Proposal of a quantitative PCR-based protocol for an optimal Pseudomonas aeruginosa detection in patients with cystic fibrosis

Background The lung of patients with cystic fibrosis (CF) is particularly sensitive to Pseudomonas aeruginosa. This bacterium plays an important role in the poor outcome of CF patients. During the disease progress, first acquisition of P. aeruginosa is the key-step in the management of CF patients. Quantitative PCR (qPCR) offers an opportunity to detect earlier the first acquisition of P. aeruginosa by CF patients. Given the lack of a validated protocol, our goal was to find an optimal molecular protocol for detection of P. aeruginosa in CF patients. Methods We compared two formerly described qPCR formats in early detection of P. aeruginosa in CF sputum samples: a qPCR targeting oprL gene, and a multiplex PCR targeting gyrB and ecfX genes. Results Tested in vitro on a large panel of P. aeruginosa isolates and others gram-negative bacilli, oprL qPCR exhibited a better sensitivity (threshold of 10 CFU/mL versus 730 CFU/mL), whereas the gyrB/ecfX qPCR exhibited a better specificity (90% versus 73%). These results were validated ex vivo on 46 CF sputum samples positive for P. aeruginosa in culture. Ex vivo assays revealed that qPCR detected 100 times more bacterial cells than culture-based method did. Conclusion Based on these results, we proposed a reference molecular protocol combining the two qPCRs, which offers a sensitivity of 100% with a threshold of 10 CFU/mL and a specificity of 100%. This combined qPCR-based protocol can be adapted and used for other future prospective studies.


Background
Pseudomonas aeruginosa is the major pathogen involved in the decline of lung function in patients with cystic fibrosis (CF) [1][2][3][4][5]. Its presence in the lungs is associated with an increased mortality and morbidity of CF patients [6]. Early detection of this bacterium from respiratory tract is determinant because it ensures effective patient management [5,7,8]. Indeed, after intermittent colonization by different strains, once acquired, chronic P. aeruginosa colonization by mucoid and biofilm-growing isolates is difficult to eradicate [2,4,9,10]. Thus, the earlier the treatment toward P. aeruginosa onset, the higher the chance to efficiently control P. aeruginosa [5,7,8].
However, accurate identification of this bacterium in CF sputum by conventional microbiology techniques is known to be limited. This can be explained by a large phenotypic diversity of P. aeruginosa isolates recovered from CF patients such as loss of pigment production or exopolysaccharide production. Moreover, Singh et al. demonstrated that P. aeruginosa can form biofilms in the airways of CF patients [11]. Biofilms contain bacterial cells that are in a wide range of physiological states. One of the mechanisms contributing to this physiological heterogeneity includes the adaptation to the local environmental conditions.
For instance, bacterial cells from the deep layers of biofilm depleted of oxygen [12] can grow in anaerobic conditions. Therefore, the CF patients isolates obtained from biofilms, i.e. in anaerobic conditions, grow hardly in aerobic conditions on a conventional culture medium [13]. Another limitation of conventional culture is that P. aeruginosa can be easily misidentified with closely related Gram-negative bacilli in CF sputum [14][15][16][17][18][19].
The use of molecular techniques such as PCR could improve accurate identification of P. aeruginosa [14][15][16][17][18][19], and consequently, its early detection in CF sputum patients [20][21][22][23][24]. To date, there is no consensus for a universal protocol for the molecular detection of P. aeruginosa. Indeed, its genome is known to be highly polymorphic. Changes that can occur at the genetic level could compromise the reliability of molecular identification techniques. In particular in CF patients lungs, most of the recovered isolates are hypermutable [2,4,25,26], and show a high genetic plasticity by acquisition or loss of genes [27,28]. Since mutations or gene deletions occur on PCR target sequences, they could decrease the sensitivity of the method [29]. Moreover, horizontal genetic transfer with other bacterial species present in the CF lung niche can impact upon the specificity of the PCR [14].
In a prospective multicenter study, we aimed to assess the role of PCR for the early detection of P. aeruginosa in CF patients; we evaluated two qPCRs in detection of P. aeruginosa: a simplex qPCR targeting oprL gene [30], and a multiplex qPCR, targeting gyrB and ecfX genes [14]. The sensitivity and the specificity of both qPCRs were initially evaluated testing a large panel of P. aeruginosa isolates and closely related non-P. aeruginosa gram-negative bacilli isolates from CF patients. Then, the two different qPCRs ability in detection of P. aeruginosa were tested ex vivo, i.e in CF sputum samples. Finally, we were able to propose a promising reference protocol combining these two qPCRs for an optimal detection of P. aeruginosa in clinical setting.

