Experimental Confirmation that an Uncommon rrs Gene Mutation (g878a) of Mycobacterium tuberculosis Confers Resistance to Streptomycin

ABSTRACT

The effective treatment of patients diagnosed with drug-resistant tuberculosis is highly dependent on the ability to rapidly and accurately determine the antibiotic susceptibility profile of the Mycobacterium tuberculosis isolate(s) involved. Thus, as more clinical microbiology laboratories advance toward the use of DNA sequence-based diagnostics, it is imperative that their predictive functions extend beyond the well-known resistance mutations in order to also encompass as many of the lower-frequency mutations as possible. However, in most cases, fundamental experimental proof that links these uncommon mutations with phenotypic resistance is lacking. One such example is the g878a polymorphism within the rrs 16S rRNA gene. We, and others, have identified this mutation within a small number of drug-resistant isolates, although a consensus regarding exactly which aminoglycoside antibiotic(s) it confers resistance to has not previously been reached. Here, we have employed oligonucleotide-mediated recombineering to introduce the g878a polymorphism into the rrs gene of Mycobacterium bovis BCG, a close relative of M. tuberculosis, and demonstrate that it confers low-level resistance to streptomycin alone. It does not confer cross-resistance to amikacin, capreomycin, or kanamycin. We also demonstrate that the rrs^g878a mutation exerts a substantial fitness defect in vitro that may at least in part explain why clinical isolates bearing this mutation appear to be quite rare. Overall, this study provides clarity to the phenotype attributable to the rrs^g878a mutation and is relevant to the future implementation of genomics-based diagnostics as well as the clinical management of patients in whom this particular polymorphism is encountered.

INTRODUCTION

Despite the availability of effective antibiotic therapy for close to 80 years, human tuberculosis (TB) resulting from infection with Mycobacterium tuberculosis remains one of the leading causes of mortality in low-income countries worldwide (1). Indeed, the evolution of multidrug-resistant TB (MDR-TB) and extensively drug-resistant TB (XDR-TB) severely limits the effectiveness of current treatment programs, with close to 500,000 new MDR cases being reported each year (2). Thus, as well as being the leading cause of death due to a single bacterial infection, M. tuberculosis also has the unenviable distinction of causing the most antimicrobial resistance-related deaths. However, with the introduction of the newly approved, all-oral regimen known as BPaL (bedaquiline, pretomanid, and linezolid), there is renewed optimism that we may now be able to treat pulmonary TB cases that were previously considered incurable (3).

While a great technological advance, current culture-independent molecular tests for identifying antibiotic resistance in M. tuberculosis, including the PCR-based Gene Xpert and GenoType MTBDR systems, are limited to the identification of the most common of the known, or well-established, resistance-conferring mutations. In addition, depending on the particular version of the system in use, they may also be limited to the detection of resistance to just 1 or 2 antibiotics (e.g., rifampicin [RIF] and/or isoniazid [INH]) (4–6). In contrast, next-generation sequencing (NGS)-based approaches show tremendous potential for the unbiased detection of resistance mutations for DNAs prepared directly from patient sputum samples or from a primary culture. However, at present, NGS approaches are still somewhat limited by virtue of the fact that there are many examples of low-frequency mutations or so-called “unexplained” resistance where poorly characterized mutations are detected that, while they may be suspected/predicted to confer resistance (with various levels of confidence), have never actually been experimentally proven to confer phenotypic antibiotic resistance within a laboratory setting (7–11). The ability to confirm, or discount, as many of these lower-frequency mutations as possible would serve to increase the certainty with which antibiotic resistance and susceptibility predictions are able to be made based solely on genomic data. In turn, this would greatly enhance the clinician’s ability to correctly apply the most appropriate drug combinations as early as possible after a positive TB diagnosis is made, thus limiting treatment failure and further resistance development.

One example of an unconfirmed mutation was recently detected in our laboratory for a single MDR-TB isolate that is part of the McGill University/RI-MUHC strain collection (12). As well as being resistant to RIF and INH, the strain was also classified as being resistant to streptomycin (STR) by the Laboratoire de Santé Publique du Québec (LSPQ). As part of a separate study, we decided to identify the genetic basis for resistance to each of these antibiotics in this particular strain and sequenced the main candidate loci, including the rpsL, rrs, and gidB genes, mutations in which are most frequently responsible for phenotypic STR resistance (10). The only mutation detected in any of these gene sequences was a G-to-A SNP (single nucleotide polymorphism) at position 878 of rrs that encodes the 16S RNA sequence. More typical rrs mutations linked to STR resistance include the a514c and c517t SNPs (10, 13). Also interesting was the fact that this mutation seems to be rarely recorded in either DNA sequence databases or the published literature. In fact, and as discussed below, in the small number of studies where this g878a SNP has been reported, there appeared to be some level of confusion regarding exactly which aminoglycoside antibiotics it may be associated with resistance to, namely, amikacin (AMK), capreomycin (CAP), kanamycin (KAN), and/or STR (14–18). As such, we considered that it might be beneficial to the TB community for our group to determine experimentally whether or not this SNP contributes to resistance to one or more of these antibiotics currently in use as second-line treatments against drug-resistant TB. Through the specific introduction of this point mutation into the antibiotic-sensitive Mycobacterium bovis bacillus Calmette-Guérin (BCG) background via oligonucleotide-mediated recombineering (recombination-mediated genetic engineering) (19, 20), here, we demonstrate for the first time that the g878a rrs mutation confers low-level resistance to STR alone.

