A subset of 6 isolates (2u, 12c, 14h, 19f, 20h, and 28c) representing the breadth of genomic diversity observed within the Papworth cystic fibrosis outbreak clusters 1 and 2 (6) were selected. Illumina sequencing reads from each of these isolates were assembled into sets of contigs by using Velvet software (21). These contigs were combined with draft genome sequences of the Seattle cystic fibrosis outbreak and available whole-genome sequences of M. abscessus subsp. massiliense (Table 1) and subjected to whole-genome multiple sequence alignments by using Mugsy software (22). Core segments of the alignment that are shared among all isolates included in the analysis were identified and concatenated by using Phylomark software (23). Concatenated nucleotide sequences, including single-nucleotide polymorphisms (SNPs), were then used for construction of a neighbor-joining phylogenetic tree by using MEGA software (24). The use of microbial samples and data was approved by the ethics committees at each of the institutions involved.

To replicate data from the Papworth cystic fibrosis outbreak clusters 1 and 2 (6) by using a similar approach, we mapped sequencing reads from the subset of 6 Papworth isolates, together with reads with from the 3 Seattle cystic fibrosis isolates and soft tissue strain CRM-0020 from Brazil (Table 1), onto the M. abscessus type strain ATCC 19977^T reference genome by using BWA software (25). Variants, including SNPs, were called by using GATK software (26) and filtered for quality. The SNP panel was used for construction of a neighbor-joining phylogenetic tree by using MEGA software. The resulting tree replicated the topology of clusters 1 and 2 and showed that the Seattle isolates are most closely related to cluster 2.

Standard PCR and sequencing strategies were used to amplify and analyze partial sequences of the rpoB (723 bp) (27,28) and secA1 (465 bp) (29) genes. In addition, a multilocus sequence typing (MLST) scheme (29,30), including primers to 13 housekeeping genes (cya, gdhA, argH, glpK, gnd, murC, pgm, pknA, pta, pur, rpoB, hsp65, and secA1) was used to conduct electronic PCR on the panel of 20 M. abscessus subsp. massiliense genomes (Table 1). Published forward primers for cya and gdhA (30) did not amplify in silico for some M. abscessus subsp. massiliense strains; therefore, the following new primers conserved across the M. abscessus group were used: cya_F_new 5′-GCC TGC GTA AGG GTG ATG-3′ and gdhA_F_new 5′-GTG AAG CTC GCC GCC TGC-3′. Alleles from each gene were extracted and concatenated for each genome, the panel of concatenated sequences was aligned by using ClustalW software (31), and the core segments of the alignment were used for construction of a neighbor-joining phylogenetic tree by using MEGA software.

A core genome phylogenetic tree (Figure 1) showed a tight cluster of the 3 Seattle cystic fibrosis outbreak strains. The Seattle cystic fibrosis cluster was closely related to the 2 cystic fibrosis clusters described for the Papworth outbreak (6) and the Birmingham, UK, cystic fibrosis isolate 47J26 (9). Furthermore, the Seattle and Papworth cystic fibrosis outbreak strains showed some relatedness to strains CRM-0020 and GO-06 derived strains (known collectively as BRA-100) isolated during an epidemic of soft tissue infections in Brazil (32) and the M. abscessus subsp. massiliense M18 strain from Malaysia (10).

The cumulative size of core segments of Mugsy alignments provides information on relatedness among groups of strains compared. The core genome reduces in size as more genomes are added; an expected major decrease occurs after addition of more distant strains to the group. The average genome size of cystic fibrosis outbreak strains was 4.81 Mb for Seattle (n = 3) and 4.97 Mb for Papworth (n = 6). The Seattle and Papworth cystic fibrosis outbreak strains (n = 9) shared a core genome of 4,264,844 nt, which is almost unchanged by including the Birmingham cystic fibrosis strain 47J26 (n = 10; 4,264,127 nt). Addition of the soft tissue outbreak strain CRM-0020 from Brazil (n = 11) (32) decreased the core to 4,231,390 nt, and adding the related outbreak strain GO 06 from Brazil (n = 12) (8,33), led to an additional decrease in the core genome to 4,043,718 nt. As expected, including unrelated available clinical M. abscessus subsp. massiliense strains (n = 20, including M139 with ambiguous subspecies taxonomic assignment (12), (Table 1), reduced the core genome size to 3,869,950 nt. Further addition of M. abscessus subsp. abscessus (n = 2) and M. abscessus subsp. bolletii (n = 2) genomes (Table 1) reduced the core to 3,828,656 nt.

