Evaluation of a multiplex PCR for bacterial pathogens applied to bronchoalveolar lavage

Study subjects

The present study was performed prospectively at the Silkeborg County Hospital, Silkeborg, Denmark, in its Dept of Internal Medicine, which has 100 beds. Silkeborg County Hospital is the only hospital in the region and serves a basic population of 100,000 inhabitants in the area. As previously described 14, all immunocompetent adult patients hospitalised for LRTI during weekdays between September 1997 and August 2000 were consecutively included in the present study. Thus, the included patients were unselected adult patients with LRTI requiring hospitalisation. The clinical diagnosis of LRTI required that the patient had fever (rectal temperature ≥37.6°C within 48 h of inclusion in the study) and/or an increased leukocyte count (≥11×10⁹ leukocytes·L^-1) in peripheral blood on admission, together with increased focal symptoms from the lower airways with at least one of three newly developed symptoms of increased dyspnoea, increased coughing and/or increased sputum purulence. Patients with known malignancy and patients with oxygen saturation <85% with a maximum of 1 L nasal oxygen were not included. Chest radiograph and spirometry were performed. Demographic, comorbidity and smoking data were collected, and information on prior antibiotic treatment for any diagnosis was obtained from either the individual patient or the patient's general practitioner.

Adult patients who consecutively underwent FOB for suspected malignancy were enrolled as controls. Information about prior antibiotic treatment and smoking data were collected.

Bronchoscopy

All patients and controls underwent a standardised FOB with BAL 15 within 24 h of admission. Prior to the FOB, the mouth and pharynx were anaesthetised with a solution of 4% lidocaine. The fibre bronchoscope was introduced through the nose or through the mouth and additional anaesthetics (1% lidocaine solution) were applied as the scope passed the larynx and trachea. A sterile, thin tube was introduced in the working channel of the bronchoscope. The tip of the scope was wedged in an appropriate segment in the bronchus, and thereafter lavage was performed. A segment of bronchus affected by a new pulmonary infiltrate, as seen at a prior chest radiograph, was chosen for lavage. In the LRTI patients without radiographically observed infiltrates and the control patients, the middle lobe was chosen for lavage. Between one and three portions of 60 mL of isotonic sodium chloride solution were used for lavage and the aspirated fluid was collected in one single portion for microbiological analyses.

Conventional microbiological investigations

BAL fluid and sputum samples from the LRTI patients and BAL fluid from the controls were analysed with culture at the Dept of Clinical Microbiology, Aarhus University Hospital (Aalborg, Denmark), within a maximum of 6 h from the time of sampling. The specimens were cultured on 5% horse blood agar and chocolate agar with semiquantitative determinations by dispersion of 1 and 10 μL on each half of the plate. The plates were incubated in 5% carbon dioxide at 35°C for 24–48 h. Bacterial identification was performed according to standard microbiological methods 16. The cut-off limit for a positive BAL or sputum culture result was 10² colony-forming units (cfu)·mL^-1 sample. After culture, the BAL fluid was frozen at -20°C.

BAL samples and throat swabs, which were placed in two sucrose phosphate chlamydial transport media, from the LRTI patients where sent to the Dept of Bacteriology, Mycology and Parasitology, Statens Serum Institute (Copenhagen, Denmark), and were analysed with singleplex PCR (sPCR) assays for M. pneumoniae, C. pneumoniae, Chlamydophila psittaci, and Legionella spp. 17, 18. In addition, the BAL samples and throat swabs were cultured for C. pneumoniae 19.

From the LRTI patients, blood samples were collected for culture with a Bactec blood-culturing system (BioMérieux, Marcy-ĺEtoile, France) at the Dept of Clinical Microbiology, Aarhus University Hospital. Nonfrozen urine samples were sent to the Dept of Bacteriology, Mycology and Parasitology, Statens Serum Institute, and were analysed for pneumococcal capsular polysaccharides by countercurrent immunoelectrophoresis 20, and for Legionella pneumophila serogroup 1 antigen by enzyme immunoassay (Biotest AG, Dreieich, Germany).

