Challenges of Antibacterial Discovery - PMC

Summary: The discovery of novel small-molecule antibacterial drugs has been stalled for many years. The purpose of this review is to underscore and illustrate those scientific problems unique to the discovery and optimization of novel antibacterial agents that have adversely affected the output of the effort. The major challenges fall into two areas: (i) proper target selection, particularly the necessity of pursuing molecular targets that are not prone to rapid resistance development, and (ii) improvement of chemical libraries to overcome limitations of diversity, especially that which is necessary to overcome barriers to bacterial entry and proclivity to be effluxed, especially in Gram-negative organisms. Failure to address these problems has led to a great deal of misdirected effort.

Contact us to discuss your requirements of methyluracil ointment factory. Our experienced sales team can help you identify the options that best suit your needs.

The antibacterial product pipeline has not been empty during this time but has been filled with improved versions of previously registered classes. Many of these were true improvements, adding bacterial spectrum, safety, simpler dosing regimens, and most importantly, activity insusceptible to specific resistance mechanisms acting on the parent compound. A number of drugs with improved activity against resistant pathogens, such as oritavancin, iclaprim, and ceftobiprole, have reached the FDA but have met with regulatory problems, largely due to inadequacy of trials in proving noninferiority. Telavancin has been approved, and development of ceftaroline continues. Finding a derivative with an exploitable advantage over the original drug is not an easy path but is a process whose starting material has acceptable pharmacological properties and whose modification may be approached rationally. While the same caveats of meeting pharmacological and toxicological standards apply to this effort as to novel drug discovery, the leap is much greater for novel discovery, as it requires that the leads meet a tremendous number of criteria.

It is due in large part to this discovery void that Big Pharma has been withdrawing from research in the area, even though there has certainly been recognition of the continuing need for new antibacterials to combat the rise of resistant organisms.

Walsh has noted that “no major classes of antibiotics were introduced” between 1962 and 2000 and refers to the interim as an innovation gap ( 115 , 378 ). This understates the problem. The latest registered representatives of novel antibacterial classes, linezolid, daptomycin, and the topical agent retapamulin, were indeed introduced in 2000, 2003, and 2007, respectively, but these chemical classes (oxazolidinones, acid lipopeptides, and pleuromutilins) were first reported (or patented) in 1978 ( 124 ), 1987 ( 86 ), and 1952 ( 275 ), respectively. A timeline (Fig. ) of dates of discovery of distinct classes of antibacterials (as opposed to dates of introduction) illustrates that there have been no (as yet) successful discoveries of novel agents since 1987. There is a discovery void of unknown extent rather than a gap. While there are a small number of novel compounds in the early clinical phase that might portend the end of this hiatus, in most cases their eventual developmental success is unclear. Is the void due to a lack of innovation? While the simple definition of innovation is the act of introducing something new, the word implies creativity, intent, and experimentation. Almost all the discoveries shown in Fig. (with the exception of trimethoprim, monobactams, fosfomycin, and carbapenems) were serendipitous, made by screening fermentation products or chemicals for inhibition of bacterial growth (empirical screening). Not especially innovative, but it worked. In fact, since those last discoveries in the 1980s, there has been a great deal of creative, rational, technologically cutting-edge screening for and efforts at design of new antibacterials. But so far, little has reached serious development. The problem with the conduct of antibacterial discovery since the early 1980s is not a lack of innovation.

The challenges to antibacterial discovery have kept the output of novel antibacterial drug classes to extraordinarily low levels over the past 25 years, even though discovery programs have been in place at large and small pharmaceutical companies as well as academic laboratories over this period. This review focuses on the scientific challenges to the discovery of novel small-molecule antibacterials rather than on the commercial and regulatory considerations, which are well covered in a number of reviews ( 186 , 197 , 301 , 303 , 345 ). Rate-limiting steps to the discovery process are discussed, and some perspective on avenues to address those limitations is offered. An underlying thesis of this review is that the bleak picture of antibacterial discovery is due to an expenditure of effort and resources on non-rate-limiting steps of the process. While it is easy to find compounds that kill bacteria, it is hard to find novel antibacterial classes worthy of development. If new molecular entities with desirable properties and specificity had been discovered commonly throughout the past 25 years, it seems likely that large pharmaceutical companies (Big Pharma) would have viewed the area as productive and continued with antibacterial discovery. Even if unlimited money were poured into discovery and problematic regulatory guidelines were improved and stabilized, then it is probable that novel discovery would still be stymied because scientific obstacles remain to be overcome.

If success as a monotherapeutic is indeed due to multitargeting (or targets encoded by multiple genes) and single-targeted agents, prone as they are to single-step mutation to target-based resistance ( 347 , 388 ), are not optimal for monotherapy, then there are grave implications for antibacterial discovery. The impact of endogenous resistance (that occurring by antibiotic selection in the pathogen) on antibacterial drug discovery and development is covered below.

It remains a hypothesis that the low potential for target-based resistance is causally related to the success of multitarget agents in monotherapy. This is supported by the inverse, that most single targeted antibacterials in the clinic are indeed subject to single-step high-level resistance selection and are used as part of combination therapies, especially in therapy of M. tuberculosis or as topical agents (see Tables 3 and 4 of reference 337 ). Of course, all antibacterials with even moderate spectra have multiple homologous targets in that they must inhibit enzymes or bind to structures that are present and varied among the bacterial species of that spectrum.

The clinically used agents that target rRNA in their inhibition of protein synthesis provide another avenue of support for the hypothesis. These include macrolides, lincosamides, chloramphenicol, oxazolidinones, tetracyclines, aminoglycosides, and pleuromutilins (the last is not included in Table because it is not yet used systemically). They are useful in monotherapy against organisms that contain multiple copies of rRNA genes. Against the slow-growing mycobacteria, however, which contain only a single rRNA cistron ( 36 , 188 ), they are used in combination with other agents because single base changes in the rRNA gene lead to high-level resistance. With Helicobacter pylori, which contains 2 rRNA cistrons ( 55 ), clarithromycin resistance arises during therapy ( 2 , 376 ), and heterozygous strains display a resistant phenotype ( 376 ). Of course, anti-H. pylori therapy generally involves 2 or more antibacterial agents, although not strictly due to resistance development. For enterococci, it has been shown that MICs of linezolid-resistant isolates are highly correlated with the percentage of rRNA cistrons mutated ( 237 ). In a way similar to the rRNA case, the FQs which have dual targets in standard pathogens have only a single target (DNA gyrase) in both Mycobacterium tuberculosis and H. pylori ( 160 ) and do yield to single-step resistance ( 178 , 195 ).

Only two of the commonly used antibacterial classes actually target multiple different enzymes in a given species. The beta-lactams target the penicillin binding proteins (PBPs) ( 34 , 93 , 130 , 348 ), and the fluoroquinolones (FQs) target the catalytic subunit of DNA gyrase (GyrA) and topoisomerase IV (ParC) ( 65 , 112 , 189 ). The multitarget hypothesis was offered before the second target of the FQs, topoisomerase IV, was recognized ( 112 , 189 ). The FQs had appeared to be an exception to the rule ( 340 ), since resistance to the FQs was not extensive in the clinic by the early 1990s. This finding thus served to support the hypothesis.

The fact that successful systemic antibacterials have multiple molecular targets or targets encoded by multiple genes has been evident for the past 20 years ( 50 , 73 , 204 , 337 , 339 , 340 , 347 ). This is illustrated in Table , where the currently used systemic monotherapeutic agents and their targets are listed. These antibacterials are not subject to high-level target-based resistance by single genetic changes in the host. The hypothesis is that these agents are successful monotherapeutics and not subject to such resistance because they are multitargeted ( 339 , 340 ).

Many challenges to candidate selection and subsequent development of antibacterials, including pharmacological properties, pharmacokinetic/pharmacodynamic (PK/PD) analysis, and toxicities (both mechanism and chemistry-based), are common to all drug discovery. They have been approached, addressed, and overcome by more-standardized medicinal chemistry magic for many generations of successful antibacterials and other human health drugs and are addressed only minimally in this review.

In regard to target selection, the emphasis here on the importance of choosing targets by their low propensity for rapid resistance selection may seem a narrow view of the problem. There are a number of other important considerations involved in choosing specific targets for rational antibacterial discovery projects. These include (i) essentiality to the organism of the function, enzyme, or structure so that inhibition of enzyme action or blockage of the function leads to inhibition of bacterial growth or, better, death; (ii) conservation of structure of the target enzyme across bacterial species sufficient to provide a useful antibacterial spectrum; (iii) a lack of structural homology with the same or similar functions in the mammalian host in order to avoid mechanism-based toxicity; and (iv) in common with other areas of human drug discovery, “druggability” of the chosen target, in that there should exist sites on the target enzyme or structure that small drug-like molecules can bind to and, in so doing, exert a biological effect. These are important considerations and, in practice, generally lead to the selection of single enzymes as targets to pursue. However, one of the theses of this review is that single enzymes may not make good antibacterial targets due to their potential for rapid resistance development. This possibility has largely been neglected in the course of recent antibacterial discovery, to its detriment, and thus it is spotlighted here.

The purpose of this review is to underscore and illustrate some of those problems unique to the discovery and optimization of novel antibacterial agents that have adversely affected the output of the effort over the past 20 years. These are the rate-limiting steps of the antibacterial discovery process and can be divided into two main areas: (i) proper target selection, specifically the necessity of pursuing molecular targets that are not prone to rapid resistance development; and (ii) limitation of chemical diversity, especially that which is necessary to overcome barriers to bacterial entry and proclivity to be effluxed, especially in Gram-negative organisms.

The direction of novel antibacterial discovery research (as opposed to that of improving upon established classes) in the past 20 years has been to deploy an array of new technologies, based on genomics, bioinformatics, structural biology, and various high-throughput methods, in an effort to transform the giant leap of novel discovery into doable quantum steps. Indeed, the allure of the rational, engineering-oriented, stepwise application of techniques to make the discovery process a turnkey system is understandable. The concept has been to define broad-spectrum (or more species-specific) targets, screen for or design inhibitors, and then hope to address the subsequent obstacles of bacterial entry, non-mechanism-based toxicity, serum binding, pharmacokinetics, etc., in a piecemeal manner. But this approach has apparently not worked.

