A nested polymerase chain reaction for detection of Legionella pneumophila in clinical specimens


Corresponding author and reprint requests: Sverker Bernander, Department of Clinical Microbiology, Karolinska Hospital, S-171 76 Stockholm, Sweden Tel: +46 8 7293539 Fax: +46 8 308099


Objective: Because presently used methods for diagnosis of Legionella pneumonia lack sufficient sensitivity and sometimes specificity and rapidity, the detection of Legionella spp. by amplification of nucleic acids might be valuable. However, performing polymerase chain reaction (PCR) on clinical samples such as sputum is difficult because of the presence of extraneous DNA and inhibitors of the reaction. An attempt to circumvent these problems was made.

Method: A nested PCR method was devised using primers from the mip gene of Legionella pneumophila. This PCR was tested on pure cultures of legionellae and clinical isolates of other bacteria. Clinical samples (bronchoalveolar lavage fluid, bronchial aspirate and sputum) from patients who suffered from legionellosis and samples from patients who suffered from other causes of pneumonia were also tested.

Results: The PCR was specific for L. pneumophila and no non-Legionella bacteria reacted. Ten to 50 colony forming units of Legionella in the sample could be detected. Twenty-two of 25 clinical samples were positive among patients suffering from pneumonia proven to be due to L. pneumophila serogroups 1, 3, 4, 5 and 6. Two of the three negative samples were from patients who had been treated with adequate therapy for at least 2 days and were culture negative. However, nine other culture-negative samples were PCR positive, of which seven came from patients who had been treated for 3–7 days. All pneumonia patients in the control group proved negative in PCR. A commercial kit for DNA preparation from clinical samples, based on absorption of nucleic acids to silica gel, was superior to the traditional phenol/chloroform extraction and increased the rapidity, simplicity and sensitivity of the procedure.

Conclusions: A nested, simplified and rapid PCR method using mip primers proved to be more sensitive than culture and as sensitive and specific as other PCR procedures previously reported.


The polymerase chain reaction (PCR) has been used for the detection of legionellae in water [1,2]. It has also been applied to clinical material, usually from the respiratory tract [3–8], but also serum and urine [9,10]. In one early study, using primer sequences of the macrophage infectivity potentiator (mip) gene, L. pneumophila could be detected at a level of 10–20 CFU/mL in bronchoalveolar lavage (BAL) fluid [3]. Mip gene primers were also used in two studies employing the commercial EnviroAmp kit (Perkin-Elmer), which was originally designed for detection of Legionella spp. in water [4,5]. Uldum studied different combinations of mip and 5S rRNA primers, of which some were less specific than others [7]. Other mip primers were used for detection of Legionella DNA in serum from patients with pneumonia [9]. The 5S rRNA primers of the EnviroAmp kit were utilized in a study on urine samples from experimentally infected guinea pigs and from human patients with confirmed legionellosis [10]. In two recent reports, DNA in clinical specimens was amplified by PCR with primers specific for the 16S rRNA gene of L. pneumophila [6,8]. The above mentioned methods have in common that they require a final hybridization step in order to achieve adequate sensitivity and specificity.

Another approach to PCR technology performs the reaction in two steps with two sets of primers in a so-called nested configuration. This method was originally applied in the detection of Mycobacterium leprae, Borrelia burgdorferi and herpes simplex virus [11–13]. The procedure increases sensitivity and specificity in DNA-rich material, since the risk of amplifying non-specific sequences is decreased, and the final hybridization step can then be omitted. A large clinical validation of a nested system has been reported for B. burgdorferi [14]. The use of a nested system has been reported in testing one patient suffering from a L. pneumophila serogroup 2 infection [15]. In two other reports, nested primers have been evaluated for detection of Legionella spp. in an experimental setting and on samples of potable water [16,17].

In the present study, a nested PCR method was developed for the detection of L. pneumophila in clinical samples from patients suffering from proven legionellosis and other pneumonias. Preliminary reports on the principle of the method and its application to clinical specimens were given by the authors at the 7th Meeting of the European Working Group on Legionella Infections (EWGLI) in Greece, May 1992, and at the 9th Meeting of EWGLI in Italy, June 1994.