Multiplex PCR

Frozen BAL samples from the LRTI patients and controls were sent to the Dept of Clinical Microbiology, Örebro University Hospital (Örebro, Sweden), for mPCR. A detailed description of the development and procedures of the mPCR is presented in a previous paper 12. mPCR was performed under blinded conditions for simultaneous identification of specific genes for S. pneumoniae (lytA, 229 bp), M. pneumoniae (P1, 483 bp), C. pneumoniae (ompA, 368 bp), and H. influenzae (16S rRNA, 538 bp). The following primers were used: lytA, 5′-CGGACTACCGCCTTTATATCG-3′ and 5′-GTTTCAATCGTCAAGCCGTT-3′; P1, 5′-ACTCGGAGGACAATGGTCAG-3′ and 5′-CAAACCCGGTCTTTTCGTTA-3′; ompA, 5′-ACACGATGCAGAGTGGTTCA-3′ and 5′-TGTTTACAGAGAATTGCGATACG-3′; and 16S rRNA, 5′-TCCTAAGAAGAGCTCAGAGAT-3′ and 5′-TGATCCAACCGCAGGTTCC-3′. The concentration of all primers was 50 μM. Briefly, DNA from 0.2–0.5 mL BAL fluid was extracted using the automatic MagNa Pure LC DNA-Isolation system (Roche Diagnostics, Mannheim, Germany). Extracted DNA (10 μL) was added to a 40-μL mixture to give a total mPCR mixture volume of 50 μL, which included 90 nM of each of the two lytA primers and 250 nM of each of the other six primers, 800 μM of the deoxyribonucleoside triphosphate mixture (Perkin-Elmer Applied Biosystem, Norwalk, CT, USA), 1.5 U of AmpliTaq Gold DNA Polymerase (Perkin-Elmer Applied Biosystem), and 10× PCR buffer (Perkin-Elmer Applied Biosystem). PCR amplification was performed on a GenAmp PCR system 9600 (Perkin-Elmer Applied Biosystem) with the following parameters: 94°C for 10 min followed by 40 cycles at 94°C for 30 s, 58°C for 30 s, and 72°C for 1 min, with the last cycle concluding with 72°C for 7 min, before storage at 4°C. The amplified sample was examined by electrophoresis on an agarose gel containing ethidium bromide. The result was compared with a DNA molecular weight marker. In every PCR run, one negative control and positive controls for each one of the four investigated bacteria were included. Samples positive for the 16S rRNA gene were considered positive for H. influenzae if they were also positive in a verification PCR for the P6 gene, using the same PCR protocol. The P6 PCR mixture included 125 nM of each P6 primer (5′-TTGGCGGWTACTCTGTTGCT-3′ and 5′-TGCAGGTTTTTCTTCACCGT-3′). All mPCR-negative samples were spiked with a reference strain of S. pneumoniae and were re-run by the PCR protocol for control of inhibition.

By serial dilution of bacterial strains, the detection levels of the mPCR have been shown to be 10² cfu·mL^-1 of sample for S. pneumoniae and H. influenzae, and 10³ genome copies or inclusion-forming units per millilitre of sample for M. pneumoniae and C. pneumoniae 12.

Criteria for aetiologies and exclusion of aetiologies

The criteria for LRTI aetiologies used in the present study were as follows. S. pneumoniae: positive blood culture, positive urinary antigen test, positive BAL culture, or positive sputum culture; H. influenzae: positive blood culture, positive BAL culture, or positive sputum culture; M. pneumoniae: positive BAL or throat swab sPCR; C. pneumoniae: positive BAL or throat swab sPCR, or positive BAL or throat swab culture. These criteria were used in the calculations of sensitivities and specificities of BAL mPCR. If no available reference analysis was positive for a pathogen, the pathogen was ruled out as an aetiological agent in the specificity analysis.

Ethics

The study was performed according to the Declaration of Helsinki II and was approved by the local ethical committee; all participating patients gave written consent.