RESULTS

MIC analysis and sequence determination.

Initial drug susceptibility testing (DST) by the LSPQ classified the lineage 4 Montreal M. tuberculosis isolate 57001 as being resistant to RIF, INH, and STR. To more accurately assess the MICs of these compounds for this strain, we conducted broth microdilution (BMD) assays in a 96-well format (Table 1). In this manner, we determined the MICs to be 250 μg/mL for RIF, 0.24 μg/mL for INH, and 4 to 5 μg/mL for STR, which served to confirm the initial report of the local public health laboratory. Next, to identify the genetic basis for resistance to each of these antibiotics, PCR products corresponding to the genes that are most frequently associated with mutations conferring resistance to these compounds were amplified and sequenced. These included products corresponding to the 81-bp rifampicin resistance-determining region (RRDR) of rpoB; the katG, inhA, and inhA promoter sequences (associated with INH resistance); as well as the rpsL, rrs, and gidB genes, mutations in which are most frequently responsible for resistance to STR (10, 21). In this manner, we identified that the strain possessed the S450W RpoB mutation that is known to confer RIF resistance and the −15c/t inhA promoter mutation consistent with the low level of INH resistance observed for strain 57001. However, the strain lacked any of the commonly reported rpsL, rrs, or gidB alleles associated with resistance to STR. The only variation from the wild-type (H37Rv) sequence that we could identify in any of these sequences was a G-to-A SNP at position 878 (g878a) of the rrs 16S rRNA gene. This corresponds to nucleotide 1472729 of the complete NCBI H37Rv reference sequence (GenBank accession number NC_018143.2 [2012 release]).

TABLE 1.

In vitro MIC values for M. tuberculosis strains H37Rv and 57001^a

M. tuberculosis strain	MIC (μg/mL)
	INH	RIF	STR	AMK	CAP	KAN
H37Rv	0.04	0.012–0.025c	0.63	0.63	1.25	3.13
57001	0.24	250	4–5b	0.63	1.25	3.13

^aAll MIC values were determined by liquid BMD assays and are presented in micrograms per milliliter. Details regarding the number of replicate assays and the range of antibiotic concentrations tested can be found in the text.

^bA separate dilution series was used in this case, leading to a slightly different MIC value.

^cDifferent values were obtained in independent assays.

A BLAST search of the NCBI nucleotide database was somewhat surprising in that there appeared to be only 4 strains bearing the identical rrs g878a SNP within the database. Likewise, a search of the literature identified relatively few publications making reference to the g878a mutation. What was also striking was that there was no obvious consensus among these articles regarding exactly which antibiotic resistance profile the g878a mutation might be associated with. As discussed below, these articles variably referred to the strain(s) bearing this mutation in the context of either AMK, CAP, KAN, or STR, and there was certainly no experimental confirmation of these largely genomics-based epidemiological studies (14–18). As shown in Table 1, when comparing H37Rv to strain 57001 in broth microdilution assays, we observed no difference in the MIC values obtained with AMK, CAP, or KAN.

Introduction of the g878a rrs mutation into wild-type antibiotic-sensitive M. bovis BCG-Danish via oligonucleotide-mediated recombineering.

In an effort to clarify the role of the g878a rrs SNP in aminoglycoside resistance, we decided to precisely engineer this mutation within the chromosome of a nonpathogenic mycobacterial species that is very closely related to M. tuberculosis, namely, the M. bovis BCG-Danish strain. BCG has been employed as a surrogate for studying the mechanisms of antibiotic activity and resistance on numerous occasions to date (22–25), and BCG strains bearing defined resistance mutations have also recently been proposed as a quality control or reference tool for use in clinical mycobacteriology laboratories (26). In our case, the choice to use BCG-Danish over an antibiotic-sensitive M. tuberculosis strain (such as H37Rv) was also encouraged by the shortage of personal protective equipment (PPE) in non-health-care settings during the first year of the coronavirus disease 2019 (COVID-19) pandemic.

