Abstract
Mycobacterium chimaera is an opportunistic environmental mycobacterium, belonging to the Mycobacterium intracellulare complex. Although most commonly associated with pulmonary disease, there has been growing awareness of invasive M. chimaera infections following cardiac surgery. Investigations suggest world-wide spread of a specific M. chimaera clone, associated with contaminated hospital heater-cooler units used during the surgery. Given the global dissemination of this clone, its potential to cause invasive disease, and the laboriousness of current culture-based diagnostic methods, there is a pressing need to develop rapid and accurate diagnostic assays, specific for M. chimaera. Here, we assessed 354 mycobacterial genome sequences and confirmed that M. chimaera is a phylogenetically coherent group. In silico comparisons indicated six DNA regions present only in M. chimaera. We targeted one of these regions and developed a TaqMan qPCR assay for M. chimaera with a detection limit of 10 CFU in whole blood. In vitro screening against DNA extracted from 40 other mycobacteria and 22 bacterial species from 21 diverse genera confirmed in silico predicted specificity for M. chimaera. Screening 33 water samples from heater cooler units with this assay highlighted the increased sensitivity of PCR compared to culture, with 15 of 23 culture negative samples positive by M. chimaera qPCR. We have thus developed a robust molecular assay that can be readily and rapidly deployed to screen clinical and environmental specimens for M. chimaera.
Introduction
Mycobacterium chimaera is an environmental mycobacterium and infrequent pathogen, most commonly linked with pulmonary disease (1-8). Interest in M. chimaera has heightened with global reports of invasive infections (including endocarditis and vascular graft infections associated with the use of LivaNova PLC (formerly Sorin Group Deutschland GmbH) Stöckert 3T heater-cooler units during cardiac surgery. The most plausible hypothesis for this widespread contamination is a point-source outbreak, although the underlying causative factors are not currently known (9-16). Phylogenetic comparisons of 16S–23S rRNA internal transcribed spacer (ITS) sequences, and/or partial rpoB or hsp65 sequences (2, 5-7, 17, 18) suggest M. chimaera as a distinct entity within the M. intracelluare complex (6) and two recent population genomic analyses have confirmed this relationship (8, 13). The complete 6,593,403 bp genome sequence of M. chimaera ANZ045 revealed a single circular 6,078,672 bp chromosome and five circular plasmids ranging in size from 21,123 bp to 324,321 (8). M. chimaera is slow-growing, therefore current culture-based laboratory methods, followed by Sanger sequencing of amplicons for one or more combinations of conserved sequence regions, or line-probe hybridization assays are not amenable to timely and specific detection of this pathogen. This delay carries significant clinical, health provision, and medico-legal implications as patients may be exposed to contaminated machines during this turn-around time of up to 6-8 weeks. A rapid and reliable diagnostic tool is urgently needed to support clinical management of patients and to establish the efficacy of heater-cooler unit decontamination procedures. Here we addressed this issue by using comparative genomics to identify DNA sequences present in M. chimaera and absent from other mycobacteria. We describe the initial development and validation of a sensitive, specific and quantitative PCR assay for identification of M. chimaera in both clinical and environmental samples.
Materials and methods
Bacterial strains and genome sequences
Mycobacterium chimaera strain DMG1600125 (a 2016 HCU isolate from New Zealand) was used for spiking experiments (8). The mycobacterial genome sequences used in this study are listed in Table S1. M. chimaera was grown on Brown and Buckle whole-egg media, Middlebrook 7H9 broth or Middlebrook 7H10 agar (Becton Dickinson) supplemented with 10% (v/v) oleic acid albumin dextrose complex (OADC; Difco) or Middlebrook 7H10 agar. Cultures were incubated without shaking at 37°C. M. chimaera colony counts were obtained by spotting 3 µL microliter volumes of six, 10-fold serial dilutions of a M. chimaera culture suspensions in quintuplicate on two Middlebrook 7H10 agar plates. The colonies were counted after incubation for four weeks at 37°C.
