Nature PublishingGroup PERSPECTIVES organisms become widespread

26 Jun Nature PublishingGroup PERSPECTIVES organisms become widespread

O P I N I O N — A N T I – I N F E C T I V E S

Where will new antibiotics come from?

to combat successive waves of resistant pathogens. Analogously, zithromax and biaxin are second-generation variants of the natural product erythromycin. Third-gener-ation macrolides, the 3-ketolides, have now been approved in Europe and are under review in the United States for clinical

Christopher Walsh

There is a constant need for new antibacterial drugs owing to the inevitable development of resistance that follows the introduction of antibiotics to the clinic. When a new class of antibiotic is introduced, it is effective at first, but will eventually select for survival of the small fraction of bacterial populations that have an intrinsic or acquired resistance mechanism. Pathogens that are resistant to multiple drugs emerge around the globe, so how robust are antibiotic discovery processes?

Every antibiotic that is introduced into clinical use has a limited shelf life as it selects for bacte-ria that have some intrinsic or acquired mecha-nism of resistance. Although these bacteria are rare (for example, 1 in 108), in the continuing presence of the selecting antibiotic the resis-tant bacteria become more populous than their dying neighbours. The timeline for the development of clinically significant resis-tance is multifactorial and depends on several parameters, such as the quantity of antibiotic used, how widely it is prescribed, the fre-quency with which sub-therapeutic levels of antibiotic are used to select for resistant organ-isms, the reservoirs of existing resistance mech-anisms, the number of mutations required for resistance to emerge in a killing target and the fitness of the resistant organisms1,2.

Resistance has developed to every main class of antibiotic, both natural and synthetic, over the course of 1 year to more than a decade3 after the first clinical use. Resistance, then, has proven not to be a matter of if, but a matter of when. As a new antibiotic moves into significant clinical use, the clock starts ticking down on its useful lifetime as resistant

organisms become widespread. The number of new antibiotics that are needed to treat the same bacterial infection also increases reciprocally. For example, in the 1940s, staphylococcal infections were treated with first-generation penicillins. Within a year, resistance had appeared in Staphylococcus aureus, and a decade later,β-lactam resistancehad spurred the development and introduc-tion of methicillin. Methicillin-resistant

S. aureus(MRSA) became so prevalent that,in 1986, vancomycin became a front-line antibiotic to treat MRSA infections. Since then, MRSA have evolved to vancomycin-resistant S. aureus (VRSA), and the quin-upristin/dalfopristin combination (Synercid) and then linezolid (Zyvox) were approved in the late 1990s as new therapeutic options2,3.

Current molecules

In 2002, the global antibiotic market was esti-mated at US $25 billion4 and six antibiotics each topped US $1 billion in sales3. This set of six antibiotics comprises twoβ -lactams, ceftriaxone (Rocephin) and an amoxicillin/ clavulanate combination (Augmentin), two

MACROLIDES, azithromycin (Zithromax) andclarithromycin (Biaxin), and two fluoro-quinolones, ciprofloxacin (Cipro) and lev-afloxacin (Levaquin). These six antibiotics reflect three structural classes, which have been the mainstay of antibiotic scaffolds for four decades. Ceftriaxone (Rocephin) is a third-generation cephalosporin and Augmentin is a combination of theβ -lacta-mase inhibitor clavulanate and the second-generation penicillin amoxicillin — reflecting the enormous effort in semi-synthetic improvements toβ -lactam natural products

approval. In turn, the best-selling fluoro-quinolones are second-generation mole-cules, and newer, third-generation versions, such as gatifloxacin, have been introduced in recent years. Other antibacterial drugs, such as tetracyclines and isoniazid, still have useful roles in therapy but they are not the focus of this perspective.

Innovation gap

Although the history of antibiotic develop-ment demonstrates the remarkably successful

mining ofNATURAL PRODUCT SCAFFOLDS by gen-

erations of medicinal chemists to meet the challenges of resistance development, it also emphasizes the ongoing, cyclical need for innovation. Indeed, it raises the question of whether progressive tinkering with antibiotic scaffolds has been hiding an innovation gap, as a dramatic gap existed between the introduction of quinolones in 1962 and the next new structural class of antibiotic, the oxazolidinone linezolid (Zyvox), which was approved almost 40 years later(TIMELINE).

