houses. Samples from the water heater and from the most frequently used faucet were taken in each house, whereas 274 water samples and a similar number of swabs were obtained from showers in 209 houses (two of the 211 did not have any shower, 57 had two showers, and four had three showers). In total, 63 houses (29.9%) had at least one positive sample for L. pneumophila serogroups I-6 by culture. Water heaters were colonized in 61 houses (28.9%). Hot water obtained from faucets was positive in 19 houses (9.0%). The proportion of colonized water samples from shower heads was 21 of 274 (7.7%), whereas only six of the 274 swabs were culture positive (2.2%). Samples from 209 houses were also examined by IFA. In addition to the two houses where IFA was not done on any sample (both houses had two showers), samples from three other faucets were not tested by IFA. According to these assays, 132 houses (63.2%) had at least one sample positive for L. pneumophila serogroups 1-6. These bacteria were identified by IFA in 17.7% (37 of 209) of the water heaters, whereas corresponding proportions
M. Alary and J.R. Joly:
21
Legionella pneumophila in Hot Water
Table 1. Sensitivity and specificity of indirect immunofiuorescence when compared with culture in samples from different water devices Shower heads
Fluorescence Positive Negative Total Sensitivity Specificity
Water heaters
Faucets
Water
Swabs
Culture
Culture
Culture
Culture
Positive
Negative
Positive
Negative
Positive
Negative
Positive
Negative
11 49 60a
26 122 148
4 15 19
36 150 186
4 17 21
58 191 249
1 5 6
31 233 264
18.3% 82.4%
21.1% 80.6%
19.0% 76.7%
16.7% 88.3%
One sample not tested by IFA was positive by culture.
were 19.4% (40 of 206) for faucets, 23.0% (62 of 270) for water samples from shower heads, and 11.9% (32 of 270) for swabs.
Table 2. Sensitivity and specificity of indirect immunofluorescence compared with culture when contamination of at least one site in a given house was found Culture
Comparison of culture and indirect immunofluorescence. Although !FA was performed on 955 of the
970 samples, this analysis was restricted to 953 samples. Legionellaceae strains isolated from two water samples (one from a water heater and one from a faucet) were lost prior to speciation. Since the study was restricted to L. pneumophila serogroups 1-6, these samples were excluded from comparisons between culture and IFA. The sensitivity of IFA compared with culture was between 16.7% and 21.1% for specific sampling sites, whereas specificity varied between 76.9% and 88.3% (Table 1). However, when the ability ofimmunofluorescence to detect contamination by Legionellaceae at any site in a given house was considered, its sensitivity compared with culture was 61.3%, whereas its specificity was 36.3% (Table 2). In the 38 houses with at least one positive sample by IFA and culture, the same serotypes were identified by both methods in four instances (10.5%), whereas partial concordance (identification of the same serotype but also presence of other serotypes either by culture or by fluorescence or by both) was found in nine (23.7%). Thus, there was partial or total agreement in only 34.2% of the houses found positive by both techniques. The corresponding proportions were 45.5% (n = 11) for the water heater, 25% (n = 4) for the faucets, and 40% (n = 5) for scrapings or water samples from shower heads. There was no statistically significant difference between the different measures of sensitivity of IFA
Fluorescence Positive Negative Total Sensitivity Specificity
Positive
Negative
38 24 62"
93 53 146
Total
131 77 208 h
61.3% 36.3%
One house with at least one positive sample by culture was not tested by IFA. b One house was excluded from this comparison because the only strain of Legionella isolated from it by culture was lost prior to identification.
performed over the quartiles of bacterial counts (Table 3). However, sensitivity was slightly higher when bacterial counts were in the highest quartile (29.6%), but even when this sensitivity was compared with that of all other quartiles combined, statistical significance was not achieved (chi-square 1 d.f. = 2.74, p-value = 0.10).
