Quantitative real-time PCR tests for diagnostic and prognostic purposes in cases of legionellosis

Authors


Corresponding author and reprint requests: M. Maurin, Laboratoire de Bactériologie, CHU de Grenoble, Université Joseph Fourier, BP 217, 38043 Grenoble, Cedex 9, France
E-mail:mmaurin@chu-grenoble.fr

Abstract

Clin Microbiol Infect 2010; 16: 379–384

Abstract

The usefulness of two quantitative real-time PCR assays (qrt-PCRmip targeting Legionella pneumophila, and qrt-PCR16S targeting all Legionella species) performed on lower respiratory tract (LRT) samples for diagnostic and prognostic purposes in 311 patients hospitalized for community-acquired pneumonia (CAP) in Rhône-Alpes (France) was evaluated. The Now Legionella urinary antigen test (UAT) from Binax (Portland, ME, USA) was used as a reference test. Samples were divided into two groups. Group A included 255 CAP patients admitted to Chambery hospital in 2005 and 2006. The Now Legionella UAT was positive in 14 patients. Sensitivities, specificities, positive predictive and negative predictive values for both qrt-PCR tests were 63.6, 98.7, 77.7 and 97.4%, respectively. Group B included 56 consecutive legionellosis patients diagnosed during a 4-year period (2003–2006) at the Grenoble University Hospital. The qrt-PCR16S and qrt-PCRmip displayed a sensitivity of 82.14 and 80.4%, respectively. Among the 70 legionellosis cases, L.  pneumophila serogroup 1 was isolated in 15; qrt-PCRmip was positive in another 36, suggesting L. pneumophila infection, whereas the Legionella species involved could not be determined in the remaining 19 cases. The Legionella burden in LRT samples at the time of admission was determined in 46 patients using qrt-PCR16S tests, 44 for qrt-PCR mip groups A and B patients. It varied from 1.9 to 8.35 log10 DNA copies/mL of LRT sample for qrt-PCR16S and from 1.9 to 8.11 log10 DNA copies/mL of sample for qrt-PCRmip. High bacterial loads in LRT samples at hospital admission were significantly associated with higher Fine classes, the need for hospitalization in an intensive care unit and for prolonged hospitalization.

Introduction

Legionellosis is a pneumonia caused by inhalation of Legionella species-contaminated aerosols [1]. It is responsible for 2–15% of all community-acquired pneumonias (CAP) requiring hospital admission [2]. Of the 50 species and 72 serogroups belonging to the genus Legionella, L. pneumophila serogroup one (sg1) is responsible for at least 80% of human infections [1–3]. Mortality rates of 10–15% are usually reported [1,2] but may be higher in immunocompromised patients.

The severity of legionellosis depends primarily on the immune status of the patient and the speed of diagnosis and administration of an appropriate antibiotic therapy [1,4]. Culture-based diagnosis of legionellosis remains fastidious and its sensitivity is poor [1]. Direct fluorescence antibody staining of Legionella spp. in respiratory samples is considered unreliable [1,2,5,6]. Serological diagnosis lacks both sensitivity and specificity [5,7], and only provides a retrospective diagnosis. In recent years, L. pneumophila urinary antigen tests (UATs) have become reference tests allowing rapid legionellosis diagnosis [5,8–14]. These tests are considered highly sensitive, although they may be negative at the early stage of legionellosis, and highly specific, although urinary antigens may persist for months in some patients [12,15]. However, these tests detect primarily L. pneumophila sg1 antigens and are much less sensitive in detecting other Legionella serogroups or species [5,8,12]. PCR-based techniques have been developed as rapid diagnostic tools, potentially allowing amplification of DNA from all Legionella species and serogroups [5,16–27], but these tests are poorly standardized.

In the present study, taking the Now Legionella UAT as a reference test, we evaluated the relative sensitivities and specificities of two qrt-PCR assays for legionellosis diagnosis in CAP patients requiring hospitalization. Additionally, in a subset of legionellosis patients, we quantified L. pneumophila DNA burden in lower respiratory tract (LRT) samples at hospital admission and tentatively correlated bacterial loads with disease severity.

Materials and Methods

Patients

CAP patient inclusion criteria were as follows: patients > 18 years old, admitted to hospital with clinical and radiological findings suggestive of pneumonia (fever, cough, expectoration, dyspnoea, thoracic pain, new or progressive infiltrate on chest X-ray), in whom the infection occurred outside the hospital setting. The Fine class, a pneumonia severity index [28], was determined at the time of admission.

