Variants associated with Bedaquiline (BDQ) resistance identified in Rv0678 and efflux pump genes in Mycobacterium tuberculosis isolates from BDQ naïve TB patients in Pakistan

Mutations in the Rv0678, pepQ and atpE genes of Mycobacterium tuberculosis (MTB) have been reported to be associated with reduced antimycobacterial susceptibility to bedaquiline (BDQ). Resistance conferring mutations in treatment naïve MTB strains is likely to have implications for BDQ based new drug regimen that aim to shorten treatment duration. We therefore investigated the genetic basis of resistance to BDQ in MTB clinical isolates from BDQ naïve TB patients from Pakistan. In addition, mutations in genes associated with efflux pumps were investigated as an alternate mechanism of resistance. Based on convenience sampling, we studied 48 MTB clinical isolates from BDQ naïve TB patients. These isolates (from our strain bank) included 38 MDR/pre-XDR/XDR (10 BDQ resistant, 8 BDQ intermediate and 20 BDQ susceptible) and 10 pan drug susceptible MTB isolates. All strains were subjected to whole genome sequencing and genomes were analysed to identify variants in Rv0678, pepQ, atpE, Rv1979c, mmpLS and mmpL5 and drug resistance associated efflux pump genes. Of the BDQ resistant and intermediate strains 44% (8/18) had variants in Rv0678 including; two reported mutations S63R/G, six previously unreported variants; L40F, R50Q and R107C and three frameshift mutations; G25fs, D64fs and D109fs. Variants in efflux pumps; Rv1273c (G462K), Rv0507c (R426H) and Rv1634c (E198R) were found to be present in drug resistant isolates including BDQ resistant and intermediate isolates. E198R in efflux pump gene Rv1634c was the most frequently occurring variant in BDQ resistant and intermediate isolates (n = 10). We found RAVs in Rv0678 to be commonly associated with BDQ resistance. Further confirmation of the role of variants in efflux pump genes in resistance is required so that they may be incorporated in genome-based diagnostics for drug resistant MTB.

re-purposed drugs have been introduced to improve outcomes of patients with drug resistant tuberculosis [1]. Bedaquiline (BDQ); a novel diaryl quinolone, has been shown [2] to be effective against both actively replicating and dormant Mycobacterium tuberculosis (MTB). The increased efficacy of anti-mycobacterial regimens that include BDQ as treatment for MDR TB is well illustrated [2][3][4]. BDQ along with delamanid (DLM) was given accelerated approval by the United States Federal Drug Authority (U.S F.D.A) for treating M/XDR TB [2], highlighting the role of BDQ in circumventing the challenge of treating drug resistant TB. As of 2018, the World Health Organization (WHO) has included BDQ as an essential drug to be used along with fluoroquinolones and linezolid for treating MDR TB [5].
Studies investigating mutations conferring resistance to BDQ, derived in vitro [6] or in clinical isolates with high minimum inhibitory concentrations (MICs) posttreatment [6,7], have identified resistance associated variants (RAVs) arising in atpE, Rv0678, and pepQ genes. Genomic variants D28N and A63V in atpE gene in two BDQ resistant clinical isolates have also been reported [8]. Moreover, RAVs in Rv0678 have been reported in BDQ-naïve MDR TB population [7,9], with one study reporting 4.6% of BDQ-naïve MDR MTB isolates carrying RAVs in Rv0678 and pepQ genes [9]. Worse treatment outcomes are reported in patients infected with MTB strains possessing baseline or emergent Rv0678 variants [10,11]. Similarly, co-occurrence of loss of function mutations in Rv0678 and mmpL5 in MTB isolates has also been reported [12]. These data underscore the need to continue surveillance of BDQ resistance and to establish presence of RAVs in Rv0678 and mmpLS-mmpL5 in baseline MDR TB isolates from patients before starting therapy. Keeping in mind the diversity of RAVs in Rv0678 [9,13,14] and variations in associated MICs [9], mutation hotspots should be catalogued and characterized. Furthermore, genomic variants associated with MICs to BDQ need to be explored towards interpreting molecular results with confidence. Moreover, the recently published catalogue of resistance associated genetic variants by the WHO [15] does not categorize any of the variants in Rv0678, Rv1979c, and mmpL5 to be associated with BDQ resistance due to the rarity of the variants reported in these genes. As more data is collected on RAVs for BDQ there will be greater information available regarding the sensitivity of these mutations in predicting a drug resistant genotype.
Efflux pumps (EPs) are recognized as contributing to resistance in MTB [16][17][18]. EPs in general have been shown to be to be associated with low level intrinsic drug resistance [19]. Earlier studies [18,20] have identified SNPs in efflux pumps in XDR MTB suggesting that increased expression of SNPs may have a role in drug resistance. Resistance to BDQ is reported to arise through over-expression of the Resistance Nodulation Cell Division Super Family (RND) efflux pump mmpLS-mmpL5, linked to mutations arising in the transcriptional repressor Rv0678 gene [21]. Presence of nonsynonymous single nucleotide polymorphisms (nsSNPs) in efflux pump genes other than mmpLS-mmpL5 have not been reported in BDQ resistant MTB isolates to date [22].
Through this research, we aimed to study RAVs in BDQ naïve population of pre-XDR/XDR MTB isolates from Pakistan, using whole genome sequencing (WGS) to simultaneously identify RAVs in the known genes contributing to BDQ resistance. We further aimed to analyse genomes of BDQ resistant/ intermediate strains for variants in drug resistance associated efflux pump genes.