Bacterial collection
Thirty-six P. aeruginosa isolates, including mucoid and non mucoid forms, were obtained from 31 sputum samples of CF patients and from 5 samples of non CF patients (blood, n = 1; stool, n = 1; urine, n = 1; sputum, n = 1; peritoneal fluid, n = 1), attending three French University Hospitals, the CHRU of Brest (n = 3), the CHU of Nantes (n = 26), and the GHSR of Saint Pierre, La Réunion (n = 2). The reference strain P. aeruginosa CIP 76.110 was also included in the study.
Identification of bacterial isolates was previously conducted based on phenotypical and morphological criteria (colony morphology, pigmentation, lactose fermentation, oxidase activity checked with 1% tetramethyl p-phenylenediamine dihydrochloride, sensitivity to antibiotics). Atypical P. aeruginosa isolates, for which difficulties of identification were encountered, were further analyzed with biochemical tests [API 20NE system (bioMérieux, Marcy l'Etoile, France), ID 32GN (bioMérieux)], or with the gram-negative bacillus identification card on VITEK 2 Compact (bioMéreux). All non-P. aeruginosa gramnegative bacillus isolates were identified by 16S rRNA gene sequencing as previously described [31].
All bacteria were stored at -80°C. Bacteria from frozen stocks were grown aerobically at 37°C for 24 to 48 hours on Muller-Hinton medium (bioMérieux).

P. aeruginosa detection and quantification by sputum samples culture CF patients and sample processing
Fourty-six sputa were selected in line with our study objective. These CF sputum samples have been collected from 34 patients (median age: 11 years, range: 4-29, 53% female) attending the CF center of Roscoff (France), between March 2008 and May 2012. At the time of CF patients inclusion, all of the patients were P. aeruginosa free for at least one year. More precisely, according to the Leeds definition [32], ten of them were never and 22 were free (Table 1). Each sputum sample was mixed with equal volume of dithiothreitol (Digesteur® Eurobio, Courtaboeuf, France) and incubated at room temperature for 30 min. For isolation of P. aeruginosa, liquefied sputa were immediately processed. For molecular detection of P. aeruginosa, two one-milliliter aliquots of every liquefied sputum were stored at -80°C.

P. aeruginosa isolation
Ten μl of liquefied sputum pure and diluted into 1/1000, were inoculated and incubated onto several non selective and selective media for P. aeruginosa isolation, including Columbia blood agar supplemented with 5% defribinated horse blood (Oxoid, Dardilly, France), Columbia chocolate agar (Oxoid), and cetrimide agar (Oxoid). All media were incubated aerobically at 37°C for five days and monitored daily. All different morphotypes of bacterial colonies were identified phenotypically with  [33,34]. Quantification was conducted based on the colony forming unit (CFU) counts and the dilution ratio of the plate.

P. aeruginosa detection and quantification by quantitative PCR (qPCR) DNA extraction
For each isolate of the bacterial collection, 1 ml of a 0.5 McFarland suspension was extracted. For each sputum sample, one of the two 1 ml-aliquots was treated by 5 min of sonication using a bath sonicator (Elamsonic S10, Singen, Germany). After a 10 min-centrifugation (5000 g), the pellet was suspended in 200 μl of DNA free water. Ten μl of the IC2, an internal control provided in the DICO Extra r-gene™ kit (Argène, Verniolle, France), were added in each sample and, for each batch of extraction, in 200 μl of DNA free water as a negative control. DNA was extracted using the QIAamp DNA Minikit® (Qiagen, Courtaboeuf, France) according to the instructions of the manufacturer ("Tissue protocol") with elution volumes of 100 μl.

