Genome Sequencing of Mycobacterium abscessus Isolates from Patients in the United States and Comparisons to Globally Diverse Clinical Strains

Abstract

Nontuberculous mycobacterial infections caused by Mycobacterium abscessus are responsible for a range of disease manifestations from pulmonary to skin infections and are notoriously difficult to treat, due to innate resistance to many antibiotics. Previous population studies of clinical M. abscessus isolates utilized multilocus sequence typing or pulsed-field gel electrophoresis, but high-resolution examinations of genetic diversity at the whole-genome level have not been well characterized, particularly among clinical isolates derived in the United States. We performed whole-genome sequencing of 11 clinical M. abscessus isolates derived from eight U.S. patients with pulmonary nontuberculous mycobacterial infections, compared them to 30 globally diverse clinical isolates, and investigated intrapatient genomic diversity and evolution. Phylogenomic analyses revealed a cluster of closely related U.S. and Western European M. abscessus subsp. abscessus isolates that are genetically distinct from other European isolates and all Asian isolates. Large-scale variation analyses suggested genome content differences of 0.3 to 8.3%, relative to the reference strain ATCC 19977^T. Longitudinally sampled isolates showed very few single-nucleotide polymorphisms and correlated genomic deletion patterns, suggesting homogeneous infection populations. Our study explores the genomic diversity of clinical M. abscessus strains from multiple continents and provides insight into the genome plasticity of an opportunistic pathogen.

INTRODUCTION

Nontuberculous mycobacteria (NTM) represent a diverse group of environmental and pathogenic bacteria that are increasing in clinical prevalence in the United States (1) and other countries (2, 3). NTM are thought to be acquired primarily through environmental exposure (4), as they reside in water, biofilms, and soil environments (5, 6), although a few recent studies provide evidence of possible person-to person transmission among individuals with cystic fibrosis (CF) (7, 8). Mycobacterium abscessus is the second most clinically prevalent NTM species in pulmonary NTM infections (1, 9, 10), with Mycobacterium avium complex (MAC) being the most prevalent. M. abscessus infections are challenging to treat because they are innately resistant to many antimicrobials, including some that are effective against Mycobacterium tuberculosis and other mycobacteria (11–13). Acquired antibiotic resistance has been observed in some M. abscessus strains, with mechanisms of resistance ranging from mutational resistance to aminoglycosides (14) to mutational (15) and inducible (16) resistance to macrolides.

The current taxonomy of M. abscessus recognizes two subspecies, M. abscessus subsp. abscessus and M. abscessus subsp. bolletii (17), although recent population and comparative genomic studies support the three previously recognized subspecies, i.e., M. abscessus subsp. abscessus, M. abscessus subsp. massiliense, and M. abscessus subsp. bolletii (18–21). This delineation is important because clinical studies suggest differing treatment outcomes for patients infected with M. abscessus subsp. abscessus versus M. abscessus subsp. massiliense (22, 23). The increased recognition of M. abscessus as an emerging pathogen has reinforced the need to better understand the population structure of clinical M. abscessus at the subspecies and genome levels. Previous population studies of clinical M. abscessus used primarily multilocus sequencing typing (MLST) or pulsed-field gel electrophoresis (PFGE), but these methods have limited resolution for within-subspecies strain typing. We hypothesized that a high-resolution analysis of globally diverse clinical M. abscessus isolates could reveal distinct pathogen lineages and might uncover genetic components relevant to mycobacterial disease.

Whole-genome sequencing (WGS) and phylogenomic analysis are important methods for studying population structure and genomic evolution of bacterial pathogens, as they enable high-resolution analysis of genetic variants ranging from single-nucleotide polymorphisms (SNPs) to large-scale deletions. Analogous to MLST, core genome analyses compare genomic positions that are shared among all isolates, to create phylogenies of isolate populations. For example, a recent retrospective study of M. abscessus isolates derived from cystic fibrosis (CF) patients in the United Kingdom utilized WGS and core genome analyses to compare patient isolates and provided the first suggestive evidence of person-to-person transmission of M. abscessus subsp. massiliense (8). Our subsequent studies using WGS and core genome analysis revealed high levels of genetic relatedness between a subgroup of United Kingdom CF clinic isolates and M. abscessus subsp. massiliense strains from an epidemic of skin infections in Brazil (24) and an M. abscessus outbreak in Seattle, Washington (25).

