Lessons from Seven Decades of Antituberculosis Drug Discovery

Abstract

Despite massive global efforts tuberculosis rates continue to climb and drug-resistance rates are rising to alarming levels. Discovering new agents for treating this bacterial pathogen poses unique challenges, but these challenges have been faced throughout the entire modern history of research into anti-infectives. This review looks back at every decade since the 1940s and summarizes the most important drugs developed during each decade highlighting the lessons learned during these successful medicinal chemistry programs. Looking forward we must accelerate the integration of these past lessons with the impressive advances that have been made in the basic understanding of the biology of this disease.

1. Introduction

Tuberculosis (TB) has historically been a pernicious killer of humans, evidence for TB infections dates back to the Pleistocene more than 10,000 years ago [1]. It is not surprising then that TB has been a high priority target for antimicrobial chemotherapy since the very dawn of antibiotic development. Despite that priority and decades worth of effort, TB continues to exact a toll of nearly 2 million lives every year and still requires 6–8 months of chemotherapy to effect a cure with a combination of our best four drugs. Massive global programs have been mounted to identify and treat TB patients, and official targets were adopted in the Millennium Development Goals, but progress in reducing TB prevalence and mortality has been sluggish and the number of new cases continues to rise every year [2]. There are also clear threats to the limited progress that has been made in controlling the disease, including a trend towards increasing numbers of multidrug resistant (MDR) TB cases (defined as resistance to isoniazid and rifampicin, the two most important front-line agents) [3, 4]. There are currently around 500,000 cases of MDR TB globally but less than 10% of these are treated with expensive combinations of so-called second-line drugs [5]. With few exceptions these agents tend to be less well tolerated by patients, are less efficacious, and are much more expensive. These agents however, represent a line of last defense to contain the spread of MDR disease globally. By 2015 the World Health Organization’s “Global Plan to Stop TB” aims to ensure universal access to these second-line agents for the treatment of MDR TB [6]. However, even with the relatively few patients that currently have access to second-line therapy, strains are already rapidly appearing that show resistance to these drugs of last resort [7]. These strains, dubbed extensively drug resistant (XDR), have appeared rapidly in many countries and this emergence can be expected to accelerate as wider access to second-line agents is implemented [8, 9].

So despite the importance of the disease, and the intensity of the efforts that have gone into developing new therapeutics for its treatment, we are on the verge of a return to the prechemotherapy era as the emergence of drug resistance outpaces our efforts to develop new agents. This review will take the form of a series of case studies selected from each of the seven decades of medicinal chemistry that has given rise to the antitubercular agents in current use or late-stage development. Each decade will focus around a series that appeared most promising at the time and try to draw out the key lessons that emerged from the development program that led to a drug. Across these seven decades we will attempt to discern the guiding principles that led the programs and follow their evolution as we set the scene for where TB drug discovery is today.

The fundamental biology of Mycobacterium tuberculosis (Mtb) has advanced extraordinarily in the same time period, in some ways to such an extent that TB has now become a model organism for the development and application of new tools in genomics and systems biology. Unfortunately, as will be apparent by the end of this review, such advances have had little or no impact on the drug discovery process. The most important advances in drug development in the recent decade have derived from the same basic tools in use in the 1940s, and applications of modern methods of structure-based drug design with “genomically-informed” targets have been spectacularly unsuccessful. The failure to capitalize on our success in understanding the basic biology of the pathogen by coupling this to effective medicinal chemistry approaches is a failure of both the biologists and the chemists working on this important disease. Some suggestions for potential solutions to this will form the final part of this review but new and innovative solutions are urgently needed as we are clearly losing the battle.

