Plasmid-mediated drug resistance in mycobacteria: the tip of the iceberg?

ABSTRACT

Macrolides, such as clarithromycin, are crucial in the treatment of nontuberculous mycobacteria (NTM). NTM are notoriously innately drug resistant, which has made the dependence on macrolides for their treatment even more important. Not surprisingly, resistance to macrolides has been documented in some NTM, including Mycobacterium avium and Mycobacterium abscessus, which are the two NTM species most often identified in clinical isolates. Resistance is mediated by point mutations in the 23S ribosomal RNA or by methylation of the rRNA by a methylase (encoded by an erm gene). Chromosomally encoded erm genes have been identified in many of the macrolide-resistant isolates, but not in Mycobacterium chelonae. Now, Brown-Elliott et al. (J Clin Microbiol 61:e00428-23, 2023, https://doi.org/10.1128/JCM.00428-23) describe the identification of a new erm variant, erm(55), which was found either on the chromosome or on a plasmid in highly macrolide-resistant clinical isolates of M. chelonae. The chromosomal erm(55) gene appears to be associated with mobile elements; one gene is within a putative transposon and the second is in a large (37 kb) insertion/deletion. The plasmid carrying erm(55) also encodes type IV and type VII secretion systems, which are often linked on large mycobacterial plasmids and are hypothesized to mediate plasmid transfer. While the conjugative transfer of the erm(55)-containing plasmid between NTM has yet to be demonstrated, the inferences are clear, as evidenced by the dissemination of plasmid-mediated drug resistance in other medically important bacteria. Here, we discuss the findings of Brown-Elliott et al., and the potential ramifications on treatment of NTM infections.

COMMENTARY

The genus Mycobacterium currently comprises 195 species, with at least 12 subspecies, according to the International Code of Nomenclature of Procaryotes (https://www.bacterio.net/genus/mycobacterium, accessed 20 June 2023). Historical classifications focused on Mycobacterium tuberculosis and Mycobacterium leprae, the causes of tuberculosis and leprosy, respectively, while all other mycobacterial species were inaptly classified as “nontuberculous” mycobacteria (NTM). Many of these NTM cause various types of diseases including pulmonary infection, skin and soft-tissue infection, lymphadenitis, and septicemia. Despite their non-threatening name, NTM are the cause of the majority of mycobacterial infections in the United States. The species most frequently isolated in clinical mycobacteriology laboratories include the Mycobacterium avium complex (MAC; primarily, M. avium and Mycobacterium intracellulare), Mycobacterium abscessus subsp. abscessus, Mycobacterium kansasii, Mycobacterium gordonae (rarely pathogenic), Mycobacterium xenopi, Mycobacterium fortuitum, and Mycobacterium chelonae (Dr. Salfinger, unpublished data). In contrast to M. tuberculosis complex isolates, NTM are not required to be reported nationally; this makes an accurate assessment of the prevalence of NTM infections even more difficult to calculate (1, 2). In support of this, Prevots et al. reported on NTM lung disease prevalence at four integrated healthcare delivery systems and found that 80% of the isolates were MAC and 12% were M. chelonae or M. abscessus (3). The Cystic Fibrosis Patient Registry lists 9,796 individuals who had a mycobacterial culture performed in 2021. Nine hundred and thirty-seven (9.6%) had a mycobacterial species isolated one or more times, with the majority identified as MAC (46%) or M. chelonae or M. abscessus (35%) (https://www.cff.org/sites/default/files/2021-11/Patient-Registry-Annual-Data-Report.pdf). As whole-genome sequencing becomes more standardized, we anticipate that the number of identified NTM clinical isolates will continue to grow and NTM will become an increasing burden on public health.

Nontuberculous mycobacteria are opportunistic human pathogens that are widespread in the environment. The basis for their widespread presence in soils and natural and human-engineered waters lies primarily in their innate resistance to disinfectants, ability to form biofilms, and adaptability to fluctuating environmental conditions (4). These characteristics also mean that human exposure to NTM is frequent and partly explains their increasing prevalence in infections of immunocompromised patients and in patients with chronic lung disease (4). This increase in prevalence, combined with the innate drug resistance of NTM, has resulted in guidelines requiring antimicrobial susceptibility testing in samples that are considered clinically significant (5). Clinical Laboratory Standards Institute guidelines describe phenotypic assays and methods to detect molecular drug resistance markers. However, these are limited to a few NTM species and genes associated with resistance to aminoglycoside and acquired macrolide resistance (rrl gene) for MAC and M. abscessus, and inducible-macrolide resistance [erm(41)] for M. abscessus subspp. abscessus and bolletii (6). It is recommended that M. abscessus subsp. abscessus isolates be tested for the presence of the erm(41) gene to determine clarithromycin resistance (7).