The ATCC Legionella strains that were tested in pure culture are listed in Table 1. In addition to these, 12 clinical isolates consisting of L. pneumophila serogroup (sg) 1, six strains, L. pneumophila sg 3, one strain, L. pneumophila sg 5, one strain, L. pneumophila sg 6, three strains, and L. bozemanii sg 2, one strain, were tested. The following non-Legionella strains (clinical isolates) were tested: Bacillus cereus, Bacillus sp. (not identified to species level), Citrobacter freundii, Enterobacter cloacae, Enterococcus faecalis, Escherichia coli, Flavobacterium sp. group II b, Haemophilus influenzae, H. parainfluenzae, Klebsiella pneumoniae, Moraxella (Branhamella) catarrhalis, Mycobacterium tuberculosis, Pseudomonas aeruginosa, Staphylococcus aureus, Stenotrophomonas maltophilia, Streptococcus pneumoniae and Streptococcus pyogenes. The unidentified Bacillus sp. was isolated from a patient with Legionnaires’ disease (LD) during the Västerås outbreak in 1979 ([18] and unpublished observations). Bacilli with the same characteristics had also been found together with legionellae in the cooling tower that caused the outbreak.

Table 1.  ATCC strains of Legionella tested with nested PCR
SpeciesStrain and serogroup (sg)PCR result
Legionella pneumophilaPhiladelphia 1, sg 1ATCC 33152+
Legionella pneumophilaTogus 1, sg 2ATCC 33154+
Legionella pneumophilaBloomington 2, sg 3ATCC 33155+
Legionella pneumophilaLos Angeles 1, sg 4ATCC 33156+
Legionella pneumophilaDallas 1, sg 1ATCC 33216+
Legionella pneumophilaChicago 2, sg 6ATCC 33215+
Legionella pneumophilaChicago 8, sg 7ATCC 33823+
Legionella pneumophilaConcord 3ATCC 35096+
Legionella pneumophilaIN-23-G1-C2, sg 9ATCC 35289+
Legionella pneumophilaLeiden 1, sg 10ATCC 43283+
Legionella micdadeiTATLOCKATCC 33218
Legionella bozemaniiWIGA, sg 1ATCC 33217
Legionella bozemaniiToronto 3, sg 2ATCC 35545
Legionella longbeachaeLongbeach 4, sg 1ATCC 33462
Legionella longbeachaeTucker 1, sg 2ATCC 33484
Legionella dumoffiiNY-23ATCC 33279
Legionella gormaniiLS-13ATCC 33297

L. pneumophila, Philadelphia 1, sg 1 (ATCC 33152) was used as a control in each PCR run. Legionella strains were grown on BCYEα agar for 3 days at 37°C and suspended in ultrafiltered sterile water to a concentration of 108/mL. Non-Legionella strains were grown according to the requirements of each species and suspended in ultrafiltered sterile water to a concentration of 108/mL. The cell concentration was measured by a visual count and also checked by a viability count in the different dilutions before preparing DNA. The sensitivity of the method was tested with dilutions in water, BAL fluid and sputum.


Routine specimens from patients with suspected legionellosis were tested by culture and/or detection of L. pneumophila sg 1 urinary antigen. Respiratory tract samples from those patients who proved positive in either of the two tests were retested by PCR. The interval between initiation of adequate therapy (erythromycin and in two cases sparfloxacin) and sampling from the respiratory tract was recorded. Samples from 23 consecutive patients who suffered from proven clinical pneumonia were also tested by culture, urinary antigen and PCR. These patients were not serologically tested for Legionella but were considered to suffer from pneumonia caused by other infectious agents because of results from other tests. Another 10 samples consisting of BAL fluid or bronchial aspirate were obtained from nine consecutive patients suffering from pneumonia who had not responded to antibiotic therapy. These patients were tested for Legionella with the above methods and the indirect immunofluorescence antibody test (IFAT) on serum. One culture-positive postmortem sample from lung tissue that grew L. longbeachae sg 2 was tested by PCR.

Culture and urinary antigen detection

Respiratory tract specimens were cultured on BCYEα, MYW and BMPAα agars [19]. Mucous specimens were treated with an equal volume of 10% dithiothreitol (Sputolysin, Behring Diagnostics, La Jolla, USA). Culture was then performed on the undiluted specimen and on an acid-treated dilution (1:10 in 0.2 M HCl/KCl for 5 min). The plates were incubated at 37°C for 10 days. Urinary antigen was detected with the Equate Legionella Urinary Antigen RIA kit (Binax, Portland, Me, USA). IFAT was performed in some cases at The Swedish Institute for Infectious Disease Control, Stockholm.

