THE FUTURE CHALLENGES FACING THE DEVELOPMENT OF NEW ANTIMICROBIAL DRUGS

THE FUTURE CHALLENGES FACING THE DEVELPMENT F NEW ANTIMICRBIAL DRUGS Anthony Coates*, Yanmin Hu*, Richard Bax and Clive Page The emergence of resistance to antibacterial agents is a pressing concern for human health. New drugs to combat this problem are therefore in great demand, but as past experience indicates, the time for resistance to new drugs to develop is often short. Conventionally, antibacterial drugs have been developed on the basis of their ability to inhibit bacterial multiplication, and this remains at the core of most approaches to discover new antibacterial drugs. Here, we focus primarily on an alternative novel strategy for antibacterial drug development that could potentially alleviate the current situation of drug resistance targeting non-multiplying latent bacteria, which prolong the duration of antimicrobial chemotherapy and so might increase the rate of development of resistance. ANTIMICRBIAL AGENTS This term includes antibiotics and chemically derived agents. *Department of Medical Microbiology, St George s Hospital Medical School, Cranmer Terrace, London SW17 RE, UK. BioTherapies Ltd, ctagon House, Fir Road, Bramhall, Stockport, Cheshire, SK7 2NP, UK. Biosyn, Inc., Building 13, 1800 Byberry Road, Huntingdon Valley, Pennsylvania 19006, USA. Sackler Institute of Pulmonary Pharmacology, GKT School of Biomedical Sciences, King s College London, London SE1 9RT, UK. Correspondence to A.C. e-mail: acoates@sghms.ac.uk doi:10.1038/nrd940 The market size for ANTIMICRBIAL AGENTS is now greater than US $25 billion per year. However, the global emergence of RESISTANCE T ANTIMICRBIAL AGENTS is increasingly limiting the effectiveness of current drugs. This review covers four approaches to the development of new systemic antimicrobial agents: classic screening; structural changes to existing agents; genome hunting; and a new route that targets non-multiplying, LATENT bacteria. To provide a background for the reader, elementary information about antimicrobial agents and the first three of these developmental approaches is included, but they are not covered in depth, and references to more comprehensive reviews are provided. The primary focus of this Review is the last, and relatively novel, approach targeting non-multiplying latent bacteria which is hoped to lead to new drugs that will reduce the rate of emergence of resistance to antimicrobial agents. For example, CLINICALLY LATENT BACTERIA prolong therapy, as they are not killed, or are killed only slowly or partially by existing antimicrobial agents. New drugs that target latent bacteria could reduce the duration of chemotherapy, and thereby probably reduce the rate of development of resistance as well as increase patient compliance, which would lower the rate of emergence of resistance further. In addition, these drugs might reduce the potential period of suboptimal therapy that is associated with the emergence of antimicrobial resistance, and could diminish the rate of emergence of chromosomal mutations that could lead to resistance in clinically latent bacteria. Background Before the introduction of ANTIBITICS in the 1940s and 1950s, patients who had bacteraemia for example, with Streptococcus pneumoniae had a low chance of survival 1, and the mortality from tuberculosis was 50% 2. Antibiotics radically changed this bleak prognosis, and new classes of antimicrobial agents rapidly entered the market in the 1950s and 1960s. Unfortunately, this led to over-confidence that infectious disease would be eradicated. In addition, the large costs of research, combined with difficulties in discovering new, broad-spectrum antimicrobial drugs with previously unexploited modes of action, discouraged pharmaceutical companies from this area, and many left the field. No new classes of antibiotics were produced in the 37 years 3 between the introduction of nalidixic acid in 1962 and linezolid in 2000; all of the antibacterial agents that entered the market during this period were modifications of existing molecules. To make NATURE REVIEWS DRUG DISCVERY VLUME 1 NVEMBER 2002 895

Table 1 Main classes of antibiotics Class β-lactams Penicillins Cephalosporins First generation Second generation Third generation Fourth generation Carbapenems Monobactams β-lactamase inhibitors Aminoglycosides Tetracyclines Rifamycins Macrolides Lincosamides Glycopeptides Streptogramins Sulphonamides xazolidinones Quinolones thers RESISTANCE T ANTIMICRBIAL AGENTS A microbe that survives, for example, treatment with an antimicrobial agent (at or above the MIC) by altering its genome is resistant to that drug. The progeny of that microbe will also be genetically resistant to the agent. LATENT Existing but hidden. CLINICALLY LATENT BACTERIA A hidden infection with a pathogen that might involve microbial growth, which is balanced by host control mechanisms, so that the infection remains below the threshold of infectious disease expression. Conversely, the pathogen might be nonreplicating. It is not usually possible to distinguish between replicating and non-replicating bacteria in vivo. Examples Penicillin G, penicillin V, methicillin, oxacillin, cloxacillin, dicloxacillin, nafcillin, ampicillin, amoxicillin, carbenicillin, ticarcillin, mezlocillin, piperacillin, azlocillin, temocillin Cepalothin, cephapirin, cephradine, cephaloridine, cefazolin Cefamandole, cefuroxime, cephalexin, cefprozil, cefaclor, loracarbef, cefoxitin, cefmetazole Cefotaxime, ceftizoxime, ceftriaxone, cefoperazone, ceftazidime, cefixime, cefpodoxime, ceftibuten, cefdinir Cefpirome, cefepime Imipenem, meropenem Astreonam Clavulanate, sulbactam, tazobactam Streptomycin, neomycin, kanamycin, paromycin, gentamicin, tobramycin, amikacin, netilmicin, spectinomycin, sisomicin, dibekalin, isepamicin Tetracycline, chlortetracycline, demeclocycline, minocycline, oxytetracycline, methacycline, doxycycline Rifampicin (also called rifampin), rifapentine, rifabutin, bezoxazinorifamycin, rifaximin Erythromycin, azithromycin, clarithromycin Lincomycin, clindamycin Vancomycin, teicoplanin Quinupristin, daflopristin Sulphanilamide, para-aminobenzoic acid, sulfadiazine, sulfisoxazole, sulfamethoxazole, sulfathalidine Linezolid Nalidixic acid, oxolinic acid, norfloxacin, pefloxacin, enoxacin, ofloxacin/levofloxacin, ciprofloxacin, temafloxacin, lomefloxacin, fleroxacin, grepafloxacin, sparfloxacin, trovafloxacin, clinafloxacin, gatifloxacin, moxifloxacin, sitafloxacin Metronidazole, polymyxin, trimethoprim matters worse, the number of scientists involved in basic research in this area was reduced. Bacteria exploited this window of opportunity by progressively developing resistance to all commonly used antibiotics, and each year the