INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The genus Legionella includes 42 species to date (5, 8, 16, 21, 30). Nineteen species have been recognized to be occasional human pathogens causing Legionnaires' disease and Pontiac fever. These species are Legionella anisa, L. bozemanii serogroups 1 and 2, L. cincinnatiensis, L. dumoffii, L. feeleii serogroups 1 and 2, L. gormanii, L. hackeliae serogroups 1 and 2, L. jordanis, L. lansingensis, L. longbeachae serogroups 1 and 2, L. lytica, L. maceachernii, L. micdadei, L. oakridgensis, L. parisiensis, L. pneumophila serogroups 1 to 14, L. sainthelensi serogroup 2, L. tucsonensis, and L. wadsworthii (2, 3, 11, 12, 14, 15, 18, 23, 24, 25, 29, 30). L. pneumophila is the most frequent species isolated from patients with either community- or hospital-acquired legionellosis. Other species, mainly L. micdadei, account for approximately 15% of cases of Legionella pneumonia and are more often reported in cases of Pontiac fever. However, the importance of those non-pneumophila species in human disease may be underestimated. Phenotypic tests currently in use for the identification of Legionella species, like cell wall fatty acid and ubiquinone analyses by chromatographic techniques, are relatively cumbersome and time consuming and do not allow the identification of all species. Direct fluorescent antibody typing (DFA) often leads to cross-reactions. The DNA-DNA hybridization technique is too technically demanding for routine application, and its use is limited to taxonomic studies in reference centers (2, 30).

Several recently developed DNA analysis techniques offer alternative approaches for the identification of microorganisms for which determination of phenotypic characteristics is an unsatisfactory identification method. These techniques for the identification of Legionella strains include analysis of the intergenic 16S-23S ribosomal spacer region (ISR) (8, 9, 13, 21, 22), random amplified polymorphic DNA (RAPD) analysis (1, 15, 16), and sequence analysis of the amplified mip gene (20). These three methods discriminate among the majority of Legionella species examined to date. Similar performances were obtained with molecular phenotypic techniques based on numerical analysis of whole-cell protein sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (27) or of fatty acid methyl ester (FAME) profiles (5).

Analysis of transfer DNA-intergenic length polymorphism (tDNA-ILP), based on PCR amplification of spacers between tRNA genes, initially shown by Welsh and McClelland to distinguish species of pyogenic streptococci (28), was successfully applied to species identification of bacteria belonging to the genera Acinetobacter (6), Staphylococcus (17), Listeria (26), and Streptococcus (4). In this study, we evaluated a method of tDNA-ILP analysis using laser scanning of fluorescently labeled amplicons for identification of all clinically significant species of Legionella, except for L. lytica, a species which requires amoebal cocultivation techniques.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains. Type (n = 19) and reference (n = 16) strains were purchased from the National Collections of Type Cultures (NCTC), the American Type Culture Collection (ATCC), and the Collection of Bacterial Strains of Institut Pasteur (CIP) (Table 1). Clinical (n = 31) and environmental (n = 32) strains from the Centre National de Référence des Legionella, Lyon, France, and the Centre de Référence des Legionella, Brussels, Belgium, were previously characterized by a variety of biochemical tests, including direct fluorescent antibody, cell wall composition, protein profile, RAPD, ISR, and DNA-DNA hybridization analyses as described in Table 1.

tDNA-ILP PCR assay. Bacteria were cultured on BCYE agar at 37°C for 3 days in sealed plastic bags. Genomic DNA was extracted from cells using, sequentially, lysozyme, guanidium thiocyanate, ammonium acetate, chloroform/isoamyl alcohol, isopropanol, and ethanol, as previously described (19). PCR was carried out with tDNA consensus primers T5A (5'-AGTCCGGTGCTCTAACCAACTGAG) and T3B (5'-AGGTCGCGGGTTCGAATCC) (28). Primer T5A was 5'-end-labeled with carbocyanine dye Cy5 (Pharmacia Biotech, Roosendaal, The Netherlands). Fifty microliters of PCR reaction solution contained 1.25 U of Taq polymerase (Cetus Corp., Emeryville, Calif.), 1× PCR buffer (50 mM KCl and 10 mM Tris-HCl [pH 8.3]), 1.5 mM MgCl₂, 0.5 µM concentrations for each primer, 0.2 mM concentrations for four deoxynucleoside triphosphates and 5 µl of DNA extract. The reaction mixture was overlaid with one drop of mineral oil. Amplification conditions were as previously described (4).

