REVIEWS

FINDING NEW TRICKS FOR OLD DRUGS: AN EFFICIENT ROUTE FOR PUBLICSECTOR DRUG DISCOVERY Kerry A. O’Connor and Bryan L. Roth Abstract | With the annotation of the human genome approaching completion, public-sector researchers — spurred in part by various National Institutes of Health Roadmap Initiatives — have become increasingly engaged in drug discovery and development efforts. Although large and diverse chemical libraries of ‘drug-like’ compounds can be readily screened to yield chemically novel scaffolds, transforming these ‘chemical probes’ into drugs is a daunting endeavour. A more efficient approach involves screening libraries of approved and off-patent medications; both phenotypic- and molecular target-based screening of ‘old drugs’ can readily yield compounds that could be immediately used in clinical trials. Using case studies, we describe how this approach has rapidly identified candidate medications suitable for clinical trials in disorders such as progressive multifocal leukoencephalopathy and amyotrophic lateral sclerosis. This approach has also led to the discovery of the molecular targets responsible for serious drug side effects, thereby allowing efficient ‘counter-screening’ to avoid these side effects.

Departments of Biochemistry, Psychiatry, Neurosciences, Comprehensive Cancer Center and National Institute of Mental Health Psychoactive Drug Screening Program, 2109 Adelbert Road, Case Western Reserve University Medical School, Cleveland, Ohio 44106, USA. Correspondence to B.L.R. e-mail: [email protected] doi:10.1038/nrd1900

Now that the human genome has been sequenced and its annotation is approaching completion1, publicsector researchers have been urged to focus their attention on exploiting this information for the purposes of drug discovery2,3. The National Institutes of Health’s (NIH) Molecular Libraries Initiative, for example, proposes “to expand the availability, flexibility, and use of small-molecule chemical probes for basic research”3— an effort that has become the subject of intense debate within academia and the private sector4–6. Although efforts such as the NIH Molecular Libraries Initiative do not propose to provide new drugs for human disease — except in “exceptional circumstances”3 — it is likely that many ‘drug-like’ molecules will be discovered. It is implied that some of the ‘chemical probes’ discovered and validated by these emerging public efforts will eventually be optimized by commercial partners for therapeutic uses. In large measure, the reticence for entering fullfledged drug discovery and development efforts in the public sector stems mainly from the recognition of the

NATURE REVIEWS | DRUG DISCOVERY

enormous costs and risks associated with therapeutic drug discovery. Indeed, current estimates for successful launch of a single new medication are in excess of US$800 million7 — nearly an order of magnitude higher than that for the NIH Molecular Libraries Initiative. Given the extraordinarily high costs and risks associated with therapeutic drug discovery, it is exceedingly unlikely that any single public-sector research group will successfully see a novel chemical ‘probe’ become a ‘drug’. The main approach of the NIH Molecular Libraries Initiative and similar public-sector small-moleculebased screening centres is to screen vast libraries of chemically diverse scaffolds for novel actions3,6. Typically, either PHENOTYPIC SCREENS or MOLECULAR TAR GETBASED SCREENS are performed in a high-throughput screening (HTS)-like fashion with more than 100,000 chemically diverse compounds screened and ‘hits’ subsequently identified and validated (FIG. 1). Currently, a large number of public-sector groups TABLE 1 seem

VOLUME 4 | DECEMBER 2005 | 1005

REVIEWS

a

b Multiple drug plates containing chemically diverse

Plate cells and incubate with various compounds

compounds

Analyse phenotypic changes using microscopy

No effect

Phenotype 1: cell death Screen against a single molecular target

Find ‘hits’, optimize and validate data

Phenotype 2: cell growth

Custom 2 1

3

4

5

Plate 1 6 7

8

9

10

11

12

Deconvolute data to determine molecular target producing phenotypic change

Phenotype 3: change in cellular morphology

m1 receptor partial agonists

A B C D E F

Find ‘hits’, optimize and validate data

G H

Figure 1 | The road to drug discovery: phenotypic versus single-target approaches. a | The phenotypic approach begins with the hypothesis-based selection of compounds to test. This selection process can be applied to drugs currently on the market. Compounds can be selected for their binding profiles with certain proteins of interest (for example, transporters and receptors), known therapeutic uses, or structural similarities to known therapeutic compounds. Once selected, these compounds are then tested against a cellular read-out to identify ‘hits’. A ‘hit’ is a compound that has activity at the particular phenotypic readout examined. The compounds can be further optimized if the molecular target can be identified. Alternatively, if the compounds used already represent approved medications, investigators can directly attempt in vivo validation studies in suitable animal models and/or small-scale proof-of-concept clinical trials. b | The single-molecular-target approach is the typical approach in which a library of approved medications is simply screened for interactions at a single molecular target. ‘Hits’ can be further optimized or used directly for in vivo studies. Shown are typical screening results from a campaign we undertook to identify muscarinic M1 receptor partial agonists for the treatment of cognitive impairment that can occur in schizophrenia. N-desmethyl-clozapine — a major clozapine metabolite — was identified as a potent muscarinic M1 receptor agonist50.

PHENOTYPIC SCREENS

A type of chemical screen in which cell responses, such as motility or membrane ruffling, are monitored as an index of drug action. Frequently, phenotypic screens are carried out using high-throughput microscopy and automated image capture and analysis. MOLECULAR TARGETBASED SCREENS

A type of chemical screen in which a single molecular target, such as a receptor or enzyme, is screened using a simplified readout such as fluorescence.

