Introduction

Microorganisms belonging to the Legionella genus can cause a severe life-threatening form of pneumonia known as Legionnaires’ disease (LD). The infection is acquired by inhalation of contaminated aerosols and manifests as sporadic cases or outbreaks described worldwide [1]. Due to fastidious growth requirements, LD was not described until 1976 when L. pneumophila was isolated for the first time after an outbreak in Philadelphia, PA, USA [2, 3]. To date, more than 50 Legionella species have been described; however, L. pneumophila still remains the most common cause of LD. Among 16 different serogroups of L. pneumophila, serogroup 1 (sg1) alone accounts for approximately 85 % of culture-confirmed cases in Europe and is responsible for almost all outbreaks for which the infecting strain has been isolated and characterised [4].

Isolation of legionellae by culture remains the diagnostic gold standard; however, culture-based LD diagnosis requires experience and dedicated media supplemented with cystine, it is time-consuming and often hampered by over-growth of fast-growing microorganisms. The detection of L. pneumophila antigen in urine is the most commonly used diagnostic method but, as current tests are only reliable for sg1 strains, not all infections can be detected. Molecular techniques are reported to be sensitive and specific in detecting both L. pneumophila [5] and sg1 strains [6]. Nevertheless, the detection of Legionella DNA is still not considered by the ECDC as confirmed evidence of LD [7]. Following the formation of the ESCMID Study Group for Legionella Infections (ESGLI) in 2012, the design and validation of a qPCR assay was identified as a priority to standardise molecular diagnosis of L. pneumophila infections across Europe. During the first ESGLI meeting (Dresden, Germany, September 2012), the preliminary results of a triplex qPCR assay detecting L. pneumophila and sg1 DNA, combined with an internal process control (IPC), were presented to delegates coming from Legionella reference laboratories in Europe. In November 2012, a survey was distributed among ESGLI members, and 30 laboratories from 25 countries expressed their interest in such a test being developed. ABI 7500 (Life Technologies), LightCycler 480 Instrument II (Roche) and Rotor-Gene Q (Qiagen) were, at the time, used by 23 out of 30 laboratories. Consequently, a multi-centre international validation of the newly developed qPCR method was initiated for these three platforms.

Materials and methods

qPCR design

A new L. pneumophila assay targeting the macrophage infectivity potentiator gene (mip) was designed for this study. Briefly, the 402-bp sequence of the 59 mip alleles available at the time of the study in the L. pneumophila Sequence-Based Typing (SBT) database [8] was downloaded and aligned using ClustalW [9]. Oligonucleotides were designed in conserved regions, allowing a maximum of one mismatch in either the primer or probe binding sites using Primer3 [10]. Sequences of subsequently described mip alleles were added to the ClustalW alignment; thus, a total of 74 mip alleles obtained from 9423 L. pneumophila strains were included in this study.

The new mip assay was combined with available sg1 and IPC assays, targeting wzm [6] and gfp [11], respectively. The sequence of primers and probes used in this study is listed in Table 1.

Table 1 Summary of oligonucleotides used in this study (FP forward primer, RP reverse primer, HP hybridisation probe)
Full size table

In silico analysis

mip oligonucleotides were tested for L. pneumophila specificity against the NCBI database using BLASTn [12]. The presence of mismatches possibly affecting the performance of the qPCR assay was analysed in silico comparing the oligonucleotide sequences to the ClustalW alignment of all 74 mip alleles.

The presence of wzm was determined in 395 finished or draft L. pneumophila genomes (including 372 sg1 and 23 non-sg1 strains) using BLASTn in order to test for sg1 specificity. In the case of the draft genomes, de novo assembles were used which had been constructed using Velvet [13]. The ClustalW alignment of 17 wzm alleles characterised from these genomes was analysed to identify mismatches in the oligonucleotide binding sites.

qPCR assays

The final concentration of primers, probes and additional MgCl2, and the amount of gfp DNA are listed in Table 2; Taq polymerase was activated at 95 °C for 2 min (ABI 7500) or 3 min (LightCycler 480 Instrument II and Rotor-Gene Q), followed by 45 cycles of denaturation at 95 °C for 15 s and annealing/extension at 60 °C for 30 s (LightCycler 480 Instrument II and Rotor-Gene Q) or 1 min (ABI 7500).