In addition to phylogenomic comparisons of the core genome, WGS enables the identification of large-scale insertions and deletions (often referred to as the accessory genome). These genomic regions can include phage genes and plasmid elements and may represent genomic islands acquired through horizontal gene transfer (26). Their loss or gain can be attributed to environmental adaptations that can influence antibiotic resistance, virulence, or host range restrictions (26, 27). A recent study of M. abscessus subsp. massiliense isolates, for example, found that the absence of a cluster of glycopeptidolipid genes within a 24.8-kb genomic deletion conferred rough versus smooth colony morphology (28). Genome content variations in lung-associated opportunistic pathogens range from 22% in Staphylococcus aureus (29) to 10% in Pseudomonas aeruginosa (30) but are currently uncharacterized for M. abscessus clinical isolates. Knowledge of genome content variations among M. abscessus isolates could provide clues to genetic mechanisms contributing to virulence, antibiotic resistance, and transmission and may suggest biomarkers of diagnostic utility.

Here we report the genome sequences and genomic features of 11 Mycobacterium abscessus isolates derived from eight patients with pulmonary NTM disease in the United States. The sequenced strains were evaluated for in vitro susceptibility to 19 drugs. Using the complete genome of the M. abscessus subsp. abscessus type strain ATCC 19977^T as a reference, we identified core genome SNPs in the United States-derived isolates as well as isolates from the United Kingdom, France, Brazil, Malaysia, and China, to evaluate the global genetic population structure of clinical M. abscessus. In our analysis, we detected large-scale genomic polymorphisms, reflecting gene content variations and genome plasticity, among closely and distantly related isolates. Lastly, we performed longitudinal genomic comparisons of isolates derived from individual patients, to examine infection homogeneity and genetic mutations that may arise during M. abscessus lung infections.

MATERIALS AND METHODS

Bacterial DNA isolation and PCR genotyping.

M. abscessus isolates were grown in Middlebrook 7H9 liquid medium for 5 days, and genomic DNA was extracted using standard protocols (31). Gene-specific primers were used to amplify a partial segment of the RNA polymerase beta subunit (rpoB) gene (32) and the hsp65 gene (19) for DNA sequencing using an ABI 3730xL genetic analyzer. The erythromycin ribosomal methylase gene erm(41) target was amplified using specific primers (16), and gel electrophoresis was carried out using the Lonza FlashGel system, to determine the absence or presence of a 273-bp deletion within the erm(41) gene.

In vitro drug susceptibility testing.

National Jewish Health (NJH) isolates were evaluated for drug susceptibility to a panel of 19 drugs using the microdilution method (33), and results are reported as MICs. NJH2 and NJH3 were not included in the testing because they did not grow in the testing medium.

Genome resequencing.

Approximately 1 μg of total genomic DNA was used for library preparation for SOLiD fragment sequencing, according to the manufacturer's protocol. Fragments were size selected between 100 and 250 bp, and genome sequencing was performed using the SOLiD 5500 platform (Life Technologies, Carlsbad, CA); 75-bp single-end reads were produced, and sequence reads were filtered using the default purity filter threshold in the Lifescope genome analysis software (Life Technologies). Only purity-filtered reads were used for downstream analyses.

Single-nucleotide polymorphism detection and annotation.

For strains with next-generation resequencing data, including all NJH isolates and United Kingdom isolates reported by Bryant et al. (8) (see Table 2), sequence reads were mapped to the M. abscessus reference genome sequence (34), which includes a 5,067,172-bp chromosome and a 23,319-bp plasmid, using the Lifescope genome analysis software (Life Technologies, Carlsbad, CA) whole-genome resequencing pipeline for SOLiD data or the Bowtie mapping algorithm for Illumina data (35). SNP and reference base calls were identified with the pileup program in SAMtools version 0.1.7 (36). SNPs were filtered using a custom Perl script and the following parameters: SNP quality score of >20, minimum of 10× read depth, >50% of base calls supporting the variant base, and <25% of variant calls occurring at the beginning or end of fragment reads.

TABLE 2.

