The incidence of non-tuberculous mycobacteria (NTM) infections has increased rapidly over the past decade coincidentally with an increase in the number of immune dysfunction cases, as well as the number of endoscopies, surgeries, transplantations and other invasive operations (Griffith et al., 2007; Lai et al., 2010; Hoefsloot et al., 2013; Marras et al., 2013; Brode et al., 2014; Prevots and Marras, 2015). Among these, M. abscessus infections are particularly challenging. M. abscessus is one of the most antibiotic-resistant pathogens known (Nessar et al., 2012). It is resistant to most first-line antibiotics, and naturally resistant to antitubercular agents. Consequently, even treatment with a combination of antibiotics over a long period of time is often ineffective (Jarand et al., 2011).

The 2007 American Thoracic Society guidelines recommended the inclusion of clarithromycin (CLA) in combination therapy for M. abscessus infections (Griffith et al., 2007). In recent years, however, CLA efficacy was found closely related to bacterial genotype (Wallace et al., 1996; Nash et al., 2009; Bastian et al., 2011; Lee et al., 2014; Brown-Elliott et al., 2015; Rubio et al., 2015). Two kinds of genotype-related CLA resistance patterns have been confirmed. (1) A point mutation in the 2058/2059 (E. coli numbering) locus of the 23S rRNA (rrl) gene confers acquired resistance. (2) An intact erm(41) gene, which exhibits a T/C polymorphism at the 28th nucleotide, confers inducible resistance to CLA when thymidine is the nucleotide [erm(41)T28]. The CLA-induced erm(41)T28 gene product methylates ribosomal 23S mRNA, preventing CLA binding to the ribosome (Bastian et al., 2011). Additionally, there are two sensitivity patterns: (1) When cytidine is the 28th nucleotide [erm(41)C28], there is a loss of both adenine methylating function and CLA resistance. (2) Deletion of erm(41) nucleotides 64 and 65, and deletion of erm(41) nucleotides 159–432; both deletions result in the loss of gene function (M type) (Nash et al., 2009). M. abscessus genotype and CLA phenotype are correlated. To determine the CLA resistance phenotype (i.e., sensitive, acquired resistance or induced resistance), it is necessary to conduct in vitro susceptibility testing on day 3 (early reading time, ERT) and day 14 (late reading time, LRT) after CLA treatment.

Amikacin is also a recommended, primary therapeutic agent for treatment of M. abscessus infections. One underlying mechanism for amikacin resistance seems to be an A to G mutation in nucleotide 1408 of the rrs gene (Nessar et al., 2011). Taken together, these patterns indicate that M. abscessus antibiotic susceptibility studies should focus on the following genotypes: (1) erm(41) [including erm(41)T28, erm(41)C28 and M type], (2) the 2058/2059 rrl gene mutation, and (3) the 1408 rrs mutation.

Identification of the genotype of infectious M. abscessus is instructive when selecting antimicrobial agents for treatment. Recently, Mougari and co-workers correlated the results of standardized antibiotic susceptibility testing with M. abscessus genotyping (Mougari et al., 2016). Their analyses confirmed that rrl, erm(41) and rrs genotypes accounted for and predicted the resistance to CLA and amikacin exhibited by clinical M. abscessus isolates in France. Therefore, therapeutic approaches to treating M. abscessus infections should consider the genotype involved although the character of the etiologic agent is frequently not known in such detail. China has a large number of patients with M. abscessus infection. But so far, there is no systematic study on the relationship between susceptibility and genotype of M. abscessus clinical isolates in China. Recent articles have only explored the relationship between susceptibility and subtype (Pang et al., 2015), or just mentioned that the erm(41) genotype was related to clarithromycin resistance after 14 days' incubation, without detailed genetic information or susceptibility test results (Nie et al., 2015). Moreover, the genotype and antibiotic susceptibility of M. abscessus may vary worldwide (Harada et al., 2012; Lee et al., 2014; Brown-Elliott et al., 2015; Nie et al., 2015; Pang et al., 2015; Rubio et al., 2015; Cowman et al., 2016; Garcia de Carvalho et al., 2016; Mougari et al., 2016). The study described herein was undertaken to establish the relationship between antibiotic susceptibility and M. abscessus genotype in China, and to compare it to that found in other countries.

One hundred sixty-two M. abscessus isolates were collected, including 123 subsp. abscessus and 39 subsp. massiliense. M. abscessus subsp. bolletii was not considered in this study due to its general absence in China. One hundred eight erm(41)T28 sequevar and 15 erm(41)C28 sequevar were among subsp. abscessus. The susceptibilities of these erm(41) sequevars to the 10 antibiotics tested were shown in Table 1. While a relatively large percentage of the erm(41)T28 sequevar exhibited both acquired and inducible resistance to CLA, only 3 isolates exhibited resistance to amikacin. Neither an rrs 1408 mutation nor any other meaningful SNP was found in these 3 isolates. No correlation was observed between genotype and susceptibility to the 8 other antibiotics tested. The MIC50 and MIC90 values for all 10 antibiotics were shown in the Supplementary Table 1. The phylogenetic tree of 162 M. abscessus was also constructed, with information of genotypes and clarithromycin/amikacin-resistant types for each strain labeled (Supplementary Figure 1).