need for new antibiotics becomes more pressing 4. Now, every country in the world has antibiotic-resistant bacteria. In some regions, 25% of the most common cause of communityacquired pneumonia, S. pneumoniae, are resistant to penicillin 5, and more than 70% of bacteria that give rise to hospital-acquired INFECTINS in the United States resist at least one of the main antimicrobial agents that are typically used to fight infection 6. Until the recent development of new antimicrobial agents 7 9, the antibiotic vancomycin was the antibiotic of last resort for the treatment of multiple drug-resistant Staphylococcus aureus, and now the first case of fully vancomycin-resistant S. aureus has been reported 10. The scale of the problem is large, as in many Western countries, antimicrobial agents are prescribed by community doctors at the rate of 4 5 courses per person each year 7,8, and 20% of all patients who enter hospital have or acquire an infection 9. There is no obvious solution to this problem, although the development of new antimicrobial agents, the extension of life of current agents by improved educational methods, vaccination and other methods of disease control are being pursued. Recently, new classes of antimicrobial agents have entered clinical practice 4,11,12. New compounds arising from expanding research technology, such as genomics, might not be available for some years 11.However,the past record of rapid, widespread emergence of resistance to newly introduced antimicrobial agents indicates that even the new families of antimicrobial agents will have a short life expectancy. Response to the spread of resistant bacteria Basic research is central to the fight against the increasing threat of bacterial infectious disease. However, many universities have reduced the number of specialists in this field, particularly in antimicrobial resistance, resulting in few faculty members on whom to rebuild the research base 13,14. An understanding of bacterial physiology is the starting point, after which the powerful tools of molecular biology can be applied. The development and use of new antimicrobial agents is the most obvious way to combat the emergence of antimicrobial resistance. However, extending the life of current antibiotics could be achieved by more appropriate use of existing antimicrobial agents. For example, overuse of antibiotics to treat patients with sore throats, which are most often caused by viruses, and the treatment of the first attack of a middle ear infection in children over the age of two years, are areas in which the overall antibiotic use in the world could be reduced. Less use of antimicrobial agents would lead to a reduction in resistance and an extension to the effective life of these agents 15. Indeed, ~50% of antimicrobial agents are used for purposes other than treating human disease in particular veterinary medicine and agriculture and there has been increasing concern that such use is hastening the emergence of antibiotic resistance 16,17. The use of antibiotics as growth promoters in animal food is particularly controversial, and several antibiotics or their close relatives that are considered to be particularly important for human health, such as avoparcin, which is very similar to vancomycin, have been banned from use in animal foodstuffs. ther avenues of response include improved infection control 18. Clearly, in the hospital setting, where one in ten patients acquire an infection during their stay, better infection control could actually reduce the number of infections and so limit the need for antibiotics 19,20. The development of new vaccines is also being actively pursued. For example, the introduction of the Haemophilus influenzae type B vaccine has markedly reduced the incidence of infections in children due to this organism 21,22. However, some bacterial infections are more difficult to prevent by vaccination than others, and it is unlikely, in the foreseeable future, that vaccines will substantially reduce the world s consumption of antimicrobial agents. 896 NVEMBER 2002 VLUME 1 www.nature.com/reviews/drugdisc

R H NH H R S NH H S N N R CH H Penicillins Cephalosporins H H 2 N H H H H H 2 N H H H ANTIBITICS Naturally derived antimicrobial agents. INFECTIN The multiplication and growth of pathogens in host tissues or on host epithelia. Me Erythromycin A H H H H NMe 2 H H Tetracycline H 2 N H H 2 N H Tobramycin NMe 2 NH 2 NH 2 Me Me N F NH 2 Sulphomethoxazole Changes in public policy. The need for change has been gathering support, particularly in recent years. For example, in the United States, the 1999 Kennedy Frist Bill 23 identified resistance to antimicrobial agents as a major public health problem, leading to the establishment of an inter-agency task force. A report by a House of Lords Select Committee in the United Kingdom 13 also identified the potential danger to public health of the rise in resistance to antimicrobial agents. However, substantial new government funding for the fight against resistance has not materialized in either the United States or the United Kingdom, although given the state of basic research in this area, it could be argued that it is better to start small and grow in a few years time as more scientists enter the field. Alternatively, a better policy might be to increase funding now on the basis that it is difficult for referees to spot the winners early on, and by widening the funding net, the golden egg might be caught sooner rather than later. Concern about resistance has prompted the US Centers for Disease Control to start a campaign to target the problem by setting up a new internet site 24 N H H H Rifampin Me H 2 N Linezolid S Trimethoprim H N N NH N N NH 2 Figure 1 The chemical structures of selected important antimicrobial agents. Me H H NH N N N Me to increase awareness and to help provide ways to prevent antimicrobial resistance. Information on antimicrobial resistance is provided in a myriad of websites, an example of which is supplied in REF. 25. Drug regulation. The regulatory situation is not encouraging. Two years ago, the US Food and Drug Administration (FDA) began tightening the statistical standard that companies must meet to show the efficacy of antimicrobial agents in clinical trials. This means that trial size and duration must be significantly increased, thereby raising costs. Unfortunately, clinical trials at present do not precisely define the value of an antimicrobial agent, only that it is within preset statistical limits of non-inferiority. What is required is not tighter definitions of equivalence, but a change in the evolution of clinical development 26,27. Shifts in the FDA position on whether to grant new drug applications (NDAs) have lasting impacts on the willingness of