Globally diverse clinical M. abscessus isolates used for phylogenomic analyses

Subspeciesa	Isolate	Isolation date	Country of origin	Sequence accession no.b	Reference
MAB	ATCC 19977T	1957	France	NC_010397.1	34
MAB	6G0728	2003	USA	AKUS00000000	Unpublished
MAB	6G0125	2010	USA	AKUF00000000	Unpublished
MAB	3A0119R	2010	USA	AKUX01000000	Unpublished
MAB	3A0122R	2002	USA	AKUY00000000	Unpublished
MAB	V06705	2005	France	AUMY00000000	48
MAB	6a	2007	UK	ERR114968	8
MAB	7a	October 2007	UK	ERR115041	8
MAB	7b	October 2007	UK	ERR114965	8
MAB	7c	October 2007	UK	ERR114966	8
MAB	9a	November 2007	UK	ERR114986	8
MAB	9b	November 2007	UK	ERR115102	8
MAB	9c	December 2007	UK	ERR114974	8
MAB	9d	December 2007	UK	ERR115104	8
MAB	1b	2008	UK	ERR119107	8
MAB	23a	2010	UK	ERR115039	8
MAB	M152	NRc	Malaysia	AKVT00000000	49
MAB	M94	NR	Malaysia	AJGG00000000	50
MAB	M93	2010	Malaysia	AJGF00000000	51
MAB	9808	1998	China	ANAR00000000	Unpublished
MBOL	BD	2004	France	AHAS00000000	52
MBOL	M24	NR	Malaysia	AJLY00000000	53
MMAS	5S0304	1998	USA	AKTX00000000	Unpublished
MMAS	4S0116S	2008	USA	AKVE00000000	Unpublished
MMAS	4S0116R	2008	USA	AKVD00000000	Unpublished
MMAS	47J26	2009	UK	AGQU00000000	54
MMAS	CRM0020	2006	Brazil	ATFQ00000000	55
MMAS	19m	2010	UK	ERR115051	8
MMAS	M154	2010	Malaysia	AJMA00000000	20
MMAS	CCUG49998T	2004	France	AKVF00000000	56

^aMAB, M. abscessus subsp. abscessus; MBOL, Mycobacterium abscessus subsp. bolletii; MMAS, M. abscessus subsp. massiliense.

^bGenome sequences were obtained from the National Center for Biotechnology Information (NCBI) or the European Nucleotide Archive (ENA).

^cNR, not reported.

For M. abscessus strains with publically available draft genomes (including strains 6G0728, 6G0125, 3A0119R, 3A0122R, V06705, M152, M94, M93, 9808, BD, M24, 5S0304, 4S0116S, 4S0116R, 47J26, CRM0020, M154, and CCUG49998T) (see Table 2), we performed multigenome alignments of each draft genome to the M. abscessus reference genome sequence (34). Whole-genome alignments and single-nucleotide polymorphism (SNP) identification were performed with the progressiveMauve algorithm in Mauve 2.3.1 (37).

Genotype matrices for core genome comparisons were created for chromosomal positions for which high-quality variant and/or reference calls were available for all isolates. Genomic positions with ambiguous bases and/or missing data were excluded from the analyses. High-confidence SNP sites were annotated as genic or intergenic with a SQLite database and custom DBI-Perl scripts, using the gene annotation provided for the reference genome (34). Genic SNPs were further annotated for amino acid-level changes using ANNOVAR software (38).

Phylogenomic analysis.

To elucidate the phylogeny of all 41 isolates, high-confidence genotype data at 2,479 core genomic positions were concatenated into a FASTA sequence for each strain using a custom Perl script. A neighbor-joining (NJ) phylogeny was estimated using the observed differences between the concatenated sequences, with 1,000 bootstrap replicates, in SeaView 4.4.2 (39). To further resolve the clade of 18 U.S./European isolates, a high-resolution genotype data set of 128,074 core genomic positions was generated and a NJ phylogeny was estimated as described above.

Large-scale polymorphism analyses.

For 11 NJH isolates and 9 United Kingdom isolates with next-generation sequence data, sequence reads were mapped to the complete genome of the type strain ATCC 19977^T. Read counts were normalized for GC content bias using a modified version of a previous method (40). Sequence coverage values were estimated with a custom Perl script that counts all uniquely mapped reads in sliding, nonoverlapping, 1-kb windows across the entire ATCC 19977^T chromosome (5,068 total windows). Read counts for all 18 strains were converted to z-scores, and the genome-wide normalized read counts were clustered by hierarchical clustering using a Pearson correlation distance metric and average linkage with the Multiple Experiment Viewer (MeV) Java package (41). Pearson correlations of genome-wide deletion patterns of same-patient isolates were performed with the R statistical package (42).

Putative large-scale deletions were identified as contiguous regions greater than 30 kb with z-scores of less than −2.0. To determine the sequence homology of six identified deletion regions, genomic sequences were extracted from the reference genome with a custom Perl script and were queried against the NCBI nonredundant nucleotide database using the Blastn algorithm. BLAST results were filtered and sequence homology was defined with the following criteria: E value of 0.0, ≥70% sequence identity, and ≥30% query coverage.