Discussion This study shows that L. pneumophila serogroups 1-6 are common in domestic hot water devices. Indeed, in nearly 30% of all domestic water heaters we were able to recover these bacteria. In addition, over 13% of the houses studied were positive by culture
22
CURRENT MICROBIOLOGY Vol. 25 (1992)
Table 3. Sensitivity of IFA according to the quartile distribution of bacterial counts (per 100/xl of concentrated sample) in 106 samples positive by culture for L. pneumophila serogroups 1-6
Bacterial counts Quartile 1 Quartile 2 Quartile 3 Quartile 4
Range
Median within quartile
N
IFA+
Sensitivity a (%)
1-3 4-100 101-1660 > 1660
2 20 730 4760
26 27 26 27
5 4 3 8
19.2 14.8 11.5 29.6
a Chi-square 3 d.f. = 3.25; p-value = 0.36.
for these bacteria in at least one sample taken at the faucet or the shower. The contamination of the domestic potable water environment reached 63% when only IFA results were considered. Such a high proportion of positive samples could have been expected with direct immunofluorescence, since cross-reactions with other bacteria and nonspecific fluorescence have previously been suggested [2, 6, 10]. In our study, although the poor specificity of IFA when compared with culture may be attributable to a lack of sensitivity of the latter technique, its low sensitivity suggests a poor reliability of IFA. Indeed, in water heaters, only 11 of the 60 culture-positive samples were also positive by IFA. In faucets and shower heads, these figures were, respectively, four of 19 and five of 27. Although the total amount of water examined by culture or by IFA was different, there was no significant increase in the sensitivity of the latter technique when bacterial counts were higher. Sensitivity of IFA was only 30% when the expected number of bacteria present in the sample was at least 80 viable cells per 5/xl of centrifugated water. Finally, even when devices with positive samples by IFA and by culture were considered, concordance between serotypes identified by both methods was very poor. Using an L. pneurnophila serogroup 1-specific monoclonal antibody, Colbourne et al. found a high specificity for IFA [2, 3]. Indeed, in a large proportion of culture-negative, IFA-positive samples, L. pneumophila could be recovered by heat shock. Because we did not use heat shock in this study, it is extremely difficult to compare our specificity results with those obtained by these authors. However, in their study, only 2% of faucets were positive by culture, whereas in our study 9% were positive. This higher proportion, as well as frequently positive cultures on other samples (heater, shower heads), allowed us to evaluate more precisely the sensitivity of IFA (between 16.7% and 21.1% for individual
sites). Additional differences between both studies also make comparisons difficult. Hence, their study was restricted to L. pneumophila serogroup 1, whereas ours included the first six serogroups of L. pneumophila. However, when we restricted our analysis to L. pneumophila serogroup 1, our results remained essentially identical (sensitivity of 10% and specificity of 63% when contamination of at least one water sample in a given house was considered). Finally, these differences could also be attributable to the use of different monocional antibodies, which may not react with all strains of L. pneumophila serogroup 1. However, in a previous study we showed that the monoclonal antibodies used in this study correctly identified 309 of 312 strains of L. pneumophila serogroup 1 [11]. In a recent study, Vickers could not identify Legionella pneumophila by direct immunofluorescence assay (DFA) with a commercial monoclonal antibody [21]. These authors hypothesized that this could be owing to the epitope identified by the later monoclonai antibody. The antibodies used in our study react with the lipopolysaccharides of Legionella pneumophila, the antigens recognized by the other conjugates studied by Vickers et al. Thus, our results cannot be compared with those obtained by these authors. The description of polymerase chain reaction (PCR) applied to Legionellaceae is very recent [16]. In addition, PCR technology for these bacteria was used under very strict laboratory conditions but has not yet been used under field conditions, where its performance is totally unknown. However, the application of this method to our samples would only have added a new set of data for the purpose of comparisons. Culture, by definition, is specific, and a positive result has frequently been used as the gold standard. More recently, positivity by multiple nonculture tests was occasionally used as the gold standard for test comparisons. This is especially true for fastidious organisms such as Chlamydia tracho-
M. Alary and J.R. Joly: Legionella pneumophila in Hot Water
23
matis [14]. A similar approach could be used in fur-
mophila from water supplies: comparison of methods based
ther studies of laboratory methods for the detection of Legionellaceae in environmental samples. The exact sensitivity of culture techniques for the detection of Legionellaceae in man-made water environments remains unknown and is probably far from ideal. However, the results obtained in this study strongly suggest that IFA is not an adequate alternative for the identification of L. pneumophila in hot water systems.
