INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Forty additional species and one genomospecies of the genus Legionella have been described since the initial description of Legionella pneumophila (2, 5, 10). Many of these species originate from environmental sources and have not been implicated in human disease. Genetic techniques, including whole-genome DNA-DNA hybridization, as well as computer-assisted whole-cell protein profiling, appear to be the only methodologies to identify strains of the species of this genus, since most species cannot be identified by using conventional biochemical and physiological characteristics (4, 8, 11, 23, 24, 27).

Fatty acid analysis has been shown to be a useful alternative or adjunct to phenotypic and genetic methodologies for the identification of many bacteria and has been used extensively for the identification of clinically important bacteria, having become established in many laboratories involved in taxonomy and diagnostic microbiology (13, 18, 22, 26). Particular attention must be paid to the influence of culture conditions on bacterial fatty acid composition, but standardization of media and growth conditions leads to highly reproducible fatty acid profiles that may be used as chemotaxonomic markers to distinguish strains of different species (21, 25). Fatty acid analysis has been used to discriminate the species of the genus Legionella (12, 15, 16, 17, 28), but there have been no recent studies that include all the known species of this genus or that have used a standardized procedure that can be used by many laboratories for the tentative identification of isolates.

In this study, we used one commercial system to analyze the fatty acid profiles of all validly described Legionella species and genomospecies, with the exception of L. lytica (10), to evaluate the ability of fatty acid composition to differentiate Legionella spp. and to produce a database to facilitate the automatic identification of isolates.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and culture conditions. Seventy-six type and reference strains, representing Legionella species and one genomospecies, and 199 isolates of the genus Legionella were used in this study (Table 1). The isolates obtained from France, the United Kingdom, and the United States were identified by the laboratory of origin or by other laboratories (1, 9, 23, 24), while the remaining isolates had been previously identified by indirect immunofluorescence and/or numerical analysis of whole-cell protein profiles in our laboratory (27). Cultures were maintained at 80°C in 5.0% (wt/vol) yeast extract containing 15.0% (vol/vol) glycerol. Legionella strains were spread, as recommended by a Microbial ID, Inc. (MIDI; Newark, Del.) protocol (Microbial Identification System) on buffered charcoal-yeast extract agar (7). All plates were incubated in sealed plastic bags, to prevent evaporation, at 36 ± 1°C for 72 ± 2 h, in a humidified incubator.

FAME analysis. Cells were harvested, and the fatty acid methyl esters (FAMEs) were obtained by saponification, methylation, and extraction, as described previously (14) and as recommended by the MIDI standard protocol. The separation of FAMEs was achieved by using a Hewlett-Packard model 5890 gas chromatograph with a flame ionization detector fitted with a 5% phenylmethyl silicone capillary column (0.2 mm by 25 m; Hewlett-Packard), controlled by the Sherlock Single Tower Library Generation software (version 1.06; MIDI). The carrier gas was high-purity H₂, the column head pressure was 60 kPa, the septum purge rate was 5 ml/min, the split ratio was 55:1, and the injection port temperature was 300°C. The temperature of the oven was programmed to increase from 170 to 270°C in 5°C/min increments. Peaks were integrated, and FAMEs were quantified and identified by using the peak-naming table component of the MIDI software package (Aerobe Method versions 3.8 and 3.9) Quantities were expressed as percentages of the total named FAME peak area.

Numerical analysis and mean profiles. A dendrogram, based on the fatty acid patterns of the strains, was generated by clustering the Euclidean distances of the fatty acids, using mapped features to increase differentiation, with the unweighted pair group method with arithmetic average algorithm that is provided by the MIDI software package. Clusters were delineated in the dendrogram to differentiate each species. The mean profile of each cluster in the dendrogram was calculated on the basis of the individual profiles of all strains in the cluster. For the unclustered strains, a mean profile was obtained from two or more replicate analyses performed on different occasions.

Reproducibility. Reproducibility was tested by repeated analyses of a standard quantitative FAME mixture (MIDI). The reproducibility of the fatty acid profiles was determined by analyzing 129 strains, on two or more occasions under standardized growth conditions, and by the inclusion of a control strain (L. pneumophila OLDA; ATCC 43109) in each batch of strains analyzed. The coefficient of variation, measured as (standard deviation/mean) × 100, was calculated for each fatty acid representing at least 20% of the total fatty acid content in the mean profile (21).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

The coefficient of variation [(standard deviation/mean) × 100)] for each of the most abundant fatty acids was less than 12% among replicate analyses. Furthermore, the control strain, L. pneumophila ATCC 43109, had a maximum coefficient of variation of 4%.

At least 68 different fatty acids were detected in the 275 strains tested, but 15 of them appeared, as minor components, in fewer than 1% of the strains (Table 2). Seven fatty acids were detected in all strains, while nine other fatty acids appeared in at least 72% of the strains. These fatty acids were the most common fatty acids detected in the strains examined and may, therefore, represent the qualitative fatty acid pattern of the genus Legionella.