major pharmaceutical companies to invest in research and development (R&D) of antimicrobial agents. Big companies will readjust their R&D programmes to concentrate on what they consider will give them a larger return of their investment; for example, lifestyle drugs. Conversely, small biotechnology companies can go out of business if their key product fails. Pharmaceutical industry response. The situation is bleak. Roche has left the field, GlaxoSmithKline, Bristol-Myers Squibb and Eli Lilly have downsized their activity and several biotechnology companies have either gone out of business or stopped doing research in this area 4,11,28. Paradoxically, the emergence of resistance to newly introduced antimicrobial agents and the realization that these drugs might become useless in a few years is one factor that discourages companies from developing new antimicrobial agents. However, several drugs each still earn more than US $1 billion per year, such as Augmentin (amoxicillin/clavulanate potassium; Glaxo- SmithKline), Cipro (ciprofloxacin hydrochloride; Bayer) and Zithromax (azithromycin; Pfizer). The main classes of antimicrobial agents are shown in TABLE 1, and several structures of important agents are shown in FIG. 1. Several antimicrobial agents are now under development. TABLE 2 shows recently licensed drugs and drugs in Phase I III trials. For example, quinpristin/ dalfopristin 8,9,11 reached the market in 1999, the carbapenem Invanz (ertapenem sodium; Merck) in 2001 and the ketolide Ketek 8 (telithromycin; Aventis Pharmaceuticals) in 2002. Many of these antimicrobial agents are specifically active against Gram-positive organisms; for example, linezolid, daptomycin, ramoplanin, oritavancin, dalbavancin, AZD2563, BAL5788 and AR-100. All the others are broad spectrum, also covering Gram-negative bacteria. f the twelve compounds not yet marketed, six are being developed by large companies, four are intravenous only and one is oral (faropenem). The first new type of drug to be launched in 35 years, Zyvox (linezolid; Pharmacia) 7,29, which is an oxazolidinone, reached the market in 2000. Unfortunately, resistance to linezolid has already started to emerge 30,31. NATURE REVIEWS DRUG DISCVERY VLUME 1 NVEMBER 2002 897

GRWTH Accumulation of biomass. How do antimicrobial agents work? Antimicrobial agents target essential components of bacterial metabolism, thereby disabling the bacteria 32,33 (FIG. 2). For example, the β-lactams, such as penicillin or cephalosporins, inhibit cell-wall synthesis 34. ther bacterial targets of antimicrobial agents are DNA gyrase 35 (quinolones), DNA-directed RNA polymerase 36 (rifampicin), protein synthesis 37 (macrolides, chloramphenicol, clindamycin, aminoglycosides, tetracyclines and oxazolidinones) and enzymes 38,39 (sulphonamides and trimethoprim). Some antibiotics, such as penicillin, are active against only a narrow spectrum of bacteria, whereas others, such as ampicillin, inhibit a broad range of Gram-negative and Gram-positive bacteria. The effects on bacteria can also differ. Some agents, such as tetracycline, inhibit bacterial GRWTH that is, they are bacteriostatic whereas others, such as penicillin, kill bacteria that is, they are bactericidal. Using a combination of antimicrobial agents can lead to increased activity compared with the activity of each antimicrobial agent alone; for example, trimethoprim and sulphamethozale are synergistic 38. However, care should be taken to use combinations that are not antagonistic. How do bacteria become resistant? Bacteria have lived on Earth for more than 3 billion years. During this time, they have encountered a wide range of naturally occurring antibiotics; for example, those produced by other microbes, such as Penicillium notatum, which makes penicillin 40,41. To survive, bacteria have developed a seemingly inexhaustible range of Table 2 Antibacterial drugs in clinical development Product Class Spectrum Phase IV/oral Indications Company (Licensor) Quinupristin/ Streptogramin Gram-positive Marketed IV VRE, SSTIs, blood-stream Aventis Dalfopristin (excluding E. faecalis) infections Moxifloxacin Fluoroquinolone Broad-spectrum Marketed IV, oral RTIs, SSTIs Bayer Gatifloxacin Fluoroquinolone Broad-spectrum Marketed IV, oral Community-acquired RTIs, Bristol-Myers Squibb, SSTIs, UTIs Grunenthal (Kyorin) Linezolid xazolidinone Gram-positive Marketed IV, oral RTIs, SSTIs, VRE blood- Pharmacia stream infections Telithromycin Ketolide Gram-positive, Marketed ral Community-acquired RTIs Aventis RTI pathogens Ertapenem Carbapenem Broad-spectrum Marketed IV CAP, intra-abdominal infection, Merck & Co (excluding non- acute gynaecological infections fermenters) (US and Europe), SSTIs, UTIs (US) Gemifloxacin Fluoroquinolone Broad-spectrum Phase III/ IV, oral Community-acquired RTIs, LG Chemical pre-reg SSTIs, UTIs Daptomycin Lipopeptide Gram-positive Pre-reg IV (oral SSTIs Cubist (Lilly) Phase I) Garenoxacin Des F(6) Broad-spectrum Pre-reg IV, oral Community-acquired RTIs, Bristol-Myers Squibb quinolone SSTIs, UTIs; intraabdominal infection (Phase III) Ramoplanin Glycolipo- Gram-positive Phase III ral, non- Prevention of VRE bacteraemia BioSearch Italia, depsipeptide absorbed in neutropaenic patients Genome Therapeutics Faropenem Penem Broad-spectrum Phase III ral RTIs, SSTIs Bayer ritavancin Glycopeptide Gram-positive Phase III IV SSTIs, HAP bloodstream InterMune (Lilly) infections ABT-773 Ketolide Gram-positive, Phase II/ IV, oral Community-acquired RTIs Abbott RTI pathogens Phase III Dalbavancin Glycopeptide Gram-positive Phase III IV SSTIs, bloodstream infections BioSearch Italia, Versicor Tigecycline Glycylcycline Gram-positive, Phase III IV SSTIs, HAP, CAP, intra- Wyeth (GAR 936) Gram-negative, abdominal infections UTIs, anaerobes VRE AZD2563 xazolidinone Gram-positive Phase I IV, oral RTIs, SSTIs, VRE AstraZeneca bloodstream infections BAL5788/ Cephalosporin Gram-positive Phase I/ IV SSTIs, HAP, catheter- Basilea Pharmaceutica BAL9141 (including MRSA), Phase II related bacteraemia (Roche) Gram-negative (and probably CAP) AR-100 Diamino- Gram-positive Phase I Unknown MRSA, RTIs, SSTIs Arpida Ltd (Roche) pyrimidine (including MRSA), RTI pathogens CAP, community-acquired pneumonia; E. faecalis, Enterococcus faecalis; HAP, hospital-acquired pneumonia; IV, intravenous; MRSA, methicillin-resistant Staphylococcus aureus; RTI, respiratory-tract infection; SSTI, skin skin structure infection; UTI, urinary-tract infection; VRE, vancomycin-resistant enterococci. Data obtained from G. A. Halls, Medical Marketing Services, 34 Ledborough Lane, Beaconsfield, Buckinghamshire HP9 2DD, UK. 898 NVEMBER 2002 VLUME 1 www.nature.com/reviews/drugdisc

