Introduction

Rapidly growing mycobacteria (RGM) infections constitute a serious public health concern worldwide, particularly in East Asia, and the proportion of RGM among nontuberculous mycobacteria (NTM) is high1. The prevalence of infections caused by the Mycobacteroides abscessus group (MAG), a major group of RGM, has increased in Japan2. Several mycobacterial species causing RGM infections have a natural resistance to several antimicrobials, rendering standard treatment regimens inefficient3. Several gene mutations related to drug susceptibility or resistance in RGM have been reported, including: erm(41) C28 sequevar, which is related to macrolide susceptibility4; rrl, which is associated with acquired resistance to macrolides4; rrs, which affects aminoglycoside resistance5; and gyrA and gyrB, which encode a quinolone resistance-determining region (QRDR) related to emerging quinolone resistance6.

Of these genes, the most important is erm(41), which is involved in the macrolide-induced resistance of MAG. When MAG is exposed to a macrolide, the erm(41) gene is expressed, and the product of this gene methylates the macrolide binding site on 23S rRNA7. This inhibits the action of the macrolide, resulting in drug resistance. To accurately evaluate this induced resistance, it is necessary to wait until the 14th day after the start of the drug susceptibility test8.

There are differences in the erm(41) gene sequence among MAG subspecies.

In M. abscessus subsp. massiliense (MMA), there is a deletion in the erm(41) gene, and its function is lost9. Therefore, MMA is macrolide-susceptible, whereas M. abscessus subsp. abscessus (MAB) and M. abscessus subsp. bolletii (MBO), which do not have this deletion, are often macrolide-induced resistant9. However, it is known that this macrolide-induced resistance is lost because of the T28C mutation in the erm(41) gene, even in the erm(41) gene without the deletion4.

Previous reports in the USA showed that there were 10 different sequevar types in the erm(41) gene, and that the sequevar types with C28 (type 2, 3, and 5) were macrolide-susceptible, whereas the other sequevar types with T28 (type 1, 4, 6, 7, 8, 9, and 10) were mostly macrolide-resistant10. Because differences in the sequence of the erm(41) gene are useful in predicting susceptibility to macrolides, the Clinical Laboratory Standards Institute (CLSI) recommends that the sequevar type of the erm(41) gene be evaluated in MAB11. However, in actual clinical practice in Japan, the erm(41) sequevar type is rarely determined, strictly due to the labor and cost required for testing, and these epidemiological data are unknown.

Susceptibility of RGM to antimicrobials remains controversial for multiple reasons. First, because of the high degree of phylogenetic similarity between different RGM, accurate species identification requires detailed genetic analysis. Previously reported large-scale antimicrobial susceptibility tests have not always identified RGM species with sufficient accuracy12. The minimum inhibitory concentration (MIC) breakpoints of only 11 antimicrobials are described in CLSI M24A-28, and the MIC values of other antimicrobials that have been measured according to the CLSI method are not sufficiently evaluated. Second, recent reports demonstrate that susceptibility to macrolides correlates with the response rate13, 14, and that the use of azithromycin, imipenem, and amikacin is associated with good therapeutic results15 for pulmonary infections caused by MAG. However, the correlation between breakpoints proposed by CLSI and treatment outcomes remains unclear for most antimicrobials in many settings of RGM infection. Gathering information regarding the MIC values of antimicrobials that can be used as therapeutic options for treating infections caused by RGM species is essential. Third, epidemiological information related to the gene mutations involved in antimicrobial resistance is scarce.

Therefore, the current study aimed to determine the MIC of 24 antimicrobial agents for clinically isolated RGM and record relevant epidemiological and genetic information to identify potential therapeutic agents.

Results

The details of 15 species that were identified are shown in Table 1. Eleven isolates, M. abscessus subsp. abscessus (MAB) (5), M. abscessus subsp. massiliense (MMA) (3), M. chelonae (2), and M. senegalense (1) grew poorly in the culture medium at five days after the start of susceptibility test, and thus we could not obtain MIC data for these isolates.