Determination of the lower detection threshold
To determine the lower detection threshold, six dilution ranges were realized with six different P. aeruginosa isolates. One range was prepared with the reference strain (CIP 76.110), two with a mucoid and a non-mucoid isolates from a sputum sample of a CF patient, and three with three isolates from three non-CF patients (urine, n = 1; blood, n = 1; stool, n = 1). Ten fold iterative dilutions from 0.5 McFarland calibrated P. aeruginosa suspensions provided a full concentration range extending from 10 0 to 10 8 CFU/mL. The nine dilutions of the range were tested 30 times. To determine the exact inoculum of each dilution range, a plate counting was carried out on a Mueller-Hinton medium (bioMérieux) incubated from 24 to 48 hours at 30°C. A mean of the results was calculated taking into account the sum of all assays.

Ethics
The Comité de Protection des Personnes Ouest VI approved the protocol. All of the patients and their relatives gave written informed consent. The collection of archival specimens was registered with the French Ministry of Research and the Agence Régionale de l'Hospitalisation, No. DC-2008-214.

Results
In vitro characteristics of the oprL and gyrB/ecfX qPCR Sensitivity The two qPCRs showed 100% sensitivity. At the concentration of 10 6 CFU/mL, all the 37 P. aeruginosa isolates were detected by the two qPCRs. The cycle treshold (Cq) mean was 24.8 and 24/28.2 respectively for the oprL qPCR and the gyrB/ecfX qPCR.

Specificity
The specificity of the oprL qPCR was evaluated at 73%. At the concentration of 10 6 CFU/mL, eleven isolates out of the 41 non-P. aeruginosa gram-negative bacillus isolates, corresponding to six different species, were amplified by the oprL qPCR. The six species responsible for cross-reactions were A. xylosoxidans, B. cenocepacia, B. multivorans, E. meningoseptica, Roseomonas spp., and S. maltophilia (Table 3). By considering the gyrB/ecfX qPCR positive when at least one of the two targeted genes was amplified, the specificity was calculated at 90%. Four out of the 41 isolates corresponding to four different species induced false positive reactions in at least one of their assays (Table 3): C. indologenes, F. oryzihabitans, P. putida and P. stutzeri. No species cross-reacted with both qPCRs. In this manner, combining oprL and gyrB/ecfX amplifications allowed achieving 100% specificity.

Lower detection threshold
The lower detection threshold of the oprL qPCR was evaluated at 10 CFU/mL. Given a positive multiplex PCR when at least one of the two probes was detected, the detection threshold of the gyrB/ecfX qPCR was evaluated at 730 CFU/mL.

Consistency between in vitro and ex vivo experiments
The theoretical threshold calculated from in vitro experiments was totally consistent with the observed threshold from ex vivo experiments. Indeed, oprL qPCR assays performed ex vivo identified one hundred times more bacterial cells than culture-based methods did. Thus, the theoretical lower detection threshold of oprL qPCR of 10 CFU/mL calculated from in vitro cultures is equivalent to a threshold of 1E + 03 CFU/mL if applied ex vivo. This was verified on 9 culture-/PCR + samples for which the quantification by oprL qPCR retrieved a mean of quantification of 997.3 CFU/mL. The theoretical lower detection of the multiplex qPCR was found at 7.3E + 02 CFU/mL in vitro. Ex vivo, the amplification conducted on the sputum samples showed a positive signal for at least one target (gyrB or ecfX) for all of the P. aeruginosa-positive sputa with quantification above 7.3E + 02 CFU/mL (n = 38). On the contrary, below 7.3E + 02 CFU/mL, the majority (5 of 8 samples) of the sputa that were P. aeruginosa-positive by oprL PCR, were P. aeruginosa-negative by multiplex PCR.
To conclude, the theoretical thresholds of both qPCRs were verified on the sputum samples.