Figure 2 Antibacterial drug targets. There are five main antibacterial drug targets in bacteria: cell-wall synthesis, DNA gyrase, metabolic enzymes, DNA-directed RNA polymerase and protein synthesis. The figure shows the antimicrobial agents that are directed against each of these targets. In the case of protein synthesis, aminoglycosides and tetracyclines inhibit 30S RNA, and macrolides, chloramphenicol and clindamycin inhibit 50S RNA. mrna, messenger RNA. antibiotic resistance mechanisms (FIG. 3) (for reviews, see REFS 33,42,43). So, it is not surprising that they quickly became resistant to all the antimicrobial agents that have been developed over the last 50 years 44. However, all antibacterial agents are not equal different agents induce resistance at different rates. Factors that determine whether resistance develops are complex and interdependent, and include mechanisms of action, whether the antibacterial drug is a concentrationor time-dependent killing agent, the potency against the population of bacteria and the magnitude and duration of the available serum concentration 45. For example, β-lactam resistance among group A streptococci has not developed so far. By contrast, other antimicrobial agents, such as rifampicin, readily select for resistance. It has been proposed 46 that antimicrobial agents such as rifampicin, which target single enzymes, are the most prone to the development of resistance, whereas agents such as penicillin, which inactivate several targets irreversibly, might generate resistance more slowly. As bacteria have encountered natural antibiotics, such as β-lactams and macrolides, in the environment, it is logical that resistance determinants to natural products will have evolved and will be disseminated by horizontal transfer. However, it was hoped that resistance to synthetic antimicrobial agents, such as fluoroquinolones and linezolid, would be slow to emerge, but unfortunately, resistance to synthetic agents has arisen quite quickly 30,31,47,48. It seems that once an antimicrobial agent is widely used in the human population, resistance at least in some species of microbe soon develops. Resistance to antimicrobial agents could be due to an innate property of the bacterium, or a consequence of mutation or gene transfer 49. Resistance often arises in the commensal whole bacterial pool, and the longer that suboptimal levels of antimicrobial agent are in contact with the bacteria, the more likely the emergence of resistance (FIG. 4). The main mechanisms of genetic resistance are shown in FIG. 3: namely, inactivation of the drug; modification of the site of action (enzyme, ribosome, cell-wall precursor); modification of the permeability of the cell wall; overproduction of the target enzyme; and the bypass of the inhibited steps 33,42,43,50 54. Unfortunately, most important infections contain non-multiplying bacteria that are tolerant to all known antimicrobials 55 60 (FIG. 5), although the molecular NATURE REVIEWS DRUG DISCVERY VLUME 1 NVEMBER 2002 899

Figure 3 Mechanisms of genetic resistance to antimicrobial agents. Bacteria have developed, or will develop, genetic resistance to all known antimicrobial agents that are now in the marketplace. The five main mechanisms that bacteria use to resist antibacterial drugs are shown in the figure. a The site of action (enzyme, ribosome or cell-wall precursor) can be altered. For example, acquiring a plasmid or transposon that codes for a resistant dihydrofolate reductase confers trimethoprim resistance to bacteria 52. b The inhibited steps can be by-passed. c Bacteria can reduce the intracellular concentration of the antimicrobial agent, either by reducing membrane permeability, for example, as shown by Pseudomonas aeruginosa 53, or by active efflux of the agent 54. d They can inactivate the drug. For example, some bacteria produce β-lactamase, which destroys the penicillin β-lactam ring 50,51 (FIG. 1). e The target enzyme can be overproduced by the bacteria. MULTIPLICATIN Genomic growth and segregation into a new self-propagating unit. mechanisms are not fully understood. Furthermore, non-multiplying bacteria can mutate in the presence of antimicrobial agents, become genetically resistant and then start to multiply in the presence of the agent 61. When faced with this array of resistance mechanisms, it is clear that a high level of ingenuity is needed in the development of new, effective antimicrobials. The development of new antimicrobial agents Classic screening. In 1929, Alexander Fleming 40 observed the inhibition of MULTIPLICATIN of bacteria on the surface of a solid agar plate around a contaminating fungus, which eventually led to the clinical introduction of penicillin in the 1940s. Since then, the inhibition of actively multiplying bacteria has been the assay with which potential antimicrobial agents are selected. Screening of large numbers of natural and synthetic compounds results in the selection of a lead compound, a process that we call the classic screening approach (for reviews, see REFS 62,63). Early antibiotics, which were developed by screening in the 1940s and 1950s and beyond, were all natural products, which have been a vitally important source of new antimicrobial agents. Recent interest has been rekindled in the screening approach by the discovery of naturally occurring antimicrobial peptides. These include peptides known as cathelicidins, which are found in mammalian skin and other tissues 64, and piscidins, which are found in the 900 NVEMBER 2002 VLUME 1 www.nature.com/reviews/drugdisc

a Antibiotic Antibiotic All killed Resistant bacteria b Antibiotic Antibiotic Antibiotic Antibiotic Figure 4 Prolonged chemotherapy can lead to an enhanced rate of emergence of resistance to antibacterial agents. a Most multiplying bacteria are killed by bactericidal antibiotics, such as penicillin. b Surviving bacteria that are not killed can, over a period of days, produce resistant clones, which are then selected by the antimicrobial agent. This results in the resistant clone becoming predominant. mast cells of fish. Another new route for the development of natural products from soil has been proposed 65. It is thought that 99% of soil bacteria are non-cultivable. So, researchers are now extracting bacterial DNA from soil samples, cloning large fragments into, for example, bacterial artificial chromosomes, expressing in a host bacterium and screening the library for new antibiotics. This could open up the exciting possibility of a large untapped pool from which new natural products could be discovered. Chemical synthesis can be used to make fundamentally new structures that might act at different bacterial targets to those already identified. Drugs arising from such approaches include the oxazolidinones and ketolides (TABLE 2). Potentially, new antimicrobial agents