Non-tuberculous mycobacteria (NTM) are a large group of mycobacteria that are positive for acid-fast staining except Mycobacterium tuberculosis complex (MTBC) and M. leprae. More than 160 species of NTM have been identified, and 40 of these were deemed as pathogenic or opportunistic pathogens.

There has been a continuous increase in incidences and disease burden of NTM in many regions^{[1, 2]}. The incidence of NTM in the United Kingdom increased from 0.9 per 100, 000 population in 1995 to 2.9 per 100, 000 population in 2006 and continued to increase to 7.6 per 100, 000 in 2012^{[3, 4]}. The nationwide surveillance of tuberculosis (TB) in China showed that the average NTM isolation rate increased from 4.3% (29/682) in 1979 to 11.1% (49/441) in 2000 and 22.9% (83/363) in 2010^{[5, 6]}.

Most NTM strains are intrinsically resistant to the first- and second-line anti-TB agents and difficult to eliminate with treatment regimens commonly used for TB^{[7, 8]}. The resistance profile of NTM is highly species-specific and the clinical course and treatment response of NTM disease may be very variable^[9]. Hence, antibiotics prescribed for NTM treatment must have demonstrated activities against the specific species in the antibiotic susceptibility testing. As NTM antibiotic susceptibility testing is time-consuming, demanding several days [rapidly growing mycobacteria (RGM)] or weeks [slowly growing mycobacteria (SGM)] to obtain results depending on the bacterial growth time, more data need to be accumulated on the antibiotic susceptibility and resistance mechanism of NTM species to allow rapid detection of antibiotic-resistant genotype and for clinical applications.

Macrolides are the preferred choice of the limited antimicrobial agents for the treatment of NTM infections. The in vitro susceptibility test and in vivo clinical treatment outcome of macrolides are well consistent^{[1, 10]}. Resistance to macrolide class of antibiotics is mainly conferred by three different mechanisms as follows: substitution and modification of the 23S rRNA target site by various methyltransferases; drug efflux; and inactivation of the molecules by esterases, hydrolases, transferases, or phosphorylases enzymes^{[11, 12]}. However, only two well-known mechanisms have been reported for mycobacteria, namely, mutations in the 23S rRNA-encoding gene, rrl, at positions 2, 058/2, 059 (Escherichia coli numbering) for acquired resistance and erm genes encoding 23S rRNA methyltransferases for natural resistance^[9]. The erm genes have been described only in certain NTM species and have five classes as follows: erm(37) of M. tuberculosis complex, erm(38) of M. smegmatis, erm(39) of M. fortuitum, erm(40) of M. mageritense and M. wolinskyi, and erm(41) of MABC and M. fukienense^{[13, 14]}. Aside from the two mechanisms above, Ag85 mutant of M. smegmatis displayed increased sensitivity to erythromycin (ERY), indicating that the defects in the enzymes and proteins involved in maintaining the cell wall integrity may increase the susceptibility of the organism to macrolides^[15]. Some of genes encoding putative ATP-binding cassette (ABC) transport systems could be involved in macrolide export^[16]. However, more evidence is needed to demonstrate the role of Ag85 and ABC transport systems in the development of macrolide resistance in mycobacteria. In this study, 310 NTM clinical isolates belonging to six of the most common species in China were tested to comprehensively compare the susceptibility of NTM to four commonly used macrolides in clinical settings and elucidate the roles of the two known resistance mechanisms of NTM.

A total of 310 identified NTM clinical isolates from six provincial TB hospitals of China, including Anhui, Fujian, Jiangxi, Inner Mongolia, Hunan, and Sichuan Provinces, were included in the antibiotic susceptibility testing. All strains had been re-verified up to the species level by sequencing of the genes rrs, rpoB, hsp65, ITS, and sodA. All tests performed in this study were conducted in the laboratory of Branch of Tuberculosis, National Institute for Communicable Disease Control and Prevention.