Prior to use, we reconfirmed that the BCG-Danish strain was fully susceptible to the antibiotic compounds of interest for this study (i.e., AMK, CAP, KAN, and STR) (Table 2). Seventy-base-pair oligonucleotides targeting the relevant portions of the rrs and rpoB genes and containing the desired g878a (rrs) and c1349t (rpoB) SNPs at their center were cointroduced into the BCG-Danish background via electroporation at a ratio of 10:1, respectively. The rpoB oligonucleotide was originally designed for use as a recombineering control as it encodes the well-characterized S450L substitution commonly associated with acquired RIF resistance. However, its use turned out to be fortuitous as it allowed us to select for our recombineered clones that had taken up both oligonucleotides (STR/RIF resistant) from among the large background of spontaneous STR mutants that appeared following growth and selection at the relatively low concentration of 1 μg/mL STR. Note that this amount of STR was chosen based on the MIC value observed with M. tuberculosis isolate 57001, i.e., 4 to 5 μg/mL. We reasoned that spontaneous mutants, independent of the recombineering process, were also being selected for at 1 μg/mL STR due to the fact that we observed equivalent numbers of CFU following plating of the no-DNA control as well as an additional control in which a 70-bp oligonucleotide containing the wild-type rrs sequence was used in place of that containing the g878a SNP (also delivered in conjunction with the oligonucleotide bearing the mutant rpoB allele). Sanger sequencing of the PCR products spanning the targeted rrs sequence for a selection of 12 clones isolated following the cotransformation of the rrs^g878a and rpoB^c1349t oligonucleotides and plating at 1 μg/mL STR confirmed that mutations other than the g878a SNP were being selected for in this case. Additional sequencing revealed that all 12 clones contained one of four distinct gidB polymorphisms: W45* (stop), 352ins^g (insertion), 468ins^a, or 532ins^g. Notably, a broad range of gidB mutations have previously been associated with low-level STR resistance in M. tuberculosis, which is consistent with the apparent ease of their in vitro selection here at 1 μg/μL STR (10, 27).

TABLE 2.

Representative in vitro MIC values for the aminoglycosides AMK, CAP, KAN, and STR against the BCG strains included in this study^a

Strain	MIC(s) (μg/mL)
	AMK	CAP	KAN	STR (7H9 liquid)	STR (7H10 agar)
BCG-Danish	0.08	0.31, 0.4b	0.8, 1.6c	0.16	0.5
BCG-RpoBS450L (clones 2 and 3)	0.08	0.31	0.8	0.16	0.5
BCG-rrsg878a RpoBS450L (clones 1 and 15)	0.08	0.16, 0.2b	0.8, 1.6c	1.25	8

^aAll MIC values are presented in micrograms per milliliter. Details regarding the number of replicate assays and the range of antibiotic concentrations that were tested can be found in the text. For STR, both solid (7H10-OADC) and liquid (7H9-ADC) MIC assays were performed. All other compounds were tested by liquid-based BMD assays.

^bA separate dilution series was used in this case, leading to a slightly different MIC value.

^cDifferent values were obtained in independent assays.

After plating the original posttransformation outgrowth cultures (enriched at 1 μg/mL STR) onto 7H11-OADC (oleic acid-albumin-dextrose-complex) plates supplemented with either 2 μg/mL RIF or 1 μg/mL STR plus 2 μg/mL RIF, colonies were obtained only for the transformation that included both the g878a rrs- and c1349t rpoB-containing oligonucleotides. In contrast to plating on 1 μg/mL STR alone, neither of the control transformations (no-DNA and wild-type rrs oligonucleotide controls) resulted in even a single colony in this case. Five clones from each of the successful platings were sequence confirmed to have incorporated the two expected SNPs within their rrs and rpoB genes, respectively. A single clone from each independent plating (BCG-Danish_rrs^g878a rpoB^c1349t clones 1 and 15) was then selected for downstream analysis after confirming via PCR that it had been cured of the pNitET (KAN-resistant [KAN^r]) recombineering plasmid. In addition, we also sequence confirmed that these clones did not carry any additional, unwanted mutations within their gidB or rrs genes that could have confounded our subsequent analyses.

Introduction of the g878a rrs mutation results in resistance to STR but not other relevant aminoglycosides used in the treatment of TB.

