Validation and in-field testing of a new on-site qPCR system for quantification of Legionella pneumophila according to ISO/TS 12869:2012 in HVAC cooling towers

Legionella pneumophila, found in engineered water systems such as HVAC cooling towers, poses a significant public health risk. Culture, though routinely used to quantify L. pneumophila, has several disadvantages including long turnaround time, low sensitivity, and inter-laboratory variability. In this study, we validated the performance of an on-site quantitative polymerase chain reaction (qPCR) detection system for L. pneumophila in accordance with International Standards Organization Technical Specification 12869:2012. We evaluated specificity, limit of detection and quantification, and calibration curve linearity. Additionally, we evaluated whole system recovery and robustness using samples taken from taps and evaporative cooling towers. We then compared the system’s performance against laboratory culture and laboratory qPCR across 53 cooling towers in a 12-week in-field study. We found that concordance between on-site qPCR and culture was both laboratoryand site/sample-dependent. Comparison of laboratory qPCR with on-site qPCR revealed that laboratory results were highly variable and showed little concordance. Some discordance may be explained by time delay between sample collection and testing (‘shipping effect’) which may lead to inaccurate reporting. Overall, our study highlights the value of on-site qPCR detection of L. pneumophila, demonstrates that laboratories are prone to misreporting results due to shipping effects, and reveals significant discordance between laboratory qPCR and culture. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/wh.2019.252 ://iwaponline.com/jwh/article-pdf/17/2/237/846039/jwh0170237.pdf Shaimaa Ahmed Urszula Liwak-Muir Danielle Walker Agnes Zoldowski Alan Mears Sergey Golovan Steve Mohr Paul Lem (corresponding author) Chris Harder Spartan Bioscience Inc., 2934 Baseline Road, Suite 500, Ottawa, ON K2H 1B2, Canada E-mail: paul.lem@spartanbio.com This article has been made Open Access thanks to the generous support of a global network of libraries as part of the Knowledge Unlatched Select initiative.


INTRODUCTION
Legionella is a common water-based pathogen in man-made engineered water systems in developed countries (Vinson L. pneumophila strains are responsible for approximately 95% of all cases of LD (Walser et al. ; Kirschner ) and the sources of contamination have frequently been identified as HVAC (heating, ventilation, and air conditioning) evaporative cooling towers and domestic hot water systems (Walser et al. ; van Heijnsbergen et al. ).
Legionella is difficult to control due to its ability to replicate in protozoan hosts and its tendency to exist in biofilms, both of which contribute to its resistance to disinfectants (Kim et  There are numerous methods available for Legionella detection, but the most widely used methods are culture and quantitative polymerase chain reaction (qPCR) (Lucas & Fields ; Whiley & Taylor ). The advantage of culture is it detects viable and culturable bacteria. In contrast, qPCR is significantly more sensitive than culture, but it is thought that this is due to detection of dead cells, extracellular DNA, and ). The study concluded that culture plating significantly underestimated Legionella counts, was highly variable between laboratories, and had a significant false negative rate (Lucas et al.  Each test cartridge includes qPCR primers and a probe that are designed against a highly conserved region of the L. pneumophila macrophage infectivity potentiator (mip) gene (Benitez & Winchell ). The test cartridge also contains an internal positive control to detect the presence of qPCR inhibitors in the sample, and to identify reagent degradation and contamination. Negative controls are performed during manufacturing of the sealed test cartridge, which is opened just prior to use.

Verification of inclusivity and exclusivity
The on-site qPCR system was verified for analytical speci- In addition to testing against 40 bacterial strains, the specificities of the L. pneumophila primers and probe were assessed in silico for 15 serogroups of L. pneumophila.

Verification of linearity of the qPCR calibration curve
The calibration curve of the on-site qPCR system was For the overall result to be valid, the PCR efficiency was required to be between 75 and 125%, corresponding to a slope of regression between -4.115 and -2.839.