Genomic DNA extraction methods, M. chimaera culture and environmental isolation
Purified M. chimaera genomic DNA for TaqMan assay validation was extracted from 50 mg wet weight cell pellets as described (19) and measured by fluorimetry using the Qubit and the High Sensitivity DNA kit (Thermofisher). For spiking experiments in blood, M. chimaera DNA was extracted from 100 μL volumes of whole blood, using the Qiagen Blood & Tissue DNA extraction kit. Purified DNA was eluted from the columns in a 200 μL volume of 10 mM Tris (pH 8.0) (Qiagen). Total bacteria were concentrated from 30-1000 mL volumes of water collected from heater-cooler units by filtration through 47 mm, 0.22 μM mixed cellulose ester Millipore membranes. Immediately after filtration, membranes were aseptically placed in sterile 50 ml plastic tubes and stored at −70° C. DNA was extracted from membrane concentrate using the MoBio PowerWater DNA isolation kit following the manufacturer’s instructions (MoBio) with an additional physical disruption step consisting of 2 × 20 sec at 5000 rpm in a Precellys 24 tissue homogenizer. To prevent cross contamination, a sterilized filtration device was used for each sample and sterile, distilled water extraction blanks were filtered and processed (100 mL volumes) at a frequency of one for every 10 test samples. Culture isolation of M. chimaera from 50 mL volumes of water samples was undertaken as described (20).
Population structure and phylogenetic analysis
Snippy v3.1 (https://github.com/tseemann/snippy) was used to align Illumina sequence read data or de novo assembled contigs from M. chimaera and related mycobacterial genomes against the fully-assembled, complete MC_ANZ045 reference genome to call core genome single nucleotide polymorphism (SNP) differences and generate pairwise sequence alignments. Hierarchical Bayesian clustering (hierBAPS) was performed using these core whole-genome SNP alignments as input to assess population structure (a prior of 6 depth levels and a maximum of 20 clusters was specified) (21), with phylogenies inferred using FastTree v2.1.8 under a GTR model of nucleotide substitution (22). Pairwise SNP analysis between groups of genomes was performed using a custom R script (https://github.com/MDU-PHL/pairwise_snp_differences). Recombination detection was performed using ClonalFrameML v1.7 (23).
In silico subtractive hybridization and target identification
To identify regions of DNA present in M. chimaera but absent from other mycobacteria, the Illumina sequence reads of 46 M. chimaera isolates from Australia and New Zealand and eight publicly available M. intracellulare genomes (Table S1) were aligned using BWA MEM v0.7.15-r1140 (https://arxiv.org/abs/1303.3997) to a complete M. chimaera reference genome (MC_ANZ045) (8). The read depth at each position was examined to identify those positions in the reference genome that were present across all M. chimaera isolates but absent from all M. intracellulare genomes. These genomic regions were extracted from MC_ANZ045 and compared against the NCBI Genbank non-redundant (nt) nucleotide database using NCBI BLAST v2.5.0 (CITE?) with parameters -remote -max_target_seqs 100 -task blastn -outfmt “6 std qcovs staxid ssciname”. Resulting BLAST hits that were missing a taxon name were retrieved from the NCBI taxonomy database using the taxon id. Ignoring BLAST hits against bonafide M. chimaera sequences, the query alignment positions for every hit were extracted and were used to obtain all the sequence segments that had no hits against the Genbank ntxs database and that were greater than 500 bp in length. For this, bedtools complement and getfasta tools were used (24). The sequence segments thus obtained were considered candidate M. chimaera-specific genomic regions. The presence of these regions across a wider collection of M. chimaera was assessed by downloading all M. chimaera genome sequence reads present in the NCBI sequence read archive SRA as of October 2016 (Table S1) and processing through the Nullarbor pipeline v1.2 (https://github.com/tseemann/nullarbor). The output information was used to filter out poor quality or non-M. chimaera reads sets based on G+C content significantly below 66%, an average read depth below 30, a total contig length above 8Mb, predicted rRNA genes greater than four, or a total of sequence aligned to the reference genome below 70% (Table S1). Using Snippy again, all M. chimaera genomes identified above were mapped to a version of the MC_ANZ045 reference genome in which the non-M. chimaera-specific sequence regions had been hard masked. The resulting multiple sequence alignment was parsed using a custom Perl script to identify those M. chimaera-specific regions that were present in all M. chimaera genomes. These DNA sequences were inspected further for development of M. chimaera TaqMan PCR diagnostic assays. TaqMan primers and probes (Sigma Oligonucleotides) were designed using Primer3 (25) and Primer-BLAST against NCBI nt database was used to check that the primers and probes designed were specific to M. chimaera. TaqMan probe 1970-P were labeled with the fluorescent dye 6-carboxyfluorescein (FAM) at the 5′ end and a nonfluorescent quencher at the 3′ end (Sigma Oligonucleotides). To assess the context of these M. chimaera-specific regions, AlienHunter v1.4 was used to screen the MC_ANZ045 genome for DNA compositional bias, indicative of horizontally acquired DNA (26).