Broth microdilution assays in a 96-well format and incorporating 2-fold serial dilutions of AMK, CAP, KAN, and STR were used to examine the relative impact of the g878a rrs mutation on the MICs obtained for the two recombineered clones (1 and 15) in comparison to the parental BCG-Danish strain. To control for any unforeseen effect that the introduction of the S450L RpoB mutation may have had on the MICs obtained in the presence of these aminoglycoside antibiotics, two RIF-resistant clones selected from the mutant rpoB/wild-type rrs oligonucleotide (control) transformation at 2 μg/mL RIF were also tested in MIC assays relative to the BCG-Danish wild type. In this manner, we established that the S450L RpoB substitution had no measurable impact on the response to AMK, CAP, KAN, or STR (Table 2). Thus, we could be confident that any MIC alterations that we observed when testing the recombineered rrs mutants were solely the result of introducing the g878a SNP.

As shown in Table 2, the introduction of the g878a SNP resulted in a reproducible 8-fold increase in the MIC for streptomycin with respect to the wild-type and RIF-resistant controls (0.16 versus 1.25 μg/mL). On each occasion that they were tested, both mutant clones (1 and 15) behaved identically in these assays. As the MIC was raised above the critical breakpoint concentration reported by the WHO for STR in liquid broth culture (MGIT culture) (1 μg/mL) (28), this result supports our hypothesis that the g878a rrs SNP confers acquired resistance to STR. In contrast, we saw no evidence that this SNP confers cross-resistance to any of the other major aminoglycosides used in the treatment of antibiotic-resistant TB: AMK, CAP, or KAN. If anything, we noted what appears to be a minor (2-fold) increase in susceptibility to CAP (Table 2). The increase in STR resistance that we noted for liquid cultures was subsequently corroborated on solid 7H10 medium when comparing the STR MICs for wild-type BCG-Danish, rrs^g878a mutant clones 1 and 15, and RpoB^S450L mutant clone 2. As indicated in Table 2, the introduction of the g878a rrs SNP led to a 16-fold increase in this case, raising the STR MIC from 0.5 μg/mL to 8 μg/mL, 4 times above the STR critical concentration value used in the agar proportion method based on the use of solid 7H10 medium (28).

Introduction of the g878a rrs mutation into BCG-Danish impacts fitness when grown in the absence of antibiotics.

Although the g878a mutation clearly confers low-level STR resistance, we were curious as to why we had some difficulty in obtaining recombineering clones bearing this mutation in the absence of the secondary selection step in the presence of RIF. We reasoned that the mutation may be deleterious to the strain in some manner that impacts its relative growth and fitness. To examine this hypothesis, we carried out standard in vitro growth curves in liquid 7H9-albumin-dextrose complex (ADC) broth to compare the relative growth of the g878a mutant clones (1 and 15) to those of the parental BCG-Danish strain and the rpoB mutation control in the absence of added antibiotics. As can be seen in Fig. 1A, the introduction of the g878a SNP into rrs leads to a substantial reduction in the growth rate, above and beyond that observed with the S450L RpoB mutation alone that is commonly associated with RIF resistance in M. tuberculosis. Over the first 96 h, where the growth rates were mostly linear for all the strains examined, the relative doubling times were as follows: 28.7 h for wild-type BCG-Danish, 32.3 h for BCG-Danish_rpoB^c1349t (average for clones 2 and 3), and 39.9 h for BCG-Danish_rrs^g878a rpoB^c1349t (average for clones 1 and 15).

FIG 1.

Introduction of the rrs^g878a mutation results in a growth/fitness defect in vitro. (A) Growth curves carried out in 7H9-ADC medium (in the absence of added antibiotics) comparing wild-type M. bovis BCG-Danish (BCGwt) against two RIF-resistant (BCG-Danish_rpoB^c1349t) (S450L clones 2 and 3) and two STR/RIF-resistant (BCG-Danish_rrs^g878a rpoB^c1349t) (g878a rrs clones 1 and 15) clones obtained by recombineering. Two separate OD₆₀₀ readings were obtained for all 5 cultures at each of the time points indicated. For the final time point, all cultures were diluted 1:3 prior to reading. The plotted OD₆₀₀ values represent those of the undiluted cultures. Standard deviations are indicated by crosshairs. For the wild type and both of the g878a rrs mutant clones, the data are representative of results from two independent growth assays. (B) Growth curves carried out in 7H9-ADC medium (no antibiotic) comparing STR/RIF-resistant strains carrying three distinct combinations of mutations, BCG-Danish_rrs^g878a rpoB^c1349t (g878a rrs clone 1) (as described above), BCG-Danish_rpoB^c1349t rpsL^a263g (RpsL K88R clone 33), and BCG-Danish_rpoB^c1349t rpsL^a128g (RpsL K43R clone 36). As described above, two separate OD₆₀₀ readings were obtained for all cultures at each of the time points indicated, and the data are representative of results from two independent growth assays. At the final time point, cultures were diluted 1:3 prior to reading. Standard deviations are indicated by crosshairs. In panels A and B, the inset tables indicate the doubling times (G) for each strain calculated over the first 96 h.