Table 1 Distribution of rapidly growing mycobacteria (RGM) species by specimen type.
Full size table

Characteristics of antimicrobial susceptibilities of RGM species

Other than MMA, which was susceptible to both amikacin and clarithromycin, MAB and M. abscessus subsp. bolletii (MBO) were susceptible to only amikacin (Table 2). Although M. fortuitum was resistant to macrolides, it was susceptible to amikacin, imipenem, fluoroquinolones, and trimethoprim/sulfamethoxazole (Table 3). Only three isolates were not susceptible to fluoroquinolones. Most M. chelonae isolates were susceptible to clarithromycin. However, the proportion of isolates intermediate and resistant to aminoglycosides, imipenem, cefoxitin, and fluoroquinolones was high. Additionally, we found that 46% of M. chelonae strains were susceptible to tobramycin (Table 3). M. mageritense isolates showed remarkably high resistance to clarithromycin and amikacin, but were susceptible to fluoroquinolones, imipenem, and cefoxitin (Table 3). The results of the antimicrobial susceptibility test and MICs for other rare RGM species are shown in Table 4 and Table S2, respectively. Amikacin and linezolid were the most effective against the 15 isolated RGM species (Table 4).

Table 2 Antimicrobial susceptibility of Mycobacteroides abscessus group (MAG) strains.
Full size table
Table 3 Antimicrobial susceptibility of major rapidly growing mycobacteria (RGM) strains other than M. abscessus group (MAG).
Full size table
Table 4 Characteristics of antimicrobial susceptibility of rapidly growing mycobacteria (RGM) species.
Full size table

We also investigated the MICs against RGM for antibacterial drugs for which CLSI did not set breakpoints. The MIC50 of sitafloxacin was the lowest among all the fluoroquinolones for all RGM species (Tables 2, 3). Except in M. fortuitum and M. wolinskyi isolates, the MIC50 of arbekacin was the lowest among aminoglycoside antimicrobials (Tables 2, 3). The MIC50 of cefmetazole was equal to or lower than that of cefoxitin for all RGM species, although the values were similar (Tables 2, 3). The MIC50 of rifabutin was lower among MAB than among MMA. The proportion of isolates with MIC ≤ 2 mg/L for rifabutin was significantly higher than that of MMA isolates [MAB: 50/178 (28.1%) vs. MMA: 23/130 (17.7%); p = 0.041] (Table 2). Faropenem had a higher MIC50 than imipenem among all RGM isolates except M. iranicum (Tables 2, 3).

Relationship between MAB erm(41) sequevar type and susceptibility to clarithromycin

The sequence of erm(41) was obtained from 180 isolates, and the relationship between MAB erm(41) sequevar and clarithromycin MIC was determined (Table 5). For the remaining three isolates, we could not obtain any sequence data. None of the MAB isolates had a truncated erm(41) sequevar, whereas 2 of 133 MMA isolates had a functional erm(41) T28 sequevar. The clarithromycin (late-reading-time [LRT]) MICs for these two isolates were 0.5 mg/L and 8 mg/L, respectively. The erm(41) gene sequences of 131 MMA isolates were identical. In this survey, the proportion of the C28 sequevar in MAB was 12.2% (22/180), all of which were type 2. Several new sequevar types were identified in our isolates; the two most common of these new isolates were named jpn1 and jpn2. These new sequevars were similar to type 10, and all isolates were resistant to clarithromycin. The single nucleotide polymorphisms of erm(41) in each sequevar are shown in Table S3. Of the 158 isolates of the T28 sequevar, 7 showed clarithromycin MIC ≤ 4 mg/L, including isolates of types 1, 6, 7, 8, and 10.