Nucleotide sequence accession numbers.

Sequences were submitted to GenBank under accession numbers SRX641283 (NJH2), SRX641284 (NJH3), SRX641291 (NJH4), SRX641292 (NJH5), SRX641293 (NJH6), SRX641294 (NJH7), SRX641295 (NJH9), SRX339602 (NJH8), SRX641281 (NJH1), SRX641302 (NJH10), and SRX339603 (NJH11).

RESULTS AND DISCUSSION

Clinical and microbiological attributes of patients and isolates.

M. abscessus isolates examined in this study were obtained from sputum or bronchoalveolar lavage (BAL) fluid samples from eight patients referred to National Jewish Health (NJH), in Denver, Colorado, for management of chronic M. abscessus-related pulmonary infections between 2009 and 2011 (Table 1). All except one patient had a ≥2-year history of M. abscessus pulmonary infection, and all patients had received prolonged treatment with multiple antimycobacterial drugs, including macrolides and aminoglycosides. Isolates NJH1 to NJH4 were sampled longitudinally from the same patient at three time points during a 6-month period; isolates NJH2 and NJH3 were individual colonies collected at the second time point. The primary residences of all patients were in the northern or eastern United States, with the exception of one individual from Puerto Rico (isolate NJH7). Patients were primarily female (6/8 patients) and greater than 60 years of age (7/8 patients), and the majority of patients (6/7 patients) were smear negative at the time of sputum collection. Four of the patients had adult cystic fibrosis (isolates NJH1, NJH5, NJH8, and NJH9), and two were carriers of a CFTR polymorphic allele (isolates NJH7 and NJH11).

TABLE 1.

Clinical Mycobacterium abscessus isolates sequenced in this study and patient information

Patient	Isolate	Sample collection date	Species identificationa	Sourceb	Sexc	AFB smear resultd	Age (yr)	Cystic fibrosis
1	NJH1	September 2010	MAB	Sputum	F	−	66	Yes
1	NJH2	October 2010	MAB	Sputum	F	−	66	Yes
1	NJH3	October 2010	MAB	Sputum	F	−	66	Yes
1	NJH4	March 2011	MAB	Sputum	F	−	66	Yes
2	NJH5	2009	MAB	Sputum	F	−	62	Yes
3	NJH6	2011	MAB	Sputum	F	−	75	NR
4	NJH7	2010	MAB	BAL fluid	F	+	71	Carrier
5	NJH8	2011	MAB	Sputum	M	NR	41	Yes
6	NJH9	2011	MAB	BAL fluid	F	−	66	Yes
7	NJH10	2010	MAB	Sputum	F	−	72	No
8	NJH11	2009	MMAS	Sputum	M	−	79	Carrier

^aIsolates were identified as either M. abscessus subsp. abscessus (MAB) or M. abscessus subsp. massiliense (MMAS) by sequencing portions of the rpoB and hsp65 genes and by amplification of the erm(41) gene.

^bIsolates were derived from sputum or bronchoalveolar lavage (BAL) fluid samples.

^cF, female; M, male.

^dAFB, acid-fast bacillus; NR, not reported.

All isolates were initially identified to the subspecies level by Sanger sequencing of the rpoB and hsp65 genes. Based on sequence homology to type strains, isolates NJH1 to NJH10 were identified as M. abscessus subsp. abscessus, and isolate NJH11 was identified as M. abscessus subsp. massiliense. All isolates were also evaluated for the erm(41) deletion (16). In agreement with the rpoB and hsp65 results, isolates NJH1 to NJH10 exhibited a full-length erm(41) amplicon consistent with M. abscessus subsp. abscessus, while NJH11 showed a smaller, deletion-containing amplicon consistent with M. abscessus subsp. massiliense.

Genomic sequencing and core genome relationships among diverse clinical isolates.

Genomic sequence reads from the pulmonary-derived NJH clinical isolates were mapped to the reference genome of the M. abscessus type strain ATCC 19977^T (34), which was initially isolated from a knee infection with abscess-like subcutaneous lesions (43). The NJH isolates were deep-sequenced at levels of 97- to 214-fold coverage (see Table S1 in the supplemental material). Sequence reads from only one isolate, NJH9, showed substantial read mapping to a mercury resistance plasmid present in the reference strain, indicating that this plasmid is not widespread in the isolates studied.