Four macrolides used in this study, including ERY, clarithromycin (CLAR), azithromycin (AZM), and roxithromycin (ROX), were purchased from Sigma-Aldrich Co. (St Louis, MO). Minimum inhibitory concentrations (MICs) were determined with the broth microdilution method based on the Alamar Blue Assay (MABA). Briefly, fresh cultures were completely ground and their densities were adjusted to 0.5 McFarland (1.5 × 10⁸ cells/mL) with saline. The suspension was 1:200 diluted with culture medium supplemented with or without 10% oleic acid-albumin-dextrose-catalase (OADC) (BD, Franklin Lakes, NJ, USA) for SGM and RGM, respectively. Serial two-fold dilutions of the antibiotic solutions were prepared in Mueller-Hinton broth (Difco, Detroit, MI, USA) and inoculated with bacterial dilutions. The final reaction system comprised 100 μL antibiotic solution and an equal volume of the bacterial suspension in each well. The plates were sealed and incubated at 37 ℃. The indicator (20 μL Alamar Blue mixed with 50 μL of 5% Tween-80) was added when the drug-free control which was checked daily showed a color change from blue to pink. MIC testing usually ended at 3 to 6 days for RGM and 7 to 11 days for SGM. The lowest drug concentration that inhibited the strain growth and prevented color change was recorded as the MIC value. All tests for each strain were repeated twice. The dilution concentration range was 0.125-256.000 μg/mL for ERY, ROX, and AZM and 0.063-128.000 μg/mL for CLAR. The interpretive criteria of each drug were 8 μg/mL for RGM and 32 μg/mL for SGM, as suggested by the Clinical and Laboratory Standards Institute (CLSI, M24-A2)^[17].

Boiled DNA template from the fresh cultures was used. The primer set for the amplification of rrl gene was rrl-F, 5'-CCT GCA CGA ATG GCG TAA CG-3' and rrl-R, 5'-CAC CAG AGG TTC GTC CGT C-3'^[18]. The in-frame primer sets used to detect erm genes and the reference sequence used for primer design were as follows: erm(37)-F, 5'-CGG TGA GCT CGT GTT TGA CAT C-3' and erm(37)-R, 5' -AGG CCG ACG GTC AGG GTG AAC C-3' (GenBank accession No. AE000516); erm(38)-F, 5'-GAA ATC GTC TCG CGC ACA AAC-3' and erm(38)-R, 5'-TGC TGA CCA ACG TCG TCG AAG-3' (GenBank accession No. AY154657); erm(39)-F, 5'-AGT TCA TCA CGG CCG GCA TGA G-3' and erm(39)-R, 5'-ATC GAA CAA CGC CAC CCA CTG-3' (GenBank accession No. AY487229); erm(40)-F, 5'-TTG ACG GCC ATC GAG ATC GAC-3' and erm(40)-R, 5'-GAC GGT GTG ATG CCG TTG TG-3' (GenBank accession No. AY570506); erm(41)-F, 5'-GCA CTG CGC GAG AAG CTG GCA-3' and erm(41)-R, 5'-GCG GTG GAT GAT GGA AAG-3' (GenBank accession No. EU590124). The sequencing primer set of entire erm(41) was 5'-GCA CTG CGC GAG AAG CTG GCA-3' and 5'-GCA CTG CGC GAG AAG CTG GCA-3'. PCR products of rrl and erm(41) were sequenced (Tsingke BioTech, Beijing, China) and mutations were identified with alignment using MEGA 7.0 software.

The antibiotic susceptibility profiles and MIC ranges of 310 NTM clinical isolates are shown in Tables 1 and 2, respectively. The antibiotic resistance of the four macrolides was in the order ERY, AZM, ROX, and CLAR (high to low) and showed obvious species specificity. M. fortuitum and M. gordonae were the two species with the most distinct difference. M. fortuitum isolates were almost 100.0% resistant to all four macrolides, with 32 μg/mL MIC₅₀ and ≥ 128 μg/mL MIC₉₀. Hence, the use of macrolides may be excluded for the treatment of infections caused by M. fortuitum. On the contrary, M. gordonae strains were essentially sensitive to macrolides with ≤ 4 μg/mL MIC₉₀. The clinical isolation rate of M. massiliense was much lower than that of M. abscessus, and only four isolates were included in this study. Except for the two ERY-resistant isolates, all M. massiliense were sensitive to the four macrolides, which showed high antibacterial efficiencies. The resistance observed was similar for other three species, M. abscessus subsp. abscessus, M. avium, and M. intracellulare. Except for ERY, AZM, ROX, and CLAR showed a resistance rate lower than 24% and MIC₅₀ lower than 8 μg/mL.

To explore the role of target-site mutation in the development of macrolide resistance in NTM, the 23S rRNA coding gene, rrl, was sequenced and aligned for 143 strains with different MIC levels selected from 310 NTM clinic isolates that were tested for macrolide resistance.

Only one M. intracellulare strain harbored A2058T mutation and three M. abscessus subsp. abscessus harbored A2059G mutations. These four strains were highly resistant to all four macrolides and exhibited an MIC value as follows: ERY> 256 μg/mL, AZM> 256 μg/mL, ROX> 256 μg/mL, and CLAR ≥ 64 μg/mL.

