Comparing genotypes and antibiotic resistance profiles of Mycobacterium abscessus and Mycobacterium massiliense clinical isolates in China

This study aimed to investigate differences in the antimicrobial susceptibility of members of the Mycobacterium abscessus complex (MABC): subsp. massiliense and subsp. abscessus, and to identify associations between strain genotypes and antimicrobial resistance phenotypes. A total of 383 clinical MABC isolates (subsp. abscessus: n = 218, 56.9%; subsp. massiliense: n = 163, 42.6%; subsp. bolletii: n = 2, 0.5%) were characterised using multilocus variable number tandem repeat (VNTR) typing and drug susceptibility testing. Most isolates exhibited susceptibility to amikacin, clarithromycin and azithromycin but resistance to cefoxitin and minocycline was statistically more associated with isolates unclustered by VNTR type. The Simpson's diversity indexes of VNTR typing for M. abscessus and M. massiliense isolates were 0.999 and 0.997, respectively. Genotyping of M. abscessus and M. massiliense isolates by VNTR may provide valuable information for predicting resistance phenotype.


Introduction
Nontuberculous mycobacteria (NTM) have been increasingly recognised as causative agents of opportunistic infections of humans, generally affecting the lungs, skin and soft tissue and visceral or disseminated in severely immunosuppressed individuals [1]. The organisms are generally classified according to their rapid, or slow, rates of growth on culture media [2]. Mycobacterium abscessus is the most frequently encountered rapid grower among NTM causing pulmonary infections, and accounts for 65-80% of cases [3]. The group is characterised by extensive antimicrobial drug resistance which often results in unsatisfactory clinical treatment outcomes [4].
M. abscessus complex (MABC) isolates have been further differentiated on the basis of rpoB sequences into three closely related subspecies, namely: M. abscessus, Mycobacterium massiliense and M. bolletii [5]. In spite of several shared taxonomic characteristics, the subspecies exhibit distinctly different antibiotic resistance phenotypes and treatment outcomes [6]. Differences are particularly pronounced between M. abscessus and M. massiliense, especially in the treatment response rates of patients to clarithromycin (CLA)-based antibiotic therapy, where greater efficacy was evident for pulmonary patients infected with M. massiliense than with M. abscessus [2,7]. Similarly, M. massiliense was reported to have greater susceptibility to doxycycline, although this was based on small numbers of M. abscessus isolates tested [2]. Nevertheless, since these results suggested that subspecies and possibly strain genotypes, may be predictive of antibiotic resistance phenotype, further testing of this concept was conducted in this larger study.
Variable number tandem repeat (VNTR) analysis has been used as a suitable tool for genetic fingerprinting of several bacterial species, including the pathogenic mycobacteria [8]. Indeed, a recent study revealed that specific VNTR genotypes were associated with drug resistance in isolates of both M. tuberculosis and M. intracellulare [9]. However, an association of drug resistance profiles within the M. abscessus complex has not to our knowledge been previously described.
In this study, isolates of the species complex were characterised to the subspecies level and correlations of drug resistant profiles with subspecies analysed, with the ultimate goal of identifying predictive associations between strain genotypes and drug resistance phenotypes.

Clinical isolates and species identification
In total, 383 MABC isolates were recovered from clinical specimens collected between 2014 and 2016 from patients in two specialised TB treatment hospitals in China. All patients providing isolates met the microbiological and clinical criteria of the American Thoracic Society for a diagnosis of NTM pulmonary disease [1]. The study was approved by the Chinese Center for Disease Control and Prevention.
Clinical samples were cultured on Löwenstein-Jensen (L-J) medium after alkali treatment with 4% NaOH. After 4 days of incubation, colonies were scraped from the surface of the medium for the preparation of genomic DNA as previously reported [10]. Identification of subspecies was determined by multilocus sequence typing of genes encoding 16S rRNA, hsp65, rpoB and 16S-23S rRNA internal transcribed spacer sequence loci [11].

Antimicrobial susceptibility testing
The drug susceptibility profiles of isolates were determined by a broth microdilution method according to the guidelines of the Clinical and Laboratory Standards Institute (CLSI) [12]. Briefly, freshly grown colonies from L-J slants were vortexed with glass beads in saline. The resulting suspension was adjusted to 0.5 McFarland and further diluted 1:200 in cation-adjusted Mueller-Hinton broth before being added to 96-well microtitre plates containing serial doubling-dilutions of antimicrobial agents. After incubation at 37°C for 72 h, the minimum inhibitory concentration (MIC) was defined as the lowest concentration of antibiotic to inhibit visible growth of mycobacteria.
VNTR analysis VNTR typing of isolates was performed by the method of Wong et al. [15], but with the exclusion of locus TR20 due to amplification failure; analysis was therefore based on 17 loci. The discriminatory power of VNTR typing was calculated using Simpson's index of diversity as previously described [16]. Clustered strains were defined as isolates with the same genotype, compared with 'unclustered' isolates of different genotypes.