1006 | DECEMBER 2005

to have the capacity to carry out near-industrial scale HTS campaigns. An outgrowth of these efforts has been the emerging field of chemical genomics in which chemical probes are used “for activating or inactivating protein functions, thereby providing resources that help discern the functions of gene products in normal and disease cells, as well as in tissues”6,8. An alternative approach probes multiple targets simultaneously using smaller libraries of chemically diverse compounds. One approach that we have pioneered has been dubbed RECEPTOROMICS, in which, ultimately, the entire complement of ‘receptors’ in the genome is screened in a massively parallel fashion9,10. As will be discussed below, this approach has yielded several important discoveries that have accelerated therapeutic drug discovery efforts. In a similar vein, KINOMICS has emerged as an approach for systematically identifying and screening the full complement of kinases for drug discovery purposes11–13. As receptors

| VOLUME 4

and kinases represent the ‘most druggable’ targets in the genome, representing 5% and 2.8% of the genome, respectively14, screens focused on these segments of the genome are most likely to yield new drugs14. Both of these more focused approaches suffer from the risks related to optimizing and validating novel chemical probes for therapeutic uses. We would like to highlight an approach more likely to be successful in reaching the ultimate goal of providing new drugs, one in which already available medications — most of which are off-patent — are simultaneously screened using various in vitro and in vivo model systems. This approach utilizes existing medications that are subsequently used as probes for preclinical molecular target- or phenotype-based drug discovery efforts. This differs from the standard ‘repositioning’ approaches that have recently been reviewed15. Repositioning typically occurs when an interesting side effect of an approved medication is

www.nature.com/reviews/drugdisc

REVIEWS

Table 1 | Selected public-sector small-molecule screening resources Name of screening resource

Website of resource

Users

Comments

NIMH Psychoactive Drug Screening Program

http://pdsp.cwru.edu

A/G/NP

Free and confidential screening to qualified users

ChemBank

http://chembank.broad.harvard.edu

A/G/NP

Large database of chemical compounds available

NIH Chemical Genomics Center

http://www.genome.gov/12512295

G

Advertises screening capacity of up to 1 × 106 compounds per day

Scripps Research Institute HTS Facility

None yet

NP/A

To be installed summer 2005

Rockefeller University HTS Core Facility

http://www.rockefeller.edu/ highthroughput/highthroughput.php

A/NP

One of the oldest NP HTS facilities

Kansas University HTS Lab

http://www.hts.ku.edu

A

Funded by NIH COBRE mechanism

Johns Hopkins University HTS Core

http://www.molecularinteraction.org/ HIT%20center.htm

A

A, academic; C, commercial; COBRE, Centers of Biomedical Research Excellence; G, government; HTS, high-throughput screening; NIH, National Institutes of Health; NIMH, National Institute of Mental Health; NP, non-profit.

RECEPTOROMICS AND RECEPTOROME

Receptorome refers, ultimately, to the entire complement of ‘receptors in the genome’, although it is more commonly applied to that portion of the genome encoding G-proteincoupled receptors. Receptoromics represents tools and methods of analysis to study the receptorome. KINOMICS AND KINOME

The kinome refers to the entire complement of protein kinases in a particular genome. Kinomics represents tools and methods of analysis to study the kinome.

identified during clinical trials, which then provides the impetus for a new therapeutic indication. The best known example is that of sildenafil (Viagra; Pfizer), which was initially developed as an anti-angina medication but had the side effect of producing prolonged penile erections in human volunteers15. Repositioning can also arise from technology platforms established to identify repositioning opportunities, such as CombinatoRx’s cHTS system (see REF. 16). However, under either of these conditions a compound is first identified for a known target, then similar compounds are screened against that target. Our proposed approach, by contrast, blindly screens existing compounds against a multitude of targets, and therefore identifies either possible therapeutic benefits or side effects in a non-biased fashion. Both of these approaches, repositioning and screening of approved medications, benefit from the information obtained from the original clinical trials. When screening current medications for novel applications, most of the identified ‘hits’ will already have been used in humans and will have detailed information related to dosing, in vivo pharmacokinetics and toxicity, and so clinical trials can be performed sooner and more economically using the more limited resources of the public sector. According to the 2005 Pharmaceutical Industry Profile17carried out by the Pharmaceutical Researcher Manufacturers of America (PhRMA), member companies spent US$38.8 billion in research and development in 2004. One-third of these funds were spent on preclinical/prehuman research and development. In the ideal case using this paradigm, a novel target for an off-patent medication will be discovered and its therapeutic indication validated using public-sector resources. Because the medications are off-patent, the cost savings for consumers are potentially enormous because of savings related to medication cost. Furthermore, the majority of research and development costs for de novo therapeutics are due to

NATURE REVIEWS | DRUG DISCOVERY

the large dependence on medicinal chemistry, pharmacokinetic and toxicology expertise. In the public or academic setting, the infrastructure for this type of large-scale therapeutic development is absent using the approach we are describing. The benefit of the approach we describe of screening current medications for novel therapies bypasses the dependence on such an infrastructure. Additionally, because of the inherent financial opportunities associated with trials for orphan diseases (which allows for 7 years of exclusivity even with off-patent medications) abundant financial opportunities will exist for smaller companies that use this approach. In the following sections, we provide several case histories that validate this approach and suggest that this approach will allow the public sector to achieve the largest return on its investment. Case histories