Does the innovation gap still exist? Inspection of the antibacterial drug candidates that are being advanced through clinical trials4 indicates that it does. Many of the agents that are in development at present continue to be minor modifications of the three structural scaffolds described above. These second- and third-generation molecules can be extremely important for achieving improvements in effi-cacy and pharmacokinetics, but they do not have new molecular structures. There are also variants of other approved classes of antibacte-rial drugs, such as the tetracyclines and gly-copeptides of the vancomycin and teicoplanin family, that have been modified to regain effi-cacy against the pathogenic bacteria that, at present, cause the most common infectious

NATURE REVIEWS|MICROBIOLOGY VOLUME 1| OCTOBER 2003|65

.jpg”>

© 2003 Nature Publishing Group

P E R S P E C T I V E S

Timeline | Introduction of new classes of antibiotic into clinical practice

Quinolones

Sulfa drugs target

Phenyl propanoids

Macrolides

target nucleic-

Streptogramins target

folate metabolism

target ribosomes

target ribosomes

acid replication

ribosomes (for example,

(for example,

(for example,

(for example,

(for example,

Synercid, a quinupristin and

trimethoprim)

chloramphenicol)

erythromycin)

ciprofloxacin)

darfopristin combination )

1936

1940

1949

1950

1952

1958

1962

2000

β -Lactams target

Polyketides

Aminoglycosides

Glycopeptides target

Oxazolidinones

envelope synthesis

target ribosomes

target ribosomes

envelope synthesis

target ribosomes

(for example,

(for example,

(for example,

(for example,

(for example,

ampicillin)

tetracycline)

gentamicin)

vancomycin)

linezolid)

Note the innovation gap between 1962 and 2000. Example drugs of each structural class were not necessarily introduced on the dates shown. Modified fromREF. 3 © (2003) ASM Press.

Synthetic antibacterials.The second line ofantibiotic discovery has come from synthetic chemistry — that is, producing antibacterial agents from structures that are not found in nature. From the historical tradition of the medicinal chemical search for ‘magic bullets’ came the sulfa drugs. This drug family are

ANTIMETABOLITEStoρ-aminobenzoate, a sub-strate for an essential enzyme in the folate biosynthetic pathway. The sulfa drugs have been in continuous anti-infective use for the longest time period, having been in use since their introduction in the 1930s. An existing version is sulphamethoxazole, which is often formulated in combination with tri-methoprim, an inhibitor of a second enzyme in the folate biosynthetic pathway (and indi-rectly of DNA biosynthesis) to generate co-trimoxazole9. The second group of synthetic

diseases. The lipopeptide antibiotic dapto-mycin, at present in new drug approval review, acts by a mechanism that is distinct from the lipoglycopeptide teicoplanin and could fill a niche against the vancomycin-resistant entero-cocci (VRE), in addition to other pathogens5. The second indicator that the clinical pipeline for antibiotics is empty is the fact that few large pharmaceutical firms are active in the antibac-terial infectious disease arena. The exit and/or significant de-emphasis of many pharma-ceutical companies (for example, Roche,

GlaxoSmithKline, Bristol-Myers Squibb and Lilly)4 from this therapeutic area in the past 15 years reflects not only a mix of economic and market projections, but also a partial to com-plete failure of research programmes that have been built on existing models to find new leads that are robust enough to become clinical candidates.

Antibiotics: two lines of discovery

Antibiotic discovery efforts over the past 70 years can be divided into two streams with distinct origins3(FIG. 1).

Natural products.The discovery of the anti-bacterial activity of natural products facilitated the development of assays for the purification and characterization of penicillins, cephalo-sporins, aminoglycosides, tetracyclines, ery-thromycin and related macrolides, and vanco-mycin and teicoplanin. This strategy was dominant from 1940–1960 in the ‘golden era’ of antibiotic discovery by microbiologists. Antibiotic natural products tend to have complex architectural scaffolds and densely deployed functional groups (substituents that are likely to interact with biological targets) for the specific interaction with, and recognition by, targets in pathogenic bacteria. These complex structures have tended to mean that

large-scale production is by fermentation rather than by practical total syntheses and that oral bioavailability can be limited in these large molecules. However, these complex scaf-folds have been excellent platforms for re-modelling and re-engineering by medicinal chemists to create subsequent generations of semi-synthetic variants of antibiotic natural products in the marketplace today.

