Characterization of Mutations Conferring Resistance to Rifampin in Mycobacterium tuberculosis Clinical Strains

Resistance of Mycobacterium tuberculosis to rifampin (RMP), mediated by mutations in the rpoB gene coding for the beta-subunit of RNA polymerase, poses a serious threat to the efficacy of clinical management and, thus, control programs for tuberculosis (TB). The contribution of many individual rpoB mutations to the development and level of RMP resistance remains elusive.

ABSTRACT

Resistance of Mycobacterium tuberculosis to rifampin (RMP), mediated by mutations in the rpoB gene coding for the beta-subunit of RNA polymerase, poses a serious threat to the efficacy of clinical management and, thus, control programs for tuberculosis (TB). The contribution of many individual rpoB mutations to the development and level of RMP resistance remains elusive. In this study, the incidence of mutations throughout the rpoB gene among 115 Mycobacterium tuberculosis clinical isolates, both resistant and susceptible to RMP, was determined. Of the newly discovered rpoB mutations, the role of three substitutions in the causation of RMP resistance was empirically tested. The results from in vitro mutagenesis experiments were combined with the assessment of the prevalence of rpoB mutations, and their reciprocal co-occurrences, across global M. tuberculosis populations. Twenty-two different types of mutations in the rpoB gene were identified and distributed among 58 (89.2%) RMP-resistant strains. The MICs of RMP were within the range of 40 to 800 mg/liter, with MIC₅₀ and MIC₉₀ values of 400 and 800 mg/liter, respectively. None of the mutations (Gln429His, Met434Ile, and Arg827Cys) inspected for their role in the development of RMP resistance produced an RMP-resistant phenotype in isogenic M. tuberculosis H37Rv strain-derived mutants. These mutations are supposed to compensate for fitness impairment incurred by other mutations directly associated with drug resistance.

INTRODUCTION

Rifamycins comprise a group of structurally complex macrocyclic antibiotics originally isolated in the late 1950s from culture filtrates of Streptomyces mediterranei, currently under the name of Amycolatopsis mediterranei. The most prominent member of this group is rifampin (RMP), a semisynthetic derivative of a naturally occurring rifamycin B discovered in 1965 and subsequently introduced for clinical use in 1968. Mycobacteria exhibit variable susceptibility to RMP, with most of the slowly growing species being moderately to highly susceptible. The drug is extremely effective against Mycobacterium tuberculosis, for which the MIC values fall within a narrow range of 0.1 to 0.4 mg/liter (1, 2). The high potency of RMP against M. tuberculosis is due to its strong bactericidal effect on metabolically active, continuously replicating bacilli within the walls of lung cavities synergistically enhanced by sterilizing action on both intracellular and semidormant bacilli residing in the caseous foci of tuberculous granulomas. The activity of RMP, later coupled with that of pyrazinamide (PZA), allowed the course of tuberculosis (TB) treatment to be shortened considerably from 18 to 24 months to 6 months (3). Consequently, since the early 1970s, RMP has become a mainstay in the treatment of TB and is now a critical component of all short-course (6- and 8-month) chemotherapeutic regimens for TB recommended by the World Health Organization (WHO) (4).

Resistance of tubercle bacilli to RMP alone (RMP monoresistance) is rare, which is in contrast to what is reported for another major anti-TB drug, isoniazid (INH). At least two reasons may explain this. One is that in a sensitive population, resistance to RMP arises at a rate of approximately 1 in 10⁸ bacilli, whereas resistance to INH exists 100 times more frequently, at a rate of 1 in 10⁶ bacilli (5). The other reason may relate to the fact that, unlike INH or streptomycin (SM) in the early years of TB chemotherapy, RMP, except for short-term use for prophylaxis, has never been used alone but has always been used as a part of multidrug regimens. Nevertheless, RMP monoresistance occurs albeit at a low frequency. For instance, in Western Europe, the rates of monoresistance are below 0.4% and 1.2% of new and previously treated TB cases, respectively (6). Among all TB cases with resistance to RMP reported annually, those recognized as being monoresistant are the fewest. The bulk of RMP resistance exists in conjunction with INH resistance, thus defining a multidrug-resistant (MDR) phenotype, which portends higher rates of treatment failure and mortality (7). Hence, RMP resistance serves as a surrogate marker for the detection of MDR-TB, whose incidence in 2014 was estimated to be 490,000 cases or 5% of the total TB caseload globally (8).

Numerous studies have shown that usually more than 90% of RMP-resistant M. tuberculosis strains harbor mutations in the rpoB gene coding for the β-subunit of the DNA-dependent RNA polymerase (RNAP), which is a molecular target of rifamycins. The great majority of the rpoB mutations are located within an 81-bp central segment of the 3,519-bp-long gene, called the RMP-resistance-determining region (RRDR), encoding 27 amino acids and corresponding to codons 426 to 452 in M. tuberculosis or 507-533, according to the consensus numbering scheme of RNAP from Escherichia coli (9–23).

Identification of mutations conferring drug resistance in M. tuberculosis is now at the core of the design of most of the new diagnostic tools for fast and reliable drug susceptibility testing (DST), offering an attractive alternative to time-consuming and expertise-demanding conventional methods. The purpose of this study was to investigate the type and frequency of mutations throughout the rpoB gene among 115 M. tuberculosis clinical isolates presenting with either an RMP-monoresistant, MDR, or pansusceptible phenotype and to resolve the potential role of the newly discovered rpoB mutations in the causation of RMP resistance.

RESULTS

rpoB mutation patterns.