Aside from 2, 058/2, 059 sites, a total of 193 point mutations different from those in the E. coli reference sequence were detected in 143 NTM isolates, wherein 134 sites were common in all NTM isolates tested (Table 3). The other 35 sites existed only in a few of the six NTM species (Table 4). These showed species specificity but were unrelated to macrolide MICs. M. abscessus subsp. abscessus and M. abscessus subsp. massiliense belonged to the same fast-growing M. abscessus species. Their sequences were identical at all sites mentioned above. Although these two subspecies could not be distinguished from each other with rrl gene, the two sites G2191A and T2221C allowed differentiation between these species and other four species. The genetic variation in M. avium and M. intracellulare that belonged to the same M. avium complex (MAC) was also very similar. Seven sites, including, G2140A, G2210C, C2217G, T2238C, T2322C, T2404C, and A2406G, were specially carried by these two species, while G2321A was only observed in M. intracellulare. Three sites, A2192G, T2358G, and A2636G, were distinct for M. fortuitum, while G2152A was specific for M. gordonae. The remaining 24 sites were harbored in one or two NTM isolates and these were mutations without any statistical significance. Although we failed to identify any macrolide resistance-related 23S rRNA mutations in this study, these genus- and species-specific sites on rrl gene mentioned above may facilitate strain identification up to the species level.

While detecting 23S rRNA mutations, the presence of five types of erm genes, erm(37) to erm(41), related to mycobacteria were screened in these 143 NTM isolates. PCR amplicons of expected size (-750 bp) with erm(39) primer sets were observed in all M. fortuitum. Amplicons of the expected size (-650 bp) with erm(41) primer sets were obtained in all 46 M. abscessus subsp. abscessus isolates, while those of smaller size (-300 bp) were detected in four M. abscessus subsp. massiliense isolates. Sequencing result showed that erm(41) of M. massiliense had two deletions (nucleotides 64 and 65; 276 bp after nucleotide 158) as compared with M. abscessus and, thus, lacked most of the functional domains of ribosomal RNA adenine dimethylases. No erm gene was detected in either of the three SGM species, M. avium, M. intracellulare, or M. gordonae.

As T/C polymorphism at 28th nucleotide on erm(41) correlated with inducible macrolide resistance^{[19, 20]}, we further compared the 3-day and 14-day antibiotic susceptibility of strains that carried erm(41) genes. Three M. abscessus subsp. abscessus strains harboring A2059G substitution were excluded, as their 3-day MICs had already reached the highest concentration. As shown in Table 5, 34 of 43 M. abscessus subsp. abscessus isolates harbored thymine at 28th nucleotide position (T28 sequevar) corresponding to Trp10 in the amino acid sequence. The other nine isolates were C28 sequevar (cytosine, Arg10). MICs of all T28 sequevar isolates reached the highest detection macrolide concentration at 14 days, irrespective of these strains being sensitive or resistant to macrolide at 3 days. In contrast, MICs of the nine C28 sequevar isolates were essentially the same at 3 and 14 days. Thus, erm gene was responsible for inducible macrolide resistance of M. abscessus subsp. abscessus; the resistance observed at 3 days was out of the interpretable range for 23S rRNA and erm gene. Although all four M. massiliense isolates were T28 sequevars, these were sensitive to macrolides and had no change in MICs upon prolongation of the antibiotic susceptibility test to 14 days owing to the deletion and loss of function of erm(41).

Pulmonary infections with NTM are becoming an increasing concern in many countries. Macrolides are one of the most important drugs used for the treatment of NTM, especially MAC infections. In the present study, the result of an in vitro antibiotic susceptibility testing confirmed the efficacy of macrolides against NTM. The resistant rates and MIC₅₀ of all strains except M. fortuitum to AZM, ROX, and CLAR were lower than 24% and 8 μg/mL, respectively. In addition, these drugs were two- to four-fold more active than ERY in vivo^[21]. CLAR was the most potent agent against NTM strains, with a resistance rate of 8.1% (25/310), and served as the only agent with high clinical efficacy in the susceptibility test^[22]. These results are in line with those reported by Foo et al.^[23], wherein CLAR inhibited M. abscessus and MAC at a resistance rate of 20% and 10%, respectively. AZM was administrated once daily and is beneficial for patient compliance, especially for the long-term treatment over 6 months commonly required in mycobacterial infections^[24]. High efficiency of AZM against NTM was also observed. However, comparative studies with CLAR and AZM for NTM treatment are limited. In our study, AZM concentration required to inhibit NTM was higher than that of CLAR. However, some conflicting results were observed. Choi et al.^[25]. found that AZM is a weaker inducer of erm(41) gene expression than CLAR and should therefore be preferred for M. abscessus infections. In contrast, the findings reported by Maurer^[26] contradict this suggestion, as high median MICs of ≥ 256 μg/mL on day 12 were observed for the two drugs. Thus, further studies are warranted to determine the clinical efficacy of CLAR and AZM against NTM. The activity of ROX was reported to be comparable to that of CLAR both in vitro and in vivo in animal models^[27]. Consistent with former reports, we observed a similar activity for ROX in our results. ROX was successfully used for the treatment of cutaneous M. chelonae infections^[28].

