Antibiotics at the crossroads

Carl Nathan

The public takes for granted that the phar-maceutical industry can anticipate society’smedical needs and meet them. This faith isnowhere more evident than in the expecta-tion that antibiotics are readily available totreat bacterial infections. After all, infec-tious diseases are the second-leading causeof death worldwide and the third-leadingcause of death in economically advancedcountries. But surprisingly, despite growingbacterial resistance to existing drugs, anti-biotic development in the pharmaceuticalindustry is steeply declining (see chart,right)1–3. This new problem is convergingwith an old one — the scarcity of anti-biotics to treat diseases prevalent mainly inpoorer regions.

The emerging crisis in wealthy nationsand the long-standing crisis in poor nationsresult from the same causes — economic,regulatory and scientific — each exacerbatedby the problem of antibiotic resistance.Government agencies and professional societies have addressed the latter prob-lem3–6, but little has changed. We need newapproaches, beginning with the recognitionthat the antibiotic crises of wealthy and poornations are the same. The challenge is this:what can we do about the level of antibioticresearch and development, which has longbeen insufficient to meet the needs of mostpopulations, and now is plummeting? Causes of this decline are reviewed below,followed by ‘blue sky’ proposals for a moreconstructive approach to the permanentstruggle with infectious disease. The focus ison drugs, but vaccination has a major role toplay in reducing dependence on antibiotics.

Economic pressuresWith respect to profit margins, financialmarkets hold the pharmaceutical businessto a higher standard than almost any otherindustry. The demand for blockbusterdrugs pressures companies to focus onlong-term treatment of chronic conditionsin preference to brief treatments for bacterialinfections7. Most of the antibiotics thatmajor firms make are designed for broad-spectrum activity, so that they can be usedby as many patients as possible. This short-ens the market life of an antibiotic — aswidespread use of an antibiotic hastens theemergence of resistance against it. To wardoff resistance, physicians are urged to sparetheir use. With profits thus restrained in themedical arena, pharmaceutical firms sendroughly half their antibiotic output to the

food industry5. Pork, fowl, fish and dairyproducers use antibiotics to maintain stockand foster growth. This selects for resistantbacteria, which can find their way intohuman populations — hastening the demiseof the drug and making once-treatableinfections incurable4–9.

Industry’s retreat from developing newantibiotics is leading to a loss of expertise inboth practical and theoretical aspects ofantibiotic biology1. As industry reassigns orretires its microbiologists, academia will inturn train fewer5. When wealthy societiesdemand a resumption of antibiotic research,itwill take years to rebuild the knowledge base.

Better business modelsThe Global Alliance for Tuberculosis DrugDevelopment, a not-for-profit agency, isbuilding a case to persuade industry thatmoderate profits can be made by developing

commentary

NATURE | VOL 431 | 21 OCTOBER 2004 | www.nature.com/nature 899

antibiotics for a disease prevalent in poorregions10. AstraZeneca has opened a researchcentre for anti-tuberculosis drug develop-ment in India (www.astrazenecaindia.com)and GlaxoSmithKline has assembled a teamto work on drugs for tuberculosis and malar-ia (www.gsk.com/financial/reps02/CSR02/GSKcsr-7.htm). Although these are positivesteps, these initiatives are not enough toequip us to treat infections endemic in poorregions, nor do they address the emergingshortage of antibiotics for bacterial infec-tions in wealthy nations.

Academic scientists are making rapidadvances in the chemical biology of infectiousdiseases. But they lack access to medicinalchemistry, pharmacology and the expertise to turn ‘hits’ into drug leads, or those leads into drugs.

What is needed is a new player on thescene: a not-for-profit drug company. Theprofit sector could provide leadership.Encouraged by tax incentives, industry couldgive sabbaticals to its scientists and execu-tives to work at a not-for-profit firm in rotation. Many in the pharmaceutical indus-try would like nothing better than to con-tribute personally to an endeavor in whichtheir company (as a whole) is constrainedfrom engaging.