Antibiotics as neuroprotective agents. The first antibiotic was identified by Alexander Fleming in 1929 when a culture plate of Staphylococcus bacteria became contaminated by a mold. Penicillin was merely the first in a series of antibiotics classified as β-lactam antibiotics, due to their functional requirement of an intact β-lactam ring structure. This class of antibiotics, which includes aminocillins, cephalosporins, carbapenems and monobactams, acts via inhibition of bacterial cell-wall synthesis. The widespread use of this class of antibiotics has grown exponentially since its discovery more than 70 years ago. According to the Centers for Disease Control and Prevention’s National Center for Health Statistics, 126 million antimicrobial prescriptions were written in ambulatory care settings in 2000 alone18. More recently, β-lactams were found to also have a role in regulating glutamate levels, via interaction with the glutamate transporter GLT1 (also known as EEAT2)19. GLT1 is responsible for the inactivation

VOLUME 4 | DECEMBER 2005 | 1007

REVIEWS and prevention of neurotoxicity via reuptake of the primary excitatory neurotransmitter glutamate20. This astroglial protein has been implicated in multiple neurological disorders, including amyotrophic lateral sclerosis (ALS), stroke, brain tumours and epilepsy21.

The discovery of the neuroprotective effects of β-lactam antibiotics represents an exemplary case study on drug screening methodology. Rothstein and colleagues19 systematically screened 1,040 FDAapproved compounds for interaction with GLT1. The screens were performed in multiple model

b Identification of JC virus co-receptor

a Normal viral life-cycle

JC virus coreceptor identified as 5-HT2A Clathrincoated pit

5-HT1A 5-HT1B 5-HT1D 5-HT1E 5-HT2A 5-HT2C 5-HT3 5-HT5 5-HT6 5-HT7 D1 D2 D3 D4 D5 α2A α2B α2C α1A α1B M1-muscarinic M2-muscarinic M3-muscarinic M4-muscarinic M5-muscarinic γ-A γ-B BZP NMDA PCP SERT NET DAT H1 H2 H3 H4 EP-3 EP-4 CB1 Ca2+ channel NAR a2/b2 NAR a2/b4 NAR a3/b2 NAR a3/b4 NAR a/b2 NAR a7 β1-AR β2-AR I1-imidazoline mGlu1 mGlu2 mGlu4

Atypicality 5-HT2A Weight gain 5-HT2C

Endosome Endocytosis of JC virus with 5-HT2A receptors

Antipsychotic D2 efficacy

Adrenergic Side effects

Degradation

Muscarinic

JC viral Transcription DNA Translation Viral replication

Lysosome Infection

Weight gain

c Infection blocked by 5-HT2A antagonist

H1

Pretreated with 5-HT2A receptor antagonist before JC virus infection Clathrincoated pit

Internalized vesicles

Internalized receptors are sorted to endosome

1 nM 10 nM 100 nM 1,000 nM 10,000 nM

Ki

Antagonist treatment causes endocytosis of cell surface 5-HT2A receptors

l e ne ine ne ine ine ine ne ine ine ido ine zine ol a az ido ep ido tiap zap zap hixe daz xap per naz i t m pr ras Zot per e r i n e o o o l o l i h a o u L ro i ip ip C Th s Q Ol Ha lup lorp Z Ar Ri Th F Ch

Endosome Lysosome Degradation

=

= ASN

Figure 2 | Identification of the 5-hydroxytryptamine 2A receptor as the JC virus co-receptor via a combination of phenotypic and receptorome-based screening reveals a novel strategy for treating progressive multifocal leukoencephalopathy. a | The presumed normal life cycle of JC virus in which it interacts with cell-surface glycoproteins51 and with an unknown G-protein-coupled receptor (GPCR) that can be blocked by antipsychotic drugs28. Following internalization via the endosome pathway, virions are released and a new cycle of infection begins leading, ultimately, to progressive multifocal leukoencephalopathy (PML). b | A receptorome profile of typical and atypical antipsychotic drugs shows a large number of potential molecular targets with which antipsychotic drugs interact, which identified the 5-hydroxytryptamine 2A (5-HT2A) receptor as the likely JC virus co-receptor. c | A likely mechanism by which 5-HT2A receptor antagonists block JC virus infection via antagonist-induced intracellular sequestration to remove potential co-receptors from the cell surface. If other GPCRs are identified as viral co-receptors, this mechanism could lead to a novel viral treatment strategy. Panel b reprinted, with permission, from REF. 10 © Macmillan Magazines Limited.