could result from combinatorial chemistry 4,11,12. However, the compounds identified so far have been shown to be either too toxic or too insoluble to allow them to be developed as drugs. ne drawback is that although inhibitors of enzymatic reactions in vitro can be found, the compounds cannot be converted into effective inhibitors of bacterial cells. The overall problem with classic screening coupled with assays that use actively multiplying bacteria is that bacterial resistance arises soon after the new antimicrobial agent is widely used in the community. This means that it is necessary to produce new antimicrobial agents that have different mechanisms of action at regular intervals. Structural changes to existing drugs The emergence of antibiotic resistance led to the synthesis of many derivatives of existing antibiotics in the hope of discovering drugs that would be effective against resistant strains a process that has been dubbed molecular roulette. Whole families of drugs have been made (TABLE 1) that are based on, for example, penicillin (the -cillins), erythromycin (macrolides), imipenem (carbapenems) and nalidixic acid (quinolones) 66,67. Cefuroxime is an early example of a second-generation cephalosporin that was launched by Glaxo in the United Kingdom in 1978. It had the code name 826/359 because the Glaxo medicinal chemists had tried 826 separate groups on position 3 and 359 groups on position 7 (REF. 68). New compounds in the β-lactam, macrolide, tetracycline and quinolone families were produced that have advantages in spectrum, potency 69 and/or pharmacology 70, such as improved oral absorption or longer half-life. TABLE 2 shows new antibacterial drugs that have been registered in the past two years or are in clinical development. The four quinolones, four glycopeptide-like molecules, one carbapenem and one cephalosporin are derived from molecules that were first launched in 1962, 1967, 1942 and 1963, respectively. The two ketolides, two oxazolidinones, the deformylase inhibitor, the combination of quinopristin/dalfopristin and the diaminopyrimidone are fundamentally novel and are different from other marketed compounds. It is interesting to note that these newer compounds originate from traditional routes; there are no new antimicrobial agents in development at present that originate from fundamentally new approaches, such as bacterial genomics (see below). As described in a recent review of new antibacterial drugs, the short-term response from the pharmaceutical industry and biotechnology speciality pharmaceutical companies to bacterial resistance is, and has been, focused on structurally altering existing molecules 11. NATURE REVIEWS DRUG DISCVERY VLUME 1 NVEMBER 2002 901

a Multiplying Non-multiplying Antibiotic treatment Little or no effect All killed Antibiotic treatment All killed b Antibiotic treatment Antibiotic treatment Bacterial number Bacterial number Time Time Figure 5 Non-multiplying bacteria. a Bacteria can exist in two states multiplying and non-multiplying. Most infections contain both multiplying and non-multiplying organisms. During a course of antimicrobial agents to treat an infection, all the multiplying bacteria are quickly killed, leaving a pool of non-multiplying organisms that survive. When the levels of antimicrobial agent fall between doses, the non-multiplying bacteria spin off multiplying ones, which are then killed by the next dose of the drug. b Antibiotics, such as penicillin, can kill multiplying bacteria within hours, but are either ineffective or only partially effective against non-multiplying organisms. An interesting twist is the genetic manipulation of bacteria, such as Streptomyces spp. 71,72 and Escherichia coli 73, in such a way that they make a hybrid antimicrobial agent based on, for example, erythromycin. This technology is based on the principles of heterologous gene expression, molecular biology and the promiscuous nature of enzymes that are involved in the production of secondary metabolites, such as antibiotics (for reviews, see REFS 74 77). The advantage of this method is that numerous changes can be made to the antimicrobial agent by relatively cheap genetic means. Another intriguing approach is to merge natural-product biosynthesis with combinatorial solid-phase chemistry. Recently, this approach has been used to produce novel cyclic peptides related to the antibiotic tyrocidine that have improved properties as potential drugs 77. The main disadvantage of synthesizing derivatives of existing agents is that, because the parent molecule led to bacterial resistance, the new one will too, with a high likelihood of cross resistance, which further reduces the effectiveness of related antimicrobial agents. Molecular roulette alleviates the problem of bacterial resistance to a point, but sooner or later, the bacteria win. Genome hunting Many bacterial genomes have been sequenced 78 in recent years, and this has opened up the possibility of finding suitable enzyme targets against which inhibitors can be developed (FIG. 6). This topic has been reviewed extensively 12,79 87, and will be discussed only briefly here. The underlying hypothesis is that it might be possible to avoid encountering previously generated resistance by 902 NVEMBER 2002 VLUME 1 www.nature.com/reviews/drugdisc

SURVIVAL Maintenance of viability. seeking new targets for antimicrobial agents, such as proteins that are crucial for the SURVIVAL of the bacterium 87. Some pharmaceutical companies are screening potential targets for enzyme inhibitors that block vital metabolic pathways that are not hit by existing antimicrobials 88. Many new approaches are being investigated, often by start-up companies. These include efflux-pump inhibitors and quorum-sensing signalling systems 89, which are used to sense and respond to changes in cell population density. In this era of genomic science, new technology has caused a rapid growth in knowledge about biological function integrated with DNA sequence data. Genome sequencing, bioinformatics, combinatorial chemistry and high-throughput screening have led to exaggerated claims that we can soon expect new classes of clinically useful antimicrobials 4. At present, however, there are no products, even in early development, derived from these technologies. The promise, as detailed by various companies five or more years ago, was that these new technologies would result in a plethora of new targets and new compounds. Furthermore, it was argued that resistance to these compounds would be unlikely, because mutations in the targeted genes that are responsible for function are not compatible with viability. Concern has been expressed that too much focus has been concentrated on the throughput, with not enough focus on