Mycobacterium avium complex genomics and transmission in a London hospital

Background Non-tuberculous mycobacteria (NTM) are environmental microorganisms and opportunistic pathogens in individuals with pre-existing lung conditions such as cystic fibrosis (CF) and non-CF bronchiectasis. While recent studies of Mycobacterium abscessus have identified transmission within single CF centres as well as nationally and globally, transmission of other NTM species is less well studied. Methods To investigate the potential for transmission of the Mycobacterium avium complex (MAC) we sequenced 996 isolates from 354 CF and non-CF patients at the Royal Brompton Hospital (London, UK; collected 2013–2016) and analysed them in a global context. Epidemiological links were identified from patient records. Previously published genomes were used to characterise global population structures. Results We identified putative transmission clusters in three MAC species, although few epidemiological links could be identified. For M. avium, lineages were largely limited to single countries, while for Mycobacterium chimaera, global transmission clusters previously associated with heater-cooler units (HCUs) were found. However, the immediate ancestor of the lineage causing the major HCU-associated outbreak was a lineage already circulating in patients. Conclusions CF and non-CF patients shared transmission chains, although the lack of epidemiological links suggested that most transmission is indirect and may involve environmental intermediates or asymptomatic carriage in the wider population.


Introduction
Non-tuberculous mycobacteria (NTM) are ubiquitous environmental microorganisms found in soil and water, and are considered opportunistic pathogens in humans. Individuals with pre-existing genetic or acquired lung diseases such as cystic fibrosis (CF), non-CF bronchiectasis and COPD are more prone to NTM disease, although individuals with no known immune dysfunction can also present with NTM infections [1][2][3]. Globally, disease due to NTM infections is increasing in prevalence; the estimated prevalence of NTM disease in the USA rose from 2.4 per 100 000 in the early 1980s to 15.2 per 100 000 in 2013 [4]. NTM infections may be progressive and treatment requires prolonged multidrug therapy [5]. Treatment is often unsuccessful due to an absence of antimicrobial agents with low toxicity and effective in vivo activity against NTM species [1]. pathogens in the past three decades [6][7][8]. Another member of the MAC, Mycobacterium chimaera, has also been implicated in numerous global infections associated with cardiothoracic surgery, with the source of infections linked to LivaNova heater-cooler units (HCUs) contaminated during their manufacture [9][10][11].
Until recently the prevailing hypothesis was that NTM infections were due to independent acquisitions from environmental sources such as soil, contaminated drinking water distribution systems and household plumbing [12,13]. Recent studies of M. abscessus in CF patients have, however, identified indirect patient-patient transmission within a single CF centre as well as the presence of globally circulating clones of M. abscessus among CF patients worldwide [14][15][16][17][18]. An observational study of M. abscessus across England showed that these dominant clones are also found in patients with other chronic respiratory diseases, but was unable to identify epidemiological links for most closely related isolates, suggesting environmental acquisition [19]. However, a wider analysis demonstrated that transmission networks involve both people with CF and those without, and that these networks are global. It is therefore likely that transmission is complex, involving multiple patient cadres as well as environmental intermediates [20]. In the case of M. chimaera, a global genetic analysis of 250 isolates from patients, HCUs and the factory of origin suggested a point-source contamination during manufacture causing global distribution followed by localised transmission [11].
While transmission of MAC between CF patients in the USA has been investigated [21], little work has been done to examine whether transmission of MAC occurs between patients with CF, non-CF bronchiectasis or other chronic respiratory diseases. Using a large collection of longitudinal isolates collected from patients attending the Royal Brompton Hospital (London, UK), the aims of this study were to characterise the population structure of MAC, to identify potential transmission chains involving patients with CF and other non-CF lung conditions, and to place the Royal Brompton Hospital isolates in a global context.

