Challenges Ab

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Ann. N.Y. Acad. Sci.

ISSN 0077-8923

A N N A L S O F T H E N E W Y O R K A C A D E M Y O F SC I E N C E S
Issue: Antimicrobial Therapeutics Reviews

Challenges of antibacterial discovery revisited


Michael N. Gwynn, Alison Portnoy, Stephen F. Rittenhouse, and David J. Payne
Antibacterial Discovery Performance Unit, Infectious Diseases Center of Excellence for Drug Discovery, GlaxoSmithKline,
Collegeville, Pennsylvania

Address for correspondence: David J. Payne, 1250 South Collegeville Road, Collegeville, Pennsylvania 19426.
David.J.Payne@gsk.com

The discovery of novel antibiotic classes has not kept pace with the growing threat of bacterial resistance. Antibiotic
candidates that act at new targets or via distinct mechanisms have the greatest potential to overcome resistance;
however, novel approaches are also associated with higher attrition and longer timelines. This uncertainty has
contributed to the withdrawal from antibiotic programs by many pharmaceutical companies. Genomic approaches
have not yielded satisfactory results, in part due to nascent knowledge about unprecedented molecular targets,
the challenge of achieving antibacterial activity by lead optimization of enzyme inhibitors, and the limitations of
compound screening libraries for antibacterial discovery. Enhanced diversity of compound screening banks, entry
into new chemical space, and new screening technologies are currently being exploited to improve hit rates for
antibacterial discovery. Antibacterial compound lead optimization faces hurdles associated with the high plasma
exposures required for efficacy. Lead optimization would be enhanced by the identification of new antibiotic classes
with improved tractability and by expanding the predictability of in vitro safety assays. Implementing multiple
screening and target identification strategies is recommended for improving the likelihood of discovering new
antibacterial compounds that address unmet needs.

Keywords: antibacterial discovery; novel antibiotic; gene essentiality; HTS; DNA-encoded library; boron chemistry

decades by synthetic tailoring into new-generation


Introduction
agents1 that generally offer incremental improve-
The effectiveness of all antibiotic agents in cur- ments over existing therapies.
rent use is compromised by the prevalence of Novel-class antibiotic agents have greater poten-
drug resistance.1,2 Increasingly, multiple drug- tial to address the deficiencies of existing classes
resistant strains of Gram-positive and Gram- of antibiotics and are vitally important to ad-
negative pathogens (so called “superbugs”) are en- dress the ever-increasing problem of bacterial resis-
countered that are resistant to even last-line agents tance.6 Unfortunately, development and marketing
(e.g., vancomycin).2,3 The main classes of antibi- approval of new antibiotics have not kept pace with
otic drugs most commonly used today were dis- the growing public health threat of drug-resistant
covered, largely by empirical screening, over 50 bacteria2,3,7 and examination of the current indus-
years ago and exploit a limited range of bacterial try pipeline shows a marked innovation gap.7 There
physiology: cell wall biosynthesis (␤-lactams, gly- has been a decline in new antibiotic approvals over
copeptides), cell membranes (daptomycin, colistin), the past 20 years7 and existing antibiotics are los-
type II topoisomerases (fluoroquinolones), ribo- ing effectiveness more rapidly than they can be re-
somes (macrolides, aminoglycosides, tetracyclines, placed. This is due in part to the extensive cost
the oxazolidinone linezolid, streptogramins), tran- and timelines associated with drug development (8–
scription (rifampicin), and folate biosynthesis (sul- 15 years8,9 and $800M10 ). The availability of low-
fonamides and trimethoprim).4,5 These classes of cost generic agents, coupled with the potential that
agents represent a relatively small set of chemical use of novel agents may be reserved exclusively for
scaffolds that have been modified over the past few the treatment of infections due to highly resistant

doi: 10.1111/j.1749-6632.2010.05828.x
Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 
c 2010 New York Academy of Sciences. 5
Challenges of antibacterial discovery Gwynn et al.

pathogens, can reduce the financial returns asso- (e.g., faropenem), or have an alternative safety
ciated with antibiotic development. The numbers profile;11,a
of approved or late-phase development novel-class (ii) Agents that act at an established target but
antibacterials are even lower,7 with only two sys- contain modifications to an existing chemical
temic antibiotics based on novel chemical scaffolds scaffold that modify spectrum or overcome re-
achieving marketing approval in the past 40 years: sistance (e.g., anti-methicillin-resistant Stap-
the protein synthesis inhibitor linezolid (an oxazo- hylococcus aureus (MRSA) fluoroquinolones,
lidinone) and the lipopeptide daptomycin.1,7 ketolides, and cephalosporins);12
Considering the significant unmet medical need (iii) Agents that act at a known target but exploit
for new antibiotics to address bacterial resistance, a new mechanism of action, binding site, or
why is the pharmaceutical pipeline so poorly new chemical space (e.g., novel topoisomerase
equipped to handle this important and growing de- inhibitors13–16 and pleuromutilins17 ); and
ficiency? The answer lies in the difficulties associ- (iv) Novel chemical classes that act at a
ated with discovering novel antibacterials and the completely new target (e.g., peptide de-
reality that the most innovative approaches (e.g., formylase inhibitors, leucyl tRNA synthetase
new targets or new chemical space) have longer inhibitors).18
development times when factoring in target vali-
Compounds that act at a previously unexploited
dation and lead generation. With greater burden
target, or via new mechanisms at existing targets,
of unknowns, new targets or new chemical space
have the greatest potential to address bacterial drug
also carry the highest risk of failure and attrition.
resistance because no precedence for clinical resis-
Yet, novel agents offer the greatest potential against
tance due to target-mediated or drug-inactivation
bacterial resistance, so we must continue to apply
mechanisms exists. There is greater risk, however,
what we have learned over the past decade and
when developing novel mechanism or novel target
use the latest technological tools to support the
agents. The earlier in the drug development cycle
development of transformational medicines. This
where novel approaches are attempted, the greater
review addresses the discovery hurdles associated
the risk of failure and attrition due to the unknown
with small molecule antibacterial agents. Alternative
properties waiting downstream. For example, pur-
therapeutic approaches, such as exogenous antibod-
suing a novel target leaves room for surprises regard-
ies, vaccines, and small molecule immune modu-
ing the essentiality of the target across the bacterial
lators, are associated with unique difficulties and
spectrum. Even when essentiality is established, can
opportunities that are not fully covered in this
adequate bacterial cell penetration be achieved for
paper.
antibacterial activity by an enzyme inhibitor? Can
A drug called “novel” a suitable molecule be optimized to include physic-
Every antibiotic development candidate is consid- ochemical properties that warrant candidate selec-
ered “novel” by those who make them, but the mul- tion for clinical development? What unknown toxi-
tiple approaches to developing new compounds do cology, clinical safety, or tolerability considerations
not carry equal potential to overcome preexisting might be uncovered by working in a new chemical
resistance mechanisms, nor are they associated with space?
equal development risk. As represented below, there
Higher risk affects investment
is a spectrum of innovation that ranges from devel-
Many factors have contributed to the reduced indus-
opments within established classes, to completely
try investment in antibacterial discovery, including
new bacterial pharmacophores and molecular
targets.
Spectrum of innovation for new antibacterials:
a
Faropenem, an orally administered penem antibacterial
(i) Agents that act at an established target and marketed in Japan, received a nonapprovable letter by the
within preexisting chemical space but have new U.S. Food and Drug Administration (FDA), requesting
physicochemical properties compared to exist- additional studies for all indications, including superior-
ing agents of the same class such that they can ity studies for acute bacterial sinusitis and acute exacer-
be delivered via a new route of administration bation of chronic bronchitis.

