Comparison between real-time PCR, conventional PCR and different staining techniques for diagnosing Pneumocystis jiroveci pneumonia from bronchoalveolar lavage specimens

Pneumocystis jiroveci, formerly known as Pneumocystis carinii f. sp. hominis, is a fungal pathogen that causes Pneumocystis pneumonia (PCP), a common and serious opportunistic infection in immunocompromised patients. Diagnosis of PCP in the laboratory was, until a few years ago, dependent on visualization of Pneumocystis organisms in stained preparations of appropriate respiratory specimens using the Giemsa and GomoriGrocott techniques. Sensitivity of the staining techniques, although acceptable (7092 %) for bronchoalveolar lavage (BAL) specimens (Wakefield et al., 1990; Cregan et al., 1990), was low (3578 %) for aspirates and induced sputum samples (Zaman et al., 1988; Wakefield et al., 1991; Lipschik et al., 1992). In the last 10 years, PCR (Wakefield et al., 1990) has considerably increased sensitivity of detection of Pneumocystis, which is now 86100 % in BAL, aspirates and induced sputum specimens (Roux et al., 1994; Cartwright et al., 1994; Caliendo et al., 1998). However, one problem with this technique is that it is frequently positive in patients with asymptomatic carriage, with a rate of 221 % (Sing et al., 1999, 2000; Nevez et al., 1999; Helweg-Larsen et al., 2002; Takahashi et al., 2002; Maskell et al., 2003). Real-time PCR, recently described by Larsen et al. (2002), could have a role in distinguishing between colonization and clinical disease. The aim of this study was to compare sensitivity and specificity of different techniques (standard staining, conventional PCR and real-time PCR) used in diagnosis of PCP and, as suggested by Larsen et al. (2002), to propose an adapted cut-off value for differentiating between carriage and PCP using real-time PCR. Patients and clinical samples.
Samples for this study came from all patients seen in the University Hospital of Saint Etienne between January 2002 and July 2003 presenting with clinical symptoms of pulmonary infection associated with immunosuppression, justifying a search for Pneumocystis in BAL.

From 150 patients with immunosuppression due to different diseases and/or immunosuppressive treatments, 173 BAL samples were obtained. Samples comprised 138 specimens retrieved by washing with 50 ml 0.9 % NaCl (BAL-50) and 35 retrieved by washing with 150 ml 0.9 % NaCl (BAL-150). Of these 150 patients, 19 were HIV seropositive and the rest were seronegative.

Sample preparation and classical staining.
All BAL samples containing mucous material had twice their volume of 0.9 % NaCl added and they were then mixed vigorously for 5 min. To avoid non-specific inhibition of PCR, no detergents or mucolytic agents were used in this step. Samples were centrifuged at 3000 g for 10 min and pellets were resuspended in 1/10 of the starting volume. A portion of the resuspended pellet (100 µl) was used to prepare smears for Giemsa and GomoriGrocott staining and smears were examined by two microscopists experienced in Pneumocystis diagnosis. Additional samples of 200 µl of the resuspended pellets were stored at 4 °C until used for DNA extraction and amplification.

DNA extraction, conventional PCR and real-time PCR.
To detect Pneumocystis DNA in BAL samples, 200 µl of each resuspended pellet was used for DNA extraction with the QIAamp DNA Mini kit (Qiagen) according to the manufacturer's recommendations.

Conventional PCR.
PCR was performed according to Wakefield et al. (1990). The PCR mixture (50 µl) contained 10 pmol of each primer, pAZ102E and pAZ102H, derived from the mitochondrial large subunit rRNA (mtLSUrRNA) gene, 200 µM dNTPs (dATP, dCTP, dGTP and 3 x dUTP), 3 mM MgCl₂, 5 µl 10 x PCR buffer, 1.25 U Taq DNA polymerase, 1 IU uracyl DNA glycosylase (PCR Core Kit Plus; Roche Molecular Biochemicals) and 10 µl purified DNA. Denaturing, annealing and extension times were 1 min each, at 95, 58 and 72 °C, respectively. DNA samples were amplified for 40 cycles. The specific product (364-bp fragment) was separated on a 3 % agarose gel (Metaphor agarose; Tebu) and detected after staining with ethidium bromide under UV illumination.