Patients and controls

During the study period, 167 patients fulfilled the study criteria for LRTI. Eight patients declined participation, thus 159 patients were included in the prospective study. BAL fluid from 156 patients was available for mPCR and these patients were included in the present study. The LRTI patients had a median age of 63 yrs (range 26–90 yrs), 48 (31%) patients were current and 62 (40%) were previous smokers, and the mean pack-yrs history in the current and previous smokers was 27 yrs. A chronic lung disease was documented in 72 (46%) patients, and the mean forced expiratory volume in one second on day 7 from inclusion was 58% of predicted. New infiltrates were identified by radiography in 87 (56%) LRTI patients. The median interval between illness onset and FOB was 8 days; FOB was performed within 2 weeks of illness onset in 72% of the patients, and FOB was performed within 24 h of hospitalisation in all patients. In 103 (66%) patients, antibiotics had been taken within 7 days prior to FOB. The median length of hospital stay was 8 days (range 1–40 days). One patient died within 30 days of admission.

Out of the 51 controls included, 12 were excluded as they were diagnosed by radiography as having pulmonary infection with infiltrates (n = 9), lung abscess (n = 1), empyema (n = 1), and bronchiectasies (n = 1). One control was excluded as no BAL culture was performed, and another two controls were excluded as they had taken antibiotics during the preceding 7 days. Of the remaining 36 controls, 22 had lung malignancies and 14 had no pathology identified by FOB or radiological examinations. The median age of the 36 controls was 63 yrs (range 30–77 yrs), and 32 (89%) were current or previous smokers.

Identification of pathogens in patients with LRTI

In 55 LRTI patients (35%), any of the four studied pathogens was considered to be the aetiology, i.e. S. pneumoniae in 18 (12%) patients, H. influenzae in 30 (19%) patients, both S. pneumoniae and H. influenzae in two (1.3%) patients, M. pneumoniae in five (3.2%) patients, and C. pneumoniae in no patients.

S. pneumoniae was identified by blood culture in six (3.9%) out of 152 patients tested, by urinary antigen test in nine (6.3%) out of 142, by BAL culture in 10 (6.4%) out of 156, and by sputum culture in three (5.3%) out of 56. Two of those with positive sputum culture were also BAL-culture positive for S. pneumoniae. H. influenzae was identified by BAL culture in 31 (20%) out of 156 patients tested and by sputum culture in six (11%) out of 56, but in no blood culture. Five out of the six patients with positive sputum culture were also BAL-culture positive for H. influenzae. M. pneumoniae was identified in five patients by both BAL sPCR and throat swab sPCR. BAL sPCR was performed in 154 patients and throat swab sPCR in 156 patients. No patient was positive for C. pneumoniae, although BAL sPCR was performed in 155 patients, throat swab sPCR in 155, BAL culture in 154, and throat swab cultures in 154.

No patient was positive for Legionella spp. by urinary antigen test (154 tested), BAL sPCR (155 tested), or throat swab sPCR (156 tested). No patient was blood- or sputum-culture positive for Staphylococcus aureus or Gram-negative enteric bacilli. However, BAL culture identified S. aureus in 10 patients, Gram-negative enteric bacilli in 10 patients, Moraxella spp. in five patients, Neisseria meningitidis in three patients, and Enterococcus faecalis in one patient. sPCR identified C. psittaci in one patient.

Streptococcus pneumoniae

S. pneumoniae was an aetiological agent in 22 (14%) patients by the conventional tests, and was identified by BAL mPCR in 44 (28%) of the 156 LRTI patients. With all tests included, S. pneumoniae was identified in 30% of the LRTI patients. Table 1⇓ demonstrates moderate-to-high sensitivity and negative predictive value, but low specificity and positive predictive value of BAL mPCR for S. pneumoniae. Among the six patients with pneumococcal bacteraemia, S. pneumoniae was identified by BAL mPCR in all six patients and by BAL culture in one patient. Likewise, of the nine patients with positive urinary antigen test, S. pneumoniae was identified in seven by BAL mPCR and in none by BAL culture. Among the 87 patients with infiltrates shown by radiography, mPCR for S. pneumoniae had a sensitivity of 94% (16 out of 17) and a specificity of 73% (51 out of 70).