6 Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences.
Gwynn et al. Challenges of antibacterial discovery

cost, timelines, attrition, and the potential for a bet- genomes of relevant pathogens are compared to re-
ter return on investment for other disease areas that veal genes that are highly conserved among bacteria
require longer term treatment. These factors can be but are absent or poorly conserved in the human
unpalatable to investors and contributes to phar- genome. While certain pathogen-specific genes
maceutical companies increasingly abandoning the may be of interest, for example Mycobacterium
search for truly novel agents. The already low rate of tuberculosis, Helicobacter pylori, Pseudomonas
development candidates launched as new medicines aeruginosa, and Staphylococcus aureus, genes that
(8% in 2000–2002)19 largely reflects compounds are well conserved across the diversity of pathogens
within existing classes, and may underestimate the that can be encountered in therapeutic indications,
contest faced by novel class therapeutics. The tech- such as respiratory infection, wound infection, and
nical difficulties associated with novel antibacterial hospital acquired infections, offer greater commer-
discovery to produce differentiated medicines fac- cial attraction to balance the cost of discovery and
tor strongly into the cost–benefit calculations. Im- drug development.
proving the probability of success is paramount Based on genomic analyses, it has been estimated
to securing renewed investment in antibacterial that less than 300 bacterial genes have appropriate
discovery. conservation across species and in vitro essential-
The technical challenges to identifying new an- ity to represent potential broad-spectrum targets.20
tibacterial leads that address novel genomic targets As of January 2010, genomes for over 100 bacterial
have been previously described8 based on the ex- pathogen species and hundreds of individual strains
perience of evaluating over 300 genes for essential- have been reported.21 Additional genomes may help
ity and conducting greater than 70 high through- refine understandings of genetic conservation and
put screens over a period of 7 years starting in spectrum but the number of good molecular tar-
1995. Productivity in generating new leads was very gets will likely diminish as more is learned about
low, and this represents only part of the overall individual target exploitability. Although genomic
drug discovery process with all of the attendant databases define all potential new bacterial targets
attrition at each stage. In this paper, the techni- based on their genes, other important structures
cal challenges in validating new bacterial targets like bacterial cytoplasmic and outer membranes
and in achieving successful screening will be re- (targeted by Daptomycin and colistin, respectively)
viewed along with strategies that increase opportu- and peptidoglycan (bound by vancomycin-class an-
nities for discovering new class inhibitors for novel tibacterials), represent targets that are not defined by
targets. discrete gene products. Nongenomic starting points
(such as empirical whole cell screening described
Challenges with validating targets later) may be needed to find agents with mecha-
At the outset of the bacterial genomic era in the nisms not defined by reductionist approaches. In
mid 1990s, new target discovery was the major fo- addition, the full functionality of many highly inte-
cus for antibacterials and was anticipated to address grated genes products may be best screened for in-
rate limiting steps in antibacterial discovery. While hibitors in the context of functional pathways, such
genomics may yet transform antibacterial discov- as cell division.22
ery, the learning curve and translational challenges
have proven greater than initially appreciated. The Target essentiality
current challenges for antibacterial lead generation Essentiality for in vitro growth is the usual pre-
include validating the potential targets discovered requisite for a gene product to be regarded as a
over the past decade, understanding the risks asso- potential antibacterial target for inhibitors. How-
ciated with unprecedented targets, and identifying ever, essentiality cannot be extrapolated across bac-
inhibitor leads that have the potential to be devel- terial species and sometimes not even across most
oped into drugs. strains of the same species without additional sup-
porting data. While target heterogeneity remains
Target “prospecting” a consideration for any drug discovery effort, the
Previously unexploited targets are initially extent of microbial diversity means that bacterial
prospected by bioinformatic analysis in which targets exhibit far greater phylogenetic range than

Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences. 7
Challenges of antibacterial discovery Gwynn et al.

human or other mammalian targets, hindering the can undermine essentiality even for targets that have
identification of novel targets that are valid across been validated across multiple species and strains,
bacterial groups. As the premise for a new antibacte- and therefore need to be anticipated. Such “sur-
rial is its effectiveness against the majority of strains prises” were particularly hazardous at the outset
in a disease indication, a failure due to microbial di- of the genomic era. Today, our greater maturity
versity in the target would result in an unacceptable of knowledge helps reduce risk through essentiality
level of “built in” drug resistance. This is illustrated testing in multiple species and the use of “deep se-
by the following examples where bioinformatic and quencing” and bioinformatics analysis across mul-
in vitro genetic assessment of essentiality and pre- tiple strains and species. Highly cost-effective Next
dicted spectrum were confounded by phylogenetic Generation Sequencing technologies will broaden
variation or other factors. our knowledge of microbial target heterogeneity
The protein synthesis gene product methionyl and the frequency in clinical settings. For exam-
tRNA synthetase (MetRS) is one example of a target ple, Harris et al.26 describes the application of Next
believed to be broad spectrum based on high se- Generation DNA sequencing technology to com-
quence homology between S. pneumoniae, S. aureus, pare the near complete genomes of 63 strains of
and Escherichia coli. Potent inhibitors were identi- S. aureus.
fied and optimized for antibacterial activity against Brinster et al.27 provides another salutary lesson
strains of S. aureus and Enterococcus sp.; however, in target validation using the fatty-acid synthesis
approximately 30% of S. pneumoniae clinical iso- (FASII) pathway by demonstrating that exogenous
lates were found to be resistant due to the presence fatty acids, provided in both in vitro and in vivo
of an additional nonhomologous MetRS gene prod- systems, allow microbial pathogens to fully bypass
uct insensitive to the inhibitor class.23 The gap in S. inhibition of a pathway that was otherwise widely
pneumoniae coverage was unacceptable for treat- viewed as essential. Though the conclusions remain
ment against respiratory infections, though poten- controversial,28,29 this work reminds us that targets
tial for this class remains for treatment of MRSA broadly validated using optimized in vitro condi-
infections. tions might not always remain essential when ap-
Literature supporting essentiality of the E. coli plied to in vivo systems.
fatty acid biosynthesis gene product FabI (enoyl-
ACP-reductase) across all bacterial pathogens led to “Tractability” of bacterial targets
the exploration of FabI for use as a broad-agent tar- Not all bacterial targets have equal potential to be
get. FabI was later found to be absent from some exploited for use by antibacterial agents, even if they
bacteria whose function was encoded by nonho- were all equally essential and conserved. Most suc-
mologous genes. Despite the identification of leads cessful drug targets are associated with deep active
with potent activity against S. aureus, the absence site clefts30 that offer potential for strong inhibitor
of activity against S. pneumoniae meant this target binding. Other targets, such as those associated with
could never achieve broad-spectrum or respiratory- protein–protein interactions, are less tractable and
spectrum activity.24 are thus far devoid of examples of successful ex-
Another lesson learned about pursuing genomic ploitation for antibacterials.22 The bacterial cell di-
targets is the case of duplicate genes in some species. vision septation process, for example, is a complex
UDP-N -acetylglucosamine-enolpyruvyl trans- interplay of essential proteins involving multiple
ferase (MurA) is the clinically validated antibiotic protein–protein interactions22 and is an attractive
target of fosfomycin. Du et al.25 reported that, target in terms of bacterial specificity and the poten-
unlike E. coli, several Gram-positive species actually tial to be accessed by extra-cytoplasmic inhibitors
have two forms of essential MurA gene products. for some of the components. While complex pro-
An effective inhibitor must therefore inhibit both tein interactions have not been successfully tar-
Mur enzymes in order to have antibacterial activity, geted in the past, examples are emerging of small-
representing an additional challenge for lead molecule inhibitors of protein–protein interactions,
optimization.25 and the scientific community is beginning to un-
The examples above show that unexpected chal- derstand which interactions are more amenable to
lenges, such as nonhomologous and duplicate genes, antagonism.31

