Nontuberculous mycobacteria (NTM; referring to mycobacterial species other thanM. tuberculosis complex andM. leprae) are ubiquitous environmental organisms that can cause chronic pulmonary infections in susceptible individuals [1,2], particularly those with pre-existing inflammatory lung diseases such as cystic fibrosis (CF) [3]. The major NTM infecting CF individuals around the world isMycobacterium abscessus; a rapidly growing, intrinsically multidrug-resistant species, which can be impossible to treat despite prolonged combination antibiotic therapy [1,3–5], leads to accelerated decline in lung function [6,7], and remains a contraindication to lung transplantation in many centers [3,8,9].

Until recently, NTM infections were thought to be independently acquired by individuals through exposure to soil or water [10–12]. As expected, previous analyses from the 1990s and 2000s [13–16] showed that CF patients were infected with unique, genetically diverse strains ofM. abscessus, presumably from environmental sources. We used whole genome sequencing at a single UK CF center and identified two clusters of patients (11 individuals in total) infected with identical or near-identicalM. abscessus isolates, which social network analysis suggested were acquired within hospitalvia indirect transmission between patients [17]; a possibility further supported by genomic sequencing [18] of a separateM. abscessus outbreak in a Seattle CF center [19].

Given the increasing incidence ofM. abscessus infections in CF and non-CF populations reported globally [3,20,21], we investigated whether cross-infection, rather than independent environmental acquisition, might be the major source of infection for this organism and therefore undertook population-level, multinational, whole genome sequencing ofM. abscessus isolates from infected CF patients, correlating results with clinical metadata and phenotypic functional analysis of isolates.

We generated whole genome sequences for 1080 clinical isolates ofM. abscessus from 517 patients, obtained from UK CF clinics and their associated regional reference laboratories, as well as CF Centres in the US (UNC Chapel Hill), the Republic of Ireland (Dublin), mainland Europe (Denmark, Sweden, The Netherlands), and Australia (Queensland). We identified 730 isolates asM. a. abscessus, 256 isolates asM. a. massiliense, 91 isolates asM. a. bolletii, with three isolates (from 3 different patients) containing more than one subspecies.

Phylogenetic analysis of these sequences (using one isolate per patient), supplemented by published genomes from US, France, Brazil, Malaysia, China, and South Korea (Table S1), was performed and analysed in the context of the geographical provenance of isolates (Figure 1;Figure S1). As done previously [17], we obtained maximum likelihood phylogenetic trees demonstrating separation ofM. abscessus into three clearly divergent subspecies (M. a. abscessus, M. a. bolletii, andM. a. massiliense), challenging recent reclassifications ofM. abscessus into only two subspecies [22].

Within each subspecies, we found multiple examples of deep branches (indicating large genetic differences) between isolates from different individuals, consistent with independent acquisition of unrelated environmental bacteria. However, we also identified multiple clades of near-identical isolates from geographically diverse locations (Figure 1), suggesting widespread transmission of circulating clones within the global CF patient community.

To investigate further the relatedness of isolates from different individuals, we analysed each subspecies phylogeny for the presence of high density phylogenetic clusters (seeSupplementary Methods [23]). We identified multiple dense clusters of isolates, predominantly within theM. a. abscessus andM. a. massiliense subspecies (Figure 2A), indicating the presence of dominant circulating clones. We next excluded clusters found in only one CF centre from further analysis to remove related isolates that might have been acquired from a local environmental point source. We found that most patients (74%) were infected with clustered, rather than unclustered, isolates, principally fromM. a. abscessus Cluster 1 and 2, andM. a. massiliense Cluster 1 (Figure 2B). The median branch lengths of almost all clusters found in two or more CF centers was less than 20 SNPs (range 1-175 SNPs), indicating a high frequency of identical or near identical isolates infecting geographically separate individuals.

To determine how much of the genetic relatedness found within clusters was attributable to recent transmission, we first examined the within-patient genetic diversity ofM. abscessus isolates from single individuals. In keeping with our previously published results [17], we found that 90% of same-patient isolates differed by less than 20 SNPs, while 99% of same-patient isolates differed by less than 38 SNPs (Figure S2). We therefore classified isolates from different individuals varying by less than 20 SNPs as indicating ‘probable’, and those varying by 20-38 SNPs as indicating ‘possible’, recent transmission (whether direct or indirect). We thereby identified multiple likely recent transmission chains in virtually all multi-site clusters ofM. abscessus (Figure 2B), and across the majority of CF centers (Figure S3).

We next examined the global distribution of clustered isolates and found that, in all countries, the majority of patients were infected with clustered rather than unclustered isolates (Figure 2C;Table S2), suggesting frequent and widespread infection of patients with closely related isolates. Moreover, the three dominant circulating clones,M. a. abscessus Clusters 1 and 2, andM. a. massiliense Cluster 1, were all represented in the USA, European, and Australian collections of clinical isolates, indicating trans-continental dissemination of these clades.

We then compared the genetic differences between isolates (measured by pairwise SNP distance) as a function of geography. As expected from our previous detection of hospital-based transmission ofM. abscessus [17], average genetic distances were significantly shorter forM. abscessus isolates from the same CF center than those from different CF centers within the same country or from different countries (Figure 2D). However, we also detected numerous examples of identical or near-identical isolates infecting groups of patients in different CF centers and, indeed, across different countries (Figure 2D), indicating the recent global spread ofM. abscessus clones throughout the international CF patient community.