Verification of lower limit of detection and limit of quantification
The lower limit of quantification (LOQ) and limit of detection (LOD) for the on-site qPCR system were verified according to Section 10.4 'Verification of the PCR limit of quantification' and Section 10.5 'Verification of the PCR limit of detection' of ISO/TS 12869:2012, respectively.
A dilution from a secondary standard of L. pneumophila was made to 10 6 GU/μL, and then dilutions were made down to the LOD of 2 GU/reaction. LOD is defined as the concentration at which at least 90% of the results are positively detected. The dilution step was repeated by multiple operators. The LOQ was tested by multiple operators on multiple days at 20 GU/reaction.
Verification of the entire on-site qPCR Legionella detection system The whole system (concentration and qPCR) was verified by assessing recovery and robustness using real-world water matrices from cooling towers. This verification addresses the objectives in Section 10.6 'Recovery method' and Section 10.7 'Robustness' of ISO/TS 12869:2012, respectively.
Recovery was calculated as the percentage of qPCR fluorescence signal post concentration compared to the signal generated by directly amplifying the water sample without concentration (direct qPCR).
To verify that recovery was not affected by matrix, we tested distilled water, tap water, and cooling tower water that was known to be free of L. pneumophila DNA. These water samples were artificially contaminated with dilutions of a stock suspension of L. pneumophila (ATCC 33152).
Three input concentrations were tested corresponding to 20, 100, and 250 GU/mL. Each concentration was made using different replicate serial dilutions from the same stock suspension. For each concentration, at least three separate 22-mL spiked samples were run by several operators.

L. pneumophila in HVAC samples
Samples to be externally evaluated were collected over a 12week period from 51 HVAC cooling towers in the Canadian cities of Ottawa, Toronto, and Montreal. These samples were collected weekly from their designated system location on their scheduled day ( Figure 1). Two out of the 51 towers were shut down due to operational issues and alternative towers were brought on-line in the same facility. As a result, a total of 53 towers were tested. Individual test results from these new towers were included in the weekly testing analyses (by culture and on-site qPCR). However, for the week-over-week analyses, the four towers affected were considered as discrete.

In-field water sample collection and preparation
Prior to starting this study, all operational towers were tested by building operators at start-up with qPCR, weekly with Figure 1 | Study design and water sample collection overview. Weekly on-site qPCR testing was performed on 53 HVAC towers (619 individual samples). Of these samples, 307 were sent for laboratory culture testing. Of the 307 samples sent for laboratory culture testing, 61 were also sent for laboratory qPCR testing. dipslides, and every 4 weeks with laboratory culture testing (Public Works and Government Services Canada ).
Since the on-site qPCR system was being evaluated against these existing practices, there was some heterogeneity in terms of culture laboratories (and culture methods) used by different buildings. During the in-field study, HVAC  Table 1.
In addition to these regularly scheduled monthly culture tests, extra samples were collected for culture testing, such that all towers were tested every 2 weeks on average. After week 5 of the study, a selection of samples that demonstrated a positive on-site qPCR result of >40 Genomic Units per milliliter (GU/mL) were sent for additional culture testing in an external laboratory.

Quantification of L. pneumophila in HVAC samples by external laboratory qPCR
During the study, 61 water samples were shipped to the following external laboratories after being collected as  Table 2.
Concordance between on-site qPCR, external laboratory qPCR, and culture for L. pneumophila-

positive HVAC samples
To test the concordance of on-site qPCR against external laboratory methodologies, 19 water samples that had been reported positive by on-site qPCR were shipped to three

Categorization of test results
Test results for qPCR and culture were categorized as either positive or negative. Concentrations <10 GU/mL were considered to reflect cooling towers under control (described as negative in our results). This is also reflective of current standards for Legionella monitoring as a properly controlled tower (Public Works and Government Services Canada ). Samples with a concentration of 10 GU/mL were considered as positive, which would require additional monitoring or action such as potentially shutting down the tower. This threshold was also selected to normalize the results from external laboratories and to account for their variable limits of detection (Tables 1 and 2).

Statistical analysis
For statistical analysis, the GU/mL values were expressed as decimal logarithms. Statistical analysis was performed according to the recommendations in ISO/TS 12869:2012.
Linear correlation between datasets generated by the three different methodologies (on-site qPCR, laboratory culture, and laboratory qPCR) was performed using the Pearson correlation coefficient. Chi-square (χ 2 ) tests were performed to compare multiple populations to determine if there was a statistical difference (p-value < 0.05).

Verification of inclusivity and exclusivity
All 15 L. pneumophila serogroups in the inclusivity panel were positively detected by the Spartan Cube ® (Table S1).
All 25 microbial species in the exclusivity panel were not detected (Table S2). (Tables S1 and S2 are available with the online version of this paper.)