To summarize, after the Golden Age, antibacterial discovery became target oriented and largely abandoned natural product sources. Big Pharma evidently weighed the costs of maintaining the resources for natural product programs against the low probability of useful output and opted for the synthetic chemical route. Targets were pursued first as a means of dereplication in natural product screening and later to provide a rational basis for discovery and as a route to avoiding cross-resistance with other drugs, as discussed below.

Natural product screening (at least for novel antibacterials) waned with the low output of good leads, the advent of high-throughput liquid handling-based screening methods, for which crude microbial fermentation broths are a poor fit (since they require labor- and time-intensive culture isolation, fermentation, and extraction to produce a relatively limited number of samples), and the rise in the screening of chemical libraries, especially combinatorial chemicals. Antibacterial discovery largely became limited to screening these chemical libraries. This was not a fruitful source, as discussed below.

Much of early industrial antibacterial screening was carried out by cohesive groups that did natural product fermentation and both designed and ran the screens. The scientific direction and prioritization of resources were done within the group. But changes in the pharmaceutical industry led, in many cases, to a modularized system that is still more or less in effect. Drug discovery programs for different therapeutic areas (such as infectious diseases, cardiology, oncology, immunology, etc.) are generally organized such that biology and sometimes chemistry are committed to that area, but other functions (screening, animal testing, pharmacology, structural biology, etc.) may be shared. Since resources are always limiting, their allocation became a relatively high-level management decision (often at a remove from bench science), weighing the value to the company of a therapeutic area, the probability of success, the proximity to the “cutting edge” of current technology, and the ability of the scientists and their managers to push specific programs. For example, antibacterial discovery groups had to compete with other therapeutic areas for the opportunity (a so-called “slot”) to screen natural products. The awarding of natural product screening slots came to be based on the perceived attractiveness of the target and its amenability to downstream biochemical and physical analysis. Those antibacterial screens designed to find primarily novelty (over inhibitors of a specific target) were often given low priority. It could be argued that finding novelty is the goal of natural product screening for antibacterials and that concentration on a small number of preselected “desirable” targets (for which inhibitors might or might not be present) is an inefficient use of the natural product resource. Screening strategies for novelty among natural products are noted in a later section.

Importantly, in 1977, at a time when the output of novel antibiotic classes had decreased, the low-hanging fruit having been found, Cohen proposed rational chemotherapy of infectious organisms through a search for inhibitors of specific enzymes in the target organism ( 77 ). This, along with the growing ability to clone genes and manipulate bacterial strains to enhance whole-cell phenotypic screens for inhibitors of specific targets (and eventually allow the production of purified proteins which could be used for in vitro screening and assays), turned the whole of antibacterial discovery toward more target-directed screens.

In an effort to make dereplication easier, starting by the early 1960s ( 126 ), screening methods were modified in order to limit the hits to subsets of all possible antibiotic compounds. For example, many screens were developed over the years to detect inhibitors of the pathway of peptidoglycan (cell wall) synthesis ( 126 , 278 , 333 ). Each time a hit in such a screen was detected, it could be compared for biological and chemical similarity to the previously discovered cell wall synthesis inhibitors. Thus, pathway- or rudimentary target-based screening arose in part for dereplication purposes but also because certain pathways (cell wall and protein synthesis) appeared to be common targets for useful antibiotics. Furthermore, it was early recognized that cell wall inhibition was a very selective antibacterial target. The only clinically useful antibacterial classes discovered through directed screening thus far (monobactams, carbapenems, and fosfomycin) were discovered in these cell wall pathway screens.

The earliest history of antibacterial chemotherapeutic discovery was via screening dyestuffs and other chemicals for selective antibacterial activity, yielding salvarsan and the sulfa drugs. When the folate pathway inhibited by the sulfas was better understood, more directed screening of pyrimidine derivatives and analogs for inhibition of the bacterial folate pathway produced trimethoprim, an inhibitor of dihydrofolate reductase ( 57 , 158 , 375 ). But the so-called “Golden Age” of antibacterial discovery involved screening of natural products, especially fermentation broths and extracts of microorganisms, simply for the ability to inhibit growth of bacterial organisms of interest (pathogens or surrogates), without regard to their mechanism of action. This has been termed empirical screening ( 71 , 342 ). Selectivity was generally tested in secondary assays of toxicity in animals. This worked admirably for a number of years, as the most common antibiotics (natural product-derived antibacterials) were discovered and rediscovered rapidly. The prevalence of production of “common” antibiotics among standard Actinomycetes has been estimated by Baltz ( 22 ). To efficiently discard such previously described compounds, methods of so-called “dereplication” were quickly developed to identify them ( 1 , 104 , 352 ).

The fosfomycin case raises the possibility that endogenous resistance to single-enzyme targets may be avoided if drug levels at the infection site can be kept high without toxicity and/or the mutants are unfit or of low virulence. Clearly, fosfomycin has been successful (although for a limited indication), but is the fosfomycin/MurA scenario more broadly applicable? Should single-enzyme targets be avoided altogether, or has in vitro analysis of resistance frequencies had an unnecessary chilling effect on discovery programs within industry? Has the awareness of the potential for resistance to single-target agents led to the early demise of programs that would otherwise have proceeded—to optimization or even to the clinic? This is a chastening thought.

Although most single-enzyme-targeted agents are used in combination or topically and thus avoid rapid endogenous resistance development, there are a few exceptions, such as fosfomycin ( 272 ), which has been used successfully (outside the United States) against urinary tract infections (UTIs). Why is there a lack of clinically relevant endogenous fosfomycin resistance? Fosfomycin targets UDP-N-acetylglucosamine enolpyruvyl transferase (MurA), the first committed step of peptidoglycan synthesis, forming a covalent adduct with an active site cysteine ( 181 ). Despite its covalent and irreversible action, its activity appears to be highly selective, and it has a very low toxicity (50% lethal dose [LD 50 ] in mice of >20 g/kg of body weight when dosed orally [ 126 ]). M. tuberculosis is naturally resistant to fosfomycin due to the existence of an aspartate instead of cysteine at that site, and the aspartate-containing enzyme is highly active in Escherichia coli ( 190 ). Thus, it appears that the cysteine is not required for enzyme activity, and theoretically, a mutation consistent with cell growth could occur at that site, leading to fosfomycin resistance. However, there have been no reports of in vitro selection of spontaneous murA mutants resistant to fosfomycin (although one was selected after mutagenesis and counterselection against uptake mutants [ 393 ]). Rather, selection with fosfomycin in the laboratory results in a relatively high frequency of resistance due to loss of the α-glycerophosphate (glpT) or hexose-P (uhp) active uptake system ( 272 ), and the high rate of uptake mutants (as great as 10 −7 per generation) may obscure selection of murA mutants. These uptake mutants have been reported to be nonvirulent and slow growing ( 272 , 389 ) and, hence, likely to be unstable in an infecting population. Furthermore, fosfomycin treatment leads to high urinary levels of drug (>1 mg/ml for 12 h after oral dosing [ 326 ]), which would likely kill off preexisting mutants (of the target or uptake type) resistant to lower concentrations; additionally, the total organismal load in uncomplicated UTI is probably <10 8 (based on calculations for E. coli in reference 272 ). In the clinic, fosfomycin resistance generally is due to covalent formation of a fosfomycin-glutathione adduct by FosA, FosB, and FosX enzymes which have spread by HGT ( 52 , 113 , 252 ).

If they preexist in a population, mutations conferring resistance will be selected for by treatment with a dose of inhibitor that kills off the parental strain but to which the mutant strain is resistant. Under conditions of drug treatment, then, such mutants compete well and are fit relative to their dead siblings. In the absence of drug, many forms of resistance can exact a fitness cost, such that mutants will be slowed in growth rate and will not compete well with the nonmutant, sensitive parental strain ( 9 ). However, further compensatory mutations can often occur that reduce the fitness cost of the original mutation, and these will tend to stabilize the resistance mutation in the population ( 8 ). It should be noted that hypermutators also will play a role in the appearance of these compensatory mutations ( 388 ). It is the complex balance of these events and the pressures of repeated selection with antibacterial agents that control the overall rate of evolution of resistance that occurs upon clinical introduction of a new agent ( 241 ). The occurrence of compensatory mutations has been studied for a number of drugs, both in vitro and in clinical isolates, by Andersson and coworkers (for example, see references 198 , 230 , 231 , 268 , 273 , and 292 ). For new and novel drug candidates, how should the problem of fitness be addressed? In vitro methods, including hollow-fiber “pharmacodynamic infection” models, have been described ( 9 , 99 , 144 , 241 ) and can be used profitably, but animal models for resistance selection and competitive fitness should also be standardized and applied ( 39 , 220 , 241 ). Since it is difficult to predict the impact resistance would have in the clinic when an inhibitor is already in hand, it should be apparent that predicting low resistance potential for a given target in the absence of an inhibitor is much more problematic. It is possible to predict, however, by using microbial genetics, that inhibition of a particular target might lead to a bypass event at a relatively high frequency [see “Peptidyl deformylase” below].

Optimally, it should be determined (i) whether single mutational events that raise the MIC above clinically relevant drug levels can occur in a target organism and (ii) whether strains containing these mutations are sufficiently fit and virulent to survive and be infective in the absence of selective pressure. This is easier said than done.

What does a specific frequency portend for the future potential of endogenous resistance development in the clinic? Generally, a frequency of <10 −10 is sought because organisms in an infection can reach 10 9 cells/ml ( 256 , 281 ) or 10 10 cells in an infected individual ( 98 ). However, this may not be stringent enough. As noted above, the use of hypermutator strains can help to reveal the range of endogenous resistance. While these can give up to 1,000-fold higher resistance frequencies than normal ( 256 ), their use may be particularly relevant, since a significant percentage of antibiotic-resistant clinical isolates, especially those from chronic infections, have been shown to be hypermutable strains ( 138 , 146 , 241 , 256 , 359 ). The pressure of selection for mutations will itself select for hypermutators ( 235 ).

