INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Bacteria of the genus Legionella cause community-, travel-, and hospital-acquired pneumonia in humans, usually via inhalation of aerosols formed from man-made water systems, in which the bacteria have been enriched. Legionella pneumophila serogroup 1 is the most important causative agent, especially in community outbreaks and travel-acquired infections. Among the 1,442 cases of Legionnaires' disease reported from 28 European countries in 1998, L. pneumophila serogroup 1 accounted for 60%, other or undetermined serogroups of L. pneumophila accounted for 34.4%, and other species (L. micdadei and L. bozemaniae) accounted for the remaining 5.6% of cases (24).

In Finland, with a population of 5 million, about 10 cases of Legionella pneumonia are reported annually (http://www.ktl.fi/ttr/tt9599_33_63.pdf). Apart from a hospital outbreak involving four patients in 1995, the cases have been sporadic. According to a study of 52 Finnish legionellosis cases from 1982 to 1992 (21), 44% of patients were immunosuppressed due to ongoing or recent therapy with immunosuppressive agents, 23% had other underlying diseases, and 33% had no predisposing conditions. Seventy percent of the infections in previously healthy persons were associated with travel, and all were caused by L. pneumophila serogroup 1. Legionellosis was nosocomial in 73% of the immunosuppressed patients. Among those patients, L. pneumophila serogroup 1 caused only 26% of the cases; L. pneumophila serogroup 6 was recognized as the causative agent in 35%, and non-L. pneumophila Legionella species were so recognized in 9%.

The diagnosis of legionellosis is optimally confirmed by isolation of Legionella spp. from lower-respiratory-tract specimens. Although absolutely specific, this approach is rather slow and insensitive, as in only 21.6% of the reported European cases in 1998 was the diagnosis based on successful culturing of Legionella organisms (24). Serologic analysis can be used only retrospectively and may be of limited value in immunocompromised patients. During recent years, screening kits for Legionella antigens in urine have become available (5). Alternatively, many groups have developed and evaluated assays to detect Legionella DNA in respiratory tract specimens and sometimes also in serum or urine (1, 6-8, 9-12, 14-16, 23). PCR is more time-consuming and complex to perform than antigen detection tests but, especially if culturing fails, offers the possibility to use the same sample for identification and/or typing of the agent by molecular methods.

Here we describe the development and validation of a PCR assay for the detection of L. pneumophila DNA in respiratory specimens. In addition to validating this assay, we adapted the procedure to a new real-time capillary PCR format (Lightcycler; Roche Molecular Diagnostics, Mannheim, Germany), which allows the amplification and detection to take place in less than 1 h and therefore significantly accelerates the screening of clinical specimens.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacteria. Unless stated otherwise, the experiments were done using the L. pneumophila type strain (ATCC 33152), which was used for quality control throughout. A total of 47 other L. pneumophila serogroup 1 strains were used in this study, including 40 strains of the phase II panel of the European Working Group on Legionella Infections (EWGLI) (4), 10 of which were studied as duplicates. The strain collection also included 25 L. pneumophila strains representing serogroups 2 to 12, as well as 27 non-L. pneumophila strains representing 20 Legionella species. In addition, 21 bacterial strains either phylogenetically related to the genus Legionella or possibly present in sputa or bronchoalveolar lavage fluids were used in specificity tests. The Legionella strains were a kind gift from Hannele Jousimies-Somer, National Public Health Institute, Helsinki, Finland. The sources of the other strains are listed in Table 1. Legionella strains were grown on buffered charcoal-yeast extract (BCYE; Oxoid Ltd., Basingstoke, United Kingdom) agar plates at 37°C for 48 to 72 h. Other bacteria were cultivated on standard media supporting optimal growth. The clinical strains were identified by standard methods.

DNA purification from bacteria. Cultured bacterial cells were suspended in 200 µl of phosphate-buffered saline and digested with proteinase K (0.1 mg/ml, 56°C, 2 to 17 h). DNA was extracted with two phenol-chloroform-isoamyl alcohol extractions, washed once with ether, and precipitated with sodium acetate-ethanol. Purified DNA was dissolved in 100 µl of sterile water. The DNA concentration was determined by measuring the absorption at 260 nm and adjusted to 100 µg/ml.