1008 | DECEMBER 2005

| VOLUME 4

www.nature.com/reviews/drugdisc

REVIEWS systems, each closely approximating in vivo conditions, to assure proper functional relevance. Initial drug screens were carried out in organotypic spinal cord slices, which mimic cellular metabolism and cell–cell interactions present in vivo. More than 20 compounds were found to increase GLT1 protein expression by more than twofold, including penicillin and cephalosporin. After discovering that this class of compounds increased protein expression, a smaller subgroup was then screened in astrocytes and non-neuronal tissues expressing the GLT1 promoter. These studies allowed determination of the specificity and time-dependent effects of β-lactams on GLT1. To further elucidate the clinical potential of this effect, in vivo assays were performed in normal rats, as well as in mouse models of neurological disease states. This series of experiments provides both molecular and in vivo evidence to support the conclusion that β-lactams could have potential as neuroprotective agents, as well as their antibiotic effects. In addition, this study provides pharmacological evidence to support this conclusion. Although β-lactams such as ceftriaxone increased GLT1 protein expression, non-β-lactam antibiotics such as fluconazole had no effect, indicating a specific effect of this sub-class of compounds19. This group subsequently investigated the direct neuroprotective effects of ceftriaxone, such as cell survival under ischaemic conditions and extended lifespan in transgenic mouse models of neurodegenerative diseases. In doing so, these studies indicate the potential use of antibiotics in slowing the progression of, or preventing, neurodegenerative diseases such as ALS19. These findings and the proposed clinical trials have also generated controversy22 because ceftriaxone was previously reported, in at least two Italian clinical trials, to be ineffective in curing ALS23,24. These earlier trials can be criticized because of inadequate statistical power, trial design (for example, not double-blinded) and trial length22. A Phase II/III ‘bridging study’ is now in preparation to begin in 2005 to test this hypothesis more adequately (see The Robert Packard Center for ALS Research at Johns Hopkins in Further information). At the very least, these studies support GLT1 as a potential molecular target for commercial drug discovery efforts for ALS and related disorders. Atypical antipsychotics, PML and Tysabri. Here the recent discovery of the potential role of serotonergic antagonists in blocking the cellular entry, and therefore spread, of the human polyomavirus JC virus (JCV), in HIV-infected patients and individuals who have been administered natalizumab (Tysabri; Elan Biogen Idec) will be discussed. In addition, the possibility that other viruses use a similar cellular apparatus to invade host cells will be discussed. Progressive multifocal leukoencephalopathy (PML), a usually fatal disorder characterized by demyelination, typically occurs in immunosuppressed

NATURE REVIEWS | DRUG DISCOVERY

patients, such as those with AIDS25. Recently, multiple cases of PML have been reported in patients treated with natalizumab for multiple sclerosis26 and Crohn’s disease and in individuals who have been treated with tumour-necrosis factor-α (TNFα) antagonists27 for rheumatoid arthritis. JCV has been shown to infect oligodendrocytes and astrocytes, cells within the central nervous system (CNS), which results in the initiation of PML (FIG. 2). Recently, Elphick and colleagues28 found that this infection of glial cells could be blocked by antipsychotic drugs (which are normally used to treat schizophrenia). Importantly, selective serotonin (5-hydroxytryptamine: 5-HT) 5-HT2A receptor inhibitors, such as M100907, along with antipsychotic drugs that also are potent 5-HT2A receptor antagonists — including clozapine and chlorpromazine (see FIG. 2 and PDSP Ki Database in Further information) — were able to block infection of glial cells by JCV. In addition, infection could be blocked by 5-HT2A receptor-neutralizing antibodies, suggesting that the 5-HT 2A receptor is probably a co-receptor for JCV (FIG. 2). Finally, Elphick et al.28 demonstrated that JCV was co-internalized with 5-HT2A receptors in cells that overexpressed receptors tagged with green fluorescent protein. It is likely that 5-HT2A receptor antagonists block JCV infection by inducing internalization of 5-HT2A receptors and loss of cell-surface receptors (FIG. 2)29, rather than by directly blocking viral receptor binding, but further studies will be needed to test this hypothesis. Previous medications developed to treat PML have largely failed due to poor bioavailability within the CNS. However, because 5-HT2A receptor antagonists are often used to treat psychological disorders, they could provide a novel direction of drug development to treat HIV-infected patients. Unfortunately, drugs such as clozapine and chlorpromazine have serious and occasionally fatal side effects and are not likely to be well tolerated by immunocompromised individuals. To identify compounds with high affinity for 5-HT 2A receptors that can be easily tolerated, we searched the National Institute of Mental Health (NIMH)-Psychoactive Drug Screening Program’s Ki Database (see Further information; FIG. 3) for compounds with sub-nanomolar affinity for 5-HT 2A receptors. Identified compounds included many atypical antipsychotic drugs, as expected, along with several antihistamines, including cyproheptadine (K i = 0.46 nM) and mirtazepine (K i = 2.0 nM), which we have identified as potent 5-HT2A receptor antagonists (FIG. 3). Cyproheptadine is a well-tolerated, generic medication frequently used to treat HIV-wasting syndrome30, which preliminary studies have indicated suppresses JCV infection in vitro (W. Atwood, personal communication). Cyproheptadine or mirtazepine would, therefore, qualify as a candidate medication for prophylactic treatment of PML, which occurs as a consequence of HIV infection as well as treatment with various biological response modifiers including natalizumab and TNFα.

VOLUME 4 | DECEMBER 2005 | 1009

REVIEWS

KiDB

PubChem

N

NIMH-PDSP 86

RFU (x 1,000)

76 66 56 46 36 26 16 0

5

10

15

20

25

30

35

40

45

50

55

60

Time (sec)

Clinical trials Pretreated with 5-HT2A receptor antagonist before JC virus infection Clathrincoated pit

Internalized vesicles

Internalized receptors are sorted to endosome =

=

Antagonist treatment causes endocytosis of cell surface 5-HT2A receptors

Endosome Lysosome Degradation

ASN

Figure 3 | Identification of a potentially prophylactic treatment for progressive multifocal leukoencephalopathy. Because antipsychotic drugs have serious and potentially life-threatening side-effects, they are not suitable for prophylactic treatment of progressive multifocal leukoencephalopathy (PML). A search of a public database of drug–receptor affinity values (see PDSP Ki Database Search Page in Further information) facilitated the identification of cyproheptadine (Periactin) — an off-patent sub-nanomolar affinity 5-hydroxytryptamine 2A (5-HT2A) receptor ligand (structure via link to PubChem). The middle panel shows that 1 nM cyproheptadine is able to abolish 5-HT2A receptor signalling, and the bottom panel shows the presumed mechanism by which cyproheptadine might block JC virus infection (courtesy of National Institute of Mental Health-Psychoactive Drug Screening Program (NIMH-PDSP)). Clinical trials are currently being planned to test this hypothesis. Because PML is now also linked to treatment with various immunomodulators, including natalizumab (Tysabri; Elan/Biogen Idec), prophylactic treatment with 5-HT2A receptor antagonists such as cyproheptadine could be considered. RFU, relative fluorescent units.