Twenty-two different types of mutations in the rpoB gene were identified and distributed among 14 (93.3%) RMP-monoresistant and 44 (88%) MDR M. tuberculosis strains (Fig. 1 and Table 1; see also Table S1 in the supplemental material). Thus, a total of 58 (89.2%) RMP-resistant strains presented an altered rpoB allele. All mutations were divided into three groups, depending on whether they occurred only in MDR (i), only in RMP-monoresistant (ii), or in both MDR and RMP-monoresistant (iii) strains. The first group was the most abundant, with 16 different mutations accounting for 33 MDR strains. In this group, a substitution, G1303T, conferring an Asp435Tyr amino acid change, showed the highest frequency (8 strains or 24.2%), followed by G1302T (Met434Ile) and C1333T (His445Tyr) substitutions, detected in 4 strains (12.1%) and 3 strains (9.1%), respectively.

FIG 1.

Schematic representation of the distribution of mutations in the rpoB gene, including the RMP-resistance-determining region (RRDR). The nucleotide numbering is based on the rpoB gene sequence from the M. tuberculosis H37Rv reference strain (TubercuList [http://genolist.pasteur.fr/TubercuList/]). Nucleotide coordinates for the bases flanking the RRDR are approximated. Numbers in circles correspond to mutation numbering in Table 1. Shaded circles denote mutations that occurred in rifabutin-resistant but not rifabutin-susceptible strains.

TABLE 1.

Mutations detected among 50 MDR and 15 RMP-monoresistant M. tuberculosis clinical strains under study

Mutation	Mutated residuea			No. of strains inhibited by an RMP MIC (mg/liter) ofb:					Total no. of strains
	nt	aa (M. tuberculosis)	aa (E. coli)	40	100	200	400	800
1	C309T	Asp103Asp	Asp72Asp	2		1	4	4 + 1	11 + 1
2	G508T	Val170Phe	Val46Phe				2		2
3	del1275-80CGGCAC*	del426-7Gly-Thr	del507-8Gly-Thr					1	1
4	C1284A	Ser428Arg	Ser509Arg				1		1
5	G1287C	Gln429His	Gln510His			1			1
6	G1302T	Met434Ile	Met515Ile			1	3		4
7	G1303T	Asp435Tyr	Asp516Tyr		1	2	4	1	8
8	A1304G	Asp435Gly	Asp516Gly					1	1
9	A1304T	Asp435Val	Asp516Val					2	2
10	C1333G	His445Asp	His526Asp				1	1	2
11	C1333T	His445Tyr	His526Tyr			1	1	1	3
12	CA1333-4TG	His445Cys	His526Cys			1	1		2
13	A1334G	His445Arg	His526Arg				1	1	2
14	A1334T	His445Leu	His526Leu			1			1
15	G1343T	Arg448Leu	Arg529Leu				1	1	2
16	C1349G	Ser450Trp	Ser531Trp					1	1
17	C1349T	Ser450Leu	Ser531Leu		1		6	15 + 10	22 + 10
18	T1355C	Leu452Pro	Leu533Pro				1		1
19	A1462G	Ile488Val	Ile569Val				1		1
20	G1482A*	Leu494Leu	Leu575Leu					1	1
21	C2479T*	Arg827Cys	Arg914Cys				1		1
22	T3225C	Ala1075Ala	Ala1283Ala			1	1	3	4 + 1
No mutation	NA	NA	NA	5 + 1		1			6 + 1

^aCodon numbers are reported using either the M. tuberculosis or E. coli numbering system. Asterisks indicate novel mutations (not previously reported in the literature); nonsynonymous mutations with an as-yet-undefined association with RMP resistance are shown in boldface type. nt, nucleotide sequence; aa, amino acid sequence; NA, not applicable; del, deletion; RMP, rifampin.

^bUnderlined and nonunderlined numbers refer to RMP-monoresistant and MDR strains, respectively.

Three types of rpoB mutations were found solely in RMP-monoresistant strains. The mutations were all point nucleotide substitutions with resulting amino acid replacements at codon 435 (A1304T [Asp435Val]), codon 450 (C1349G [Ser450Trp]), and codon 452 (T1355C [Leu452Pro]) and occurred in two, one, and one strains, respectively.

Three single-nucleotide polymorphisms (SNPs) were observed in both MDR and RMP-monoresistant strains. Twenty-two (44%) MDR and 10 (66.7%) RMP-monoresistant strains had a missense mutation at codon 450 (C1349T [Ser450Leu]). The other two mutations, i.e., C309T (Asp103Asp) and T3225C (Ala1075Ala), were silent mutations and occurred in single RMP-monoresistant strains and 11 (22%) and 4 (8%) MDR strains, respectively.

Of the 44 MDR strains with an altered rpoB gene sequence, 24 (54.6%) harbored mutations at only one codon, while the remaining 20 (45.4%) strains had mutations at either two codons (14 strains or 31.8%) or three codons (6 strains or 14.6%). Among the 14 RMP-monoresistant rpoB mutants, 12 (85.7%) were single mutants, and 2 (14.3%) were double mutants.

No mutations at the rpoB locus were detected in 7 RMP-resistant (6 MDR and 1 monoresistant) strains and all 50 pansusceptible strains.

rpoB mutations and MIC of RMP.

Upon MIC determination, five MDR strains carrying a wild-type (wt) rpoB gene and two further MDR strains with a silent C309T (Asp103Asp) mutation grew at an RMP concentration of 20 mg/liter, but not 40 mg/liter, as upon initial susceptibility testing. Thus, the MICs of RMP for MDR strains fell within the range of 40 to 800 mg/liter, with MIC₅₀ and MIC₉₀ values of 400 and 800 mg/liter, respectively. All RMP-monoresistant strains showed MICs of 800 mg/liter against the drug, except for one strain, whose MIC was 400 mg/liter. When analyzing the whole RMP-resistant population, the MIC₅₀s and MIC₉₀s were 400 and 800 mg/liter, respectively. Both these values were higher for strains exhibiting additional resistance to rifabutin (27 [54%] MDR and 10 [66.7%] RMP-monoresistant strains) than for strains with no such coresistance detected (MIC₅₀ of 800 versus 200 mg/liter and MIC₉₀ of 800 versus 400 mg/liter). This difference was statistically significant (P < 0.001).

