ABSTRACT
Very few studies have examined drug susceptibility of Mycobacterium kansasii, and they involve a limited number of strains. The purpose of this study was to determine drug susceptibility profiles of M. kansasii isolates representing a spectrum of species genotypes (subtypes) with two different methodologies, i.e., broth microdilution and Etest assays. To confirm drug resistance, drug target genes were sequenced. A collection of 85 M. kansasii isolates, including representatives of eight different subtypes (I to VI, I/II, and IIB) from eight countries, was used. Drug susceptibility against 13 and 8 antimycobacterial agents was tested by using broth microdilution and Etest, respectively. For drug-resistant or high-MIC isolates, eight structural genes (rrl, katG, inhA, embB, rrs, rpsL, gyrA, and gyrB) and one regulatory region (embCA) were PCR amplified and sequenced in the search for resistance-associated mutations. All isolates tested were susceptible to rifampin (RIF), amikacin (AMK), co-trimoxazole (SXT), rifabutin (RFB), moxifloxacin (MXF), and linezolid (LZD) according to the microdilution method. Resistance to ethambutol (EMB), ciprofloxacin (CIP), and clarithromycin (CLR) was found in 83 (97.7%), 17 (20%), and 1 (1.2%) isolate, respectively. The calculated concordance between the Etest and dilution method was 22.6% for AMK, 4.8% for streptomycin (STR), 3.2% for CLR, and 1.6% for RIF. For EMB, INH, and SXT, not even a single MIC value determined by one method equaled that by the second method. The only mutations disclosed were A2266C transversion at the rrl gene (CLR-resistant strain) and A128G transition at the rpsL gene (strain with STR MIC of >64 mg/liter). In conclusion, eight drugs, including RIF, CLR, AMK, SXT, RFB, MXF, LZD, and ethionamide (ETO), showed high in vitro activity against M. kansasii isolates. Discrepancies of the results between the reference microdilution method and Etest preclude the use of the latter for drug susceptibility determination in M. kansasii. Drug resistance in M. kansasii may have different genetic determinants than resistance to the same drugs in M. tuberculosis.
INTRODUCTION
Mycobacterium kansasii, an opportunistic pathogen and causative agent of pulmonary and extrapulmonary infections, is one of the six most frequently isolated nontuberculous mycobacterial (NTM) species worldwide (1). To date, seven M. kansasii subtypes (I to VII), along with two intermediate (I/II) and atypical (IIb) types, have been identified (2, 3). Most of the disease-related strains belong to type I, while the others have usually been linked to environmental sources. Infections due to NTM, including M. kansasii, have been on the rise recently, which is due to increasing numbers of susceptible hosts, progress in microbial identification, and greater awareness of NTM disease (4).
Studies on drug susceptibility of M. kansasii are still very sparse, and those few available have evaluated a limited number of strains (5–12). For drug susceptibility testing (DST) of M. kansasii, as of other mycobacteria, the Clinical and Laboratory Standards Institute (CLSI) recommends the microdilution method in Mueller-Hinton medium. Optionally, agar proportion assays can be performed (13).
Currently, only susceptibility tests for rifampin (RIF) and clarithromycin (CLR) are advocated by the CLSI for M. kansasii; only in the rare event of resistance to RIF is it advised that researchers determine susceptibility to isoniazid (INH), ethambutol (EMB), streptomycin (STR), amikacin (AMK), co-trimoxazole (SXT), rifabutin (RFB), moxifloxacin (MXF), linezolid (LZD), and ciprofloxacin (CIP) (13, 14). M. kansasii isolates are usually susceptible to those agents if the current CLSI breakpoints are applied (14, 15).
Whereas studies on DST are lacking, those concerning genetic determinants of drug resistance in M. kansasii are virtually absent. So far, only three studies have addressed this issue, with a focus on RIF and CLR resistance. Similar to M. tuberculosis, mutations in rpoB and rrl genes were identified as a major mechanism behind resistance to these agents in M. kansasii (12, 16, 17).
The purpose of this study was to determine drug susceptibility profiles of M. kansasii representing a spectrum of species genotypes (subtypes) with broth microdilution and Etest methods. Furthermore, a panel of nine genetic loci associated with drug resistance in M. tuberculosis was screened for the presence of mutations in drug-resistant M. kansasii isolates.