The effects of M. abscessus subtype, erm(41) sequevar and incubation time on the MIC values of CLA were determined (Table 2). There were 66 CLA-sensitive (MIC ≤2 mg/mL) and 11 CLA-intermediate (MIC = 4 mg/mL), M. abscessus subsp. abscessus isolates [14 erm(41)C28 and 63 erm(41)T28 with no rrl 2058/2059 mutation] accessed at ERT. Forty-six acquired resistance isolates (MIC >8 mg/mL) accessed at the same time included 45 erm(41)T28, 2 with an rrl 2058/2059 mutation, and 1 erm(41)C28 isolate with an rrl 2058/2059 mutation. There were 29 CLA-sensitive and 1 CLA-intermediate subsp. massiliense isolates with no rrl 2058/2059 mutation, and 9 acquired resistance isolates, 7 of which had an rrl 2058/2059 mutation. Analysis of the rrl mutation pattern revealed 7 isolates with a 2058 mutation (5 isolates exhibiting a A to C mutation and 2 isolates exhibiting a A to G mutation); 2 isolates exhibited a 2059, A to C mutation. The remaining two resistant subsp. massiliense isolates, which exhibited no rrl 2058/2059 mutation, were analyzed to determine the presence of an 18 bp insertion mutation in the rpIV gene (encoding 50S ribosomal protein L22), no such insertion was found (data not shown).

There were 15 CLA-sensitive subsp. abscessus isolates [14 erm(41)C28 and 1 erm(41)T28] and 108 (inducible) resistant isolates [1 erm(41)C28] assessed at LRT. Twenty-eight CLA-sensitive and 11 CLA-resistant subsp. massiliense were also determined at LRT (Table 2).

Sixty-two subsp. abscessus isolates, all of which were erm(41)T28, and 2 subsp. massiliense exhibited inducible resistance to CLA assessed at LRT. The correlation between CLA susceptibility and M. abscessus genotypes isolated in China was shown in Table 3. The same correlation for M. abscessus isolated in Spain, France and the US, gleaned from data reported by other investigators (Brown-Elliott et al., 2015; Rubio et al., 2015; Mougari et al., 2016), was shown for comparison purposes (Supplementary Table 2).

This study was the first undertaken to correlate antibiotic susceptibility with the genotypes of M. abscessus isolated in China. The proportion of M. abscessus isolates that acquired resistance to CLA (ERT assessment) was relatively high (33.95%, 55/162) compared to other countries: Japan, 12.79% (11/86); Korea, 15.84% (64/404); UK, 6.54% (35/535); France, 9.09% (15/165); US, 2.51% (9/358); and Brazil, 5.55% (2/36). The countries included in this comparison consisted of those where M. abscessus infections were most frequently studied and, consequently, were the largest sources of available data over the last 5 years (Harada et al., 2012; Lee et al., 2014; Brown-Elliott et al., 2015; Nie et al., 2015; Pang et al., 2015; Rubio et al., 2015; Cowman et al., 2016; Garcia de Carvalho et al., 2016; Mougari et al., 2016). Undoubtedly, uneven economic development, and differences in the production, management and clinical use of antibiotics contribute to this disparity in rates of CLA resistance observed among countries.

Further analysis indicated that only 3 of the ERT resistant subsp. abscessus isolates had an rrl 2058/2059 mutation. The rrl SNP distribution of the remaining 43 isolates was scattered and irregular, but all isolates belong to the erm(41)T28 sequevar. As erm(41)T28 is an inducible resistance genotype, we speculated that resistance may have occurred in a large portion of these isolates as a consequence of contact with macrolide agents such as CLA prior to isolation. This speculation is based upon the following considerations. According to Chinese community-acquired pneumonia (CAP) guidelines, macrolides were once the primary agents of choice to treat a high proportion of mycoplasma pneumoniae and mixed, atypical pathogens infections (Xu et al., 2007). Consequently, many Chinese CAP patients including patients enrolled in the present study may have a history of macrolide use. Additionally, China has a large number of patients with bronchiectasis, which is linked closely with incidences of NTM infection (Chu et al., 2014). Notably, macrolides are widely recommended for long-term treatment of bronchiectasis patients (Fan and Xu, 2014). Therefore, prior use in treating bronchiectasis may also be a factor contributing to the high rate of macrolide resistance exhibited by pathogenic M. abscessus isolated in China. Confirming this speculation will require detailed histories of the patients and their antibiotic use, as well as analyses of the erm(41) expression level and 23S rRNA methylation status of the infectious organism. This work is ongoing in our laboratory.