The balance in antibiotic research and dis-covery programmes over the past one or two decades has shifted away from antibiotic nat-ural products, reflecting two lines of thought.

First, after 50 years of intensive screening, the number of novel natural products with useful antibiotic potency and spectrums might have run out. Second, advances in high-throughput

COMBINATORIAL CHEMISTRYmethodologies can

generate large libraries of synthetic compounds that are starting points for the optimization of novel synthetic antibacterials.

What about contemporary natural product discovery efforts? Are they indeed tending asymptotically to an unproductive limit? It is likely that marine microbial communities, including the cyanobacteria, have not been systematically collected and fully evaluated for bioactive compounds. Even more com-pelling is the fact that only a small portion of microbes are culturable by present methods and that 90–99% of theMETAGENOME is not being sampled6. This has led to efforts to clone 50–100-kilobase fragments of DNA, express the encoded genes heterologously and screen for novel bioactive compounds. As the biosyn-

thetic genes of thePOLYKETIDE andNON-RIBOSOMAL

PEPTIDEantibiotics are known to occur in geneclusters7,8, this approach is expected to capture full metabolic pathways for the synthesis of antibiotics. These efforts are only in the early phases and have not yet revealed many novel structures.

antibacterial drugs are the fluoroquinolones10, which block DNA gyrase and a related DNA topoisomerase (topoisomerase IV) in DNA replication and repair11. Fluoroquinolones are front-line drugs globally, and were used to treat recent anthrax cases in the United States.

Linezolid (Zyvox), the first of the most recently introduced antibiotic class oxazolidi-nones, continues the tradition of synthetic molecules being successfully developed into antibiotics for clinical use.

It could be argued that the paucity of novel structural scaffolds, whether synthetic or nat-ural, that can be developed into new antibiotics might continue to be the rate-limiting step in the years ahead. This might hold true both for the initial validation that novel targets in path-ogenic bacteria can be specifically inhibited by hits from molecule libraries, and for the devel-opment of screening hits into successful leads that can enter clinical development.

According to some observers, the source of new antibiotics is unlikely to come from com-binatorial libraries as there has not been a sig-nificant increase in new structural leads from

‘diversity-oriented’ synthetic efforts, which have aimed to create diversity at many branch points in the molecular structure. The first-generation synthetic libraries in the early to mid-1990s provided quantity but not quality.

Many early libraries of synthetic compounds produced by combinatorial chemistry might not have had sufficient architectural complex-ity or functional group density to be good, or even adequate, starting ligands to produce hits in the range of 1–10µM — the range at which screening thresholds are typically set. It could be argued that an architectural complexity and functional group density approaching that of natural products should be a goal of synthetic library construction, to provide a population of molecules that have a

66|OCTOBER 2003|VOLUME 1 www.nature.com/reviews/micro

© 2003 Nature Publishing Group

.gif”>P E R S P E C T I V E S

chance of moderate- to high-affinity recogni-tion by specific regions of target macromole-cules. Several efforts to produce libraries in accordance with such criteria have recently been reported and reviewed12–14. If these prove to be fruitful starting points in diverse biolog-ical screens, the continued evolution of syn-thetic library complexity and diversification of reaction schemes to build diverse synthetic chemical libraries can be expected.

However, two countering observations can be made. First, fluoroquinolones and the oxa-zolidinones, although amply endowed with nitrogen and oxygen functional groups in the

tradition ofHETEROCYCLIC DRUGS, are mostly

two-dimensional (that is, flat) molecules and do not exemplify complex three-dimensional architecture on their own. Second, the molec-ular mass of most of the orally active drugs is less than 500 daltons15, and biologically rele-vant structural complexity would have to be achieved within this size constraint if oral activity of new antibiotics is to be routinely achieved.

Another variant of the synthetic approach is to re-programme the polyketide and non-ribosomal peptide pathways by swapping protein domains, modules and subunits between sets of polyketide synthase and non-ribosomal peptide synthetase ‘assembly lines’, to mix and match the monomer units that are selected and allow chain elongation of novel intermediates16. Combinatorial biosynthetic diversification can also occur at the level of

generating alternative monomers for the assembly lines and in post-assembly line reac-tions that alter redox states and add deoxy-sugars to aglycones — steps that are often crucial for imparting antibiotic activity to the aglycone scaffolds17. So far, these combinator-ial biosynthetic approaches have been imple-mented to introduce small numbers (2–4) of changes and to produce small numbers (tens to hundreds) of variant products16, rather than producing millions of variant structures.