The resistance frequency (or rate) depends upon the concentration of the selecting inhibitor. If the inhibitor has a single target, it may require plating at a relatively high multiple of the MIC to detect target-based resistance, since at lower multiples, mutations that affect permeability and efflux functions occur at relatively high frequencies and may predominate. With some single-enzyme inhibitors, such as rifampin, single base changes can raise the MIC 32,000-fold ( 17 ). If the inhibitor has multiple targets with various sensitivities to inhibition (but within a narrow concentration range), as with the FQs, then the increment of MICs possible with changes to one of the targets can be small and will be related to the difference in intrinsic sensitivities of the targets. That is, if the most sensitive (primary) target is inhibited sufficiently at an external concentration of 0.01 μg/ml to prevent growth (MIC = 0.01 μg/ml) and the secondary target is inhibited at 0.04 μg/ml, then even a 100-fold decrease in the sensitivity of the primary target would raise the MIC no higher than 0.04 μg/ml. In fact, this illustrates the benefit of having multiple targets. Even though mutations in the most sensitive target (GyrA or ParC, depending on the species and drug being tested) occur at significant frequencies, high-level FQ resistance requires multiple mutational events ( 98 , 171 , 354 ).

A number of reviews have described useful methods for ascertaining resistance frequency (number of resistant organisms in a given population) or rates of resistance (number of mutational events leading to resistance per bacterium per generation) to a given antibacterial in the laboratory ( 240 , 281 , 314 ). O'Neill and Chopra ( 281 ) give practical information on preclinical evaluation of novel antibacterials, including important directions for evaluation of resistance potential in vitro. Martinez et al. ( 241 ) emphasize that such measurements should be made under a variety of growth conditions. Several authors recommend the use of hypermutator strains to determine the range of possible endogenous resistance mutations ( 241 , 256 , 282 ). The determination of mutation rates by fluctuation tests ( 227 , 281 , 314 ) avoids “jackpots,” which can occur when single saturated cultures are plated and may distort determinations of mutation frequency. It also demonstrates, as originally intended by Luria and Delbruck ( 227 ), that mutation to resistance can occur before selection is applied. Mutation frequencies to significant levels of resistance (between 10 −6 and 10 −9 ) usually indicate a single target. The higher rate would generally be due to resistance via loss of a function, which can occur through deletion, insertion, or base changes at many sites in the gene encoding that function. The lower frequency (10 −9 ) would indicate that resistance is due to a limited number of allowable base changes at a single site.

In summary, it seems that in the initial stages of antibacterial discovery, endogenous resistance, that which is selected for by the lead compound in the pathogen itself, is critical. Successful development of such compounds will depend on whether endogenous resistance compromises monotherapy. What level of resistance selection in vitro is compatible with advancement of a lead to clinical candidate status?

Thus, it may be misleading to apply lessons learned from the patterns of resistance development via HGT to expectations for inhibitors of new targets. It might be more reasonable to expect the patterns of resistance development to single-enzyme inhibitors that are seen with drugs used in therapy of M. tuberculosis or, for that matter, HIV, where the drugs are all single targeted or subject to high-level resistance via single mutations. It is clear for M. tuberculosis that resistance is not due to the exogenous resistome, since it lacks plasmids and does not participate in HGT, but to endogenous resistance arising through mutation of individual clones of M. tuberculosis ( 131 , 262 , 349 ). As a consequence of single targeting of drugs for these infections, successful therapy for M. tuberculosis and HIV has evolved to use combinations of these agents. Indeed, since the standard of care for M. tuberculosis and HIV is treatment with combinations, the resistance potential of new single-targeted agents for treatment of those pathogens is not as problematic as it might be for more-standard pathogens. In contrast to the case with standard pathogens, a number of interesting new anti-M. tuberculosis agents are in various stages of development ( 229 , 374 ). Is combination therapy a feasible path for development of new single-targeted agents?

For natural products, the range of times between introduction and first report of transmissible resistance in pathogens has been very large: resistance arose immediately for β-lactams, as β-lactamases were seen in Staphylococcus aureus very soon after the broad introduction of penicillin ( 191 ), while vancomycin resistance in enterococci took 33 years to be recognized in the clinic ( 206 ). For the synthetic antibacterials, the first transmissible resistance in pathogens was recognized for sulfonamides, fluoroquinolones, and oxazolidinones in 23, 11, and 6 years, respectively ( 194 , 219 , 242 ). Thus, the prospects for long-term avoidance of resistance to a novel synthetic agent are not rosy, but a few years might be expected to elapse before exogenous resistance mechanisms come into play. However, even this short period may be further abbreviated if endogenous resistance occurs more rapidly, perhaps even during therapy, as could be expected with a single enzyme target.

What relevance does this have to the discovery and development of new antibacterial agents? Obviously, novel drugs intended for development must not be cross-resistant with existing therapies. For at least the last 20 years, the general answer to the challenge of avoiding cross-resistance has been to search for inhibitors of molecular targets that had not previously been “exploited,” that is, they were not the targets of previously developed agents ( 6 , 29 , 49 , 61 , 168 , 253 ). This presupposes that existing resistance mechanisms to the drug classes in use are target directed—which some are—but many are class specific (Table ). Of course, inhibitors of these new targets would eventually fall to exogenous resistance from the resistome. How long would that take?

Studies of resistance genes from antibiotic-producing species that are theorized to be a reservoir for HGT raise the possibility that antibiotics derived from natural products are more likely to be susceptible to such a preexisting set of resistance mechanisms than are totally synthetic drugs ( 61 ). However, it has been shown that among antibiotic-producing genera, resistance determinants for the synthetic sulfonamides ( 82 ), oxazolidinones ( 219 , 324 ), and FQs ( 85 ) are not rare. Thus, the strong implication is that resistance via horizontal transfer from environmental organisms will eventually compromise both known and still-undiscovered antibacterial agents, whether derived from natural products or synthetic.

The collective genomic repertoire of possible mechanisms of resistance to antibacterial agents, via chemical modification or breakdown of antibiotics, target protection, efflux, or specific changes to the target, has been termed the antibiotic “resistome” ( 85 ). The types of exogenous and endogenous resistance mechanisms acting on marketed (nonmycobacterial) antibacterials are summarized in Table . Recent reviews have generally emphasized the role of the exogenous resistome and HGT in the spread of clinically important antibiotic resistance ( 78 , 84 , 85 , 239 , 241 , 390 ). Indeed, most of the mechanisms which have played major roles in resistance to the standard monotherapeutic agents have arisen in this way, with the notable exception of the FQs. According to the multitarget hypothesis, these drugs are useful monotherapeutically precisely because of their low susceptibility to high-level, single-step endogenous resistance development, although chromosomally encoded resistance via efflux or reduced permeability or changes to multiple targets (as with FQs) may compromise these drugs in a stepwise fashion (as noted in Table ). Woodford and Ellington ( 388 ) discuss the importance of mutation in the development of resistance and make the important distinction between those antibiotics to which resistance can arise rapidly in the laboratory (such as rifampin, streptomycin, and fusidic acid) and compromise their use in monotherapy and clinically useful monotherapeutic agents (such as FQs and linezolid) to which resistance may arise via stepwise mutation.

It is an irony of the antibacterial discovery process that finding resistance to a lead due to a single base change in a target can validate the mechanism of antibacterial action of the compound but may well signal that the lead is unfit for development, or at least for monotherapy.

The strongest demonstration that a particular cellular molecule is the proximal target of an inhibitor is to change that putative target in such a way as to prevent interaction with the inhibitor and to show that this blocks further downstream sequelae. With single-enzyme-targeted agents, this may be accomplished by direct selection and mapping and/or sequencing of mutations which increase the MIC significantly, as discussed above. It may also be demonstrated by replacing the gene for the putatively targeted enzyme with one known to encode an insensitive enzyme, as shown with LpxC inhibitors ( 26 ). If the target enzyme is known to be present or essential in only specific species, then a lack of activity of an inhibitor against other species is supportive evidence of the specificity of action, as seen with ClpXP ( 70 ), FabI ( 152 ), and LpxC ( 283 ). With a strong target hypothesis (as when the agent is selected as an enzyme inhibitor), interpretation of resistance selection results is relatively straightforward. The best evidence, as noted above, is the isolation of the mutant target enzyme and demonstration that the resistance mutation leads to reduced inhibition of (or binding to) the altered target in vitro.

Recently, a variety of methods based on the use of arrays (transcriptional [ 121 ], translational [ 23 ], hypersensitization [ 96 ], overexpression [ 394 ], stress response [ 24 ], and synergy/antagonism [ 397 ] arrays) have been described that may be useful in identifying the antibacterial mechanism of action of an inhibitor. These methods have also been used for evaluation of leads from enzyme inhibitor programs or with activities discovered through phenotypic or empirical whole-cell screening. In each of these methods, patterns of the effects of known antibacterial activities on the components of the array are established, and unknowns are then compared to these patterns to identify inhibitors with previously described mechanisms of action. With a hypersensitization array ( 96 ), for example, a set of strains, each underproducing a single target, is exposed to the unknown inhibitor, and those strains which are hypersensitive to the inhibitor are noted. In a few cases, only a single strain will be hypersensitive, indicating the probability that the underproduced enzyme is the target. However, most often several strains will be sensitized to various extents, which may indicate inhibition of a step in a particular pathway. In each type of array, complex patterns are often seen that are difficult to interpret, necessitating further dissection by phenotypic and other genetic means.

Overexpression of a target may be accomplished by cloning of the expected target behind a regulatable promoter or by use of an overexpression library of random potential targets (which are then screened for resistance to the inhibitor) ( 211 ). A rise in the MIC of an inhibitor of an overproduced putative target indicates that the overproduced enzyme can indeed bind the inhibitor and lower its effective intracellular concentration, but it does not ensure that the candidate enzyme is the cause of growth inhibition, only that the inhibitor can bind to that enzyme: another MIC-determining target may be present. For example, curcumin, a spice component with a long history of dietary and medical use in Asia, was shown to inhibit FabI (enoyl reductase) of E. coli in vitro, and its MIC against E. coli was raised >7-fold by overexpression of FabI. However, other investigators showed that curcumin bound to FtsZ and inhibited Z-ring assembly in Bacillus subtilis ( 307 ). Curcumin has also been shown to inhibit the human enzymes glyoxylase I ( 319 ) and monoamine oxidase B ( 308 ), as well as HIV integrase ( 246 ). This type of pattern indicates that curcumin is a promiscuous inhibitor (and probable binder) and that caution should be exercised in target attribution by overexpression.