1010 | DECEMBER 2005

| VOLUME 4

www.nature.com/reviews/drugdisc

REVIEWS

Table 2 | Commercial and non-commercial sources of available medication Name

Library characteristics

Website

Prestwick Chemical Co.: Prestwick Chemical Library

1,120 off-patent with >85% marketed pharmaceuticals

http://www.prestwickchemical.com

Microsource: The Spectrum Collection

2,000 biologically active and structurally diverse compounds from libraries of known drugs, experimental bioactives and pure natural products

http://www.msdiscovery.com/spect.html

Sequoia

A large collection of available medications — many currently ‘on-patent’ and includes compounds not normally available elsewhere

http://www.seqchem.com

NIH Brain Bioactive Compound Collection (BIBCC)

A planned large collection of available medications, known CNS active compounds as well as compounds with likely CNS activity. Collection will be available via the NIH Roadmap Initiative.

http://nihroadmap.nih.gov

CNS, central nervous system; NIH, National Institutes of Health.

COUNTERSCREENING

The approach of screening candidate medications against an array of molecular targets to identify ‘off-target’ actions.

When side-effects are lethal: ‘Fen/phen’ and the human valvulopathy receptor. Recently, the inability of regulatory agencies to rapidly identify the serious and potentially life-threatening side effects of approved drugs has received a great deal of attention. In large measure, this failure comes from ignorance of the molecular and cellular mechanisms responsible for serious side effects and the subsequent inability to reliably predict them. Large pharmaceutical companies do not ordinarily have the scientific resources nor the motivation to uncover the cellular and molecular mechanisms of drug toxicities of approved drugs. Additionally, it does not seem that the US FDA and similar regulatory agencies have the adequate scientific and monetary resources to perform the sort of basic discovery-based research needed to reveal the mechanisms responsible for serious side effects of approved medications. The recent withdrawal of rofecoxib (Vioxx; Merck) and many other cyclooxygenase-2 (COX2) inhibitors because of serious cardiovascular side effects is a pertinent example31. The inability to identify the ‘toxic target’ responsible for serious side effects of prescribed medications makes it impossible to avoid these targets via COUNTERSCREENING in subsequent drug discovery efforts. We will use the case of the ‘Fen/Phen’ (fenfluramine/phentermine) appetite suppressant combination as a pertinent case in which the identification of the ‘toxic target’ using available medications as probes has allowed the development of a new class of appetite suppressant medications. In 1997, Connolly and colleagues32 described valvular heart disease (VHD) as a common and serious side effect of the widely prescribed appetite suppressant combination of fenfluramine and phentermine. Later reports indicated that VHD was specifically associated with fenfluramine and its isomer dex-fenfluramine33 — findings that led to the voluntary withdrawal of fenfluramine and dex-fenfluramine and >US$20 billion in potential liability against Wyeth Pharmaceuticals. However, it was not until 3 years later that a mechanism was identified.

NATURE REVIEWS | DRUG DISCOVERY

Using a systematic approach, we analysed the receptorome profile of fenfluramine and its active metabolite norfenfluramine34. Additionally, we examined several other off-patent medications known to induce VHD, including methysergide and its active metabolite methylergonovine34,35. Because fenfluramine was known to induce 5-HT release, we also screened other drugs that elevate levels of 5-HT but are not known to be associated with VHD, including fluoxetine (Prozac; Eli Lilly) and its main metabolite norfluoxetine. This receptorome screen allowed the rapid identification of the 5-HT2B receptor as the likely culprit for the side effects of fenfluramine and other drugs that induce cardiac valvulopathy. A similar conclusion was reported by Fitzgerald and colleagues36 who independently showed an association between 5-HT2B receptor agonist potency and valvulopathic propensity of several approved medications. Subsequently, the 5-HT2B receptor was identified as the likely molecular target responsible for fenfluramine-induced pulmonary hypertension — a fatal side effect of fenfluramine administration37. Since then, we used a commercially available library (see TABLE 2 for sources of libraries containing approved medications) to determine whether other approved medications might be likely to induce VHD of the fenfluramine-type by virtue of 5-HT2B receptor agonism. Pergolide and dihydroergotamine were identified by us as potential valvulopathogens35, whereas cabergoline was identified as a 5-HT2B receptor agonist by others38. Subsequently these medications have been identified by the US FDA and others as valvulopathogens and their use has been severely restricted39–41. The receptorome screen that identified the 5-HT2B receptor as the target for the side effects of norfenfluramine also identified the 5-HT2C receptor as the likely site responsible for the appetite-suppressing actions of norfenfluramine (FIG. 4). Indeed, for many years it had been appreciated by others that 5-HT2C receptor agonists are anorectic42,43 and that the anorectic actions of fenfluramine in vivo are produced, in part, by