In an attempt to correlate different rpoB mutations with MICs of RMP, strains with the C1349T (Ser450Leu) mutation were shown to display higher levels of RMP resistance than strains without this mutation. The geometric mean (GM) RMP MICs for these two strain populations were 640.2 and 171.7 mg/liter, respectively (P < 0.001). In contrast, strains with the G1303T (Asp435Tyr) mutation had MICs of RMP that were lower than those for rpoB mutants not having this alteration, with GM MICs of 308.4 and 402.7 mg/liter, respectively (P = 0.04).

Six nonsynonymous mutations (Table 1), either previously or newly (in this study) described, whose roles in the causation of RMP resistance have been uncertain were inspected by exploring the interaction of the respective RpoB mutant proteins with RMP through homology modeling and molecular docking analysis. The amino acid residues affected by these mutations are shown in Fig. 2.

FIG 2.

Model of the spatial structure of the M. tuberculosis wt RpoB protein. Bound RMP is displayed in red. Arrows indicate residue numbers. The amino acids of interest are shown as spheres. Atoms are colored according to the following scheme: gray for C, blue for N, red for O, and orange for S.

In silico docking analysis.

The impact of the six selected mutations on the electrostatic potential in the RMP binding pocket (RBP) is presented in Table 2. The majority of substitutions induced a moderate effect, with a high score for Gln429His and the highest score for Arg827Cys. Whereas the Gln429 residue belongs to the RBP, the Arg827 residue is located distantly from the RBP. The lowest scores were evidenced for del426-7Gly-Thr and Ile488Val. Based on the electrostatic potential analysis, the latter two mutations were excluded from the mutagenesis study. To further investigate the impact of the selected mutations on the structure of RpoB, the binding energy and affinity of RMP for the protein variants were calculated and are shown in Fig. 3. All but one (Gln429His/Asp435Tyr) of the mutants displayed a lower affinity for RMP than the wt protein. However, the effect of most of the mutations was limited. After the Ser450Leu and His445Arg mutations, only the Arg827Cys substitution was shown to be clearly effective in lowering the affinity of RpoB for RMP.

TABLE 2.

Changes in electrostatic potentials of amino acids interacting with RMP

Mutationa	Electrostatic potential (kT/e)	ΔEb
wt	−450
del426-7Gly-Thr	−460	91.5
Gln429His	−338	231.5
Met434Ile	−495	93
Asp435Tyr	−547	129
Arg448Leu	−519	114.5
Ser450Leu	−463	96
Ile488Val	−492	91.5
Arg827Cys	−533	472.5

^aResidues interacting with RMP were inspected in wt RpoB, six selected mutations, and two control mutations (i.e., associated with low [Asp435Tyr] and high [Ser450Leu] MICs of RMP).

^bΔE was calculated as the difference between the total electrostatic charge of amino acids in mutant RpoB and the total electrostatic charge of RBP amino acids in the wt peptide.

FIG 3.

Affinity of RMP for wild-type (WT) and mutant RpoB proteins. An increase of the inhibitor constant predicted by docking simulation (K_i) (in nanomoles) indicates a loss of binding between RMP and the RpoB protein. Analysis involved six selected mutations (del426-7Gly-Thr, Gln429His, Met434Ile, Arg448Leu, Ile488Val, and Arg827Cys), alone (marked in light green) or in combination with mutations that originally co-occurred with them (dark green), and those co-occurring mutations alone. Among the latter, mutation Ser450Leu served as a positive control (red). Results for the Asp435Tyr mutation, reported to produce low RMP MICs, are given in blue. The results are presented as means ± standard deviations (SD) of data from three independent dockings.

Construction of M. tuberculosis mutants and their phenotypes.

Four mutations used for site-directed mutagenesis were chosen based on the highest scores upon predicted changes in the electrostatic charge of the RBP of the RpoB protein. Three M. tuberculosis H37Rv strain-derived mutants, harboring a Gln429His, Met434Ile, or Arg827Cys mutation, were produced. Despite several attempts, the Arg448Leu mutant could not be obtained. As a control, an M. tuberculosis H37Rv mutant with a rpoB Ser450Leu mutation, previously confirmed to generate RMP resistance, was used (24). None of the three isogenic mutants were resistant to RMP (MIC < 40 mg/liter), according to the proportion method. The control Ser450Leu mutant was resistant to 800 mg/liter (Table 3).

TABLE 3.

RMP susceptibility profiles of rpoB mutant strains obtained in this study

Nucleotide mutation	Amino acid mutation	MIC (mg/liter)
wt		<10
G1287C	Gln429His	<40
G1302T	Met434Ile	<40
C2479T	Arg827Cys	<40
C1349T	Ser450Leu	800

Construction of a database of M. tuberculosis rpoB gene sequences.

A great majority (92.3%) of M. tuberculosis strains from which rpoB gene sequences were retrieved originated from 29 countries (see Table S2 in the supplemental material). The geographical distribution of strains in our data set closely resembled the global distribution of M. tuberculosis isolates reported by the WHO for 2016, except that the African region was overrepresented (Fig. S1). Noteworthy, most (74%) of the strains came from only 3 countries, namely, Peru, Russia, and South Africa. These strains are likely to include several lines of strains that are clonal descendants of the common ancestor. Our data set represented all seven known M. tuberculosis spoligotype-based lineages. As expected, modern lineages (Euro-American, East Asian, and East African-Indian) accounted for a vast majority of strains, that is, 97.5% in total. These results corroborate the previously described limited distribution of ancient lineages (25, 26) (Fig. S2). Nearly two-thirds (64.5%) of strains from the database were virtually identified as being RMP resistant.

Retrieval of gene sequences and tests for selective pressure.

