Mycobacterium avium Complex Diversity within Lung Disease, as Revealed by Whole-Genome Sequencing

To the Editor:

Lung disease resulting from Mycobacterium avium complex (MAC) is a vexing and increasing clinical problem. MAC is generally thought to represent either M. avium or Mycobacterium intracellulare; however, there are more than 10 species within the complex (1). Speciation is almost never performed in the clinical microbiology laboratory, as commercial probes resolve only to the complex level. We examined the species and subtypes of longitudinal MAC isolates from 35 patients with MAC lung disease via whole-genome sequencing (WGS). We hypothesized that patients with relapse after treatment would be more likely to exhibit different strains, reflecting reinfection, compared with patients who were being observed off antibiotic treatment or who were still receiving therapy.

Results

From January 1, 2010, to June 30, 2017, there were 37,521 Mycobacterium cultures performed at the University of Virginia Clinical Laboratory, and 617 grew MAC. For longitudinal analysis, we required that patients met American Thoracic Society lung disease criteria per chart review and had at least two isolates available separated in time by at least 30 days. This amounted to 95 longitudinal isolates from 35 patients (range, 2–8 isolates per patient). Clinical and demographic information is shown in Table 1. Most patients had nodular bronchiectatic disease without cavities (77%). During the study period, 11 patients were receiving macrolide-containing antibiotic therapy, 16 patients were being observed off treatment, and eight patients were treated and then relapsed. Each isolate was a sweep of MAC colonies from 7H10 plates that was stored frozen, recultured, and underwent DNA extraction and sequencing, using Illumina Nextera XT libraries with a 2 × 150 paired end protocol (Illumina NextSeq). All work was approved by the University of Virginia Institutional Review Board for Human Subjects Research (#20243).

Table 1.

Clinical and Demographic Data

	All (N = 35)	Off Therapy (n = 16)	On Therapy (n = 11)	Relapse/Reinfection (n = 8)	P Value
Sex, M	17 (49%)	8 (50%)	4 (36%)	5 (63%)	NS
Median age at first culture (range), yr	60 (10–87)	35 (10–80)	66 (34–87)	62 (31–86)	<0.05
Median BMI (IQR)	24 (20–27)	22 (21–28)	21 (20–23)	26 (24–27)	NS
Smoking history					<0.05
Current or former	14 (40%)	2 (13%)	7 (64%)	5 (63%)
Never	21 (60%)	14 (87%)	4 (36%)	3 (37%)
MAC disease type					NS
Cavity present	8 (23%)	3 (19%)	4 (36%)	1 (13%)
Noncavitary-nodular bronchiectasis	27 (77%)	13 (81%)	7 (64%)	7 (87%)
Underlying conditions					<0.05
None	11 (31%)	1 (6%)	6 (55%)	4 (50%)
CF	12 (34%)	9 (56%)	2 (18%)	1 (13%)
COPD	5 (14%)	2 (13%)	3 (27%)	0
Other lung disease	6 (17%)	4 (25%)	0	3 (38%)

Definition of abbreviations: BMI = body mass index; CF = cystic fibrosis; COPD = chronic obstructive pulmonary disease; IQR = interquartile range; MAC = Mycobacterium avium complex; NS = not significant.

Taxonomic classification of WGS data from the isolates was performed using Kraken v1.0 (2), followed by Bayesian reestimation of relative abundances using Bracken (3) (Figure 1). Samples clustered on their Bray-Curtis dissimilarity distances, using Ward’s method, into four groups: one predominantly M. avium (n = 24, 25%), one M. intracellulare (n = 42, 44%), one Mycobacterium chimaera (n = 15, 16%), and a mixed cluster (n = 14, 15%). The isolates in the M. avium cluster were constituted by almost exclusively M. avium reads (>97%), as was the case for the M. chimaera isolates (>94%), whereas isolates within the M. intracellulare cluster were under heterogenous (reads classified to M. intracellulare ranged from 70% to 95%; P < 0.05). Sanger sequencing of the 16S rRNA gene was performed on 25 isolates from 5 species and confirmed 100% concordance with the majority species identified by WGS.

Figure 1.