If the MIC of a given inhibitor is reduced when its putative target is downregulated, then the enzyme in question is likely to be essential, and it may indeed be responsible for the antibacterial activity of the compound. However, reduction of one enzyme can sensitize another enzyme (perhaps the true target) to inhibition (for example, if they are both members of the same pathway) ( 96 ). Furthermore, even if the underproduced enzyme is a target of the inhibitor, that does not preclude the existence of other, less-sensitive targets, nor does it even establish that the underproduced enzyme is responsible for setting the MIC when it is produced at its normal level. For example, underexpression of FabI sensitized S. aureus to a thiopyridine inhibitor of FabI, but overexpression did not raise the wild-type MIC ( 214 ). Furthermore, MMS analysis showed preferential inhibition of RNA and DNA synthesis over that of fatty acids. Thus, underexpression alone is inadequate to prove that the underexpressed enzyme is the MIC-determining target of inhibition.

Frequently used genetic methods implicating inhibition of a particular enzyme as the antibacterial mechanism of action of a compound are underexpression of the expected target protein, in order to sensitize the organism to an inhibitor, and overexpression to yield resistance. Hypersensitization by target underexpression has been demonstrated by various methods, including use of tightly downregulatable promoters directing the synthesis of reduced amounts of the target protein (via reduced transcription) ( 95 ) and upregulated production of antisense RNA, which leads to reduced protein expression ( 118 , 398 ). But there are caveats.

In order to prove that antibacterial activity is due to specific target inhibition, several avenues are possible. With targeted screening, there is already a starting hypothesis for the mechanism of action and molecular target. Thus, initial work may be directed toward demonstration of phenotypes that should be associated with inhibition of that target, such as morphological changes (e.g., filamentation for inhibitors of FtsZ [ 81 , 277 ]), stress responses (stringent response for inhibitors of tRNA synthetases [ 38 ]), and specific promoter induction (gyrase inhibition leads to homeostatic upregulation of gyrase promoters as well as SOS promoters [ 7 ]). With these studies, it is necessary to use a wide variety of negative controls to show that the tested phenotypes are not caused by other classes of inhibitors. An important method of ascertaining the pathway of inhibition is measurement of effects on macromolecular synthesis (MMS). This is done most straightforwardly by use of radiolabeled tracers of DNA, RNA, protein, cell wall, and fatty acid/lipid synthesis, where the dose-response relationship for the test compound (and control inhibitors) is measured at a fixed time of incubation ( 4 , 381 ). For specific inhibitors of one of the pathways of macromolecular synthesis, incorporation of precursors of the end product of that pathway will be inhibited preferentially. If all pathways are inhibited within a narrow concentration range, then a nonspecific mechanism of inhibition, such as membrane lysis or energy poisoning, is likely.

Hits from screens for enzyme inhibition are generally tested early on for antibacterial activity. If the hits are chemically tractable, exploratory medicinal chemistry may be instituted to improve enzyme-inhibitory potency and solubility and, if no or poor whole-cell activity is present, to improve MICs. Throughout this optimization process, it is important to ascertain whether antibacterial activity tracks with enzyme-inhibitory potency. Such tracking may not be seen for perfectly legitimate reasons, as the parameters for net bacterial accumulation are not likely to track with enzyme inhibition. Regardless of the proportionality of MIC to inhibitor potency at the enzyme level, antibacterial activity should be shown to be dependent upon enzyme inhibition throughout the optimization process. Often, in discovery programs, data will be generated that show a general or even good structure-activity relationship (SAR) between enzyme inhibition and the MIC. This is supportive evidence (at best), but it does not show causality. A demonstration of causality is especially critical with a target for which there have not been any antibacterial inhibitors described, where the process of linking whole-cell activity (MIC) to inhibition of the enzyme is critical for target validation. This can be done in a number of ways, as noted below and reviewed by O'Neill and Chopra ( 281 ).

While it is not necessary that an antibacterial discovered during targeted screening hit solely the desired target, or even that target at all, the raison d'être of target-directed screening is that specific bacterium-selective and hence nontoxic inhibitors will be discovered in this manner. While it should be obvious that an inhibitor discovered in a general empirical screen for growth inhibition must be shown to be selective and not kill through nonspecific (and likely cytotoxic) activity (such as detergency, alkylation, energy poisoning, etc.), this is equally important for a compound identified via in vitro enzyme inhibition. This linkage has not been made in a number of cases (as shown below), and eventual determination that the antibacterial activity was not causally linked to enzyme inhibition might have contributed to termination of the program.

How did screening and design with the new targets turn out? An investigation of this question will benefit from discussion of the process of analyzing hits that arise from targeted screens and design programs, with an emphasis on compounds identified by in vitro biochemical screens and assays of enzyme inhibition or binding.

The number of potential target enzymes (selected as essential in bacteria but not humans, with a broad or useful spectrum) has been estimated to be ∼160 by Payne et al. ( 293 ). Lange and coauthors ( 204 ) list 16 enzyme classes that are targets of commercialized antibacterials, in addition to the nonenzyme targets rRNA, lipid II, membranes, and DNA. Thus, there are a significant number of “new” targets that have been nominated for screening and/or inhibitor design. It should be noted, however, that while these “new” targets may not have been screened explicitly for inhibition previously (although most were, in the pregenomic era), empirical screens for whole-cell growth inhibition should have implicitly screened for them.

The focus on targets for discovery led to the deployment of intensive campaigns for target evaluation, to the development of high-throughput screening (HTS) platforms, and to programs of virtual ligand screening and rational structure-based drug design (SBDD). While most of the programs discussed and tabulated below come from screening efforts, SBDD should play more of a role in the future. Virtual ligand screening and a number of its successes in human health drug discovery are reviewed by Villoutreix et al. ( 377 ). A highly relevant recent review by Simmons et al. focuses on rational discovery of antibacterials by SBDD ( 341 ). Antibacterial SBDD is based on the extensive and growing number of sequenced bacterial genomes and solved crystallographic and nuclear magnetic resonance (NMR) structures of bacterial proteins and their bound ligands. Through a variety of algorithms, it makes in silico predictions for docking of new ligands (compounds and fragments, both real and theoretical), which can then be prioritized and tested for enzyme inhibition, structural interactions, and other biological readouts (such as whole-cell activity, solubility, or bioavailability). The process is iterative and heuristic. Two potent inhibitors of HIV protease, nelfinivir ( 182 ) and amprenavir ( 369 ), were developed through such iterative SBDD.

As noted above, potential antibacterial targets would traditionally be defined as essential, distinct from related mammalian structures/enzymes, present in a useful spectrum of bacteria, such that an inhibitor might be reasonably used for therapy of a clinical indication (such as community-acquired pneumonia [CAP]), and possessing a reasonable potential for druggability. At least for protein targets, most of these parameters (aside from essentiality) can be ascertained by in silico methods of bioinformatics and structural analysis. Even the potential for multiple targets sharing active site sequence homologies or protein motifs may be addressed by ever more sophisticated analytical tools, as noted in Multitargeting.

RECENT RECORD OF SINGLE-ENZYME- TARGETED AGENTS

The Opacity of the Discovery Process

What has been the demonstrable record of antibacterial discovery of novel agents in the past decade or so? Since the clinical pipeline of novel agents has been so small, the overall record is clearly not good. But what avenues have been pursued, and why have they failed? In general, the reasons for failure, or even the fact of failure, have not been revealed in the literature. While this could be attributed to the general opacity of the industrial drug discovery process due to commercial concerns of intellectual property and restriction of information that might move stock prices, recounting of failed initiatives is infrequent in the academic literature as well. The patent literature may reveal the industrial seriousness with which projects are regarded. In fact, while reasons for failure may be veiled, it should be noted that success is also often kept from view. In biotech and academe, early advances are often published: the former to secure investor financing, the latter for intellectual pursuit but also to validate grant support. In Big Pharma, however, it is often the case that results of active programs are not publicized or published until they approach the clinic. Thus, publications may recount many dead ends (without advertising them as such or explaining why they did not go forward), as there is little publication on actual leads until they are dropped or until they enter the clinic.

We must often rely on retrospectives of internal programs, such as that of Payne and colleagues, who summarized lessons learned from 70 targeted screening campaigns at GlaxoSmithKline (GSK) (293), or on external subjective analyses such as this and other reviews (111, 204, 209, 280, 302, 335). In the following sections, which review novel discovery efforts, it is generally unknown if or why the projects were dropped. Was it due to high resistance frequency, insufficient potency, an inability to overcome a high protein binding level, a lack of efficacy in animal models due to instability, metabolism, or otherwise poor pharmacokinetics, a nonspecific mechanism of action, undisclosed toxicities, or a poor spectrum making development clinically and/or commercially infeasible, or did lead optimization stop early with the recognition of poor lead structures? It would certainly be instructive to know more.

A possibly instructive survey involves the output of targeted discovery programs or programs that produced inhibitors with purportedly identified targets. This survey should serve to exemplify some of the problems of discovery and development and, furthermore, illustrate that inhibitors of antibacterial targets are discoverable but that few have been optimized sufficiently to be chosen for development. Programs attempting to exploit ∼35 targets are noted in the text and in Table , among which only 3 (MurA, RNA polymerase, and DNA gyrase) are targeted by drugs already used in human therapy.

TABLE 3.