VOLUME 4 | DECEMBER 2005 | 1011

REVIEWS that APD 356 suppresses appetite and induces weight loss without detectable echocardiographic changes in obese humans (see Arena Pharmaceuticals in Further information). From Alzheimer’s disease to cancer

With the commercial availability of libraries of approved medications for in vitro and in vivo

EP-3 EP-1 Adenosine A2 Adenosine A1 α1B α1A β2 β1 α2C α2B α2A AMPA PCP NMDA H2 H1 Oxytocin V3 V2 V1 BZP GABAB GABAA DAT NET SERT M5 M4 M3 M2 M1 Kappa Delta MU D5 D4 D3 D2 D1 5-HT7 5-HT6 5-HT5 5-HT2C 5-HT2B 5-HT2A 5-HT1E 5-HT1D 5-HT1B 5-HT1A

5-HT2C receptor agonism44. These observations have led to the proposition that 5-HT2C receptor-selective agonists that are devoid at 5-HT2B receptor activity will be safe and effective appetite suppressants 45. Indeed, many pharmaceutical companies now have active programmes for developing 5-HT2C receptor agonists devoid of activity at 5-HT2B receptors, and one, Arena Pharmaceuticals, has recently reported

(+/–) Fenfluramine HCl

Ki (nM) 0.1 1 10 100 1,000 10,000

5-HT2B

(+) Fenfluramine HCl (–) Fenfluramine HCl Methysergide maleate Ergotamine tartrate Phentermine Fluoxetine Norfluoxetine (+/–) Norfenfluramine (+) Norfenfluramine (–) Norfenfluramine

Intraperitoneal accumulation (% of 5-HT max)

Side effects = 5-HT2B

Efficacy = 5-HT2C NPY

120

Arcuate nucleus

5-HT Pergolide DHE

100

Tuberomammilary nucleus

80

Leptin 60 40 20

Raphe nucleus

H3

0

Histamine

–20 –13

–12

–11

–10

–9

–8

–7

log [drug]

US FDA ‘black box’ warning

–6

–5

–4

H1

–3

5-HT2C 5-HT2C receptor activation suppresses appetite [AU: OK?] APD 356

NPY Paraventricular hypothalamic nucleus

Figure 4 | How to make a safe and effective anorectic agent. The top panel shows the receptorome profile that identified the 5-hydroxytryptamine 5-HT2B receptor as the molecular target responsible for the side effects of fenfluramine and the 5-HT2C receptor as the molecular target likely to be responsible for the anorectic actions of fenfluramine. Subsequently, the medications pergolide and dihydroergotamine (DHE) were identified in vitro and in humans as being valvulopathic. Importantly, Arena Pharmaceuticals has developed a 5-HT2C receptor-selective agonist that proved effective at suppressing appetite in a Phase II clinical trial without inducing heart-valve pathology as assessed by echocardiography. The bottom right figure represents multiple pathways involved in appetite control in mammals. In addition to the 5-HT2C receptor pathways described, additional hormones and neurochemicals have been described in the regulation of appetite control. Leptin interacts with leptin receptors within the central nervous system resulting in stimulation of hypothalamic anorexigenic pathways. Increased levels of histamine have also ben shown to suppress appetite via these pathways.

1012 | DECEMBER 2005

| VOLUME 4

www.nature.com/reviews/drugdisc

REVIEWS screening campaigns, potential therapies have been identified for several indications. For example, non-steroidal anti-inflammatory drugs (NSAIDs; for example, ibuprofen) and peroxisome proliferator-activated receptor-γ (PPARγ; thiazolidinediones) inhibitors have emerged as potential therapeutic agents for Alzheimer’s disease46,47 suitable for clinical trials. A different approach by Kukar et al.48, using a small library of 300 compounds (of which many were approved medications), identified several new compounds that mimic Alzheimer’s disease by elevating amyloid 42-β production. This chemical genetic approach has identified potential pathways involved in Alzheimer’s disease that could subsequently be modulated by small molecules. Likewise in cancer chemotherapeutics, the generic medication fluphenazine — which has potent activities at various 5-HT and dopamine receptors — was identified as a candidate medication for multiple myeloma49 with clinical trials in progress. Immune Control has proposed that the 5-HT1B receptor antagonist activity of fluphenazine (Ki~300 nM; see PDSP Ki Database Search Page in Further information) might be responsible for its actions. This approach has begun to be widely adopted by the commercial

1.

2. 3.

4. 5. 6.

7. 8.

9. 10.

11.

12.

13.

14. 15.

16.