The results of the antibiotic susceptibility testing demonstrated the great difference in sensitivity between different species, particularly between M. gordonae and M. fortuitum as well as M. abscessus subsp. abscessus and M. abscessus subsp. massiliense. In our study, most of M. abscessus and M. massiliense were susceptible to macrolides on day 3. However, resistance was evident only in M. abscessus after 14 days of incubation. Truncated erm(41) was observed in M. massiliense and erm(41) polymorphism in M. abscessus was related to inducible macrolide resistance. These results support the findings that macrolide-containing regimens were more effective for the treatment of M. massiliense infections than M. abscessus^[29]. In our study, 77.8% M. fortuitum were found to be resistant to CLAR and all carried erm(39). Esteban et al.^[30] showed that 84.3% (75/89) M. fortuitum clinical isolates harbored erm and only 52.8% of these were resistant to CLAR. This variation may be associated with the differences in the methodology employed and regions. Thus, genetic polymorphism of erm(39) gene may exist in M. fortuitum, necessitating further studies. SGM species were more susceptible to macrolide than RGM, probably owing to the absence of erm genes in these species. We found that M. avium seemed less susceptible than M. intracellulare to macrolide, consistent with the results of previous reports^{[31, 32]}. As macrolide susceptibility of MAC has been correlated with clinical treatment outcomes, one may expect more therapeutic failures in treating M. avium infections than M. intracellulare infections. These species-specific sensitivity suggests the importance of the identification of NTM isolates to subspecies level to design specific treatment regimens. The species-specific sites of rrl gene, such as G2321A of M. intracellulare, A2192G, T2358G, and A2636G of M. fortuitum, and G2152A of M. gordonae, identified in this study may be employed as a rapid identification method for NTM.

The most striking observation of this study is the low detection rate of the two well-known macrolide resistance mechanisms. Several reports have suggested that point mutations at 2, 058/2, 059 position in rrl gene were the most common mechanisms to confer macrolide resistance in MAC strains and were observed in> 90% of the reported resistant mutants^[22]. However, a significant fraction of resistant strains who failed to harbor any previously identified mutations were observed in former reports. In this study, the proportion of strains with 2, 058/2, 059 substitutions in 23S rRNA was particularly low and none of other substitutions related to macrolide resistance was identified. This observation was similar to M. abscessus strains isolated from Korea, wherein most strains with acquired resistance had no 2, 058/2, 059 substitutions in rrl^[33]. In this study, erm(39) and erm(41) were detected in all M. fortuitum and MABC and inducible resistance was observed in relevant sequevars. However, the resistance observed at 3 days could not be explained by 23S rRNA and erm mechanisms. Therefore, the well-known macrolide resistance mechanisms failed to cover NTM isolates and meet the requirement for developing fast diagnostic methods for resistant strains to guide rational and individualized medication. The unknown mechanisms responsible for macrolide resistance in NTM remain to be elucidated.

LI Fu designed and performed the experiments, analyzed data, and prepared the manuscript; LI Gui Lian, PANG Hui, LIU Hai Can, XIAO Tong Yang, LI Shuang Jun, and LUO Qiao participated in the experiments; WANG Rui Bai and WAN Kang Lin designed and reviewed the draft of the manuscript. All authors read and approved the final manuscript.

The authors thank LI Jun Lian (Xinjian Chest Hospital, Urumqi, Xinjiang, China), WANG Qing (Anhui Chest Hospital, Hefei, Anhui Province, China), TONG Chong Xiang (Lanzhou Chest Hospital, Gansu, China), FU Jun (Nanchang Centers for Disease Control and Prevention, Jiangxi, China), YANG Jun (Sichuan Center for Disease Control and Prevention, China), and SU Yun Kai (Inner Mongolia Institute for Tuberculosis Control and Prevention, Hohhot, China) for their contributions to this study through the collection of clinical isolates.