Discussion and conclusion
Several studies have suggested that qPCR is superior to culture for detecting early colonization of P. aeruginosa in CF sputum [20,[22][23][24]. Today, the main goal is to have an optimal protocol as the gold standard for the molecular detection of P. aeruginosa. Therefore, we performed in vitro and ex vivo evaluation of two qPCRs, one targeting the oprL gene and the other targeting simultaneously gyrB and ecfX genes [14,30]. Numerous DNA targets have been described for the amplification of P. aeruginosa [15,17,19,[34][35][36], of these three housekeeping genes, oprL, gyrB and ecfX have been reported to be reliable targets in the detection of P. aeruginosa [14,19,30,35].
The first criterion for an optimal technique in early detection of P. aeruginosa in CF sputum is related to the choice of the PCR format and its optimization. Today, the DNA molecules counting of a particular sequence in a complex sample can be achieved with exceptional accuracy and sensitivity sufficient to detect a single molecule [36]. As underlined by Deschagt et al. [35], the choice of PCR format may influence the performance of the molecular detection. We chose a probe-based assay, which is known to be more sensitive and specific than the SYBR Green-based qPCR [35].
The second criterion is a good sensitivity to prevent false negative results. Despite wide genetic variability of P. aeruginosa isolates recovered from CF patients [2,4,[25][26][27][28], results of previous studies aiming at detecting P. aeruginosa by PCR have been encouraging. In our study, both evaluated qPCRs showed an excellent sensitivity covering all the tested panel of P. aeruginosa isolates. Focusing on the lower detection threshold, the difference was significant between the two qPCR assays with a detection threshold of 10 CFU/mL for the oprL qPCR versus 730 CFU/mL for the multiplex PCR. The sensitivity of the in vitro oprL qPCR in our study was higher than that recommended by the French guidelines, i.e. a detection threshold of 10 2 CFU/mL for CF sputum sample [37].
The third criterion needed for early P. aeruginosa detection technique, in particular, for molecular one, is to have a high specificity to prevent false positive amplification. When looking at a large panel of genes described in the literature e.g. oprI, oprL, rrl, ecfX, gyrB, or rrs, specificity varied from 74% to 100% [14,17,[34][35][36]38]. In our study, specificity of the oprL qPCR was evaluated at 73% versus 90% for the multiplex PCR. Four previous studies have tested the specificity of the oprL primer pairs and found different values ranging from 87% to 100% [22,34,35,38]. Again, previous studies looking at gyrB and ecfX genes found a better specificity (100%) than in our study [14,35]. Different reasons could explain these discrepancies. Firstly, our specificity could have been influenced by a larger panel of closely related non P. aeruginosa gram-negative bacilli (41 isolates including 16 different species). Secondly, all the bacterial isolates (except one reference strain) were recovered from clinical samples (CF or non CF) or from environmental samples. These isolates, which were recovered from CF could have undergone genetic exchange with other species in the natural CF microenvironment, especially P. aeruginosa, influencing the specificity of the molecular method [38]. Thus, specificity in previous studies could have been overestimated [14,34,35,38]. As highlighted by Anuj et al. [14,35], the higher specificity of our results for the multiplex PCR may be explained by the fact that we amplified at least 2 DNA targets. The use of two probes simultaneously seems to improve the specificity, providing at the same time the detection and the confirmation of the presence of P. aeruginosa [14,19]. Interestingly, our bacterial species that cross-reacted with the oprL qPCR did not do so when oprL qPCR was combined with the multiplex PCR thus allowing 100% specificity.