In a separate assay, we then compared g878a mutant clone 1 (BCG-Danish_rrs^g878a rpoB^c1349t) against two RIF/STR-resistant clones that were generated by selecting a BCG-Danish_rpoB^c1349t (S450L) RIF-resistant mutant on 2 μg/mL STR. These clones were identified as carrying the archetypal STR resistance-conferring mutations RpsL^K88R (clone 33) and RpsL^K43R (clone 36). As shown in Fig. 1B, the recombineered strain bearing the rrs^g878a SNP shows a clear growth/fitness defect relative to the strains carrying the two most common mutations associated with high-level STR resistance. Although not compared “head-to-head” within the same growth assay, by comparing the two curves shown in Fig. 1A and B, we also noted that the introduction of the RpsL^K43R mutation seemingly improved the growth rate of the resulting double mutant strain (RpoB^S450L RpsL^K43R) (clone 36) relative to that of the RIF-resistant RpoB^S450L single mutant (clones 2 and 3). The RpsL^K43R mutation is by far the most common mutation associated with STR resistance in clinical M. tuberculosis isolates, particularly among MDR isolates, which is consistent with its low fitness cost (10, 29, 30). In addition, Spies et al. previously noted that among a small sample of MDR/STR-resistant isolates, those with the RpsL^K43R allele showed growth enhancement (31). Indeed, this type of unexpected epistatic interaction whereby dually resistant strains bearing particular combinations of mutations show improved fitness over the corresponding monoresistant strains has been described previously in relation to RIF and ofloxacin resistance (32). Nevertheless, as it was not the main objective of the current study, we have not investigated this phenomenon in relation to the RpoB^S450L and RpsL^K43R alleles any further at this stage.

Although we had previously noted that M. tuberculosis strain 57001 grew quite slowly in liquid culture relative to the H37Rv laboratory strain, for example, we decided to investigate its growth characteristics more systematically in light of the defect that we observed with BCG mutant clones 1 and 15 described above. As such, we compared M. tuberculosis strains H37Rv, 57001, HN878 (33, 34), and 63117. Like 57001, 63117 is an MDR-TB isolate from Montreal that is also resistant to STR. However, instead of the g878a rrs mutation, it has the K43R RpsL allele that confers high-level STR resistance in addition to the S450L RpoB and −15c/t inhA mutations conferring resistance to RIF and INH, respectively. Both HN878 and 63117 are lineage 2 M. tuberculosis strains, with HN878 being antibiotic sensitive. As shown in Fig. 2, M. tuberculosis strain 57001 (rrs^g878a) has a clear growth defect relative to the other 3 strains, with a calculated doubling time of 36.4 h, in comparison to 20.4 h for H37Rv, 30.5 h for HN878, and 30.4 h for 63117. Nevertheless, unlike the BCG strains engineered as described above, it needs to be kept in mind that these particular M. tuberculosis strains are not isogenic, which precludes any precise assessment of the role of individual mutations in the growth patterns observed.

FIG 2.

Relative in vitro growth of M. tuberculosis strain 57001 bearing the rrs^g878a allele. Growth curves were carried out in 7H9-ADC medium (in the absence of added antibiotics) comparing antibiotic-sensitive M. tuberculosis strains H37Rv and HN878 with the MDR/STR-resistant strains 57001 and 63117. Additional strain details can be found in the text. Two separate OD₆₀₀ readings were obtained for all 4 cultures at each of the indicated time points, and the average data from two independent experiments are shown. For the 96-h time point, all cultures other than strain 57001 were diluted 1:4 prior to reading. At the 144-h time point, strain 57001 was diluted 1:4, and the rest were diluted 1:8 prior to reading. The plotted OD₆₀₀ values represent those of the undiluted cultures. The inset table indicates the doubling time (G) for each M. tuberculosis strain over the first 96 h.

In summary, our data support the hypothesis that the g878a rrs mutation results in a substantial fitness defect, at least under standard in vitro conditions in the absence of antibiotics. However, at this point, we cannot exclude the possibility that the fitness defect attributed to the g878a rrs mutation may manifest itself only in the context of strains also bearing the S450L RpoB mutation (or similar).