A not-for-profit firm could pursueresearch differently, protecting its intellectu-al property by filing patents, but also adver-tising its work openly,with the goal of licensingthe intellectual property gratis to any com-pany or agency that commits to produce anddistribute the resulting drugs on a basis thatwould serve the needs of patients and society.For example, distribution in low-incomemarkets could be on a for-cost basis whereasdistribution in wealthy markets couldremain for-profit.

Biotechnology firms are beginning to‘translate’ the ideas of academic researchersinto drugs,but it is difficult for small firms tomount a world-class effort at medicinalchemistry and pharmacology, especiallynow that expertise in antibiotic developmentis scarce. A not-for-profit drug companycould perform these services in exchange fora share of revenues from sales in high-income countries, coupled with a commit-ment that the drugs be distributed on a not-for-profit basis elsewhere.

The best established way to delay theemergence of antibiotic resistance is to useone or more drugs in combination —known as combination therapy. A potentialsource of income for a not-for-profit, indus-try-supported drug company could there-

Antibiotics at the crossroadsAre we making the right choices to bring new drugs to the marketplace?

0

10121416

2468

1983–1987

1988–1992

1993–1997

1998–2002

2003–2004

Total number of new antibacterial agents

The lack of new drugs leaves children in developingcountries especially vulnerable to disease.

A.C

RU

MP,

TD

R,W

HO

/SP

LC

.MO

LLO

Y/S

PL;

SOU

RC

E:C

LIN

.IN

FEC

T .D

IS.

21.10 Commentary 899 NEW MH 18/10/04 3:50 pm Page 899

© 2004 Nature Publishing Group

fore be contract work for private firms toidentify effective drug combinations at anearly stage. In fact, a not-for-profit firmcould test the idea that shared knowledgeallows early identification of targets (usuallymicrobial enzymes) whose combined inhi-bition is lethal to the bacteria.

The majority of the funding for a non-profit firm would probably have to comefrom government and foundations. Taxincentives could encourage the for-profitsector to furnish services in kind or at cost,including equipment, supplies, chemicals,clinical development, and regulatory andlegal services. Manufacturing could be contracted to factories in low- and middle-income countries.

One can foresee many problems, such asback-flow of drugs from poor regions towealthy ones. Such problems will be easier tomanage than asking twenty-first centurysocieties to accept nineteenth-century deathrates from infection3.

Regulatory obstaclesAnother major disincentive in the develop-ment of new antibiotics is the current systemof regulatory requirements that discriminateagainst their approval1,5. In the United States,companies must demonstrate that a newantibiotic is superior to existing agents whenused against infections caused by drug-sen-sitive strains. Existing agents are so effectiveagainst drug-sensitive strains that a newantibiotic is unlikely to be much better thanan older one. Yet testing new antibioticsagainst infections caused by antibiotic-resis-tant bacteria is exceptionally difficult, aspatients with serious drug-resistant infec-tions have usually been treated with otherantibiotics before resistance is documented.In short, the regulatory system is geared to

generic standards of safety and efficacy — it makes no allowance for the spe-cific case of antibiotic resistance. Companieshave withdrawn from developing productsagainst which they believe the regulatory system discriminates.

Smarter regulationsIn agreement with recent recommendations3

of the Infectious Disease Society of America(see news feature on page 892), I believeregulatory requirements and patent incen-tives should be revised to encourage thepharmaceutical industry to develop newantibiotics. New antibiotics should beapproved if they meet three tests. First, thesafety profile is acceptable for the severity ofthe infection; second, the drug is effective inpatients against antibiotic-sensitive bacte-ria; and third, the drug is effective in vitroagainst bacteria that are resistant to one ormore existing antibiotics used to treat thatinfection. After a new antibiotic isapproved, its clinical efficacy should bemonitored in patients who are infected withbacteria resistant to previously approvedantibiotics. This information should beposted on the Internet as it is collected.