A recent meta-analysis evaluated the success rate of macrolide-containing combination therapy in M. abscessus lung disease: 34% of patients with new M. abscessus subsp. abscessus infection and 54% of patients with M. abscessus subsp. massiliense infection achieved sustained sputum culture conversion (8). Thus, this combination therapy has enormous potential for treating NTM infections and is a therapy that should be rigorously implemented to ensure that drug-resistant variants are prevented. We note that full identification of the subspecies level for M. abscessus is clinically important, as antimicrobial susceptibility varies between subspecies. For example, Lange et al. noted that M. fortuitum, and some M. abscessus subspecies, have functional erm genes (encoding macrolide resistance), while M. chelonae was thought not to contain an erm gene, so that macrolides were considered fully active against this latter species (9, 10). However, this general statement has already changed because researchers at the University Health Science Center in Tyler, Texas have discovered a new erm gene in M. chelonae.

Brown-Elliott et al. describe a collection of M. chelonae patient isolates that have acquired inducible resistance to macrolides (up to 16 µg/mL) (10). Whole-genome sequencing of one isolate, with high levels of macrolide resistance, identified a 137-kb plasmid that was absent in isolates conferring low (1 µg/mL) or intermediate (4 µg/mL) levels of resistance to clarithromycin. Genome annotation of the plasmid called pMchErm55 identified a gene with similarity (71%–65% amino acid identity) to annotated erm genes in other rapidly growing mycobacteria (RGM) and Rhodococcus. erm genes mediate inducible macrolide resistance by methylating 23S ribosomal RNA (11, 12). The authors termed the plasmid allele erm(55)^P to distinguish it from other erm genes. pMchErm55 was similar to plasmid sequences, found in databases, isolated from other macrolide-resistant RGM, Mycobacterium obuense and Mycobacterium iranicum, suggesting resistance might be plasmid borne and transferrable, although direct evidence is still needed.

Using primers specifically designed to detect the erm(55)^P gene, the authors found similar (but not identical) genes on the chromosomes of four independent, highly macrolide-resistance isolates of M. chelonae. Whole-genome sequence of these isolates showed they lacked pMchErm55, and instead contained chromosomal copies of erm(55). Two isolates contained a chromosomal insertion of 37 kb, compared with reference strains, and this insertion contained an erm allele that was designated erm(55)^C to indicate that, while similar (85% amino acid identity) to erm(55)^P , it is found on the chromosome. In two other isolates, an erm(55) gene with 82% identity to erm(55)^P was found in a 2-kb segment of DNA that likely constitutes a transposon; the 2-kb segment includes erm(55) and a likely transposase, and the two genes are flanked by direct repeats, often the hallmark of transposon insertion. This allele was called erm(55)^T to reflect its potential association with a transposon.

In summary, all the isolates with high-level clarithromycin resistance contained either plasmid or chromosomal copies of erm(55), and lacked copies of a second allele, erm(41), and point mutations in 23S rRNA known to confer resistance. This association suggests that this new erm(55) allele confers high-level macrolide resistance in M. chelonae and other RGM.

These data are strongly supportive but do not formally prove the association of erm(55) with clarithromycin resistance, nor the mobility of the plasmid. As microbiologists, we would turn to Koch’s postulates, but as molecular biologists we would invoke Stanley Falkow’s “Molecular Koch’s postulates” (13, 14). The association of erm(55) and high-level macrolide resistance is compelling but not proven. There could be other novel unannotated genes on the plasmid or chromosome that confer or promote resistance. Ideally, a simple proof would involve mutating or deleting erm(55)^P from pMchErm55 and demonstrating that a pMchΔErm55-containing strain is clarithromycin susceptible. Cloning and expressing the erm(55) (P, T, and C) alleles back in this susceptible strain should now restore macrolide drug resistance, establishing erm(55) as the “causative agent.”