on the guinea-pig and culture media. J Hyg Camb 91:179-187 9. Fliermans CB, Cherry WB, Orrison LH, Thacker L (1979) Isolation of Legionella pneumophila from non-epidemic related aquatic habitats. Appl Environ Microbiol 37:1239-1242 10. Flournoy DJ, Belobraydic KA, Silverg SL, Lawrence CH, Guthrie PJ (1988) False positive Legionella pneumophila direct immunofluorescent monoclonal antibody test caused by Bacillus cereus spores. Diagn Microbiol Infect Dis 9:123-125 11. Joly JR, Ramsay D (1985) Use of monoclonal antibodies in the diagnosis and epidemiologic studies of legionellosis. Clinics Lab Med 5:561-574 12. Joly JR, Chert YY, Ramsay D (1983) Serogrouping and subtyping of Legionella pneumophila with monoclonal antibodies. J Clin Microbiol 18:1040-1046 13. Joly JR, Boissinot M, Duchaine J, Dural M, Rafrafi J, Ramsay D, Letarte R (1984) Ecological distribution of Legionellaceae in the Quebec city area. Can J Microbiol 30:63-67 14. Lefebvre J, Lapierre H, Rousseau H, Masse R. (1988) Comparison of three techniques for detection of Chlamydia trachomatis in endocervical specimens from asymptomatic women. J Clin Microbiol 26:728-731 15. Leinbach ED, Winkler HH, Wood DO, Coggin JH (1983) Improved facility and sensitivity in the use of guinea pigs for the isolation of Legionella pneumophila from cooling tower water. Appl Environ Microbiol 45:1119-1121 16. Mahbubani MH, Bej AK, Miller R, Haff L, Dicesare J, Atlas RM (1990) Detection of Legionella with polymerase chain reaction and gene probe methods. Mol Cell Probes 4:175-187 17. Neil MA, Gorman GW, Gilbert C, Roussel A, Hightower AW, McKinney RM, Broome CV (1985) Nosocomial legionellosis, Paris, France. Evidence of transmission by potable water. Am J Med 78:581-588 18. Pasculle AW, Feeley JC, Gibson RJ, Cordes LG, Myerowitz RL, Patton CM, Gorman GW, Carmack CL, Ezzel JW, Dowling JN (1980) Pittsburg pneumonia agent: direct isolation from human lung tissue. J Infect Dis 141:727-732 19. Roberts KP, August CM, Nelson JD (1987) Relative sensitivities of environmental Legionellae to selective isolation procedures. Appl Environ Microbiol 53:2704-2707 20. Thorpe TC, Miller RD (1980) Negative enrichment procedure for isolation of Legionella pneumophila from seeded cooling tower water. Appl Environ Microbiol 40:849-851 21. Vickers RM, Stout JE, Yu VL (1990) Failure of a diagnostic monoclonal immunofluorescent reagent to detect Legionella pneumophila in environmental samples. Appl Environ Microbiol 56:2912-2914 22. Wadowsky RM, Yee RB (1981) Glycine-containing selective medium for isolation of Legionella pneumophila from environmental samples. Appl Environ Microbiol 42:768-772
ACKNOWLEDGMENTS
The authors thank Nicole Garneau for collection of the water samples and Danielle Ramsay for laboratory procedures. This study was funded by the Canadian Electric Association (CEA grant no. 818U670).