TaqMan quantitative PCR
TaqMan PCR mixtures contained 2 μl of template DNA, 0.4-μM concentrations of each primer, a 0.2 μM concentration of the probe, SensiFAST Probe Lo-ROX (1x) mix (Bioline), and TaqMan exogenous internal positive control (IPC) reagents (Applied Biosystems) in a total volume of 20 µl. Amplification and detection were performed with the Mx3005P (Stratagene) using the following program: 40 cycles of 95°C for 10 s and 60°C for 20 s. DNA extracts were tested in at least duplicate, and negative and positive template controls were included in each run. Standard curves were prepared using eight, 10-fold serial dilutions of M. chimaera genomic DNA at an initial concentration of 120 ng/µL, tested in triplicate. The percentage PCR amplification efficiency (E) for the TaqMan assay was calculated from the slope (C) of the standard curve E = (10^(-1/C))*100. Cycle threshold values for unknown samples were converted to genome equivalents by interpolation, with reference to the standard curve of Ct versus dilutions of known concentrations of M. chimaera genomic DNA. The mass in femtograms of a single M. chimaera genome was estimated as 6.59 fg, using the formula M = (N)*(1.096e-21), where M = mass of the single double-stranded M. chimaera NZ045 reference genome and N = 6593403, which is the length of the M. chimaera NZ045 reference genome, and assuming the average MW of a double-stranded DNA molecule is 660 g/mol. Analyses were performed using Graphpad Prism v6.0h.
Results
Assessment of M. chimaera population structure
To identify DNA segments present only in M. chimaera genomes we first assessed the phylogenetic coherence of “M. chimaera” as a species. Using 96 mycobacterial genome sequences, comprising 63 M. chimaera genomes from North America, Australia, and New Zealand, and 33 other related, publicly available mycobacteria from the Mycobacterium avium-intracelluare complex, we conducted whole genome pairwise comparisons of the 96 taxa to the M. chimaera ANZ045 complete reference chromosome. The 63 M. chimaera genomes included 49 HCU-associated and 14 previously described patient isolates, not all of which were associated with Stöckert 3T HCU contamination (8, 12). These comparisons identified 448,878 variable nucleotide positions in a 2,340,885 bp core genome. A robust phylogeny inferred from the alignments strongly suggested that M. chimaera forms a monophyletic lineage within the M. intracellulare complex (Fig. 1A) (8, 13). Bayesian analysis of population structure (BAPS) using these same data confirmed this clustering (Fig. 1A). Interestingly, this assessment indicated that a publicly available isolate originally identified as Mycobacterium intracellulare (strain MIN_052511_1280) was in fact M. chimaera. The mean number of SNPs between any pair of the 63 M. chimaera isolates, and MIN_052511_1280 (BAPS-3), not adjusted for recombination, was 115 SNPs (range: 1 - 3,024 and IQR: 13 - 31), highlighting restricted core genome variation within this species, particularly given the large 6.5 Mb genome size. In comparison, the mean number of SNPs between 15 M. intracellulare-complex genomes (BAPS-2) was 24,134 SNPs (range: 13 - 39,109 and IQR: 14,780 - 33,138) (Fig. 1A). We then extended this analysis to assess an additional 257 publicly available M. chimaera and related mycobacterial genome sequences (Table S1). Pairwise whole genome comparisons of this larger data set were performed against the ANZ045 reference genome. Population structure analysis indicated 303 mycobacterial genomes fell within BAPS-3 (Table S1). Pairwise comparisons were again performed against the ANZ045 using only these 303 genomes, with five other genome sequences from BAPS-2 included for context. The alignment was filtered to remove sites from the alignment affected by recombination and a phylogeny was inferred from the resulting 10,166 variable nucleotide positions (Fig. S1. Fig. 2). The mean number of SNPs between the 303 genomes was 268 (range: 0 - 3,211 and IQR: 9 - 62). This analysis confirmed that M. chimaera does indeed form a monophyletic lineage, providing a robust genetic definition for the species. As previously reported, HCU-associated isolates from around the world formed a distinct sub-clade within this lineage (Fig. 2) (8, 13).