Table 5 erm(41) sequevar type and clarithromycin minimum inhibitory concentration (MIC) of M. abscessus subsp. abscessus (MAB).
Full size table

Relationship between rrl gene mutation of MAG and susceptibility to clarithromycin

Among the 37 MAG isolates with acquired macrolide resistance, the proportions with rrl mutations were 2/24 for the MAB T28 sequevar, 0/2 for the MAB C28 sequevar, 1/2 for MAB unknown, 1/1 for MBO, and 3/8 for MMA (Table 6). However, the rate of rrl mutation among MMA isolates that acquired macrolide resistance was higher than that of the MAB T28 sequevar, although not significantly (MAB T28 sequevar: 2/24 [8.3%] vs. MMA: 3/8 [37.5%]; p = 0.085).

Table 6 Frequency of rrl mutation in 37 Mycobacteroides abscessus group (MAG) strains.
Full size table

Relationship between rrs gene mutation and susceptibility to amikacin

rrs (A1408G) mutations were not found among the 73-amikacin non-susceptible (MIC ≥ 32 mg/L) isolates (MAG, M. chelonae, and M. mageritense).

Quinolone resistance of M. fortuitum and its mechanism

Of the three isolates of M. fortuitum that were resistant to ciprofloxacin, only one had a mutation in gyrA. In the mutant strain, the gyrA gene resulted in S83W amino acid substitution (TCG → TGG). None of the ciprofloxacin-susceptible isolates had mutations in gyrA and gyrB.

Discussion

In this study, we accurately identified 15 species of RGM from clinical isolates obtained from different locations around Japan. We characterized the susceptibility of these isolates to 24 antimicrobials, including tigecycline, sitafloxacin, rifabutin, and cefmetazole; none have defined MIC breakpoints in the CLSI, but they may have potential as therapeutic agents for RGM infections. We investigated not only MAG antimicrobial susceptibility, but also several gene mutations involved in antimicrobial resistance and prepared a summary of the susceptibility of the remaining 14 species of RGM.

The proportion of the C28 sequevar in MAB isolated from lower respiratory specimens (LRS) has been reported to be approximately 16‒35%4, 10, 16, 17. However, in some previous Japanese reports, the ratio of the C28 sequevar among MAB from LRS was very low at 4.2% (2/48)18. In our survey, it was 12.2% (22/180), which is higher than that in the previous report18. In Japan, it is necessary to continue to evaluate whether the proportion of the C28 sequevar in MAB is lower than those in other countries.

A previous report from the USA indicated that sequevar types 4, 6, 7, 8, 9, and 10 (all T28 sequevars) may be associated with macrolide-induced resistance10. However, similar assessments outside of the USA have not been conducted so far. Among the 180 MABs in our study, only 4 isolates with erm(41) sequevar types 6, 7, 8, and 10 were susceptible to clarithromycin. Our data were generally consistent with the previous report10. Therefore, it was suggested that these sequevars are macrolide-resistant. So far, CLSI has recommended the determination of the erm(41) sequevar type for evaluation of induced macrolide resistance in MAB11, and our results support this recommendation. Further investigations on the relationship between sequevar types and macrolide resistance in other regions are required.

The rrl gene mutation is more likely to occur in the MAB C28 sequevar and MMA than in the MAB T28 sequevar among clarithromycin-acquired resistant strains in MAG4. In our survey, we found a similar trend but could not show a significant difference. Among MAG, more than half of the macrolide-acquired resistance occurred by mechanisms other than rrl gene mutation. The exact mechanism remains to be investigated.

Additionally, none of the amikacin non-susceptible isolates in our survey had the rrs gene mutation. A previous French study of antimicrobial susceptibility in 165 isolates of MAG showed that 7/8 strains with amikacin MIC > 64 mg/L had a rrs A1408G gene mutation16, which suggested that amikacin MIC > 64 mg/L is a criterion to suspect amikacin-acquired resistance16. In our survey, only one isolate of MAG showed MIC > 64 mg/L, and none of the isolates showed rrs mutation. MAG isolated in Japan may have fewer amikacin-acquired resistant isolates than those isolated in France.