Literature Cited 1. Alary M, Joly JR (1991) Risk factors for contamination of domestic hot water systems by Legionelleae. Appl Environ Microbiol 57:2360-2367. 2. Colbourne JS, Dennis PJ (1989) The ecology and survival of Legionella pneurnophila. J Inst Water Environ Man 3:345-350 3. Colbourne JS, Dennis PJ, Trew RM, Berry C, Vesey G (1988) Legionella and public water supplies. Water Sci Technol 20:5-10. 4. Edelstein PH (1981) Improved semi-selective medium for isolation ofLegionella pneumophila from contaminated clinical and environmental specimens. J Clin Microbiol 14:298-303 5. Edelstein PH (1982) Comparative study of selective media for isolation of Legionella pneumophila from potable water. J Clin Microbiol 16:697-699 6. Edelstein PH, Mckinney RM, Meyer RD, Edelstein MA, Krause JC (1980) Immunologic diagnosis of Legionnaires disease: cross reactions with anaerobic and microaerophilic organisms and infections caused by them. J Infect Dis 141:652-655 7. Feeley JC, Gibson RJ, Gorman GW, Lanford NC, Rasheed JK, Mackel DC, Baine WB (1980) Charcoal-yeast extract agar: primary isolation medium for Legionella pneumophila. J Clin Microbiol 10:437-441
8. Fitzgeorge RB, Dennis PJ (1983) Isolation of Legionella pneu-
Current Microbiology 9 Springer-Verlag New York Inc, 1992
Comparison of Culture Methods and an Immunofluorescence Assay for the Detection of Legionella pneumophila in Domestic Hot Water Devices Michel Alary 1 and Jean R. Joly 2 IDepartments of Social and Preventive Medicine and 2Microbiology, Facult6 de M6decine, Universit6 Laval, Qu6bec, Canada
Abstract. The objective of this study was to compare an indirect immunofluorescence assay
with culture methods for the identification of Legionella pneumophila serogroups 1 to 6 in hot water samples taken from domestic environments. Hot water samples were obtained from the water heater, the shower heads, and the most frequently used faucet of 211 private houses. Concentrated water samples were inoculated on buffered charcoal yeast extract agar (BCYE) and on a semi-selective culture medium (GPV). Colonies with a morphology similar to that of Legionellaceae were subcultured on BCYE and on blood agar plates; those that grew on the former but not the latter were further characterized and identified by direct immunofluorescence techniques. The concentrated samples were also smeared on multiple-well microscope slides and tested by indirect immunofluorescence with monoclonal antibodies against L. pneumophila, serogroups 1 to 6. Of the houses studied, 30% were found to contain culturable L. pneumophila in at least one water sample, whereas 63% were positive by indirect immunofluorescence. The sensitivity of this assay compared with culture varied from 16.7-21.1%, and its specificity was between 76.7% and 88.3% depending on the sample source (water heater, shower heads, or faucet). In the 38 houses with at least one positive sample found by both immunofluorescence and culture, total or partial agreement between serogroups identified by both techniques was only 34%. The results obtained in this study strongly suggest that indirect immunofluorescence is not an adequate alternative for the identification of L. pneumophila in hot water systems.
Initial studies examining water samples for the presence of Legionellaceae relied heavily on animal inoculation (guinea pig) for isolation [9, 17]. These techniques were meticulous, expensive, and time consuming. They were, however, essential, since semi-selective culture media were not available and Legionellaceae were readily overgrown by other environmental bacteria [5]. These early procedures appeared to lack sensitivity [15]. Although a few authors proposed methods to improve the sensitivity of animal inoculation [15, 20], these procedures became obsolete with the development of semi-selective media. Fitzgeorge and Dennis (1983) [8] showed that direct plating of water samples on semi-selective media was the method of choice for the isolation of Legionella pneumophila from water samples and that the recovery of these bacteria was independent of the water source from which the sample was
taken. The two semi-selective media used in that study were described byEdelstein (1981) [4] and by Wadowsky and Yee (1981) [22]. These media were based on the charcoal yeast extract medium described by Feeley et al. (1980) [7] and subsequently improved by Pasculle et al. (1980) [18]. Results obtained by Roberts et al. (1987) [19] suggest that the use of different culture media might be necessary to recover maximum numbers of Legionellaceae. Direct immunofluorescence techniques have been used to identify the presence of Legionellaceae in water samples obtained from aquatic environments [9]. However, this technique has not been largely adopted by investigators because of potential specificity problems. Indeed, cross-reactions with other bacteria and nonspecific fluorescence may occur with the direct fluorescence assay [2, 6, 10]. Recently, an indirect immunofluorescence assay (IFA) has been proposed to overcome this potential
Address requests to: Dr. Jean R. Joly, Groupe de recherche en ~pid6miologie, H6pital du Saint-Sacrement, 1050, chemin Ste-Foy, Qu6bec, Qc, Canada, G1S 4L8.