Unless new antibiotics are used in combi-nation, resistance8 against them will quicklyarise.A new pre-approval test should be devel-oped for antibiotics intended to treat persis-tent or recurrent infections,such as tuberculo-sis and malaria, which require sustainedadministration.The manufacturer should runpreclinical tests to determine how the newagent interacts with existing antibiotics used to treat that disease (one from each class).Thisinformation, combined with the drug’s phar-macokinetic profiles,should help regulators todevelop new post-approval requirements fortreating patients. First, the manufacturer

should specify lists of agents, at least one ofwhich should be used together with the newdrug.Second,the manufacturer should moni-tor clinical efficacy and incidence of drug resis-tance when such combinations are used inpractice.

Patent life should be extended for antibi-otics of new chemical classes directed at newtargets2,analogous to US patent extensions fordrugs developed to treat rare genetic diseases.

In addition, all new antibiotics should bebanned from widespread administration tohealthy animals. It remains for the rest of theworld to embrace an enforceable version ofthe ban enacted in the European Union in1998 (ref. 6), or alternatively, to provide taxincentives with the same effect.

Stalled scienceIt would be simplistic to blame marketforces and regulatory requirements alonefor the antibiotic crisis. There is anotherand more surprising cause — industrialresearch and development has mainly produced variants of older antibiotics,when new drugs are sorely needed. Over thepast few decades, only two new chemicalentities have entered clinical practice asantibacterial agents, and only one whosetarget is in a new biochemical class1,2,8,9. It issurprising that the well has gone dry,despite heavy investment to dig it deeperusing combinatorial chemistry and compu-tational biology. Although genomic analysesare revealing hundreds of potential targetsin pathogens, it remains a fact that almostall agents used to treat bacterial infectionseither have unknown enzymatic targets ortarget just four classes of enzymes — thoseinvolved in synthesis of protein, nucleic acids,cell walls or folate8. How did yield declinewhile knowledge grew and tools improved?

900 NATURE | VOL 431 | 21 OCTOBER 2004 | www.nature.com/nature

AstraZeneca's research operation in Bangalore brings much-needed investment and expertise to India in the search for anti-tuberculosis drugs.

I.M

UK

HE

RJE

E/A

FP/G

ET

TY

IM

AG

ES



The answer may lie in a set of premises thatwere so successful that they hardened intodogmas. In a time of rapid intellectual expan-sion, dogmas are constraints.

Fresh approachesThere is no longer any reason to confineourselves to drugs that inhibit the synthesisof protein, nucleic acids, cell walls andfolate simply because such drugs have beenso successful. We must find new microbialtargets. First, synthesis offers too narrow aset of targets. Macromolecules such as DNAand protein have life cycles. Birth need notbe the only point of intervention, as pro-cessing, repair and degradation are alsopoints of vulnerability. This is the rationalebehind efforts to target the proteasome inMycobacterium tuberculosis12.

Another broad set of targets are theenzymes of core metabolism (intermediarymetabolism, energy generation andmicronutrient acquisition) in the bacteria.Athird set of targets (overlapping the othertwo) are the enzymes that enable thepathogen to resist the defences of the host.After all, evolution has had more time thanscientists to select chemicals to killpathogens; the host has reactive oxygenintermediates, reactive nitrogen intermedi-ates and pore-forming peptides in its arsenal.Of course, evolution has also strengthenedmicrobial defences against the host’schemistries. But we can aid the host’s immunity by usingantibiotics that disablepathogens’ resistancemechanisms12,13.

The goal of anti-biotic development isinhibition of essentialenzymes — those thepathogen needs to sur-vive. But survivewhere? Traditionally,tests to determinewhat enzymes areessential have beenin rich, highly oxygenated cul-ture medium. The conditions facingpathogens in thehost during many infections — especiallythose that are persistent — can be drasticallydifferent from and more demanding thansuch conditions in vitro. Not only do meta-bolic niches in vivo differ from culture brothin many ways (for example, in oxygen, iron,pH and carbon source), but the immune system acts to suppress the pathogen’s repli-cation and damage many of its molecules.Successful pathogens adapt by expressing adifferent set of genes than they do in culture14. Accordingly, a different repertoireof genes is essential in vivo than in vitro9,11,15.In short, essentiality is conditional and the

conditions defining essentiality are multiple.For example, many infections, includingtuberculosis, enter phases of latency — astate of equilibrium between the bacteriumand the host response. The agents now usedto treat tuberculosis kill rapidly growing bacteria in culture within hours. In contrast,treatment of tuberculosis in people takes at least 6–9 months of daily combinationtherapy, because the microbial targets thatare essential in exponential growth phasemay not be so critical during latency or persistence5.