DISCUSSION

Although the g878a rrs mutation that is the focus of this investigation appears to be relatively uncommon among antibiotic-resistant patient isolates, one of the primary motivations for our study was to clarify which, if any, of the second-line aminoglycoside antibiotics it conferred resistance to. We felt that clarification was necessary in this case due to the potential confusion that could arise based on surveying the currently available literature. For example, of the 5 articles that we identified as reporting the g878a SNP, 3 associated its presence with isolates that were STR resistant (15, 17, 18). Only one of these studies looked at resistance to other aminoglycosides in addition to STR (CAP and KAN) but did not find an association with the g878a SNP (15). We also found 2 studies that examined the g878a rrs SNP in the context of aminoglycosides other than STR. In one of these studies, a single isolate bearing the g878a polymorphism was shown to be susceptible to AMK, CAP, and KAN (14). Next, a South African study of pre-XDR- and XDR-TB patients identified 21 isolates with the g878a mutation (16). However, the distribution of the mutation with respect to the reported resistance phenotypes was quite variable: 10 of the isolates were CAP monoresistant, 4 were KAN monoresistant, 4 were cross-resistant to AMK/CAP/KAN, 2 were AMK/CAP cross-resistant, and the remaining isolate was resistant to CAP/KAN. None of these isolates were examined for STR resistance for reasons that were not given. Based on the frequency of CAP resistance, the g878a rrs SNP was reported by those authors as a new mechanism of resistance to CAP. As such, it was recommended that the g878a mutation be included in new molecular assays to increase the sensitivity of CAP resistance detection (16). Finally, in addition to the 4 isolates identified by BLAST searching of the NCBI database, we also identified 4 distinct MDR-TB isolates included within the PATRIC (Pathosystems Resource Integration Center) database that are reported to carry the g878a rrs allele (35). Three of these are classified as being STR resistant, while all four are listed as being susceptible to AMK, CAP, and KAN. Overall, we felt that it was quite important to investigate this polymorphism at the experimental level in an attempt to generate conclusive data directed at addressing the nagging question regarding its precise clinical relevance.

Although there are relatively few published studies where we find the g878a rrs mutation mentioned, it is interesting to note that in each case, the M. tuberculosis isolates involved are either MDR or XDR. This is despite the fact that STR monoresistance is second only to INH monoresistance in terms of global frequency (29). Whether this observation reflects some form of cryptic epigenetic interaction between specific mutations in rpoB and the rrs^g878a mutation, for example, or whether it reflects a strong bias toward the sequencing and reporting of MDR/XDR-TB clinical isolates at present rather than those that are monoresistant is not clear. Alternatively, it may reflect the reality that STR has been relegated to use only in second-line regimens when AMK cannot be used. We do note, however, that our engineered BCG-Danish_rrs^g878a rpoB^c1349t strain exhibits a substantial fitness defect in vitro whereby its observed doubling time is 1.24 times that of the rpoB^c1349t RIF-resistant mutant strain. This finding suggests that in the absence of any compensatory adaptation, cells that arise with the g878a rrs polymorphism are likely to be at a distinct competitive disadvantage in the presence of other resistant clones that do not exhibit a fitness defect to the same degree. This scenario might well explain why M. tuberculosis isolates containing the g878a mutant rrs allele are quite rare among clinical isolates appearing in the NCBI and PATRIC databases as well as in the published TB literature. With this in mind, it is interesting to note that we also identified a fitness defect for M. tuberculosis strain 57001 (rrs^g878a), with a doubling time 1.19 times that of HN878/63117 and 1.76 times that of H37Rv.

In terms of an underlying mechanistic basis that could potentially explain how the g878a rrs SNP may contribute to resistance to STR, we note that position 878 of the M. tuberculosis 16S rRNA molecule is equivalent to position 885 within a highly conserved region of the Escherichia coli 16S rRNA sequence (36–40). In E. coli, residue 885 (G) base pairs with residue 912 (C) at the base of helix 27, a structure implicated in tRNA selection in both prokaryotes and eukaryotes (41). Cross-linking, footprinting, and mutagenesis experiments have all demonstrated that STR binds to this same area, specifically to residues 912 to 915, which form what is referred to as the “915 region” (36, 42, 43). Moreover, at least two independent studies have shown that mutations introduced into this 915 region, including a C-to-T mutation at position 912, reduce STR binding to the ribosome, resulting in low-level resistance to this antibiotic (36, 44, 45), analogous to the phenotype that we report here for both the M. tuberculosis isolate 57001 and our recombineered BCG-Danish_rrs^g878a rpoB^c1349t strain. By inference, we hypothesize that a G-to-A substitution at position 878 within the M. tuberculosis or BCG rrs sequence will prevent its base pairing to the complementary cytosine residue at position 905 (equivalent to E. coli residue 912). In turn, this disruption has the potential to perturb the organization of the 915 region in a manner that may impede the binding of STR, thereby leading to resistance (Fig. 3). Notably, each of the other 3 aminoglycosides examined in this study, namely, AMK, CAP, and KAN, has been shown to bind to a distinct region of the 16S rRNA molecule known as the “A-site” (aminoacyl-tRNA site) that comprises a portion of helix 44 (46, 47). Thus, mutations causing resistance to these compounds all tend to be localized around rrs position 1400 (10, 27, 29).