There were 9 CLA-resistant subsp. massiliense isolates assessed at ERT, only 7 exhibited an rrl 2058/2059 mutation. Since subsp. massiliense possesses an incomplete/nonfunctional erm(41) (M type), the mechanism underlying resistance of the 2 M type erm(41) strains with no rrl 2058/2059 mutation remains to be determined. However, it has been reported that M type isolates will ultimately acquire resistance provided long-term CLA exposure; the underlying mechanism(s) appears to be a change in 50S ribosomal subunit structure (Mougari et al., 2017). Following exposure, 97.5% (39/40) M type strains exhibited an rrl 2058/2059 mutation; 2.5% (1/40) had an 18 bp insertion mutation in the rpIV gene, which encodes the L22 ribosomal protein. In the present study, CLA resistance in subsp. massiliense was principally due to an rrl 2058/2059 mutation, which occurred in 77.8% (7/9) of isolates; no rpIV mutation was found in the 2 resistant isolates that remained.

To further understand the relationship between M. abscessus genotype and CLA susceptibility typical of different geographic regions, regions with detailed genotypes and CLA susceptibility data were selected and the data compared with those obtained in China (Table 3 and Supplementary Table 2) (Brown-Elliott et al., 2015; Rubio et al., 2015; Mougari et al., 2016). Sixty to hundred percent of the acquired, CLA-resistant M. abscessus isolated in Europe and the US, but only 18.2% in China, expressed an rrl mutation. Most acquired, CLA-resistant isolates in China might be macrolide-induced erm(41)T28 strains. The vast majority (96.9–100%) of the induced CLA-resistant strains were erm(41)T28. Only 2 M type strains in our study appeared induced resistance, the mechanism is unknown. One CLA-resistant erm(41)C28 was isolated. Studies suggest, that CLA resistance can be induced in erm(41)C28 sequevar by long-term exposure to higher CLA concentrations (Maurer et al., 2012; Mougari et al., 2017). This might require a mutation in the rrl gene, which is consistent with our findings.

CLA-sensitive isolates assessed at LRT belonged mainly to the erm(41)C28 and M type sequevars; one, however, belonged to erm(41)T28. The mechanism(s) underlying the CLA sensitivity of the erm(41)T28 isolates remains to be elucidated.

M. abscessus isolates that belong to the erm(41)C28 or M type sequevar, which do not express an rrl 2058/2059 mutation, are probably susceptibility to CLA treatment. In the case of patients infected with erm(41)T28 sequevar, treatment options should consider geographic location and patient medical history; response to treatment and recovery should be closely monitored.

Only 3 (1.9%, 3/162) of the M. abscessus isolates recovered in China exhibited resistance to amikacin, neither an rrs 1408 mutation nor any other meaningful SNP was determined. Consequently, amikacin appears to have a higher efficiency rate than CLA in treating M. abscessus infections in China. The requirement for frequent and long-term intramuscular or intravenous administration undoubtedly impacts negatively on the clinical use of amikacin. Recently, Olivier et al. reported the results of a phase II clinical trial in which long-term liposomal amikacin inhalation combined with traditional therapy improved the outcomes of the sputum culture and 6-min walk test in NTM infected patients (Olivier et al., 2016). These findings suggest that amikacin may be a good, alternative for M. abscessus treatment.

Our results are consistent with those obtained in other studies with the exception that most M. abscessus isolates obtained in China were resistant to imipenem and cefoxitin (Griffith et al., 2007; Mougari et al., 2016). Therefore, a better treatment for M. abscessus infections in China might be CLA combined with amikacin or linezolid, while European and American researchers recommended the use of cefoxitin as an alternative agent (Griffith et al., 2007; Harada et al., 2012; Nie et al., 2015; Pang et al., 2015; Cowman et al., 2016; Mougari et al., 2016). Imipenem should not be considered for early treatment in China. No resistance-related genotypes for the other 8 antibiotics tested were found.

The rrl and erm(41) genotypes are predictive of the resistance of M. abscessus clinical isolates to CLA. China has a large number of CLA-resistant M. abscessus isolates with erm(41)T28 sequevar. Treatment of M. abscessus infections should be based upon a comprehensive consideration of factors that include genotype and geographic location.

HC, BL, SY, and ZZ, conceived and designed the work; WL, SY and LL collected the bacterial isolates and compiled the data; WM, BL and LL analyzed and interpreted the data; HC, ZZ and XX revised the manuscript for important intellectual content. All authors approve the final draft of this manuscript. HC is accountable for all aspects of the work.