2. The 1940S: STREPTOMYCIN

The discovery of penicillin by Alexander Fleming in 1928 had energized a search for other antibiotics from microorganisms. Amongst those searching was Albert Schatz, a graduate student at Rutgers University in the laboratory of Selman Waksman (who would later win the Novel Prize for the discovery of streptomycin 1, Fig. (1)). In 1943 Schatz and Waksman first reported the isolation of streptomycin and its activity against Mtb [10]. Immediately after the manuscript describing the in vitro activity appeared, clinicians at the Mayo Clinic requested a sample and began an animal efficacy study in infected Guinea pigs and upon observing a positive result they injected the first human TB patient with streptomycin by November the same year [11]. Within three years streptomycin was being produced by Merck & Co. and what is widely credited as the first randomized controlled clinical trial ever performed was underway by the Medical Research Council in Britain [12]. The trials were well underway before the structure of streptomycin, the first aminoglycoside to enter clinical usage, was even determined [13]. Although these initial trials were to end largely in disappointment because of the rapid emergence of resistance, subsequent trials using combinations of streptomycin and paminosalicylate were much more successful and combination chemotherapy for TB was born. The subsequent isolation of kanamycin (2) in 1957 [14] and production of its inactivation resistant semisynthetic L-hydroxyaminobuteroyl amide amikacin (3) in 1972 [15] led to use of these more active aminoglycosides over streptomycin Fig. (1). Although kanamycin was not selected for use specifically in treatment of TB, it is now widely used and there is some evidence for superior efficacy of kanamycin over other aminoglycosides [16].

Fig. (1).

Streptomycin (1) discovered in 1943 and the aminoglyco-sides discovered subsequently such as kanamycin (2) in 1957 and amikacin (3), a semisynthetic derivative of kanamycin made in 1972 to circumvent the emerging problem of resistance arising from inactivation through modification of the amine, are still used in TB therapy particularly for treating MDR infections.

The urgency with which streptomycin went from discovery to the clinic is unmatched in any subsequent development program and while regulatory standards have certainly changed since the 1940s there is a clear need for this kind of urgency in contemporary efforts. The evidence for efficacy in humans that came from even the first usage (a single patient named Patricia T) motivated a groundswell of activity to manufacture and test streptomycin and resulted ultimately in the first useful therapeutic regimens. It is also worth noting that streptomycin was not evaluated in mouse models, which fail to recapitulate many important aspects of human TB pathology but instead went through Guinea pigs that do [17].

Streptomycin also serves as a reminder that the chemical space occupied by most “drug-like” molecules poorly overlaps the space occupied by most TB drugs. Perhaps it is unsurprising then that existing libraries of compounds, optimized for conventional drug-like properties, are proving to be disappointing sources of new leads for discovery programs. Natural products have been a potent source of antimicrobial agents, in fact more than two-thirds of clinically used antibiotics are natural products and natural products remain a promising avenue for the development of future drugs [18]. Despite this there are few contemporary efforts to identify potential new natural products active against TB and even fewer to explore new biological niches that might harbor unique agents.

3. The 1950S: ISONIAZID

Isoniazid 5, Fig. (2) was actually first synthesized in 1912, a direct follow-on of Emil Fischer’s fortuitous discovery of phenylhydrazine and the powerful reducing properties of hydrazine in the 1870s (which contributed to his Nobel Prize in 1902) [19, 20]. Forty years later it was rediscovered as an antituberculosis drug, its earlier discovery a stunning disappointment to the three pharmaceutical companies who were simultaneously trying to file for patent protection for the use of isoniazid in the treatment of TB [21]. Hydrazides were being widely explored at the time because of the availability of large stocks of hydrazine confiscated from Germany’s V2 rocket fuel stocks following World War II [22]. These acyl hydrazides, however, were originally just intermediates in the synthesis of analogs of thiosemicarbazones (such as thiacetazone (4) Fig. (2) which has also been used to treat TB) whose antitubercular activities had been described by Domagk more than a decade earlier [23]. Two groups, one at Hoffmann La Roche Laboratories led by Herbert Hyman Fox and the other led by Harry L. Yale at the Squibb Institute, independently discovered that the hydrazide intermediate was more active than the thiosemicarbazone product [24–26]. Fox’s team was building on the observation that nicotinamide was bacteriostatic to Mtb and was trying to make a thiosemicarbazone derivative of nicotinamide. Importantly hundreds of these newly synthesized analogs were screened directly in infected mice, not in in vitro assays.

Fig. (2).