To investigate the phylogenomic relationships of the NJH M. abscessus isolates, compared to globally diverse clinical isolates, we assembled 30 publically available M. abscessus genomes from the United States, the United Kingdom, France, China, Malaysia, and Brazil for our analysis (Table 2), including 11 representative M. abscessus isolates from a recent retrospective study of CF patients in the United Kingdom (8). First, we performed a neighbor-joining (NJ) phylogenetic reconstruction for 2,479 core genome positions (2,320 genic and 159 intergenic) that were shared across all 41 isolates (Fig. 1A). The resulting phylogeny supports the three previously recognized M. abscessus subspecies as monophyletic, consistent with population studies utilizing MLST genotyping (19, 21), and the NJH isolates grouped into expected clades consistent with subspecies identifications based on rpoB, hsp65, and erm(41) findings. Our phylogeny is also consistent with previously reported, independent, phylogenomic comparisons (8, 20, 24, 25, 44).

FIG 1.

Phylogenomic comparison of the National Jewish Health (NJH) M. abscessus clinical isolates from the United States, compared with global strains. (A) Genotype data at 2,479 core genome positions, relative to the M. abscessus subsp. abscessus ATCC 19977^T reference genome, were used to estimate the phylogeny among 41 M. abscessus isolates using the neighbor-joining algorithm. The phylogenomic tree supports previously recognized subspecies as monophyletic groups. Gray bars and asterisks, isolates acquired from the same patients. (B) Higher-resolution phylogeny of the 18-isolate U.S./European M. abscessus subsp. abscessus cluster was created using genotype information at 128,074 core genome positions. Isolation dates are included. Only 320 (0.25%) of 128,074 core genome sites vary among these isolates.

Within the subspecies of M. abscessus subsp. abscessus included in our analysis, we observed two major subclades (Fig. 1A). The basal subclade included all Asian isolates, one NJH isolate (NJH8), and six isolates from three CF patients in the United Kingdom. This clade was genetically distinct from a cluster of 18 isolates exclusively from the United States or Europe, including the ATCC 19977^T type strain (43), nearly all of the NJH isolates, four additional U.S. isolates, one French isolate, and four isolates from the United Kingdom. In this analysis, the 18 clustered isolates were distinguished by only 9 of 2,479 core genome SNP sites (0.36%). An additional 198 SNP sites discriminated the adjacent Puerto Rican isolate (NJH7). These results suggest that, among the M. abscessus subsp. abscessus strains analyzed, the Asian isolates are divergent from most U.S. isolates, while the United Kingdom isolates are divided between the two genetic subgroups. Further sampling and future studies are needed to validate these regional subgroups.

The lack of SNP variations in the U.S./Europe M. abscessus subsp. abscessus cluster is intriguing, given the temporal and geographic disparities among the isolates. To further resolve the 18 clustered isolates and to test for temporal or geographic groupings, a high-resolution phylogeny was constructed using genotype information at 128,074 core genome nucleotide positions (Fig. 1B). A total of 320 SNPs (320/128,074 positions = 0.25%) were observed in this data set, similar to the proportion of variations observed in the initial phylogeny. The majority of SNP variations occurred in the French isolate, V06705, which was the most divergent in the cluster. Overall, the branching patterns did not distinguish temporal or geographic groups, suggesting independent acquisitions of the isolates. Therefore, the high genetic relatedness may represent a dominant clinical M. abscessus subsp. abscessus subgroup or may reflect similarities in the isolates' geographic or environmental niches.

In vitro drug susceptibility results and variants associated with resistance.

Nine of 11 NJH clinical isolates were evaluated for in vitro drug susceptibility to 19 antimicrobial drugs. These drugs span several drug classes, and the results are reported as MICs (Table 3). Most isolates exhibited resistance to imipenem, amoxicillin-clavulanic acid (Augmentin), two of four aminoglycosides (tobramycin and gentamicin), four generations of cephalosporins, two of three tetracyclines (doxycycline and minocycline), and two fluoroquinolones (ciprofloxacin and moxifloxacin), confirming the innate resistance of M. abscessus to several drug classes (11, 13). Drugs that showed in vitro activity against the M. abscessus strains included the aminoglycosides amikacin and kanamycin and the macrolides clarithromycin and azithromycin, which are commonly used to treat pulmonary M. abscessus infections (12), as well as the less commonly prescribed tigecycline.

TABLE 3.

