Analysis of drug-susceptibility patterns and gene sequences associated with clarithromycin and amikacin resistance in serial Mycobacterium abscessus isolates from clinical specimens from Northeast Thailand

Mycobacterium abscessus is an important infectious agent highly associated with drug resistance and treatment failure. We investigated the drug resistance situation of M. abscessus in Northeast Thailand and the possible genetic basis for this. Sixty-eight M. abscessus clinical isolates were obtained from 26 patients at Srinagarind Hospital during 2012–2016. Drug susceptibility tests and sequencing of erm(41), rrl and rrs genes were performed. Mycobacterium abscessus was resistant to 11/15 antibiotics (nearly 100% resistance in each case). Partial susceptibility to four antibiotics was found (amikacin, tigecycline, clarithromycin and linezolid). Non-massiliense subspecies were significantly associated with clarithromycin resistance (p<0.0001) whereas massiliense subspecies were associated with tigecycline resistance (p = 0.028). Inducible clarithromycin resistance was seen in 22/68 (32.35%) isolates: 21 of these isolates (95.45%) belonged to non-massiliense subspecies and resistance was explicable by the T28C mutation in erm(41). Inducible clarithromycin resistance was found in one isolate of the massiliense subspecies. Acquired clarithromycin resistance explicable by the A2271G/C mutation of rrl was seen in only 7/16 (43.75%) of strains. Inducible and acquired resistance mechanisms can be interchangeable during the course of infection. Rrs mutations were not associated with amikacin resistance in our study. Antibiotic resistance in subspecies of M. abscessus was reported from Northeast Thailand. Known resistance-associated mutations cannot explain all of the resistance patterns observed.

Introduction Nontuberculous mycobacteria (NTM) do not cause tuberculosis (TB) but can nevertheless cause life-threatening diseases. Although NTM infection exhibits similar characteristics to TB in terms of symptoms and AFB staining, different antibiotics are required for treatment. Some species of NTM have been found to be resistant to many types of antibiotics [1].
Mycobacterium abscessus is a rapidly growing mycobacterium that causes a wide spectrum of diseases in humans, including pulmonary, skin and soft tissue, and disseminated disease [2]. In addition, it is uniformly resistant to the standard anti-TB agents and is highly associated with treatment failure [3]. Mycobacterium abscessus has been classified into three subspecies based on whole-genome sequencing analysis: M. abscessus subsp. abscessus, M. abscessus subsp. bolletii, and M. abscessus subsp. massiliense [4].
The antibiotic regimen recommended for treatment of M. abscessus infection (ATS/IDSA guidelines 2007) is a multidrug macrolide-based therapy and hence clarithromycin is regarded as a key antibiotic [2]. In addition, natural susceptibility of M. abscessus to amikacin has been reported and hence inclusion of this antibiotic in the treatment regimen has been suggested [5]. Two main classes of clarithromycin resistance are acquired and inducible resistance [6,7]. Acquired clarithromycin resistance is associated with point mutations at positions 2,058 and 2,059 of the rrl gene encoding the 23S rRNA [6]. Inducible resistance is conferred by the erm (41) gene, coding for erythromycin ribosomal methylases [7,8]. Several studies have investigated the roles of erm (41) and rrl of M. abscessus in conferring resistance to clarithromycin [9][10][11][12], but with inconsistent results [13]. These mutations did not explain all clarithromycinresistant M. abscessus strains [13] and other mechanisms might exist. Amikacin resistance is also partially explained by point mutations at positions 1,406 to 1,409 in the rrs gene encoding 16S rRNA [8,14,15].
Studies in a number of countries have reported the drug resistance situation for M. abscessus [16][17][18]. The few reports of the drug-resistance situation in M. abscessus infection in Thailand date to before 2005 [19][20][21]: further study to supplement and update these is needed.
Serially isolated strains of a bacterial pathogen provide a good model to investigate the acquired drug resistance mechanism and to monitor changes in drug susceptibility during the course of an NTM infection. So far, there has only been only one report (from Spain) investigating clarithromycin resistance in four serially isolated strains of M. abscessus [22].
We aim to update information on the drug resistance situation and to investigate mutations of genes associated with clarithromycin and amikacin resistance in serial isolates of M. abscessus from patients in Northeast Thailand.