Three strains included in the virtual database contained two frameshift mutations each (see Table S3 in the supplemental material), compensatory for each other, so that the rpoB reading frame was maintained. Hence, all the strains contained genes presumably encoding functional RpoB proteins. The virtual database contained strains carrying various mutations identified among clinical strains used in this study (Table S4). When we analyzed if our mutations are linked to other SNPs in rpoB, we found that all strains (n = 7) with mutations at codon 429 always carried mutations at codon 435. Likewise, all strains (n = 26) with mutations at codon 827 also carried mutations at codon 450, and finally, all strains (n = 5) with mutations at codon 434 unexceptionally harbored another SNP in rpoB (Table S3). Importantly, strains identified for each of the above-mentioned codons belong to various lineages of M. tuberculosis, indicating that these mutation patterns arose independently (Table S5).

For tests of selection, 312 alleles of the rpoB gene containing SNPs were found. Polymorphism counts of unique sequences were determined with DnaSP v5 (Table S3). Within the group of alleles containing SNPs, there were 178 segregating sites at the amino acid level. The average number of differences was 4.045, and nucleotide diversity was 0.00115. The most abundant sequence in our data set was that homologous to M. tuberculosis H37Rv (GenBank accession number NC_000962). It was represented by 646 (17.07%) strains. Among the other most abundant sequences (shared by ≥200 strains), compared to H37Rv, 483 (12.76%) strains had a C1349T (Ser450Leu) mutation, 384 (10.15%) strains had a T3225C synonymous mutation, 371 (9.8%) strains had C1349T (Ser450Leu) and T3225C mutations, and 207 (5.47%) strains had an A1304T (Asp435Val) mutation.

The mean ratio of nonsynonymous to synonymous substitutions (dN/dS ratio) across all codons was 0.76, indicating that the rpoB gene is predominantly under negative selection (Table S6). When tested for individual codons with HyPhy, 13 codons were found to be under purifying selection. These codons are expected to be important for the proper functioning of the RpoB protein and can be exploited in future research for the development of novel drugs inhibiting RpoB in M. tuberculosis (Tables S6 and S7). Twenty codons were found to be under positive selection, suggesting their role in escaping selective pressure caused by the use of antibiotics (Table 4 and Table S6).

TABLE 4.

Amino acid sites under positive selection in RpoB of Mycobacterium tuberculosis

Codon	Method(s)a	dN−dSb	Identified in clinical strains in this study	Reference(s)
				Association with drug resistance	Association with compensatory mutations
45	FUBAR	2.51	No		55, 88, 89
265	FUBAR	1.09	No
400	FUBAR	2.50	No	80, 90	55
428	FUBAR	1.09	Yes	80	55
429	FUBAR	1.25	Yes	80	55
430	FUBAR, SLAC	2.26	No	91	55
432	FEL, FUBAR, MEME	2.99	No	91	55
435	FEL, FUBAR, MEME, SLAC	6.03	Yes	91	55
437	FUBAR	1.09	No	80	55
441	MEME	0.72	No	91	55
445	SLAC	6.59	Yes	56	55
448	FEL, FUBAR, MEME	2.01	Yes	80	55
450	SLAC	8.14	Yes	56, 91	55
491	FUBAR	2.41	No	92	55, 88
496	FUBAR	1.99	No		55, 88
672	FUBAR	1.50	No		55
674	FUBAR	1.09	No
695	FUBAR, MEME, SLAC	4.02	No		55
835	FUBAR	1.45	No		55, 93
957	FUBAR	1.13	No

^aFor methodological information, see the virtual database screens and tests for selection in Materials and Methods.

^bRate of nonsynonymous (dN) to synonymous (dS) nucleotide substitutions as calculated with MEGA 7.0.

DISCUSSION

As evidenced in a number of studies, the most frequently mutated codon in the rpoB gene is codon 450 (corresponding to codon 531 of the E. coli numbering system), and the predominant mutation in this codon is Ser450Leu (C1349T) (27). This mutation ranks first in RMP-resistant strains globally, although its frequencies differ between different geographic locations. The mutation is highly prevalent in Kazakhstan (80.9%) (28), Taiwan (66.7%) (29), Italy (59.5%) (30), and China (59.1%) (15); equally common in South Korea (53.1%) (20), Brazil (52.4%) (31), Bangladesh (52.3%) (12), and Australia (52%) (9); and somewhat rarer in Spain (42.6%) (32), Russia (41.5%) (33), Mexico (40.4%) (34), Vietnam (37.8%) (35), and Hungary (31%) (36).

After codon 450, the next rpoB codons at which mutations occur most abundantly in RMP-resistant strains are codons 435, 445, and 452. The mutation frequencies at these codons are within the ranges of 1.1 to 20.4%, 6.8 to 32%, and 0 to 8%, respectively (9, 12, 15–17, 20, 23, 28, 32, 33, 35, 37). In this study, the most frequently affected codons were codon 450 (50.8% of all RMP-resistant strains), codon 435 (16.9%), and codon 445 (15.4%). A single Ser450Leu substitution predominated, totaling 49.2% of all RMP-resistant strains tested.

The differences in the mutability of rpoB codons and the prevalence of specific rpoB mutations can be explained in terms of fitness costs associated with the adverse effects that these mutations possibly exert on RNAP functions and DNA transcription. This is best illustrated for the Ser450Leu mutation. Strains with this substitution were shown to impart no fitness deficit relative to other rpoB mutants (38). The selective advantage of these mutants makes them persist and spread more successfully in human populations. Thus, the overall fitness of different rpoB mutants translates into their clinical frequency. This correlation also exists for mutations conferring resistance to isoniazid and streptomycin, with low-cost resistance mutations being katG Ser315Thr and rpsL Lys43Arg, respectively (38). This type of mutation is particularly expected in strains with MDR and extensively drug-resistant (XDR) phenotypes, in which the accumulation of resistance-associated mutations would normally be detrimental to the survival of the bacilli. One observation confirming this is the exceptionally high proportion of the rpoB Ser450Leu mutation among XDR strains (39–41).