As reported previously18, 19, M. fortuitum was resistant to clarithromycin; however, it was susceptible to aminoglycosides, carbapenems, and fluoroquinolones in our study. Previous reports suggest that, in M. fortuitum, a serine residue at the 83rd position of gyrA constitutes QRDR and contributes to susceptibility to fluoroquinolones compared with other NTMs6. However, to date, only one report has shown quinolone resistance due to mutations in gyrA19. There has been no report of mutations in a serine residue at the 83rd position of gyrA. Fluoroquinolone resistance was found in 3 of 85 (3.5%) isolates in our study, and the S83W amino acid substitution was present in one of the three isolates. Our result also suggests that fluoroquinolone resistance can occur based on genetic changes other than QRDR mutations, and it is necessary to clarify the resistance mechanism in the future. In Japan, fluoroquinolones are being overused20, and there is a concern regarding the increase of fluoroquinolone-resistant isolates in M. fortuitum. Because M. fortuitum shows induced resistance to macrolides, fluoroquinolones play an important role in the treatment of M. fortuitum infections as an oral antibiotic. There is a great concern regarding treatment efficacy with the increase in resistant isolates.

Among M. chelonae isolates, resistance to clarithromycin was found in approximately 10% of isolates, consistent with previous reports18, 21. Previous reports seem to indicate regional variability in tobramycin susceptibility, ranging from 54% in the UK21 to 83% and 17% in Japan18, 22. In our study, approximately 40% of the strains were tobramycin-susceptible, an intermediate value between the values reported by the two previous reports from Japan. In addition, no rrs mutations were found in amikacin non-susceptible isolates. Arbekacin may be a potential therapeutic for isolates that are less susceptible to amikacin and tobramycin.

M. peregrinum was susceptible to most of the tested antimicrobials. M. mageritense isolates were resistant to clarithromycin, as has been previously reported23, and showed a low susceptibility to amikacin, although none had a rrs gene mutation (3 isolates showed an amikacin MIC > 64 mg/L). The mechanism of M. mageritense resistance to amikacin remains to be investigated. Conversely, it showed good susceptibility to quinolones, cefoxitin, and linezolid.

There are few reports on antimicrobial susceptibility for other rare RGM species using a sufficiently high number of clinical isolates. There is only one study involving M. mucogenicum and M. immunogenum reporting that most of the isolates were susceptible to linezolid, amikacin, and trimethoprim/sulfamethoxazole, while showing a poor susceptibility to clarithromycin24. In our study, although the number of isolates was small, we could show the tendency of antimicrobial susceptibility for rare RGM species. These rare RGM species tended to be susceptible to linezolid, quinolones, and trimethoprim/sulfamethoxazole.

Although there have been no reports regarding the MIC of arbekacin in RGM, this antimicrobial showed the lowest MIC among the aminoglycosides for almost all RGM species in this study (Tables 2, 3). The MIC50 value of sitafloxacin is reported to be lower than that of other fluoroquinolones in MAG, M. fortuitum, and M. chelonae18. However, in this study, we showed that the effect of sitafloxacin was similar on the 15 RGM species (Tables 2, 3). Cefmetazole and cefoxitin, cephamycin-based antimicrobials, had similar MICs, consistent with previous reports25, 26 (Tables 2, 3). In countries such as Japan, when patients cannot be administered cefoxitin, cefmetazole may be an option for RGM treatment. In recent years, rifabutin has attracted attention as an oral treatment for MAG27, 28, but, so far, there have been few reports of MICs measured by micro-dilution using cation-adjusted Mueller–Hinton broth medium28. Here, we have not only measured rifabutin MICs for many isolates using this standard method, but also showed that MICs were lower for MAB than for MMA (Table 2). MAB has a very high resistance rate not only to clarithromycin but also to fluoroquinolone; thus, finding an alternative orally administrated therapeutic option is essential. A detailed evaluation is required in the future to determine whether rifabutin will be an effective orally administered therapeutic option.