FIG 3.

Predicted structure of the 16S rRNA helix 27 region in M. tuberculosis (Mtb) and M. bovis BCG based on that of E. coli (top). Note that the E. coli and M. tuberculosis/BCG sequences are identical in this region aside from a single U to C substitution at position 897 in the latter. Canonical base pairs within the helix are indicated by lines, while “wobble” (G-U) base pairing is indicated by dots. The position of the g878a rrs mutation is indicated in red (bottom). The relative nucleotide numbers for both the E. coli and M. tuberculosis/BCG 16S RNA sequences are included at 10-bp intervals (900, etc.). The 915 region involved in STR binding is circled in the E. coli panel. (Schematic representations adapted from reference 40 and the RiboVision2 website [http://apollo.chemistry.gatech.edu/RiboVision/index.html] [39].)

In summary, through a combination of genetic recombineering and in vitro MIC assays, here we have, for the first time, experimentally confirmed that the presence of the clinically relevant rrs^g878a mutation causes low-level STR resistance. However, by itself, it does not alter susceptibility to the other second-line injectable aminoglycosides used in the treatment of TB. Of note, at the time of submission of the manuscript, we became aware of a new WHO publication entitled Catalogue of Mutations in Mycobacterium tuberculosis Complex and Their Association with Drug Resistance (48). Regarding STR resistance, the catalogue lists 12 group 1 mutations (i.e., those associated with resistance at the highest confidence level based on the statistical thresholds and rules applied), one of which is the rrs^g878a mutation examined here. The PPV/SOLO value (positive predictive value of the mutation occurring independently within the set of genes/promoter regions examined) of this mutation for the identification of STR resistance is quoted as 85.7% (95% confidence interval [CI], 63.7 to 97.0%), while for AMK, CAP, and KAN, the value is 0.0%. Thus, our experimental work serves effectively to validate the genomics-based associations reported for the g878a mutation within this extensive and valuable catalogue (48). In addition to providing an important point of clarification regarding the precise role of this SNP, the knowledge generated by our study is also relevant in light of a recent call for the reinstatement of STR for use in the treatment of drug-resistant TB caused by isolates that exhibit low-level STR resistance yet are highly resistant to AMK and KAN (27). Theoretically, any M. tuberculosis isolates identified as bearing the rrs^g878a SNP in addition to the canonical rrs 1400 region mutations conferring high-level resistance to AMK and KAN would fall into this category and could potentially be treated successfully with STR.

MATERIALS AND METHODS

Bacterial strains and culture.

The MDR M. tuberculosis isolate 57001 was classified as belonging to the Euro-American lineage (lineage 4) in a previous molecular epidemiological study of Montreal TB patient isolates collected between January 2001 and May 2007 (12). For initial drug susceptibility testing of the isolate, the radiometric Bactec 460TB system was used. The standard critical concentration of STR used with this system is 2 μg/mL. The M. bovis BCG-Danish strain was kindly provided by Marcel Behr (RI-MUHC, Montreal). All strains were grown either in liquid 7H9 medium (Difco) supplemented with 10% ADC (8.1 g L⁻¹ NaCl, 50 g L⁻¹ bovine serum albumin [BSA] fraction V [Calbiochem], 20 g L⁻¹ glucose), 0.2% glycerol (Sigma-Aldrich), and 0.05% Tween 80 (Sigma-Aldrich) or on 7H10 or 7H11 agar (Difco) supplemented with 10% OADC (as described above for ADC, plus 0.6 mL L⁻¹ oleic acid and 3.6 mM NaOH) and 0.5% glycerol. The antibiotics RIF, INH, and KAN were purchased from Sigma-Aldrich, while AMK and CAP were purchased from Cayman Chemical Company (supplied by Cedarlane).

DNA purification, PCR, and sequencing.

M. tuberculosis genomic DNA was isolated and purified according to a protocol described previously by Pelicic et al. (49). For PCR-based screening of BCG clones picked into 7H9-ADC medium, boiled culture lysates were prepared (e.g., 250 μL of culture heated at 90°C for 30 min, followed by centrifugation and resuspension of the pelleted material in 50 μL Tris-EDTA [TE]). Taq DNA polymerase, 10× reaction buffer, MgCl₂, and deoxynucleoside triphosphates (dNTPs) were obtained from Thermo Fisher Scientific. PCR was carried out according to standard protocols except for when amplifying a portion of the KAN resistance cassette, where 5% dimethyl sulfoxide (DMSO) was also included in the reaction mixtures. Primers used in this study for PCR and sequencing are shown in Table S1 in the supplemental material, a number of which are based on those reported previously by Rowneki et al. (21). Sanger sequencing of PCR products was carried out at the Centre d’Expertise et de Services Génome Québec (Montreal).