understanding the output 90. Indeed, because the DNA sequencing field has progressed so rapidly, large data sets are subject to errors, ambiguities and incompleteness. However, the maturity and status of microbial genomics is improving as people are trained, databases are constructed and improved analytical tools are invented to handle the flood of information 91. This has resulted in several new bacterial targets being identified, many of which are considered essential to the life of bacteria. At present, the field could be described by the statement target rich and compound poor. ther new approaches include antisense peptide nucleic acids that can specifically inhibit E. coli gene expression and growth 92. Chemical modifications improve the potency by at least two orders of magnitude, while retaining target specificity. This new functional genomic method opens up exciting possibilities for drug discovery. A recent article 91 suggests that rapid advances are being made in part owing to bioterrorism. Using bacterial genetics and genomics, structural biology and assay development, it is hoped that new products will arise that will increase the useful life of approaches that search for inhibitors of cell-wall targets 93. This is an exciting new area that has yet to bear fruit in the antibacterial field. Inhibition of an enzyme pathway does not equate to a new antimicrobial agent, and many more steps must be successfully completed before a new antimicrobial drug is produced 11. A fundamental reassessment The current state of antimicrobial chemotherapy is unsatisfactory owing to the relentless emergence of resistance to antimicrobial drugs within a few years of introduction into the market place 12, which reduces their life expectancy considerably compared with other medicines. New antibacterial agents must be launched. However, this assumes that there are a limitless number of new drugs waiting to be discovered, which is almost certainly not true. A pessimistic view is that, like fossil Identification of a target gene Clone gene and express; purify and crystallize recombinant protein Determine structure by X-ray crystallography and identify active site Screen small molecules computationally against active site and/or against recombinant protein and optimize binding Test in whole bacteria Figure 6 Development of new antimicrobials through genome hunting. Numerous bacterial genomes have been sequenced (38 bacterial genomes have been completed or are being sequenced at the Wellcome Trust Sanger Institute 78 ). A bacterium such as Escherichia coli has ~4,000 genes. The development of a new antimicrobial drug starts with the identification of a target gene. This is achieved in various ways. For example, gene knockout in combination with upregulated expression in a DNA microarray in response to exposure to the antimicrobial agent. The gene is then expressed, and the recombinant protein purified. It can then be crystallized for structure determination by X-ray diffraction analysis, and the active site of the target can be identified (other ways of acquiring molecular-structure information include NMR and molecularstructure predictions). Theoretical small chemical moieties can then be fitted into the active-site pocket and tested for binding to the recombinant protein. These are usually optimized further before testing for activity against whole bacteria. Alternatively, if an assay for the protein activity is available, high-throughput screens with large small-molecule libraries can be carried out, with the most promising candidates usually being optimized and then tested against whole bacteria. NATURE REVIEWS DRUG DISCVERY VLUME 1 NVEMBER 2002 903

DRMANCY A reversible state of low metabolic activity in a unit that maintains viability. The nonculturable form of Micrococcus luteus is an example of the dormant state. RESUSCITATIN Transition from a temporary state in which the specified unit had lost the capacity to multiply, to a state in which multiplication can take place. INFECTIUS DISEASE When an infection causes a disease, such as bacterial pneumonia. It is usually associated with the potential to transmit the causative pathogen to other people. STATINARY PHASE This is a growth phase that is slow or non-multiplying and is observed in vitro. Stationary phase bacteria are widely used in in vitro models. fuels, the supply of new antimicrobial agents will one day dry up. Conversely, the optimist might suggest that the low output of absolutely new families of antimicrobial agents during the past 35 years will be reversed by the discovery of, for example, new antibiotics in the 99% of microbes in the biosphere that are non-cultivable. What can be done? The most obvious question is: are we barking up the wrong tree? In particular, is it best to target fast-growing bacteria during antimicrobial discovery? So far, all current antimicrobial agents have been developed against multiplying bacteria (FIG. 4; TABLE 1). However, multiplication is not the main state in which microbes exist similar to humans, they spend most of their time not multiplying. This nonmultiplying state is highly resistant to all known antimicrobial drugs (FIG. 5) and is a universal feature of the bacterial kingdom. This state is resistant to a wide range of environmental stresses, such as heat, and represents a bacterial survival mode. The resistance is reversible, in that when non-multiplying bacteria start to multiply they become sensitive to antimicrobial agents again. This type of resistance is not genetic; rather it is phenotypic, and is sometimes called tolerance. The physiological characteristics of non-multiplying bacteria have been intensively studied, which has revealed a world of remarkable complexity, adaptability and resistance to attack, in great contrast to bacterial reproduction 94. Non-multiplying bacteria exist in three forms: spores, such as those found in the Bacillus spp. 95,96 ; DRMANT, such as Micrococcus luteus 97,98,which requires specific bacterial RESUSCITATIN factors to restart multiplication; and clinically latent 60,99 104, in which most bacterial species exist. A pathogen in a clinically latent state causes a hidden infection that might involve microbial growth, albeit slowly multiplying, which is balanced by host control mechanisms so that the infection remains below the threshold of INFECTIUS DISEASE expression 105. Alternatively, the pathogen might be non-multiplying. It is not usually possible to distinguish between multiplying and non-multiplying bacteria in vivo, and these two states probably coexist. The term STATINARY PHASE is sometimes used to describe slow or non-multiplying bacteria in vitro. The rest of this Review deals with clinically latent bacteria, and the rationale and potential