**3. DNA Gyrase: A Validated Drug Target**

Topoisomerases are enzymes that participate in the overwinding or underwinding of DNA. They are commonly split into two classes, depending on the number of DNA strands cut during a single catalysis cycle: class I topoisomerases, which include eukaryotic topoisomerase I, prokaryotic topoisomerase IIIa, or reverse gyrase among others, and class II topoisomerases, whose members include (but are not restricted to) topoisomerase IV and DNA gyrase [76]. In *M. tuberculosis*, DNA gyrase is a heterotetrameric nanomachine, consisting of two GyrB and two GyrA subunits, and is essential for DNA replication, transcription, and repair in living cells [77].

*M. tuberculosis* is an exception in the prokaryotic world because of the presence of only one type I and one type II topoisomerase, whereas most other eubacteria have two enzymes of each type. In the canonical situation, DNA gyrase and topoisomerase IV, both type IIA topoisomerases, have distinct specific activities: DNA gyrase removes positive supercoils that accumulate ahead of

replication forks [78], whereas topoisomerase IV decatenates replication intermediates [79]. Thus, in *M. tuberculosis*, DNA gyrase being the unique topoisomerase IIA, it must be able to carry out both activities in vivo [80]. Furthermore, with no other topoisomerase IIA in *M. tuberculosis*, the DNA gyrase is the unique target of antibiotics from the fluoroquinolones (FQ) family (quinolones with higher cellular penetration efficiency), which bind and stabilize the DNA–DNA gyrase complex. Stabilization of this ternary complex leads to the cytotoxic accumulation of double-strand cleaved DNA fragments within the cell, inducing bacterial death [81]. Crystallographic structures of the isolated domain and the cleavage core, either in its *apo* form or co-crystallized with several FQ, brought crucial information to elucidate the mode of binding of these drugs [82–87]. The ratio FQ/DNA–DNA gyrase in the ternary complex is two FQ molecules per complex, with one molecule binding in each cleavage site, wedging between the two ends of the cleaved DNA strands on both halves of the complex, in a drug-binding pocket whose walls are formed by DNA base stacking (Figure 2A) [82]. The emergence of strains with multidrug resistance (MDR) phenotype, or extensively-drug resistance (XDR) phenotype in which drug resistance includes this family of antibiotics, has made improving the activity of this family of drugs by using structural data to be of utmost importance. The main goals of these studies are to increase specific drug/enzyme interactions, but also to find new molecules that inhibit this validated target without cross-resistance with FQ [88].

**Figure 2.** Inhibitors of DNA gyrase. (**A**) Top view of the X-ray structure of *M. tuberculosis* gyrase cleavage core in complex with DNA and moxifloxacin. Protein is shown in cartoon representation with transparent surfaces, with GyrB in blue and GyrA in pink. DNA is shown in orange, with moxifloxacin as red spheres (pdb: 5BS8). (**B**) Top view of the model of a complex of GSK000 (green spheres) with *M. tuberculosis* gyrase cleavage core and uncleaved DNA (in orange), based on the crystal structure of GSK299423 with *Staphylococcus aureus* gyrase (pdb: 2XCR). The same representation, as in panel A, was used for the protein. (**C**) Ball-and-stick representation of the ligands colored by element. (**D**) Generic chemical structure of the novel bacterial topoisomerase inhibitor (NBTI) series of compounds. Analog 1 is a classical NBTI, whereas analogs 2 and 3 combine the three favorable features for *M. tuberculosis* DNA gyrase inhibitor (MGI).

A significant amount of effort has thus been invested in discovering new *M. tuberculosis* gyrase inhibitors, mostly centered on the ATP- and DNA-binding sites, and particularly using in silico methods based on the emerging gyrase structural information, as extensively reviewed by Nagaraja and coworkers [89]. Although the ATP binding sites have been less successfully exploited as antibacterial targets, with the exception of the natural products, such as coumarins [90,91], it is worth noting that a candidate drug emerged from these studies. SPR720 (Table 3) is an orally bioavailable prodrug of SPR719, an aminobenzimidazole inhibitor of both gyrase and topoisomerase IV. Indeed, this novel class of antimicrobials targets the ATPase subunits of gyrase and topoisomerase IV, and it was optimized using SSBD and structure-activity relationship (SAR) studies of potency against both Gram-positive and some Gram-negative bacterial species [92,93]. Crystal structures of novobiocin (Table 3) bound to *Escherichia coli* GyrB subunit served as a starting point for ligand optimization [94], while further optimization of the metabolic profile led to the identification of SPR720 (formerly VXc-486) [95], whose suitability to be a drug candidate for the treatment of tuberculosis and non-tuberculosis mycobacterial infections was thoroughly evaluated [96,97]. The compound completed Phase I clinical trials in 2019 (study ID NCT03796910, sponsored by Spero Therapeutics, LLC), aimed at evaluating the safety, tolerability, and pharmacokinetics (PK) profile in healthy volunteers. Moreover, an in vivo combination of SPR720 with rifampicin and pyrazinamide (two first-line drugs used in TB treatment) showed comparable efficacy to the combination of three drugs, including moxifloxacin, rifampicin, and pyrazinamide [98].