Mycobacterial isolates, identification and culture
This study was conducted on MTB strains isolated from clinical specimens of BDQ naïve patients (January 2015 to June 2019) and stored as part of the Aga Khan University Hospital Mycobacteriology laboratory strain bank. Using convenience-based sampling, BDQ intermediate and resistant isolates (n = 18), as well as randomly selected BDQ susceptible MTB isolates; MDR/pre-XDR (n = 20), and pan drug susceptible MTB isolates (n = 10) were included in the study.
Frozen vials of MTB isolates that had been stored at -80 °C were thawed, and sub-cultured in MGIT (Becton Dickinson, USA), on Lowenstein Jensen agar slants, and on Middlebrook 7H10 with OADC (Oleic acid, albumin, catalase, dextrose) enrichment (BBL Microbiology Systems, Cockeysville, MD, USA) medium. Following sub-culture, isolates were re-confirmed to be MTB using MPT 64 Antigen test using BD MGIT TBc Identification Test kits (Cat. no 245159) and p-nitro benzoic acid test (PNB) [23].
Based on testing by the BMD method, isolates were categorized as susceptible, intermediate or resistant to BDQ. The following cut-offs values were used for interpreting susceptibility to BDQ: isolates with MIC ≤ 0.12 µg/ml were categorized as susceptible, whereas those with an MIC ≥ 0.5 µg/ml were considered resistant. Isolates with a BDQ MIC of 0.25 µg/ml were categorized as intermediate in accordance with studies [9,30].
BDQ susceptibility in MDR/pre-XDR/XDR MTB isolates with intermediate or resistant phenotypes had further been confirmed using 7H10 Agar Proportion and MGIT 960 method [29] (Additional File 1).
MDR strains were categorized as pre-XDR and XDR according to the updated definition of WHO [31]. MDR/ RR-TB strains that were also resistant to any fluoroquinolone were considered as pre-XDR. While XDR-MTB strains included pre-XDR strains resistant to either or both of the group A drugs; LZD or BDQ.

DNA extraction and Whole Genome Sequencing
Genomic DNA (gDNA) was extracted from second passage of MTB grown on LJ slants using CTAB method [32]. DNA (µg/ml) was quantitated using Qubit High sensitivity assay (Invitrogen). 0.5 ng/µl of input gDNA was used for library preparation, that was carried out using Nextera XT DNA library kit (15,032,355,15,052,163). Prepared libraries were then sequenced using the Illumina MiSeq platform (Illumina). Sequence was performed on pair end reads of 2 × 250 bp using the MiSeq reagent kit v2 [33]. Furthermore, 10 of the BDQ R/I pre-XDR/XDR MTB isolates that had low average genome coverage (< 10x) were re-sequenced using the Illumina MiniSeq platform (Illumina). MiniSeq high output reagent cartridge (15,073,286) was used.