for therapeutically targeting them. Targeting non-multiplying bacteria Non-multiplying bacteria prolong treatment. In bacterial pneumonia, for example, the microbes consist of at least two populations that exist simultaneously: multiplying and non-multiplying (FIG. 7). Multiplying bacteria are killed quickly by antimicrobial agents, whereas in humans and animals, non-multiplying or slowly multiplying bacteria tolerate repeated doses of antimicrobial agents and lead to the need for a conventional prolonged course of drugs 58 60,106 111. The pool of non-multiplying bacteria spin off multiplying ones, and this prolongs the treatment period 111 114. If the antimicrobial agents are stopped before the pool of multiplying bacteria has been substantially reduced or eliminated, clinical relapse will occur. The numbers of non-multiplying bacteria that survive depends on the dose, the duration and the type of antimicrobial agent, the species of bacteria and the population dynamics of the bacteria (biofilm or individual, separated cells). For example, drugs such as carbenicillin, tobramycin and isoniazid are relatively inactive against some non-multiplying bacteria 106 109. Conversely, penems can kill several types of non-multiplying bacterium, and Nocardicin A kills stationary phase E. coli 115,116. Some non-multiplying Gram-negative organisms are mostly killed by quinolones, leaving a fraction of invulnerable persisters 107,110, and rifampicin kills most, but not all, clinically latent Mycobacterium tuberculosis 60. Problems with prolonged treatment. Resistant bacteria are often present in the healthy human commensal bacterial flora 117 122. Prolonged suboptimal bactericidal concentrations can lead to the emergence of resistance, not usually in the target pathogen, but in the normal flora in the gut, skin and throat 123. Long courses of antimicrobial agents are more likely to encourage the emergence of resistance than shorter courses 124,125 (FIGS 4 and 7). Long courses of antimicrobial agents tend to kill fast-growing susceptible bacteria, but select for resistant ones and induce mutations that favour resistance (FIG. 4). Non-multiplying bacteria will tend to survive and, interestingly, probably have an enhanced ability to mutate to resistance 126 128. For example, non-dividing E. coli continually mutates to ciprofloxacin resistance during a seven-day exposure to the agent 129.Longterm incubation of Pseudomonas aeruginosa in the presence of bacteriostatic tetracycline increases the mutation rate by several orders of magnitude 130.In Salmonella typhimurium, carbon starvation increases the rate of mutation to rifampicin resistance 128,but does not lead to a general increase in mutation rates. So, nonmultiplying bacteria might be one of the sources of resistant bacteria. Prolonged courses of antimicrobial agents are also associated with a reduction in patient compliance, which leads to an increased rate of resistance 131 133. If the patient takes the drug intermittently, only suboptimal levels are achieved in the tissues. During the six-month therapy for tuberculosis, patient compliance is crucial to prevent resistance 112,134. Patient compliance also has a lesser, but nevertheless significant, role in the treatment of other types of infectious disease 131 133. A possible solution. No antimicrobial agents have been specifically developed against non-multiplying bacteria. ne possible solution to the current problem might be to shorten the duration of chemotherapy by targeting non-multiplying bacteria with new antimicrobial agents (FIG. 7). Drug libraries should be screened against clinically latent bacteria to discover new antibacterial drugs that kill them. It is generally accepted that chemotherapy should try to eradicate disease-causing bacteria in the patient as quickly as possible 135. Ideally, one dose of an antimicrobial agent should have the least possible impact on the normal flora. It is possible that even those bacteria 904 NVEMBER 2002 VLUME 1 www.nature.com/reviews/drugdisc

in the normal flora that survive a one-dose therapy will not mutate at the chromosomal level to higher levels of resistance, providing the period of exposure to the antimicrobial agent is very short 125 (FIG. 7).So, one-dose therapy should reduce the rate of emergence of resistance against itself by shortening the period of therapy (which will increase patient compliance 136 ) and reducing the number of chromosomal mutations. These drugs should have short half-lives so that they do not linger in the body, which should minimize Figure 7 Killing non-multiplying bacteria with one-dose therapy. a In a patient with a bacterial infectious disease, such as pneumonia, both multiplying and non-multiplying bacteria are present. The antimicrobial agent kills the multiplying bacteria but does not kill the non-multiplying ones. When the level of the drug dips between doses, metabolism in the surviving non-multiplying bacteria spurts into activity, and some start to multiply. The next dose of antimicrobial agent, which raises the level of drug, kills these multiplying organisms. The number of bacteria in the non-multiplying pool gradually declines, but enough survive to repeat the cycle many times. If antimicrobial treatment is stopped too early, non-multiplying bacteria can start to multiply again, which leads to reactivation of infectious disease. Eventually, when the non-multiplying pool is either eliminated or can be mopped up by the immune system, the patient is cured. However, prolonged treatment with antimicrobial agents, particularly with suboptimal concentrations or in the presence of partially resistant bacteria, leads to resistance in the commensal bacteria of the bowel, mucosa and skin. ver the period of chemotherapy, these resistant bacteria increase in number, and can be transmitted to other people, or cause disease with a resistant organism in the parent host. b Drug discovery that is aimed at this non-multiplying population should result in shorter periods of chemotherapy, perhaps one dose. This should lead to a lower level of resistance in the commensal flora and could prolong the effective life of new antimicrobial agents. A drug that could kill the non-multiplying bacteria should lead to an overall improvement in the rate of emergence of resistance, not only because it should reduce chromosomal resistance in clinically latent bacteria, but also because it should improve patient compliance. NATURE REVIEWS DRUG DISCVERY VLUME 1 NVEMBER 2002 905