Analysis of tDNA-ILP. Amplification products were analyzed as previously described (4). Briefly, DNA fragments were separated by electrophoresis through acrylamide/bisacrylamide denaturing gels (ReadyMix Gel, A.L.F grade; Pharmacia) run for 4 h on an ALFexpress automated laser fluorescent DNA sequencer (Pharmacia). A fluorescein-labeled molecular marker (Cy5 Sizer 50-500; Pharmacia) was used as an external size marker. Samples including 1.25 µl of the PCR products, 5 µl of gel loading solution (Pharmacia), and 0.4 µl each of 50- and 1,000-bp internal reference standards were denatured at 100°C for 3 min in a water bath and loaded into the gel. Fluorescence densitograms were produced by using Fragment Manager software (Pharmacia), normalized by alignment first with external standards and then with internal standards and visually compared. Normalized densitogram data were exported to and analyzed with the GelCompar software (Version 4.1; Applied Maths, Kortrijk, Belgium). Matrices of Pearson product moment correlation coefficients between pairs of PCR patterns were used for construction of a dendrogram by using the unweighted-pair group method with averages (UPGMA).

Reproducibility. Duplicate bacterial lysates of each type strain were coamplified in the same PCR experiment and in separate PCR experiments. These repeat samples were analyzed in the same gel for interrun and interextract pattern reproducibility. Intergel reproducibility was assessed by analyzing one amplicon (L. pneumophila subsp. pneumophila) in three different gels. The reproducibility of normalized tDNA-ILP patterns was evaluated visually and by Pearson product moment correlation coefficients.

Computer-assisted pattern identification. The Identification module of GelCompar software was used to create a library of 35 units, each of which consisted of the tDNA-ILP pattern of a type or reference strain of Legionella. Patterns from each clinical or environmental Legionella strain were compared to all units in the library. The most probable identification proposed by using the software corresponded to the unit showing the highest similarity of Pearson coefficients with the unknown profile (best match), followed by the second and highest similarity coefficient with another unit (second best match).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Reproducibility. Interextract, interrun, and intergel comparisons of tDNA-ILP patterns showed complete reproducibility by visual analysis of fluorescence densitograms. The number and size of DNA fragments (densitogram peak positions) were highly reproducible. Small variations occurred only in amplified product concentration (peak heights). Normalized tDNA-ILP patterns showed a mean 94.8% similarity coefficient (range, 91 to 97.3%) for interextract and interrun comparisons of the 18 species. Intergel comparison of tDNA-ILP patterns showed a mean 94.4% similarity coefficient (range, 93 to 96.4%). Based on these data, the threshold of pattern identity was defined at 91% similarity (Fig. 1).

View larger version (39K): [in this window] [in a new window]   FIG. 1.   Dendrogram of tDNA-ILP pattern Pearson correlation product moment coefficient relatedness of type and reference strains of Legionella species constructed by UPGMA. The threshold of pattern identity level is indicated by a vertical dotted line. Discrepancies between visual comparison and computer analysis are indicated by an asterisk.

Visual comparison of tDNA-ILP patterns. Type and reference strains (n = 35) of the 18 species of Legionella produced 15 distinct tDNA-ILP patterns (Fig. 2). These patterns consisted of 5 to 10 DNA fragments, which varied in size between 66 and 376 bp. Easily distinguishable specific patterns were produced by the type strains of the 14 following species: L. hackeliae, L. micdadei, L. maceachernii, L. wadsworthii, L. dumoffii, L. sainthelensi, L. cincinnatiensis, L. longbeachae, L. lansingensis, L. gormanii, L. jordanis, L. feeleii, L. oakridgensis, and L. pneumophila. However, L. anisa, L. bozemanii, L. tucsonensis, and L. parisiensis displayed very similar patterns (Fig. 2). Closer analysis of the fragment sizes allowed the distinction of L. tucsonensis based on the presence of a peak of 103 bp from L. parisiensis that produced a specific peak of 105 bp. L. anisa and L. bozemanii displayed visually identical patterns (Fig. 3). Strains of different serogroups could not be distinguished within a species. Subspecies of L. pneumophila displayed similar patterns (Fig. 1).

View larger version (32K): [in this window] [in a new window]   FIG. 2.   Fluorescence densitograms of Legionella type and reference strains.

View larger version (16K): [in this window] [in a new window]   FIG. 3.   Fluorescence densitograms of L. anisa, L. bozemanii, L. tucsonensis, and L. parisiensis type strains. Dotted lines indicate DNA fragment size.

Computer-assisted analysis of tDNA-ILP patterns. At the 91% similarity cutoff level, 26 clusters were obtained with Pearson similarity coefficients of the normalized tDNA-ILP patterns of the 35 type and reference strains, which displayed between 21 and 98% similarity (Fig. 1). Non-pneumophila strains (16 type strains and 1 reference strain of L. sainthelensi serogroup 2) clustered separately in 15 clusters. L. anisa clustered in the same branch as L. bozemanii serogroup 1, as did L. cincinnatiensis and L. sainthelensi serogroup 2. Among the four non-pneumophila species, which included two serogroups, L. feeleii and L. longbeachae patterns were common to both serogroups, whereas L. hackeliae and L. bozemanii serogroups 1 and 2 could be distinguished separately, due to reproducible peak height variations between visually similar patterns.