Work on the DNA-binding site has also been productive, as GlaxoSmithKline (GSK) created a new class of type IIA topoisomerase inhibitors, called novel bacterial topoisomerase inhibitors (NBTIs) (Table 3) [99]. The crystallographic structures of the DNA-bound *Staphylococcus aureus* DNA gyrase with either NBTI or FQ, all obtained at high resolution, revealed that the binding sites for each drug are different [100]. Indeed, the NBTI 'bridges' the DNA and a transient non-catalytic pocket on the two-fold axis at the GyrA dimer interface, remaining close to the active sites and FQs binding sites (Figure 2B). As NBTIs display relatively poor activity against the *M. tuberculosis* DNA gyrase, the antitubercular profiles of 3000 compounds, representative of the chemical diversity of this family, were evaluated by high-throughput phenotypic screenings in vitro and in vivo [101]. MIC determination on *M. tuberculosis* H37Rv showed a high hit rate (68% of compounds with MIC values lower than 10 μM), with the most potent derivatives matching or even improving the MIC values for currently used TB drugs, including last-generation FQs. Overall, 29% of the compounds had MICs of < 1 μM, and 18% had MICs of < 0.1 μM. By using structural data provided by the work of Bax et al., the general structures of these TB-active compounds were divided into three different regions, each of them interacting with one of the three topologically important target-gyrase interacting regions [100]. While a left-hand side (LHS) is responsible for key contacts with the gyrase DNA substrate, a right-hand side (RHS) is embedded into the enzyme, potentially contributing to the protein target selectivity (Figure 2D). Last, a central linker unit (CU) establishes key interactions with the gyrase and offers the opportunity to modulate key physicochemical properties (Figure 2D). These SAR observations helped to schematize synthesis and to rationalize how to balance antimycobacterial potency with oral exposure, safety, and synthetic complexity, leading to the identification of the 7-substituted-1,5-naphthyridin-2-one core as a privileged LHS binder, the N-ethyl-4-aminopiperidines as a linker, and monocyclic aromatic rings with different substitution patterns as the best RHS binding option, in contrast to NBTIs bearing bicyclic rings as the RHS (Figure 2D, analogs 2 and 3) [101]. This work led to the identification of a subclass of naphthytidone/aminopiperidine-containing compounds that displayed activity against *M. tuberculosis*, both in vitro and in the mouse model, known as '*M. tuberculosis* DNA gyrase inhibitors' (MGIs) due to structural and activity differences with respect to NBTIs [101]. More recently, the mechanism of action of two such compounds, i.e., GSK000 and GSK325 (Figure 2C; Table 3), was assessed on the *M. tuberculosis* gyrase, showing that MGIs greatly enhanced DNA cleavage mediated by the bacterial enzyme, but they induced only single-stranded DNA breaks [102]. Their mechanism of action involves stabilizing covalent gyrase-cleaved DNA complexes and appears to suppress the ability of the enzyme to induce double-stranded breaks. Furthermore, these compounds maintained activity against mutant versions of DNA gyrase, bearing the three most commonly observed FQs resistance mutations, but displayed no activity against human topoisomerase IIα [102], suggesting good potential

as candidate drugs, especially in the presence of FQ resistance. Modeling studies were carried out by the authors using the crystal structure of the *M. tuberculosis* gyrase cleavage core and the NBTI crystal structure complex of *S. aureus* gyrase [100] (Figure 2B). Most importantly, these studies allowed to confirm that FQs and MGIs do not share the same binding site and bind in a mutually exclusive manner (Figure 2), considerably reducing the risk of developing cross-resistance phenotypes. μ


**Table 3.** *M. tuberculosis* DNA gyrase inhibitors.

μ

μ

μ

ND: not determined.

#### **4. General Remarks**

Several challenges need to be overcome for therapeutic molecules to be active against *M. tuberculosis.* Due to its metabolic adaptability and its known capabilities to occupy different niches inside the human host, from free aerobic bacteria to granulomas, an ideal molecule should be able to act against bacteria in very different, environmentally adapted states. The unique structure and composition of the *M. tuberculosis* cell wall are also well known to act as a barrier for potentially active compounds. To be active against latent and active infections, an ideal drug candidate should thus be able to reach bacteria in all the tubercular lesions and niches inside the human host [103], permeate the mycobacterial cell wall or be actively internalized, and deliver the desired bioactivity in bacteria under different metabolic states. To overcome these limitations and reduce the risk of further resistance development, TB drugs are delivered in combinations, starting from the standard DOTS (directly observed therapy short-course). Drug combinations, however, have to take in due account the bioavailability and pharmacokinetic properties of the single components and their associations, as extensively treated in this same journal issue [104].