In silico genome comparisons to identify M. chimaera specific sequences
A subset of 46 genome M. chimaera genome sequences as defined above became the ‘training set’ to find DNA segments present only in M. chimaera (Fig. 2, Table S1). Mapping of DNA sequence reads to the ANZ045 reference genome (refer methods) allowed the identification of 159 genomic segments >500 bp in length and covering 510,924 bp that were present across the 46 M. chimaera isolates and absent from eight M. intracellulare isolates (Fig. 1A). BLAST comparisons of the 159 segments against all entries in the NCBI Genbank nt database and removal of any non-M. chimaera specific sequence reduced the number to 37 segments (covering a total of 37,890 bps). A larger validation set comprising the 63 M. chimaera genomes described in Fig. 1A and 242 additional, publicly available M. chimaera genomes that satisfied our above phylogenomic inclusion criteria was then screened (Table). Six of the 37 M. chimaera-specific regions (SR), covering a total of 8,292 bp, were present in all 305 M. chimaera genomes (Fig. 1B). The six SRs ranged in length from 531 bp to 4,641 bp, the majority overlapping predicted chromosomal protein-coding sequences (Fig. 1B, Table 1). Inferred functions of these CDS are summarized (Table 1). The regions were scanned for sequence polymorphisms and one of these regions (SR1) that was 100% conserved among all M. chimaera, was selected as a template for the design of a TaqMan assay (Fig. 2, Table 2). SR1 spans two predicted CDS that DNA composition analysis and gene annotation predicted lay within a 35 - kb putative prophage or integrative mobile element. A 79 bp TaqMan amplicon (assay ID: 1970) was designed within a 2934 bp CDS (predicted function: unknown) (Table 2).
TaqMan assay specificity testing
The above in silico analyses predicted the TaqMan assay would be diagnostic for the presence of M. chimaera. To test this prediction, DNA was prepared from 42 mycobacteria (including two M. chimaera isolates) and 22 other bacteria from 21 different genera. A pan-bacterial 16S rRNA PCR was first performed to ensure detectable bacterial DNA was present. All 64 DNA samples were positive by 16S rRNA PCR (data not shown) but only the two M. chimaera isolates were 1970-P TaqMan assay positive, supporting the in silico predictions that these assays are specific for M. chimaera (Table S2).
TaqMan assay efficiency and sensitivity testing
To establish the limit of detection and amplification efficiency for the 1970-P TaqMan assay, 10-fold serial dilutions of purified M. chimaera ANZ045 genomic DNA were tested in triplicate. The 1970-P assay showed excellent performance characteristics, with a very good linear response across five orders of magnitude, R2 values >0.99, amplification efficiencies of 94%, and a detection limit around 20 genome equivalents (Fig. 3A). The detection limit was also assessed using dilutions of M. chimaera culture spiked into whole blood. Again, the assay showed excellent performance characteristics in this simulated clinical condition, with an absolute detection limit around 10 CFU (equivalent to 100 CFU/mL of blood). These experiments indicate 1970-P is a suitable qPCR assay for sensitive and quantitative detection of M. chimaera DNA.
Detection of M. chimaera DNA in environmental samples
A key requirement for this assay is the ability to screen water and biofilm samples from heater-cooler units for M. chimaera, to determine if maintenance procedures have removed the bacteria from contaminated units, or prevented contamination. Having assessed the sensitivity and specificity of the assay with laboratory-prepared samples, we next explored performance with environmental samples. We screened concentrates from 33 water samples obtained from heater-cooler units at seven hospitals and HCU distributors in our region, that had been assessed by culture for M. chimaera. A total of 25 of 33 samples were positive by 1970-P TaqMan PCR, with estimated M. chimaera concentrations ranging from 2-102,000 GE/mL of water (Table 3). Using the culture results as a ‘gold standard’, the negative predictive value for the TaqMan assays was high (100%) with all eight culture positive samples also positive 1970-P TaqMan PCR (Table 4). However, there was poor correspondence between culture negative samples and qPCR. Fifteen samples negative by culture returned 1970-P TaqMan positive results, with Ct values for some of these samples less than 24, indicating high M. chimaera concentrations above 10,000 GE per milliliter of original sample (Table 3). Heterotrophic colony count (HCC) at 37ºC is used as a general indicator of water cleanliness, and in some settings may be used as a surrogate indicator of decontamination effectiveness (27). However, we observed a poor correlation between HCC and the presence of M. chimaera as measured by qPCR (Spearman’s ρ=0.2813, p=0.1128) (Table 3), suggesting that HCC is not a suitable surrogate for the presence or absence of M. chimaera.