RESULTS
Broth microdilution.MIC ranges, median MICs, MIC50s, and MIC90s are all summarized in Table 1. All of the isolates tested were susceptible to RIF, AMK, SXT, RFB, MXF, and LZD. Among drug-resistant isolates, 83 (97.6%) were resistant to EMB, 17 (20%) to CIP, and 1 (1.2%) to CLR. The highest detected MIC for INH was 4 mg/liter and was evidenced in five (5.9%) M. kansasii isolates. The MICs for STR and DOX varied widely (1 to >64 mg/liter and 1 to >16 mg/liter), whereas the ethionamide (ETO) MICs were within a narrow range of <0.3 to 1.2 mg/liter. Twelve (12/20; 60%) isolates with the highest MICs for DOX (≥16 mg/liter) were also resistant to CIP.
MIC values obtained for different M. kansasii subtypes (n = 23) and clinical isolates from Poland (n = 62) with broth microdilution methodc
No statistically significant differences were observed between MICs of different drugs and presence or absence of M. kansasii disease (P value of >0.05). The only two exceptions were INH and ETO, whose MICs were significantly lower in patients without M. kansasii disease (P value of <0.03).
Etest.The Etest results of the 62 M. kansasii clinical isolates are presented in Table 2. All isolates tested had SXT MICs of >32/608 mg/liter (trimethoprim/sulfamethoxazole), i.e., at least 16 times that in the dilution method.
MIC values obtained for clinical isolates from Poland (n = 62) with Etest methodb
The median Etest MICs for RIF and EMB were 0.008 mg/liter and 0.5 mg/liter, 30 times lower than the median MICs for these two drugs according to the dilution method. Only two isolates had Etest MICs above the breakpoint level (4 mg/liter) defined for the dilution method.
The Etest MICs of CLR were within a broad range of values (<0.016 mg/liter to >256 mg/liter), quite similar to what was observed for the dilution method (MIC range, <0.06 to >64 mg/liter).
Higher activities of AMK and STR than KAN were evidenced by the MIC50 and MIC90 values (for AMK and STR, 1.5 and 4 mg/liter, respectively, as opposed to 2 and 12 mg/liter for KAN).
Overall, for all seven drugs tested with both microdilution and Etest methods, the MIC values determined with these two methods were significantly different (P value of ≤0.001) (Fig. 1).
Scattergram results comparing MICs determined by Etest with those determined by the broth microdilution reference method (log2 scale).
Mutation profiling.The results of sequence analysis are summarized in Fig. S1 to S9 in the supplemental material. Since all but two (type V) isolates were identified as EMB resistant with the dilution method, sequence analysis of the EMB resistance-associated loci (embB and embCA) was limited to only 8 isolates, representing 8 distinct M. kansasii types. For M. kansasii type IV (EMB resistant) and V (EMB susceptible), the sequences at both loci were identical. All M. kansasii subtypes had a G918C nucleotide change, translated into an M306I substitution, in the EMB resistance-determining region (ERDR), as it is referred to in M. tuberculosis. This mutation was always coupled with G406P (GGC1216-1218CCG or GGC1216-1218CCC) and M423I (G1269C) alterations (Table S11).
The 17 CIP-resistant isolates (type I) had no mutations in the quinolone resistance-determining region (QRDR) of the gyrA and gyrB loci compared with CIP-susceptible isolates (i.e., with the lowest CIP MICs).
The only detected CLR-resistant isolate (type I) differed from its susceptible counterparts by a single mutation (A2266C) in the rrl gene.
Five isolates (type I) with the highest STR MICs (≥16 mg/liter) had their rrs and rpsL gene sequences determined. Upon comparison with the corresponding sequences of the low-MIC isolates (≤2 mg/liter), a single polymorphism in the rpsL gene (A128G and K42R) was detected in one of the high-MIC isolates.
Five isolates (type I) with the highest INH MIC (4 mg/liter) were screened for mutations in the inhA and partial katG gene, with the corresponding sequences of the five low-MIC (1 mg/liter) isolates as a reference. All 10 of these isolates shared identical sequences.