Evaluation of genes known to confer BDQ resistance in MTB isolates
We screened for variants in genes known to confer resistance to BDQ including; Rv0678, atpE, pepQ, mmpLS-mmpL5, Rv1979c.

Lineage of MTB isolates
Lineage of the MTB isolates was determined by running the Fastq files generated on TB Profiler version 3 [40].

Results
Forty-eight MTB isolates from BDQ naïve patients were sequenced. Of these, based on BMD method, 18 MDR/pre-XDR/XDR isolates were having resistance or intermediate resistance to BDQ. Phenotypic and genotypic DST data for rifampicin (RIF), isoniazid (INH), fluoroquinolones and linezolid (LZD) for the study strains (n = 48) is provided in separate additional files (Additional Files 1 and 3).
The average coverage of the sequenced isolates ranged from 9.9-230.3x (Additional File 3). According to lineage characterization, 42  Previously unreported RAVs in Rv0678 gene were noted in 6/10 BDQ-R MTB isolates. These included three frameshifts at amino acid positions; 25 fs in S10, 64 fs in S4 and 109 fs in S3 and three missense mutations at amino acid positions; L40F in S5, R50Q in S6 and R107C in S9 (Table 1). Additionally, previously reported variants at amino acid position 63 [7] were observed in two BDQ intermediate isolates; S63R in S16 and S63G in S20. Variants in Rv0678 were not found in any of the BDQ susceptible MTB isolates tested. A silent mutation A18A in the atpE gene of XDR-MTB isolate S2 was also identified, as was a missense mutation; R7Q in the pepQ gene of pan drug susceptible MTB isolate SS3 (Table 1).
CFZ resistance was identified in 1/10 BDQ resistant and 2/8 BDQ intermediate strains (Additional File 1). Variant in the Rv0678 gene were found in isolate with resistance to both CFZ and BDQ (Table 1).