Computer-assisted identification. All 63 clinical and environmental strains were correctly identified to the species level by matching the tDNA-ILP patterns with those of the 35 reference units, including the 2 L. anisa and the 4 L. bozemanii strains (Table 1). There was a significant difference between the best and second match (mean difference, 12%; range, 4% to 33%) (Table 1). L. feeleii, L. bozemanii, and L. longbeachae strains were further correctly identified to the serogroup level, whereas L. pneumophila strains of various serogroups were correctly identified to the species level but not to the serogroup level.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, we showed that tDNA-ILP analysis was a simple and rapid method which allowed the identification of diverse environmental and clinical strains of the 18 Legionella species described as human pathogens until today. This excellent performance was related to the method used for fragment analysis. Since tDNA spacers vary in size between species by only a few base pairs, we analyzed spacers amplified using fluorescently labeled primers and analyzed DNA fragments by PAGE on a DNA sequencer to improve both resolution and reproducibility, as previously reported (4, 6, 17). Indeed, analysis of the densitometric curves provided by the ALFexpress system was considerably easier than that of the patterns exhibited on agarose gels (17). Furthermore, the resolution of the electrophoretic profiles was occasionally affected by minor variations in the porosity of the agarose (17). Moreover, we used a sophisticated pattern analysis software to discriminate between Legionella species showing subtle variations. Computer-assisted analysis was more accurate for the identification of unknown strains by matching tDNA-ILP patterns with reference patterns, as previously discussed by Vaneechoutte et al. (26). Other molecular techniques for the identification of Legionella species also included computer analysis of results (1, 5, 15, 16, 20, 21, 27). These molecular genotypic and phenotypic methods are at least as discriminatory as tDNA-ILP analysis, including RAPD (1, 15, 16), ISR (21), sequence determination of the mip gene (20), SDS-PAGE of whole-cell proteins (27), and FAME (5) analyses. However, ISR analysis was shown to be less discriminatory in other studies which differed in the sequence of the primers used (8, 13).

Two species, L. anisa and L. bozemanii, could not easily be distinguished by tDNA-ILP analysis. These species belong to the bluish-white autofluorescent group which includes closely related species known to be difficult to distinguish (8, 10, 14). RAPD (1, 15, 16), mip sequence (20), and FAME (5) analyses easily identified these species whereas HinfI restriction fragment length polymorphism analysis of the ISR DNA fragments (21) and numerical analysis of specific portions of the SDS-PAGE protein profiles (27) were required by those techniques to reach equivalent performance. However, clinical and environmental strains of L. anisa and L. bozemanii species were correctly identified using computer matching of tDNA-ILPs, despite the visually similar profiles and their theoretical overlap based on a conservative lowest-similarity threshold derived from stringent reproducibility tests. This suggests that this general threshold may in fact have been defined too conservatively for some species. In contradiction to DNA-DNA hybridization tests (3), RAPD (16), ISR (27), and SDS-PAGE analyses of whole-cell proteins, tDNA-ILPs could not differentiate two of the type strains of the three subspecies of L. pneumophila.

The topology of the phenogram of tDNA-ILP patterns above species level did not exhibit congruence with the phylogenetic trees that were derived from sequence homologies of the 16S rRNA gene (7, 14). This discrepancy can be explained by the small size of the genomic region examined for polymorphism by tDNA-ILP assay. We previously reported a similar discrepancy for tDNA-ILP patterns of viridans group streptococci (4).

Our results were obtained in 8 h, starting with 72-h-old bacterial growth on solid medium, a time interval as short as that described for the fastest genotypic assays. For example, 10 h was required to identify Legionella strains by using RAPD assays (1, 15, 16), not taking into account the additional step of restriction of 16S-23S rRNA amplified spacer fragments required for distinguishing the bluish-white autofluorescent species (20). Likewise, Fry and Harrison (8) reported a 17-h electrophoresis time to separate 16S-23S rRNA amplified spacer fragments. The expertise required for tDNA-ILP analysis was equivalent to that reported for other PCR-based identification assays. Our method appeared more convenient than alternative genotypic assays for the following reasons: no quantification of DNA was needed in contrast to other protocols (16, 21), and DNA profiles were captured numerically without the need to photograph the gels and to scan the picture as required elsewhere (1, 8, 16, 21, 27). Furthermore, the assay described here also offers the advantage of broad applicability to the routine clinical laboratory identification to the species level of other common bacterial pathogens, including staphylococci (17) and streptococci (4), by using the same PCR primers, amplification conditions, electrophoresis, and software for pattern recognition.

It would be nice if identification of Legionella performed with this assay could be obtained directly from clinical samples. However, this approach was not evaluated for the following reason: tDNA-ILP PCR using consensus primers performed on specimens collected from the respiratory tract would produce mixed profiles corresponding to the amplification of DNA from commensal bacteria in the oropharynx.

In conclusion, tDNA-ILP analysis of Legionella strains is a useful technique which compares well in terms of efficiency with recently described molecular methods for identification of environmental and clinical strains of Legionella species described as potential human pathogens.