Patients were considered legionellosis cases when clinical signs compatible with CAP were present, and a Legionella strain was grown in culture and/or a Now Legionella UAT was positive.

Patients were divided into two groups. Group A included consecutive patients hospitalized for CAP from November 2004 to March 2006 at the General Hospital of Chambery, France. Group B included consecutive legionellosis cases, hospitalized at the University Hospital of Grenoble during a 3-year period (2003–2006).

Microbiological investigations

Usual microbiological investigations were carried out at hospital admission: two sets of blood cultures (Bactec 9240; Becton Dickinson, Grenoble, France); culture of LRT samples [i.e. bronchial aspirations and/or bronchoalveolar lavage (BAL) specimens and/or sputum samples], including Legionella culture on buffered charcoal yeast extract medium supplemented with α-ketoglutarate (Oxoid, Wesel, Germany); an indirect immunofluorescence assay (IFA) for detection of serum antibodies directed against L. pneumophila sgs1–6 (Meridian Diagnostics Inc., Cincinnati, OH, USA); the Now Legionella UAT (Binax, Portland, ME, USA) performed on urine samples 25-fold concentrated by selective ultrafiltration (Minicon B15; Millipore Corp., Bedford, MA, USA). Additionally, two Legionella qrt-PCR assays were performed on LRT samples, either tested immediately after collection or frozen at −80°C for several days to 1 month before testing.

qrt-PCR assays

DNA was extracted from the patients’ LRT samples using QIAamp DNA Mini kit (QIAGEN, Hilden, Germany). Two qrt-PCR assays previously described in the literature [29], which we had used to quantify Legionella spp. in water samples [30], were used. Analytical sensitivities, specificities, and quantification limits of these techniques have been previously described [29,30]. The qrt-PCR16S amplifies a 386-bp portion of 16S ribosomal RNA-encoding genes of all Legionella species [29,30]. The qrt-PCRmip amplifies a 186-bp portion of the mip (macrophage internalization potentiator) gene of all serogroups of L. pneumophila [29,30]. Primers, probes and amplification protocols were as previously described [30], and results obtained with the Light Cycler 2.0 instrument (Roche, Meylan, France). A 374-bp internal inhibitor control (3000 copies per reaction) was used in all experiments [30].

Mip and 16SrRNA gene amplification and sequencing

In patients with positive qrt-PCR tests but negative cultures, we attempted to amplify and sequence nearly complete mip and 16S rRNA genes directly from LRT samples, in order to identify the Legionella species involved. PCR amplification was performed using the mip gene primers and the procedure previously reported by Ratcliff et al. [31] and Stolhaugh et al. [32] and using the following primers for the 16SrRNA gene: Lg16SFw (5′-TTAACACATGCAAGTCGAACGG-3′) and Lg16SRv (5′-ACCGGAAATTCCACTACCCT-3′).

Statistical analysis

Using qrt-PCR assays, we evaluated the bacterial burden in LRT samples at the time of admission, and tentatively correlated this biological marker with age (<60 or >60 years), the Fine class at hospital admission, the need for hospitalization in an intensive care unit (ICU), and duration of hospitalization (1–14 days vs. >14 days), using the two-tailed Student’s t-test at the 95% confidence limit.

Results

Group A included 255 CAP patients, with a Fine class of 2 for 31 patients, 3 for 39 patients, 4 for 61 patients, 5 for 30 patients, and an undetermined Fine class for 94 patients.

A blood culture was positive in 14 (5.5%) cases, with isolation of Streptococcus pneumoniae in nine cases, Escherichia coli in two cases, and S. pyogenes, Klebsiella pneumoniae or Pseudomonas aeruginosa in one case each. In these patients, Legionella cultures, the Now Legionella UAT and the qrt-PCR tests were all negative.

A Now Legionella UAT was positive in 14 of 220 patients tested (6.36%) (Table 1), of whom only seven cases gave positive qrt-PCR tests. Legionella cultures from LRT samples were negative in all 199 patients tested. A single positive IFA test (IgG titre ≥256) was found in eight patients, including two with a positive Now Legionella UAT and positive qrt-PCR tests, one with a positive Now Legionella UAT but negative qrt-PCR tests, and five with negative Now Legionella UAT and qrt-PCR tests. The qrt-PCR tests were positive in nine of 195 patients tested (4.61%), including seven patients with a positive Now Legionella UAT and two patients with a negative test. These two patients were considered to be non-legionellosis cases because both Legionella culture and UAT were negative. They presented with mild disease and made a rapid recovery under ciprofloxacin therapy. They had negative acute-phase serological tests. A convalescent phase-serum was not available. We could not amplify the whole Legionella mip and 16S rRNA genes from LRT samples.