DISCUSSION
The scant knowledge about drug susceptibility in M. kansasii and molecular determinants of drug resistance in this species were behind the undertaking of this study. The data reported on drug resistance in M. kansasii in the available literature were summarized in Table 3.
Reported drug resistance of M. kansasii isolates
Drug susceptibility profiles established with broth microdilution.The four drugs which currently form the core of treatment schemes for M. kansasii infections are RIF, EMB, INH, and CLR (13, 15).
In this study, resistance to RIF was not detected among the isolates, whereas all of the isolates collected in Poland were resistant to EMB (two isolates susceptible to EMB were collected either in Germany or in the Netherlands). Previously described levels of resistance to those drugs, established with the microdilution method, varied widely from 1.9% to 56.4% for RIF and 0% to 94% for EMB (7, 8, 10, 12, 18).
Although there is no consensus on the INH breakpoint for M. kansasii, the most commonly used value has been 5 mg/liter. The frequency of INH resistance with this breakpoint was 2.9% in Spain (8) and 8% in Brazil (7). However, in a study from the Netherlands, a breakpoint set at 1 mg/liter was used, yielding all M. kansasii isolates INH resistant (10). The highest MIC of INH, detected in 5 (5/85; 5.9%) isolates in this study, was 4 mg/liter. According to the CLSI, INH MICs for untreated M. kansasii isolates are between 0.5 and 5 mg/liter, thus a concentration of 1 mg/liter for testing strains of M. tuberculosis should not be applied (13).
In patients with RIF-resistant M. kansasii disease a three-drug regimen should be administered, with a macrolide (CLR or azithromycin), fluoroquinolone (FQ; MXF), and SXT or STR (15, 19). In the case of resistance to any of these drugs, alternative agents should be considered, including aminoglycosides other than STR, tetracyclines, LZD or ETO, and RFB. Exclusion of INH and EMB in RIF-resistant cases is motivated by the observed increase in INH and EMB MICs in isolates with acquired RIF resistance (19).
All Polish isolates were found to be susceptible to MXF, SXT, RFB, CLR, and LZD. The frequencies of resistance to these drugs greatly varied in previous studies, ranging from 0 to 40.5% for MXF, 8 to 18.9% for SXT, and 0 to 50% for RFB (7–10, 12, 18, 20).
Several authors have shown high in vitro activity of CLR and LZD against M. kansasii, with less than 1% of isolates being resistant (7–10, 15, 18, 21). In one study from China, a third (32.1%) and a fifth (20.5%) of the M. kansasii isolates were described as LZD and CLR resistant, respectively (12).
High activity of ETO was evidenced in our study, with the highest detected MIC being 1.2 mg/liter. This is in line with a study by da Silva Telles et al., where only 5% of the isolates were ETO resistant, with an assumed breakpoint value of 4 mg/liter (7).
A quarter (25.8%) of our samples showed resistance to CIP. Similar or higher rates of CIP resistance (15 to 66%) had been reported earlier (7, 9, 10, 18).
Interestingly, more than two-thirds (70%) of our CIP-resistant isolates showed high DOX MICs (≥16 mg/liter). The cross-resistance between quinolones (CIP) and tetracyclines (DOX) had previously been described for several bacterial species (e.g., Escherichia coli or Enterobacter aerogenes) but, to our knowledge, not for mycobacteria. A hypothesized mechanism for this association is the broad substrate specificity of the FQ efflux systems, capable of extruding various drugs, including tetracyclines (22). Efflux pumps are important drivers of drug resistance in NTM (23).
As for aminoglycosides, resistance to AMK has seldom been reported in M. kansasii, with the percentage of resistant isolates not exceeding 6% (not detected in our study) (7, 9, 12, 18). The high (54%) frequency of AMK resistance was reported only in the Netherlands, and this might have been attributed to the lower breakpoint applied (5 mg/liter versus 32 mg/liter) (10). According to previous studies, STR showed much less activity than AMK against M. kansasii. With a breakpoint of 10 mg/liter, resistance to STR was 14% in Brazil (7) and 23% in the United States (5). In our study, nearly one-fourth (23%) of the isolates had STR MICs of >10 mg/liter.