20 specificity problem in man-made environments [2, 3]. According to these authors, nonspecific background fluorescence is reduced with IFA in comparison with the direct assay. Finally, Vickers recently reported that a diagnostic monoclonal antibody failed to detect L. pneumophila in environmental samples [21]. The objective of this study was to compare IFA by use of monoclonal antibodies, to culture for the identification of L. pneumophila in hot water samples taken from the domestic environment. Materials and Methods Sampling within the domestic environment. Between August and December 1989, 211 private houses were selected for this study. Methods of selection have been previously described [1]. In each selected house, water samples were taken from the water heater, the shower heads and the most frequently used faucet. For the water heater, the first 500 ml of water obtained from the drainage valve was used. Similar samples were taken from all shower heads and from the selected faucet. A swab of inner parts of shower heads was also taken. These swabs were transported to the laboratory in 50-ml conical tubes that contained approximately 40 ml of water obtained from the shower that was sampled. Sterile bottles were used throughout the study. Water samples were transported to the laboratory and stored at 4~ for a maximum of 12 h prior to laboratory procedures. Culture and isolation of Legionellapneumophila. The media used in this study were buffered charcoal yeast extract medium supplemented with c~-ketoglutarate (BCYE [18]) and BCYE supplemented with 3 mg of glycine/ml, 5 /xg of vancomycin/ml, and 100 U of polymixin B/ml (GPV [22]). Each lot of culture medium was tested for sterility prior to use and was also evaluated for its ability to support the growth of a wild strain of L. pneumophila. Each water sample was concentrated 100-fold by centrifugation for 30 rain at 5000 g. Thereafter, 100/xl of the concentrated sample and five tenfold dilutions were inoculated on the two culture media. Plates were incubated under normal atmosphere at 35~ -+ 2~ for 7 days and were examined at days 1, 5, and 7. Bacteria were counted on day 7. Colonies with a morphology similar to that of Legionellaceae were subcultured on BCYE and on blood agar plates; those that grew on the former but not the latter medium were further characterized and tested by a direct immunofluorescent antibody technique previously described [ 13]. Swabs from the interior of shower heads were processed as above, except that the final concentration obtained was 16-fold. Bacterial counts were adjusted accordingly. Indirect immunofluorescence assay. Monoclonal antibodies against L. pneumophila serogroups 1-6 were used for indirect immunofluorescence in this study. These have previously been described [12] and shown to be serogroup specific [11]. The analysis was restricted to these serogroups because the specificity of monoclonal antibodies against other serogroups has not been clearly established [11]. Concentrated water samples (5 /~l/well) were smeared on multiple-well microscope slides (40 wells of 2-mm diameter per slide) and allowed to dry. They were heat-fixed and placed in 1% (vol/vol) formalin in water for 10 rain. Staining was performed by
CURRENT MICROBIOLOGY Vol. 25 (1992)
placing 5 ~1 of the appropriate serogroup-specific monoclonal antibody diluted in phosphate-bfiffered saline (PBS) in each well. After 30 rain, slides were washed in PBS and stained with a rabbit anti-mouse IgG antiserum labeled with fluorescein, diluted in PBS containing 0.02% of Evan's blue. Slides were washed in PBS after 30 min, dried, and mounted. Each smear was examined for 5 min (time required for the entire well to be visualized) with an epiillumination fluorescence microscope (x400). Positive controls (L. pneumophila serogroups 1-6) were used throughout all of these experiments. Negative controls employing Pseudomonas aeruginosa were also used in this study. In order to determine whether Legionella might have been inadvertently placed on slides during the manufacturing/cleaning process, slides without any water samples were also stained. All reagents that were used in these assays were prepared with filtersterilized distilled water. Statistical analysis. Indirect immunofluorescence assay (IFA) was compared with culture for its ability (in terms of sensitivity and specificity) to identify the presence ofL. pneumophila serogroups 1-6. In addition, when Legionellae were found by both techniques, the concordance between identified serogroups was examined. Finally, because the volume of water examined by IFA was smaller than that inoculated on culture media, sensitivity of IFA was also examined over the quartile distribution of the highest bacterial counts obtained on culture media. Chi-square was used for statistical testing in this analysis.