Targets for antibiotic action

Historically, the antibacterial therapeutic arena has been a target-poor environment in terms of validated macromolecular targets with useful bacteriostatic or bacteriocidal activity in humans. The main classes of antibacterial drugs mentioned above inhibit only four classical targets: bacterial-cell-wall biosynthesis; bacterial protein biosynthesis;

DNA replication; and folate coenzyme biosynthesis (thereby blocking thymidine 5′-monophosphate and eventually DNA biosynthesis)3(FIG. 2). Multiple steps have been blocked in protein biosynthesis — macro-lides, aminoglycosides and tetracyclines all bind to ribosomal RNA, but at distinct sites on the 50S and 30S subunits18–21 — consistent with the many steps that are involved in building up polypeptide chains.

The explosion in microbial genome sequencing in the past decade has radically altered the information available about

essential genes and encoded proteins. There are genomic sequences of many important bacterial pathogens available, with almost 100 microbial genomes deposited in the

Comprehensive Microbial Resource that is run by The Institute for Genomic Research

(TIGR)22 (seeTIGR in the Online links). In turn, there have been many efforts, using a variety of genetic approaches, to define the subset of essential bacterial genes (~200–400) that would be a priori candidates for killing sites by inhibitors23,24. Comparative bacterial genomics allows further prioritization of open reading frames that are conserved in pathogens but absent in higher eukaryotes, and for which a function can often be pre-dicted by bioinformatics. Therefore, dozens of candidate target bacterial proteins can be rapidly identified and purified, and assays can be developed for effective screening. In principle, the genomics revolution has made this a target-rich therapeutic area.

In this information-rich environment, is there still an innovation gap? Which, if any, of these potential new bacterial target proteins will lead to new structural classes of natural-product-derived or synthetic antibacterials? One answer might be that given that the drug discovery/development process typically takes 6–10 years, it is still too early to tell whether success is (still) around the corner. Another answer could be that many companies have already run their compound libraries on promising targets without success and have left the antibacterial therapeutic area. It is an

aNatural products

H

H3C

H

N

N

O

OH

H2N

N

NH2

S

N

N

O

OH

O

S

O

O

O

O

OH

HN

N

Ceftriaxone

Cl

OH

COOH

(Rocefin)

O

O

O

H3C

O

CH3

HO

Cl

OH

O

O

O

N

H

H

H

O

N

N

N

HO

OH

CH3

N

N

N

CH3

OH

HN

H

O

H

O

H

N

CH3

OHO O

HOOC

O

NH2

O

O

O

CH3

OH

O

HO

OH

Vancomycin

Azithromycin

O

OH

(Zithromax)

bSynthetic products

O H

Sulphomethoxazole

N

O

S

N

O

O

O

F

COOH

NH2

O

N

CH3

O

N

N

O

H

NH2

N

N

CH3

H2NN

N

HN

O

F

Linezolid

O

O

Ciprofloxacin

CH3

(Zyvox)

Trimethoprim

Figure 1 | Structures of naturally and synthetically derived antibiotics. Antibiotic trade names are indicated in parentheses.

open question as to how many of the newly catalogued essential gene products will turn out to be robust targets. The answer will only come if molecules that are used for screening are of a high enough quality and architectural diversity in their unoptimized starting states to score promising hits as antibiotics. If the num-ber of companies that are committed to new antibiotics is small, this could be an area of real opportunity for lucrative investment. If too few companies pursue novel targets, the num-ber of discovery scientists in this field could be too low, and good opportunities could be missed because not enough target molecules or potential inhibitors are evaluated.

Targets with high probability?

There are several cases in which additional knowledge is likely to increase markedly the probability of finding novel molecules that are active against new macromolecular tar-gets3(FIG. 2). An under-explored target is the transglycosylase step of peptidoglycan assem-bly in the cell wall(FIG. 2a), in which the chemo-enzymatic synthesis of the complex lipoglycopeptide substrate, Lipid II, which