Primer design and PCR optimization. To find oligonucleotide sequences specific for L. pneumophila, a multiple alignment was done from the complete 16S rRNA and rDNA sequences at EMBL database release 52.0 (22) for six L. pneumophila strains (accession numbers M59157, M36023, M36024, X36025, M36026, and X73402) and the type strains of L. longbeachae, L. bozemaniae, and L. micdadei (accession numbers M36029, M36031, and M36032, respectively). The specificity of the candidate primers for all bacterial sequences in the database was verified by FastA analysis (18). Primers were purchased from Medprobe AS, Oslo, Norway.

Sample preparation. About 1 ml of sputum samples was treated with proteinase K, and 200 µl was used for DNA extraction. BAL fluids were concentrated (8,000 × g, 5 min), and DNA was extracted from 200 µl of the concentrate. Two DNA extraction systems were applied: the standard phenol-ether purification described above (without the precipitation step) and a commercial kit based on the adsorption of DNA to silica particles (High Pure PCR template preparation kit; Roche Molecular Diagnostics).

Determination of the analytical sensitivity of the assay. To determine the lowest number of L. pneumophila cells detectable by the assay, serial dilutions of L. pneumophila cells were used to spike sputum samples or BAL fluids. The artificially seeded sputum samples were prepared by pooling sputum samples from four or five different patients, adding an equal volume of 0.1% dithiothreitol (Sigma), vortexing for 5 min, and incubating for 15 min at room temperature. The homogenate was divided into 1-ml aliquots, which were spiked with 2 to 200,000 Legionella cells. One-milliliter aliquots of pooled and homogenized sputum or of pooled BAL fluid were concentrated as described above, and DNA was extracted with either the phenol-ether procedure or the High Pure PCR template preparation kit. To directly compare the ability of the two systems to release and purify template DNA for amplification, each spiked sample was divided in two, an equal volume was treated with each DNA preparation procedure, and 5 µl was used as a template in the PCR (or 2 µl for the Lightcycler PCR). Statistical analysis of the detection limits for the two DNA extraction methods was performed with the Kruskal-Wallis test.

Reproducibility of Lightcycler PCR. The within-run and between-run variations in the Lightcycler PCR system were estimated by running samples of one dilution series in two different sputum backgrounds twice, two replicates in each run, with purified L. pneumophila DNA as the standard. The run-to-run variations in the Lightcycler system were also assessed by calculating the average crossing points and T_ms from the three standard samples in 20 consecutive runs.

Clinical specimens. Seventy-seven respiratory specimens (11 sputum samples and 66 BAL fluid samples) from 71 hospitalized patients with symptoms of acute pneumonia were analyzed by L. pneumophila PCR. DNA was isolated by the phenol-ether method described above. Eighteen specimens were divided on arrival in the laboratory, and about 1 ml was processed with the High Pure PCR template preparation kit. All specimens were tested for PCR inhibition by amplification of the human growth hormone gene as described previously (19). All specimens were cultivated on BCYE agar plates with MWY and BMPA- supplements (Oxoid) at 37°C for 14 days to screen for the presence of Legionella DNA. The identification of Legionella species was based on typical cellular fatty acid profiles in gas-liquid chromatography (3).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Primers and optimization of PCRs. Primers Lpneu2 and Lpneu3 (Table 2) produced the expected 245-bp amplicon from L. pneumophila DNA, while the control reactions with DNA from P. aeruginosa and S. maltophilia remained negative. Originally, a polymerase isolated and purified from an Escherichia coli strain carrying a plasmid with the cloned DNA polymerase gene from Thermus brockianus was used (DyNAzyme II F-501L DNA polymerase). Optimized PCR with DNA from bacteria produced either a single 245-bp band or no band at all. However, when DNA from simulated sputum samples was used as a template, bands of about 500 bp were detected, especially when no or few L. pneumophila cells were used for spiking. No cross-amplification of DNA from normal flora was detected when AmpliTaq Gold enzyme was used.

