Comparison of Sample Preparation Methods for Detection of Legionella pneumophila in Culture-Positive Bronchoalveolar Lavage Fluids by PCR

Abstract

A prospective study was conducted on 25 Legionella pneumophila culture-positive and 98 culture-negative bronchoalveolar lavage fluid samples to compare two DNA preparation methods: a rapid modified Chelex-based protocol and a proteinase K method. PCR was found to be more sensitive with the Chelex-based method (P = 0.03). No difference was found concerning the inhibition rate.

Among the bacteria responsible for atypical pneumonias, Legionella is one of the main agents involved in severe forms of the disease. Since a good prognosis depends in part on specific early treatment, rapid identification of the causative agent is therefore of great importance. Unfortunately, Legionella is a slow-growing bacterium, and serological diagnosis is generally delayed. For these reasons, molecular methods and especially nucleic acid amplification techniques (20) may represent a useful tool for detection of Legionella in clinical samples. Among the rapid diagnostic methods for evidencing Legionella, DNA amplification by PCR has been reported by several groups to be very effective on various biological samples (4–9, 11–14, 17). To be useful for clinical diagnosis, the format of such a PCR test needs (i) to detect samples containing low numbers of bacteria, (ii) to be easy and rapid, and (iii) to avoid as much as possible false-negative results due to inhibition of the amplification step. The latter aspect is of special importance concerning bronchoalveolar lavage (BAL) fluids, which are very heterogeneous and may contain various substances able to inhibit Taq polymerase such as large amounts of host DNA, heme, acidic polysaccharides and laboratory reagents (3). Numerous protocols have been reported for BAL fluid preparation before DNA amplification (4–6, 10), but, up to now, no comparison of different sample preparation methods has been performed on a significant number of relevant clinical samples.

The aims of our study were (i) to compare on culture-positive BAL fluid samples the sensitivities of two different DNA preparation methods—a rapid modified Chelex-based method and a standard proteinase K method previously reported by some of us for detection of Legionella (4)—and (ii) to determine the rates of Taq polymerase inhibition of these two methods in BAL fluid samples.

The two DNA preparation methods were first tested on artificially seeded BAL fluid samples to compare the sensitivity levels obtained for detection of Legionella by in vitro DNA amplification. First, 10 colonies of a 3-day culture of Legionella pneumophila serogroup 1 on BCYE-α agar (2) were suspended in 1 ml of sterile distilled water. The number of CFU was determined by plating 100-μl samples of serial dilutions of the culture on BCYE-α agar plates; then, a 100-μl culture sample was serially diluted at 4°C in a 1-ml volume of BAL fluid containing no Legionella spp. as proven by culture on BCYE-α agar. The artificially seeded BAL fluids obtained were then processed in parallel by the following two protocols.

The first protocol used was a proteinase K lysis method previously described (4). Briefly, a 1-ml aliquot of each BAL fluid specimen was mixed for 15 s with an equal volume of phosphate-buffered saline in a 2-ml microtube and centrifuged for 10 min at 9,500 × g. This wash step was repeated once. The pellet obtained was treated with 50 μg of proteinase K (Sigma, Saint Quentin Fallavier, France), 0.5% (vol/vol) Nonidet P-40 (Boehringer Mannheim, Meylan, France), and 0.5% (vol/vol) Tween 20 (Boehringer Mannheim) in 500 μl of 10 mM Tris-HCl (pH 8)–50 mM KCl–50 mM MgCl₂. DNA was then purified by phenol-chloroform and ethanol precipitation. The air-dried pellet was then resuspended in 40 μl of 10 μM Tris HCl–1 μM EDTA buffer and was heated for 10 min at 95°C. Half of the processed sample (20 μl) was subjected to 40 cycles of amplification in a 100-μl volume. DNA amplification and Southern blot hybridization were performed with primers Lpm-1 and Lpm-2 and detection probe Lpm-3 as previously described (4), except for the Taq polymerase which was purchased from Gibco-BRL (Cergy Pontoise, France). Under these conditions, a sensitivity of 25 CFU/ml was obtained by the proteinase K lysis method (Fig. 1A).

FIG. 1.

Sensitivity of the detection of Legionella in BAL fluids by DNA amplification according to the sample preparation method used. Serial dilutions of the culture of L. pneumophila serogroup 1 ATCC 33152 were seeded in 1-ml BAL fluid samples and subjected to 40 amplification cycles with primers Lpm-1 and Lpm-2 and then to Southern blot analysis after hybridization with the detection probe Lpm-3. (A) Proteinase K protocol; (B) optimized Chelex protocol. CFU per milliliter: lane 1, 2.5 × 10³; lane 2, 5 × 10²; lane 3, 2.5 × 10²; lane 4, 50; lane 5, 25; lane 6, 5; lane 7, 2.5; lane 8, 5 × 10⁻¹; lane 9, 2.5 × 10⁻¹. Lane 10, no bacterium.

In the second DNA preparation method, the BAL pellet obtained after two wash steps in phosphate-buffered saline was resuspended in 500 μl of a 5 to 20% (wt/vol) solution of Chelex 100 resin (Bio-Rad, Richmond, Calif.) in autoclaved distilled water, as described by de Lamballerie et al. (1), in 10 mM Tris–HCl (pH 8.0)–0.1 mM EDTA–0.1% sodium azide (11), or in 0.5% (vol/vol) Nonidet P-40–0.5% (vol/vol) Tween 20–50 μg of proteinase K per ml. The samples were then mixed vigorously in a rotary shaker for 30 s and then incubated at 55 or 98°C for 30, 60, or 120 min.