Preparation of DNA for PCR

Suspensions of pure bacterial cultures in ultrafiltered sterile water were heat lysed. Ten microliters was then used for PCR. BAL fluid, bronchial aspirate and sputum were prepared according to two methods as follows without centrifugation:

  • 1Two hundred and fifty microliters of sample material was added to an equal volume of lysis buffer (20 mM Tris, 100 mM KCl, 100 mM MgCl2, 1% Nonidet P40, 1% Tween-20, pH 8.2) and 20 μL of proteinase K (10 mg/mL). After incubation at 55°C for 60–120 min and subsequent cooling, the sample was extracted once with 500 μL of phenol/chloroform/isoamyl alcohol (50:49:1) according to a protocol described previously [14]. Ten microliters of the final DNA preparation in TE buffer (10 mM Tris, 0.1 mM EDTA, pH 8.0), diluted 1:10, was used as a template in the PCR assay.
  • 2The QIAamp Tissue Kit (QIAGEN, GmbH, Hilden, Germany) was used for rapid preparation of DNA from clinical specimens. This method is based on the principle of adsorbing nucleic acids onto a silica gel membrane in the presence of chaotropic salts. Carbohydrates, proteins and inhibitors pass through. The manufacturer's instructions were followed. Two hundred microliters of the clinical material was added to the recommended amount of lysis buffer containing proteinase K. The nucleic acids were finally eluted in 100 μL of TE buffer, pH 9. Ten microliters of this preparation was used in the PCR.

Primers for PCR amplification

The assay was performed as a nested PCR. The target genome is the gene coding for the macrophage infectivity potentiator protein (mip), which has been mapped and which has been shown to be an important virulence factor in L. pneumophila [20]. The external primers were the same as those used by Mahubani et al., reported to be specific for L. pneumophila, all serogroups [21], and to consist of the following sequences: 5′-GCT ACA GAC AAG GAT AAG TTG-3’(920–940) and 5′-GTT TTG TAT GAC TTT AAT TCA-3’(1568–1548). They amplify a 649-bp product. The internal primers were specially chosen for this study, with the sequences 5′-CAT GCA AGA CGC TAT GAG TG-3’(1021–1040) and 5′-CAA GTT GAT CCA GCT GGC AT-3’(1423–1392) giving rise to a 403-bp product.

PCR amplification

The PCR mixture (50 μL) contained 10 mM Tris, (pH 8.8), 50 mM KCl, 0.1% Triton X-100, 4 mM MgCl2, 0.2 mM of each nucleotide and 1.6 units of DNA polymerase (DynaZymeTM, Finnzymes OY, Espoo, Finland). External primers were added to a concentration of 0.12 μM and the internal primers to a concentration of 1.6 μM. The first step of the nested PCR, with the prepared DNA and the external primers added, was performed with 30 cycles, each consisting of denaturation at 95°C for 30 s (5 min in the first cycle), annealing at 55°C for 30 s, and extension at 72°C for 1 min (5 min in the last cycle). In order to diminish non-specific amplified products the reaction mixtures and prepared PCR tubes were kept on ice prior to loading the DNA Thermal Cycler (Perkin Elmer Cetus). The second step with the internal primers was run under the same conditions as the first step except that 5 μL of amplified DNA from the first step diluted 1:10 was used as a template. Ten microliters of the amplified product from the second step was run on a 3% agarose gel, stained with ethidium bromide and photographed under UV light. Lanes with a distinct band at the exact level of the positive control band (403 bp) were regarded as positive.

Each run contained a positive control prepared from L. pneumophila sg 1 in ultrafiltered sterile water. Two aliquots of every clinical sample were extracted and amplified. One of these was spiked with 50 cells of the control strain and run separately from the test aliquot in order to exclude false negativity because of inhibitors. Each run also contained a negative sputum sample and several blank tubes, positioned between the test samples and containing the PCR mix but with no template added, in order to detect possible contamination. To prevent cross-contamination, the different steps in the procedure were physically separated and the amplified product was handled in a hood in a separate room. This hood was decontaminated after each assay. No equipment was moved between the separate areas. Gowns and gloves were changed between each step in the procedure.

Specificity of the amplificate

Two amplificates, one from a known strain of L. pneumophila sg 1, the other from a clinical sample in a patient who had grown L. pneumophila sg 6, were sequenced to check the specificity. The sequencing was performed with a Taq DyeDeoxyTM Terminator Cycle sequencing kit and the ABI PRISM 373 Sequencer (Perkin-Elmer Applied Biosystems Division, Foster City, Ca, USA).


Lysates of L. pneumophila produced distinct bands (Figure 1, Table 1). The other Legionella species tested proved negative except for a weak band produced by L. bozemanii sg 2 (clinical strain). The sensitivity of the test was in the range 10–50 CFU of Legionella in the sample volume from the original material, which corresponds to 1–5 CFU in the PCR test. The result was the same when legionellae were diluted in BAL fluid or liquefied sputum. All non-Legionella bacteria lacked specific bands.