8 Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences.
Gwynn et al. Challenges of antibacterial discovery

Discovery of successful antibacterial targets and efflux systems that can reduce antibacterial activ-
inhibition mechanisms have often been preceded ity by orders of magnitude.35 Targets outside the
by the identification of “tool compounds” that bacterial cytoplasmic membrane obviate penetra-
were found by nontarget-based approaches, such tion barriers to the inhibitor reaching the target,
as whole cell screening. These tool compounds in- but the choice of essential, well-conserved broad-
crease confidence in finding new inhibitors for the spectrum targets outside the membrane is greatly
target with antibacterial activity that can be ex- reduced. ␤-lactam antibiotics exemplify successful
ploited in lead optimization programs. Exploitation targeting outside the cytoplasmic membrane and
of the bacterial cell division protein FtsZ was in part achieve a superb spectrum of activity, but this ex-
encouraged by knowledge of effective inhibitors of ample has not been matched. An example of a po-
the homologous human cytoskeleton protein, ␤- tent new-target, narrow-spectrum (P. aeruginosa)
tubulin, such as taxol. A new small molecule in- agent that addresses an exposed outer membrane
hibitor of the GTPase activity of FtsZ had potent target involved in its assembly is the synthetic pro-
antistaphylococcal activity and efficacy in animal tein epitope mimetic (PEM) POL7080, which is cur-
models of infection.32 Where no previously known rently in preclinical development and is described
inhibitors of targets exist, extensive screening cam- later.36
paigns have frequently failed to find exploitable
inhibitors.8 Why some targets appear to be less Spontaneous resistance to new antibacterials
amenable for finding inhibitors is not always ob- Target selection for antibiotics also has a complex
vious, and might be a function of the diversity of temporal aspect: selective pressure generated by
the compounds screened or the physiology of the clinically used classes of antibacterials will alter bac-
targets.33 terial population genetics. Proliferative diseases, in-
Ribosome and gyrase targets, for example, have cluding bacterial infection, are different from other
multiple classes of published leads and marketed therapeutic areas in that emergence of resistance
drugs, whereas other essential gene products have diminishes effectiveness of therapies with time.
no known inhibitors despite a long history of an- Single-enzyme targets are particularly susceptible
tibacterial research. Pursuing novel chemical classes to spontaneous resistance.37 Examples of agents that
of inhibitors for tractable, pharmacologically vali- rapidly became compromised as single-agent thera-
dated targets by new mechanisms or binding modes pies due to resistance include: fosfomycin, novo-
that overcome preexisting resistance mechanisms, biocin, fusidic acid, and rifampicin. Agents with
such as DNA replication, ribosome function, cell- nonsingle-enzyme targets fare better against the de-
wall biosynthesis, and ␤-lactamase inhibition, is a velopment of resistance. These include ␤-lactams
rational strategy that reduces the risks associated (which act at multiple penicillin binding proteins),
with unprecedented targets and exploits their estab- fluoroquinolones (which target the homologous
lished druggability.8,15,34 topoisomerases DNA gyrase and topoisomerase IV
In addition to activity at the bacterial target, enzymes), vancomycin-class antibacterials (which
antibiotic compounds face the added requirement target the peptidoglycan substrate rather than the
of penetrating across bacterial cell membranes, processing enzymes), and a wide range of ribo-
or for intracellular pathogens, penetrating across somally targeted antibacterials (which experience
both mammalian and bacterial cell membranes. low target-mediated spontaneous resistance because
The requirement to cross challenging biological rRNA is encoded by multiple genes). These are no-
barriers, while not unique to antibacterials (e.g., tably also the targets for most clinically used an-
central nervous system active agents and some tibacterials that have stood the test of time.
anticancer drugs must cross the blood–brain bar- Most antibacterial classes in wide clinical use
rier), does place significant demands on the dis- benefit from acting at molecular targets associ-
covery and lead optimization of antibiotics and ated with intrinsically reduced levels of sponta-
may contribute to the lower success rates of an- neous resistance. Novel, single-enzyme bacterial
tibacterials. Simultaneous to the challenge of pas- targets can still be considered for development,
sive diffusion across membrane(s), Gram-negative but the risk of high spontaneous resistance caus-
bacteria in particular express powerful multidrug ing expensive failure during early clinical use is

Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences. 9
Challenges of antibacterial discovery Gwynn et al.