The differences in the rpoB mutation profiles between the RMP-monoresistant and MDR M. tuberculosis strains remain elusive. Findings from the very few studies that analyzed these two phenotypic groups separately were consistent in that the most common mutation, for both RMP-monoresistant and MDR strains, was Ser450Leu, followed by His445Tyr. Whereas the former mutation occurred at a higher proportion in MDR strains (58.4% versus 36.4%, or 64.7% versus 57.1%), the latter one was more prevalent among RMP-monoresistant strains (36.4% versus 13.4%, or 28.6% versus 0%) (42, 43). However, the significantly uneven proportions of the two strain groups may add an important bias to these results.

In this study, 16 mutation types were located within the RRDR of the rpoB gene (codons 426 to 452). Of these, all but one (del1275-80CGGCAC) had been reported previously, and all (at codons 428, 435, 445, 450, and 452) but four mutations (Gln429His, Met434Ile, Arg448Leu, and del1275-80CGGCAC) had already been shown to confer RMP resistance (12, 15, 20, 35, 44). The role of these four mutations in the development of RMP resistance could not be unambiguously established, as they co-occurred with other mutations known to confer RMP resistance, both in this study and in previous ones (12, 45, 46).

Six different mutation types outside the RRDR were observed in 20 (30.8%) RMP-resistant strains. Three of these mutations were synonymous, two of which (C309T [Asp103Asp] and T3225C [Ala1075Ala]) were described previously (16, 47) and one of which (G1482A [Leu494Leu]) was hitherto unreported. Mutations C309T (Asp103Asp) and T3225C (Ala1075Ala) are unlikely to be associated with RMP resistance, as they have been found in both RMP-resistant and -susceptible strains (47, 48). When analyzed with the virtual database, codons 103 and 1075 were found to be under purifying selection, suggesting that their maintenance over the course of evolution is probably because of their importance for the proper functioning of RpoB.

Of the three nonsynonymous mutations located outside the RRDR, two (Val170Phe and Ile488Val) had already been described (10, 16, 21, 49), and one (Arg827Cys) was identified for the first time. Among the non-RRDR rpoB mutations detected, only Val170Phe was definitively proven to confer RMP resistance in M. tuberculosis (16). Mutation Ile488Val was found only once prior to this study and, as in the present report, in a single MDR strain, together with Ser450Leu, a well-known RMP resistance mutation. Due to this coincidence, the contribution of the Ile488Val mutation to RMP resistance remains unclear.

A closer look has to be taken at 9 strains, whose rpoB gene sequences were either of the wild type (7 strains) or silently mutated at one codon (C309T [Asp103Asp]) (2 strains). Seven of these strains, although classified as being RMP resistant upon initial susceptibility testing, failed to grow at a critical RMP concentration (40 mg/liter) in the MIC determination assay but were still able to grow at 20 mg/liter of RMP, a concentration 4 times higher than that tolerated by a control strain, H37Rv (5 mg/liter). It thus seems that these strains are borderline resistant to RMP and that the mechanism responsible for this type of resistance must not necessarily involve alterations in the rpoB gene. One MDR strain and one RMP-monoresistant strain, presenting with high MIC values of 200 and 800 mg/liter, respectively, also had a wild-type rpoB gene. Strains resistant to RMP that lack rpoB mutations are rare but have been acknowledged in the literature (47, 50, 51). The mechanisms underlying RMP resistance in such strains may involve alterations in the drug's permeability or metabolism or modifications in other RNA polymerase subunits. Regarding the latter, Comas et al. recently showed that RMP-resistant strains without rpoB mutations do not contain mutations in rpoA or rpoC either. Mutations at either of these two loci were found only if a mutation in the RRDR was present (47). On the other hand, Wang et al. discovered a mutation (A191C [Asp64Ala]) in the Rv2629 gene, carried by over 99% of RMP-resistant strains, including all those with no rpoB mutations detected. Since Rv2629 is hypothesized to encode one component of a drug efflux pump, the newly described mutation may change the structure of the pump and decrease the uptake of RMP in RMP-resistant strains (50). Neither of the two high-MIC and wt rpoB strains from this study harbored the A191C substitution (data not shown).

Resistance to RMP is usually caused by mutations in the RMP binding pocket (RBP) of RNAP, which replace an amino acid with a compact side chain with a larger one and thereby block the access of the relatively inflexible RMP molecule to the RBP (24, 52). Expectedly, there is a strong correlation between the MIC values of RMP and the distance of the mutated residues to the binding site of the drug in the RpoB structure (53).

According to this rule, most of the analyzed mutations in the rpoB gene should have a noticeable effect on RMP resistance, since the majority of the affected residues are located in the vicinity of RBP (Fig. 2). However, while in silico experiments have predicted important alterations within the RBP affecting RMP-simulated binding, the mutations failed to produce RMP resistance in isogenic mutants. Of the four mutations selected for in vivo model experiments, three affected residues inside (Gln429 and Arg448) or close to (Met434) the RBP, and one mutation involved a distant amino acid (Arg827). Given the localization and the size of the introduced amino acid, the Gln429His substitution should have the most distinct effect on RMP docking (24, 52). Also, codon 429 was found to be under positive selection, and it was previously associated with the acquisition of RMP resistance and a compensatory mechanism (54, 55). Despite structural and electrostatic changes in the RBP, this mutation resulted in only a slightly diminished affinity of RpoB for RMP upon docking simulation, and the isogenic Gln429His mutant was susceptible to RMP. In this study, as in the virtual database, strains with a mutation at codon 429 always harbored an additional mutation at codon 435. Hence, mutations at codon 429 may act as compensatory mutations for those at codon 435, conferring RMP resistance. Although the Met434 residue does not directly interact with RMP, its proximity to the RBP combined with noticeable changes, upon replacement, in the total electrostatic energy of the RBP and the affinity for the drug prompted us to include this mutation in the mutagenic study. The MIC value of the isogenic mutant was below 40 mg/liter, indicating susceptibility to RMP. Similarly to codon 429 mutations, strains with a mutated codon 434 always had alterations at other codons (codon 430 or 435), as evidenced in the virtual database. Therefore, we hypothesize that a codon 434 mutation also serves as a compensatory mechanism. The Arg448Leu mutation replacing arginine with a smaller leucine had a greater effect on relaxing RMP-RpoB affinity than the Gln429His substitution. This affinity was calculated for this mutant alone and for a double mutant (Arg448Leu/His445Arg), as originally found in a clinical strain. The affinity of the latter was much lower than those of the single mutants, indicating that both residues (His445 and Arg448) are involved in the RMP-RpoB interaction albeit with a more pronounced role of the His445 amino acid. Both codons 445 and 448 were found to be under positive selection and had previously been associated with RMP resistance and a compensatory mechanism (54–56). Interestingly, despite several attempts, the Arg448Leu mutant could not be obtained. This mutation was not found in the virtual database, although other variants (Arg448Lys and Arg448Gln) had been reported. It thus seems that the amino acid at position 448 is important for the proper functioning of RpoB and that mutation Arg448Leu alone is unstable and requires additional modifications in the RpoB structure.