Table 2 Final concentrations of forward primers (FP), reverse primers (RP), hybridisation probes (HP) and additional MgCl2, and amount of gfp DNA used in this study
Full size table

PCR reactions were performed with ABI 7500 in a final volume of 50 μL containing 25 μL of Platinum® Quantitative PCR SuperMix-UDG with ROX (Life Technologies) and 5 μL of template using 0.2-μL optical tubes (Life Technologies). The results were analysed with the Auto Threshold function in the green, yellow and red channels using v2.0.6 of the ABI 7500 software.

PCR reactions were performed with LightCycler 480 Instrument II in a final volume of 20 μL containing 10 μL of 2× iQ Multiplex Powermix (Bio-Rad) and 5 μL of template using 0.2-mL thin-wall 8-tube strips (Biozym TC). The results were analysed in the green, yellow and red channels using v1.5.039 of the LightCycler 480 Instrument II software and interpreted by the second derivative maximum method with crossing points automatically calculated. Colour compensation was activated for the green and yellow channels because of an observed crosstalk.

PCR reactions were performed with Rotor-Gene Q in a final volume of 20 μL containing 10 μL of Rotor-Gene Multiplex PCR Kit (Qiagen) and 5 μL of template using 0.1-mL thin-walled tubes (Qiagen). The results were analysed in the green, yellow and red channels using v2.1.0.9 of the Rotor-Gene Q software. The instrument was set to ignore fluorescent signals in the first 15 cycles and to use 0.025 as the threshold value.

L. pneumophila sg1 DNA standards

Two commercially available L. pneumophila DNA quantification standards (i.e. Minerva Biolabs, Cat.-No. 52–0101 and LGC Standards, Cat.-No. SRM_LEGDNA_01) prepared from DNA of the Philadelphia-1 strain (NCTC 11192T) were used in this study.

Efficiency

For each qPCR platform/kit combination, five experiments were performed on separate days. In each run, four dilutions (i.e. 50,000, 5000, 500 and 50 gu/reaction) of L. pneumophila DNA standard were tested in triplicate together with a no template control. Efficiency values were obtained from the slope of the standard curve using the qPCR machine software. For each qPCR platform/kit combination, the mean and standard deviation of the efficiency values obtained from the five runs were calculated. Mean values between 90 % (slope = −3.1) and 110 % (slope = −3.6) were considered acceptable [14].

Limit of detection

For each qPCR platform/kit combination, five experiments were performed on separate days. In each run, five dilutions of the L. pneumophila DNA standard (namely, 50, 40, 30, 20 and 10 genome units) plus a no template control were tested. The limit of detection (LoD) was determined statistically by Probit analysis [14].

Validation panel

A panel of microorganisms (see the Supplementary Material) consisting of 36 L. pneumophila reference strains (including all 16 described serogroups), 46 L. pneumophila isolates representing the 46 mip alleles available in the authors’ collection, 52 non-pneumophila Legionella species (total of 60 strains) and 96 non-Legionella species (total of 109 strains) was analysed on the Rotor-Gene Q to test assay specificity. Legionellae were cultured on buffered charcoal yeast extract (BCYE) agar (Oxoid) at 37 °C for 48–72 h in a moist atmosphere. Non-Legionella strains were grown on Nutrient Agar (Oxoid) at 37 °C for 24 h or specific media/conditions according to growth requirements. Genomic DNA was extracted using the Wizard Genomic DNA Purification Kit (Promega) and quantified with a Qubit® Fluorometer (Life Technologies) and/or a NanoDrop ND-100 spectrophotometer (Thermo Scientific). L. pneumophila DNA extracts were tested at a concentration of ca. 10 pg/reaction, while extracts from other species were tested at a concentration of ca. 1 ng/reaction.