Discussion
Since 2012 there have been a small but increasing number of case reports of invasive infection with M. chimaera in individuals who have undergone surgical procedures requiring cardiac bypass (8, 10-16). Almost all cases have involved placement of prosthetic valves or other prosthetic material and are linked to use of a specific type of heater cooler unit (HCU) in the bypass procedure (8, 13, 14). As contamination of these HCUs may have occurred at or near the time of manufacture and the machines are widely exported, exposure to M. chimaera during cardiac bypass surgery is an emerging issue in infection control that is not restricted by region or country (8). However, while it appears that generation of aerosols when machines are contaminated may be relatively common, so far the likelihood of infection for any exposed individual is very low (13). From a clinical perspective, this poses a major diagnostic challenge. Post-cardiac surgery M. chimaera infection has a long incubation period, non-specific symptoms and can be misdiagnosed as a steroid-requiring inflammatory condition with potentially disastrous consequences (10, 15). Moreover, there is a significant case fatality rate even when correctly identified and current opinion is that early accurate diagnosis will be key to achieving the best treatment outcomes (28). Given the non-specific symptoms and low prior probability of infection in a large exposed population, clinicians urgently need access to a specific and sensitive test for M. chimaera cardiac and extra-cardiac infection. Infection control practitioners have a different but equally challenging problem with respect to surveying and cleaning contaminated HCUs. In this report, we describe and validate a DNA target that is diagnostic for the presence of M. chimaera. We have shown under simulated conditions that it can accurately detect M. chimaera in human blood samples at very low concentrations and outperform culture in detecting M. chimaera in specimens obtained from contaminated HCUs.
The European Centre for Disease Prevention and Control recommends M. chimaera identification is performed by sequencing at least two conserved fragments among 16S–23S rRNA ITS, 16SrRNA, rpoB and hsp65 [21,22]. Some laboratories have also used the MIN-2 probe in the INNO-LiPA Mycobacteria v2 line probe assay (6, 18). Here we simplify this suite of tests, with a M. chimaera-specific PCR assay, that has the advantages of providing a rapid yes/no result and an estimate of bacterial concentration. The test could be further enhanced with multiple DNA targets. The other five M. chimaera-specific regions reported here could be used to develop additional diagnostic targets (Fig. 1B). We envisage that this test will be used in conjunction with efforts to culture M. chimaera from specimens, as isolates are required for WGS to establish the genetic relatedness of isolates, and potentially for antimicrobial susceptibility testing (although antimicrobial resistance is not thought to be a major problem). As a previous risk assessment by Public Health England suggested a possible legionellosis risk for staff and patients, we propose that our assay should form part of a ‘HCU panel’ along with Legionella PCR (28, 29). While we have designed an assay to detect all M. chimaera, it may also be possible to detect specific M. chimaera lineages. For instance, it may be possible to use DNA deletion polymorphisms to discriminate among intra-species lineages, as in the Mycobacterium tuberculosis complex (30). We are currently exploring this possibility.
The infection risk posed by the water reservoirs within HCUs and the need for regular maintenance has been long recognized (29). Heterotrophic colony counts (HCC) are being considered as surrogates to assess the microbiological quality of HCUs (27, 31). We (like others) found a poor correlation between the presence of M. chimaera and HCC in water samples from HCUs (27), with examples of M. chimaera concentrations of 60,000 GE/mL when HCC were below the limit of detection (Table 3). More evaluation of the 1970-P assay is required, but our data suggest HCC may have an unacceptably high false-negative rate, which significantly reduces its utility for measuring the effectiveness of HCU decontamination procedures.
Screening HCU water samples with our TaqMan assay indicated the widespread presence of M. chimaera and a poor correlation with culture. There are several potential explanations for these observations. Despite the extensive in silico assessments, it is possible that the qPCR assay lacks specificity for M. chimaera, or that the culture method lacks sensitivity, or perhaps DNA from M. chimaera is still present but source organisms are no longer viable. Given the extensive in silico validation undertaken here to ensure target specificity, and the high prior probability that these HCU water samples contained M. chimaera, the discrepancy between culture and qPCR might best explained by either lower sensitivity of the mycobacterial culture method or PCR detection of intact DNA from non-viable M. chimaera. The later explanation is perhaps the most likely given that some of these PCR positive-culture negative samples were obtained from HCUs subjected to extensive decontamination procedures involving extended heating above 70ºC.
In summary, we have developed a new diagnostic tool for rapid, sensitive and specific detection of M. chimaera to help address the urgent need to screen patient and HCU samples.
Funding information
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. DAW, BPH and TPS are supported by National Health and Medical Research Council Fellowships GNT1123854, GNT1105905 and GNT1105525 respectively.
Supplementary material
Tables:
Table S1: Mycobacterial genome sequences used in this study
Table S2: Summary of TaqMan assay in vitro specificity screening
Acknowledgments.
We thank submitting health-care facilities and laboratories for providing water samples and mycobacterial isolates. We are grateful to Chris Coulter for provision of materials and critical review of the manuscript.