TargetFunctionNonantibacterial enzyme inhibitor(s)Antibacterial inhibitor(s) (some with SAR)Inhibitor(s) with some genetic/phenotypic evidence (but insufficient for proof)Inhibitor(s) with probable additional targetWell-validated antibacterial inhibitor(s)MurACell wallT6362R (107)Compound 1 (28)Three compounds (33)MurIGlu racemase

d

-Glu analogsb (88)Pyrazolopyrimidinedione (89)Cell wallPyridodiazepine amines (129)MurGCell wallRhodanines (163)WalK/WalRCell wall regulationThiazolidinones,c benzamides,c pyrimidinonec (305)UppSBactoprenol synthesisTetramicc and tetronicc acids (297)FtsZCell divisionZantrins (236)C-seco-taxanesb (164)3-Methoxybenzamide (MBO) (277)Dichamanetin (372)PC190723b (81, 151)Viridotoxin (379)Cmpd 2 (MBO derivative)b (150)4-Aminofurazan (173)FabI (FabL and -K)Fatty acid synthesisLuteolinc (396)Imidazoles (154)Thiopyridines (214)Aminopyridines (177, 258, 294)Enoyl-reductasesAquastatins (203)AFN1252a (185), MUT37307a,b(106),

{"type":"entrez-nucleotide","attrs":{"text":"CG400549","term_id":"34399433","term_text":"CG400549"}}

CG400549 (290)Curcuminc (396)FabKFatty acid synthesisPhenylimidazole derivatives of 4-pyridoneb (193), AG205 (357)Enoyl-reductasesCoaACoenzyme APantothenic acid derivativesb (74)AccA, AccDAcetyl-CoA carboxylaseMoiramide B (120, 121)Pyridopyrimidines (255)Pyrrolidinediones (122)LpxCLipid A deacetylaseUDP analogsc (25)BB-78485 (76)Hydantoinsb (80)CHIR-090b (247)LptD (Imp)LPS translocasePeptidomimetic (350)RpoB and -CRNA polymeraseSB-2 (rhodanine) (10)CBR703 (15)Ureidothiphenes (14)Ef-TuTranslation elongationIndole-dipeptide (176)LeuRSLeu tRNA synthetaseABX (benzoxaborole) (123)MetRSMet tRNA synthetaseREP 8839 (139, 175)PheRSPhe tRNA synthetaseThiazolylurea-sulfonamides (38)Spirocyclic tetrahydrofurans (157)PdfPeptidyl deformylaseActinonin (68), BB-3497a (75)Ro 66-6976, Ro 66-0376 (66)LBM-415a (209), N-alkyl urea hydroxamic acids (145)MapMet aminopeptidasePyrazolediaminesc (109)Catechol thiazoles, catechol thiophenes (382)PolCDNA polymerase IIICAnilinopyrazolopyrimidinonesb (4)Anilino-uracilsb (53, 221)HB-EMAUb (59), EMAIPUb (199)DnaGDNA primaseBenzopyrimido-furans,c pyridothieno-pyrimidinecMenAMenaquinone synthaseHydrazine or tertiary or secondary amine-containing mimics of demethylmenaquinoneb (200)ClpXPProteaseCyclic peptide IXP1 (70)DnaKChaperoneTetradecanoyl-aminomethylbenzoyl-

l

See also:
Frequently Asked Questions - DCIPS - Department of Defense
HPMC for Tile Adhesive
The Ultimate Buyer's Guide for Purchasing Carboxylic nitrile latex

If you are looking for more details, kindly visit dimethyl urea hardeners.

-isoleucineb (213)His kinasesSignal transducersThieno-pyridine (132)Open in a separate window

MurB to MurF Enzymes as Antibacterial Targets

Many attempts have been made to find inhibitors of cytoplasmic enzymes of the peptidoglycan synthesis pathway, including MurB (UDP-N-acetylenolpyruvylglucosamine reductase) and the UDP-N-acetylmuramyl-amino acid ligases MurC, -D, -E, and -F, which sequentially add l-Ala, d-Glu, meso-diaminopimelic acid (m-DAP) (in most rods) or l-Lys (in most cocci), and d-Ala-d-Ala, respectively, to UDP-muramic acid. These have been considered good antibacterial targets because they are part of the synthetic pathway of the essential macromolecule peptidoglycan and were themselves shown to be essential by conditional mutations (223-226). It would have been expected that inhibitors of these enzymes would be found in the many phenotypic screens for inhibitors of the cell wall pathway (126, 278, 333, 336), but they were not. Merck and Versicor (later Vicuron, and then Pfizer) reported using a “one-pot” in vitro screen for inhibitors of MurA to MurF, but no inhibitors arising from them have been reported (67, 387). In fact, while some inhibitors of the enzymes MurB through MurF have antibacterial activity, in no case has that activity been shown to be due to inhibition of the Mur enzymes in vitro.

Many of the reported inhibitors of MurB to MurF have been described in earlier reviews (105, 196, 334, 336). Phosphinate transition-state analogs of the MurC to -F enzymes were synthesized by various groups in the 1990s, and some of these showed potent enzyme inhibition (at low nanomolar levels for some, illustrating druggability of the targets) but no antibacterial activity, presumably due to a lack of cell entry (127, 254, 310, 362, 403). A number of inhibitors of MurB, -C, -D, and -F, found via screens for enzyme inhibition or binding, had micromolar (in some cases, low micromolar) 50% inhibitory concentrations (IC50s) and no (or weak in one case [370]) antibacterial activity (11, 30, 103, 141, 364, 370).

The thiazolidinone core of the MurB inhibitors reported by workers at Bristol-Myers-Squibb was postulated to be a diphosphate mimic (11). These inhibitors were not antibacterial, but replacement of the thiazolidinone with an imidazolinone led to gain of antibacterial activity (48). There was a correlation between MIC and enzyme-inhibitory activity in a small series, but no further proof of causality was demonstrated. This is one of many cases in the literature where the project was left (as far as reported) at the preproof stage.

A number of synthetic projects undertaken by the Wyeth group (12, 119, 201, 212, 234, 395) were directed toward finding inhibitors of multiple enzymatic steps in the Mur pathway, with the idea that hitting multiple targets would lessen ultimate resistance selection. For many of these, the lead compound was discovered in a “one-pot” MurA-to-MurF screen (234), probably similar to those of Merck and Versicor. Hits from that screen were assayed against the individual enzymes, and leads were chosen for expansion. Where tested, these inhibitors were reported to be subject to abrogation of antibacterial activity by 4% bovine serum albumin (BSA). These campaigns all yielded sets of compounds with various spectra of target inhibition, from inhibiting a single target (among MurA to MurF) to inhibiting sets of one, two, three, or four enzymes. Some were shown to inhibit synthesis of soluble peptidoglycan in whole cells, but no other MMS inhibition was measured (119, 395). The lack of this negative control lowers the value of the finding. Examples of these compounds and the enzymes they preferentially inhibit are shown in Fig. .

Open in a separate window

The active sites of the Mur ligases (MurC to MurF) have been shown to have homology (110), so inhibitors of multiple Mur ligases are not unexpected. It is not impossible that a single molecule (especially a diphosphate mimic) could inhibit a range of enzymes, including MurA or -B plus some of the Mur ligases, since all of these enzymes have a UDP-sugar as a substrate. With a goal similar to that of Wyeth, a group at the University of Ljubljana discovered rhodanine derivatives that were balanced inhibitors of MurD, -E, and -F, one of which had extremely weak antibacterial activity (363) (Fig. ). Recent analyses by Baell and Holloway showed a number of promiscuous, panassay-interfering (PAIN) compound classes (18), among which rhodanines were prime culprits. While none of the Wyeth compounds appear to match Baell and Holloway's list of PAIN compounds, it seems reasonable to require further evidence that the inhibition seen in assays of a set of enzymes is selective and not a reflection of promiscuity of the compound class.

The best pieces of evidence for MurB to -F inhibitors having any target-related activity in whole cells are the MurF inhibitors of Baum et al. (31, 32). The 4-PP compound (32) and the diarylquinoline DQ1 (Fig. ) (31) were shown to moderately inhibit E. coli MurF (IC50, 24 to 29 μM) and to cause an intracellular build-up of the pool of UDP-muramyl-l-Ala-d-Glu-m-DAP, the MurF substrate, and a decrease in the UDP-muramyl-pentapeptide product, as would be expected for inhibitors of MurF. Treatment with DQ1 eventually led to cell lysis. Overproduction of MurF led to normalization of pools but did not affect the MIC of DQ1. Thus, these inhibitors can enter the cell (permeable E. coli) and exert an inhibitory effect on MurF, but the antibacterial effect was not shown to be due exclusively or at all to MurF inhibition. This is in contrast to the potent benzothiophenyl-(morpholine-4-sulfonyl)-benzamide inhibitor (Fig. ) of Streptococcus pneumoniae MurF (IC50 = 22 nM) described by Stamper et al. (351), which had no activity on permeabilized E. coli or S. aureus. Unfortunately, no MIC testing on S. pneumoniae was reported. Was the lack of activity due to a lack of cell entry or to some other factor?

The fact that the MurB to -F enzymes lack validation (as useful antibacterial targets) with inhibitors, even though they have been shown genetically to be essential, is curious and has been commented on in the literature (141, 336, 351, 368). One speculative possibility, against which there is no clear evidence, is that the action of the pathway is concerted, perhaps performing as a multienzyme complex with channeling of intermediates, the active site(s) being inaccessible to inhibitors. Other possibilities include the lack of a rate-limiting step within that part of the pathway or upregulation of the pathway by inhibition of one of the steps, either of which might lead to the necessity for very strong inhibition of the cellular enzyme, perhaps by 99% or more. Presumably, irreversible and/or covalent inhibitors would overcome these obstacles. While generally avoided in human drug discovery programs, covalent inhibitors (e.g., the β-lactams) have clearly been useful in the antibiotic field. Interestingly, in this regard, a treatise on leadlikeness and unlikeness (312) notes that such irreversible and covalent inhibitors would be considered false-positive and nonleadlike compounds in biochemical assays, in contrast to those found in biological assays (such as the β-lactams).

Targets of Inhibitors Discovered by Enzyme Screening or Design

Table intends to display the stages of compound validation reached for leads from programs based on discovery and optimization of single-enzyme inhibitors as antibacterials. The table covers programs reported during the past 10 years and is organized by the enzyme targeted. The compounds (or series of compounds) are categorized as in vitro enzyme inhibitors (i) that have no reported antibacterial activity; (ii) where antibacterial activity is seen, but for which no further correlation than SAR has been shown; (iii) for which some phenotypic and/or genetic evidence has been accumulated to link inhibitor and target but which have not eliminated other target possibilities; (iv) that are likely to affect the putative target intracellularly, but for which there is a high probability that another target (or nonspecific activity) exists; and (v) whose enzyme inhibition is well validated to be the sole antibacterial mechanism by combinations of critical genetic, biochemical, and phenotypic means. The classes of validated inhibitors that have reached clinical trials are noted in the table and are discussed in a later section.