Genotype-specific differences in phenotypes.Exposure to drug selection pressure may predispose human disease-associated M. kansasii types (I and II) to a drug-resistant phenotype. Li et al. showed CLR resistance to be associated with M. kansasii type I (12). An interesting observation from our study was that only type V M. kansasii isolates were susceptible to EMB. This difference may relate to different cell wall compositions, rendering the bacilli more sensitive to EMB. Apart from that, and as demonstrated by others, no significant differences were found between drug susceptibility profiles and disease-related M. kansasii subtypes except for those for ETO (7). A subtle correlation between INH and ETO and M. kansasii disease was found in our study. The median MICs of both of these drugs were significantly higher in patients with M. kansasii disease (P value of <0.03).
Etest in susceptibility profiling.Due to good correlation between drug susceptibility results and clinical response, microdilution in Mueller-Hinton medium is the CLSI-recommended drug susceptibility platform for five of the most frequently isolated and clinically relevant NTM species, including M. kansasii, M. avium, M. xenopi, M. fortuitum, and M. abscessus (13, 14). Since the Etest has only sporadically been used for NTM, its laboratory and clinical usefulness is doubtful.
An important purpose of this study was to explore the congruence of the results obtained with the reference dilution and Etest methods. Those few studies which addressed this issue in M. abscessus, M. fortuitum, and M. chelonae produced conflicting conclusions. Whereas Woods et al. (24) observed a high concordance, at a rate of 80% or 90%, between the MIC values of CLR, AMK, and CIP obtained with the Etest and microdilution methods, the others found the Etest MICs of those drugs differed up to 16-fold from those calculated using the proportion method, with overall concordance rates of 72% to 87% for AMK, 63 to 77% for CLR, and 36 to 78% for CIP (25, 26).
For M. kansasii, only one study has looked at the MIC values measured by the agar proportion and Etest methods (on Löwenstein-Jensen medium). The results were in agreement in 95 to 100% of cases (27).
Our study is the first to compare the reference broth microdilution method with the Etest for M. kansasii. In general, the results from these two methods were widely discrepant for EMB, INH, SXT, RIF, CLR, and STR, with calculated MIC level concordance below 5%. It was somewhat higher, but still low (22.6%), for AMK.
DST methods, including broth microdilution, agar proportion, and Etest, vary by the medium used, supplements and inoculum added, incubation conditions, and endpoint interpretations (4). All of those features presumably influence the test outcomes.
Genetic determinants of drug resistance.The genetic background of drug resistance in NTM, including M. kansasii, has only anecdotally been addressed and largely explained through extrapolation of findings from studies on M. tuberculosis. The only two mutations possibly associated with drug resistance were A2266C in the rrl gene and A128G (K42R) in the rpsL gene, found in single M. kansasii isolates resistant to CLR, and with the highest STR MICs established with broth microdilution (>64 mg/liter). Mutations in the rrl gene have been associated with high levels of macrolide resistance among M. kansasii and M. avium isolates, with mutations at rrl codon 2058 or 2059 found with frequencies of 56.2% (9/16) and 92.3% (24/26), respectively (12, 28). In contrast, only 1.4% (2/140) of CLR-resistant M. abscessus isolates had mutations at this locus (29). The A128G (K42R) substitution is the leading alteration among STR-resistant M. tuberculosis isolates, with a frequency of detection of 24 to 64% (30–32). However, it has yet to be described in NTM (33).
In M. tuberculosis, most of the EMB resistance-associated mutations are located in the embB gene (34). In our study, all M. kansasii types harbored M306I, G406P, and M423I amino acid substitutions, frequently described in EMB-resistant M. tuberculosis strains but unreported in NTM, including M. kansasii, M. abscessus, M. fortuitum, and M. avium (35, 36). As these alterations were found in both EMB-resistant (types I to IV and VI) and EMB-susceptible (type V) isolates, they cannot be considered specific for EMB resistance in M. kansasii. However, they might be responsible for reduced susceptibility of M. kansasii to EMB, since EMB-susceptible isolates presented MICs around the breakpoint level (4 mg/liter), as opposed to EMB-resistant isolates with MICs of 8 to >16 mg/liter.
