Tuberculosis (TB), caused by Mycobacterium tuberculosis complex (MTBC), is a major threat to public health worldwide (World Health Organization, 2016). In 2015, an estimated 10.4 million new TB cases emerged, and 1.4 million died of the disease globally (World Health Organization, 2016). Although the incidence and mortality rates of TB are slowing declining, the current advances in TB control remains impeded by the alarming increase in reports of drug-resistant TB, especially multidrug-resistant TB (MDR-TB), defined as resistant to at least rifampicin (RIF) and isoniazid (INH) (Gandhi et al., 2010; Jamieson et al., 2014). RIF is a cornerstone of TB treatment and is a key factor in determining the treatment efficacy of the treatment regimens (Somoskovi et al., 2001; Pang et al., 2013a). As more than 90% of RIF-resistant TB strains are simultaneously resistant to INH, RIF resistance is considered as a promising marker for the diagnosis of MDR-TB, and RIF-resistant cases are also eligible for MDR-TB treatment if the INH susceptibility is inaccessible in clinical practice (Garcia de Viedma et al., 2002; Pang et al., 2013a).

Rifabutin (RFB) is a semisynthetic derivate of rifamycin S (Marsili et al., 1981; Uzun et al., 2002), and is part of the rifamycin family together with RIF. Despite sharing several of the common properties of RIF, RFB exhibits more potent efficacy against M. tuberculosis, as well as M. avium complex and M. leprae (Kunin, 1996). In addition, due to fewer interactions with protease inhibitor drugs, RFB is recommended to be used as an alternative to treat MTB in HIV-infected patients (Berrada et al., 2016). Resistance to both RIF and RFB is largely associated with mutations in an 81-bp RIF resistance determining region (RRDR) between rpoB codons 507 and 533 of MTB (Rukasha et al., 2016), and mutations within codons 516, 526, and 531 are responsible for up to 90% of RIF-resistant strains (Pang et al., 2013a). High-level cross-resistance between the two rifamycins is noted by numerous studies (Beckler et al., 2008; Berrada et al., 2016; Rukasha et al., 2016), whereas not all mutations within the RRDR display the same loss of rifamycin susceptibility, and only some specific rpoB mutations confer RFB resistance in MTB (ElMaraachli et al., 2015). Hence, a detailed analysis of the association between RFB susceptibility and genetic mutations within rpoB gene would provide new insights for guiding RFB-based therapeutic regimens. In China, due to the limited data regarding this issue, RFB is not routinely used in the treatment of RIF-resistant TB cases, which make the cases lose the opportunity to access the potentially effective RFB. In this study, we aimed to investigate the cross-resistance between RIF and RFB among clinical MTB isolates. In addition, the correlations between specific rpoB mutations and the minimum inhibitory concentrations (MICs) of RIF and RFB were analyzed, which will lay a foundation to establish criteria to predict in vitro susceptibility of MTB against RFB by molecular detection of rpoB gene.

Although high-level cross-resistance between RIF and RFB has been repeatedly observed in MTB isolates, many studies have demonstrated that specific rpoB mutations are associated with differential resistance to RIF and RFB (Jamieson et al., 2014; ElMaraachli et al., 2015). Our data were in line with previous findings, which have consistently shown that the rpoB mutations S531L, H526D, H526R, H526C, and D516V conferred phenotypical resistance to both RIF and RFB, whereas amino acid substitution at codons L511P, D516G/Y, S522L, H526Y/L/D/N, S531Q, and L533P were associated with phenotypic resistance to RIF and susceptibility to RFB. In addition, we detected 11 different double mutations within rpoB from a total of 26 strains studied, including 3 infrequently encountered double mutations that have not been described elsewhere (Sandgren et al., 2009). Interestingly, a single specific mutation in codons 511, 516, and 526 was linked to the susceptibility to RFB for MTB, while the strains with these double mutations irrelevantly conferring RFB resistance produced RFB-resistant phenotype. The increased resistance to rifamycin contributes to the loss of binding affinity between rifamycin and RpoB protein (Pang et al., 2013a). The high MIC values of Ser531Leu and His526Asp mutants are due to the low affinities to the rifamycin molecules. For MTB isolates with specific mutations in codons 511, 516, and 526, we hypothesize that the structural alternations of the rpoB may be mediated by the presence of these amino acid substitutions. Despite being no correlated with RFB resistance, these structural alternations could result in loss of RFB susceptibility compared with isolates harboring wild-type rpoB genotype. Although the exact reason is unknown, the potential synergistic effect between double specific mutations may contribute to the conversion of RFB susceptibility among MTB isolates. It is important to noted that the majority of these RFB-resistant isolates with double mutations had low RFB MICs (≤4 mg/L), which may reflect the moderate synergistic effect of the two mutations within rpoB. Further structural analysis of RFB and these mutant RpoBs will be required to confirm our hypothesis.