Results Culture and immunofluorescence for samples from the houses. In total, 970 samples were taken from 211 houses. Samples from the water heater and from the most frequently used faucet were taken in each house, whereas 274 water samples and a similar number of swabs were obtained from showers in 209 houses (two of the 211 did not have any shower, 57 had two showers, and four had three showers). In total, 63 houses (29.9%) had at least one positive sample for L. pneumophila serogroups I-6 by culture. Water heaters were colonized in 61 houses (28.9%). Hot water obtained from faucets was positive in 19 houses (9.0%). The proportion of colonized water samples from shower heads was 21 of 274 (7.7%), whereas only six of the 274 swabs were culture positive (2.2%). Samples from 209 houses were also examined by IFA. In addition to the two houses where IFA was not done on any sample (both houses had two showers), samples from three other faucets were not tested by IFA. According to these assays, 132 houses (63.2%) had at least one sample positive for L. pneumophila serogroups 1-6. These bacteria were identified by IFA in 17.7% (37 of 209) of the water heaters, whereas corresponding proportions
M. Alary and J.R. Joly:
21
Legionella pneumophila in Hot Water
Table 1. Sensitivity and specificity of indirect immunofiuorescence when compared with culture in samples from different water devices Shower heads
Fluorescence Positive Negative Total Sensitivity Specificity
Water heaters
Faucets
Water
Swabs
Culture
Culture
Culture
Culture
Positive
Negative
Positive
Negative
Positive
Negative
Positive
Negative
11 49 60a
26 122 148
4 15 19
36 150 186
4 17 21
58 191 249
1 5 6
31 233 264
18.3% 82.4%
21.1% 80.6%
19.0% 76.7%
16.7% 88.3%
One sample not tested by IFA was positive by culture.
were 19.4% (40 of 206) for faucets, 23.0% (62 of 270) for water samples from shower heads, and 11.9% (32 of 270) for swabs.
Table 2. Sensitivity and specificity of indirect immunofluorescence compared with culture when contamination of at least one site in a given house was found Culture
Comparison of culture and indirect immunofluorescence. Although !FA was performed on 955 of the
970 samples, this analysis was restricted to 953 samples. Legionellaceae strains isolated from two water samples (one from a water heater and one from a faucet) were lost prior to speciation. Since the study was restricted to L. pneumophila serogroups 1-6, these samples were excluded from comparisons between culture and IFA. The sensitivity of IFA compared with culture was between 16.7% and 21.1% for specific sampling sites, whereas specificity varied between 76.9% and 88.3% (Table 1). However, when the ability ofimmunofluorescence to detect contamination by Legionellaceae at any site in a given house was considered, its sensitivity compared with culture was 61.3%, whereas its specificity was 36.3% (Table 2). In the 38 houses with at least one positive sample by IFA and culture, the same serotypes were identified by both methods in four instances (10.5%), whereas partial concordance (identification of the same serotype but also presence of other serotypes either by culture or by fluorescence or by both) was found in nine (23.7%). Thus, there was partial or total agreement in only 34.2% of the houses found positive by both techniques. The corresponding proportions were 45.5% (n = 11) for the water heater, 25% (n = 4) for the faucets, and 40% (n = 5) for scrapings or water samples from shower heads. There was no statistically significant difference between the different measures of sensitivity of IFA
Fluorescence Positive Negative Total Sensitivity Specificity
Positive
Negative
38 24 62"
93 53 146
Total
131 77 208 h
61.3% 36.3%
One house with at least one positive sample by culture was not tested by IFA. b One house was excluded from this comparison because the only strain of Legionella isolated from it by culture was lost prior to identification.
performed over the quartiles of bacterial counts (Table 3). However, sensitivity was slightly higher when bacterial counts were in the highest quartile (29.6%), but even when this sensitivity was compared with that of all other quartiles combined, statistical significance was not achieved (chi-square 1 d.f. = 2.74, p-value = 0.10).
Discussion This study shows that L. pneumophila serogroups 1-6 are common in domestic hot water devices. Indeed, in nearly 30% of all domestic water heaters we were able to recover these bacteria. In addition, over 13% of the houses studied were positive by culture
22
CURRENT MICROBIOLOGY Vol. 25 (1992)
Table 3. Sensitivity of IFA according to the quartile distribution of bacterial counts (per 100/xl of concentrated sample) in 106 samples positive by culture for L. pneumophila serogroups 1-6
Bacterial counts Quartile 1 Quartile 2 Quartile 3 Quartile 4
Range
Median within quartile
N
IFA+
Sensitivity a (%)
1-3 4-100 101-1660 > 1660
2 20 730 4760
26 27 26 27
5 4 3 8
19.2 14.8 11.5 29.6
a Chi-square 3 d.f. = 3.25; p-value = 0.36.