The best results were obtained when DNA was extracted with a 5% solution of Chelex 100 in water or in sodium azide for a 30-min incubation. Increasing the resin concentration to 20% did not produce any improvement in sensitivity, nor did increasing the incubation time to 2 h. Significantly better results were obtained with an incubation temperature of 55°C, and a significant decrease of the sensitivity was also observed if lysis was performed at boiling temperature instead of 55°C. This might be explained by fragmentation of Legionella DNA at temperatures higher than 55°C. Under these optimized conditions, the detection threshold of Legionella in BAL fluids was reproducibly estimated to be 5 CFU/ml (Fig. 1B).

Then, a total of 25 BAL fluid samples, consecutively collected at two separate university hospitals (Lyon and Strasbourg, France) and previously found to be positive by culture for L. pneumophila, were each processed by the two optimized DNA preparation methods described below. L. pneumophila serogroups 1, 3 to 5, 8, and 10 were isolated from these BAL fluid samples with concentrations ranging from 10¹ to >10⁵ CFU/ml (Table 1). Among the 25 samples tested, 17 (68%) were found to be positive by DNA amplification after DNA preparation by the proteinase K lysis method whereas 22 specimens (88%) were positive by the optimized Chelex protocol. Compared to the results obtained by the proteinase K method, the optimized Chelex protocol showed better results (P = 0.03) by Fischer’s test using StatXact-3 software (Cytel Software Corporation, Cambridge, Mass.). This confirms PCR as an efficient tool with high sensitivity, among the direct-diagnosis tests for this disease. In this way, direct immunofluorescence and nucleic acid probe assays are usually reported to have a sensitivity of 50 to 70% among culture-proven samples (15). Urinary antigen assay has a sensitivity of 80%, but for L. pneumophila serogroup 1 infection only (16).

TABLE 1.

PCR positivity in relation to the Legionella load of the BAL fluid sample and the sample preparation method used

Legionella CFU/ml	No. of BAL fluid samples tested	No. of PCR-positive samplesa
		PK method	Chelex method
101	6	1	3
2 × 101 to 102	9	7	9
2 × 102 to 103	4	3	4
>2 × 103	6	6	6
Total (%)	25	17 (68)	22 (88)

^aAll samples positive by the proteinase K (PK) method were positive by the Chelex method. 

All DNA samples positive by the proteinase K method were positive when prepared by the Chelex method. Sensitivity of PCR by the latter preparation method should be particularly emphasized since, of the six culture-positive samples containing as few as 10 Legionella bacteria/ml, three were positive by the Chelex protocol whereas only one was positive by the proteinase K protocol. In the same way, all nine BAL fluid samples containing 2 × 10¹ to 10² CFU/ml were positive by Chelex whereas only seven of the nine BAL fluid samples were positive after proteinase K preparation (Table 1).

Then, we investigated the rate of Taq polymerase inhibition in BAL fluid samples by testing a total of 98 BAL fluid samples previously found to be negative for Legionella by culture on BCYE-α agar and by PCR. A quantity of a BAL fluid DNA preparation equal to that previously tested for the presence of Legionella DNA was amplified in a second PCR run after 1 pg of L. pneumophila serogroup 1 total DNA was added to the amplification mix. Samples giving no signal after PCR as well as samples giving a significantly lower signal than expected were considered inhibited. In this way, the overall Taq polymerase inhibition rate was 15% for all the preparation methods used (Table 2). This inhibition rate for the proteinase K lysis method is in accordance with the data reported by Maass (10), who showed that 12% of 75 samples were inhibited when the same DNA preparation procedure was used. This enhances the requirement for systematic testing of PCR-negative samples for inhibitors when BAL fluid samples are subjected to PCR.

TABLE 2.

Rate of Taq polymerase inhibition in BAL fluid specimens in relation to sample type

BAL fluida	No. of samples tested	No. of samples inhibited (%)
		PKb method	Chelex method
Standard	30	3 (10)	3 (10)
Bloody	27	5 (18)	3 (11)
Cellular	24	2 (8)	4 (16)
Bloody and cellular	17	5 (29)	5 (29)
Total	98	15 (15)	15 (15)

^aStandard, no macroscopically detectable blood and low number of cells (no smear or a faint smear on agarose gel); Bloody, macroscopically visible presence of blood; Cellular, giving a strong smear on agarose gel. 

^bPK, proteinase K. 

Among the 98 BAL fluid samples tested, some had a higher probability of PCR inhibition: 27 contained macroscopically visible blood, 24 contained numerous host cells (DNA preparation giving a visible smear on agarose gel), and 17 contained both blood and a high number of host cells. Although the difference was not statistically significant, it is noteworthy that the lowest rate of inhibitors (10%) was observed for BAL fluid samples with normal cellularity (no visible smear on agarose gel) and without macroscopically visible blood whereas the highest rate (29%) was observed for samples containing macroscopically visible blood and numerous host cells. The presence of blood alone did not seem to have an influence on the rate of inhibitors with the Chelex method compared to the results observed with the proteinase K method. This result is in accordance with previous data concerning bloody samples (18, 19). In another way, the proteinase K method seems to be less sensitive than the Chelex method to the inhibitory effect of the presence of numerous cells in the sample. However, the Chelex protocol is quicker and easier to perform. These characteristics make this method less prone to contamination from sample to sample and more suitable for a diagnostic application.