Verification of calibration curve
Analysis of the on-site qPCR system's calibration curve  (Table S3, available online).
Verification of lower limit of detection and limit of quantification The limit of detection (LOD) of the on-site qPCR system was verified at 2 GU/reaction (Table S4). Similarly, the limit of quantification (LOQ) was verified at 20 GU/ reaction (Table S5) with an accuracy at the LOQ (E LQ ) of 0.15. (Tables S4 and S5 are available online.) Verification of the entire on-site qPCR system Results showed that recovery of the on-site qPCR system was not affected by matrix conditions in the tested samples (Table 3) In-field study results for the on-site qPCR system and laboratory culture   gave a laboratory culture positive result that was negative by on-site qPCR (Figure 2(a)). The majority of discordant results (21% of total) consisted of a positive on-site qPCR result that was negative by laboratory culture. Overall, concordance between results was laboratory-dependent ( Figure 2(b)).  (Figures 3(a) and 3(b)). All samples tested by laboratory qPCR experienced a shipping delay of 24-72 h.

Concordance between laboratory culture and laboratory qPCR
A subset of 61 samples (61/307) was tested by both laboratory culture (three different laboratories) and laboratory qPCR (four different laboratories). Concordance was poor poor concordance (1/19) with on-site qPCR ( Table 6).  (Table 7). These findings clearly indicate that there is a significant 'shipping effect' and that the time delay between sample collection and analysis can have a large impact on quantification.

Shipping effects on in-field samples
In order to confirm the shipping effect on L. pneumophila quantification, samples that were identified as positive by on-site qPCR were evaluated with three different methodologies: delayed on-site qPCR, laboratory qPCR, and laboratory culture. Relative to the original on-site qPCR result, the 'shipping effect' was again found to be substantial and was consistent across methodologies. Approximately 72% of samples displayed degradation, 15% showed no change, and 13% showed growth (Table 8). There were no statistically significant differences between the three methodologies (χ 2 p-value ¼ 0.89) confirming the universality of the 'shipping effect' in these samples.
Week-over-week L. pneumophila growth The potential significance of weekly versus monthly testing was evaluated. From weekly testing, it was observed that rapid L. pneumophila growth events occurred in 11/20 text indicates positive (10 GU or CFU/mL) results that were obtained by more than one method for a given sample.
positive cooling towers. These towers experienced between 3-and 21-fold growth over 7 days. Furthermore, the effect of testing weekly, bi-weekly, every 3 weeks, and monthly was analyzed to identify the number of transitional events from <10 to 10 GU/mL that would be missed with reduced testing frequency. Testing every 3 or 4 weeks would miss half of the events, whereas biweekly testing would miss approximately one-third (Table 9). The calibration curve of the on-site qPCR system was verified across a dynamic range of 20-20,000 GU/reaction.
As reflected by in-field testing, this dynamic range covers concentrations that are relevant to real-world results, such as the threshold of 1,000 CFU/mL that is found in various Legionella testing regulations around the world (Peter et al. ). The lower limit of detection for the qPCR assay was verified to be 2 GU/reaction and 8 GU/mL for the entire system. This LOD is similar to the 5-10 GU/reaction in previously validated L. pneumophila qPCR assays (Collins et al. ; Omiccioli et al. ).
Verification of the entire system with water samples suggests that it is not affected by the matrix, which was further investigated through in-field testing. This is impor- This finding suggests that the on-site qPCR system may be a superior detection method compared to culture (Whiley & Taylor ). Laboratory culture results were similar but generally under-reported positives compared to on-site qPCR: 9% of (χ 2 p-value ¼ 0.70).  The findings of this study strongly suggest that on-site qPCR is able to accurately detect and quantify L. pneumophila in HVAC cooling towers and has the potential to significantly reduce public health risk compared to existing testing methods. Given that the on-site qPCR system is comparable to culture, its test results can be used within existing standards and action levels and is an important addition to current testing methods.

CONCLUSIONS
A new on-site qPCR detection system for L. pneumophila has been developed that provides immediate results in less than 1 hour. This validation study has shown that the system meets the objectives of ISO/TS 12869:2012 and performs as well as previously published qPCR assays.
In the HVAC cooling towers monitored in this study, we found that on-site qPCR was more sensitive and detected more positive towers than culture. Furthermore, the concordance between on-site qPCR and culture was significantly higher than that observed between external laboratory qPCR and culture. However, the degree of concordance was both laboratory-and tower-dependent. Comparable results between positive on-site qPCR and culture suggested that the on-site detection system is not prone to over-quantification due to the presence of dead bacteria or free DNA.
Additionally, we demonstrated that shipping time-delay had a significant impact on Legionella enumeration in HVAC water samples regardless of methodology. Furthermore, we showed that on-site qPCR was a more reliable and rapid method of Legionella quantification compared to laboratory culture and that increasing the frequency of testing greatly improved response time to elevated levels of Legionella.