Drug susceptibility of NJH M. abscessus clinical isolates to a 20-drug panel, reported as MICs

Drug	MIC (μg/ml)									No. of resistant isolates/total no. of isolatesa
	NJH1	NJH4	NJH5	NJH6	NJH7	NJH8	NJH9	NJH10	NJH11
Amikacin	16	≤8	≤8	≤8	>64	32	16	≤8	32	1/9
Kanamycin	16	≤8	≤8	≤8	>64	32	16	≤8	16	1/9
Tobramycin	16	16	8	8	>16	16	8	16	>16	6/9
Gentamicin	>16	8	4	8	>16	>16	16	8	>16	5/9
Imipenem	16	>16	>16	8	8	8	16	>16	>16	6/9
Cefoxitin	64	64	64	32	64	64	32	64	64	9/9
Ceftriaxone	>64	>64	>64	>64	>64	>64	>64	>64	>64	9/9
Cefotaxime	>64	>64	>64	>64	>64	>64	>64	>64	>64	9/9
Cefepime	32	>32	>32	>32	>32	32	>32	>32	>32	9/9
Clarithromycin	0.5	0.5	1	2	1	1	1	>32	0.5	1/9
Azithromycin	≤16	≤16	32	32	32	32	≤16	>256	≤16	1/9
Amoxicillin/clavulanic acid	>32/16	>32/16	>32/16	>32/16	>32/16	>32/16	>32/16	>32/16	>32/16	9/9
Ciprofloxacin	8	>8	8	4	>8	>8	8	>8	8	9/9
Moxifloxacin	>4	>4	>4	>4	>4	>4	>4	>4	>4	9/9
Trimethoprim/sulfa	>4/76	>4/76	>4/76	>4/76	>4/76	>4/76	>4/76	>4/76	>4/76	9/9
Doxycycline	16	>16	>16	>16	>16	>16	>16	>16	>16	9/9
Minocycline	8	>8	>8	>8	>8	>8	>8	>8	>8	9/9
Tigecycline	1	0.5	≤0.25	2	1	1	0.5	2	1	0/9
Linezolid	>16	16	>16	16	>16	>16	16	>16	>16	9/9

^aDetermination of isolates with resistant phenotypes was based on previously defined MIC thresholds specific to each drug.

Among the NJH isolates tested, only two exhibited different susceptibility phenotypes. NJH7 was the only isolate with observed in vitro resistance to amikacin and kanamycin. Based on our analysis, this resistance can be explained by the presence of a 16S rRNA mutation that was previously associated with aminoglycoside resistance in M. abscessus and Mycobacterium chelonae isolates and corresponds to the A1408G mutation in Escherichia coli (14). NJH10 was the only isolate with observed resistance to clarithromycin and azithromycin. This isolate does not have a resistance-conferring mutation in the 23S rRNA (15) and therefore may contain a novel mechanism of resistance.

Large-scale polymorphic genomic regions in M. abscessus clinical isolates.

Given the high levels of genetic relatedness in the core genomes of the clustered U.S./European isolates, we also investigated large-scale genomic variations among a subset of clustered and nonclustered isolates. We analyzed sequence coverage in 1-kb windows relative to the ATCC 19977^T reference genome and identified 1-kb windows with substantially low coverage (<2-fold), indicative of deletions or highly divergent regions of the genome. Among the 19 isolates analyzed, we observed a range of 0.3 to 8.3% of the reference genome that was not represented in the clinical isolates (Table 4). The least divergence was observed among clustered isolates (0.3 to 2.8%), and greater divergence was observed among nonclustered isolates (5.9 to 7.1%). As expected, the most divergence was observed in the M. abscessus subsp. massiliense isolate (NJH11; 8.3%). This diversity is similar to genome content differences observed among clinical M. tuberculosis isolates (45) and illustrates the genome plasticity and diversity among U.S. and European M. abscessus isolates.

TABLE 4.

Large sequence polymorphisms between 19 clinical M. abscessus isolates and the reference type strain, ATCC 19977^T