The core genome of 10 strains representing the Papworth cystic fibrosis (n = 6), the Seattle cystic fibrosis (n = 3), and soft tissue CRM-0020 (n = 1) outbreaks comprised 4,231,390 nt. Strain GO 06 was excluded from the analysis because its genome harbors a large number of ambiguous nucleotides and an unusual hybrid appearance with fragments of M. abscessus subsp. massiliense and M. abscessus subsp. abscessus sequences. Strains 47J26 and M18, isolated from the sputum of a cystic fibrosis patient in Birmingham, UK, and a lymph node sample from a patient in Malaysia, respectively, were related to the outbreak strains (Figure 1). However, no information was available about any epidemiologic link between cystic fibrosis strain 47J26 to reported or unpublished outbreaks, and no clinical information was available about the patient from whom strain M18 was isolated. Therefore, both strains were excluded from the SNP analysis. Nevertheless, SNPs for these 3 strains at positions relevant to the outbreak strains are shown in the Technical Appendix.

A total of 293 identical SNPs in the core segments of Mugsy alignments were shared by the 10 outbreak strains but were different in available M. abscessus subsp. massiliense strains not related to outbreaks (Figure 2; Technical Appendix). Of the 293 SNPs, 95 gave rise to nonsynonymous mutations in several genes, including virulence factors (mammalian cell entry and yrbE proteins), transcriptional regulators (TetR family), and lipid metabolism genes (Technical Appendix). Eleven SNPs were shared only by Papworth and Seattle cystic fibrosis outbreak strains (n = 9), including nonsynonymous mutations in the preprotein translocase secA1 and a putative lyase (Figure 2; Technical Appendix). Sixteen SNPs were shared only by the 3 Seattle cystic fibrosis outbreak strains, including nonsynonymous mutations in a mycobacterial large membrane protein (MmpL) family involved in lipid transport and virulence (34) and genes involved in amino acid and energy metabolism (Figure 2; Technical Appendix). Eighty-six SNPs were present only in strain CRM-0020 (soft tissue outbreak) from Brazil, including nonsynonymous mutations in an MmpL family protein; transcriptional regulators; and lipid, amino acid, and energy metabolism genes (Figure 2; Technical Appendix).

Having shown high-level relatedness among Papworth and Seattle cystic fibrosis outbreak strains and their relatedness to the soft tissue outbreak strains from Brazil, we also searched for genomic regions ≥200 nt outside the core genome that were specific to subsets of isolates. A single region of ≈11.5 kb was unique to the Papworth cystic fibrosis isolates (n = 6) and encompassed 2 conserved hypothetical proteins, 2 phage integrase family proteins, and an MmpL family protein. Alignment of the MmpL family protein with distinct MmpL proteins described above for the Seattle cystic fibrosis outbreak and the Brazil soft tissue outbreak showed diversity at several amino acid residues in all 3 proteins.

No region was unique to the Seattle cystic fibrosis isolates (n = 3). The soft tissue isolate CRM-0020 from Brazil harbored several large unique regions, including a previously described broad–host-range IncP-1β plasmid (35) and 3 regions (contigs) of 5 kb, 10.6 kb, and 79 kb of unknown origin encoding almost exclusively hypothetical proteins.