Case definition
Among the 26 patients, 20 were regarded as suffering true NTM infection based on the criteria in the ATS/IDSA guidelines 2007 [2]; isolation from a sterile site (n = 10), had radiological data, isolation from sputum �2 times, with clinical symptoms and also supported by history of NTM treatment (n = 10). The remaining six patients were previously treated TB patients who did not match the criteria of true infection were therefore defined as possible colonization cases.

Drug susceptibility testing
Drug susceptibility testing was performed using a RAPMYCOI Sensititre 96-well plate (Sensititre, Trek Diagnostic Systems, United Kingdom). The plates were incubated at 30˚C for 5 days (except for the clarithromycin susceptibility test). Clarithromycin susceptibility was read at 3, 5 and 14 days. A reading at day 3 was used to test for inducible resistance according to the protocol previously described [7]. Inducible resistance was inferred by changes in MIC values from susceptible at day 3 to resistant at day 14. Strains with resistance status since day 3 were regarded as demonstrating acquired resistance. Interpretation of the results followed the Clinical and Laboratory Standards Institute (CLSI M24-A2) guidelines. Staphylococcus aureus ATCC29213 and Mycobacterium abscessus ATCC19977 were used as control strains.

DNA extraction
All M. abscessus isolates were subcultured onto Löwenstein-Jensen media and incubated at 37˚C for 7 days. Genomic DNA from M. abscessus isolates was extracted from multiple loopfuls of M. abscessus colonies using the cetyl-trimethyl-ammonium bromide-sodium chloride (CTAB) method [23].

PCR and gene sequencing
We selected 3 genes (rrs, erm (41) and rrl) to investigate the resistance-related genetics of two antibiotics, amikacin and clarithromycin (inducible and acquired resistance in the latter). Quick Taq HS DyeMix (Taq-based 2x master mix PCR reagent containing all components for PCR except primers and template DNA) were used in PCR for sequencing. The PCR conditions for amplification of all genes were as follows; pre-denaturation at 95˚C for 5min; 35 cycles of 95˚C for 1 min, 63˚C (64˚C for erm (41) and for rrl) for 1 min, 72˚C for 1.5 min and final elongation at 72˚C for 10 min. Table 1 lists the primer sequences that were newly designed in our study. The PCR products were sent for sequencing using Sanger sequencing (BioBasic Inc., Canada) with an Automated Sequencer ABI Prism 3730XL.  [7,24,25]. Comparisons of drug susceptibility patterns between massiliense and non-massiliense subspecies were calculated as percentage and proportion. Drug susceptibility patterns and sequences of associated genes were analyzed in serially isolated strains exhibiting changing drug susceptibility (within-patient analysis) and between drug-susceptible vs drug-resistant groups of isolates (between group analysis). A correlation analysis of SNPs and MIC level of antibiotics was performed. A chi-square test or Fisher's exact test were used for the comparison of categorical variables. P<0.05 was considered significant. Demographic and epidemiological data, as well as the clinical information, were analyzed. All statistical analyses were performed using SPSS version 16.0.

Drugs resistance situation of M. abscesses in Northeast Thailand
The drug susceptibility pattern of M. abscessus isolates from clinical specimens of 26 cases (n = 68) is shown in Table 1. Our M. abscessus isolates exhibited nearly 100% resistance to 11/ 15 antibiotics. Partial susceptibility was noted to only 4 antibiotics (82.35% of isolates were susceptible to amikacin, 50% to tigecycline, 48.53% to clarithromycin and 14.71% to linezolid) ( Table 2).