Figure 1.

Agarose gel electrophoresis of nested PCR products. Lane 1: DNA molecular weight marker VIII (Boehringer Mannheim GmbH, Mannheim, Germany). Lanes 2, 4 and 6: three aliquots from the same DNA preparation of a sputum sample from a patient with Legionella infection. Lanes 3, 5, 7 and 10 are negative controls. Lane 8: a negative sputum. Lane 9: the same sample as in lane 8, but with 50 CFU of L. pneumophila sg 1 (ATCC 33152) added. Lane 11: positive control of heat-lysed L. pneumophila sg 1 (ATCC 33152).

Twenty-three patients were positive for Legionella infection by culture or L. pneumophila sg 1 urinary antigen assay or both (Table 2). PCR was performed on 25 specimens (four BAL fluid, 11 sputum, nine bronchial aspirate, and one pleural fluid) from these patients. Of the 25 specimens, 15 were from patients who suffered from pneumonia caused by L. pneumophila sg 1, three by sg 3, one by sg 4, one by sg 5, and five by sg 6. The 14 culture-positive specimens grew L. pneumophila sg 1 in six cases, sg 3 in three, sg 5 in one, and sg 6 in five. Thus, in nine patients with sg 1 infection, diagnosis relied on a positive urinary antigen test. Two of these patients had been serologically tested but found to be negative in a serum sample from the acute phase. The patient with sg 4 infection showed a fourfold titer rise and proved weakly positive in the urinary antigen test but was negative in sputum culture.

Table 2.  Results of PCR testing on patients with Legionella pneumonia in relation to previous antibiotic treatment: 25 samples from 23 cases
 No. of samples
Test results2 days3–7 daysTotal
  1. aNo information about therapy for one patient.

  2. bIncludes one bronchial aspirate overgrown by Serratia sp. and one pleural fluid. Both patients proved culture positive in an earlier sample. Seven specimens from patients with positive urinary antigen test.

  3. cBoth specimens from patients with a positive urinary antigen test, of whom one was positive in serology (IFAT) for L. pneumophila sg 4.

Culture +/ PCR +10213a
Culture +/PCR -101
Culture -/PCR +279b
Culture -/PCR -112c

Positive bands of amplificate in clinical samples were distinct (Figure 1). Non-specific bands were always considerably weaker than the specific band. Non-specific bands were seen in some negative samples due to the presence of extraneous DNA. They did not interfere with the reading. The results are given in Table 2, including data regarding the interval between initiation of adequate therapy and sampling for PCR and culture. Notably, 7/9 culture-negative but PCR-positive samples were collected 3–7 days after the start of adequate treatment, whereas the culture positives in 11/13 evaluable cases were sampled at an earlier stage. Five clinical specimens proved PCR positive when DNA was prepared by the QIAamp method instead of by phenol/chloroform extraction. The test was run again on three aliquots of the same DNA preparation from each sample in three patients, who were urinary antigen positive but PCR negative initially. One or two of these aliquots from each patient then became positive, showing an uneven distribution of the DNA template in samples with few bacteria (Figure 1). Contamination was excluded, since blank tubes between the sample tubes were negative.

Seven of the 11 culture-negative samples were sputum and one was pleural fluid. However, five of these samples and the pleural fluid were PCR positive. Two culture- and PCR-positive samples, one sputum and one bronchial aspirate, produced only two or three colonies on each plate, and another sputum sample that proved PCR negative only produced one colony of L. pneumophila sg 5 on each of two plates. Ten of 14 culture-positive specimens were BAL fluid or bronchial aspirate.

The 33 samples from pneumonia patients who were negative on Legionella testing (in 10 cases including serology) proved negative in PCR. The postmortem sample from lung tissue that grew L. long-beachae sg 2 showed no bands in PCR.

Sequencing 250 bp of the amplificate showed full agreement with the sequence published previously [18], except for 1 bp.


Methods used presently for diagnosis of Legionella infection lack sufficient sensitivity and sometimes also specificity [22]. Culture is the most specific method, but sensitivity is variable. Serology is not of value in the acute phase of disease and the serologic methods have not been satisfactorily validated in the diagnosis of non-L. pneumophila sg 1 infections. Direct immunofluorescence (DFA) of specimens from the respiratory tract using monoclonal fluorescein isothiocyanate (FITC)-conjugated antibody is specific but insensitive and needs much experience to read. Detection of Legionella antigen in urine using enzyme immunoassay (EIA) or radioimmunoassay (RIA) is very specific, but the commercial systems only detect L. pneumophila sg 1 infections. Therefore, there is a need for other rapid and specific methods that are more sensitive than those now used.