These results were successfully validated by the sputum samples of CF patients from the never or free categories according to the definition of Leeds [32]. The ex vivo experiments put forward a significant difference between the culture-based quantification and the qPCRbased quantification. In average, the qPCR detected 100 times more CFU of P. aeruginosa than the culture did. This could be explained by different hypotheses. First, the difference in utilized sputum volumes contributes to this discrepancy. Indeed, only 10 μl were cultured whereas 1 ml was extracted for the qPCR. The lowest concentration that theoretically can be detected by qPCR equals the presence of one genome (i.e. equivalent to one CFU) per qPCR reaction mixture. Using 1 ml of 10-fold concentrated sputum by centrifugation and extraction (elution volume of 100 μl) and 4.5 μl for the PCR reaction (final volume of 25 μl), the detection limit of our molecular diagnosis is ≈22 CFU/mL. In comparison, the lowest concentration that theoretically can be detected by culture is 100 CFU/mL. Second, given the phenotypic diversity of P. aeruginosa isolates and the large diversity of species found in pulmonary microbiota, the detection of P. aeruginosa by culture in CF sputum is a hard task [14][15][16][17][18][19]. Moreover, culture in aerobic conditions can fail in the detection of some isolates adapted to anaerobic conditions of the CF lung niche [13], or of non-cultivable isolates present in the bacterial biofilm [39].
Another explanation could be that qPCR detects P. aeruginosa DNA, i.e. not only live bacteria but also dead cells [40]. As CF patients are chronically treated with antibiotics, one can suppose that dead bacteria are significantly present in the pulmonary tract. In a study lead by Deschaght et al. in 2009, no difference in sensitivity between culture and oprL qPCR was found [41]. Their study was conducted on eight artificial P. aeruginosa-positive sputum pre-liquefied samples thus skipping the sample homogenization step, one of the cornerstones in amplification-based technique. Our ex vivo application of the two qPCR assays with real samples took into account the sample homogenization. It also put forward the importance of having a controlled amplification assay in particular to avoid false negatives due to inhibitors or a bad extraction. Indeed, the DNA-extraction method has been shown to be a critical step in the PCR performances [41]. In our study, we chose the DICO Extra r-gene kit, a totally Figure 1 Proposal of a molecular protocol integrating two qPCR formats (targeting oprL and gyrB/ecfX genes) for an early detection of Pseudomonas aeruginosa in sputum samples of patients with cystic fibrosis. The oprL qPCR is applied in screening because of its good sensitivity. In case of a doubtful or a positive result, the gyrB/ecfX qPCR is applied in a second time. Interpretation of the gyrB/ecfX qPCR takes into account the quantification found with oprL qPCR. Below the detection threshold of 730 CFU/mL, the oprL qPCR prevails over the gyrB/ecfX qPCR. Conversely, beyond this threshold, the gyrB/ecfX qPCR prevails over the oprL qPCR.
artificial DNA, as internal control, which prevents from contamination during procedure handling, and allows to test extraction and amplification at the same time.
Altogether, our study showed that the oprL qPCR offers a good sensitivity whereas the multiplex PCR offers a good specificity. Based on these data, we decided to combine these two qPCR assays and proposed a molecular protocol for an optimal detection of P. aeruginosa by qPCR in CF sputum as follows (Figure 1). The oprL qPCR can be applied in screening because of its good sensitivity. In case of a doubtful or a positive result, we can proceed to the multiplex PCR. Interpretation of the multiplex PCR takes into account the quantification found with oprL PCR. Below the detection threshold of 730 CFU/mL, the oprL qPCR prevails over the multiplex PCR. Conversely, beyond this threshold, the multiplex PCR prevails over the oprL qPCR. Overall, this combined molecular protocol offers a sensitivity of 100% with a threshold of 10 CFU/mL and a specificity of 100%.
This qPCR-based combined protocol can be adapted for instance in a subgroup of non-sputum producing patients and used for other future prospective studies. Indeed, the initial colonization of P. aeruginosa often occurs in CF patients who do not produce sputum (e.g. mainly children). This qPCR format should therefore be tested on the sample secretions routinely obtained from, e.g. deep throat swabs or endolaryngeal suction. This study was supported by a grant from the French Cystic Fibrosis Association "Vaincre la Mucoviscidose" (contract No. RCO 1773). This study was presented in part at the 4 th Congress of European Microbiologists FEMS, 26-30 June 2011, Geneva, Switzerland. The authors thank Jocelyne Caillon, and Alain Michault for providing some of the isolates used in this study. We are indebted to Zarrin Alavi for critical reading of the manuscript.