Isoniazid: progenitors and progeny. Thiacetazone (4), an example of the thiosemicarbazides synthesized originally by Domagk in the 1940s and 50s gave rise to isoniazid (5), an intermediate in the synthesis that was discovered to be more active than the products. Many derivatives of isoniazid were tested in the clinic, including iproniazid (6) that ultimately proved to be less effective than isoniazid despite radical changes in the moods of many sanatoria patients. Ultimately iproniazid was found to inhibit monoamine oxidase (MAO) and gave rise to a whole family of MAO inhibitors such as isocarboxazid (7), tranylcypromine (8), and phenelzine (9) that were used for decades as antidepressants.

Derivatives of isoniazid continued to be tested for improvements, including in clinical trials that led to the fortuitous discovery of “hydrazide therapy” for depression. This therapy arose through the incidental observation made in clinical trials of an isopropyl derivative of isoniazid (iproniazid, (6)) for the treatment of TB that patients receiving this drug became deliriously happy [27]. Iproniazid was later shown to be an inhibitor of monoamine oxidase and gave birth to the family of monoamine oxidase inhibitors Fig. (2) used for decades to treat depression (although less commonly used today because of the well known side effects of this family of molecules).

The mechanism of action of isoniazid is very complex, the molecule is a pro-drug that is oxidatively activated by an endogenous mycobacterial catalase-peroxidase before forming a covalent adduct with nicotinamide (and possibly other cellular molecules) [28, 29]. Although the precise molecular events that cause cell death are still less than perfectly clear, the requirement for activation of the molecule is very clear. Isoniazid is not unique amongst TB drugs in requiring activation by a cellular enzyme prior to exerting its lethal activity, this property is shared by the thioamides (widely used second-line agents including ethionamide, prothionamide and thiacetazone (4)) that require activation by a monooxygenase [30] and another front-line agent, pyrazinamide (which requires enzymatic hydrolysis of an aryl amide to liberate pyrazinoic acid [31]). Of course most modern chemical libraries are largely constructed to avoid including molecules with latently reactive functionality and would be unlikely to contain compounds such as acyl hydrazides or thiosemicarbazones.

The discovery program that led to this current front-line agent offers a few key lessons; (1) always test your intermediates, (2) always look for unexpected side-effects with an eye to alternative uses, and (3) screening for molecules with reactive functionalities specific for bacterial metabolism has been productive. The last point is perhaps the most difficult to conceptualize in a modern development program but it is worth emphasizing that mycobacterial metabolism (and hence the range of potential prodrugs that could be activated) is still not well understood and remains a potential weak spot for the creative chemist.

4. The 1960S: RIFAMPICIN

The rifamycins were first reported from the Dow Lepetit Laboratories (Milan, Italy) in 1957 as a collection of active metabolites produced by a soil microbe (first designatedStreptomyces mediterranei but later renamed as Amycolatopsis rifamycinica) [32]. These “ansamycins” (from the latin “ansa” for “handle” referring to their aliphatic chain connecting an aromatic nucleus) were an extremely complex mixture of five natural products (A-E according to their mobility on paper chromatography). The only one of these that was stable enough to isolate was “B” which was both a minor product and showed the lowest antibacterial activity. Nonetheless the structure of this compound 10, Fig. (3) proved valuable in understanding the core rifamycin structure. It was observed that rifamycin B undergoes a spontaneous “activation” process in aqueous solution that involved conversion to rifamycin O (11) [33] and deglycoylation to form rifamycin S (12) [34]. Mild reduction of the quinone of rifamycin S yielded rifamycin SV (13), the first rifamycin to be used clinically albeit limited to parenteral and topical administration. Even with this inconvenient route of administration, the impact of rifamycin SV on a range of bacterial infections was clearly dramatic, more than 150 papers describing the potency of this material appeared between 1961 and 1965 on infections ranging from Staphylococcus to TB and leprosy [35].

Fig. (3).

The long road from the isolation of rifamycins to an orally available agent. Rifamycin B (10) was the only stable molecule that could be isolated from the fermentation broth of the microbe that produced a potent antimycobacterial agent. The poor activity of rifamycin B could be improved by “activation” to rifamycin O (11) followed by deglycoylation to form rifamycin S (12) which could be reduced under mild conditions to form rifamycin SV (13) which was used parentally to treat TB. The key observation (highlighted) that allowed the production of rifampicin (15) was a Mannich reaction allowing the production of 3-formylrifamycin SV (14).