Table 1.   Laboratory results for legionellosis diagnosis in group A and group B patients
Group A patientsPositive/tested%Result/tested when Now AUT
PositiveNegative
  1. aNow UAT, Binax Now Legionella urinary antigen test; IFA, immunofluorescence assay (positive if IgG titre >256); Legionella culture and qrt-PCR tests were performed on patients’ lower respiratory tract samples.

 Now UATa14/2206.4  
 IFAa8/2203.63/145/206
 Legionella culturea0/19900/110/188
 qrt-PCR testsa9/1954.67/112/184
Group B patients
 Now UATa56/56100  
 Legionella culturea15/5228.8  
 qrt-PCR 16Sa46/5682.1  
 qrt-PCR mipa41/5180.4  

When considering the 165 patients for whom both the Now Legionella UAT and qrt-PCR test results were available, and taking the former test as a reference, the results of both qrt-PCR16S and qrt-PCRmip tests combined gave relative sensitivities, specificities, positive predictive and negative predictive values of 63.6% (95% CI 63.1–64.2), 98.7% (95% CI 98.6–98.8), 77.7% (95% CI 77.2–78.2) and 97.4% (95% CI 97.2–97.6), respectively.

Group B included 56 legionellosis patients, with a Fine class of 2 for four patients, 3 for 13 patients, 4 for six patients, 5 for ten patients, and the Fine class was undetermined for 23 patients. Blood cultures remained sterile for these 56 patients. LRT samples taken from 52 patients yielded a L. pneumophila sg1 strain in 15 cases (28.8% sensitivity) (Table 1). All 15 cases gave a positive qrt-PCR16S test, and 14 gave a positive qrt-PCRmip test (qrt-PCRmip was not done in the remaining patient). The qrt-PCR16S and qrt-PCRmip displayed sensitivities of 82.1 and 80.4%, respectively (Table 1).

We were able to determine a L. pneumophila DNA burden in LRT samples at the time of admission in 46 cases using the qrt-PCR16S test and in 44 cases using the qrt-PCRmip test. The mean ± SD of log10 DNA copies/mL of LRT sample were 6.70 ± 7.52 for qrt-PCR16S (range 1.9–8.35 log10 DNA copies/mL), and 6.48 ± 7.29 for qrt-PCRmip (range 1.9–8.11 log10 DNA copies/mL). The bacterial load was >103 DNA copies/mL of sample in 29 of 46 (63.0%) and 30 of 44 (68.2%) patients, as determined by the qrt-PCR16S and qrt-PCRmip tests, respectively.

There was a linear relationship between bacterial loads determined by both qrt-PCR16S and qrt-PCRmip (Fig. 1). The mean ± SD of log10 DNA copies/mL of LRT sample was higher in Legionella culture-positive patients (7.28 ± 7.81 for qrt-PCR16S and 7.08 ± 7.59 for qrt-PCRmip) than in culture-negative patients (4.18 ± 4.76 for qrt-PCR16S and 4.27 ± 4.81 for qrt-PCRmip), although the difference was not statistically significant.

Figure 1.

 Correlation between Legionella bacterial loads in lower respiratory tract (LRT) samples from legionellosis patients as determined by qrt-PCR16S and as determined by qrt-PCRmip.

In a subset of 32 group B patients, for whom clinical and epidemiological data were available, a high L. pneumophila bacterial load in LRT samples at hospital admission was significantly correlated with higher Fine classes (4 or 5 vs. 2 or 3) (p <0.01), with the need for admission to an ICU (p <0.01) and with hospitalization >2 weeks (p 0.01), but not with age (Fig. 2). Evidently, the first two criteria were correlated since most patients with a Fine class of 4 or 5 on hospital admission (i.e. 11/16, 68.7%) were hospitalized in an ICU.

Figure 2.

 Correlation of Legionella bacterial loads in lower respiratory tract (LRT) samples from legionellosis patients at the time of hospital admission with patients’ age (<60 or ≥60 years), the Fine class (pneumonia severity index), the need for hospitalization in an intensive care unit (ICU), and duration of hospitalization (1–14 days vs. >14 days). Open bars: qrt-PCR 16S quantification. Filled bars: qrt-PCR mip quantification. Statistical analysis using two-tailed Student’s t-test, at the 95% confidence limit.

