*2.2. Mode of Action*

Advances in next generation sequencing technologies and the continued reduction in cost has placed whole genome sequencing (WGS) firmly into the drug discovery pipeline from mechanism of action identification through to monitoring drug resistance post approval [29]. Illumina sequencing (Illumina Inc., San Diego, CA, USA) generates short sequence reads (typically 150 bp long), and while the majority of sequencing data generated today is from Illumina platforms, alternative methodologies are also useful to antimicrobial drug discovery. Long read sequencing (~kilobases), available through platforms such as PacBio (Pacific Biosciences, Menlo Park, CA, USA) and Oxford Nanopore (Oxford Nanopore Technologies Ltd, Oxford, UK), are especially suited for sequencing repetitive regions and structural variations. WGS has been exploited to identify mutations that occur in spontaneous drug resistant colonies growing on solid agar containing a novel antimicrobial compound of interest, typically between 5-fold and 10-fold the minimum inhibitory concentration. Using drug exposure as the selective pressure, this methodology may reveal mutations that affect drug action, in the drug target itself or in gene functions that influence drug activation or efflux. Andries et al. in the discovery of bedaquiline, employed WGS of *M. smegmatis* and *M.tb* bedaquiline-resistant mutants cultured in vitro to demonstrate that this diarylquinoline targets the product of *atpE*; a subunit of the mycobacterial ATP synthase anchored in the mycobacterial membrane [30]. Kundu et al. further characterized

this mode of action, showing that bedaquiline binds to the epsilon subunit [31]. Mutations can, however, be identified in other non-target genes that confer resistance, expanding understanding of drug action but complicating the interpretation of results. For example, in a murine model of TB infection, mutations in the putative Xaa-Pro aminopeptidase *pepQ* were identified that conferred low level resistance to both bedaquiline and clofazimine, neither of which target PepQ directly [32]. While the function of PepQ remains to be determined in *M.tb*, Almeida et al. postulated that this mutation prevented protein degradation of the efflux pump MmpL5, enhancing drug efflux and inducing resistance to bedaquiline and clofazimine [32].

WGS has also been used to reveal the mechanism of action of repurposed licensed pharmaceuticals in *M.tb*. Rybniker et al. identified lansoprazole from the Prestwick library of 1280 FDA approved drugs as protective to lung fibroblasts in an *M.tb* intracellular model of infection [33]. Three lansoprazole resistant mutants were sequenced; each mutant showed the same SNP, a substitution of leucine for proline, in the β-subunit of the cytochrome bc1 complex gene *qcrB*, a key element of the mycobacterial respiratory chain. To demonstrate that exposure to this commonly prescribed proton pump inhibitor did not select for lansoprazole-resistant *M.tb*, Rybniker et al. sequenced 13,559 *M.tb* Complex clinical isolates revealing only one *M. bovis* isolate with a mutation in *qcrB* [34]. Of course, this methodology to define the target of drugs with unknown mechanisms of action is only fruitful if drug-resistant colonies can be raised, which is not always the case. This was highlighted in the discovery of the novel antimicrobial drug, teixobactin from the bacterium *Eleftheria terrae*. Here, a novel iChip culture method was used to capture previously unculturable bacteria in their natural soil environment, which led to the identification of the cell wall-targeting agent teixobactin [35]. Neither *Staphylococcus aureus* nor *M.tb* drug resistant colonies could be generated even at sub-inhibitory concentrations. The authors hypothesized that this indicated general cell toxicity of the compound; however, no mammalian cell toxicity was observed. Subsequent biochemical approaches indicated that this compound exerted its antimicrobial effects through interactions with peptidoglycan pre-cursors—mostly lipid II. Therefore, the inability to generate spontaneous resistance in vitro is likely due to teixobactin inhibiting multiple non-protein targets. Like many other antimicrobial drugs, teixobactin was isolated from an environmental microorganism. It is therefore possible that bacteria in close proximity to *E. terrae* in the soil may be resistant to teixobactin, offering a scenario where teixobactin resistance mechanisms could be discovered [36]. While the inability to detect drug resistance in the laboratory does not mirror the complexity of TB clinical disease alongside host pharmacokinetic and pharmacodynamic factors, the concept of resistance-proof antimicrobial drugs is very attractive and could transform the destructive antimicrobial drug to evolution of drug resistance cycle.