Spurred on by the mounting evidence for great clinical potential of this agent the Lepetit Laboratories team, led by Piero Sensi in Milan, teamed up with Ciba-Geigy in Basel and over the next 18 years produced several hundred semisynthetic analogs of rifamycins with the goal of obtaining oral bioavailability [36]. The obvious sites for derivatization (either of the secondary alcohols on the ansa chain, or the hydroxyl at C-8 of the aromatic nucleus) all proved to be essential for activity. The key to achieving their goal was the realization that changes in the C-3 position of the hydroquinone nucleus were well tolerated in terms of activity, and that this position was a site for selective aminoalkylation using the Mannich reaction of formaldehyde with a wide range of secondary amines [37]. The resulting N-substituted aminomethylrifamycins were the first compounds in this series with oral bioavailability but proved susceptible to oxidation yielding 3-formylrifamycin SV (14). This pivotal compound was a highly valuable intermediate that allowed access to an array of imines, hydrazones, oximes, and hydrazide-hydrazones. This ultimately allowed the synthesis of the N-amino-N’-methylpiperazine hydrazone of rifamycin SV which was found to be both the most orally bioavailable and the least toxic derivative and rifampicin (15) was born [38].

The history of the rifamycins illustrates yet again the promise and pitfalls of natural products. It is worth noting that no other drug has so radically transformed chemotherapy of TB as rifampicin, the entire modern “short-course” regimens are a direct result of the introduction of this agent [39–41]. Prior to this therapy was 12–18 months so the introduction of rifampicin effectively halved the duration of chemotherapy. The effort that went into transforming this from a natural product to a semisynthetic, orally available drug was massive however. More medicinal chemistry effort went into this single agent than is being applied in the entire field of TB drug discovery today. It is hard to imagine an effort of comparable magnitude being mounted today around such a complex scaffold, yet perhaps that is precisely what should be considered.

5. The 1970S: A Lost Decade

The impact of the success of the rifampicin-containing regimens in shortening the duration of therapy gave rise to a naïve optimism that TB was finally on the way out. Many sanatoria, where TB patients would have formerly been sent to recover or perish, were being closed down. Rates of TB in the developed world plummeted with the new four-drug combination regimens and research into tuberculosis, including new drugs, all but stopped. The situation was not unique to TB but extended to most antibiotics and the resulting “innovation gap” has only been partially reversed with the emergence of many diverse drug-resistant pathogens [18]. The lesson from this decade is a repeating theme that complacency leads to new drug resistance and lost lives.

6. The 1980S: OXAZOLIDINONES

N-aryl substituted 1,3-oxazolidin-2-ones have been known since the 1950s as antidepressant monoamine oxidase inhibitors and the antibacterial activity of these compounds was uncovered initially in 1978 in a search for novel antibacterials for use against drug-resistant Staphylococci by researchers at E. I. Dupont de Nemours and Co [42]. Although never specified the discovery of this class was likely by whole-cell screening, and shortly after it was announced in 1987 other pharmaceutical companies began medicinal chemistry programs around this nucleus [43]. The subsequent chemical optimization that took place was directed at both whole cell activity and at protein synthesis explicitly using either whole cell readouts or in vitro protein synthesis assays. In the initial hit series reported by the Dupont group e.g. 15 and 16, Fig. (4) they made the following observations regarding the Structure Activity Relationships (SAR): first that an N-aryl group was required and that a para electron withdrawing group improved activity, second that the Sconfiguration at C-5 was required for antibacterial activity, and that an acetamidomethyl substituent was optimal at this same position. Early attempts at optimization aimed to reduce the conformational flexibility of the system by constraining this in tricyclic or bicyclic systems such as 17 or 18 [44].

Fig. (4).

Antitubercular oxazolidinones. Early hits in the series such as 15, 16, 17 and 18 showed promising antibacterial activity but toxicity issues showed up almost immediately. The key insight that saved the series was the nearly equivalent PK and activity of compounds 16 and 17 demonstrating that STR and SAR were distinct Linezolid (19) ultimately resulted as an important broad — spectrum agent and newer agents such as PNU—100480 (20) are currently being examined again for antitubercular activity.