arguably higher for single-enzyme targets, and also Specialized vendors are now available who can
presents an additional risk associated with unprece- add diversity for screening collections by filling
dented targets. Careful preclinical assessment of in perceived diversity “gaps” present in preexist-
spontaneous resistance and virulence of laboratory ing collections.41 Further enrichment of synthetic
mutants is important when evaluating new drug compound screening libraries is based on use of
leads in order to understand the potential for rapid Diversity-Oriented Synthesis (DOS), which applies
reduction in therapeutic effectiveness due to emer- combinatorial synthesis of diverse small molecules
gent resistance. within biologically relevant and underexploited
chemical space,42 and Galloway et al.43 describe
Screening strategies to address target an example applied to antibacterial lead discov-
challenges ery. Biology-Oriented Synthesis (BIOS) centers on
Following target identification and validation, the generation of small compound libraries based
screening for molecules with inhibitory activity at on scaffolds of proven biological relevance, natural
the selected target is needed; however, the out- products, and drugs.44 The aims of DOS and BIOS
put of high throughput screening (HTS) campaigns are to generate a relevant diversity of scaffolds and
against novel bacterial targets has been inadequate functionalities,42 while avoiding the downsides of
to support the antibacterial pipeline.8 This was at- traditional natural product screening, such as longer
tributed in part to the screening of available chem- discovery process and low synthetic tractability.
ical libraries that were limited in relevant diversity In addition to biologically inspired diversity,
for antibacterial discovery. Marketed antibacterials there is also renewed interest in natural products
do not generally follow Lipinski’s “rule of five,” and themselves as a source of screening leads. Thirty-
are on average less lipophilic and slightly larger than four percent of all small molecule New Chem-
many other drug classes.38 Fortunately, advances in ical Entity (NCE) approvals between 1981 and
small molecule libraries and the creation of innova- 2006 were either natural products or semisyn-
tive screening technologies could potentially trans- thetic derivatives.45 Natural products have an even
form the identification of suitable molecules for new greater relevance in the history of antibiotic dis-
targets.39 covery.46–48 Despite that success, pharmaceutical re-
search into natural products declined during the
HTS libraries and enhanced diversity past two decades. At the start of the HTS era,
Pharmaceutical industry small molecule screening many companies moved from natural-product ex-
collections are generally composed of around one tract libraries toward more convenient synthetic
million compounds, derived from a number of compound libraries. The reemerging interest in nat-
sources. HTS in the new genomic era was unsuc- ural products for lead discovery is partly driven
cessful in finding hits for many targets (particularly by the realization that competing technologies,
in the antibacterial arena), thus driving the need such as combinatorial chemistry, have failed to
for compound screening collections with more bio- deliver new drug leads in significant numbers.41
logically relevant and diverse compounds.39 Several Analysis of chemical space associated with mar-
sources of compounds have been used for expanding keted drugs, natural products, and molecules made
and enhancing screening collections. Compounds from combinatorial synthesis demonstrated that the
from previous and current lead optimization efforts structures of drugs in use today more closely re-
across diverse therapeutic areas are useful within semble those of natural products.49 In addition, re-
screening collections as they have demonstrated rel- cent analytical technological advances, such as spec-
evance to biological systems and tend to have desir- troscopy and HTS, have been crucial breakthroughs
able physicochemical properties, such as solubility, in separation of natural products and in determin-
stability, and lack of unwanted reactivity. However, ing structures of screening hits.46–51 A recent study
while such molecules can be “drug-like,” they can looking at hit rates from HTS campaigns with a
also represent a limited range of scaffolds and they collection at Novartis suggests natural products are
tend to have higher molecular weight than smaller, a more diverse class than traditional synthetic and
“lead-like” molecules with greater potential as combinatorial molecules, and that natural products
inhibitors.40 have the highest hit rate.52

10 Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences.
Gwynn et al. Challenges of antibacterial discovery

The enduring value of enhancing the diversity of and selectivity data, before functional biochemical
HTS screening collections to industry drug pipelines assays are developed. This rapid and inexpensive
remains undetermined given the lengthy research platform for the discovery of small molecule drug
and development time scale, but investment in these leads has the potential to revolutionize the discovery
approaches by both industry and academia suggests of small-molecule modulators of biological targets
promise for the future. including bacterial targets.
Specialized chemical libraries
Fragment-based drug discovery
The potential for specialized chemical libraries to
One complementary approach to HTS includes
yield promising novel small molecule leads and ther-
“fragment screening,” which has become a main-
apeutics is illustrated by the boron chemistry plat-
stream discovery strategy in the pharmaceutical in-
form of Anacor Pharmaceuticals (Palo Alto, CA).
dustry over the last few years. Fragment-based drug
Boron as boric acid is ubiquitous in the natural
discovery (FBDD) screens molecules less than ∼300
environment and known to be safe in many every-
Da (compared to standard HTS of molecules from
day items and is a trace mineral believed to con-
450 to 500 Da). The “hit rates” are higher than with
tribute to growth, brain function, and inflamma-
traditional HTS, but the affinities are typically rela-
tory response.53 Boron-containing molecules are
tively low (␮M–mM range). The higher hit rates are
increasingly being explored for their chemothera-
achieved because smaller, less complex molecules
peutic potential. The properties and electrophilicity
have lower chances of a mismatch that would other-
of boron and boronic acids confer to boron-based
wise negate binding.40 Chemists can later enhance
compounds the ability to undergo novel interac-
affinity by increasing structural complexity with the
tions with biological targets.54 In December 2009,
benefit of structure-guided design.
the commencement of a Phase I clinical study was
FBDD is typically conducted in conjunction with
announced for AN3365, a novel boron-based, small
high-resolution structural techniques. X-ray crys-
molecule drug candidate in development for the
tallography is most widely used for protein structure
treatment of hospital infections caused by Gram-
determination and advances have solved technical
negative bacteria, including E. coli, K. pneumoniae,
constraints to enable crystallographic techniques to
and Enterobacter spp. AN3365 targets the bacterial
be used on a higher throughput basis and serve
enzyme leucyl tRNA synthetase, a previously unex-
as part of fragment screening strategies. The high-
ploited target.55 The potential to act at this bacterial
resolution structural data can then be used as a start-
target was presaged by the activity of this class of
ing point for lead optimization and also forms part
inhibitor against a fungal homolog, and illustrates
of an iterative process in the creation of larger, more
an exciting new mechanism of action.56,57
potent compounds.30
DNA-encoded small molecule libraries FBDD approach has been successful across a
Clark et al.58 report a technology for the syn- range of targets, either when the target is not
thesis and screening of an 800-million-member amenable to HTS or as a parallel strategy to increase
DNA-encoded library, in which small molecules probability of success. FBDD-related discovery ap-
are covalently attached to an encoding, double- proaches have been credited with at least 13 exam-
stranded oligonucleotide. The library was interro- ples of molecules entering Phase I and Phase II clin-
gated by affinity selection and underwent decon- ical trials; however, none of these examples thus far
volution through DNA “barcode” sequencing of are antibacterials.59 A recent application to antibac-
the library, resulting in the discovery of inhibitors terial lead discovery is reported in an elegant study
for two human enzymes: Aurora A kinase and p38 by Pfizer, which demonstrated the use of FBDD and
MAP kinase. This process increases library screening virtual screening to identify antibacterially active
throughput by several orders of magnitude over pre- leads targeting biotin carboxylase, a component of
vious methods with only modest instrumentation the multienzyme complex acetyl-CoA carboxylase,
and protein requirements. By using high through- which catalyzes the first committed step in fatty
put sequencing techniques, the selection output can acid biosynthesis. This further validates biotin car-
be deconvoluted to yield families of ligands with boxylase as a tractable antibacterial target and il-
emerging trends of structure–activity relationships lustrates the applicability of FBDD in antibacterial

Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences. 11
Challenges of antibacterial discovery Gwynn et al.

discovery.60 FBDD was used with virtual potential spectrum and of selectivity (with respect
screening—in which shape and electrostatic to human homologs), and supports lead optimiza-
properties were screened in silico—to identify tion with target Structure Activity Relationship
weak-binding but exploitable fragments as ATP- (SAR) and cocrystallography. Macromolecular syn-
competitive inhibitors. Lead optimization then re- thesis (RNA, protein, DNA, peptidoglycan, and fatty
sulted in up to a 3,000-fold improvement in affin- acids) encompasses more than 75% of the bacte-
ity and achievement of antibacterial activity while rial essential genome. Inhibition of these pathways
maintaining desirable physicochemical properties. can be detected using radiolabeled precursor up-
take and can be screened in high throughput mi-
Whole cell screening: pros and cons crotiter format. Genetic approaches can be applied
Most currently used antibacterial classes were dis- to rapid definition of mode of action including re-
covered by whole cell screening and are composed sistant mutant analysis, gene overexpression and
of relatively few structural families. Molecular tar- under-expression (e.g., antisense), the use of con-
get screening became fashionable, but the lack of ditional lethal mutants to identify compounds with
tractable leads, together with the known difficulty synergistic effects, and gene-expression assays (such
of converting compounds lacking whole cell activity as DNA microarrays) that identify compounds that
into ones with activity, has led to renewed consid- elicit specific stress responses.65 Sequencing tech-
eration of nontarget-based, whole cell antibacterial nology now makes it practical to locate a single point
screens. More holistic approaches, such as viable mutation in the genome of a mutant selected by a
cell screening and biochemical pathway screens, do novel antibacterial, in order to identify the molecu-
not limit the concept of “target” to a defined gene lar target.66
product, and also allow biological complexity to be A recent application of genomic methods to
screened with the potential to identify inhibitors rapidly analyze a mechanism of action was described
with unpredicted mechanisms. This is well illus- by Pathania et al.,67 who assembled an ordered,
trated by the investigation of an unexploited class of high-expression clone set of all the essential genes
natural product antibacterial from the patent liter- from E. coli and used it to systematically screen for
ature that was found to dysregulate bacterial prote- suppressors of antibacterial action. This method led
olytic machinery.61 Whole cell screening evaluates to the discovery of MAC13243, a new antibacterial
all targets essential for in vitro growth in a phys- with a unique mechanism that inhibits the func-
iological context. Although potentially less sensi- tion of the LolA protein (a key component of the
tive than molecular screens, whole cell screening lipoprotein targeting pathway in bacteria).
selects for antibacterial activity as a starting prop- Miller et al.68 used whole cell screening to identify
erty. The availability of improved methods for rul- a series of antibacterial pyridopyrimidines derived
ing out nuisance compounds62 and for defining from a “repurposed” protein kinase inhibitor com-
molecular targets of new antibacterial leads con- pound collection. The leads were found to target
tributes to the renewed efficiency of empirical whole the ATP-binding site of the fatty acid biosynthesis
cell screening, affording it a “back to the future” target, biotin carboxylase, and were effective in vitro
quality. and in vivo against Gram-negative pathogens in-
Mechanism-based whole cell screens can com- cluding Haemophilus influenzae. While the enzyme
bine a targeted approach with a biologically inte- has homology to the human counterpart, the in-
grated context. This is elegantly illustrated by the hibitors were selective for bacterial biotin carboxy-
discovery of a broad-spectrum FtsZ inhibitor in a lase as the binding mode was distinct from the pro-
mechanistic whole cell screen63 and the use of a tein kinase–binding mode of mammalian enzyme
whole cell antisense screening strategy to identify inhibitors. This illustrates the creative combination
a novel natural product inhibitor of the fatty acid of whole cell screening with the exploitation of an
biosynthesis condensing enzymes, FabF/B.64 enhanced compound collection based on inhibitors
Once a screening “hit” has been identified, defin- of a structurally related human target. This resulted
ing the molecular target is a key step before en- in a mechanistically and structurally new class of se-
tering into lead optimization. Knowing the molec- lective antibacterial leads—derived from a protein
ular target enables bioinformatic analysis of the kinase inhibitor pharmacophore.68

12 Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences.
Gwynn et al. Challenges of antibacterial discovery

ins. The results reported so far are promising; in a


Focus on spectrum
Phase II clinical trial involving 200 patients with C.
Antibacterial spectrum is a major consideration difficile infection receiving standard-of-care antibi-
when selecting a target for lead optimization. Many otics, the addition of intravenous MDX-066–MDX-
targets have the potential to be broad in spectrum 1388 treatment reduced the rate of recurrent in-
but are limited by the requirement to penetrate the fection by 70%.80 As reviewed by Bebbington and
outer membrane of Gram-negative bacteria. In ad- Yarranton,81 there are a number of antibodies to
dition, multidrug efflux systems present in many bacterial targets primarily in S. aureus, Bacillus an-
bacteria transport a wide range of chemically di- thracis, and P. aeruginosa that are currently in clin-
verse molecules out of the bacterial cell.69,70 Medici- ical studies. As Mabs are highly directed against a
nal chemists are therefore challenged with the task of specific bacterial target without a human homolog,
identifying molecules whose physicochemical prop- the safety profile using this approach is predicted
erties allow it to penetrate the cell and evade efflux, to be very good. While there are currently no FDA-
but retain potent activity at the molecular target. approved Mab products for treating bacterial in-
The design of small molecules to inhibit bacterial fections, this promising approach warrants further
efflux has also been explored as a means of broaden- investigation.
ing the antibacterial spectrum and overcoming bac- PEM technology is a new approach in drug
terial resistance. While a clinical candidate has yet discovery in use for the development of narrow-
to emerge, this approach has been validated across spectrum therapies, that may represent a novel class
a range of in vitro and in vivo assays.71 of therapeutics. As described by Obrecht et al.,82
One approach to circumvent the challenges of PEM molecules can mimic the beta-hairpin and
discovering and optimizing broad-spectrum agents alpha-helical secondary structure elements often
is to focus on agents that target a specific pathogen. observed in protein–protein interactions. In rapid
In addition to offering fewer drug discovery bar- iterative cycles that include design, synthesis, pu-
riers, narrow-spectrum agents have the potential rification, and biological profiling, optimized PEM
for reduced disruption of normal intestinal flora.72 molecules can be obtained with greatly improved
Clostridium difficile is an obvious candidate as it Absorption, Distribution, Metabolism, Excretion,
accounts for 20–30% of antibiotic-associated diar- and Toxicity (ADMET) properties compared to the
rhea and nearly all cases of antibiotic-associated col- starting natural products. The PEM approach thus
itis.73–75 Episodes of recurrent C. difficile infection efficiently generates fully synthetic compounds with
are difficult to treat and require repeated and pro- drug-like properties, which display natural product-
longed treatment. Several strategies have been taken like structural complexity. A new structure-activity
to develop narrow-spectrum agents against C. diffi- trail has provided access to derivatives with a much
cile associated diarrhea (CDAD). Fidoximicin (Opt- higher potency and selectivity against P. aeruginosa.
80) is a novel, bactericidal, macrocyclic compound One molecule, POL7080, showed potent in vitro ac-
that inhibits RNA polymerase of specific bacteria, tivity against a broad panel of Pseudomonas strains
such as C. difficile.76,77 Because of its narrow spec- (including multidrug-resistant clinical isolates from
trum, fidoximicin is believed to prevent relapse in cystic fibrosis patients) with an MIC90 = 0.25
CDAD since alterations of the normal flora of the ␮g/mL. In addition, POL7080, which is believed
gastrointestinal tract are a prerequisite for C. difficile to perturb outer-membrane biogenesis, showed po-
proliferation.78 Furthermore, fidoximicin is primar- tent activity in an in vivo septicemia mouse model
ily retained in the gut, with low levels in serum fol- against P. aeruginosa PAO1 (EC50 = 0.3 mg/kg) and
lowing oral administration, making it an attractive was 10 times more potent than gentamicin tested in
agent to treat a gut-specific pathogen.79 the same model. These data demonstrate the poten-
Monoclonal antibodies (Mabs) have also been tial utility of this new class of compound.83
added to the armamentarium of potential agents
Lead optimization
for narrow-spectrum antibacterial therapy. One of
the most advanced agents is the C. difficile Mab Identification of a lead molecule is considered a
combination MDX-066 and MDX-1388, which tar- milestone achievement although most leads re-
gets and neutralizes the two main C. difficile tox- quire substantial chemistry (optimization) in order

Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences. 13
Challenges of antibacterial discovery Gwynn et al.

to achieve the potency, selectivity, spectrum, and ADME properties and increased potential for toler-
pharmacokinetic properties necessary for an antibi- ability and toxicological effects.
otic. Traditional approaches to synthetic antibiotic The maximum concentration (Cmax) of un-
drug design focused on incremental changes to lead bound drug associated with efficacious exposure
molecules that altered the molecular structure to im- for antibiotics is often 1,000-fold higher than for
prove or reduce a particular feature, such as potency, other therapeutic areas (antibiotic Cmax range is
spectrum, absorption, distribution, metabolism, or generally 1–10 uM). Even where the primary phar-
toxicity. With the discovery of three-dimensional macokinetic parameter for efficacy is not Cmax, an-
crystal structures of essential antibacterial targets, tibiotics must still achieve plasma levels comparable
X-ray crystallography now enables improved bind- to bacterial MIC levels in order to achieve efficacious
ing predictability associated with proposed modifi- exposures. Since antibacterial activity requires pres-
cations to the molecule. A review of structure-based ence of unbound drug, it is not uncommon to find
drug design using the bacterial ribosome crystal antibacterials that are >50% unbound to plasma
structure describes the significant improvements in proteins (whereas 1% of unbound drug may be ad-
the potency and spectrum obtained for the oxazo- equate for other therapeutic areas). Therefore, an-
lidinones class of antibacterials.84 tibiotics can be associated with up to 50,000-fold
Antibiotic lead optimization has some advantages greater free Cmax levels than other therapeutics.
over other therapeutic areas in that validated in vitro High free Cmax concentrations contribute to
assays and in vivo models mitigate a great deal of the potential drug-induced arrhythmia and electro-
efficacy risk against the bacterial target at the time cardiogram QT prolongation safety hurdles facing
of drug candidate selection. On the other hand, the many antibiotic leads, because acute functional car-
high plasma exposures needed for antibiotic efficacy diovascular toxicity is generally related to Cmax.
create additional hurdles in optimizing candidate Over the past decade our understanding of certain
leads due to the difficulties associated with achieving drug related toxicities, in particular those related
high systemic exposures, such as increased toxicity to cardiovascular anomalies, have substantially im-
associated with off-target effects. proved. Drug-induced QT prolongation, which cur-
High plasma exposures are required for antibi- rently serves as a biomarker for increased risk for
otics as they must cross multiple biological barriers torsade de pointes, has been the reason why several
in order to access their target, which generally re- antibiotics have been rejected by regulatory agen-
sides within the bacterial cell. In comparison, tar- cies or withdrawn from the market.90 While such
gets for many other disease areas are represented an outcome is warranted, it sacrifices large invest-
by receptors or ligands that are outside of the cell. ments, especially when this occurs after a full clinical
This is exemplified in Table 1, which illustrates the program has been conducted. To better manage this
dose levels and therapeutic exposures needed for risk, assays such as those for hERG ion channel ef-
amoxicillin/clavulanate compared to rosiglitazone fects, as well as other in silico, ex vivo, and in vitro
(treatment of type 2 diabetes) and sumatriptan models can be implemented during the lead op-
(treatment of migraine headaches). With the need timization stage to deselect compounds that have
for relatively high plasma exposure for antibacteri- potential cardiac ion channel risk.91,92
als comes the burden of both engineering acceptable Lead optimization to reconcile all the required at-
tributes for a drug candidate remains a highly chal-
lenging, resource-intensive, and time-consuming
Table 1. Plasma concentrations associated with thera-
phase of drug discovery. Drug leads from estab-
peutic use of selected marketed drugs
lished classes have to demonstrate improvements
Drug (therapeutic dose) Plasma Cmax (␮g/mL) that adequately differentiate them from marketed
drugs enough to warrant development investment.
Amoxicillin/clavulanate 11.6 (amoxicillin)85 Leads from drug classes that exploit novel antibacte-
(875/125 mg) rial targets have greater risk for encountering unan-
Rosiglitazone (2 mg) 0.15686,87 ticipated class liabilities. Despite the widespread
Sumatriptan (25 mg) 0.02888,89 acceptance of the physical properties associated
with “drug-like” molecules, analysis of current