Discussion
The inclusion of BDQ and DLM for treating MDR TB has allowed development of effective standardized shortcourse (SSC) regimens [42,43]. However, reports of variants in genes conferring BDQ resistance, in M/XDR from BDQ naïve TB population are reported [7,9]. These variants can contribute to resistance or reduced susceptibility to BDQ leading to MDR TB patients receiving sub-optimal treatment resulting in poorer outcomes and spread of resistance in the population [44].
The strains studied belonged primarily to Lineage 3 (87.5%; 42/48) but strains from Lineages 1, 2 and 4 were also present. The predominance of Lineage 3 is reflective of the molecular epidemiology of MTB in Pakistan [45]. RAVs in Rv0678 have been reported at diverse amino acid positions including: V1A, C4R, S63R, L117R, L136P, R172W and R135G [15,46], among BDQ resistant MTB clinical strains. Six out of the eight individually occurring variants in the Rv0678 gene of isolates showing phenotypic BDQ resistance and raised MICs to BDQ were found to be novel. The phenotypic resistance to BDQ for isolates with variants at amino acid positions 109 and 63 differed between the clinical isolates from this study as compared to previously published studies [7,11]. It has been recognised that SNPs at the same amino acid position could lead to variable resistance to BDQ depending on the effect of the amino acid substituted on disrupting protein function, and whether these variants are present singly or with multiple SNPs [47]. Guidelines to delineate clinical interpretation of variants in Rv0678, Rv1979c and mmpLS-mmpL5 and its subsequent incorporation into      = previously unreported missense mutations, ε = previously reported missense mutation at amino acid position treatment strategies are required. Studies coupling treatment outcome data with phenotypic and genotypic testing results would provide stronger evidence to interpret the clinical significance of these mutations. Further, it is important to classify the intermediate category for BDQ resistance. As presented by Ismail et al. 2018 [9], patients (n = 8) with isolates having intermediate resistance (MIC = 0.25 µg/ml) were culture positive at six months of therapy. Additionally, the presence of intermediate resistance to BDQ raises concerns whether susceptibility testing at a single BDQ concentration, as in case of the MGIT 960 method, would limit understanding of the clinical challenges these phenotypes may present in context of successfully treating MDR TB. Methods such as microbroth dilution that provide a wider range of drug concentrations tested should be standardized for routine DST to capture raised MICs and intermediate resistance.
Due to the cost of fitness on MTB in vivo [46], mutations in atp E occur at a low frequency in BDQ resistant clinical MTB isolates [48]. While our study reports a synonymous mutation A18 in S2-a BDQ resistant XDR isolate, it is unlikely to be contributing to BDQ resistance as it is silent and does not translate into an amino acid substitution.
Variants in Rv1979c have been previously reported in ≤ 1% of BDQ resistant MTB isolates with no prior exposure to the drug [9]. However, association of phenotypic BDQ resistance with SNPs in Rv1979c is indeterminate [9,49]. A unique SNP (F144*) in Rv1979c gene was also seen in one of our strains (S21) that had intermediate resistance to BDQ raising the possibility that such mutation maybe contributing to resistance in these strains.
We were unable to detect RAVs in the Rv0678 gene of 55.55% (10 /18) of BDQ resistant and intermediate MTB isolates sequenced. These finding are supported by previous studies [49][50][51], suggesting that alternate mechanisms of resistance contributing to BDQ resistance need to be explored.
Our study shows presence of genomic variants in the RND efflux pump; mmpL5 (rv0676)-recognised as the target of the Rv0678 repressor protein in both BDQ susceptible and resistant isolates [6]. Published studies report an association of SNPs (V344L) in mmpL5 of a naïve BDQ resistant isolate with decreased susceptibility to antimycobacterial drugs [52]. Variants in mmpL5 could also lead to a loss of function in the efflux pump subunit; thus increasing BDQ susceptibility. This is an important consideration when for instance Rv0678 RAVs co-occur with putative loss-of-function mutations as reported [12]. However, mmpL5 mutations; I948V and T794I reported here are phylogenetic variants [41].
NsSNPs were detected in four of twenty EP genes screened in our pre-XDR/XDR MTB isolates including BDQ resistant and intermediate isolates. Literature supports the role of SNPs in EP genes in contributing towards low level resistance [53,54]. Efflux pumps; Rv0450c (mmpL4), Rv1273c, Rv0507c (mmpL2), and Rv1634c, identified to have variants in MDR/pre-XDR/ XDR MTB isolates including BDQ intermediate and resistant isolates have also previously been documented [20,55] to contribute to multi drug resistance. MmpL2 and mmpL4 gene expression is controlled by Rv0678 transcriptional repressor protein [56]. It is therefore possible to hypothesize that nsSNPs in mmpL2 and mmpL4 EP genes may impair binding of Rv0678 leading to an over-expression of these pumps, contributing towards BDQ resistance. As variants in these efflux pumps were found exclusively in drug resistant MTB isolates, including BDQ resistant and intermediate strains, they should be studied in MTB clinical isolates for their potential role as diagnostic markers of multidrug resistance. However, transcriptomic studies of MTB drug resistance isolates are required to provide a clearer understanding of the association between genetic mutations and alterations in gene expression towards development of drug resistance [57].
Variants in BDQ naïve XDR MTB isolates reported in our study, as well as previously described acquisition of BDQ resistance during therapy [58] support genotypic and phenotypic DST at baseline and during therapy to ensure appropriate management of TB cases. The absence, amongst phenotypically BDQ resistant/intermediate isolates of variants in genes known to be associated with BDQ resistance, indicates that currently both genotypic and phenotypic DST methods are required to be used in tandem for the detection of BDQ resistance in MTB. In the context of resource-limited settings including Pakistan; implementation of WGS as a diagnostic tool is likely to present technical, logistic as well as financial challenges that need to be considered at a programmatic level [59]. However, the benefits of using WGS as a companion diagnostic to phenotypic DST as part of TB management and control cannot be ignored [60,61]. Therefore, investment in human and infrastructural resources must be made to strengthen lab phenotypic and genotypic DST capacity.
One of the limitations of this study is that the bioinformatic pipeline used was not developed to screen for low frequency SNPs that could accurately indicate the presence of heteroresistance MTB population [62].

Conclusion
Previously unreported variants in Rv0678 gene of BDQ naïve MTB isolates in our strains coupled with the presence of genomic variants in EP genes in BDQ resistant and intermediate pre-XDR/XDR MTB isolates indicate the need to include genomic analysis in studying BDQ resistant strains. Our data further highlight the need to explore the role of EPs as an alternate mechanism of resistance towards including them as part of molecular screening of antimycobacterial drug resistance.