Notable improvements in activity were achieved by numerous groups working in this class of molecules (nicely summarized in [45]) but the key insight that allowed Linezolid to become a marketed drug happened at Pfizer’s laboratories in Kalamazoo, Michigan. As toxicity problems with this series were beginning to emerge this group did an experiment comparing the toxicological profile of arylmethylketone 15 with the indanone 16. Both compounds had similar activity and pharmacokinetic profiles in the rat, yet compound 15 was highly toxic, causing progressive weight loss, bone marrow toxicity and death while compound 16 had no apparent side effects [43]. This suggested strongly that the SAR and STR (Structure Toxicity Relationship) of these series were distinct and that developing potent, nontoxic compounds was, at a minimum, possible. This insight ultimately led to the development and licensure of Linezolid as a broad-spectrum antibiotic.

TB has, of course, been an incidental bystander in the story of the development of oxazolidinones. Fortunately the ultimate broad-spectrum molecule (Linezolid, 19) retains considerable activity against TB and mounting evidence of efficacy of this drug in highly drug-resistant patients, combined with two on-going prospective clinical trials (Clinical-Trials.gov Identifiers: NCT00727844 and NCT00664313) have dramatically increased enthusiasm for this class of molecules and the therapeutic potential of oxazolidinones for TB treatment. Currently several of the oxazolidinones from these historical development programs are being reexamined for anti-TB activity resulting in candidates such as PNU-100480 (20) that are re-entering clinical development specifically for TB [46–48].

The clear lesson from the oxazolidinone development program was to address toxicity issues early in a lead optimization program. While this worked beautifully, and resulted in a truly novel antibacterial class that is incredibly useful and important for some bacterial diseases, this needs to be seriously readdressed for TB. The toxicological profile of Linezolid has not been optimized for the length of time required for TB therapy and serious side effects are common when used for extended durations of time [49]. While there are several other candidate oxazolidinones currently in development for TB there is limited information on the toxicology of these to date and they were not specifically developed with an extended dosing problem in mind [48]. With the solution of the co—crystal structure of Linezolid bound to the 50S subunit of the ribosome last year there is finally hope of applying the tools of structure—based design to improve our understanding of both the SAR and the STR of the oxazolidinones and develop a more selective agent with reduced toxicity specifically for the treatment of TB [50, 51].

7. THE 1990S: NITROIMIDAZOLES

Another series of molecules that were natural product inspired are the nitroimidazoles, two examples of which are currently in Phase II clinical trials, PA—824 (23) and OPC67683 24, Fig. (5). These compounds were derived originally from metronidazole (22), which was itself derived from the natural product azomycin (21). Metronidazole has given rise to a wide variety of analogs that have found various utilities including; antiinfectives (primarily directed at anaerobes), antiprotozoans and radiosensitization of anoxic tumours [52]. Aerobic antitubercular activities have mostly been developed in the 4- and 5-substituted nitroimidazoles, particularly those containing fused five or six member oxazole or oxazine ring systems [52, 53]. Because these compounds contained an aromatic nitro substituent that was required for activity their development has been consistently plagued by concerns about genotoxicity. The Ciba—Geigy Hindustan group that initially reported the potent aerobic activity of these molecules abandoned their development because of such concerns. Logically, however, the 4- and 5- nitroimidazoles have reduction potentials far outside the reach of mammalian nitroreductases. They appear as genotoxic because of the use of a bacterial species as an indicator for genotoxicity (Salmonella typhimurium) in the classic Ames test for mutagenicity and these compounds can be reduced by bacterial nitroreductases [52]. One of the key insights in this series was that there was distinct SAR for the bacterial nitroreductase that was unrelated to the SAR for antitubercular activity. Of course a relationship between the SAR for human genotoxicity and that of the nitroreductase from S. typhimurium is extremely unlikely making this, at best, an effort to satisfy regulatory concerns that is scientifically highly dubious.

Fig. (5).

Antitubercular nitroimidazoles. Inspired by the natural product azomycin (21), nitroimidazole-containing antibiotics such as metronidazole (22) were developed and are widely used antibac-terials with anaerobic activity. Building on earlier programs at Ciba-Geigy Hindustan that first explored bicyclic nitroimidazoles for their aerobic activity against TB, PathoGenesis produced PA-824 (23) that is currently in Phase II clinical trials while Otsuka Pharmaceuticals produced OPC-67683 (24) that is like-wise in the clinic. These candidates were primarily driven by opti mization of whole cell activity, combined with activity in murine models of TB.