WGS is also shaping our understanding of drug action and drug resistance through the large-scale sequencing of clinical isolates. Consortia mapping drug-resistance conferring mutations have revealed novel mechanisms of resistance and potential novel modes of action of existing anti-*M.tb* drugs in patients [37]. The CRyPTIC Consortium and the 100,000 Genomes Project obtained over 10,000 *M.tb* clinical isolate genomes and associated phenotypes to predict phenotypic drug susceptibility to front line agents from genome sequence. The authors successfully demonstrated that susceptibility could be correlated to the presence or absence of specific antimicrobial drug resistance mutations [37]. This approach also exposed mutations with as yet unexplained roles in drug action. For example, *M.tb* isolates have been identified with mutations in the mycothiol biosynthetic genes *mshA* and *mshC* that confer high-level resistance to ethionamide and low-level resistance to isoniazid [38]. Mycothiol is a key detoxifying and reducing molecule with similar functions to glutathione, which is absent in mycobacteria [39]. Knockout mutants of *mshA* grew normally in vitro and in immunodeficient mice but were growth defective in immunocompetent mice [40], and mycothiol-deficient mutants were less susceptible to ethionamide and isoniazid [41]. Ethionamide and isoniazid are both prodrugs that target the NADH-dependent enoyl-ACP reductase InhA; however, isoniazid is activated by KatG and ethionamide by EthA [42,43]. In order to delineate the roles mycothiol biosynthetic genes play in isoniazid and ethionamide susceptibility, Xu et al. generated null mutants of each mycothiol

biosynthetic gene in *M. smegmatis* [44]. *M. smegmatis* is useful for studying the action of these two drugs, with their target InhA characterized in this bacterium to have >95% sequence identity to the *M.tb* InhA [45]. While the exact role of mycothiol in isoniazid and ethionamide resistance remains to be determined, the authors hypothesized that mycothiol is involved in the activation of isoniazid and ethionamide. *M.tb* KatG, required for isoniazid activation, likely compensates for loss of mycothiol in some settings, resulting in isoniazid susceptibility and ethionamide drug resistance in mycothiol-deficient isolates [40]. Understanding the action of current *M.tb* drugs has direct implications for the development of new inhibitors. For example, bedaquiline and clofazimine resistance can be mediated through SNPs in *Rv0678*, a transcriptional repressor of *mmpL5* and *mmpS5*, which encode an efflux pump of the resistance-nodulation-division (RND) family [46,47]. SNPs in *Rv0678* were detected in clinical isolates of *M.tb* with no prior exposure to bedaquiline or clofazimine [48]; meaning that resistance to bedaquiline was already circulating in *M.tb* isolates before the new drug was introduced to the TB clinic. While the authors could not determine the driver of these mutations, they ruled out rifampicin as a selective pressure. Studies such as these are revealing drug detoxification and efflux pathways that operate across drug class to reduce efficacy of both existing and new antimicrobial drugs.

To expand the methods available for understanding drug mechanism of action, Melief et al. constructed a library of *M.tb* strains overexpressing single genes that could be screened in a high-throughput format [49]. It was hypothesized that library mutants overexpressing the target of an antimicrobial drug should be more resistant than wild type *M.tb* to that particular drug. The library was constructed cloning each gene downstream of a tetracycline-inducible promoter. The 1733 constructs covered 40% of protein coding genes in the *M.tb* genome and contained the majority of annotated essential genes, as well as genes involved in cell wall and fatty acid biosynthesis, virulence factors, regulatory proteins and efflux. The functionality of the system was confirmed by screening the library for resistance to D-cycloserine, which identified the Alr-over-expressing mutant as the only recombinant strain that grew in the presence of the drug. Over-expression of Alr, a target of the drug, resulted in a 7-fold increase in the minimum inhibitory concentration of D-cycloserine. This library represents a new tool to discern targets and pathways that influence drug efficacy of lead compounds with unknown mechanisms of action.

## **3. Transcriptomics**