The role of mutations in gyrA and/or gyrB in FQ resistance determination in NTM is uncertain. A study on M. abscessus showed that resistance to FQ may, to some extent, relate to mutations in the same codons as those responsible for FQ resistance in M. tuberculosis (A83 in gyrA and R447 and D464 in gyrB) (37). However, only 3.7% (2/54) of FQ-resistant M. abscessus isolates were mutated at these loci compared to sequences of susceptible isolates (30). Here, 17 M. kansasii isolates resistant to CIP did not harbor any changes in the gyrA and gyrB genes.
The key role in developing resistance to INH in M. tuberculosis is played by mutations in the katG gene, with reported frequencies of 67.7% to 95% among INH-resistant strains (38). To our knowledge, literature devoted to the role of katG or inhA in the formation of the INH-resistant phenotype among NTM is absent.
Drug resistance in NTM might be acquired (drug induced) through specific mutations in a number of different genetic loci or could be preexistent (intrinsic), mediated by the cell envelope and active efflux. For instance, low permeability of the mycobacterial cell wall shields the bacilli from environmental stress and contributes to their drug-resistant phenotype. Therefore, enzymes involved in cell wall integrity (e.g., MurA, MurB, Ldt, PonA1, and PonA2) might play an important role in the development of drug resistance (39). NTM also have a broad repertoire of multidrug efflux pumps, such as P55, Tap, LfrA, and EfpA, involved in the extrusion of STR (P55, Tap), INH (P55, EfpA), RMP (P55), β-lactams, and FQs (LfrA) (4). Among other plausible mechanisms of intrinsic drug resistance in NTM are porin channels, enzymatic modification or degradation of antibiotics, and dormancy of mycobacterial cells (39).
To get better insight into the molecular mechanisms of drug resistance in M. kansasii, further investigations, involving whole-genome sequencing, are needed.
Concluding remarks.In conclusion, this study brings two important advantages to the study of M. kansasii. First, it provides DST results for a wide panel of antimycobacterials by means of two different methodologies, i.e., reference broth microdilution and Etest. Second, it gives a snapshot of the genetic background of drug-resistant and drug-susceptible phenotypes in M. kansasii.
Overall, the results presented here show high activity of RIF, CLR, AMK, SXT, RFB, MXF, LZD, and ETO against M. kansasii isolates. Discrepancies of the results between the reference microdilution method and Etest precludes the use of the latter for drug susceptibility determination in M. kansasii. Finally, drug resistance in M. kansasii may have genetic determinants somewhat similar to those of resistance to the same drugs in M. tuberculosis for CLR, STR, and EMB and others for INH and CIP. Further studies are required to pinpoint the specific mutations or mechanisms conferring drug resistance in M. kansasii.
MATERIALS AND METHODS
Isolates.A total of 85 M. kansasii isolates were included in the study. Within this number were 82 strains representing six subtypes (type I, n = 67; type II, n = 7; type III, n = 3; type IV, n = 2; type V, n = 2; type VI, n = 1) and three representing intermediate (type I/II, n = 1) and atypical (type IIb, n = 2) types. The isolates were originally collected from Poland (n = 62), the Netherlands (n = 10), the Czech Republic (n = 4), Germany (n = 3), South Korea (n = 3), Spain (n = 2), Belgium (n = 1), and Italy (n = 1). Also, the M. kansasii ATCC 12478 type strain was used. The isolates were categorized as representing (or not) M. kansasii disease according to the American Thoracic Society (ATS) diagnostic criteria (15). Since no data on patient management and long-term treatment outcomes were available, they were excluded from the analysis.
DNA extraction and genotyping.Genomic DNA was extracted using the cetyl-trimethyl-ammonium bromide (CTAB) method. For PCR-restriction enzyme analysis (PCR-REA) genotyping, protocols described by Telenti et al. and Bakuła et al. were used (40, 41).