The numerical analysis of the FAME profiles of 275 strains resulted in the formation of 33 clusters, with 18 strains remaining unclustered (Fig. 1). Nineteen of the clusters corresponded directly to Legionella species, while 13 of the unclustered strains belonged to species represented by type strains alone.

View larger version (23K): [in this window] [in a new window]   FIG. 1.   Dendrogram based on unweighted pair group average linkage of FAME profiles, using mapped features, of 275 Legionella strains. With the exceptions of Toulouse 20 no. 5, LC0870, Montbeliard A1, and Greoux 11D13, all of the unclustered strains are type strains.

L. geestiana ATCC 49504^T had a fatty acid composition that was different from those of all other Legionella strains (Fig. 1). This strain had 16:17c as the major fatty acid and higher relative proportions of i15:0 and i17:0 than of the corresponding a15:0 and a17:0 isomers, in addition to an unusually large amount of i15:0 (Table 3). Cluster 2 was composed of the 18 L. oakridgensis strains. The relative proportions of 16:17c and 18:0, associated with the lowest relative amount of a15:0, clearly differentiated L. oakridgensis from all other species (Table 3). The L. jordanis strains were grouped in cluster 3 (Fig. 1). The major fatty acid detected was a15:0, followed by i16:0. These two fatty acids also constituted the major acyl chains of L. birminghamensis (cluster 13), L. israelensis, and several species that autofluoresced blue-white. However, L. jordanis could be distinguished from the other species by the low amounts of cyc17:0 and its biosynthetic precursor, 16:17c (Table 3). The five strains identified as L. brunensis formed cluster 4 (Fig. 1), the major fatty acid detected being a15:0, followed by a17:0. These were also the major fatty acids in L. jamestowniensis. However, these two species could be distinguished by the relative amounts of i17:0, a17:19c, and the sum of cyc17:0 and 16:17c, which is higher in the type strain of L. jamestowniensis than in L. brunensis (Table 3).

Two reference strains and 14 isolates previously identified as L. micdadei were split into clusters 5 and 6 (Fig. 1), despite the fact that a15:0 was the major fatty acid in all strains of this species. Cluster 5, containing the type strain of L. micdadei, had higher amounts of i16:0 and a17:0 and lesser quantities of 16:17c than cluster 6 did (Table 3). One important characteristic of the L. micdadei strains was the presence of moderate amounts of a17:19c, previously identified as a17:17c (19, 20). This fatty acid was also detected in the type strains of L. jamestowniensis and L. lansingensis and two of the three strains assigned to L. maceachernii (cluster 7); nevertheless, a17:19c was not detected in strain Toulouse20 no. 5 (Table 3).

L. hackeliae and L. nautarum strains constituted clusters 8 and 9, respectively (Fig. 1). The mean fatty acid profiles of these species were similar (Table 3), but they could be distinguished clearly from each other by numerical analysis. L. adelaidensis and L. moravica strains formed clusters 10 and 11, respectively (Fig. 1). In both species, the major fatty acid was 16:17c, followed by i16:0 and a15:0 in similar amounts (Table 3). The L. worsleiensis strains available formed cluster 12 and had a fatty acid composition that was similar to that of L. quateirensis ATCC 49508^T, resulting in a high similarity value between these two species (Fig. 1).

Legionella genomospecies 1 (ATCC 51913) was responsible for the separation of the two L. quinlivanii strains used (Fig. 1). While strains ATCC 51913 and LC0870 (L. quinlivanii serogroup 2) had similar mean fatty acid profiles, the type strain of L. quinlivanii had a slightly different profile (Table 3), being included in cluster 14 with a Portuguese isolate (Edt-8); this strain produced strong cross-reactions between L. quinlivanii and L. feeleii when analyzed by indirect immunofluorescence.

The eight strains assigned to L. feeleii were split into clusters 15 and 30 (Fig. 1). Cluster 30 included four strains identified as L. feeleii by several methods, while cluster 15 contained both the type and a reference strain of L. feeleii. The mean profile of each cluster had different major fatty acids, but all of the strains contained two fatty acids (16:111c and alcohol 16:17c) that were not found in any other of the strains examined (Table 3).

Twenty strains belonging to the red autofluorescent species L. rubrilucens and L. erythra (cluster 17) had very similar fatty acid compositions, leading to the formation of a single cluster (Fig. 1). The type strain of L. londiniensis and 12 strains assigned to this species formed cluster 18 (Fig. 1). The major fatty acid detected in most strains was a17:0, although in some strains i16:0 and 16:17c could reach similar levels or could be even slightly higher than the level of a17:0 (Table 3). This fatty acid was also the major fatty acid in L. lansingensis, but this species could be easily distinguished from L. londiniensis by differences in the relative amounts of a15:0, i16:0, and 16:17c and by the presence of a17:19c.