Association between massiliense and non-massiliense subspecies of M. abscessus and drug-susceptibility patterns
We found that non-massiliense subspecies were significantly associated with clarithromycin resistance (p<0.0001) whereas the massiliense subspecies was significantly associated with tigecycline resistance (high resistance combined with intermediate resistance) (Table 3). However, both subspecies categories were highly susceptible to amikacin and moderately resistant to linezolid.

Changes in DNA sequences and MIC levels in isolates serially sampled from individual patients
A change in MIC levels for all 15 antibiotics was often detected between serial isolates from individuals, regardless of whether that individual was diagnosed as infected or colonized. MIC levels for 7/15 antibiotics were found to increase or to decrease over different sampling intervals in serially isolated strains (S1 Table). Drug susceptibility of serial isolates from each patient could also change from resistant to susceptible or susceptible to resistant (Fig 1 and S1 Table). Inducible and acquired resistance to clarithromycin were interchangeable (Fig 1). For two antibiotics, associated genes were sequenced (erm (41) and rrl for clarithromycin and rrs for amikacin). No particular genetic variant was associated with increased or decreased MIC levels ( Table 4) and susceptibility patterns (resistance or susceptibility) (data not shown) in amikacin and clarithromycin.
Raw data used in this study was available S4 Table.

Discussion
Mycobacterium abscessus is an important pathogen in immunocompromised and immunocompetent patients causing pulmonary infection, soft tissue infections and disseminated infection [2]. Mycobacterium abscessus is one of the most antibiotic-resistant pathogens, rendering it difficult to treat [3]. Suggested antibiotics include clarithromycin, azithromycin, amikacin, cefoxitin, meropenem, imipenem, ciprofloxacin and trimethoprim/sulfamethoxazole, with clarithromycin being the key antibiotic [2]. It is likely that M. abscessus is intrinsically resistant to ethambutol, imipenem, isoniazid and rifampicin [26]. Ethambutol resistance is associated with alterations in embB [27]. Imipenem is highly unstable, thereby affecting/limiting in-vitro DST over several days, as required for M. abscessus [28]. Rifampicin resistance is due to the expression of a rifamycin ADP-ribosyltransferase (arr) [29]. Action of an efflux pump is a possible mechanism for intrinsic isoniazid resistance [3]. We have reported here the drug-resistance situation of M. abscessus infection in Northeast Thailand with implications for antibiotic selection for treatment. Different rates of clarithromycin resistance of M. abscessus have been reported; 75.34% in China [30], 35% in Venezuela [11], but resistance is rare in England [31]. We investigated the antibiotic resistance of M. abscessus in Northeast Thailand and found that, despite high levels of resistance to a broad range of antibiotics, clarithromycin remains the most effective antibiotic with >80% of isolates being susceptible. High or very high levels of susceptibility were noted to amikacin and