Genotyping and clustering analysis
All isolates, with the exception of the two M. bolletii, were genotyped by VNTR and assigned to phylogenetic clusters. The 218 M. abscessus representatives were differentiated into 10 clusters (comprising two to nine isolates per cluster) and 188 unique genotypes (Fig. 1). Similarly, 119 of the 163 M. massiliense isolates were differentiated into unique genotypes, with the remaining 44 isolates grouped into 16 clusters (two to five isolates per cluster) (Fig. 2). The clustering rate of M. massiliense was 17.2%, compared with 9.2% for M. abscessus (P > 0.05). Analysis of the allelic diversity of VNTR loci between M. abscessus and M. massiliense (Fig. 3) isolates showed that most M. abscessus exhibited allelic diversity, with the exception of locus TR101. Notably, loci TR2 and TR137 were found to be less polymorphic and thus less discriminatory between M. massiliense isolates (h < 0.2), whereas the opposite was true for M. abscessus (h > 0.6).

VNTR genotype and drug susceptibility
Associations between VNTR genotype and drug susceptibility profiles of M. abscessus and M. massiliense ( Table 2) showed that of the 188 unclustered isolates of the former, 53 (28.2%) were resistant to FOX which was a significantly higher rate than that for clustered isolates (3/30, 10.0%, P = 0.041). Similarly, for M. massiliense, the rate of MNO resistance among unclustered isolates (65/119, 54.6%) was significantly higher than among clustered isolates (18/44, 40.9%, P = 0.029). No significant differences were evident for other drug resistance rates between clustered and unclustered isolates within M. abscessus or M. massiliense groups of isolates (P > 0.05) (Supplementary table).

Discussion
Rapidly growing mycobacteria have in recent years been increasingly implicated as causes of extensive pulmonary infections in many countries [17,18]. Due to their natural resistance to most commonly administered antibiotics, infections by MABC members, especially the prevalent subspecies M. abscessus, requires special treatment regimens [6,7]. In this study, we compared in vitro drug susceptibility profiles between isolates of these subspecies from patients in China, and explored potential associations of antimicrobial susceptibility and strain genotypes that might have predictive value for guiding antimicrobial therapy.
The key finding was that both antibiotic susceptibility profiles and genotyping results differed between M. abscessus and M.

massiliense.
Of the three cornerstone drugs CLA, AMK and FOX, used for treatment of M. abscessus infections, CLA and AMK exhibited the highest activities against both species, but with a higher percentage of CLA-resistant strains of M. massiliense than M. abscessus, as was also observed by Nie et al. [19].
However, according to Harada et al. [20], CLA resistance rates of M. abscessus eclipsed those for M. massiliense in experiments based on a 14-day incubation period rather than the conventional 3-day period. Nevertheless, our results are in agreement with Nie et al. [19] in that resistance in M. massiliense is more extensive than in M. abscessus in China. We found high resistance rates to MXF, GAT and LVX in both subspecies, particularly for LVX which was much higher than those for MXF and GAT. These findings are consistent with previous reports that MABC isolates are frequently fluoroquinolone-resistant [19]. This phenomenon may be due to the overuse of fluoroquinolones in China, although some isolates are naturally resistant to this antimicrobial class [19]. Moreover, resistant rates to SFX and TIG among M. abscessus were shown here to be significantly lower than that for M. massiliense, indicating increased resistance in the latter. By contrast, M. massiliense proved to be more susceptible to MNO compared to M. abscessus, a finding in line with Adekambi et al. [21]. VNTR typing based on 17-loci was highly discriminatory for the two subspecies with DI(Diversity Index)values of 0.999 and 0.997, respectively, slightly exceeding values obtained for Malaysian and Japanese isolates [15,22]. Of differences in TR-2 and -137 loci (DI, respectively, 0.400 and 0.392 in Malaysia, and 0.786 and 0.736 in China), allelic diversity among most other M. abscessus loci was similar to previously published data [15]. This was speculated to be possibly due to phylogenetic incongruence among strains from different geographic regions, resulting in alterations in the discriminatory power of some loci. Most VNTR loci, except for TR101, of M. massiliense showed less allelic diversity than for M. abscessus, which suggests that these loci might be more suitable for differentiation between isolates of the latter. Similarly, the higher cluster rate of M. massiliense isolates may be a reflection related to pathogenicity associated with the severity of the epidemic, a concept which warrants further exploration. Yiting Wang et al.
To our knowledge, there are no comparable reports from China which have investigated strain VNTR genotype associations with drug resistance phenotypes for M. abscessus and M. massiliense on such a large number of samples. We observed that FOX resistance was more common among unclustered than clustered strains of M. abscessus, the former being significantly associated with MNO resistance in M. massiliense. Although NTM are regarded as opportunists that cause infections in both immunocompromised and vulnerable immunocompetent hosts [23], the observed resistance to FOX and MNO may be related to the pathogenicity of MABC strains and host preferences, leading to a biased distribution of these resistances among the subspecies.
Our study had two key limitations. First, we did not analyse M. bolletii isolates, due to the small number of samples, and second, as had been done in several other studies, we used 37°C instead of 30°C for DST (Drug susceptibility testing) analysis of rapid growers.
In conclusion, these data illustrate that among MABC isolates in China, M. abscessus was the most common subspecies followed by M. massiliense. AMK, CLA and AZM exhibited high antimicrobial activities but with significantly different resistance rates to MXF, GAT, SFX, TIG and MNO, between subspecies. Additionally, 17-loci VNTR typing had high discriminatory power for differentiation of clustered and unclustered isolates. Given the wide heterogeneity of resistance profiles to currently available antimicrobials, further genotype differentiation among subspecies may prove to be predictive of antibiotic resistance phenotypes, and inform more effective antibiotic regimens to optimise    Yiting Wang et al.