The compound series was revived at a small Seattle Biotech startup (PathoGenesis) that produced many additional analogs, screening most of these in infected mice using a moderate throughput luminescent mouse model [54]. Ultimately this program produced PA—824 (23) that is currently in Phase II clinical trials [55]. In parallel, a program to identify inhibitors of mycolic acid synthesis was in progress at Otsuka Pharmaceuticals in Japan that had picked up on a related series of compounds. This group developed another candidate (OPC—67683, 24) that is also currently in Phase II trials [56, 57]. In both cases the activity was optimized against whole cells, at Otsuka in vitro, and at PathoGenesis in vivo. In both cases compounds continued to be screened in the Ames test and analogs were found that were not reduced by the S. typhimurium nitroreductases.

So the lessons that could be derived from the nitroimidazole story were that whole cell activity (or in vivo) activity was a perfectly plausible and successful way to proceed. However, both in vitro and in vivo tests only provide readouts of aerobic activity yet both pathologists and microbiologists continued to think that anaerobic activity was critical to sterilizing TB lesions [58]. Neither of the two candidates are particularly potent under anaerobic conditions, nor was there a simple way to optimize them for such activity. Subsequent work to understand the mechanism of TB killing under anaerobic conditions [59] and explore the SAR of compounds for this activity [60] revealed the full details of a unique enzymatic reduction process and the subsequent release of reactive nitrogen species leading to killing under anaerobic conditions [61]. This work has, for the first time, enabled a structure-based approach to design of a pro-drug for TB and second-generation nitroimidazoles optimized for anaerobic activity that can engage the tools of structure based design are a topic of active programs.

8. Y2K: The DIARYLQUINOLINES and BENZOTHIAZINONES

The diarylquinolines are the most recent new class of molecules to enter clinical trials for TB but the story of their discovery repeats many of the same themes we’ve encountered in the previous examples. TMC207 25, Fig. (6), the candidate diarylquinoline that has completed some Phase II testing with dramatic results [62, 63], was discovered through a whole-cell screening program run by TiboTech, a Belgian company. Unusually the drug was discovered not by screening against Mtb itself but against a fast-growing cousin M. smegmatis [64]. Little published information is available on the medicinal chemistry program that led to this candidate outside of the patent literature that provides no information on the logic that led to TMC207.

Fig. (6).

Diarylquinolines and benzothiazinones. The newest generation of TB agents, both molecules were selected in whole cell screens and resulted in important new targets for future programs. TMC207 (25) targets the ATP synthase of TB and kills even non-replicating bacteria, BTZ043 (26) targets an enzyme involved in cell wall arabinan biosynthesis.

The compound was exciting because of the identification of a new target (the c subunit in the FO domain of ATP synthase) of Mtb and the subsequent realization that ATP levels were a critically vulnerable point of attack for nonreplicating Mtb [65, 66]. Computational studies using the E. coli ATP Synthase subunits for which structural information is available provided a post hoc rationalization of the preference for the R,S stereoisomer in binding to the interface of the a and c subunits of the ATP synthase [67]. Nonetheless there is relatively little information available to suggest the rationale behind the lead-optimization chemistry and many compounds were synthesized as racemic mixtures limiting the amount of information that could be obtained.

One important lesson that emerged from this program was the utility of doing proof-of-concept studies in patients with MDR-TB disease to obtain an early indication of utility of an agent [63]. TMC207 showed a convincing effect in a trial of only 47 subjects with this difficult to cure disease, fully validating the ATP synthase target. Another lesson highlights one of the difficulties and risks of relying on whole-cell screens. TMC207 has now been recognized as a good substrate for cytochrome P-450 isoform 3A4, an isoform that is highly induced by rifampicin co-administration and studies in patients where the two agents were coadministered showed that the TMC207 level was reduced by half. While it should be possible to identify the site of metabolism and possibly block this, the limited knowledge of the SAR for the target enzyme will make this more difficult than would be the case if there was molecular information regarding the interaction of the diarylquinolines with the target ATP synthase. In this case, in common with the aminoglycosides and rifamycins, the target is an extremely complex piece of cellular machinery that is not very amenable to most modern drug discovery tools.