Systematic studies in yeast teach us thatmutations in many

genes are onlylethal when com-

bined with muta-tions in others16. Thisis also true forpathogens. We shouldtherefore target geneproducts which areessential together,even when they arenot essential individ-ually. For example,M. tuberculosis has

two genes encod-ing isocitrate lyaseenzymes, each of

which is capable ofsupporting lipid

metabolism. Disruption of both genes incombination, but neither alone, causes rapidbacterial decline in vivo(J.McKinney,personalcommunication).Why not target both lyasesat once? This idea will require fundamentalchanges in scientific and regulatoryapproaches. First, antibiotic developmentneeds to be cooperative, not competitive —an approach that a not-for-profit drug com-pany could pursue using drug candidatesfrom different manufacturers.Second,when asuccessful combination therapy involves anew unapproved drug, regulatory agenciesshould allow approval of the combination to

NATURE | VOL 431 | 21 OCTOBER 2004 | www.nature.com/nature 901

proceed based on clinical tests of the combi-nation itself, rather than insisting onapproval for each individual component asthey do now.

Another premise that handicaps antibioticdevelopment is that targets in the pathogenmust have no equivalent (homologue) in the host. This is to avoid harming the host.Yet most classes of targets inhibited byantibiotics do have host homologues, exceptthose involved in cell wall synthesis. It is time to abandon the premise.Contemporarystructural, computational and chemicalbiology should be able to engineer compounds that can harm the pathogenwithout harming the host17.

Conventional antibiotic developmenthas reached an impasse, partly because itdemands that new agents have broad-spectrum activity. This imposes severe limi-tations, as targets must be widely conservedacross pathogens — and even then only themost conserved subsites can be targeted.In contrast, it is medically preferable and will preserve the utility of the drugs longer,if antibiotics are highly specific, so that each one is used less often8.

Treating infections with pathogen-specificrather than broad-spectrum antibiotics(whenever possible) will require prior,rapid,accurate and specific diagnosis. It makes nosense to use twenty-first century technologyto develop drugs targeted at specific infec-tions whose diagnosis is delayed by nine-teenth-century methods. Advances in PCR,mass spectroscopy, quantum dot-enhancedimmunoassays, nanotechnology, instru-mentation and other technologies should beused to develop diagnostics. With furtherinvestment, doctors could expect to submitpatient specimens (such as throat swabs,blood or urine) to analysis, and receive diag-noses in many cases within minutes to hours.Today, diagnosis usually takes a day or more.Without minimizing the challenge, weshould acknowledge that pretreatment diagnosis is key to minimizing the use ofbroad-spectrum agents and keeping even inthe endless race against drug resistance8.

It is time to start applying new technolo-gies to antibiotic development. Here arethree examples. First, conventional gene dis-ruption in the pathogen does not allow oneto test the role of a gene during a given stageof infection, such as latency. If disruption ofthe gene precludes growth in vitro, then thegene-deficient mutant cannot be studied atall9. To determine whether a given target isessential in vivo we need ‘conditional geneinactivation’. This allows the investigator toturn a gene off at a particular time after infec-tion has begun, and thereby model the effectof treating the infection with an antibioticdirected against the gene product.

Second, we must not rely exclusively onscreening chemical libraries against enzymesisolated from the bacteria. Although this

Overuse of antibiotics in livestock has led to anincrease in resistant bacterial strains.