Legionella PCR EQA samples

Samples from seven ELDSNet/ESGLI and three QCMD EQA distributions were analysed on the Rotor-Gene Q in parallel to a previously validated mip assay [5]. Each distribution included ten samples, and a total of 100 EQA specimens were tested.

Clinical samples

Respiratory samples from 132 consecutive patients with evidence of legionellosis submitted over a 12-month period (16/08/2013 to 14/08/2014) were tested at Public Health England Legionella Reference Laboratory (UK) using the Rotor-Gene Q Legionella culture and DNA extraction was performed as previously described [5].

Results

mip assay

The new L. pneumophila assay was designed in a conserved area of the mip allele region with only the reverse primer binding site containing a non-conserved position (A/G), which was taken into account by including a ‘Y’ base (i.e. C/T). Among the 74 alleles, 67 contain no other variation in the oligonucleotide binding sites (Table 3), for a total of 9319 total strains (98.90 % entries) in the SBT database. The seven remaining alleles each contained one mismatch, for a total of five distinct mismatches (Table 2). Nevertheless, mip was successfully amplified from representative strains for 4/5 mismatches on all three qPCR platforms. Testing of allele 61 was not possible, as the carrying strain is not internationally available (Table 3).

Table 3 Summary of single nucleotide polymorphisms identified in mip and wzm oligo binding sites. mip analysis was performed on the 74 alleles obtained from 9423 strains deposited in the Legionella pneumophila SBT database, while wzm analysis was performed on 372 L. pneumophila serogroup 1 (sg1) sequenced genomes available to the authors
Full size table

BLASTn analysis showed the newly designed mip oligonucleotides as specific (100 % coverage and homology) for mip fragments of three putative non-pneumophila Legionella strains, namely, L. micdadei, L. fairfieldensis and L. worsleiensis. These mip sequences (AJ496274, U60163 and U60164) were downloaded from the NCBI database and analysed using the ‘Legionella species identification’ online tool [15], which uses the mip sequence to differentiate between members of the Legionella genus [16]. All three sequences were identified as belonging to L. pneumophila strains, with homologies between 99.83 and 100 %.

wzm assay

The presence and sequence of wzm were analysed in silico on 395 L. pneumophila available genomes (data not shown). All sg1 genomes (372) included wzm, while this gene was not found in non-sg1 strains. Sequence analysis of wzm revealed 17 variants, of which 14 (for a total of 347 sg1 strains) did not show variation in the oligonucleotide binding sites. As per mip validation, DNA from a representative strain for each of the four identified mismatches was extracted and successfully tested on each qPCR platform (Table 3).

Efficiency

Determination of qPCR efficiency is crucial to ensure accurate target quantification. Efficiency should be as close as possible to 100 %, while mean values between 90 and 110 % were considered acceptable [14]. An efficiency value equal to 100 % means that the target sequence is doubling at each cycle during the logarithmic phase of the reaction. A theoretically optimal PCR efficiency gives a 3.3-cycles difference between 10-fold dilutions of template. The triplex qPCR assay generated efficiency values within the required limits on all platforms for both mip and wzm (Table 4), ensuring reliable target DNA quantification within the standard curve (i.e. between 50,000 and 50 units of L. pneumophila sg1 DNA). Efficiency was calculated with Philadelphia-1 strain (NCTC 11192T) that does not contain mismatches in the mip and wzm primer binding sites. Less reliable target quantification may occur if mismatches are present.

Table 4 Average crossing points (Cp) and efficiency obtained in this study for the three qPCR platforms
Full size table

Limit of detection

Conventionally, the LoD is reported as the “estimate of the detection limit that can be achieved with 95 % confidence” [14] or as “the lowest actual concentration of analyte” (L. pneumophila DNA in this instance) “in a specimen that can be consistently detected” (e.g. in 95, 90 or 50 % of specimens tested) “under routine laboratory conditions in a defined type of specimens” [17]. The LoD results are shown in Table 5. Lower LoD 90 % and LoD 95 % values were obtained using LightCycler 480 Instrument II, but this could be a consequence of the fact that the LGC quantification standard was used, while the Minerva standard was used for Rotor-Gene Q and ABI 7500. In line with Saunders et al. [14], the values estimated with 95 % confidence (LoD 95 %) were adopted as the LoD for mip and wzm assays on all three qPCR platforms (Table 5).