The blue-white autofluorescent species formed a well-defined group, at a Euclidean distance of approximately 23 units (Fig. 1). The type strain of L. bozemanii, two reference strains, and 16 isolates previously assigned to this species were split into clusters 26 and 27. Strains assigned to L. dumoffii also formed two clusters (clusters 19 and 21). Cluster 19 did not contain the type strain of this species, but two of the three isolates had been identified as L. dumoffii (24). Five strains previously identified as L. gormanii were included in cluster 23. However, the type strain of L. gormanii remained unclustered, despite the similar fatty acid compositions of the strains of this species (Table 3).

The fatty acid composition of the type strain of L. waltersii was unique because i14:0 was the major fatty acid. Moreover, this organism and the type strain of L. gratiana, unlike all the other strains examined, possessed only trace amounts of a17:0 (Table 3).

All 58 L. pneumophila strains examined formed a single cluster (cluster 29), at a Euclidean distance of approximately 17 units (Fig. 1). Cluster 32 included the type strain of L. cincinnatiensis, two strains previously assigned to this species (E1549 and LC3936), four unidentified strains recently isolated from a Portuguese spa (designated Felg), and five other isolates (designated La) shown to be similar to the type strain of L. sainthelensi (27). Clusters 31 and 33 contained only strains previously identified as L. longbeachae and L. sainthelensi, respectively (Fig. 1). For L. cincinnatiensis, L. sainthelensi, and L. longbeachae, the major fatty acids were the same, with only small quantitative differences in the minor acyl components (Table 3).

The fatty acid profiles of eight previously unidentified isolates and one isolate initially assigned to L. geestiana (LC3644) did not conform to any of the profiles of recognized species. Legionella group I (cluster 1) comprised two of these isolates for which the presence of i15:0 and a15:0 as the major fatty acids represents a unique profile among Legionella spp. (Table 3). Two other strains (Montbeliard A1 and Greoux 11D13) had distinct fatty acid compositions, while Legionella group II (cluster 22) included five unidentified isolates with colonial blue-white autofluorescence.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The strains of the genus Legionella are notoriously difficult to identify by traditional methods. Fatty acid analysis of Legionella strains is useful for characterizing these organisms at the genus level but has been difficult to use in identifying individual species (15, 16). This study shows that the application of numerical analysis, instead of simple quantitative comparison of fatty acid compositions, coupled with the utilization of a standardized identification system greatly enhances the utilization of fatty acid profiles for the identification of individual Legionella species.

The identification of some legionellae may not be straightforward, since the strains of some Legionella species, namely, L. micdadei, L. maceachernii, L. quinlivanii, L. feeleii, L. dumoffii, L. gormanii, and L. bozemanii, form more than one cluster. The variability of the fatty acid composition of L. micdadei strains was noticed previously by Moss and Lambert-Fair (19), who showed that strain Bari 2/158 had a different fatty acid composition than the type strain of L. micdadei. Our results extend this characteristic to two other strains of L. micdadei and clearly show that a similar degree of variation can occur in other species. However, automatic identification of the isolates of these species is possible, as long as these fatty acid subgroups are taken into account in the construction of databases.

Variations in the major acyl components may also occur within one cluster, as with the L. londiniensis strains examined. Previous results for the fatty acid composition of this species were based on the type strain alone, and variations in the levels of the fatty acids could not be observed (6). This example shows the importance of examining a large number of strains, when available, for a reliable analysis of chemotaxonomic markers of a species.

Despite these variations, some species like L. pneumophila had a very stable fatty acid profile. The formation of a single cluster by a large number of L. pneumophila strains provides a suitable example of the resolution and ability of this method to distinguish this species from all other species of the genus Legionella. Some Legionella species, on the other hand, had very similar fatty acid compositions and may be difficult to differentiate by this method. The numerical analysis of FAME profiles showed a high similarity between some species, such as L. moravica, L. quateirensis, and L. worsleiensis; L. erythra and L. rubrilucens; and L. sainthelensi, L. cincinnatiensis, and L. longbeachae. Moreover, these similarities were also found by so-dium dodecyl sulfate-polyacrylamide gel electrophoresis protein profiling (27), mip gene sequencing studies (23), and intergenic 16S-23S ribosomal spacer PCR analysis (24), corroborating recent phylogenetic studies (3, 10). With the exception of the red autofluorescent strains, it was possible to differentiate all the Legionella species and unambiguously identify, by numerical analysis of FAME profiles, the vast majority of the strains used. Indeed, the identification of the vast majority of the strains that were also used in other recent studies could be confirmed by fatty acid analysis (23, 24, 27).

Nine isolates had fatty acid profiles unlike any of those of the validly described species and may represent distinct FAME groups of known species or undescribed Legionella species, as has been suggested recently by others, based on genetic typing methods (23, 24) or protein profiling (27).

The lack of strains other than the type strain of some species was probably the major difficulty in assessing the resolution of this identification method for species of the genus Legionella. The inclusion of more strains of some species, preferably from widely separated geographical areas and diverse ecological habitats, is clearly necessary for an enhanced database for this method. However, on the basis of our results, we conclude that the numerical analysis of fatty acid profiles, by a standardized system, is helpful in identifying Legionella species, especially when a large number of Legionella isolates are being screened, as is the case in many clinical or reference laboratories.