Amplification of DNA from bacterial strains. (i) Conventional PCR. DNA from all 73 L. pneumophila strains was amplified in the optimized PCR, and all PCR products also hybridized with the probe Lpneuprb (Table 1). In addition, DNA from 21 out of 27 other Legionella strains yielded the expected 245-bp band on an agarose gel. The bands were often weaker with non-L. pneumophila Legionella strains, and some remained negative for hybridization with the probe. DNA from other bacteria was not amplified.

(ii) Real-time PCR All L. pneumophila strains were amplified by the Lightcycler system. With 20 ng of purified DNA as a template, serogroup 1 strains had an average T_m of 85.58°C (standard deviation [SD], 0.18), with a mean T_m ratio (T_m of sample divided by T_m of the 20-ng standard) of 0.999 (SD, 0.002). The respective values for the other L. pneumophila serogroups were 85.37°C (SD, 0.21) and 1.000°C (SD, 0.002). The amplification of non-L. pneumophila Legionella strains was variable. They were considered positive when the T_m ratio was between 0.993 and 1.005 (average and 3 SDs for L. pneumophila serogroup 1 strains). The results were in accord with the presence or absence of the 245-bp band on the gel. The average T_m for positive non-L. pneumophila Legionella strains was 85.31°C (SD, 0.30). The results obtained by Lightcycler PCR were not always consistent with those obtained by conventional PCR (Table 1). With a high template DNA concentration (200 ng), some bacteria other than Legionella species spp. were amplified, but the products were readily distinguished from Legionella-specific products by different melting points (Table 1).

Detection limits of the assays Table 3 shows the detection limits of the assays when serial dilutions of L. pneumophila were used to spike simulated clinical specimens and then were purified by different DNA extraction methods. For sputa, the background sample had a profound effect on the detection limits, as the lowest and the highest limits were obtained with the same dilution series and with the same samples by both DNA isolation methods. When aliquots of each simulated sample were processed by the two DNA isolation methods, the detection limits achieved with the silica method were statistically significantly lower than those obtained by the phenol-ether method (P = 0.0218; Kruskal-Wallis test). In general, the sensitivity of the Lightcycler system in detecting small amounts of target DNA was comparable to that of conventional PCR. As in conventional PCR, the variations observed in the detection limits with the Lightcycler system were mainly due to the background sample, but there were also run-to-run variations in the most dilute sample detected as positive.

Reproducibility of Lightcycler PCR. Table 4 shows the results obtained from an experiment evaluating the within-run and between-run variations of the Lightcycler PCR. As determined from values obtained in 20 consecutive runs, the coefficient of variation (CV) for the crossing point was about 16% with all three standard samples. The average T_m for the standards was 85.71 (SD, 0.54; CV, 0.62), and the calculated T_m of the 20-ng standard varied from 84.97°C to 87.02°C.

Detection of Legionella DNA in clinical specimens. (i) Conventional PCR. None of the samples showed total PCR inhibition, as judged by amplification of the human growth hormone gene in a separate reaction. None of the 11 sputum samples was PCR positive for Legionella DNA. Two (3%) out of the 66 BAL fluid samples were positive for Legionella DNA in the PCR. Both bands also hybridized with the probe Lpneuprb. The PCR results were consistent with those of culturing, as L. pneumophila was isolated from both PCR-positive BAL fluid samples. One of the positive samples was processed by both tested DNA purification methods; one was a scant sample and was purified only by the routine phenol-ether method.

(ii) Real-time PCR. Clinical samples were considered positive when the T_m ratio was between 0.993 and 1.005 (average and 3 SDs for L. pneumophila serogroup 1 strains). Sixty-five BAL fluid samples and 11 sputum samples treated with phenol-ether were screened. The two BAL fluid samples growing L. pneumophila were positive, with T_ms of 86.97 and 86.45°C (T_m ratios of 0.999 and 0.994, respectively). The crossing points were 17.99 and 28.49, corresponding to about 1 ng and 2 pg of DNA in the 2-µl preparation, respectively. In terms of CFU, the concentrations would have been more than 2 × 10⁶ CFU/ml and about 20,000 CFU/ml of BAL fluid, respectively (Fig. 1). Silica-purified DNA from 18 BAL fluid samples yielded consistent results.