International Human Genome Sequencing Consortium. Finishing the euchromatic sequence of the human genome. Nature 431, 931–945 (2004). Zerhouni, E. Medicine. The NIH Roadmap. Science 302, 63–72 (2003). Austin, C. P., Brady, L. S., Insel, T. R. & Collins, F. S. NIH Molecular Libraries Initiative. Science 306, 1138–1139 (2004). Duyk, G. Attrition and translation. Science 302, 603–605 (2003). Couzin, J. Molecular medicine. NIH dives into drug discovery. Science 302, 218–221 (2003). Strausberg, R. L. & Schreiber, S. L. From knowing to controlling: a path from genomics to drugs using small molecule probes. Science 300, 294–295 (2003). Rawlins, M. D. Cutting the cost of drug development? Nature Rev. Drug Discov. 3, 360–364 (2004). Haggarty, S. J., Clemons, P. A. & Schreiber, S. L. Chemical genomic profiling of biological networks using graph theory and combinations of small molecule perturbations. J. Am. Chem. Soc. 125, 10543–10545 (2003). Armbruster, B. N. & Roth, B. L. Mining the receptorome. J. Biol. Chem. 280, 5129–5132 (2005). Roth, B. L., Sheffler, D. J. & Kroeze, W. K. Magic shotguns versus magic bullets: selectively non-selective drugs for mood disorders and schizophrenia. Nature Rev. Drug Discov. 3, 353–359 (2004). Coito, C., Diamond, D. L., Neddermann, P., Korth, M. J. & Katze, M. G. High-throughput screening of the yeast kinome: identification of human serine/threonine protein kinases that phosphorylate the hepatitis C virus NS5A protein. J. Virol. 78, 3502–3513 (2004). Johnson, S. A. & Hunter, T. Kinomics: methods for deciphering the kinome. Nature Methods 2, 17–25 (2005). In references 11 and 12 the approach for screening the ‘kinome’ is validated. Manning, G., Whyte, D. B., Martinez, R., Hunter, T. & Sudarsanam, S. The protein kinase complement of the human genome. Science 298, 1912–1934 (2002). Hopkins, A. L. & Groom, C. R. The druggable genome. Nature Rev. Drug Discov. 1, 727–730 (2002). Ashburn, T. T. & Thor, K. B. Drug repositioning: identifying and developing new uses for existing drugs. Nature Rev. Drug Discov. 3, 673–683 (2004). Keith, C. T., Borisy, A. A. & Stockwell, B. R. Multicomponent therapeutics for networked systems. Nature Rev. Drug Discov. 4, 71–78 (2005).

NATURE REVIEWS | DRUG DISCOVERY

sector15. Of recent interest, Melior Discovery (see Melior Discovery in Further information) has developed a platform of in vivo assays to identify promising compounds either alone or in combination with existing compounds (via multiplexing). This disease model-based assay has allowed Melior to identify an entirely new indication for its first drug candidate, MLR-1023. This assay system allows for rapid identification and development of novel therapeutics. Conclusions

It is our contention that the widespread use of the screening strategies highlighted here will yield the rapid identification of novel therapeutic targets that can then be exploited using more conventional drug discovery efforts by commercial entities. Additionally, the discovery of new uses for old medications will enhance the ability of non-industrial entities (for example, academic, governmental and patient advocacy groups) to bring ‘new’ and affordable treatment options forward for a number of serious diseases. We suggest that the widespread strategy of ‘teaching old drugs new tricks’ will lead to a win–win situation for the public and private sectors and, most importantly, healthcare consumers.

17. Pharmaceutical Researcher Manufacturers of America. Pharmaceutical Industry Profile 2005 [online], (2005). 18. McCaig, L. F., Besser, R. E. & Hughes, J. M. Antimicrobial drug prescription in ambulatory care settings, United States, 1992–2000. Emerging Infect. Dis. 9, 432–437 (2003). 19. Rothstein, J. D. et al. Beta-lactam antibiotics offer neuroprotection by increasing glutamate transporter expression. Nature 433, 73–77 (2005). A pivotal paper in which cleverly designed phenotypic screens are used to identify a promising treatment strategy for a serious disease involving drugs already approved for human use. 20. Rothstein, J. D. et al. Knockout of glutamate transporters reveals a major role for astroglial transport in excitotoxicity and clearance of glutamate. Neuron 16, 675–686 (1996). 21. Maragakis, N. J., Dykes-Hoberg, M. & Rothstein, J. D. Altered expression of the glutamate transporter EAAT2b in neurological disease. Ann. Neurol. 55, 469–477 (2004). 22. Beghi, E., Bendotti, C. & Mennini, T. Merits of a new drug trial for ALS? Science 308, 632–633; author reply 632–633 (2005). 23. Couratier, P., Vallat, J. M., Merle, L., Preux, P. M. & Hugon, J. Report of six sporadic cases of ALS patients receiving ceftriaxone. Therapie 49, 146 (1994). 24. Carelli, V., Liguori, R., Cordivari, C., Bianchedi, G. & Montagna, P. Ceftriaxone is ineffective in ALS. Ital. J. Neurol. Sci. 15, 66 (1994). 25. Koralnik, I. J. New insights into progressive multifocal leukoencephalopathy. Curr. Opin. Neurol. 17, 365–370 (2004). 26. Schreiner, B., Kieseier, B. C., Hartung, H. P., Hohlfeld, R. & Wiendl, H. Blocking adhesion molecules with natalizumab in multiple sclerosis. Nervenarzt 76, 999–1005 (2005). 27. Imperato, A. K., Bingham, C. O. & Abramson, S. B. Overview of benefit/risk of biological agents. Clin. Exp. Rheumatol. 22, S108–S114 (2004). 28. Elphick, G. F. et al. The human polyomavirus, JCV, uses serotonin receptors to infect cells. Science 306, 1380–1383 (2004). In this paper, the authors use a combination of phenotypic and molecular-target-based screening approaches to identify the co-receptor for JC virus, the causative agent of PML. 29. Bhatnagar, A. et al. The dynamin-dependent, arrestinindependent internalization of 5-hydroxytryptamine 2A (5-HT2A) serotonin receptors reveals differential sorting of

30.

31. 32.