Haemophilus influenzae

H. influenzae was an aetiological agent in 32 (21%) patients by the conventional tests, and was identified by BAL mPCR in 73 (47%) of the 156 LRTI patients. While the sensitivity and negative predictive value of BAL mPCR for H. influenzae were moderate to high, the specificity and positive predictive value were low (table 1⇑). Among the 87 patients with infiltrates shown by radiography, mPCR for H. influenzae had a sensitivity of 84% (16 out of 19) and a specificity of 62% (42 out of 68).

H. influenzae was identified in 20% by BAL culture and 51% by BAL mPCR in the 110 current or previous smokers compared with 22% by BAL culture and 37% by BAL mPCR (not significant) in the 27 patients who had never smoked.

Mycoplasma pneumoniae

M. pneumoniae was an aetiological agent in five (3.2%) LRTI patients. There was total agreement between BAL sPCR, throat swab sPCR and BAL mPCR for identification of M. pneumoniae in the present study (table 1⇑). None of these patients had any other pathogen identified.

Chlamydophila pneumoniae

No patient had positive sPCR or positive culture for C. pneumoniae. The patient with BAL mPCR positive for C. pneumoniae had no throat swab sPCR for C. pneumoniae performed, but BAL sPCR and the two C. pneumoniae cultures were negative.

Identification of more than one pathogen

The relationship between the different tests for S. pneumoniae and H. influenzae is shown in table 2⇓. Both of these pathogens were identified in 18 patients as follows: S. pneumoniae, but not H. influenzae, was identified in 29 patients; and H. influenzae, but not S. pneumoniae, was identified in 59 patients. Among the 22 patients with S. pneumoniae aetiology, BAL culture was positive for S. aureus in two patients and Moraxella spp. in one patient. Two patients with H. influenzae aetiology were also BAL-culture positive for Gram-negative enteric bacilli.

Diagnostic influence of antibiotic treatment prior to bronchoscopy

Among 53 patients with no antibiotics taken prior to FOB, BAL culture and BAL mPCR identified S. pneumoniae in 13% (n = 7) and 23% (n = 12), respectively, and H. influenzae in 19% (n = 10) and 42% (n = 22), respectively, of the patients (nonsignificant). A bacterial concentration of ≥10⁴ cfu·mL^-1 in BAL culture was identified in six out of seven patients identified with S. pneumoniae and in eight out of 10 patients identified with H. influenzae.

In the 103 patients with antibiotics taken prior to FOB, BAL culture and BAL mPCR identified S. pneumoniae in 2.9% (n = 3) and 31% (n = 32), respectively (p<0.001, Chi-squared test), and H. influenzae in 20% (n = 21) and 50% (n = 51), respectively (nonsignificant), of the patients. A bacterial concentration of ≥ = 10⁴ cfu·mL^-1 in BAL culture was identified in one out of three patients identified with S. pneumoniae and in 16 out of 21 patients identified with H. influenzae. One LRTI patient with pneumococcal bacteraemia had taken antibiotics prior to blood sampling and FOB, and S. pneumoniae was identified by BAL mPCR but not by BAL culture in that patient. Of the five patients with M. pneumoniae aetiology, three had taken antibiotics prior to FOB.

Bacterial identification in the control group

Table 3⇓ shows the pathogens identified by BAL culture and BAL mPCR in the control group. Among the 36 controls, either BAL culture or BAL mPCR identified S. pneumoniae in six (17%) controls and H. influenzae in 14 (39%) controls. In the controls with lung malignancy, S. aureus was identified in three controls and Gram-negative enteric bacilli in two controls. S. aureus was also identified in one control without identified pathology. No control was mPCR positive for either M. pneumoniae or C. pneumoniae.

PCR inhibition

No PCR inhibition was identified in any BAL mPCR analysis.