Methods
Bacterial isolates and study period Isolates were collected from 363 patients attending the respiratory inpatient and outpatient clinics of the Royal Brompton Hospital between January 2013 and April 2016. DNA extractions were performed on confirmed MAC cultures and the Illumina Hiseq X10 platform (Illumina, San Diego, CA, USA) was used to generate 2×150 bp paired-end reads (supplementary text S1). Sequencing reads were deposited in the European Nucleotide Archive under project PRJEB21813 (supplementary file S2).
Whole-genome sequencing and data analysis Following sample quality control, whole-species maximum likelihood phylogenetic trees were built with IQ-TREE version 1.6.5 using a previously described pipeline [22] (supplementary text S1), including published sequences for context (supplementary file S3). Pairwise single nucleotide polymorphism (SNP) distances for within-patient longitudinal isolates for each species in the Royal Brompton Hospital datasets were used to calculate thresholds for defining transmission clusters (supplementary text S1).

Patient demographics
In total, 354 patients were included in the study. The clinical and demographic characteristics of the study population are shown in table 1.  1a). In this group, most of the isolates collected were typically from a single species, with other species observed more infrequently (figure 1b). Subsequent analyses in this study will focus on the three predominant species in the dataset: M. intracellulare, M. avium (M. avium subsp. avium and M. avium subsp. hominissuis) and M. chimaera.

Species distribution
M. intracellulare 162 genomes from 37 patients were identified as M. intracellulare (figure 2a). 11 patients had CF, 17 had non-CF bronchiectasis and seven had other lung conditions (COPD n=3; interstitial lung disease (ILD) n=3; asthma n=1; congenital pulmonary airway malformation n=1). Disease metadata were missing for two patients. Genomic clustering with fastBAPs identified nine lineages, with three of these having more than 10 genomes (figure 2a). Following remapping to local references for the three largest lineages, three putative transmission clusters were identified, with the largest, Mi_FB3_1, composed of 16 patients (figure  2b and supplementary table S2). Of these 16 patients, eight had non-CF bronchiectasis, seven had CF and one had ILD. During the sampling period, four patients were treated with antibiotics, with successful outcomes in three patients. No epidemiological links were identified between patients in the year prior to isolate collection. Comparison with 77 previously published M. intracellulare isolates showed that the most closely related contextual isolate was collected in the UK in 2015/2016 [23] and formed part of a clade containing isolates from the Mi_FB3_1 cluster (figure 2c).
M. avium 405 sequenced isolates collected from 176 patients were identified as M. avium. Sequence data for all M. avium isolates were mapped to a single reference (M. avium subsp. avium 104; NC008595.1) and a phylogenetic tree constructed with M. avium subsp. paratuberculosis (DRR263663) as an outgroup (supplementary figure S2). The structure of this phylogeny showed that there were two major clades which corresponded to the two subspecies M. avium subsp. avium (n=207) and M. avium subsp. hominissuis (n=198). Each of these subspecies was analysed separately.
M. avium subsp. avium Of the 76 patients infected with M. avium subsp. avium, 15 had CF, 39 had non-CF bronchiectasis and 16 had other lung conditions (COPD n=10; ILD n=3; allergic bronchopulmonary aspergillosis (ABPA) n=3; asthma n=6), while three patients, including one smoker, had no pre-existing respiratory disease (metadata were unavailable for three patients) (figure 3a). Clustering of the 207 M. avium subsp. avium genomes identified 18 fastBAPS lineages (figure 3a). Remapping of the four largest lineages allowed the identification of seven putative transmission clusters comprising between two and 10 patients    [23]. If the SNP threshold used to calculate putative transmission clusters circulating in the Royal Brompton Hospital was applied, then one isolate from the Royal Brompton Hospital (40-1) formed part of a cluster comprised of isolates from QUAN et al. [23]. The majority (35 out of 46) of the isolates forming MAA_FB15 were collected in the USA, with most from a study investigating M. avium in the community and household water in Philadelphia (figure 4b) [24]. The pairwise distances between two of the Royal Brompton Hospital isolates and other UK isolates and isolates from the USA in MAA_FB15 were under our threshold, suggesting potential cross-Atlantic transmission (supplementary figure S4a).            figure 5b, and comprised patients with CF (n=4), non-CF bronchiectasis (n=7), COPD (n=3) and a single patient with no pre-existing lung condition (metadata were unavailable for one patient). Two pairs of patients in each of the clusters, MAH_FB8_1 and MAH_FB14_9, were found to have epidemiological links.   S7).