In the course of the last ten years, it has become increasingly clear that some of the criteria commonly used to classify a given *M. tuberculosis* biological process, in many cases, an enzymatic step in a pathway, as a 'good' target for drug development, needed to be revised. This is particularly true for in vitro genetic essentiality. Even targets confirmed to be essential by the construction of specific conditional mutants do not necessarily show a good 'druggability', even when highly potent hits become available. The case of PknB is an emblematic example: despite the gene was early shown to be essential for *M. tuberculosis* growth, first by transposon mutagenesis [43], then through attempts to generate a knock-out mutant [44], yet several research groups, including ours, had to face the problem of the lack of correlation between hit potency in vitro and efficacy in vivo [50–52]. Similar issues were also reported for other *M. tuberculosis* kinase inhibitors, in addition to the cross-reactivity sometimes reported towards eukaryotic kinases, and the cytotoxicity shown by some compounds on human cell lines [105]. In addition, although the availability of crystal structures of both PknB and PknG in complex with hits issued from medium-throughput screens (mitoxantrone and AX20017, respectively) has looked as a promising 'proof-of-concept' [48,58], further structure-based work on these kinases has, so far, failed to produce suitable drug candidates. Many reasons might be evoked to explain the lack of *M. tuberculosis* kinase inhibitors that display significant antibacterial activity, ranging from the poor capacity to penetrate the mycobacterial cell and reach their target to the kinase redundancy in *M. tuberculosis* and their, relatively poor, substrate specificity [46], which prompted groups to seek for multiple kinase inhibition [105]. Yet, the main causes should perhaps be looked for in our, still limited, knowledge of mycobacterial molecular physiology and the regulatory networks in *M. tuberculosis*, in which Ser/Thr kinases are actors of a complex interplay [46,106]. Indeed, the activation mechanisms of both PknB and PknG are still a matter of speculation, and changes to *pknB* expression, either as depletion or overexpression, have been reported to alter the bacterial growth significantly [34]. Given the known *M. tuberculosis* metabolic plasticity, whose complexity has only recently started to be elucidated, thanks to the development of genetic and 'omics' tools [107], and the adaptability of the pathogen in the course of infection, it is now largely accepted that genetic essentiality of a putative target in laboratory conditions is not necessarily an indication of chemical vulnerability [15], and even more in clinical conditions. For these reasons, phenotypic screens have increasingly been employed in TB drug development to the detriment of target and structure-based methods, including for lead optimization [15], and the vast majority of candidate TB drugs that have been able to enter clinical trials in the last years were issued from this kind of screens [17,108].

Nevertheless, structure-based and computer-aided drug design maintain a clear potential for the future development of new anti-tubercular drug candidates. For instance, a recent, very promising success of structure-based and fragment-based approaches in TB drug discovery is the development of the so-called ethionamide boosters directed against the EthR repressor, one of which has been shown to be active both in vitro and in vivo [109]. In addition, the recent developments in X-ray crystallography [110], cryo-EM [4], and integrative structural biology methods [111] will all contribute to increasing the number of tools available to tackle the challenges that lay ahead. These opportunities are well exemplified by the case of mycobacterial DNA gyrase that, considering our capacity to obtain high-resolution structures of an almost full-length form of the *M. tuberculosis* enzyme [112], and the recent, high-resolution cryo-EM structure of *E. coli* gyrase in complex with NBTI [113], let us believe that SBDD will deliver a key contribution to developing new compounds against this 'old' but validated target. High-resolution snapshots of the complete mycobacterial gyrase machinery, especially if in complex with representative members of each known family of inhibitors, might revolutionize our knowledge of this key target and substantially increase chances of improving our therapeutic bullets. For instance, combining crystallographic and cryo-EM data could allow to perform structure-guided drug design to target these flexible complexes and identify new conformations of mycobacterial gyrase that could not, otherwise, be obtained by conventional structural methods. More generally, the integration of biophysical and structural biology data, with the notable contribution of high-resolution EM, will allow the drug discovery pipelines to work on a higher complexity level that was previously not achievable, now looking at targets in their larger biological context (e.g., complexes or cellular compartments). It is, therefore, foreseeable that, despite the technical challenges, target-based and structure-based approaches will have increasing relevance in future drug discovery and will give significant contributions in the search for new tuberculosis drugs.

**Author Contributions:** All authors contributed to conceptualization, the preparation of the draft, and the review of this work. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was partially supported by institutional grants from the Institut Pasteur and the CNRS.

**Acknowledgments:** We are grateful to all our past and present colleagues from the Structural Microbiology Unit, as well as collaborators from the former MM4TB Consortium, for so many fruitful exchanges and insightful discussions.

**Conflicts of Interest:** The authors declare no conflict of interest.