The benzothiazinones 26, Fig. (6) are the most recent addition to the list of candidate molecules for TB therapy. These compounds were discovered by whole cell screening of various heterocycles [68]. The lead compound BTZ038 (26) contains an essential sulfur and an aromatic nitro group that are essential for activity. Enantiomers of its single chiral center were found to be equipotent. Again this compound uncovered yet another novel target in TB, an epimerase involved in converting lipid-linked ribose into arabinose to be used in constructing the complex arabinogalactan polymer of the mycbacterial cell envelope. Curiously, an independent whole cell screen identified another potential, structurally unrelated hit molecule that apparently targets the same epimerase [69]. This either suggests that this target is unusually sensitive or that there is some as yet unappreciated complexity in the mechanism of BTZ038.

9. Dragging TB Drug Discovery into the Modern Era

Given the history of success in programs that are driven by whole-cell activity and the repeated failures of target-led efforts to provide leads that translate into whole cell actives [63] it is unsurprising that there is an overwhelming sentiment throughout the TB community in favor of whole-cell screening. The success of TMC207 has reinforced this notion considerably and more and more this is translated as a failure of genomics to deliver what was promised. In fact, a lot of target selection based upon the genome sequence was done using “essentiality testing”. In effect this testing postulates that if a genetic knockout of a given protein fails to grow in vitro then this gene constitutes a good drug target. The misconception is that complete absence of an enzyme is equivalent to disruption of an enzyme’s function by a small molecule. In reality obtaining 100% inhibition of an enzyme using an inhibitor is rarely achievable and unless that particular enzyme happens to be the rate-limiting step in a biochemical pathway, a fraction of normal activity will often suffice to permit replication of the organism.

A second misconception lies in the nature of biochemical events that lead to cell lysis. Recent work from JJ Collins laboratory has begun to explore the underlying complexity of drug killing by antibiotics and suggests discrete mechanisms that are often very distal to the immediate target but which are the ultimate causes of cell death [70, 71]. Predicting such indirect effects of drug action from genomic information will require a considerably deeper understanding of cellular systems than we possess currently. As a result whole-cell screening for such vital sensitive steps in metabolism makes a good deal of sense but leaves the subsequent follow-on chemistry for lead optimization on the fragile ground of interpreting SAR only in terms of whole-cell activity and removes from consideration powerful technologies such as fragment-based screening and structure-based optimization.

A recent example of an attractive alternative approach to the development of antibacterials targeting Gram-negative pathogens including Haemophilus influenzae offers an attractive middle ground. In this program at Pfizer whole cell screening of an efflux compromised E. coli strain led to the identification of the pyridopyrimidines as potent lead compounds containing a pharmacophore known to bind to ATP binding sites in eukaryotic kinases [72]. These hit compounds were then used to deconvolute the bacterial target resulting in the identification of the biotin carboxylase (BC) subunit of acetyl CoA carboxylase. The Pfizer team then crystallized the BC from E. coli, H. influenzae, and Staphylococcus aureus in complex with the hits identified in the whole cell screen, information which was used to drive a structure-based lead optimization approach that resulted in orally available compounds showing in vivo activity in animal models of disease. There were several significant accomplishments in this work, selecting upfront for sensitive targets using whole-cell screening and coupling this to target identification and structural biology obtained the advantages of both approaches. Because of the depth of knowledge obtained from the structure, future problems in ADME or toxicology can be addressed in an informed way, making ultimate progression to a useful drug much more likely than if only whole-cell SAR were available. This team also capitalized on the extensive knowledge and experience of ATP binding sites of kinases. But the real beauty of this approach is that even if the series that was the hit allowing identification of the target becomes intractable, alternative scaffold can be rapidly identified using techniques such as computational and fragment-based screens. The team at Pfizer went on to do just that [73].

TB drug discovery is being rapidly outpaced by the development of resistance by the organism. Unless strategies are developed to allow chemists to engage modern tools in this process it appears unlikely that the goal of eliminating TB will ever be achieved. Worse there remains a persistent chance of an epidemic of untreatable, highly drug-resistant TB disease rising to the point of becoming a global epidemic.