Methods for amplification of DNA and RNA have become important in diagnosing infections caused by slow-growing and fastidious organisms. Such methods are very sensitive (in theory detecting one bacterial cell) and also very specific. However, they are still too labor-intensive to be practical in clinical use. This may be due to the complicated extraction methods used or to the need for hybridization with a probe in the final step of the test. We used a simplified extraction procedure, which improved the results of testing, and we also used a nested system for amplification, which effectively removes the need for hybridization.

A nested PCR method in our setting can detect 10–50 Legionella CFU in the original sample, and similar numbers when bacterial cells are added to BAL fluid or sputum. This should be quite adequate unless the method is hampered by problems in processing clinical specimens or by inhibitors in them. Positive bands were easy to read. Though negative lanes in the ethidium bromide-stained gel would sometimes show non-specific bands, this did not constitute a problem. The preparation of DNA from clinical specimens with the rapid QIAamp method increased sensitivity and rapidity. The specificity of the amplified product could be confirmed by sequencing two amplificates from different serogroups. Therefore, we do not believe that a final hybridization with an internal probe is needed. No false-positive results were obtained with either the clinical specimens or the cultured non-L. pneumophila strains. Though the use of a nested system theoretically might increase the risk of cross-contamination, this was not seen. Thus, our precautions, described above, were sufficient.

Because legionellosis is relatively rare, it was not found feasible to study samples from patients with the suspected illness consecutively. It is therefore not possible to calculate sensitivity in the usual terms. However, all culture-positive specimens, mostly sampled in an early phase of disease, were also positive in PCR, except for one culture that only grew one legionella colony on each of two plates. We have not been able to establish any reason for the negative PCR in this latter case. However, the result may be an incidental effect of random distribution in samples containing very few Legionella cells. This is substantiated by the fact that different aliquots of the same DNA preparation from culture-negative specimens could vary in positivity. It might therefore be prudent to perform PCR on more than one aliquot from each sample preparation. Seven of nine samples from patients who were urinary antigen positive, but culture negative, proved positive in PCR. Of the total nine culture-negative and PCR-positive samples, seven were from patients who had been treated adequately with erythromycin or sparfloxacin for 3–7 days. This seems to be an important advantage of the PCR technique. Therefore, the sensitivity of our PCR on respiratory tract samples is probably very high.

Since the control group of patients suffering from other causes of pneumonia was completely negative in PCR, specificity should be adequate. Not all of these had been serologically tested for Legionella, but they were clinically judged to suffer from other kinds of pneumonia. Positive urinary antigen tests per se have generally only been regarded as indicative of a presumptive diagnosis of legionellosis, but objections to this view have been expressed [23]. Though cross-reactions might occur with other Legionella serogroups, as in the sg 4 infection mentioned above, specificity is close to 1.0 in cases of clinical pneumonia. The positive urinary antigen tests in our samples were well above cut-off values.

Several amplification methods have now been reported. Those relying on the 5S rRNA gene seem to have problems with specificity in clinical specimens [7,10]. However, the primers used cover a broader spectrum of Legionella spp. Primer sequences from the mip gene are in some instances only specific for L. pneumophila, and in others seem to cover a broader range of species [3,7]. Mip genes are also found in other Legionella spp. The external primers that we used are principally specific for L. pneumophila, though we observed a weak reaction with a L. bozemanii strain. This narrow specificity might be a disadvantage but still covers approximately 90% of Legionella infections [22]. The sensitivity of our method on diluted specimens was the same irrespective of the diluent used (distilled water, BAL fluid or sputum). The QIAamp Tissue Kit seemed to remove the inhibitors from the clinical samples. Recently, the use of primers from 16S rRNA gene has been reported [6,8]. These PCRs have been run in traditional fashion, hybridizing the final product with a probe. They have shown the same level of sensitivity as ours but cover a broader range of Legionella spp. Our method has been run as a nested system. A nested system using different mip primers from ours has been reported, but in this case only one patient with a L. pneumophila sg 2 pneumonia was tested [15]. We believe that combining the use of mip primers with the use of a nested PCR may be advantageous in constructing a rapid amplification test for clinical samples without loss of specificity and sensitivity.

In summary, we have designed a nested PCR which is as sensitive and as specific as those previously described without needing a final hybridization step. Full sensitivity demands an efficient preparation of DNA from clinical samples in order to deprive the material of inhibitors to the reaction. This can be done with a commercially available kit, adsorbing DNA onto silicate gel. Further prospective studies on consecutively studied patients with atypical pneumonia are needed for full evaluation.