Discussion

The sensitivity of the Now Legionella UAT in concentrated urine samples has been evaluated and ranges from 69.6 to 97.2% for L. pneumophila sg1 infections [10,33,34]. In the present work, sensitivities of the tested qrt-PCR assays were lower than that of the Now Legionella UAT. Because the Now Legionella UAT is considered highly specific [9,15,34], false-positive results with this test may not account for this difference. However, PCR inhibition could not explain such a difference, as evidenced by amplification of the internal inhibitor control incorporated in each LRT sample. Freezing the sputum samples for a prolonged period before qrt-PCR testing may have induced partial bacterial DNA degradation. Also, for the 14 patients of the study with a positive Now Legionella UAT but negative qrt-PCR tests, only poor-quality spontaneous sputum samples were available. Finally, we found the quantification limit of both qrt-PCR tests to be approximately 80 genome copies per mL of LRT sample, and patients with lower DNA copy numbers may not have been detected.

In two group A patients, both 16S rRNA gene and mip qrt-PCR tests were positive, whereas the Now Legionella UAT was negative. We were unable to confirm these two cases by amplification and sequencing of the whole Legionella mip and 16S rRNA genes. However, in both cases, clinical and epidemiological circumstances were compatible with legionellosis, and the patients were cured with fluoroquinolone therapy. Although we considered these two patients to be false-positive cases, they may instead be true legionellosis cases.

Diagnosis of legionellosis cases due to non-L. pneumophila species remains difficult, because these species are fastidious, making them difficult to grow, and they are poorly detected by Legionella UATs [5,8,12]. Thus, sensitivity levels of only approximately 45% were reported for nosocomially acquired legionellosis cases, for whom non-L. pneumophila species are more frequently involved [12]. The use of the combination of two qrt-PCR tests potentially allowed us to detect legionellosis cases due to non-L. pneumophila species, which will typically present with a positive qrt-PCR16S but a negative qrt-PCRmip. Among the 70 patients of the study with a positive Now Legionella UAT, 15 were culture-confirmed L. pneumophila sg1 infections, and 34 culture-negative cases were probably infected with L. pneumophila because the qrt-PCRmip test was positive. We could not determine the Legionella species involved in the remaining 21 patients, because Legionella culture and qrt-PCRmip test results were negative or unavailable. We could not amplify the whole mip or 16S rRNA genes from LRT samples, but since these tests were performed at the end of the study we suspect that bacterial DNA was partially degraded, precluding amplification of large DNA fragments. However, none of the 311 CAP patients investigated presented with a positive qrt-PCR16S but a negative qrt-PCRmip. Altogether, our results indicate that Legionella species other than L. pneumophila may be infrequently responsible for CAP.

The L. pneumophila DNA load in LRT samples at the time of hospital admission was very high in most patients, which contrasts with the frequent negativity of Legionella cultures. This suggests predominance of the viable but non-cultivable state of bacteria in the airways of infected patients [35]. Interestingly, in patients with multiple LRT specimens, bacterial loads were significantly inferior (≥2 log inferior) in BAL samples compared with sputum or bronchial aspiration samples. The use of saline solution during BAL processing may lower the sensitivity of DNA detection because of respiratory sample dilution.

A high L. pneumophila DNA load in LRT samples may reflect exposure to aerosols with high bacterial loads, infection with a highly virulent L. pneumophila strain and/or poor immune response in the infected host. This biological marker was significantly correlated with severity of disease, as evidenced by a higher Fine class on admission and the need for hospitalization in an ICU, and with prolonged hospitalization (>2 weeks). Although more clinical data are needed, our study suggests that determining L. pneumophila DNA load in LRT samples may help clinicians to evaluate prognosis in legionellosis patients.

In conclusion, qrt-PCR assays were useful for confirming community-acquired legionellosis cases caused by L. pneumophila, although they had lower sensitivity than the UAT. On two occasions, however, qrt-PCR tests allowed detection of L. pneumophila DNA, when the Now Legionella UAT was negative. The Legionella UATs and qrt-PCR assays should be considered complementary in the diagnostic armamentarium for legionellosis. The qrt-PCR assays were also useful for predicting disease severity, which may be a true advantage of these techniques over available diagnostic methods and therefore warrant further evaluation.

Acknowledgements

The authors thank J. N. Del Baño, S. Moulin and A. Mollard for technical assistance, and L. Northrup for English correction.

Transparency Declaration

This work was supported by a Public Health Service grant ‘Projet Hospitalier de Recherche Clinique’ 2003, from Rhone-Alpes region, France. The authors have no relationship (commercial or otherwise) that may constitute a dual or conflicting interest.

Ancillary