14 Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences.
Gwynn et al. Challenges of antibacterial discovery

medicinal chemistry efforts, as reflected in patent subsequently discovered to be inactivated by human


applications from AstraZeneca, GlaxoSmithKline, lung surfactant, a phenomenon not clearly observed
Merck and Co., and Pfizer, indicates that recent in animals.102 The activity of daptomycin is strictly
compounds synthesized by pharmaceutical compa- dependent on the presence of physiological levels of
nies have been larger and more lipophilic than mar- free calcium. Daptomycin has a novel mechanism of
keted oral drugs. Higher lipophilicity is recognized action that results from insertion into and disrup-
as a correlate to reduced pharmacological specificity tion of the functional integrity of the Gram-positive
and potentially increased risk of future attrition due bacterial membrane. This produces a rapid loss of
to safety findings during the drug optimization and membrane potential, cessation of macromolecular
development process.93,94 Improved choices and di- synthesis, and cell death. The calcium dependence
versity of starting points for antibacterial lead opti- of daptomycin–surfactant interactions, the absence
mization programs and greater chemical tractabil- of surfactant inhibition of other classes of antibi-
ity of the classes selected would help this situation. otic, and the demonstration of direct insertion of
However, lead optimization failure is also driven daptomycin into surfactant aggregates in vitro all
by incomplete target validation as described in an support a model whereby daptomycin in the lungs
earlier section, the learning curve around toxicity is unable to efficiently distinguish between the vast
prediction, and the need to avoid a narrow drive surfactant layer and the relatively small surface area
for molecular potency that can result in a level of of the bacteria. Insertion into the surfactant layer
lipophilicity that is more prone to off-target activity. is predicted to be an essentially irreversible pro-
The ability to predict clinical effectiveness and cess resulting in sequestration and inactivation of
spectrum using in vitro and in vivo models is a the antibiotic. This effect102 represents the first ex-
distinct advantage in antibacterial lead optimiza- ample of a pulmonary-specific inactivation of an
tion efforts over other therapeutics. Animal models antibiotic. Other potential limitations include dif-
of bacterial infection have substantially improved ferences in drug concentrations at specific sites of
the ability to predict clinical efficacy of discovery infection, protein binding, and immune response.
stage antibiotics. Additionally, these models can be Local tissue concentrations at the site of infection
used to study the efficacy of a compound against may differ between animals and humans. Decades
strains that may be encountered infrequently in clin- of experience gained across these models, however,
ical studies.95,96 These models offer insight into the have minimized this risk by showing good corre-
utility of a compound across a range of pathogens, lation using blood levels with clinical outcome.95
infection sites, and resistance profiles. Over the past Differences in protein binding can also be mitigated
two decades our understanding of the pharmacoki- by using the free drug concentrations to compare
netic and pharmacodynamic (PK/PD) relationship blood concentration levels between species.96 While
of conventional antibiotics has become well estab- recognizing the potential limitations of the animal
lished. Understanding the PK/PD properties that models, there remains a high level of confidence that
drive efficacy allows selection of a clinically relevant these models provide insight to potential human
dose and dose regimen to optimize efficacy and mit- efficacy.
igate against the selection of bacterial resistance.97,98
Strategies to address the innovation gap
This is exemplified by a number of approved antibi-
otics including doripenem and linezolid for which There is now high-level political, medical, and pub-
the doses selected for use in Phase II/ III studies were lic concern about the perfect storm of increasing
based on animal derived PK/PD combined with hu- antibiotic resistance and the thin industry pipeline.
man Phase I pharmacokinetic data.99,100 For example, in 2009 the U.S. and European Com-
Animal models generally show excellent correla- munity Presidencies have established a Transatlantic
tion with clinical outcome; however, these mod- Task Force to address antimicrobial resistance,103
els do have some limitations, such as between and the Infectious Diseases Society of America
animal and human physiology. For example, the li- (IDSA) has announced a call to action for a global
poglycopeptide antibiotic daptomycin failed to meet commitment to develop 10 novel antibacterial drugs
the noninferiority criteria in severe community- by 2020 (10 × ‘20).104 Recognition of the chal-
acquired pneumonia trials.101 The antibiotic was lenge is also evident in public–private partnerships,

Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences. 15
Challenges of antibacterial discovery Gwynn et al.

including government agencies and charities (e.g., Acknowledgments


Wellcome Trust, Bill and Melinda Gates Foun-
We thank James R. Brown, Michael H. Hann,
dation) to address the public health threat
Jianzhong Huang, Andrew R. Leach, and Richard
from conventional and weaponizable pathogens,
Jarvest for valuable input, and Ricardo Macarron
a sign of the importance and difficulty of the
and David Drewry for valuable input and generous
challenge.
access to prepublication information.
However, while incentives for antibacterial drug
development and optimization of the regulatory en- Conflicts of interest
vironment are key, the principal barrier is techno-
logical. The technical challenges we described in The authors are employees of GSK.
2007 remain.8 The ability of new antibiotics to over-
come bacterial resistance is enhanced when more References
novel approaches are pursued; however, greater nov- 1. Fischbach, M.A. & C.T. Walsh. 2009. Antibiotics for emerg-
elty introduces more technical and financial risks. ing pathogens. Science 325: 1089–1093.
Furthermore, efforts at the earliest stages of drug 2. Infectious Diseases Society of America. 2004. Bad bugs,
discovery, which employ the newest technologies, no drugs: as antibiotic R&D stagnates, a public health cri-
sis brews, http://www.idsociety.org/ (accessed October 23,
can require 12–15 years of effort before a licensed 2010).
antibacterial drug becomes available on the mar- 3. Talbot, G.H., J. Bradley Jr., J.E. Edwards, et al. 2006. Bad
ket. Development of new-class antibacterial agents bugs need drugs: an update on the development pipeline
is imperative, but financial and scientific consider- from the antimicrobial availability task force of the In-
ations also drive a nearer term strategy to augment fectious Diseases Society of America. Clin. Infect. Dis. 42:
657–668.
the antibacterial armamentarium by continued ex- 4. Fernandes, P. 2006. Antibacterial discovery and
ploitation of proven antibacterial classes. This in development—the failure of success? Nat. Biotechnol.
some ways reduces the technical challenge, but on 24: 1497–1503.
the other hand only offers more incremental therapy 5. Lange, R.P., H.H. Locher, C.P. Wyss, et al. 2007. The tar-
improvements. gets of currently used antibacterial agents: lessons for drug
discovery. Curr. Pharm. Design 13: 3140–3154.
Challenges with identifying tractable new tar- 6. Royal Society Policy Document. 2008. Innovative
gets, coupled with the poor performance of HTS mechanisms for tackling antibacterial resistance, http://
to generate leads for new antibacterial targets, has royalsociety.org/Innovative-mechanisms-for-tackling-
driven the search for screening methodologies that antibacterial-resistance/ (accessed October 23, 2010).
use greater compound diversity and novel assay 7. Spellberg, B., J.H. Powers, E.P. Brass, et al. 2004. Trends
in antimicrobial drug development: implications for the
designs, as well as innovative approaches to find- future. Clin. Infect. Dis. 38: 1279–1286.
ing new targets. Promising discovery strategies in- 8. Payne, D.J., M.N. Gwynn, D.J. Holmes, et al. 2007. Drugs
clude improvements in the diversity of compound for bad bugs: confronting the challenges of antibacterial
collections for screening, such as prefractionated discovery. Nat. Rev. Drug Discov. 6: 29–40.
natural product libraries, improved synthetic di- 9. IDSA. 2008. Challenges in the pathway to an-
tibiotic approvals. http://www.idsociety.org/WorkArea/
versity libraries, DNA-encoded small molecule li- DownloadAsset.aspx?id=12134 (accessed October 23,
braries, new chemical platforms such as boron 2010).
chemistry, and fragment-based screening. “Back to 10. DiMasi, J.A., R.W. Hansen & H.G. Grabowski. 2003. The
the future” use of whole cell antibacterial screen- price of innovation: new estimates of drug development
ing with reverse genomics to define molecular tar- costs. J. Health Econ. 22: 151–185.
11. Schurek, K.N., R. Wiebe, J.A. Karlowsky, et al. 2007.
gets is an attractive strategy, and also some in- Faropenem: review of a new oral penem. Expert Rev. Anti-
dustry successes have been reported in exploiting Infect. 5: 185–198.
new narrow-spectrum and broad-spectrum molec- 12. Theuretzbacher, U. 2009. Future antibiotics scenarios: is
ular targets. Recognizing that not all targets have the tide starting to turn? Int. J. Antimicrob. Agents 34: 15–
equal potential to yield drugs, pursuing novel in- 20.
13. Mitscher, L.A. 2005. Bacterial topoisomerase inhibitors:
hibitors of validated targets, such as ribosomes and quinolone and pyridone antibacterial agents. Chem. Rev.
topoisomerases, can overcome existing resistance 105: 559–592.
mechanisms, while reducing early stage discovery 14. Black, M.T., T. Stachyra, D. Platel, et al. 2008. Mech-
risk. anism of action of the antibiotic NXL101, a novel