MIC determination.

Broth microdilution (BMD) assays were performed in 7H9-ADC medium (50). Twofold serial dilutions of antibiotics (and no-antibiotic controls) were added to 96-well microtiter plates prior to the addition of an equal volume of an M. tuberculosis or BCG culture diluted 1:100 from growing stock cultures adjusted to an optical density at 600 nm (OD₆₀₀) of 0.1. The plates were sealed in plastic ziplock bags and incubated for 7 to 14 days at 37°C. Thirty microliters of 0.01% resazurin (Sigma-Aldrich) was added to each well, and the plates were incubated for a further 4 days prior to quantifying fluorescence on a Tecan Infinite 200 Pro plate reader. All broth MIC assays were set up in triplicate or quadruplicate with two independent assays performed, except for the BCG STR MICs, where three independent assays were performed. All strains were tested together as part of the same assay in a batchwise manner. The range of antibiotic concentrations tested for M. tuberculosis were as follows: INH at 0.005 to 5 μg/mL for H37Rv and 0.12 to 125 μg/mL for 57001, RIF at 0.0004 to 0.4 μg/mL for H37Rv and 1.95 to 2,000 μg/mL for 57001, STR at 0.02 to 20 μg/mL, AMK at 0.01 to 10 μg/mL, CAP at 0.01 to 10 μg/mL, and KAN at 0.1 to 100 μg/mL. For all BCG strains, the following antibiotic concentrations were tested: STR at 0.02 to 20 μg/mL, AMK at 0.01 to 10 μg/mL, CAP at 0.01 to 10 μg/mL, and KAN at 0.1 to 100 μg/mL.

MIC determination for STR was also carried out with the BCG strains on 7H10-OADC solid medium according to the absolute concentration method, with minor modifications (51). After the addition of OADC, the 7H10 medium was kept at 55°C, and 20-mL aliquots were mixed with the corresponding amount of STR prior to distributing 1 mL of each 2-fold serial dilution into quadruplicate wells of 24-well plates. The range of STR concentrations tested was from 0.02 to 16 μg/mL, and no-antibiotic control wells were included. Each well was inoculated with 5 μL of a 1:100 dilution of cultures adjusted to an OD₆₀₀ of 0.1. Four replicate wells were tested for each strain, and the MIC was considered to be the concentration of STR for which we observed no CFU. Plates were incubated at 37°C in sealed, plastic ziplock bags and read at 14, 21, and 28 days.

Mycobacterial recombineering.

Oligonucleotide-mediated recombineering was carried out essentially as described previously by Murphy et al. (19). Briefly, BCG-Danish was transformed with the pNitET-SacB-kan plasmid (referred to as pNitET here) (kindly provided to our RI-MUHC colleague Marcel Behr by the laboratory of Chris Sassetti, University of Massachusetts Medical School) and selected on 7H11-OADC agar plates containing 20 μg/mL KAN. A single BCG::pNitET-positive clone was subsequently grown to an OD₆₀₀ of approximately 0.8 and treated for 24 h with 1 μM isovaleronitrile (Sigma-Aldrich) to induce the expression of the RecET proteins prior to the cotransformation of the 70-bp oligonucleotides (Invitrogen) (Table S1) targeting rpoB (0.1 μg) and rrs (1 μg). A control (no-DNA) electroporation was also included. The electroporated cells were allowed to recover in 10 mL of 7H9-ADC medium at 37°C with shaking for 4 days, after which a 1-mL aliquot was diluted 1:20 with 7H9-ADC medium containing STR at 1 μg/mL. Once the growth of the antibiotic-treated cultures was detected and they had reached an OD₆₀₀ of approximately 0.3 (after 15 days), 150-μL aliquots of undiluted and diluted (1:10) cells were spread onto 7H11-OADC agar plates containing either 1 μg/mL STR, 2 μg/mL RIF, or both. No KAN was added to the media. After 4 weeks of incubation at 37°C, colonies were picked into 1 mL 7H9-ADC medium without antibiotics (to further aid in curing the bacteria of the pNitET plasmid) and allowed to grow for 3 weeks, with occasional mixing. A sample of the cultures was screened for the presence of the desired rrs mutation by sequencing of the PCR products generated from boiled culture lysates. The complete loss of the pNitET plasmid from the clones of interest was also confirmed by PCR using primers specific for the KAN resistance cassette.