VIABLE Capable of multiplication. TLERANCE T ANTIMICRBIAL AGENT A microbe that survives treatment with an antimicrobial agent (at or above the minimum inhibitory concentration (MIC)) without altering its genome is said to be tolerant. For example, multiplying log-phase Mycobacterium tuberculosis is killed by sub-µg per ml concentrations of rifampicin. However, when the growth of the organism slows, it can survive. In other words, it tolerates higher concentrations of rifampicin; in certain situations one thousand times the MIC. The survivors, or persisters, become highly sensitive to rifampicin again when they re-enter the log-phase. exposure of other, commensal bacteria to the new drug. Clearly, there needs to be a balance between killing the target organism and the emergence of resistance. So, if efficacy relies on duration-dependent killing, then a longer half-life might be needed to prevent the emergence of resistance. For each agent, the optimal dosing must be identified, with detailed pharmacokinetics (PK) and pharmacodynamics (PD), to minimize the emergence of resistance. Animal models have been used extensively in the evaluation of antimicrobials, and show that the PK/PD target determining efficacy varies between different classes of antimicrobials. However, the magnitude of the target required for bacteriological efficacy is relatively similar for various sites of infection, pathogens and drugs in the same class, providing that free drug levels are used 137. ver the past decade, there have been considerable advances in the understanding of PK/PD predictors of microbiological outcomes. Measures of drug exposure, such as peak concentration and the area under the concentration versus time curve, relative to a measure of the potency of the drug for the organisms being treated for example, the minimum inhibitory concentration (MIC) are linked to clinical outcomes in both animals and humans 138,139. In addition, new agents should not be excreted in the bile, which would affect the bowel flora. Unfortunately, one-dose antibacterial drugs will have little effect on existing levels of horizontally acquired resistance, because resistance plasmids are now so highly prevalent. A case for combination therapy? If new drugs that target non-multiplying bacteria are used in combination with those that target multiplying bacteria, the emergence of antimicrobial resistance to the new drugs could potentially be slowed, and the drugs could remain useful for longer than at present. In the treatment of tuberculosis, combination therapy reduces the incidence of drug resistance 140,141, and probably lowers the emergence of resistance in P. aeruginosa 142,143 and Helicobacter pylori 144 infections, but the evidence seems to be less strong for other infections 142,145 148.Care should be taken to use combinations that are not antagonistic, and are, preferably, synergistic. Genetic resistance. Slowly growing or non-multiplying bacteria have increased rates of mutation of antimicrobial drug targets relative to the fast-growing state 126. For example, starvation, which induces a slowly multiplying state, is accompanied by an increased rate of mutation to rifampicin resistance in E. coli and S. typhimurium 127,128. The clinical importance of mutations that arise from non-dividing rather than dividing bacteria is unknown. However, the wide variety of resistant mutants that are selected during clinical infection 149 indicates that it is unlikely that one physiological state namely fast-multiplying cells is the sole originator of mutants. Clinically latent bacteria. The latent state has numerous different populations 60,150,151, is metabolically active 60,152,153, synthesizes protein 152,154, contains messenger RNA 59,155, has no growth or multiplies at a very slow rate and has a constant net VIABLE count 99,104,156. When antimicrobial agents are added to cultures of latent bacteria, there are hundreds of changes in gene expression at the mrna level, some genes being upregulated and others being downregulated (Y. H. & A. C., unpublished observations). Interestingly, among all the genes that are modulated by antimicrobial agents, only a few are related to known antibacterial resistance mechanisms. These data indicate that clinically latent bacteria might be highly adaptable. The best example of a clinically latent bacterium is M. tuberculosis. This species exists in at least two forms: multiplying and clinically latent 60,150,151,156.In a patient with active tuberculosis, many of the invading bacteria are multiplying, but in an otherwise healthy person who is carrying the bacteria (there are two billion carriers in the world 157 ), most of the bacteria are clinically latent. Although multiplying bacteria are quickly killed by antimicrobial agents, chemotherapy is required for six months 150,151,158, because the non-multiplying bacteria are tolerant to existing antimicrobials. Unfortunately, many patients stop taking their drugs after a few months, which leads to recurrence of disease and transmission to others in the community 134,135. To avoid the emergence of bacterial resistance, tuberculosis is treated with a combination of three or four drugs. If only one or two drugs are used continuously or intermittently, owing to inadequate drug regimens as a result of poor patient compliance or incorrect prescribing, bacterial resistance commonly occurs. Unfortunately, bacterial resistance is now rising 159. Animal models, such as the Cornell model 111, show that clinically latent bacteria in mice contain specific mrnas 60 ;they cannot be grown on media, but can cause disease after a period of immunosuppression. A similar situation occurs in other bacterial infections. For example, although multiplying bacteria can be killed in a test tube by one dose of an antimicrobial agent, such as tobramycin or ciprofloxacin 160,in humans, a number of doses over several days are required to cure an infection. The genetics of TLERANCE T ANTIMICRBIAL AGENTS is not well understood, although genes including hip, vncs and sula 107,161 164 are associated with tolerance and can increase the number of bacteria that survive antibacterial drugs. For example, vncs mutant S. pneumoniae 161,163 that are in stationary phase are resistant to killing by several different antimicrobial agents that hit unrelated targets, although growth of the mutant is inhibited by the drugs as effectively as in the wild type. vncs encodes a putative histidine kinase that is part of a two-component sensor regulator system. Such two-component systems monitor the environment through a sensor histidine kinase/phosphatase, which phosphorylates/ dephosphorylates a response regulator that mediates changes in gene expression. This indicates that signal transduction might be needed for antimicrobial agents to be bactericidal. Another gene that might be important is rela.an E. coli rela mutant makes nonmultiplying cells sensitive to killing by antimicrobial 906 NVEMBER 2002 VLUME 1 www.nature.com/reviews/drugdisc