Forty of these programs and leads arose from enzyme-targeted high-throughput screens. Fifteen compounds or series (marked with a dagger in Table ) resulted from synthesis of inhibitors based on known substrates (without other structural input) or optimization of previously discovered leads, most of those from empirical screening. Eleven compounds or series (marked with a double dagger in Table ) were based on structural studies of six targets, using in silico docking or similar methods for virtual screening followed by actual assays of selected compounds (SBDD). This list is certainly incomplete, despite serious effort. Any selection bias was toward compounds that had antibacterial activity with some degree of validation that the MIC was due to inhibition of the targeted enzyme in vitro.

UppS.

Among the earliest-stage target/inhibitor programs of interest are the tetronic acid inhibitors of UppS, the last enzyme in the bactoprenol (undecaprenol) pathway, which catalyzes double-bond formation during condensation of 8 isopentenyl-PP and 1 farnesyl-PP subunit to form bactoprenol-PP. A tetramic acid enzyme inhibitor was discovered by HTS and virtually docked in the enzyme active site to generate a pharmacophore model. Tetramic, tetronic, and carboxamide analogs (Table ) and derivatives of the lead were synthesized and tested for inhibition of UppS, human farnesyl-PP (FPP) synthase (FppS), Enterococcus faecalis, S. aureus, and S. pneumoniae. The best antibacterial activity, the least FPP inhibition, and a reasonable correlation of enzyme inhibition and S. pneumoniae MIC were seen with the tetramic acids, as exemplified in Fig. . More potent enzyme inhibitors appeared to have less antibacterial activity, most likely due to a lack of penetration. The mechanism of enzyme inhibition was explored using a tetramic acid probe and showed that it binds to an allosteric site near the FPP binding site (but does not bind to FPP-bound enzyme), changing enzyme conformation and preventing FPP binding (207). This illustrates that such docking studies with leads followed by in silico optimization can yield improved compounds and validate the in vitro inhibitor-target interaction. This work is promising and appears to be ongoing. While it is likely that the antibacterial activity seen is due to enzyme inhibition, substantiation of specificity is required, notably because many tetramic and tetronic acids are metal chelators and/or have cytotoxic properties (16).

Open in a separate window

WalK/WalR.

WalK/WalR (YycG/YycF) is a histidine kinase/response regulator two-component system essential in low-GC Gram-positive organisms. The system was reviewed recently (100). Both components are essential in S. aureus, B. subtilis, and E. faecalis; only WalR is essential in S. pneumoniae. WalK appears to be deletable in Streptococcus pyogenes (where it is called VicK) (216), and there are conflicting reports on WalK/WalR essentiality in Streptococcus mutans. The system appears to be a regulator of peptidoglycan synthesis. Through structure-based virtual screening targeting the autophosphorylation activity of WalK, a set of 76 compounds were chosen and tested for antibacterial activity against Staphylococcus epidermidis (305). Seven compounds had antibacterial activity, and six of these bound to WalK. The compounds were from 4 structural classes and showed antibacterial activity against strains of S. epidermidis, S. aureus, S. pyogenes, and S. mutans, in roughly the same rank order. Enzyme-inhibitory activity for the 6 enzyme binding compounds was generally correlated with the MIC against S. epidermidis. The compounds had low cytotoxicity and showed low hemolytic activity. The finding of significant antibacterial activity with this set of compounds is quite surprising—since with many nonvirtual screening efforts, enzyme inhibition and antibacterial activity are seldom correlated (see Chemistry below). Even with the good correlation of enzyme inhibition and antibacterial activity (across 3 structural classes), more data are required to identify WalK as the antibacterial target. As noted above, it seems that WalK is nonessential in S. pyogenes, so the activity against S. pyogenes, especially of the compound with the broadest spectrum (the thiazolidinone shown in Fig. ), may indicate that there are other targets being inhibited.

Many other programs that have produced inhibitors whose mechanism of antibacterial action has not yet been validated (shown to be due to inhibition of the targeted enzyme in vitro) are listed in columns 3 through 6 of Table . Rather than a continued dissection of these programs, a discussion of 2 targets, LpxC and FtsZ, with validated inhibitors follows.

LpxC.

LpxC is the second enzyme and the first committed step of the essential lipid A synthetic pathway, which is present only in Gram-negative organisms. A hydroxamic acid inhibitor of LpxC (L-573,655 in Table ) was first discovered at Merck by phenotypic screening for inhibitors of lipopolysaccharide (LPS) synthesis and later determined to be a specific inhibitor of LpxC. A program based on L-573,655 found more potent enzyme inhibitors of the E. coli enzyme with improved antibacterial activity against E. coli but no activity against Pseudomonas aeruginosa, with the best being L-161,240 (Fig. ), which was efficacious in vivo against E. coli septicemia (283). Resistance to L-573,655 and L-161,240 was seen at a frequency of 10−9; two of four mutants sequenced contained mutations in lpxC (283; N. Rafanan, S. Lopez, C. Hackbarth, M. Maniar, P. Margolis, W. Wang, Z. Yuan, R. Jain, J. Jacobs, and J. Trias, presented at the 40th Interscience Conference on Antimicrobial Agents and Chemotherapy). Programs at Versicor, British Biotech (later Oscient), and Chiron (later Novartis) followed that at Merck, all yielding compounds with hydroxamate groups necessary for coordination with zinc in the active site (76, 136, 174, 247). Activity was restricted to Gram-negative organisms, as expected for lipid A inhibitors. BB-78485 (Table ; Fig. ) selected for resistance at two sites, lpxC and fabZ (76). FabZ is required for synthesis of the fatty acids which are attached to lipid A and which are substrates of LpxA, the first step of the pathway. Presumably, overproduction of FabZ overcomes inhibition downstream in the pathway (260). The Chiron inhibitor CHIR-090 (Table ; Fig. ) was optimized for activity against P. aeruginosa and is a potent enzyme inhibitor and broad-spectrum Gram-negative antibacterial (247). A patent application (M. A. Siddiqui, U. F. Mansoor, P. A. Reddy, and V. S. Madison, U.S. patent application 2007/0167426 A1) from Schering-Plough (now Merck) covers a large number of hydroxamates with some similarity to the Chiron compound. Since the absorption of the originating companies, the fate of these specific inhibitors is not known, but it is clear that several companies have continued to work on this target. For example, the Achaogen website lists an LpxC preclinical program.

Other approaches to discovery of LpxC inhibitors are the substitution of hydantoin for hydroxamate as the zinc binding group (Table ; Fig. ), as described previously (80; U. F. Mansoor, P. A. Reddy, and M. A. Siddiqui, WIPO patent application 2008/027466 A1, 5 March 2008), and the exploration of the use of a lipophilic tail on a simple benzoic acid (Fig. ), a structure which lacks a metal binding group but has significant (2.3 μM) activity (331). Another avenue is exploration of uridine-based compounds as potential UDP analogs of the substrate (25). No antibacterial activity has yet been shown for these. Schering scientists recently described a novel LpxC screening strategy and noted that a large group of active compounds were selected (∼286 with IC50s of <5 μM) from over 700,000 compounds screened (205), but no structures were shown. Six compounds were tested for antibacterial activity, and 5 of 6 were shown to be more active on an lpxC mutant strain (which expresses a reduced amount of LpxC activity) than on its isogenic wild-type parent. As noted by the authors, this is to be expected for inhibitors of LpxC but would also be seen for compounds otherwise excluded from E. coli by the outer membrane (OM) (400).

Clearly, the target is of interest for its importance in Gram-negative organisms, especially P. aeruginosa, which are increasingly problematic in the clinic due to growing resistance. Work on the LpxC inhibitors appears to be continuing, and there are a number of drug-like leads. Will FabZ upregulation be able to suppress potent inhibition of LpxC? Once again, the question of in vitro resistance selection of single-target inhibitors arises. An LpxC inhibitor might be a good candidate for combination therapy with a current or future drug.

FtsZ.

FtsZ, a protein required for cell division in most, if not all, bacteria, has been a popular target (217), with a number of reported inhibitors (Tables and ). Haydon and colleagues at Prolysis/Biota recounted the path of optimization of inhibitors (150) based on 3-methoxybenzamide (3-MBO) (Fig. ), a known inhibitor of ADP-ribosylase, whose antibacterial activity was reversed by mutations in ftsZ (277). This account describes an excellent antibacterial discovery program, with a variety of parameters tracked throughout in order to progress to a drug-like molecule (Fig. ) with excellent pharmacokinetic and acceptable pharmacological properties, including oral bioavailability and low cytotoxicity. The best compound, whose human plasma protein binding was reduced to ∼90% from the 99.9% value of a related compound, was efficacious in a mouse model of S. aureus septicemia. The compound is not sufficiently active for clinical candidacy, but the lead has validated ftsZ as an effective antibacterial target in vivo and is suitable for further optimization. And what of resistance? Notably, resistance selection was carried out on selected compounds at each step of optimization in order to monitor on-target activity. In all cases, a mutant mapping in ftsZ was identified (though no resistance frequency was indicated). Will the program go forward despite in vitro resistance selection?

TABLE 4.