Even more enigmatic is the involvement of the Arg827Cys substitution in the causation of RMP resistance. While in silico analysis revealed its prevailing role in the RMP-resistant phenotype, over the Val170Phe mutation, with which it originally co-occurred, no significant departure of nucleotide substitution rates at this codon was detected in the virtual database. Moreover, all strains in the database with mutations at codon 827 also harbored a mutation at codon 450. The Arg827Cys isogenic mutant had an RMP MIC below the breakpoint for resistance. This again shows that mutations at codon 827 rather compensate for the loss of fitness incurred by RMP resistance due to other rpoB mutations (e.g., at codon 450).

Several studies have attempted to correlate different rpoB mutations with different levels of RMP resistance. For instance, despite different MIC cutoff values, mutations at codon 450, specifically mutation Ser450Leu, have consistently been associated with high levels of RMP resistance (29, 51, 57). This correlation was apparent in this study, where Ser450Leu mutants had mean MICs of RMP that were >3-fold higher than those observed for all other resistant strains (i.e., without this mutation). Quite conversely, the mean MIC values for the Asp435Tyr mutants were somewhat lower than those for other rpoB mutants not having this alteration. Still, the MICs for the Asp435Tyr mutants were high (MIC range, 100 to 800 mg/liter; GM MIC, 362.5 mg/liter), which is in contrast to previous studies suggesting that mutations at codon 435, including Asp435Tyr, confer low levels of RMP resistance (11, 37, 58). Also, in our virtual simulations, the RMP-RpoB affinity of the Asp435Tyr mutant was only minimally lowered compared to that of the wt protein (Fig. 3). Finally, mutations at codon 435 were found to be under positive selection, pointing to their adaptive advantage.

Given the structural relatedness and similar modes of action, one would expect that all rifamycins share common resistance mechanisms, resulting in complete cross-resistance between all drugs from this group. This cross-resistance occurs, but strains resistant to RMP and susceptible to other rifamycins have been well described (9, 48, 51, 59). In this study, 57% of RMP-resistant strains were cross-resistant to rifabutin. This rate is relatively low compared with those in previous studies, in which cross-resistance between the two drugs was observed for 72.2 to 85.4% (9, 46, 51, 59) of RMP-resistant strains. Importantly, RMP-resistant/rifabutin-susceptible patterns could be linked to certain rpoB mutations, including Met434Ile, Asp435Tyr, Asp435Val, or His445Leu (46, 51, 59), and this was confirmed in the present study. All these mutations were found exclusively in strains displaying an RMP-resistant and rifabutin-susceptible phenotype. Thus, the detection of these mutations might be indicative of not only RMP resistance but also susceptibility to rifabutin, with all its clinical consequences. The replacement of RMP by rifabutin may improve treatment outcomes for MDR- and XDR-TB patients (60). Moreover, rifabutin can be continued for HIV-coinfected TB patients receiving a protease inhibitor (PI)-based regimen, because the risk of substantial drug interactions with PIs is lower with rifabutin than with RMP (61).

In conclusion, this study, by interrogating the entire rpoB gene sequence, provides a comprehensive characterization of mutations conferring resistance to RMP in M. tuberculosis clinical strains. The key findings of this study can be summarized in three points. First, some within-RRDR mutations are not involved in the development of RMP resistance. Instead, they may act as compensatory mutations to alleviate fitness impairment incurred by other mutations directly associated with drug resistance. Second, mutations outside the RRDR are rare, and their role in the causation of RMP resistance seems to be limited. Third, strains with a wild-type rpoB gene occur, suggesting other unknown mechanisms responsible for the RMP-resistant phenotype.

MATERIALS AND METHODS

Strains and drug susceptibility testing.

A total of 115 M. tuberculosis strains were examined, comprising 15 RMP-monoresistant, 50 MDR, and 50 pansusceptible strains. All the strains were isolated from single pulmonary TB patients (91 male and 24 female patients; age range, 19 to 93 years; median age, 53 years), diagnosed in different TB dispensaries across Poland between 2004 and 2013. The study sample represented 100%, 19%, and <0.1% of all bacteriologically confirmed RMP-monoresistant, MDR, and drug-susceptible TB cases, respectively, reported in Poland over that 9-year period.

Primary isolation, culturing, and preliminary, phenotype-based identification of mycobacteria from clinical specimens were carried out in the regional TB laboratories, using standard procedures. The isolates were then sent to the National TB Reference Laboratory, where they underwent confirmatory identification to the species level according to methods described previously (62). There, DST was also performed with the conventional 1% proportion method on Löwenstein-Jensen (L-J) medium (62). For all strains resistant to RMP (growth at a critical concentration of 40 mg/liter), the MICs of RMP were determined. This was done in duplicate, in L-J medium containing incremental concentrations of the drug (40, 100, 200, 400, and 800 mg/liter). The MIC was defined as the lowest concentration of RMP that inhibited more than 99% of the bacterial growth compared to a drug-free control. The pansusceptible M. tuberculosis strain H37Rv was used as a reference.