We also searched for polymorphisms associated with macrolide and aminoglycoside resistance. The Papworth cystic fibrosis and Seattle cytic fibrosis outbreak set of strains showed an A2058C/G mutation in 23S rRNA, which conferred macrolide resistance (36) (A2058G in strains 2u and 28c representative of Papworth cluster 2 and the Seattle strains). Strains 19f, 14h, 12c, and 28a, representative of Papworth cluster 1, and Seattle strains shared the A1408G mutations in 16S rRNA, which conferred aminoglycoside resistance (37). None of these mutations were found in the soft tissue outbreak strains CRM-0020 and GO 06 from Brazil or the M18 strain.

In light of the possibility of a common ancestor and/or intercontinental transmission of strains, we identified SNPs in genes commonly used for identification of mycobacteria and an MLST scheme that could be used by clinical laboratories to assess relatedness of newly isolated strains to this global cluster. In the first approach, we retrieved rpoB sequences from the 6 genomes of representative strains of the Papworth cystic fibrosis outbreak and performed partial sequencing of the rpoB gene for selected isolates from the Seattle cystic fibrosis outbreak. We then compared these sequences with those of isolates from the outbreak in Brazil and unrelated clinical isolates comprising M. abscessus subsp. abscessus, massiliense, and bolletii, as well as other rapidly growing mycobacteria.

By using the rpoB gene MAB_3869c from the M. abscessus subsp. abscessus type strain as a reference (Table 2) described in the BRA-00 outbreak isolates from Brazil (32,33), we showed that Seattle (n = 4) and Papworth (n = 6) cystic fibrosis isolates carried the 2 rpoB SNPs (C→T at position 2569 and T→C at position 2760. However, none of the M. abscessus subsp. abscessus or subsp. bolletii or other rapidly growing mycobacterial isolates outside the M. abscessus group harbored this 2-SNP rpoB signature (Table 2). The second SNP (T→C substitution at position 2760) was present in several strains, but the combination of both rpoB SNPs (C→T at position 2569 and T→C at position 2760) was not present. Most of the 26 M. abscessus subsp. massiliense strains not related to outbreaks tested did not harbor this 2-SNP rpoB signature. However, 4 strains harbored this signature (Table 2) (29,38).

Multiple alignment of rpoB sequences among available M. abscessus subsp. massiliense genomes showed the absence of the 2-SNP rpoB signature in most strains. However, both SNPs were present in 1 strain not related to an outbreak (1S-151–0930) (Table 2).

Multiple alignment of secA1 sequences among available M. abscessus subsp. massiliense genomes showed a G→T substitution at position 820 (by using the secA1 gene MAB_3580c from the M. abscessus subsp. abscessus type strain) shared by the Papworth and Seattle cystic fibrosis outbreak strains but not by the soft tissue outbreak strains from Brazil or additional unrelated strains. Further analysis of secA1 sequences from 12 M. abscessus subsp. massiliense identified by multitarget sequencing and PCR-based typing (29,38) showed a G→T substitution at position 820 in 2 strains unrelated to the outbreak (Table 2). Those 2 strains were included among the 4 strains that had the 2-SNP rpoB signature. Although the SNPs described for rpoB and secA1 were not 100% specific markers for the outbreak strains, these SNPs could be used for first-level identification of newly isolated strains as possibly being related to cystic fibrosis clusters or soft tissue outbreak strains from Brazil to be confirmed by a second assay.

We also developed a simple MLST protocol that could be used as a second confirmatory assay. Alleles for each of 13 housekeeping genes (cya, gdhA, argH, glpK, gnd, murC, pgm, pknA, pta, pur, rpoB, hsp65, and secA1) were extracted and concatenated for each M. abscessus subsp. massiliense genome (Table 1), and the panel of concatenated sequences was used for construction of a neighbor-joining phylogenetic tree by using MEGA software. The Seattle and Papworth cystic fibrosis outbreak strains grouped together in the tree with cystic fibrosis strain 47J26 and isolate M18 from Malaysia (Figure 3). Thus, partial sequencing of rpoB and secA1 gens, followed by 13-target MLST analysis, could be used to rule out isolates as belonging to these 2 cystic fibrosis clusters.

Recent whole-genome data show deep genetic separation of 3 subspecies, ruling against grouping M. massiliense and M. bolletii under M. abscessus subsp. bolletii.