Variants (No. of cases)
Clarithromycin erm (41) 11 De11A (10)   tigecycline, which therefore are good candidates for treatment of M. abscessus infection in this region. The latter is not included in CLSI M24-A2 guidelines, yet half of the isolates had low MIC levels for tigecycline. Only 15% of isolates were susceptible to linezolid, rendering this antibiotic of limited value. However, tigecycline and linezolid were found to be the most effective combination for treatment tested in an animal model and humans [32]. We interpreted "high", "moderate" and "low" resistance for non-CLSI standardized antibiotics based on the range of the MIC obtained from the studied strains for each drug. This approach could help to identify potential antibiotics for treatment and to monitor the drug resistance situation of M. abscessus infection. Mycobaterium abscessus has been classified into three subspecies based on whole-genome sequencing analysis [4]. These subspecies differ in their susceptibility to clarithromycin [6,33]. Mycobacterium abscessus subspecies massiliense does not show inducible resistance to clarithromycin [7]. Subspecies abscessus and bolletii seem to be able to resist clarithromycin through a mechanism depending on the erm(41) gene [6,7,34]. We analyzed the association between a panel of 15 antibiotics and M. abscessus subspecies in Northeast Thailand. We found that non-massiliense subspecies were significantly resistant to clarithromycin whereas the massiliense subspecies was significantly resistant to tigecycline. Resistance to tigecycline due to an alteration of tetracycline monooxygenase was recently reported [35]. Other mechanisms of tigecycline resistance might also exist. There is a risk that the massiliense subspecies might evolve greater resistance to tigecycline similar, in the way that the non-massiliense subspecies (bolletii and abscessus subspecies) have become highly resistant to clarithromycin. These findings could be helpful for managing antibiotic treatment of M. abscessus infection where subspecies information is available.
Previously, 2 deletions (nucleotides 64-65 and nucleotides 159-432) [7,24,25] and the mutation G466A in erm (41) [33] were suggested as markers for the massiliense subspecies. We found that 4 isolates of massiliense subspecies defined by these 2 deletions did not have erm (41) G466A. Furthermore, 1 out of the 4 isolates had an inducible resistance phenotype for clarithromycin, which should not be found in this subspecies. We also found a shift by 3 bp of the 2 deletions (nucleotides 64-65 and 159-432 became nucleotides 61-62 and 156-429, respectively). This shift implies a 3-bp deletion elsewhere in the upstream region of M. abscessus strains from our study.
Our study used serially sampled isolates from individual patients. Advantages of this approach include; (i) the genetic backbone is the same in each serially sampled strain, hence controlling for genetic differences between strains (ii) the change of MIC levels during the course of treatment can be used to study the response of the pathogen to the antibiotic. Furthermore, increasing or decreasing MIC levels can be used to study within-host evolution of a drug-resistant strain. Only a single previous study has investigated gene sequences of serially isolated strains of M. abscessus. However, this was a small study (10 isolates from 4 patients) and provided little information [22].
In the 2007 American Thoracic Society guidelines, macrolide drugs such as clarithromycin are recommended as key antibiotics for treatment of M. abscessus infection [2]. No previous study has investigate changes of MIC levels in clarithromycin and amikacin in association with sequences of drug-resistance genes. We investigated drug resistance in serially isolated strains of M. abscessus and sequenced erm (41) and rrl genes for clarithromycin and rrs gene for amikacin. These antibiotics were selected because they are the principal antibiotics used in our hospital for treatment of M abscessus, are recommended in the guidelines [2] and exhibit suitable variation of MIC levels. We analyzed genetic and MIC level changes during the course of persistent infection/colonization. In serially isolated strains, MIC levels for all antibiotics frequently changed between sampling times. The response of M. abscessus to antibiotics seems not to be stable through time, possibly in response to antibiotics previously used or other factors. Besides subspecies markers, no association was found between the mutations of erm (41), rrl, and rrs genes and changes in MIC levels or the resistance phenotype for clarithromycin and amikacin.

Patterns n (%) SNPs
Massiliense subspecies (n = 35) Indels Acquired resistance (MIC�8 μg/ml at day 3) Two mechanisms (acquired and inducible) of clarithromycin resistance have been reported. Acquired resistance to clarithromycin is associated with point mutations at A2271G [36] (or at 2058-2059 in the Escherichia coli numbering system [37,38]) of the rrl gene. We found these two mutations only in 43.75% (7 strains) of all 16 isolates (5 of 8 massiliense and 2 of 8 nonmassiliense subspecies isolates) with acquired clarithromycin resistance. It seems that acquired clarithromycin resistance in M. abscessus (of both subspecies) cannot be exclusively explained  Table 7. Association between rrl gene sequence and clarithromycin susceptibility in M. abscessus (n = 64).