View larger version (28K): [in this window] [in a new window]   FIG. 1.   Amplification profiles (A), melting curves (B), and agarose gel electrophoresis patterns (C) of Lightcycler PCR with SYBR Green I. Samples: 1, water (reagent control); 2 to 4, L. pneumophila DNA standards at 100 ng (2), 20 ng (3), and 4 ng (4); 5 to 9, aliquots of BAL fluid pools spiked with 200,000 (5), 20,000 (6), 2,000 (7), or 200 (8) L. pneumophila cells before DNA extraction and the same BAL fluid pool without added bacteria (9); 10 to 13, patient samples (BAL fluids). (A) Fluorescent emissions were measured as the last step of each thermal cycle at 82°C and plotted against the cycle number. The number associated with each curve is the sample number. For quantification, the noise band was set above the background fluorescence at about 1, and the cycle at which the curve of each sample crosses the noise band is proportional to the amount of L. pneumophila DNA in the sample. (B) After amplification, fluorescence was measured while the temperature was increased at 0.1°C/s from 63°C to 95°C. The change in fluorescence with the change in temperature [d(F1)/dT] was plotted as a function of temperature. The Tm of the L. pneumophila PCR product is about 86.75°C, and nonspecific products melt at between 74 and 83°C. (C) PCR products were run on a 4% agarose gel, stained with ethidium bromide, and visualized under UV light. Lane L, 1-kb DNA ladder (Gibco); lanes 1 to 13, samples (see panel A). In addition to the specific 245-bp bands, some faint bands of other lengths are seen in BAL fluid samples with little or no L. pneumophila DNA; these correspond to the small melting curve peaks in panel B. F1, fluorescence at wavelength of the instruments channel 1.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

An optimal PCR assay for the diagnosis of legionellosis should detect L. pneumophila as well as about 20 other Legionella species that have been associated with infections in humans. The importance of Legionella spp. other than L. pneumophila is emphasized when a major portion of referred samples originate from hospitalized patients with underlying diseases. On the other hand, these patients are often colonized by other gram-negative rods (20), which must not give false-positive signals in the assay.

The first PCR systems for the detection of the genus Legionella were based on primers targeted at 5S rDNA (13), which amplified a wide range of Legionella spp. but also some Pseudomonas species. Specificity problems also were reported in attempts to detect Legionella DNA in urine with the use of modified 5S rRNA primers (14).

16S rDNA is more variable than 5S rRNA, and large databases with thousands of bacterial sequences offer a firm background for the rational design of primers for genus- or species-specific PCRs. However, we were unable to design generic Legionella sp. primers based on 16S rRNA sequences, and previously published primers have covered only part of the genus. Lisby and Dessau (12) described a PCR assay with primers partly overlapping ours and a membrane hybridization confirmation system. L. pneumophila and some other Legionella spp. were detected, but the clinically important L. micdadei and L. bozemaniae were not. Only two out of seven PCR-positive results could be confirmed by culturing or serologic analysis. Jonas et al. (8) developed a PCR-enzyme-linked immunosorbent assay with 16S rRNA primers and probe and reported a slightly different range of detected Legionella species. Eight of the 14 PCR-positive results were confirmed by culturing, and 5 out of 6 culture-negative, PCR-positive samples came from patients on high-dose erythromycin therapy which, according to our experience, explains such results (19). Cloud et al. (1) used the same primers in less stringent conditions to allow the amplification of, e.g., L. micdadei and obtained specificities of 93% with PCR alone and 98% after confirmation by sequencing.

An alternative approach for a clinical PCR, the use of the macrophage infectivity potentiator (mip) virulence gene as a PCR target, has also been applied to detecting L. pneumophila in respiratory samples (7, 10) and serum (11). Jaulhac et al. (7) used primers targeted to the conserved area of the mip gene and successfully amplified not only L. pneumophila but also L. bozemaniae and L. micdadei. L. dumoffii is the clinically most important species that remains unamplified in this assay, but it can be detected in our system. The reported detection limit in BAL fluid (25 CFU per ml after Southern blot hybridization) is comparable to ours. A commercial kit using a combination of 5S rRNA and mip primers (EnviroAmp; Applied Biosystems) and designed for the detection of Legionella DNA in water samples has been successfully applied to respiratory specimens (9, 15, 23) and to urine specimens with somewhat reduced sensitivity (6). The kit is, however, currently not commercially available.