for these bacteria in at least one sample taken at the faucet or the shower. The contamination of the domestic potable water environment reached 63% when only IFA results were considered. Such a high proportion of positive samples could have been expected with direct immunofluorescence, since cross-reactions with other bacteria and nonspecific fluorescence have previously been suggested [2, 6, 10]. In our study, although the poor specificity of IFA when compared with culture may be attributable to a lack of sensitivity of the latter technique, its low sensitivity suggests a poor reliability of IFA. Indeed, in water heaters, only 11 of the 60 culture-positive samples were also positive by IFA. In faucets and shower heads, these figures were, respectively, four of 19 and five of 27. Although the total amount of water examined by culture or by IFA was different, there was no significant increase in the sensitivity of the latter technique when bacterial counts were higher. Sensitivity of IFA was only 30% when the expected number of bacteria present in the sample was at least 80 viable cells per 5/xl of centrifugated water. Finally, even when devices with positive samples by IFA and by culture were considered, concordance between serotypes identified by both methods was very poor. Using an L. pneurnophila serogroup 1-specific monoclonal antibody, Colbourne et al. found a high specificity for IFA [2, 3]. Indeed, in a large proportion of culture-negative, IFA-positive samples, L. pneumophila could be recovered by heat shock. Because we did not use heat shock in this study, it is extremely difficult to compare our specificity results with those obtained by these authors. However, in their study, only 2% of faucets were positive by culture, whereas in our study 9% were positive. This higher proportion, as well as frequently positive cultures on other samples (heater, shower heads), allowed us to evaluate more precisely the sensitivity of IFA (between 16.7% and 21.1% for individual
sites). Additional differences between both studies also make comparisons difficult. Hence, their study was restricted to L. pneumophila serogroup 1, whereas ours included the first six serogroups of L. pneumophila. However, when we restricted our analysis to L. pneumophila serogroup 1, our results remained essentially identical (sensitivity of 10% and specificity of 63% when contamination of at least one water sample in a given house was considered). Finally, these differences could also be attributable to the use of different monocional antibodies, which may not react with all strains of L. pneumophila serogroup 1. However, in a previous study we showed that the monoclonal antibodies used in this study correctly identified 309 of 312 strains of L. pneumophila serogroup 1 [11]. In a recent study, Vickers could not identify Legionella pneumophila by direct immunofluorescence assay (DFA) with a commercial monoclonal antibody [21]. These authors hypothesized that this could be owing to the epitope identified by the later monoclonai antibody. The antibodies used in our study react with the lipopolysaccharides of Legionella pneumophila, the antigens recognized by the other conjugates studied by Vickers et al. Thus, our results cannot be compared with those obtained by these authors. The description of polymerase chain reaction (PCR) applied to Legionellaceae is very recent [16]. In addition, PCR technology for these bacteria was used under very strict laboratory conditions but has not yet been used under field conditions, where its performance is totally unknown. However, the application of this method to our samples would only have added a new set of data for the purpose of comparisons. Culture, by definition, is specific, and a positive result has frequently been used as the gold standard. More recently, positivity by multiple nonculture tests was occasionally used as the gold standard for test comparisons. This is especially true for fastidious organisms such as Chlamydia tracho-
M. Alary and J.R. Joly: Legionella pneumophila in Hot Water
23
matis [14]. A similar approach could be used in fur-
mophila from water supplies: comparison of methods based
ther studies of laboratory methods for the detection of Legionellaceae in environmental samples. The exact sensitivity of culture techniques for the detection of Legionellaceae in man-made water environments remains unknown and is probably far from ideal. However, the results obtained in this study strongly suggest that IFA is not an adequate alternative for the identification of L. pneumophila in hot water systems.