DNA extraction.

Chromosomal DNA was extracted from M. tuberculosis cultures on L-J slants by the lysozyme-proteinase K cetyl-trimethyl ammonium bromide (CTAB) method (63). The purified DNA was quantified with the NanoDropND-1000 spectrophotometer (NanoDrop Technologies).

Sequencing of rpoB.

For all strains, the complete, 3,519-bp-long rpoB gene was amplified as two partially overlapping fragments by using two primer pairs, rpoB1-F/R and rpoB2-F/R (Table 5). The 30-μl PCR mixtures contained 15 μl of the HotStarTaq master mix (Qiagen), 0.3 μM forward (rpoB1-F or -R) and reverse (rpoB2-F or -R) primers each, and 1 μl (ca. 20 ng) of template DNA. All PCRs were run on a T100 thermal cycler (Bio-Rad) under the following amplification conditions: an initial denaturation step at 95°C for 3 min, followed by 35 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 90 s, with a final extension step at 72°C for 7 min.

TABLE 5.

Primers used for amplification and sequencing of the complete rpoB gene and for amplification of mutated variants of the rpoB gene used in directed mutagenesis

Primera	Sequence (5′→3′)	Ta (°C)b	Amplicon size (bp)
Amplification and sequencing of the complete rpoB gene
rpoB1-F	CTTTGTTCGTGGTGAGCGTGAG
rpoB1-R	TCGATCGGCGAATTGGCCTGTG	55	2,193
rpoB2-F	GGTCAACCCGTTCGGGTTCATC
rpoB2-R	GTTGATCGTCTCCGGCTTTTTG	55	2,172
rpoB3-F*	ACGAATCGGTCTGGACGTCAAG
rpoB4-R*	CGCGAACCACTTGAGGTTCC
rpoB5-F*	CATGCAGCCCGAGCTTCTTG
rpoB6-R*	TCAAGGAGAAGCGCTACGAC
rpoB7-F*	TCGAGGTGAGCACGTCCTCTTC
rpoB8-R*	ACGAGGACGCGATCATCCTG
rpoB9-F*	CGTCGAAGAGCATGGCCTTG
rpoB10-R*	GAGCTGCAGGGCCTGTTGTC
Amplification of mutated variants of the rpoB gene used in directed mutagenesis
no mut-F	TCGGCGTGCGCATCGACC
no mut-R	GCACGCAGCAGCCGCTCCT	60	1,804
1287 G>C-F	CTTCGGCACCAGCCACCTGAGCCAATTCATG
1287 G>C-R	CATGAATTGGCTCAGGTGGCTGGTGCCGAAG	65	1,804
1302 G>T-F	GCTGAGCCAATTCATTGACCAGAACAACCC
1302 G>T-R	GGGTTGTTCTGGTCAATGAATTGGCTCAGC	63	1,804
1333 A>G-R	AAGTGTCGCGCACCTCGCG
1333 A>G-R	AAGTGTCGCGCACCTCGCG	65	1,852
1343 G>T-F	TGACCCACAAGCGCCTACTGTCGGCGCTGG
1343 G>T-R	CCAGCGCCGACAGTAGGCGCTTGTGGGTCA	65	1,804
2479 C>T-F	AAGGCCCGCGAGGTGTGCGACACTTCGCTG
2479 C>T-R	CAGCGAAGTGTCGCACACCTCGCGGGCCTT	65	1,470
1349 C>T-F	CGCGACACCGTCGGCGTGCG
1349 C>T-R	AAGTGTCGCGCACCTCGCG	65	1,852

^aPrimers rpoB1-F and rpoB2-R correspond to primers rpoB-pcrF and pcrB-pcrR, respectively, designed by Siu et al. (16); all the remaining primers were newly designed in this study, based on the sequences of the respective genes of M. tuberculosis reference strain H37Rv, available at the TubercuList database (http://genolist.pasteur.fr/TubercuList/). The letters “F” and “R” designate forward and reverse primers, respectively. The asterisks indicate primers used for sequencing only.

^bT_a, temperature of primer annealing.

The amplicons were sequenced by the Sanger method. A total of 12 primers, targeting both strands of the rpoB gene were employed for sequencing. The 5′ segment of the gene was sequenced with three forward (rpoB1, -3, and -5) and three reverse (rpoB1, -4, and -6) primers. Another three forward (rpoB1, -7, and -9) and reverse (rpoB2, -8, and -10) primers were used for sequencing of the 3′ segment of the rpoB gene (Table 5).

The sequence data were assembled and analyzed with ChromasPro (version 1.7.1) software (Technelysium), and the resulting consensus sequences were aligned with the corresponding wild-type (wt) rpoB sequence of the reference M. tuberculosis H37Rv strain (TubercuList [http://genolist.pasteur.fr/TubercuList/]) using the BLASTN algorithm (http://blast.ncbi.nlm.nih.gov/).

To establish whether a given mutation had or had not been described previously, the mutated allele sequences were searched against the following interactive, publicly available databases: GenBank (National Center for Biotechnology Information) (http://www.ncbi.nlm.nih.gov/); the TB Drug Resistance Mutation Database (http://www.tbdreamdb.com/index.html); MuBII (https://umr5558-bibiserv.univ-lyon1.fr/mubii/mubii-in.cgi), and the Broad Institute database (https://olive.broadinstitute.org/projects/Mycobacterium%20tuberculosis%20diversity).

Codon numbers were based on the M. tuberculosis rpoB gene strain (TubercuList [http://genolist.pasteur.fr/TubercuList/]). For comparative purposes, all rpoB codons were also reported using the E. coli numbering system.