Patterns N (%) SNPs Indels
by rrl mutations. Concordantly, no A2271G variant of rrl was found in clarithromycin resistance strains from a previous study [22]. Therefore, A2271G rrl is not a robust resistant marker in our region and other associated mutations conferring resistance might exist. T28C of erm (41) is associated with inducible resistance to clarithromycin [6,7,25] in nonmassiliense subspecies [33]. Subspecies massiliense has 2 deletions leading to dysfunction of erm (41) and so the resistance mechanism is independent of T28 erm (41) and this subspecies is usually susceptible to clarithromycin [6,7]. Our study concordantly found that all isolates of the massiliense subspecies had truncated erm (41) and 70% of them were susceptible to clarithromycin. For non-massiliense subspecies (subspecies without erm (41) deletion markers), only strains containing T28 erm(41) exhibit inducible clarithromycin resistance [33]. All of our non-massiliense subspecies isolates had T28C in erm (41) and around 75% of these had inducible clarithromycin resistance. So inducible clarithromycin resistance in our non-massiliense subspecies can be explained by T28C in erm (41). However, one isolate of the massiliense subspecies (identity based on 2 deletions in erm (41)) showed an inducible resistance phenotype (MIC at day 3 = 0.12 μg/ml, at day 14 = 16 μg/ml) in our study. A recent study reported that some M. abcessus strains with a functional erm(41) gene did not exhibit a clarithromycin inducible resistance phenotype: hence the presence of a functional erm(41) gene should not be used as a marker for inducible clarithromycin resistance [39]. However, our study found a rare strain with a non-functional erm (41) gene but exhibiting the inducible resistance phenotype. It is therefore unclear whether the inducible clarithromycin resistance phenotype is specific to non-massiliense subspecies or not and inducible resistant mechanisms other than the erm (41) gene might exist. Whole-genome sequencing analysis of these phenotypically diverse strains could unveil such complexity. Interestingly, we found that inducible and acquired resistance to clarithromycin in M. abscessus were interchangeable during the course of infection or colonization. This new finding was only possible because we applied DST to serial M. abscessus isolates from patients.
Amikacin is one of the antibiotics recommended for treatment of NTM infection and M. abscessus infection. It is a bactericidal antibiotic based on inhibition of protein synthesis though alteration of the 16S rRNA. Mutations of the 16S rRNA gene (rrs) at T1406A, A1408G, C1409T and G1491T are reportedly responsible for high resistance levels to amikacin [8,14,15]. We did not find any of these mutations associated with amikacin resistance in rrs, implying that other, undetected mutations may be responsible for conferring resistance in our region. A recent study reported an additional mechanism of amikacin resistance in M. abscessus. This was associated with mutations of aac(2 0 ) and eis2 [40]. High-throughput genomic analysis should provide an insight into the genetic mechanisms involved.
The strains used in our study were genotyped using multi-locus sequence-typing (MLST) (unpublished data, work under review). Based on 24 patients (62 isolates), only two clonal clusters (2 isolates from 2 patients and 8 isolates from 5 patients) were found. We recruited an additional two patients for the current study. The presence of such clusters may bias comparisons of phenotypic and genotypic DTS. We classified M. abscessus into massiliense and nonmassiliense subspecies only based on sequence deletions in erm (41). However, the MLST data (unpublished data) showed that the bolletii subspecies was not among our studied strains. Drug-susceptibility tests based on broth microdilution antibacterial assays were performed according to RAPMYCOI Sensititre's protocol. Although the CLSI guidelines do not specify the control strain or expected DST results for M. abscessus, our DST results were similar to those in a previous study of M. abscessus ATCC19977 [41].
In conclusion, M. abscessus was highly resistant to almost all antibiotics, but highly susceptible to amikacin and moderately susceptible to clarithromycin and tigecycline. Non-massiliense subspecies were associated with clarithromycin resistance whereas the massiliense subspecies tended towards tigecycline resistance. Around 75% of non-massiliense subspecies had inducible clarithromycin resistance explicable by T28C of erm (41) but inducible clarithromycin resistance might not be specific to non-massiliense subspecies only. The recognized mutation of the rrl gene (A2271GC) conferred acquired resistance to clarithromycin in only half of the strains. No mutation of the rrs gene associated with amikacin resistance in M. abscessus was found during our study.  Table. Raw data used in this study. (XLSX)