on the guinea-pig and culture media. J Hyg Camb 91:179-187 9. Fliermans CB, Cherry WB, Orrison LH, Thacker L (1979) Isolation of Legionella pneumophila from non-epidemic related aquatic habitats. Appl Environ Microbiol 37:1239-1242 10. Flournoy DJ, Belobraydic KA, Silverg SL, Lawrence CH, Guthrie PJ (1988) False positive Legionella pneumophila direct immunofluorescent monoclonal antibody test caused by Bacillus cereus spores. Diagn Microbiol Infect Dis 9:123-125 11. Joly JR, Ramsay D (1985) Use of monoclonal antibodies in the diagnosis and epidemiologic studies of legionellosis. Clinics Lab Med 5:561-574 12. Joly JR, Chert YY, Ramsay D (1983) Serogrouping and subtyping of Legionella pneumophila with monoclonal antibodies. J Clin Microbiol 18:1040-1046 13. Joly JR, Boissinot M, Duchaine J, Dural M, Rafrafi J, Ramsay D, Letarte R (1984) Ecological distribution of Legionellaceae in the Quebec city area. Can J Microbiol 30:63-67 14. Lefebvre J, Lapierre H, Rousseau H, Masse R. (1988) Comparison of three techniques for detection of Chlamydia trachomatis in endocervical specimens from asymptomatic women. J Clin Microbiol 26:728-731 15. Leinbach ED, Winkler HH, Wood DO, Coggin JH (1983) Improved facility and sensitivity in the use of guinea pigs for the isolation of Legionella pneumophila from cooling tower water. Appl Environ Microbiol 45:1119-1121 16. Mahbubani MH, Bej AK, Miller R, Haff L, Dicesare J, Atlas RM (1990) Detection of Legionella with polymerase chain reaction and gene probe methods. Mol Cell Probes 4:175-187 17. Neil MA, Gorman GW, Gilbert C, Roussel A, Hightower AW, McKinney RM, Broome CV (1985) Nosocomial legionellosis, Paris, France. Evidence of transmission by potable water. Am J Med 78:581-588 18. Pasculle AW, Feeley JC, Gibson RJ, Cordes LG, Myerowitz RL, Patton CM, Gorman GW, Carmack CL, Ezzel JW, Dowling JN (1980) Pittsburg pneumonia agent: direct isolation from human lung tissue. J Infect Dis 141:727-732 19. Roberts KP, August CM, Nelson JD (1987) Relative sensitivities of environmental Legionellae to selective isolation procedures. Appl Environ Microbiol 53:2704-2707 20. Thorpe TC, Miller RD (1980) Negative enrichment procedure for isolation of Legionella pneumophila from seeded cooling tower water. Appl Environ Microbiol 40:849-851 21. Vickers RM, Stout JE, Yu VL (1990) Failure of a diagnostic monoclonal immunofluorescent reagent to detect Legionella pneumophila in environmental samples. Appl Environ Microbiol 56:2912-2914 22. Wadowsky RM, Yee RB (1981) Glycine-containing selective medium for isolation of Legionella pneumophila from environmental samples. Appl Environ Microbiol 42:768-772
ACKNOWLEDGMENTS
The authors thank Nicole Garneau for collection of the water samples and Danielle Ramsay for laboratory procedures. This study was funded by the Canadian Electric Association (CEA grant no. 818U670).
Literature Cited 1. Alary M, Joly JR (1991) Risk factors for contamination of domestic hot water systems by Legionelleae. Appl Environ Microbiol 57:2360-2367. 2. Colbourne JS, Dennis PJ (1989) The ecology and survival of Legionella pneurnophila. J Inst Water Environ Man 3:345-350 3. Colbourne JS, Dennis PJ, Trew RM, Berry C, Vesey G (1988) Legionella and public water supplies. Water Sci Technol 20:5-10. 4. Edelstein PH (1981) Improved semi-selective medium for isolation ofLegionella pneumophila from contaminated clinical and environmental specimens. J Clin Microbiol 14:298-303 5. Edelstein PH (1982) Comparative study of selective media for isolation of Legionella pneumophila from potable water. J Clin Microbiol 16:697-699 6. Edelstein PH, Mckinney RM, Meyer RD, Edelstein MA, Krause JC (1980) Immunologic diagnosis of Legionnaires disease: cross reactions with anaerobic and microaerophilic organisms and infections caused by them. J Infect Dis 141:652-655 7. Feeley JC, Gibson RJ, Gorman GW, Lanford NC, Rasheed JK, Mackel DC, Baine WB (1980) Charcoal-yeast extract agar: primary isolation medium for Legionella pneumophila. J Clin Microbiol 10:437-441
8. Fitzgeorge RB, Dennis PJ (1983) Isolation of Legionella pneu-