Real-time PCR with the LightCycler.
Real-time PCR was performed with fluorescence resonance energy transfer (FRET) hybridization probes using the LightCycler. PCR was performed with LC FastStart DNA Master hybridization probes (Roche Molecular Biochemicals) in a quantitative touch-down real-time PCR as previously described by Larsen et al. (2002). Briefly, 20 µl of a mixture containing Fast Start Taq DNA polymerase, dNTPs (with dUTP instead of dTTP), 5 mM MgCl₂, 0.5 IU heat-labile uracyl-DNA glycosylase (Roche Molecular Biochemicals), 10 pg mouse DNA, 5 µl PCR extract, 1 µM of each selected primer and 0.20 µM of each hybridization probe was used. Internal control was the mouse galactose-1-phosphate uridyl transferase (GALT) gene (Costa et al., 2001). Specific primers and probes were used for co-amplification of the internal control (Costa et al., 2001) and the MSG gene of Pneumocystis (Larsen et al., 2002). A double fluorescence reading for each sample was taken at the annealing step: one for LCRed 640 used for the Pneumocystis probe (channel 2) and the other for LCRed 705 used for the internal control probe (channel 3).

Standards and external controls.
To obtain standards for real-time PCR, P. jiroveci-positive BAL samples with GomoriGrocott smear were used, and 10-fold dilutions were made to obtain 10¹10⁴ copies per capillary (Larsen et al., 2002).

Data analysis.
In the absence of a sensitive gold standard, diagnosis of PCP is difficult in certain cases. The criteria that we used for confirming an ongoing pneumocystosis were clinical findings of PCP with characteristic X-ray findings (diffuse interstitial infiltrates), dyspnoea with partial pressure of arterial oxygen <70 mmHg and response to anti-Pneumocystis treatment but resistance to other antibiotics. Based on these criteria for diagnosis, sensitivity and specificity of each laboratory technique were estimated.

Statistical assessment of these characteristics was performed using the S² test (P < 0.05 was considered significant). Statistical assessment of differences of the mean of Pneumocystis DNA concentration [log (copies) per capillary] from PCP or non-PCP cases was performed by unpaired Student's t-test (P < 0.05 was considered significant).

Of 150 patients examined, 11 (7.33 %) developed PCP. In the remaining 139, diagnosis of PCP was not confirmed. Of 11 confirmed cases, seven were HIV seropositive and four were seronegative.

Comparison of different techniques
Concerning sensitivity and specificity of the diagnostic techniques, results obtained for the two classical staining methods (where either one or both were positive) were 60 % [6/10 (the volume was insufficient for one test)] and 100 % (139/139), respectively. For conventional PCR, sensitivity and specificity were 100 % (11/11) and 87 % (121/139) and, for real-time PCR, 100 % (11/11) and 84.9 % (118/139), respectively. There was a significant difference (P < 0.01) in both sensitivity and specificity between PCR and staining techniques; staining gave excellent specificity but lacked sensitivity, while the two PCR tests were much more sensitive but could give false positives.

False-positive cases with the molecular techniques
In this study, false-positive cases using the molecular biology techniques were frequent compared with clinical diagnosis, and were similar to data published elsewhere (see Introduction). There were 18/139 (12.9 %) false positives with conventional PCR; these were also positive with real-time PCR, and there were a further three positives, giving a total of 21/139 (14.5 %) positives with real-time PCR. With the further three positives detected by real-time PCR, quantities of DNA detected were very small (1.3, 1.8 and 3.8 copies per capillary), indicating a slightly higher sensitivity of approximately 1 DNA copy per capillary. The coherence in both PCR techniques confirms the presence of P. jiroveci-specific DNA in asymptomatic patients and their probable carrier state.