The plasmid pMchErm55 is large and, in addition to erm(55), likely encodes many other proteins that increase virulence, as well as plasmid maintenance. Two features noted by the authors are the clustered operons that are predicted to encode type IV and type VII secretion systems. Type IV systems have been well documented in Gram-negative and Gram-positive bacteria as secretion systems that can export both proteins and DNA (15, 16). During the transfer of plasmids, a relaxase encoded by the conjugative plasmid nicks the plasmid and becomes covalently linked to one strand of the plasmid DNA at the origin of transfer. The protein and linked DNA are pumped through the secretion apparatus into a recipient cell, where the plasmid is re-circularized and second-strand synthesis completes the plasmid transfer. Type VII systems are a feature of mycobacteria and are often found in multiple copies on the chromosome called ESX, for ESAT-6 secretion (e.g., M. tuberculosis has five copies and M. abscessus two copies) (17, 18). The different esx loci are thought to have evolved specifically to secrete proteins through the unique mycomembrane structure. Each ESX system encodes non-redundant and diverse functions. ESX-1 is associated with different aspects of virulence, ESX-3 is essential for iron uptake and metal homeostasis, and ESX-5 and ESX-1 play roles in membrane integrity. Adding to the variety, these functions are not always conserved between species, ESX-1 and ESX-4 are required for distributive conjugal transfer in the non-pathogen, Mycobacterium smegmatis (19, 20). Thus, it has been hypothesized that while the ESX secretion apparatus has been functionally retained, the substrates it secretes have been customized to benefit its mycobacterial host in its natural environment (21). The evolution and distribution of these ESX systems has been a focus of research: What was the progenitor ESX system and how was it duplicated and spread to different mycobacteria (22)? Based on phylogenetic analyses of plasmid and mycobacterial genome sequences, the current consensus is that these systems likely evolved on plasmids (23 – 26). Thus, duplication of the esx locus on a plasmid allowed selection for advantageous mutations and evolution of novel ESX systems. Subsequent plasmid transfer into different mycobacterial species, followed by integration of this ESX locus into the chromosome, would confer novel ESX functions on the new host. Consistent with this hypothesis, mycobacterial plasmids have been described encoding an ESX system that is phylogenetically linked to one of the five core systems, implying that each plasmid class is likely to be the distributor of that ESX system. Thus, pMchErm55 belongs to this growing catalog of mycobacterial plasmids that have provided a vehicle for ESX evolution in different mycobacterial species.

Type IV systems are not normally found in mycobacterial genomes, but they have been documented on mycobacterial plasmids that also contain type VII secretion systems. The two secretion systems are hypothesized to have co-evolved on these plasmids, perhaps to facilitate plasmid dissemination (26). The best evidence of this co-evolution to drive conjugation has been documented with a large plasmid from Mycobacterium marinum, pRAW, that encodes both type IV and type VII secretion systems (27). Similar plasmids were also identified by BLAST searches in databases containing genome sequences of other slowly growing mycobacteria (SGM). Most relevant for the current study on pMchErm55, it was demonstrated that pRAW is conjugative. Ummels et al. introduced antibiotic markers into the plasmid and used this marker to select for transfer into other SGM, including M. tuberculosis, but transfer was not detected using an RGM recipient (27). Moreover, they also demonstrated that genes required for normal type IV and type VII secretion are required for successful conjugation; plasmids with insertion mutants in type IV and type VII genes were transfer defective. Based on the example of pRAW, it is likely that pMchErm55 is also conjugative, but proof is in the postulates!

The authors describe pMchErm55 as having a broad-host range because it is found in different RGM species, but its absence in SGM and other bacterial species actually reflects that, like most bacterial plasmids, it has a narrow-host range. Broad-host-range plasmids such as those from the IncQ (RSF1010) and IncP (RP1) plasmid incompatibility groups can conjugate into, and stably replicate in, phylogenetically distinct bacterial species (28, 29). These plasmids have relatively sophisticated replication and maintenance systems that are independent of host functions, allowing replication in very diverse species. The restriction of pMchErm55 to RGM most likely reflects the need for mycobacterial host proteins for pMchErm55 replication and stable maintenance. In contrast to pMchErm55, pRAW appears restricted to SGM (27). Not all mycobacterial plasmids exhibit this rapid- versus slow-growing host specificity (30), but it is possible that these large, low-copy plasmids might have evolved this host distinction because of selective forces to ensure plasmid replication mirrors that of its host. It would be interesting to create hybrid derivatives of these two classes of plasmid to map the genes defining mycobacterial RGM-SGM-host specificity.