Isolate	Country of origin	No. of genomic windows with <2-fold coveragea (% of genome)	Genetic groupb
NJH1	USA	141 (2.8)	MAB clustered
NJH2	USA	142 (2.8)	MAB clustered
NJH3	USA	142 (2.8)	MAB clustered
NJH4	USA	142 (2.8)	MAB clustered
NJH5	USA	82 (1.6)	MAB clustered
NJH6	USA	15 (0.3)	MAB clustered
NJH9	USA	25 (0.5)	MAB clustered
NJH10	USA	16 (0.3)	MAB clustered
7a	UK	85 (1.7)	MAB clustered
7b	UK	85 (1.7)	MAB clustered
7c	UK	85 (1.7)	MAB clustered
NJH7	USA	71 (1.4)	MAB clustered
NJH8	USA	297 (5.9)	MAB nonclustered
1b	UK	317 (6.3)	MAB nonclustered
9b	UK	360 (7.1)	MAB nonclustered
9c	UK	359 (7.1)	MAB nonclustered
9d	UK	359 (7.1)	MAB nonclustered
9a	UK	360 (7.1)	MAB nonclustered
NJH11	USA	419 (8.3)	MMAS

^aGenomic windows with significantly low sequence read coverage (less than 2-fold), compared to the reference strain.

^bThe three genetic groups identified in our phylogenomic analysis were clustered M. abscessus subsp. abscessus (MAB) isolates, nonclustered M. abscessus subsp. abscessus isolates, and M. abscessus subsp. massiliense (MMAS) isolates.

To test whether the clustered isolates shared the same deletions, we performed a cluster analysis of genome-wide coverage patterns (Fig. 2). The clustering results were similar to the initial core genome phylogeny (Fig. 1A), as the U.S./European clustered isolates grouped into a major clade distinct from nonclustered isolates. One exception was the Puerto Rican isolate (NJH7) which clustered with the U.S. isolates based on deletion patterns but was somewhat divergent from the U.S. and United Kingdom isolates in the core genome phylogeny (Fig. 1A). In contrast to the high-resolution phylogeny (Fig. 1B), the U.S./European clustered isolates grouped together largely by geographic region. This finding, along with the grouping of NJH7 with other North American isolates, suggests that evolution of the accessory genome may be driven by events within local microenvironments.

FIG 2.

Large-scale genomic variations among M. abscessus genomes. Large-scale genomic variations, including deletions and duplications, were identified by sequence read mapping and coverage assessment along a reference genome. Chromosome coordinates of three regions of the reference genome are indicated on the x axis. Next-generation sequence read counts were assessed along nonoverlapping 1-kb sliding windows of the reference genome (5,068 total windows). Contiguous windows with z-scores (y axes) less than −2.0 indicate putative genomic deletions or highly divergent regions, while contiguous stretches of high z-scores, greater than 2.0, indicate putative genomic duplications. The dendrogram represents the result of average linkage hierarchical clustering of genome-wide coverage values, using a Pearson correlation metric. Arrows A to F, large-scale genomic differences (>30 kb) among the clinical M. abscessus isolates.

Further analyses of six large-scale deletions (>30 kb) (Fig. 2, arrows A to F), relative to the reference M. abscessus genome, revealed potential mechanisms of accessory genome evolution among clinical M. abscessus strains (see Table S2 in the supplemental material). Regions A and F, which were deleted in all nonclustered isolates, include operons for copper ion and phosphate ion transport, respectively. Both genomic regions contain sequences that are homologous to plasmids and transposons, providing evidence for recombination or horizontal gene transfer (26). Region C, which was absent in 14 of 19 isolates, shows sequence homology only to mycobacteriophages, consistent with previous annotation as a prophage (34). Two U.S. isolates (NJH7 and NJH10) had excessive sequence coverage in this region, suggesting multiple genomic copies of prophage sequences. Regions B, D, and E show no evidence of recombination or homology to phage DNA. Region E, which was deleted in nonclustered isolates and M. abscessus subsp. massiliense (NJH11), includes a cassette of biphenyl and aromatic hydrocarbon metabolic enzymes that functionally enable degradation of environmental chemicals (46) in some cases. Overall, the results suggest that gene content variations in clinical M. abscessus are due, in part, to recombination and/or metabolic adaptations.

Genomic comparisons of longitudinally sampled isolates from individual patients.

To evaluate same-patient strain variations, we compared genomic information derived from three sets of isolates from individual patients. From the United Kingdom CF M. abscessus study (8), we studied isolates from patients 7 and 9. Isolates 7a, 7b, and 7c were collected at a single time point. Isolates 9a and 9b were collected at an initial time point during treatment, and isolates 9c and 9d were collected 1 month later. From NJH, four isolates (NJH1 to NJH4) were longitudinally isolated from a single patient at three time points during a 6-month period, including two colonies (NJH2 and NJH3) that were collected at the second time point.