Table 5 Number of L. pneumophila genome copies to produce a positive result in 50, 90 and 95 % of replicate samples
Full size table

Validation panel and EQA samples

The triplex assay was 100 % specific for L. pneumophila and sg1 DNA on the large panel of type strains (see the Supplementary Material) tested on Rotor-Gene Q. mip was successfully amplified form all EQA samples containing L. pneumophila DNA. One sample containing PBS thus intended to be negative gave a positive mip result. The same happened when the sample was tested with a previously published mip assay [5] targetting a different part of the mip gene, hence a contaminated DNA extract rather than a specificity issue was considered as the likely cause of the unexpected result. On the same set of samples, wzm was successfully amplified whenever DNA from sg1 strains was present.

Clinical samples

Respiratory samples from 132 consecutive patients with evidence of legionellosis were tested for the presence of L. pneumophila and sg1 DNA using Rotor-Gene Q. 108/128 (84 %) L. pneumophila urinary antigen-positive patients were L. pneumophila and sg1 qPCR-positive. L. pneumophila sg1 was subsequently grown from 80/108 and a further 17 of the 28 PCR-positive/culture-negative samples yielded DNA sequence typing data directly from the sample by nested SBT [18]. L. pneumophila sg1 was isolated from 1/20 PCR-negative samples; however, a second sample from the same patient was both qPCR- and culture-positive. 3/132 samples were submitted as being positive using local Legionella spp. PCRs and 1/132 was obtained from an L. pneumophila sg6 culture-proven case. All were L. pneumophila-positive and sg1-negative in this triplex qPCR. L. pneumophila was subsequently isolated from 2/3 samples (sg3 and sg5, respectively) and nested SBT detected ST1775 in the third sample; previously, this ST has only been reported from an L. pneumophila sg5 strain.

Inhibition

Under the applied testing conditions, no assay failure due to inhibition issues was observed with any specimen or culture under test. Inhibition is primarily caused by inhibitors present in the DNA template. In this study, methods yielding high-quality DNA extracts were used. Inhibition issues might occur if extraction methods yielding lower quality DNA were used.

Discussion

We describe the design and validation of a triplex qPCR for the simultaneous detection of L. pneumophila and L. pneumophila sg1 DNA on three real-time PCR platforms to aid international standardisation of molecular detection of L. pneumophila amongst ESGLI laboratories.

A rational approach using the mip allele sequences available on the L. pneumophila SBT database was applied when designing a new L. pneumophila assay; a wide range of both in silico and in vitro experiments was then performed to fully validate the newly designed mip assay (targeting L. pneumophila) and to expand the validation of the wzm assay (targeting sg1) previously published [6]. The gfp assay used as an IPC was adopted, as previously published [11].

The assay proved 100 % specific for L. pneumophila and sg1 DNA on a large panel of type strains, clinical samples and EQA samples tested on Rotor-Gene Q. The efficiency and LoD were calculated for all three qPCR platforms. The LoD was excellent for both L. pneumophila and sg1 targets, indicating that this methodology can detect low quantities of target DNA reproducibly. Inhibition issues were not observed. The international validation documented in this study demonstrates that this methodology is applicable for use in laboratories around the globe and can be performed on different qPCR platforms without loss of functionality.

Although the utility of molecular techniques in diagnosing infectious diseases is universally recognised, detection of nucleic acid is still among the laboratory criteria of a probable but not proven LD case according to the ECDC. This is currently a limitation in LD case definition. The validation of this qPCR method by the ESGLI is intended to promote the use of qPCR to detect L. pneumophila in LD cases and consideration for the inclusion of L. pneumophila DNA detection within laboratory criteria for proven LD cases should be given.