Plasmid conjugation or transformation are the most likely routes of dissemination for erm(55)^P, but what about the chromosomal copies? Did those alleles evolve on a plasmid before jumping to the chromosome, or vice versa? Was the plasmid mediating erm(55)^P dissemination a pMch progenitor or another (uncharacterized) plasmid? Erm(55)^T appears to be part of a composite transposon, but is it still an active element? The putative transposon is flanked by direct repeats consistent with transposon insertion; transposases normally cut a target with staggered nicks that, when repaired, result in a duplication of the target site. However, the authors do not indicate whether there are also inverted repeats present immediately inside the direct repeats. These are the critical binding sites of the transposase and the sites of transposon excision before it transposes to a new site. Is there a similar transposon on pMchErm55 or other plasmids that would explain how it arrived in the two sequenced M. chelonae isolates? Demonstration of its ability to transpose using a pMchΔErm-derivative as a target and selecting for transfer of erm(55)^T into a clarithromycin susceptible host is needed.

The second chromosomal copy of erm(55)^C is described as lying in a 37-kb chromosomal insertion, but details on the other genes in the insertion and the site of insertion are lacking. Is this insertion a large transposon or a cryptic mycobacteriophage? The insertion has occurred between two converging genes in such a way that it results in the maintenance of the stop codon of one gene, and the addition of just eight codons to the C-terminus of the other gene product. Certainly, these are subtle changes, and remarkably similar to the integration sites of many phages and some transposons that target conserved regions, but their insertion does not disrupt gene function (31, 32).

Future work is needed to confirm that pMchErm55 is indeed conjugative, and, perhaps, that the erm(55)^T gene is also transposable. Thus, we anticipate that additional copies of the erm(55) gene in its current and in new forms will inevitably be found in other RGM. What is more surprising is that plasmid- and transposon-mediated drug resistance has not been described more often in mycobacteria. A partial explanation is that “we” must look in the right places! Most whole-genome sequencing has been on the M. tuberculosis complex, which lacks plasmids and lacks any substantive evidence of horizontal gene transfer (HGT) as the complex is clonal (33). Other plasmid sequences are not fully accessioned, lie buried, and are yet-to-be discovered as sequence reads in various databases as described by Brown-Elliott et al. (10). However, there appear to be fewer plasmids in general in mycobacteria compared with other bacterial species. Why has there not been more plasmid dissemination? We can only speculate. Most likely, this is due to host-range restriction. Only two broad-host-range plasmids, RSF1010 and RPI, have been described to stably replicate in mycobacteria (34, 35), which suggests that the thousands of other plasmids in non-mycobacterial species just cannot be taken up by mycobacteria, despite their presence in co-existing bacteria in the environment. One likely reason is the unusual cell envelope of mycobacteria; their waxy, hydrophobic mycomembrane might simply prevent the appropriate cell-cell contact necessary for conjugation and preclude negatively charged exogenous DNA from accessing binding proteins on the cell surface that are necessary for transformation. In addition, even if a plasmid crosses into the cytoplasm, it is unlikely to stably replicate in mycobacteria because they lack the key host functions required for replication. Adding to this host specificity, is the apparent restriction of some mycobacterial plasmids to either SGM or RGM (as described above). Does the relative difference in mycobacterial growth rates prevent broader plasmid dissemination among mycobacterial hosts?

By highlighting a novel allele conferring macrolide resistance and identifying the potential for its further spread via conjugation and transposition, the authors underscore the need to recognize the potential for rapid spread of drug resistance among mycobacteria. While this article focuses on macrolide resistance, it would be naïve not to consider the spread of other antibiotic resistance genes by similar means. Plasmids and transposons are extraordinarily dynamic and can capture and mobilize large segments of DNA (36 – 38). The impact of mycobacterial plasmids on the evolution and dissemination of ESX systems already demonstrates their role as vehicles for HGT among both RG- and SG-mycobacteria. All bacteria are capable of becoming resistant to an antibiotic; it is only a matter of time and selection for that resistance mechanism to become part of a mobile element. This and other works highlight the obvious; we have only found the tip of the iceberg when it comes to the mobilome of mycobacteria and drug resistance.