False-negative cases with standard staining techniques
Of cases that had different results from the clinical and laboratory diagnoses, false-negative cases were frequent with the Giemsa and GomoriGrocott staining techniques. It is important to note that false negativity using the staining techniques was rare (1/6) in PCP cases with HIV (Table 1), but more frequent (3/4) in PCP non-HIV cases (Table 2). This difference between HIV-positive and -negative cases has also been found in previous studies, which have shown the difficulty of confirming diagnosis in non-HIV-infected cases owing to lack of sensitivity (2567 %), even in BAL specimens (Roux et al., 1994; Weig et al., 1997; Sing et al., 2000). This is worrying because, as in this study, there was a large number of PCP cases among seronegative patients and, as described by Ninin et al. (1998), deterioration is rapid, and death results in 35.4 % of cases.

Table 1. Comparison of different diagnostic methods in seven HIV-positive patients who developed pneumocystosis Dates are given in the form dd/mm/yy. ND, Not done.

Table 2. Comparison of different diagnostic methods in four HIV-negative patients who developed pneumocystosis See Table 1 for details. CLL, Chronic lymphocytic leukaemia.

DNA concentration in relation to clinical findings
In this study, one of the objectives was to try to determine a cut-off value that could discriminate between cases of PCP and asymptomatic Pneumocystis carriers. For this, it is important to have the same type of specimen for all cases. In the present study we had 173 BAL specimens divided into two groups, BAL-50 and BAL-150. No difference in sensitivity or specificity between the two groups regarding diagnosis of PCP was demonstrated, and they were therefore analysed together. However, clinically, BAL-50 was much better tolerated by patients who were severely hypoxaemic.

As shown in Fig. 1, concentrations of DNA detected in PCP patients were very high (P < 0.0001). All cases had >10³ copies of DNA per capillary, whereas all cases that did not have PCP but had detectable Pneumocystis-specific DNA had >10⁴ copies of DNA per capillary (Table 3). Using 10³ copies of DNA per capillary of BAL as a cut-off, determined by real-time PCR, increased specificity from 84.9 (118/139) to 98.6 % (137/139) without reducing the sensitivity of the technique.

(20K): Fig. 1. Quantity of DNA by clinical category. DNA concentrations in PCP cases [n = 11: , HIV-positive (n = 7); , HIV-negative (n = 4)] and non-PCP cases in BAL [n = 21: ", HIV-positive (n = 3); , HIV-negative (n = 18)] are shown. Mean concentrations: log[mean PCP (copies per capillary)] = 4.72 ± 1.01; log[mean non-PCP(copies per capillary)] = 1.50 ± 1.01; P < 0.0001.

Table 3. Distribution of clinical and biological parameters according to Pneumocystis DNA concentration by quantitative real-time PCR in BAL fluid specimens of 150 patients (173 specimens)

In addition, the limit of detection of the staining techniques was well below that of real-time PCR. Thus, all BAL specimens that were positive by the staining techniques had DNA concentrations >10⁴ copies per capillary. This quantitative difference in sensitivity (a factor of 10⁴) has already been noted in different studies. Leigh et al. (1993) defined a factor between 10⁴ and 10⁶ using serial dilution, while Ribes et al. (1997) found a factor of at least 10².

As regards a cut-off to differentiate asymptomatic carriage from PCP, there was an overlap between the two groups (Fig. 1) in spite of a significant difference. That is the reason why a grey zone between 10³ and 10⁴ copies per capillary must be proposed. This also allows for possible minor inter-laboratory differences in performing the techniques. A BAL sample with <10³ copies of DNA per capillary is likely to be a chronic carrier state but, in such cases, follow-up is necessary, as such patients could be in the early stage of an active PCP.

To conclude, real-time PCR is a rapid technique (<3 h including extraction, amplification and visualization). Having a high sensitivity as with conventional PCR, it has the added advantage of quantification to determine a cut-off that permits differentiation between carriage and disease in the majority of cases.

We acknowledge the staff of the Laboratory of Parasitology for their skilful technical assistance, and the clinical departments for sending the specimens.