Our PCR design was successful in the sense that the clinically most relevant species, L. pneumophila, L. micdadei, L. bozemaniae, and L. dumoffii, were amplified. Furthermore, a positive PCR result could be interpreted rather reliably as indicative of the presence of Legionella DNA in the sample and thus reported to the clinician in a timely manner. However, the development of a rapid confirmatory assay seems necessary. Confirmation could be based on capillary sequencing or on a hybridization probe assay by use of Lightcycler with a small panel of probes for different species.

Optimal sample processing should concentrate the possible target organisms in the specimen, release their DNA, and wash away the inhibitory compounds. Of the two DNA purification systems tested, the High Pure PCR template preparation kit achieved higher analytical sensitivity, although it remains unclear whether this result also affects clinical sensitivity. With both DNA purification methods, the detection limits for PCR remained relatively high in comparison to those reported for culturing (1 to 60 cells per ml in the artificially seeded sputum samples) (2). In clinical samples, the lower analytical sensitivity of PCR is partly compensated for by the ability of PCR to detect DNA from nonreplicating bacterial cells as well.

In the Lightcycler system, the accumulation of amplicons in the reaction capillaries can be monitored cycle by cycle using either SYBR Green I dye, which binds any double-stranded DNA, or sequence-specific detection with two fluorogenic hybridization probes. Since Legionella organisms are not considered to be part of the normal human flora, we were interested in detecting the presence or absence of Legionella DNA rather than its quantification and chose to start with the simpler and less expensive SYBR Green I format with T_m analysis of the PCR product. Adaptation of this particular PCR to Lightcycler was easy, and optimization was even easier than with conventional PCR, as the whole reaction rather than the end product could be monitored. Analysis of T_ms readily distinguished the specific and nonspecific PCR products, even if they were visualized as bands of almost equal sizes in electrophoresis. Determination of melting points was very stable in a given run and in consecutive runs, as observed by analysis of strain collections, but the long-term variation makes it difficult to set fixed limits for accepted melting points without normalization to a standard sample. According to our experiments, reliable quantification of Legionella DNA in clinical samples with this assay would be difficult due to variable sample background and a tendency to form primer-dimers and nonspecific amplification products when the concentration of the specific target is decreasing, i.e., variable amplification efficiency in different samples. However, the assay shows promise for simple screening of samples for the presence of Legionella organisms provided that the concentrations of Legionella DNA in samples from infected patients are at the level of the two positive samples in our collection and that the variation in the normalized T_m of the PCR product is as low as in this study.

The real-time assay can be completed 2 to 3 h after the sample has arrived in the laboratory, whereas the conventional assay takes at least 6 h (without hybridization). Even with a confirmatory test, the results of the real-time PCR assay could be reported during the same or the next day, depending on the transportation of the samples and the routine work flow of the laboratory. The cost of reagents, plasticware, and capillaries for the Lightcycler PCR is about $5 per reaction, as opposed to about $2 for the conventional PCR and electrophoresis detection. Labor savings via elimination of the need for electrophoresis compensate for the higher reagent cost of the real-time system. However, a larger number of clinical samples should be analyzed to ensure that annealing and amplification in the Lightcycler assay are stringent enough to be applied to the bacterially complex environment of human respiratory specimens and that the melting curve analysis reliably separates the possible by-products of the amplification reaction.

In conclusion, we describe a PCR assay which detects the presence of the clinically most important Legionella species in respiratory samples. We have included the conventional assay in the repertoire of our routine diagnostic PCR laboratory. Although the amplification seemed specific, the development of a rapid confirmatory system will improve the clinical utility of the assay. The primers are also applicable to capillary real-time PCR with SYBR Green I detection and melting point analysis of the PCR product, but further experience is needed to assess the reliability of that kind of assay as a screening tool for the presence of Legionella DNA in clinical specimens.