Target category and targetInhibitor(s)b discovered by:Empiric screeningPhenotypic screeningDNA replication and substrates    GyrANalidixic acid (137)Pyrazole derivatives (360)    GyrB (276)Novobiocin, coumermycin, clorobiocin, cyclothialidine    NdkDesdanine (318, 330)Cell wall synthesis and cell division    MurAFosfomycin (155, 181)    IspCFosmidomycin (202, 279)    WalK/WalRAranorosinol B (383)    SAV1754 (flippase? MurJ?)Compound D (263)DMPI, CDFI (166)    FtsZSanguinarine (37), curcumin (307)Transcription, translation, and chaperones    RNA polymerases (72)Rifamycins, streptolydigin, lipiarmycin, sorangicin    tRNA synthetases (RS) (73)        IleRSMupirocin        TrpRSIndolmycin        PheRSOchratoxin        ThrRSBorrelidin        LeuRSGranaticin        ProRSCispenticin    Ef-G (73)Fusidic acid    Ef-Tu (291)Kirromycin, enacyloxin, pulvomycinGE2270 (327)    PdfActinonin (68)Lipid and membrane synthesis    FabFPlatensimycin (381)    LpxCL573,655; L-161,240 (283)    AccCPyridopyrimidine (255)Open in a separate window

Inhibitors Identified after Phenotypic and Empirical Discovery

The discussion thus far has dealt with inhibitors discovered via screening for in vitro inhibition of isolated enzymes. Both phenotypic screening and empirical screening have played large parts in antibacterial discovery efforts, and lately, both have been revived, with the general failure of enzyme screening and the advent of relatively powerful methods of target identification (the array of arrays mentioned above). A number of single-enzyme inhibitors found through phenotypic and empirical screens throughout the years are noted in Table . Most of the listed natural product inhibitors are antibacterial due to specific enzyme inhibition (except in the case of the plant-derived compounds inhibiting FtsZ, which may have other mechanisms). In contrast, as illustrated in Table and discussed in Chemistry, very few enzyme inhibitors found in chemical collections have demonstrated antibacterial activity due to selective and specific inhibition of the putative enzyme target. Some exceptions, where chemical collections did yield compounds with antibacterial activity due to specific enzyme inhibition, are actinonin, a natural product that was present in chemical collections and discovered as a specific inhibitor of peptidyl deformylase (Pdf) (68), and a pyridopyrimidine compound found by empirical screening of a library of kinase inhibitors, which was found to be a specific inhibitor of the biotin carboxylase subunit of acetyl-coenzyme A (acetyl-CoA) carboxylase, AccC (255).

Phenotypic screening could be used profitably with chemical libraries, as it has the benefit of finding hits with whole-cell activity due to a relatively specific mechanism. This has been seen with the compounds DMPI (3-{1-[(2,3-dimethylphenyl)methyl]piperidin-4-yl}-1-methyl-2-pyridin-4-yl-1H-indole) and CDFII {2-(2-chlorophenyl)-3-[1-(2,3-dimethylbenzyl)piperidin-4-yl]-5-fluoro-1H-indole}, detected in a screen for synergists of a carbapenem against methicillin-resistant S. aureus (MRSA) (166). Their mechanism of action was discovered, in part, through use of an antisense hypersensitization array which implicated S. aureus open reading frame (ORF) SAV1754 as a possible target (96). SAV1754 shares homology with murJ of E. coli, which has been implicated as a lipid II flippase in cell wall synthesis (316). The compounds did indeed inhibit cell wall synthesis, and resistant mutants mapped in the indicated ORF.

Single-Enzyme Targets of Novel Inhibitors in Clinical Trials

Inhibitors of four single-enzyme targets became clinical candidates in the period from 2000 to the present. All of these targets were known to have essential bacterial functions by microbial genetic analysis before the “genomics” era. The status of these programs and of the development of specific compounds is not clear in some cases.

Peptidyl deformylase.

The history of Pdf as an antibacterial target has been reviewed recently (143, 209, 401). A unique feature of bacterial translation is that peptide chains are initiated with formylated methionine (fMet), which is synthesized by a transformylase (Fmt) acting on initiator fMet-tRNA charged with unformylated methionine. The enzyme which removes the N-terminal methionine from a number of bacterial proteins, methionyl-aminopeptidase (Map), is essential (62). Since deformylation of the fMet must be carried out by peptidyl deformylase (Pdf), before demethylation by Map, deformylation too is an essential function. The stabilization, purification, and characterization of Pdf, both biochemically and genetically, led to its proposal as an antibacterial target (245). This was followed by a rush to find inhibitors of the enzyme by most antibacterial drug discovery groups. Enzymologists, biochemists, structural biologists, and medicinal chemists were very optimistic about this target, but (by informal survey) microbial geneticists were not, since the original genetic characterization of the essentiality of Pdf also showed that cells lacking Fmt could survive the loss of Pdf. In that case, a nonformylated pathway (similar to that used by eukaryotes) was used for translation initiation. Such pdf-fmt deletions of E. coli grew very slowly (245), but faster-growing variants overgrew the cultures. Thus, it appeared that the requirement for Pdf could be bypassed, and even with reduced fitness, compensatory events appeared to improve that fitness. The genetic prediction is that inhibitors of Pdf would lead to a similar situation.

When such inhibitors were discovered (13, 66, 68, 145, 261, 401), including the previously mentioned natural product antibacterial, actinonin (Fig. ) (68), resistance to inhibitors arose as expected, at a high frequency (10−7 in permeable E. coli [13]), due to mutations in the fmt gene. Fmt loss-of-function mutations were found in other species, including S. aureus, Haemophilus influenzae, and P. aeruginosa. In S. pneumoniae, mutations of the target pdf gene were found at a low frequency, as would be expected for specific target-based “gain-of-function” resistance (209). In a detailed study of actinonin resistance in Salmonella enterica serovar Typhimurium LT2, 31 loss-of-function mutations were mapped to either the fmt or folD gene. (FolD is required for synthesis of 10-formyl-tetrahydrofolate, the formyl donor used by Fmt.) Compensatory mutations for a number of these mutants, all of which grew slowly, were selected by serial passage in rich medium until faster-growing cells dominated the populations. Intragenic suppressors of the original folD and fmt resistance mutations were identified which led to actinonin susceptibility. Interestingly, compensatory mutations leading to amplification (5- to 40-fold) of the genes encoding initiator fMet-tRNAs led to higher growth rates and retained actinonin resistance (273).

Open in a separate window

Pdf inhibitors BB83698 (Oscient) (309) (Fig. ) and LBM415 (Novartis) (209) (Fig. ) were optimized to meet criteria for clinical development (143). The phase I single-ascending-dose trial of BB83698 proceeded without problems and appeared successful (75). However, both development programs were stopped in 2004. Reasons for this are not clear, but concerns about resistance development were likely important, as was the finding of a human mitochondrial homolog of Pdf that is inhibited by inhibitors of the bacterial enzyme (208, 209). Furthermore, both of the development candidates and most other reported inhibitors had hydroxamic acid moieties as the metal chelating warhead. Hydroxamates are generally avoided by medicinal chemists for their possible promiscuity, toxicity, and instability/metabolism (117, 148, 312); however, hydroxamates have reached the market, including a histone deacetylase inhibitor licensed for therapy of certain cancers, i.e., vorinostat (233). Recently, efforts at optimizing Pdf inhibitors for therapy of M. tuberculosis, where single-target inhibitors are the norm, have been reported (298, 329).

Enoyl-reductases of FAS II.

The enoyl-reductases of bacterial fatty acid synthetase II (FAS II) and their inhibitors have been reviewed recently (222, 243). Triclosan (Fig. ) was initially thought to be a nonspecific antiseptic but was later found to target the enoyl-reductase FabI (250). With the advent of genomically driven drug discovery programs, the broadly distributed FabI enzyme was targeted in a number of programs. In the same time frame, it was discovered that certain pathogens either lacked FabI or contained FabI in addition to another nonhomologous enoyl-reductase (243). For example, S. pneumoniae uses FabK instead of FabI (152), while E. faecalis and Enterococcus faecium have both FabI and FabK (152). P. aeruginosa and Vibrio cholerae contain FabV in addition to FabI (244, 405). FabL and FabI are present in B. subtilis (153). It is interesting that FabK, FabL, and FabV encode triclosan-resistant enzymes, which Heath and Rock (152) noted would have implications for drug discovery. For broad-spectrum Gram-positive activity, such an inhibitor would have to target both FabI and FabK with similar potencies (295). Indeed, the FabI program at GSK, which was initially directed toward respiratory pathogens (293), discovered potent aminopyridine inhibitors of the S. aureus enzyme, but their lack of useful activity against S. pneumoniae, explained by poor activity against FabK, led the company to outlicense the program to Affinium (294). Most of the GSK inhibitors (Fig. ) were shown to target FabI by overexpression of the enzyme in S. aureus leading to raised MICs and by specific inhibition of lipid synthesis over other MMS in S. aureus (258, 294). In later work, antisense depression of expression of FabI in S. aureus was shown to sensitize cells specifically to the GSK FabI inhibitors (177). The accretion of SAR, biochemical, and genetic evidence along with the lack of activity against S. pneumoniae supports the specific antibacterial mechanism of FabI inhibition, although resistant mutants were not sought. Interestingly, the most potent inhibitor of the GSK aminopyridine series (294) had activity against S. pneumoniae in addition to S. aureus, was found to inhibit FabK as well as FabI, and had relatively nonspecific MMS results, and its S. aureus MIC was not raised by FabI overexpression. This illustrates the need to characterize the mechanism of whole-cell activity across a series of inhibitors, especially when activity is increased or altered significantly.

The Affinium clinical candidate AFN-1252 (Fig. ), which is closely related to its GSK progenitors, is staphylococcus specific in spectrum and has no demonstrable S. pneumoniae or S. pyogenes activity (184). No data on resistance selection by the GSK or Affinium FabI inhibitors are available. A different FabI inhibitor, {"type":"entrez-nucleotide","attrs":{"text":"CG400549","term_id":"34399433","term_text":"CG400549"}}CG400549 (Fig. ), which is based on triclosan and whose MIC has been shown to be raised by FabI overexpression, has been shown to select for resistance in a single step at low frequencies (290). The resistance mutations occurred at a single site in fabI in 10 of 13 mutants tested. The Mutabilis FabI inhibitor, MUT37307 (Fig. ), an aryloxy-phenol inhibitor based on triclosan, is highly active against staphylococci and has activity only against organisms containing solely FabI (92). Resistance was found at 10−8 at 4 times the MIC (106).