Another interesting finding of our results was that the rate of RFB-resistant isolates with double mutations was significantly lower than that of isolates with a single mutation. One possible reason for this observation is that the double mutations identified in our study majorly consisted nucleotide substitutions located in codon 511, 515, 522, and 533 of rpoB, which confer low-level RIF resistance, while have no correlation with RFB resistance (Berrada et al., 2016). In view of the low-level RIF MICs of the strains with these single mutations, the high exposure to RIF may accelerate the occurrence of the second mutation. The acquisition of genetic mutation often comes with a cost to strain fitness (Mariam et al., 2004; Vogwill and MacLean, 2015). Briefly, the mutations resulting in higher level of drug resistance are associated with the higher loss of fitness in MTB isolates (Mariam et al., 2004). Hence, the MTB strains with single mutation conferring low-level resistance prefer to have a second accumulated mutation with less loss of fitness, which might be essential for survival from interspecific competitive pressures. This serves as a reasonable explanation for the high frequency of mutations conferring low-level resistance among double mutant MTB isolates. In addition, the proportion of strains with double mutations among RR group was significantly higher than that of MDR-TB group, indicating double mutations in the core region of rpoB gene lead to higher fitness cost than a single mutation, and a high rate of strains with double mutations are excluded during the period of the further accumulation of INH resistance.

Roll-out of molecular diagnostics provides an alternative for clinicians to obtain drug susceptibility results within shorter turn-around time than conventional DST (Pang et al., 2013b). Numerous commercial kits have been developed to analyze RIF susceptibility of MTB on the basis of detecting the core region of rpoB gene (Pang et al., 2013b, 2014). However, in light of the differential resistance to RIF and RFB with various rpoB mutations in MTB isolates, the WHO-recommended GeneXpert assay and GenoType MTBDR are not suitable for predicting RFB resistance due to failure to interpret the exact mutant types of rpoB gene. Specially, several previous studies have demonstrated that RFB-containing regimen produces more favorable efficacy against RFB-susceptible MDR-TB cases compared with RFB-resistant MDR-TB receiving other DST-guided regimen (McGregor et al., 1996; Jo et al., 2013). Therefore, accurate and rapid diagnostics designed for RFB is essential for guiding RFB-based therapeutic regimens and achieving favorable treatment outcome.

There were several obvious limitations in this study. First, the Clinical and Laboratory Standards Institute recommends the use of the standardized agar proportion method for susceptibility testing of MTB isolates. Although there were strong evidences that the excellent correlation was observed between the resistance determined by agar proportion method and microplate alamarBlue assay method (Chauca et al., 2007; Yu et al., 2011), the difference in methodology may interfere with the analysis of the relationship between in vitro phenotypic resistance and genotypic mutations. Second, the MTB isolates enrolled in this study were only collected from one clinical hospital in Beijing. As the National Clinical Center on TB, patients from different regions of China seek health care in our hospital. However, the bias in sample enrollment may result in loss of representativeness. Further experiments should be performed to explore the molecular characteristics among RFB-resistant isolates from different regions of China, especially the prevalence of strain with novel double mutation identified in our study. Third, due to the sample number of the strains with rare mutation patterns, they were excluded from statistical analysis. The relationship between these mutations and the level of drug resistance therefore is still unknown. There is increasing interest in the extent to answer this question by the recruitment of a large number of strains with rare mutations in the future.

In conclusion, our data demonstrate that various rpoB mutations are associated with differential resistance to RIF and RFB. A single specific mutation in codons 511, 516 and 522 was linked to the susceptibility to RFB for MTB, while the strains with these double mutations irrelevantly conferring RFB resistance produced RFB-resistant phenotype. In addition, compared with RR group, MDR-TB has significantly higher proportion of isolates with mutations in codon 531 of rpoB gene. Further structural analysis of RFB and the rpoBs with double mutations will extend our knowledge of the RFB resistance mechanism in MTB.

WJ, YP, HH, and NC designed this study and wrote the manuscript. WJ, YP, ZZ, GJ, YM, and FH performed experiments. WJ, YP, JW, and RG interpreted the data. All authors approved the final version of the paper.