Discussion
The aims of this study were to characterise the population structure of MAC isolates collected in a London hospital, and to identify potential transmission between CF and non-CF patients. During the sampling period we identified 45 patients infected with more than one species or lineage. This suggests that many of the patients had polymicrobial infections and that certain lineages were preferentially sampled at different      times. This has been observed previously in other longitudinal samples of patients with MAC lung disease [26] and suggests that single colonies from sputum swabs may be underrepresenting the underlying diversity of MAC species in patients. Therefore, to better understand the ecology and transmission of MAC species in the lung, a deep-sequencing approach in which plate sweeps from swabs or mycobacteria growth indicator tube cultures are utilised should be considered for future studies.
Previous studies have used thresholds of 15-25 SNPs to identify putative transmission clusters in NTM species [14,19,24]. Rather than using a previously applied or arbitrary threshold, we used a method developed to define SNP thresholds in Staphylococcus aureus [27]. This allowed us to calculate a threshold for each species using the longitudinal isolates collected in the Royal Brompton Hospital. The thresholds calculated for M. intracellulare and M. avium subsp. hominissuis of 16 SNPs were similar to those previously applied [24]. While we obtained higher thresholds of 30 and 58 SNPs for M. chimaera and M. avium subsp. avium, respectively, this did not result in an inflated number of transmission clusters, as applying a threshold of 16 SNPs to the M. avium subsp. avium dataset would have resulted in only two fewer clusters. These larger values were predominantly due to high levels of within-host diversity observed in patients with both M. avium subsp. avium and M. chimaera.
Using these SNP thresholds, we were able to identify putative transmission clusters among all four of the MAC species/subspecies investigated. Broadly, two different patterns of infection were observed. Among M. intracellulare and the two M. avium subspecies, our results suggested that there have been multiple independent introductions into the human population followed by onward transmission. Our analysis of M. chimaera, however, identified a single large transmission cluster which linked 106 out of 155 (68.4%) of our patients infected with this species. Most of these transmission clusters contained both CF and non-CF patients, and a small number also contained patients with no pre-existing lung conditions.
Using our criteria, we were unable to identify potential epidemiological links for most patients included in the transmission clusters. The exception to this was the largest M. chimaera transmission cluster where we found 12 potential epidemiological links up to 1 year before sampling began. The tight infection control surrounding patients with CF, which prioritises preventing cross-infection through hygiene and segregation, means that they are unlikely to have interacted with patients with other lung conditions while in hospital. Outpatient clinics are organised so that patients are seen individually and, during inpatient care, CF patients are given individual rooms with en suite facilities. This implies that transmission is likely occurring via other pathways, such as through environmental intermediates or through a reservoir of healthy, asymptomatic carriers in the wider population. This observation is not unique and there has been considerable debate whether NTM transmission between patients is occurring, even when transmission is strongly supported by genome-based analysis [14,19,28]. The limited evidence of epidemiological links between patients in transmission clusters as well as the ubiquitous presence of NTM species in the environment, especially water supplies, has led some researchers to suggest that the dissemination of NTM lineages at a national level is associated with exposure to contaminated water supplies [19]. While there is certainly strong evidence for the same lineages being isolated from water supplies and patients in the same location [24], the absence of a single national water supply in even a comparatively small country like the UK would suggest that closely related isolates collected from different geographical locations are unlikely to be due to a single contaminated water supply. It is, however, possible that transmission networks may include local water supplies as intermediates. The alternative hypothesis that NTM populations are being maintained in apparently healthy individuals with no symptoms of NTM lung disease is equally worthy of consideration. This would explain the presence of local and long-distance transmission clusters in the absence of direct epidemiological links as well as the presence of phylogeographical structure in NTM trees.