Society’s ongoing struggle against infectious disease.

commentary

D.L

EA

H/S

PL



approach identifies chemical compoundsthat inhibit specific targets, it cannot revealwhether they would affect that target insidethe bacterium, or even if they would get intothe pathogen. We need to identify up frontthose compounds that can enter the bacter-ium and inhibit the target in its natural setting. For example, we could replace anendogenous gene encoding a potential targetwith a ‘conditional hypomorphic allele’,allowing reduction but not complete elimi-nation of the target (which could kill it).Then we could screen chemical libraries tofind compounds to which the ‘weakened’mutant is particularly sensitive9.

Third, innovative chemistry can allow usto find more potent inhibitors more quicklyand cheaply. Drug development usuallystarts with inhibitors that work at nanomolarconcentrations. Conventional screening ofcompounds rarely yields inhibitors activebelow the micromolar range. It can taketeams of chemists months to makea micromolar inhibitor athousand-fold more active.But the target enzymeitself may be able toselect weakly bindingcompounds that aremutually and cova-lently reactive fromtwo separate but com-plementary chemicallibraries.The enzymecan then catalyse theircovalent reaction into asingle new compound thatinhibits the enzyme with higheraffinity18.

Finally, we must exploit microbial diver-sity better.Many drug leads have been natur-al products developed from one bacterialorder, Actinomycetales19. But most of themicrobial universe remains unexplored,because most microbial speciesremain to be cultured. Forexample, studies of themicrobes in soil8,19 and waterand the viruses that prey onthem20 could reveal many com-pounds and enzymes that mayhelp a given species competewith others in its environment.These natural products canteach us a great deal about microbial vulner-abilities and how to exploit them.

Impossible? Think againIs it hopelessly unrealistic to imagine not-for-profit drug companies working in asmart regulatory environment, applyingfresh scientific approaches to antibioticdevelopment? Perhaps the most challengingaspect of this three-part vision is the notionof a not-for-profit drug company. Yet, atleast three models of public–private partner-ships for development of anti-infectives21 are

up and running. Each isdevoted to one or a small

number of infectiousdiseases. Their growingsuccess22 suggests thatit is possible toinvolve privateindustry in workthat society needs,but the market does

not competitivelyreward.

One model, funded insubstantial part by the Bill

and Melinda Gates Foundationand the Rockefeller Foundation,

involves small, not-for-profit drug companiesthat are virtual and physically distributed.TheMedicines for Malaria Venture(www.mmv.org) takes ideas, hits or leads,mainly from academic scientists, and uses a

contract mechanism to fundmedicinal chemistry and phar-macology in the labs of other aca-demics or pharmaceuticalfirms21. Similar approaches aretaken by the Global Alliance forTuberculosis Drug Development(www.tballiance.org) and theDrugs for Neglected DiseasesInitiative (www.dndi.org)21.

A second model of public–private partnerships involves on-site research anddevelopment funded by a major pharmaceu-tical company in conjunction with a publicpartner, such as The Novartis Institute forTropical Diseases. This is funded jointly byNovartis and the Singapore Economic Devel-opment Board (www.nitd.novartis.com). Athird model, also funded in substantial part by the Gates Foundation, is represented byThe Institute for OneWorld Health(www.oneworldhealth.org). This agencyuses donated intellectual property to operate

a small, on-site, not-for-profit drug companythat prepares vaccines or drugs for malaria,leishmaniasis, trypanosomiasis, helminthinfections and diarrhoeal diseases. Finally,the biotechnology industry appears to bepositioning itself to contribute to the public–private partnership model throughanother Gates Foundation-funded initia-tive, BIO Ventures for Global Health(www.bvgh.org).

All sectors of society, including the pharmaceutical industry, have a major stakein the control of infectious diseases, not onlyfor medical reasons, but also for global economic development and security23. It is in the interest of both rich and poor societiesthat initiatives such as those described abovegrow by orders of magnitude and broadentheir scope to include all major infectiousdiseases that the pharmaceutical industrydoes not adequately address. ■

Carl Nathan is in the Department of Microbiology& Immunology, Weill Cornell Medical College,and Programs in Immunology and MolecularBiology, Weill Graduate School of MedicalSciences of Cornell University, 1300 York Avenue, New York 10021, USA.1. Projan, S. J. Curr. Opin. Microbiol. 6, 427–430 (2003).