AFN-1252 has been in phase I clinical trials, and Mutabilis announced it would have a FabI inhibitor candidate (most likely MUT37307) in the clinic in 2009 (125), but no results have been announced. While company websites indicate that the programs are ongoing, the following findings may influence further development of these and other FAS II inhibitors. Recently, data were reported that may undermine the validity of FAS II enzymes as Gram-positive drug targets (46). This controversial work showed that Streptococcus agalactiae and S. aureus appear to be able to take up sufficient unsaturated fatty acids from human serum to obviate the essentiality of FAS II enzymes in vivo. The authors caution that the uptake of fatty acids may require time for adaptation of the infecting organisms to growth (and fatty acid uptake) in serum; hence, animal models of efficacy of FAS II inhibitors may overestimate the therapeutic effect unless they are carefully designed to allow sufficient adaptation time and organismal load. A counterargument to that paper (20) and a reply by the original authors (47) appeared even more recently. Further elucidation is needed, but the preliminary report of a fabI deletion mutant of S. aureus that can be maintained on an exogenous source of fatty acids (47) seems to militate against development of FabI inhibitors for S. aureus (or other Gram-positive?) infections. In vivo models were used to show the efficacy of the FabI inhibitors described above. AFN-1252 is active in a murine subcutaneous abscess model (384), the similar GSK inhibitors are active in a rat groin abscess model (294), and MUT37307 is active in methicillin-susceptible S. aureus (MSSA) and MRSA septicemia and thigh abscess in a murine model (106). While not all infections involve growth in the presence of serum and long-chain fatty acids may not be sufficient for supplementation at some infection sites, it should be noted that the free linoleic and oleic acid concentrations in staphylococcal abscesses are 1.2 mg/ml (4.2 mM) and 0.9 mg/ml (3.2 mM), respectively (calculated from data in reference 332), similar to those in whole serum (46).

Thus, the FabI inhibitors illustrate a number of problems that target-directed antibacterial discovery programs have encountered: the occurrence of nonhomologous enzymes with similar (complementing) activities in important pathogens can narrow the ultimate spectrum, analysis of fitness of a target by genomic and in vitro methods may miss the inappropriateness of the target in the infected host, and resistance occurs readily in vitro.

Leucyl tRNA synthetases of Gram-negative organisms.

Anacor Pharmaceuticals has been exploring boron-based chemistry for systemic antibacterials in a program partnered with GSK. First announced at a recent meeting, Anacor disclosed a prototype benzoxaborole, ABX (Fig. ), that is reported to be a specific inhibitor of the editing domain of leucyl tRNA synthetase (LeuRS) with excellent activity against Gram-negative organisms (123). ABX was shown to be active in vitro against multidrug-resistant (MDR) Gram-negative strains of Enterobacteriaceae (except for Proteeae), P. aeruginosa, and Stenotrophomonas maltophilia, with an MIC90 of 1 μg/ml, and against Burkholderia cepacia (MIC90 of 4 μg/ml), but it had little activity (MIC50 of >128 μg/ml) against Acinetobacter spp. In a neutropenic mouse thigh abscess model, ABX gave 3.5-log killing of P. aeruginosa with a dose of 30 mg/kg subcutaneously or 100 mg/kg dosed orally. No data on resistance selection or details of validation of leuRS as the antibacterial target of the compounds have been reported. Soon after the first disclosure of the program, a November 2009 news release reported that a clinical candidate, AN3365, had entered phase I trials. Presumably, it is related to ABX, but its structure and activity have not yet been disclosed; it is likely to be covered in the same patent application as ABX (S. J. Baker, V. S. Hernandez, R. Sharma, J. A. Nieman, T. Akama, Y.-K. Zhang, J. J. Plattner, M. R. K. Alley, R. Singh, and F. Rock, U.S. patent application 2009/0227541). The fact that this compound series entered the clinic so soon after its first disclosure is exciting.

RNA polymerase in C. difficile.

Although it is neither a novel target nor a newly discovered agent, nevertheless fidaxomicin is a single-target inhibitor in the late stage of clinical development (2 phase III trials have been completed at this time). Fidaxomicin is the generic name of an Actinoplanes product, variously known as lipiarmycin (Table ), tiacumicin B, and difimicin, discovered in 1975 (289). It is a specific inhibitor of transcription by RNA polymerase (328, 344), and in B. subtilis, it selects for resistant mutants mapping in rpoC (at a frequency of 2 × 10−6 in one report [344] and <10−7 in another [although 2 such mutants were indeed selected and mapped] [142]). Such mutants have MICs that are raised ≥16-fold. In the latter report (142), mutations leading to reduced permeability (or efflux?), as indicated by effects on MICs of other antibiotics, were also found. Fidaxomicin is highly active on Clostridium difficile but not on enteric Gram-negative organisms, including anaerobes. The frequency of resistance selection in C. difficile was reported to be <2 × 10−8 at 4 or 8 times the MIC (356). Fidaxomicin has a selective Gram-positive spectrum (somewhat narrower than that of vancomycin [114]) that should lead to minimal disruption of gut flora, which is desirable for treatment of C. difficile. It has thus been under development for C. difficile-associated diarrheal disease (CDAD) by Optimer Pharmaceuticals. No resistant isolates have been seen in clinical trials (134). It should be noted that levels of drug are very high in the gut, as the compound is not absorbed, reaching average levels of 104 times the MIC (325), which would be expected to suppress selection of resistant mutants. It appears to be effective in phase III trials, with relapse rates significantly lower than that of vancomycin (133).

DHFR and iclaprim.

Although iclaprim (Fig. ) is not a member of a novel class of agents, but rather an optimized trimethoprim (Fig. ), it is an example of a single-target inhibitor specifically designed to overcome target-based resistance to its progenitor. The derivatization project was begun by Roche and spun off to Arpida, who continued research and development. The added rings of iclaprim allow interaction of the compound with two binding sites different from the site altered in the dihydrofolate reductase (DHFR) target enzyme with resistance to trimethoprim (point mutation yielding a Phe98-to-Tyr98 change in the S. aureus enzyme). This change in DHFR is a main form of trimethoprim resistance in Gram-positive organisms. Thus, while trimethoprim has a low affinity for the resistant form of DHFR, iclaprim retains high affinity for the enzyme because of its additional binding sites (although it too loses affinity for the altered site) (149). Indeed, it has been reported that iclaprim is active against trimethoprim-resistant Gram-positive strains and that iclaprim does not select for resistance in a single step from wild-type or trimethoprim-resistant starting strains (frequencies of <10−10, spontaneously in S. aureus or with UV treatment in E. coli). It also appears to be very slow to yield resistance upon serial passage (little change in MIC after 15 passages) (149). While iclaprim is active against wild-type E. coli and does not yield to endogenous resistance upon selection, MIC90s for E. coli and other Enterobacteriaceae are quite high (>8) (179, 386). This is likely due to the fact that trimethoprim resistance in the enterics can be conferred by plasmids carrying many types of resistant DHFR enzymes, as opposed to Gram-positive organisms, where resistance is generally due to endogenous mutation (170). The activity of iclaprim against strains of Enterobacteriaceae with defined trimethoprim resistance mechanisms has not been reported.

Clinical development of iclaprim was begun for treatment of complicated skin and skin structure infections (cSSSI) as well as hospital-acquired pneumonia (HAP), ventilator-acquired pneumonia (VAP), and CAP. Unfortunately, iclaprim was not approved for use in SSSI in either the United States or the European Union because of a failure to meet noninferiority standards. In late 2009, Arpida, financially unable to support further trials, sold the rights for iclaprim to Acino Pharma, which plans to continue development on a limited scale for parenteral treatment of selected pneumonias.

This type of approach, the specific design of single-enzyme inhibitors to overcome target-based resistance by iterative synthesis, resistance selection, crystallography, and redesign, might provide a light at the end of the single-target dilemma if it can be applied to novel compounds (before reaching the clinic) on the basis of resistance patterns seen in vitro or in animal models.

The concept of adding chemical moieties providing extra intramolecular binding sites in order to overcome preexisting exogenous target-based resistance has indeed been applied in the development of other compounds (even those having complex targets), such as the ketolides, telavancin, glycylcyclines, cephalosporins active against MRSA (such as ceftaroline), and newer oxazolidinones. This approach has been discussed in some detail in several reviews (5, 64, 97, 337).

Summary of Challenges of Single-Target Discovery

To summarize, single-target screening and design, which have been the major efforts in antibacterial discovery for at least 15 years (the start of the genomic era), have had little success thus far. Although inhibitors for many targets have been discovered in this period, only a few new single-target agents have reached the clinic, with at least 2 candidates currently viable. It is likely that most of the programs presented in Table have ended, for unknown reasons. Indeed, in the industry, clinical candidates often appear “fully formed” in the literature or at meetings (although often presaged in the patent literature), as was the case with linezolid (45) and the new benzoxaborole Leu-tRNA synthetase inhibitors (123). A variety of problems that may have contributed to cessation of programs are in evidence. Since for many published compounds enzyme inhibition was not proven to determine the MIC, later (unpublished) work might have shown otherwise, for example, that the inhibitors were promiscuous or nonspecifically lytic or toxic. The nature of the chemical library may have been “at fault,” as discussed below. Perhaps the inhibitors were nonselective, hitting mammalian targets and yielding mechanism-based toxicity. It is likely that other factors played a part in preventing program progression as well, such as intractable low potency or solubility, a narrow spectrum, or high protein binding levels. It remains probable, however, that many of the programs that led to validated inhibitors were terminated because, where investigated, these inhibitors all were subject to selection of resistance due to a single mutational event.

What is the future of antibacterial discovery based on single-target agents? Screening and, no doubt, increased SBDD efforts will continue, but these should be carried out with due attention paid to whole-cell activity and bacterial entry (discussed below) and the potential for resistance development. Once inhibitors are designed and/or optimized and whole-cell activity due to inhibition of the putative enzyme target demonstrated, then SBDD can be used to analyze resistant enzymes and to iteratively redesign them to overcome that resistance.

Tylosin (Tylan powder) 500 mg for chickens ● poultry ● pigs ● dogs

Tylosin is an antibiotic typically used to treat bacterial infections in farm animals, but veterinarians often use it to treat certain types of chronic diarrhea in cats and dogs. The powder form is not FDA approved for use in companion animals, but it is common practice for veterinarians to prescribe this medication. The drug is also given as an injectable type.

It is prescribed for the prevention and treatment of respiratory mycoplasmosis of birds, infectious sinusitis of turkeys, as well as the treatment of dysentery and gastroenterocolitis of bacterial etiology of pigs, bronchopneumonia of calves caused by pathogens sensitive to tylosin. To combat necrobacteriosis in cattle and small ruminants.

If you want to learn more, please visit our website China 2-Bromo-4′-cyanoacetophenone factory.