To better address how NTM species are being transmitted, future studies should focus on collecting isolates from the environment as well as from the general population, perhaps focusing on smokers where there is some evidence that they may provide a reservoir for NTM species and other opportunistic pathogens [20].
The inclusion of published genomes alongside the Royal Brompton Hospital isolates allowed us to use our defined pairwise SNP thresholds to identify putative MAC lineages circulating in the UK and globally. Apart from the M. chimaera lineages that were known to be associated with HCUs, we were unable to identify large globally circulating lineages like those observed in M. abscessus. This potentially suggests different evolutionary dynamics among the MAC species or, in the case of M. intracellulare, may simply reflect a lack of comprehensive sampling. We were only able to identify a single lineage containing Royal Brompton Hospital isolates among the two M. avium subspecies that included isolates from outside the UK. This cluster comprised most of the isolates in MAA_FB15 (figure 4b). Given that most M. avium genomes have been collected in the UK or USA, a more comprehensive global sampling strategy could potentially reveal the existence of additional globally circulating clones.
When we calculated global transmission clusters for M. chimaera using a pairwise SNP threshold of 30 SNPs (supplementary table S7), we found that most clusters either consisted of isolates from across Europe or included isolates from other parts of the world such as Australia. Notably, we identified a single large transmission cluster containing 489 isolates from 12 countries that included most of the HCU-associated isolates. Previous work investigating the global population structure of M. chimaera suggested that the majority of M. chimaera transmission globally was associated with point-source contamination of HCUs during manufacturing which then led to direct infection of patients undergoing cardiac surgery [11]. Our study included isolates collected from HCUs and cardiac patients as well as patients with lung conditions such as CF and non-CF bronchiectasis. This enabled us to investigate the global population structure of M. chimaera in more detail than has previously been possible [11,29]. Strikingly, we discovered clear evidence for distinct lineages within the global phylogeny that were almost exclusively associated with patient isolates, indicating considerable onward transmission. In addition, most of the HCU-associated isolates were clustered together in a single lineage that was derived from the lineages containing only human isolates ( figure 6b). This was also true of a second, smaller, lineage that contained only HCU and environmental isolates. The results suggest that the predominant HCU-associated lineage is descended from a M. chimaera population already circulating among patients with respiratory diseases, and that a similar derivation has happened on at least one other occasion.
Potential limitations of this study are the lack of environmental samples as well as the distinct bias with respect to country of isolation in the contextual dataset. Given the lack of direct patient-patient contact and epidemiological links, inclusion of samples collected from wards and hospital water supplies could potentially have revealed additional vectors of transmission. Assembling useful contextual datasets is reliant on what is available in the public domain, and the geographical bias towards the USA and UK in these datasets is not confined to this study.
Our results showed that polymicrobial infections by different MAC species are not uncommon and that transmission is occurring between CF and non-CF patients even in the presence of strict infection controls. This suggests that all of the species and subspecies within the MAC, like M. abscessus, are capable of generating lineages that can sustain transmission within the human population, albeit to different extents. We also provided evidence that the lineage responsible for the HCU-vectored M. chimaera outbreak was derived from an existing lineage which was already circulating among patients with pre-existing lung conditions, indicating that this is also true of M. chimaera, even after excluding the extensive HCU-vectored outbreak event. To better understand the transmission dynamics and to prevent continuing circulation of MAC and other NTM species, future studies should include sampling of environmental reservoirs and potential asymptomatic carriers.
Sequencing reads have been deposited in the European Nucleotide Archive under project PRJEB21813 (supplementary file S2).