There are some limitations to our study. It was unclear whether there was prior administration of antibacterial drugs before susceptibility testing for all isolates. Some of the RGM species isolated in this study were rarely isolated to evaluate drug susceptibility. However, despite these limitations, our study reveals important epidemiological information about RGM in Japan and suggests several drugs that can be investigated as new treatment candidates. It is therefore necessary to accumulate and evaluate data from a larger set of samples and to verify the correlation between the actual therapeutic effect and the MIC values of these drugs in clinical trials.

We showed antimicrobial susceptibility profiles of 15 RGM species isolated in Japan. Amikacin and linezolid were the most effective against the 15 isolated RGM species. Arbekacin, sitafloxacin, and cefmetazole may be possible therapeutic options for RGM infections. Based on the MIC values, rifabutin may be more potent in the treatment of MAB than MMA. Clinical trials are needed in the future to validate our findings.

Methods

Clinical isolates

From January 2012 to March 2019, 509 clinical specimens [409 LRS, 87 non-lower respiratory specimens, and 13 unknown] isolated from patients in Japan (one specimen per patient) were included in this study. From the specimens, 403 strains were isolated at BioMedical Laboratories (BML), Inc., a major clinical laboratory, and 106 were isolated at 45 hospitals in Japan.

PCR and sequence analysis

Bacterial genomic DNA was extracted using ISOPLANT II (NIPPON GENE CO., LTD, Japan). The three housekeeping genes, hsp6529, rpoB30, and sodA31, of each isolate were sequenced for RGM species identification. For MAG, an additional PCR-based typing scheme32 was used for subspecies identification, and the erm(41) sequence type was determined9, 10. The sequence of rrl from MAG strains exhibiting acquired macrolide resistance was also determined33. Similarly, the sequences of the rrs gene from MAG, M. chelonae, and M. mageritense, which are amikacin non-susceptible isolates, were analyzed5. The sequences of the gyrA and gyrB genes encoding the QRDR were elucidated for all M. fortuitum isolates. All PCR procedures were performed as described previously5, 9, 29,30,31,32,33, and the primers used are shown in Table S1.

Antimicrobial susceptibility test

All strains were subcultured on trypticase soy agar with 5% sheep blood (Becton, Dickinson and Company, New Jersey) at 35 °C for 3‒5 days. Antimicrobial susceptibility testing was performed per the recommendations in the CLSI M24A-2 at 30 °C8. A nephelometer (VITEK DENSICHEK, bioMerieux, France) was used to standardize the inoculum density (0.5 McFarland standard). The MICs of 24 antimicrobial agents (tigecycline, linezolid, clarithromycin, azithromycin, arbekacin, amikacin, gentamycin, tobramycin, imipenem, doripenem, faropenem, levofloxacin, sitafloxacin, ciprofloxacin, moxifloxacin, cefmetazole, cefoxitin, ceftriaxone, cefepime, ethambutol, rifabutin, minocycline, amoxicillin/clavulanic acid, and trimethoprim/sulfamethoxazole) were measured by the micro-dilution method using cation-adjusted Mueller–Hinton broth medium (Becton, Dickinson and Company)8. The MICs of clarithromycin and azithromycin were read two times to detect induced resistance. Positive growth of the control between days 3 and 5 was defined as early-reading-time. Inducible macrolide resistance was determined on day 14 and defined as LRT. Repetition of MIC measurement, as recommended by guidelines, was performed without exception.

Statistical analysis

Statistical analyses were performed using GraphPad Prism ver. 8.2.0 for Windows (GraphPad Software, San Diego, CA, USA). Data were compared using the Chi-square test for categorical variables, whereas Fisher’s exact test was used where the assumption of the Chi-square test was violated.