In the previous core genome phylogenies (Fig. 1) with limited numbers of genomic positions (2,479 bp and 128,074 bp), all three sets of isolates showed identical genotypes among themselves within each set. Therefore, we performed in-depth core genome analyses for each patient isolate group, to explore infection homogeneity and the potential for genomic mutations during infection and treatment. Using all high-quality base calls within each group of patient isolates, we found 4 SNPs among the three patient 7 isolates across 4,971,231 genomic positions (0.00008%), 12 SNPs among the four patient 9 isolates across 4,503,047 positions (0.0002%), and 10 SNPs among the four NJH patient 1 isolates across 4,587,198 positions (0.0002%). In contrast to SNPs in the overall phylogeny (Fig. 1A), in which the minority of coding SNPs were nonsynonymous (20.9%) versus synonymous (79.5%), the majority of SNPs in the longitudinal groups were nonsynonymous changes (55.6% to 87.5%), suggesting positive selective pressures as possible driving forces of genetic mutation. Examples of genes with nonsynonymous SNPs observed among NJH1 patient isolates include an amino acid permease family protein (NCBI gene locus tag MAB_0950c), a probable DNase TatD (NCBI gene locus tag MAB_1129), a putative membrane protein MmpL (NCBI gene locus tag MAB_2303), sulfate adenylate transferase CysD (NCBI gene locus tag MAB_4181), and arsenic-transport integral membrane protein ArsC (NCBI gene locus tag MAB_4863).

To test for large-scale variations among same-patient isolates, correlations of genome-wide sequence coverage patterns were performed for all pairwise combinations of isolates within and between patient groups (Fig. 3). Results demonstrated that all same-patient isolate pairs were highly correlated, compared to different-patient isolate pairs, and we saw no evidence of isolate-specific deletions within patient groups. These data, along with minimal SNP variations, suggest homogeneous populations in the patients studied, consistent with previous longitudinal studies of M. abscessus (8, 47). The high-resolution comparisons enabled by WGS provide evidence for genetic mutations and possible adaptation within the host.

FIG 3.

Correlations of genome-wide coverage patterns among 11 M. abscessus isolates collected from three patients. Correlations of genome-wide sequence coverage patterns were performed for all pairwise combinations of 11 isolates (n = 55). Each pair was binned as between-patient isolates (n = 40) or within-patient isolates (n = 15). Isolates obtained from the same patient had much higher correlations than isolates from different patients, suggesting genetically homogeneous infecting populations in individual patients.

Conclusions.

Our study examined the genome sequences of 11 clinically relevant M. abscessus isolates from eight U.S. patients with persistent pulmonary infections that exhibited in vitro resistance to several classes of antimicrobials. Moreover, it provides a comprehensive phylogenomic assessment of U.S. M. abscessus isolates in comparison with globally diverse clinical isolates. Our core genome comparisons, across 41 M. abscessus isolates, revealed three genetic subgroups, consistent with previous subspecies classifications. We observed that most of the NJH isolates examined (9/11 isolates) belonged to a cluster of M. abscessus subsp. abscessus isolates, from primarily U.S. or Western European origins, with low core genomic diversity. This cluster was genetically distinct from most United Kingdom isolates and all Asian isolates examined in the study. Future studies with expanded populations of both clinical and environmental isolates could reveal whether this cluster represents a clinical subtype of M. abscessus or an ecological subgroup related through a shared environmental niche.

Despite the core genome similarities observed among clustered U.S. and European isolates, we found substantial accessory genome variations in the form of multiple large-scale deletions. This illustrates the resolving power of WGS in detecting not only SNPs but also the presence or absence of genomic islands, which can confer clinically important traits associated with virulence and antibiotic resistance. Overall, we observed genome content variations of 0.3 to 7.1% within subspecies (M. abscessus subsp. abscessus) and 8.5% between subspecies (M. abscessus subsp. abscessus versus M. abscessus subsp. massiliense), including one large deletion that was shared across subspecies (Fig. 2, arrow F). Functional annotation analyses of the deletion regions suggested that genome content variations in M. abscessus could be attributed to recombination and/or metabolic adaptations.

Analyses of longitudinally collected samples from individual patients showed low levels of SNP variations and no evidence of large-scale variations, suggesting homogeneous populations in the hosts studied. SNPs in the longitudinal studies were enriched for nonsynonymous mutations, compared to the overall background population, suggesting that pathogen genome evolution in the host may be under positive selective pressure from prolonged exposure to the host immune system, antibiotic treatment, or other selective pressures. Further WGS studies will be needed to explore long-term in vivo mutation rates in the host and potential implications for antibiotic targets.