Broth microdilution.Drug susceptibility testing using the broth microdilution method was performed on all 85 isolates. A panel of 13 drugs, including RIF, CLR, INH, EMB, STR, AMK, SXT, RFB, MXF, LZD, CIP, doxycycline (DOX), and ethionamide (ETO), using Sensititre SLOMYCO plates (TREK Diagnostic Systems, Cleveland, USA) was tested. Inocula were prepared according to the CLSI protocol (13) and the manufacturer's instructions, and the SLOMYCO plates were incubated at 36°C until the controls showed sufficient growth. The MICs were determined visually using an inverted mirror and read as the lowest concentration of the antibiotic yielding no visible growth. The CLSI-recommended breakpoints were applied: RIF, >1 mg/liter; CLR, >16 mg/liter; EMB, >4 mg/liter; AMK, >32 mg/liter; SXT, >2/38 mg/liter; RFB, >2 mg/liter; MXF, >2 mg/liter; LZD, >16 mg/liter; CIP, >2 mg/liter. In the absence of breakpoints for INH, STR, DOX, and ETO, only MIC values were reported for these drugs, as recommended by the CLSI (13). M. kansasii ATCC 12478, M. avium ATCC 700898, and M. peregrinum ATCC 700686 reference strains were used as controls.
Etest.For 62 M. kansasii clinical isolates from Poland, MIC values were also determined using the Etest method according to the manufacturer's instructions (bioMérieux, Durham, NC). Eight antimycobacterial drugs, available under the Etest strip formula, were employed, i.e., RIF, CLR, INH, EMB, STR, AMK, SXT, and kanamycin (KAN). The M. kansasii ATCC 12478 reference strain was used as a control.
DNA sequencing and data analysis.The search for mutations possibly associated with drug resistance was performed by PCR amplification and sequencing of nine genetic loci, i.e., rrl (CLR), katG and inhA (INH), embB and embCA (EMB), rpsL and rrs (STR), and gyrA and gyrB (CIP). All primers used for the amplification and sequencing were newly designed (see Table S1 in the supplemental material). The PCRs were performed with a TopTaq master mix kit as recommended by the manufacturer (Qiagen, Hilden, Germany).
PCR amplicons, after purification (Clean-Up; A&A Biotechnology, Gdynia, Poland), were sequenced in both directions using the same primers as that for PCR amplification and, when needed, with additional sequencing primers (Table S1). Sequence analysis was performed with MEGA8 tool using ClustalW and visualized with ESPript (42, 43).
Screening for mutations associated with resistance to all drugs but EMB was performed on M. kansasii type I isolates which, upon using the microdilution method, were declared resistant based on the published breakpoints or showed, for a given drug, the five highest MIC values if no breakpoints were available.
Sequencing of the EMB resistance-associated loci (embB and embCA) was carried out for 8 isolates, each representing a different subtype, declared EMB resistant (n = 7) or EMB susceptible (n = 1) upon using the dilution method.
The sequences obtained for M. kansasii isolates all were compared to sequences of the respective genetic loci of the M. kansasii ATCC 12478 and M. tuberculosis H37Rv reference strains and to isolates with the five lowest MICs.
Statistical analysis.The MIC50, MIC90, and geometric means (GMs) of the MICs were calculated. Nonparametric statistical methods were used. Mann-Whitney U and chi-square tests were applied for correlations between MICs and interpretation with disease status according to the ATS diagnostic criteria. The differences in the MICs between the two drug susceptibility methods employed (microdilution versus Etest method) were assessed by the Wilcoxon test. Statistical significance was set at a P value of <0.05. All calculations were performed using IBM SPSS v.23 software.
Accession number(s).All nucleotide sequences were deposited in the GenBank database (http://www.ncbi.nlm.nih.gov/GenBank/) under the accession numbers indicated in Fig. S1 to S8 in the supplemental material.
ACKNOWLEDGMENTS
This work was supported by the National Centre for Research and Development LIDER Programme (LIDER/044/457/L-4/12/NCBR/2013).
We thank Vit Ulmann, Institute of Public Health, Ostrava, Czech Republic, Miren Josebe Unzaga-Baranano, Clinical Microbiology and Infection Control, Basurto Hospital, Bizkaia, Spain, and Won-Jung Koh, Division of Pulmonary and Critical Care Medicine, Department of Medicine, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, South Korea, for kindly providing the strains for this study. We also thank Agata Podpora for assisting with the Etest analysis.
FOOTNOTES
- Received 28 August 2017.
- Returned for modification 4 December 2017.
- Accepted 20 January 2018.
- Accepted manuscript posted online 5 February 2018.
Supplemental material for this article may be found at https://doi.org/10.1128/AAC.01788-17.
- Copyright © 2018 American Society for Microbiology.