We applied Bayesian analysis [24] to date the establishment and spread of dominant circulating clones (Figure S4, S5), focusing onM. a. massiliense Cluster 1, which includes isolates from both the Seattle [19] and Papworth [17] CF Center outbreaks, as well as isolates from CF centres across England (Birmingham, London, Leicester), Scotland (Lothian, Glasgow), Ireland (Dublin), Denmark (Copenhagen), Australia (Queensland), and the USA (Chapel Hill, NC) (Figure 3A). We estimate that the most recent common ancestor of isolates infecting patients from all these locations emerged around 1978 (95% CI: 1955-1995), clearly indicating recent global dissemination of this dominant circulating clone amongst individuals with CF (Figure 3A).

Furthermore we were able to resolve individual transmission events between patients infected with dominant circulating clones through two orthogonal approaches. Firstly, using high-depth genomic sequencing of colony sweeps, we were able to track changes in within-patient bacterial diversity in sputum cultures of a single individual over time. By linking the frequency of occurrence of minority variants in longitudinal samples, we were able to define the presence of particular subclones within infected individuals, assign their likely evolutionary development (involving the successive acquisition of non-synonymous mutations in likely virulence genes;Figure 3B), monitor their relative frequencies over time, and demonstrate their transmission between patients (Figure 3B). Secondly, through longitudinal whole genome sequencing of isolates collected over time from individuals, we were able to find multiple examples of the complete nesting of one patient’s sampled diversity within another’s (Figure S6). Such paraphyletic relationships are strongly indicative of recent person-to-person transmission [25].

We next examined potential mechanisms of transmission ofM. abscessus between individuals (which our previous epidemiological analysis had suggested was indirect rather thanvia direct contact between patients [17]). We provide proof of concept for fomite spread ofM. abscessus (detecting three separate transmission events associated with surface contamination of an inpatient room by an individual infected with a dominant circulating clone;Figure S7), and also for potential airborne transmission (by experimentally demonstrating the generation of long-lived, potentially infectious cough aerosols by an infected CF patient;Figure S8).

A potential explanation for the emergence of dominant clones ofM. abscessus is that they are more efficient at infection and/or transmission. We therefore analysed clinical metadata to establish whether outcomes were different for patients infected with clustered rather than unclustered isolates. We correlated clinical outcomes with bacterial phylogeny and the presence of constitutive resistance to two key NTM antibiotics, amikacin and macrolides [26,27], acquired through point mutations in the 16S and 23S ribosomal RNA respectively (Figure 4A). We found no differences in the proportions ofM. abscessus-positive individuals diagnosed with ATS-defined NTM pulmonary disease [1], (namely the presence of two or more culture-positive sputum samples with NTM-associated symptoms and radiological changes), but did observe increased rates of chronic infection in individuals infected with clustered rather than unclustered isolates (Figure 4B). As anticipated for transmissible clones exposed to multiple rounds of antibiotic therapy, we also found high rates of constitutive amikacin and/or macrolide resistance in clustered isolates (Figure 4B). Of note, resistance to these two antibiotics did not necessarily result in a poor clinical outcomes (Figure S9), suggesting that additional bacterial factors might contribute to worse responses in patients infected with clustered isolates.

To explore differences in intrinsic virulence between clustered and unclusteredM. abscessus, we subjected a panel of representative isolates (27 clustered and 17 unclusteredM. a. abscessus; 25 clustered and 13 unclusteredM. a. massiliense) to a series ofin vitro phenotypic assays. While we found no or only minor differences between groups in their colony morphotype, biofilm formation, ability to trigger cytokine release from macrophages (Figure S10) and their overall phenotypic profile (by multifactorial analysis;Figure S11), we detected significantly increased phagocytic uptake (Figure 4C) and intracellular survival in macrophages (Figure 4D) of clustered isolates of bothM. a. abscessus andM. a. massiliense compared to unclustered controls, indicating clear differences in pathogenic potential. Moreover, infection ofSCID mice revealed significantly greater bacterial burden (Figure 4E) and granulomatous inflammation (Figure 4F) following inoculation with clustered rather than unclustered isolates ofM. a. abscessus andM. a. massiliense, confirming differences in virulence between these groups.

In summary, our results reveal that the majority ofM. abscessus infections of individuals with CF worldwide are caused by genetically-clustered isolates, suggesting recent transmission, rather than independent acquisition of genetically-unrelated environmental organisms. Given the widespread implementation of individual and cohort segregation of patients in CF centres in Europe [28], the USA [29], and Australia [30] (which have led to falling levels of MRSA,Burkholderia, and transmissiblePseudomonas infections [31–33]), we believe that the likely mechanism of local spread ofM. abscessus is via fomite spread or potentially through the generation of long-lived infectious aerosols (as identified for other CF pathogens [34–36]). Although further research is needed, both transmission routes are plausible given our findings (Figures S7, S8), and would be potentially enhanced by the intrinsic desiccation resistance ofM. abscessus. Such indirect transmission, involving environmental contamination by patients, is supported by our previous social network analysis of a UK outbreak ofM. abscessus [17] in CF patients, which revealed hospital-based cross-infection without direct person-to-person contact, and by the termination of a SeattleM. abscessus outbreak associated with the introduction of clinic room negative pressure ventilation and double room cleaning [19]. The long-distance spread of circulating clones is more difficult to explain. Importantly we found no evidence of CF patients or of equipment moving between CF centers in different countries indicating that the global spread ofM. abscessus may be driven by alternative human, zoonotic or environmental vectors of transmission.

Our study illustrates the power of population-level genomics to uncover modes of transmission of emerging pathogens and has revealed the recent emergence of global dominant circulating clones ofM. abscessus that have spread between continents. These clones are better able to survive within macrophages, cause more virulent infection in mice, and are associated with worse clinical outcomes, suggesting that the establishment of transmission chains may have permitted multiple rounds of within-host genetic adaptation to allowM. abscessus to evolve from an environmental organism to a true lung pathogen.