33. 34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

arrestins and 5-HT2A receptors during endocytosis. J. Biol. Chem. 276, 8269–8277 (2001). Summerbell, C. D., Youle, M., McDonald, V., Catalan, J. & Gazzard, B. G. Megestrol acetate versus cyproheptadine in the treatment of weight loss associated with HIV infection. Int. J. STD AIDS 3, 278–280 (1992). Couzin, J. Drug safety. Withdrawal of Vioxx casts a shadow over COX-2 inhibitors. Science 306, 384–385 (2004). Connolly, H. M. et al. Valvular heart disease associated with fenfluramine-phentermine. N. Engl. J. Med. 337, 581–588 (1997). Devereux, R. B. Appetite suppressants and valvular heart disease. N. Engl. J. Med. 339, 765–766 (1998). Rothman, R. B. et al. Evidence for possible involvement of 5-HT2B receptors in the cardiac valvulopathy associated with fenfluramine and other serotonergic medications. Circulation 102, 2836–2841 (2000). Setola, V. et al. 3,4-Methylenedioxymethamphetamine (MDMA, “Ecstasy”) induces fenfluramine-like proliferative actions on human cardiac valvular interstitial cells in vitro. Mol. Pharmacol. 63, 1223–1229 (2003). Fitzgerald, L. W. et al. Possible role of valvular serotonin 5-HT2B receptors in the cardiopathy associated with fenfluramine. Mol. Pharmacol. 57, 75–81 (2000). Launay, J. M. et al. Function of the serotonin 5hydroxytryptamine 2B receptor in pulmonary hypertension. Nature Med. 8, 1129–1135 (2002). Newman-Tancredi, A. et al. Differential actions of antiparkinson agents at multiple classes of monoaminergic receptor. III. Agonist and antagonist properties at serotonin, 5-HT1 and 5-HT2, receptor subtypes. J. Pharmacol. Exp. Ther. 303, 815–822 (2002). Flowers, C. M., Racoosin, J. A., Lu, S. L. & Beitz, J. G. The US Food and Drug Administration’s registry of patients with pergolide-associated valvular heart disease. Mayo Clin. Proc. 78, 730–731 (2003). Van Camp, G. et al. Treatment of Parkinson’s disease with pergolide and relation to restrictive valvular heart disease. Lancet 363, 1179–1183 (2004). Horvath, J. et al. Severe multivalvular heart disease: a new complication of the ergot derivative dopamine agonists. Mov. Disord. 19, 656–662 (2004). Clineschmidt, B. V., Hanson, H. M., Pflueger, A. B. & McGuffin, J. C. Anorexigenic and ancillary actions of MK-212 (6-chloro-2-(1-piperazinyl)-pyrazine; CPP). Psychopharmacology (Berl.) 55, 27–33 (1977). Dourish, C. T. Multiple serotonin receptors: opportunities for new treatments for obesity? Obes. Res. 3, S449–S462 (1995).

VOLUME 4 | DECEMBER 2005 | 1013

REVIEWS

44. Vickers, S. P., Clifton, P. G., Dourish, C. T. & Tecott, L. H. Reduced satiating effect of d-fenfluramine in serotonin 5-HT2C receptor mutant mice. Psychopharmacology (Berl.) 143, 309–314 (1999). 45. Roth, B. L. & Shapiro, D. A. Insights into the structure and function of 5-HT2 family serotonin receptors reveal novel strategies for therapeutic target development. Exp. Opin. Ther. Targets 5, 685–695 (2001). 46. Combs, C. K., Bates, P., Karlo, J. C. & Landreth, G. E. Regulation of β-amyloid stimulated proinflammatory responses by peroxisome proliferator-activated receptor alpha. Neurochem. Int. 39, 449–457 (2001). 47. Yan, Q. et al. Anti-inflammatory drug therapy alters β-amyloid processing and deposition in an animal model of Alzheimer’s disease. J. Neurosci. 23, 7504–7509 (2003). 48. Kukar, T. et al. Diverse compounds mimic Alzheimer disease-causing mutations by augmenting Aβ42 production. Nature Med. 11, 545–550 (2005). 49. Glaser, V. Immune Control pursues seotonin antagonists. Genet. Eng. News 24 [online], (2004).

1014 | DECEMBER 2005

| VOLUME 4

50. Davies, M. A., Compton-Toth, B. A., Hufeisen, S. J., Meltzer, H. Y. & Roth, B. L. The highly efficacious actions of N-desmethylclozapine at muscarinic receptors are unique and not a common property of either typical or atypical antipsychotic drugs: is M1 agonism a pre-requisite for mimicking clozapine’s actions? Psychopharmacology (Berl.) 178, 451–460 (2005). 51. Eash, S. et al. Differential distribution of the JC virus receptor-type sialic acid in normal human tissues. Am. J. Pathol. 164, 419–428 (2004).

Acknowledgements The authors thank D. Sheffler for figures related to the JC virus life cycle, W. Kroeze for editorial assistance and grants from the National Institutes of Health and the National Institute of Mental Health-Psychoactive Drug Screening Program for supporting the work in the authors’ lab.

Competing interests statement The authors declare competing financial interests: see Web version for details.

Online links DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene GLT1 OMIM: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM ALS | Crohn’s disease | epilepsy | stroke FURTHER INFORMATION Arena Pharmaceuticals: http://www.arenapharm.com PDSP Ki Database: http://kidb.case.edu PDSP Ki Database Search Page: http://kidb.case.edu/pdsp.php The Robert Packard Center for ALS Research at Johns Hopkins: http://www.alscenter.org/clinical_trials Access to this interactive links box is free online.

www.nature.com/reviews/drugdisc