2. Wenzel, R. P. N. Engl. J. Med. 351, 523–526 (2004).

3. Infectious Diseases Society of America. Bad Bugs, No Drugs;

www.idsociety.org/Template.cfm?Section=Home&CONTENT

ID=7455&TEMPLATE=/ContentManagement/ContentDisplay.

cfm (2004).

4. Centers for Disease Control and Prevention. A Public Health

Action Plan to Combat Antimicrobial Resistance ;

www.cdc.gov/drugresistance/actionplan/html/index.htm (2000).

5. Coates, A., Hu, Y., Bax, R. & Page, C. Nature Rev. Drug Discov.

1, 895–910 (2002).

6. World Health Organization Draft Global Strategy for the

Containment of Antimicrobial Resistance ; www.who.int/

emc/amr.html (2001).

7. Service, R. F. Science 303, 1796–1799 (2004).

8. Walsh, C. Nature Rev. Microbiol. 1, 65–70 (2003).

9. Miesel, L., Greene, J. & Black, T. A. Nature Rev. Genet. 4,

442–456 (2003).

10.Global Alliance for Tuberculosis Drug Development.

The Economics of TB Drug Development; www.tballiance.org/

7_6D_Publications.asp (2001).

11.Chopra, I., Hodgson, J., Metcalf, B. & Poste, G. J. Am. Med.

Assoc. 275, 401–403 (1996).

12.Darwin, K. H., Ehrt, S., Gutierrez-Ramos, J. C., Weich, N.

& Nathan, C. F. Science 302, 1963–1966 (2003).

13.Bryk, R., Lima, C. D., Erdjument-Bromage, H., Tempst, P. &

Nathan, C. Science 295, 1073–1077 (2002).

14.Schnappinger, D. et al. J. Exp. Med. 198, 693–704 (2003).

15.Sassetti, C. M., Boyd, D. H. & Rubin, E. J. Mol. Microbiol.

48, 77–84 (2003).

16.Tong, A. H. et al. Science 303, 808–813 (2004).

17.Fidock, D. A., Rosenthal, P. J., Croft, S. L., Brun, R. & Nwaka, S.

Nature Rev. Drug Discov. 3, 509–520 (2004).

18.Manetsch, R. et al. J. Amer. Chem. Soc. 126, 12809–12819(2004).

19.Keller, M. & Zengler, K. Nature Rev. Microbiol. 2, 141–150

(2004).

20.Liu, J. et al. Nature Biotechnol. 22, 185–191 (2004).

21.Nwaka, S. & Ridley, R. G. Nature Rev. Drug Discov. 2, 919–928

(2003).

22.Vennerstrom, J. L. et al. Nature 430, 900–904 (2004).

23.World Health Organization, Commisssion on Macroeconomics

and Health Macroeconomics and Health: Investing in Health for

Economic Development; www.cmhealth.org (2001).

AcknowledgementsThanks to A. Apt, K. Deitsch, H. Djaballah, B. Ganem,

W. Jorgensen, T. Kapoor, M. MacCoss, V. Mizrahi, S. Projan,

L. Quadri, K. Rhee, M. Rosenberg, D. Russell, D. Scheinberg,

D. Schnappinger, D. Tan, T. Templeton and P. van Helden for

stimulating discussions. Special thanks are due to P. Davies, S. Ehrt,

M. Glickman, B. Kelly, J. McKinney and S. Nwaka.

commentary

902 NATURE | VOL 431 | 21 OCTOBER 2004 | www.nature.com/nature

P.D

UM

AS/

EU

RE

LIO

S/SP

L

“It makes no sense touse twenty-first centurytechnology to developdrugs targeted at specific infections,whose diagnosis isdelayed by nineteenth-century methods.”

New approaches to screening chemical librariesare needed to develop antibiotics.

BSI

P,C

HA

SSN

ET

/SP

L



Documents

Antibiotics at the crossroads