Microbial community profile of Mycobacterium avium complex isolates, as determined by taxonomic classification using Kracken/Bracken. Each of the 95 isolates is organized vertically and clustered on their Bray-Curtis dissimilarity into four major groups: a predominantly Mycobacterium avium (blue), Mycobacterium intracellulare (orange), Mycobacterium chimaera (green), and a mixed “other” cluster (black), as seen on the dendrogram to the left. Each isolate’s species composition is indicated by the colored horizontal stacked bar. Each of the individual 35 patients is indicated to the right, and their longitudinal isolates are linked by a line, with the first collected isolate denoted by a filled circle and subsequent samples denoted by open circles. Patients not receiving therapy are shown in blue, receiving therapy in green, and relapse patients in red. Patients whose isolates were all identical (single nucleotide variant < 100) are shown encircled. Rx = prescription.

In addition to species identification, we examined the genetic relatedness of isolates from individual patients, using pairwise single nucleotide variant distances. Although some patients (10/35; 29%) produced identical isolates (pairwise distance < 100 single nucleotide variants) over time, the majority (25/35; 71%) produced longitudinal isolates at least 100 single nucleotide variants apart from each other, if not of an entirely different species cluster. This phenomena of MAC isolate diversity was seen not only in those with relapse (5/8 [63%]; red patient group Figure 1) but also in those being observed off of antibiotic treatment (12/16 [75%]; blue patient group) and in those receiving therapy (8/11 [73%]; green patient group). It was also seen in both patients with cavitary (6/8, 75%) and patients with nodular bronchiectatic (19/27, 70%) lung disease. Strain variation was not associated with a longer period between isolates (data not shown). Antibiotic susceptibility testing was performed for clarithromycin, rifampin, ethambutol, amikacin, moxifloxacin, and linezolid by broth microdilution, according to Clinical and Laboratory Standards Institute methodology (4), and different strains were more often of a different antibiogram than identical strains (data not shown). M. avium isolates had higher minimum inhibitory concentrations of rifampin and linezolid than predominant M. intracellulare and M. chimaera isolates (P < 0.05); otherwise, there was no difference in antibiotic resistance between the species clusters.

Discussion

Clinically, we generally assume that longitudinal MAC isolates from patients with nontuberculous mycobacterial lung disease represent the same infection. Using WGS, we found that this is frequently not the case, as there is great diversity within the MAC. The high rate of MAC diversity within a patient over time was robust to multiple bioinformatic approaches, and was supported by the antibiogram data. Moreover, the phenomena were not concentrated in patients who had clinical relapse, as we hypothesized, nor was it associated with antibiotic pressure. We can imagine several explanations for this diversity. First, some patients with MAC lung disease may have polymicrobial infections, and certain subtypes are sampled or preferentially grow over time. Such infection with multiple strains has been noted, particularly with nodular bronchiectatic disease (5). Second, there could be a high force of reinfection from the environment. Third, some variation could be spurious and reflect transient colonization as a result of specimen contamination, for example, from drinking water (6), making it difficult to know which isolates are clinically significant and which are not. Further study is needed to assess these clinical implications and whether certain species or subtypes carry greater prognostic significance.

Next, we found that although the M. avium and M. chimaera isolates were largely pure, at 94–100% of reads at the sequence level, M. intracellulare isolates were more heterogeneous, often with genomic content that mapped substantially to M. avium, M. chimaera, or other MAC species (7). This heterogeneity may suggest, again, that certain M. intracellulare infections are polymicrobial, or simply that available M. intracellulare reference genomes are poorly representative. Notably, most of our MAC isolates were predominantly M. intracellulare (43%), followed by M. avium (25%), then M. chimaera (15%), and then a mixture of others (16%). The high prevalence of M. intracellulare is similar to the studies from Texas (8), but differs from an M. avium predominance seen in other studies (9). A substantial prevalence of M. chimaera has been seen in Illinois (9).

Taken together, there appears to be abundant interspecies and intraspecies MAC diversity in patients over time with MAC lung disease. As a starting point, we believe clinical microbiology laboratories should routinely speciate isolates from patients with MAC lung disease, for instance, using matrix-assisted laser desorption/ionization–time-of-flight mass spectroscopy methods (10), to enable further clarification of this phenomenon.