Statistical analysis.

The associations between mutation data from PCR sequencing assays and data from drug susceptibility testing were assessed by unpaired Student's t test (parametric) or a Mann-Whitney U test (nonparametric). A P value of <0.05 was considered significant. For all analyses, the SPSS statistical software package (version 20.0; SPSS Inc., USA) was used.

Structural modeling and docking simulations.

The three-dimensional structural models of wt and mutant RpoB proteins were constructed by homology modeling. The M. tuberculosis RpoB sequences were threaded with the template of the Thermus thermophilus RNA polymerase structure (PDB accession number 2A69) (64). The models were obtained, verified, and optimized by using the Swiss Model server and Swiss-PDB viewer (65, 66). Residues interacting via van der Waals forces (Gln429, Leu430, Leu452, Gly453, Glu484, and Ile491) and hydrogen bonding (Gln432, Phe433, Asp435, His445, Arg448, and Ser450) with the RMP binding pocket (RBP) (67, 68) were inspected for changes in their electrostatic potential. Changes in these residues in mutant proteins were compared to wt RpoB and scored. Computations were done with the Gromos96 implementation of the Swiss-PDB viewer. The wt and mutant RpoB protein models were docked with RMP to identify changes in binding patterns by using AutoDock 4.2 (69, 70). The three best-docked conformations were selected for evaluation, and the average affinity of RMP for RpoB was calculated. The results were averages of data from three independent dockings. RpoB structures were visualized with PyMOL 0.97 software (DeLano Scientific).

Virtual database screens and tests for selection.

The rpoB gene sequences were retrieved from genome sequences of M. tuberculosis strains deposited in the Genome Database (GD) of GenBank at the NCBI. All strains included had <150 contigs per genome in order to ensure good quality of extracted sequences. The final data set consisted of 3,796 rpoB gene sequences originating from as many M. tuberculosis strains (see Table S2 in the supplemental material). Information regarding the geographic origin of the strains was retrieved from the GD. Spoligotyping of the strains was performed through a genome BLAST search for previously described oligonucleotide sequences, as previously reported for SpoTyping (71). As with SpoTyping, a maximum of one mismatch within a 25-bp spacer was allowed to be considered a hit. The accuracy of this approach was verified by comparing our results with previously reported spoligotypes for laboratory strains included in the virtual database (72, 73). The genetic lineages and families were determined with the TB-Insight (http://tbinsight.cs.rpi.edu/), TB-Lineage (74), and SPOTCLUST (75) tools.

The excerpted rpoB gene sequences were processed and analyzed with Geneious R11 (76). This program allowed alignment of sequences with MUSCLE, determination of sequence variation estimates in reference to M. tuberculosis H37Rv (GenBank accession number NC_000962), and separation of unique sequences. Resistance to RMP was determined in silico upon the detection of mutations detectable with one of the following tools: KvarQ, Mykrobe Predictor TB, PhyResSE, or TBProfiler (54, 77–80).

Sequences from 12 strains containing insertions/deletions and those with ambiguous bases were excluded. Polymorphisms of unique sequences, in terms of the total number of mutations, average number of nucleotide differences, and average nucleotide diversity, were determined with DnaSP v5 (81).

The relative ratios of nonsynonymous (dN) to synonymous (dS) nucleotide substitutions (82) in rpoB were estimated. Nonsynonymous substitutions, which may reduce the functionality of the protein, are expected to be eliminated by negative (purifying) selection (dN/dS ratio of <1), while when new mutations are adaptive, a gene is expected to be under positive (directional or diversifying) selection (dN/dS ratio of >1). Finally, when genes become nonfunctional, nonsynonymous nucleotide substitutions are expected to accumulate at a similar rate as synonymous substitutions (dN/dS ratio of ∼1) (83, 84). The rpoB gene sequences were tested for the presence of codon-specific signatures of positive and negative selection using the DataMonkey online server for the HyPhy package. A significant departure of codon-specific nucleotide substitution rates was tested with three different methods: the mixed-effects model of evolution (MEME), single-likelihood ancestor counting (SLAC), fixed-effects likelihood (FEL), and fast unconstrained Bayesian approximation (FUBAR) (85–87). Whereas MEME, SLAC, and FEL use a combination of maximum likelihood (ML) and counting approaches to infer rates of nonsynonymous (dN) and synonymous (dS) substitutions on a per-site basis for a given coding alignment and the corresponding phylogeny, FUBAR uses hierarchical Bayesian methods that implement a Markov chain Monte Carlo (MCMC) routine to ensure robustness against model misspecifications (87). Cutoff P values for MEME, SLAC, and FEL were set at 0.05; the cutoff limit for FUBAR was a P value of 0.9. Codon-specific estimates of substitution rates for each codon were obtained with MEGA 7.0 software (58).

Generation of M. tuberculosis mutants.

Selected mutations were introduced into the rpoB gene through directed mutagenesis, essentially as described previously (52). The mutated rpoB genes under the control of their own promoter were then cloned into the pMV306 integration vector and introduced into the attB chromosomal locus of the M. tuberculosis H37Rv strain, as described previously (24). Primers used in the direct mutagenesis experiments are given in Table 5. The presence of the integrated plasmid with the mutation was confirmed by PCR sequencing of the partial rpoB gene. The mutant strains were tested for RMP resistance with the proportion method, as described above.

Accession number(s).

The nucleotide sequences of the mutant rpoB genes (one sequence for each type of mutant gene) identified among RMP-monoresistant and MDR M. tuberculosis strains were deposited in the GenBank database (NCBI [http://www.ncbi.nlm.nih.gov/]) under accession numbers KP744369 to KP744373 and KJ095683 and under accession numbers KF771050 to KF771067, respectively. The sequences with new, previously undescribed mutations have the following accession numbers: KF771054, KF771061, and KF771067.