16 Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences.
Gwynn et al. Challenges of antibacterial discovery

nonfluoroquinolone inhibitor of bacterial type II topoi- 32. Haydon, D.J., N.R. Stokes, R. Ure, et al. 2008. An inhibitor of
somerases. Antimicrob. Agents Chemother. 52: 3339– FtsZ with potent and selective anti-staphylococcal activity.
3349. Science 321: 1673–1675.
15. Bax, B.D., P.F. Chan, D.S. Eggleston, et al. 2010. Type IIA 33. Silver, L.L. 2008. Does the cell wall of bacteria remain a vi-
topoisomerase inhibition by a new class of antibacterial able source of targets for novel antibiotics. Biochem. Phar-
agents. Nature 466: 935–940. macol. 71: 996–1005.
16. Miller A.A., G.L. Bundy, J.E. Mott, et al. 2008. Discovery 34. Stachyra, T., P. Levasseur, M.-C. Péchereau, et al. 2009.
and characterization of QPT-1, the progenitor of a new class In vitro activity of the ␤-lactamase inhibitor NXL104
of bacterial topoisomerase inhibitors. Antimicrob. Agents against KPC-2 carbapenemase and Enterobacteriaceae ex-
Chemother. 8: 2806–2812. pressing KPC carbapenemases. J. Antimicrob. Chemoth. 64:
17. Novak R. & D.M. Shlaes. 2010. The pleuromutilin antibi- 326–329.
otics: a new class for human use. Curr. Opin. Invest. Drugs 35. Nikaido, H. 1994. Prevention of drug access to bacterial
11: 182–191. targets: permeability barriers and active efflux. Science 264:
18. Ochsner, U.A., X. Sun, T. Jarvis, et al. 2007. Aminoacyl- 382–388.
tRNA synthetases: essential and still promising targets for 36. Page, M.G.P. & J. Heim. 2009. Prospects for the next anti-
new anti-infective agents. Expert Opin. Inv. Drug. 16: 573– Pseudomonas drug. Curr. Opin Pharmacol. 9: 558–565.
593. 37. Silver, L.L. & K.A. Bostian. 1993. Discovery and develop-
19. Gilbert, J., P. Henske & A. Singh. 2003. Rebuilding big ment of new antibiotics: the problem of antibiotic resis-
pharma’s business model. Business Med. Rep. http://www. tance. Antimicrob. Agents Chemother. 37: 377–383.
bain.com/bainweb/PDFs/cms/Marketing/rebuilding big 38. Gualtieri, M., F. Baneres-Roquet, P. Villain-Guillot, et al.
pharma.pdf (accessed October 23, 2010). 2009. The antibiotics in the chemical space. Curr. Med.
20. Hughes, D. 2003. Exploiting genomics, genetics and chem- Chem. 16: 390–393.
istry to combat antibiotic resistance. Nat. Rev. Genet. 4: 39. Drewry, D.H. & R. Macarron. 2010. Enhancements of
432–441. screening collections to address areas of unmet medical
21. NCBI Genome Project. http://www.ncbi.nih.gov/genomes/ need: an industry perspective. Curr. Opin. Chem. Biol. 14:
lproks.cgi (accessed October 23, 2010). 289–298.
22. Lock, R.L., & E.J. Harry. 2008. Cell-division inhibitors: 40. Hann, M.M., A.R. Leach & G. Harper. 2001. Molecular
new insights for future antibiotics. Nat. Rev. Drug Discov. complexity and its impact on the probability of finding
7: 324–338. leads for drug discovery. J. Chem. Inf. Model. 41: 856–
23. Gentry, D.R., K.A. Ingraham, M.J. Stanhope, et al. 2003. 864.
Variable sensitivity to bacterial methionyl-tRNA synthetase 41. Snowden, M.A. & D.V.S. Green. 2008. The impact of
inhibitors reveals subpopulations of Streptococcus pneumo- diversity-based, high-throughput screening on drug dis-
niae with two distinct methionyl-tRNA synthetase genes. covery: “chance favours the prepared mind.” Curr. Opin.
Antimicrob. Agents Chemother. 47: 1784–1789. Drug Dis. Dev. 11: 553–558.
24. Payne, D.J., W.H. Miller, V. Berry, et al. 2002. Discov- 42. Tan, D.S. 2005. Diversity oriented synthesis: exploring the
ery of a novel and potent class of FabI-directed an- intersections between chemistry and biology. Nat. Chem.
tibacterial agents. Antimicrob. Agents.Chemother. 46: 3118– Biol. 1: 74–84.
3124. 43. Galloway, W.R.J.D., A. Bender, M. Welch & D.R. Spring.
25. Du, W., J.R. Brown, D.R. Sylvester, et al. 2000. Two active 2009. The discovery of antibacterial agents using diversity-
forms of UDP-N-Acetylglucosamine enolpyruvyl trans- oriented synthesis. Chem. Commun. 18: 2446–2462.
ferase in Gram-positive bacteria. J. Bacteriol. 182: 4146– 44. Jacoby, E. & A. Mozzarelli. 2009. Chemogenomic strategies
4152. to expand the bioactive chemical space. Curr. Med. Chem.
26. Harris, S.R., E.J. Feil, M.T.G. Holden, et al. 2010. Evolution 16: 4374–4381.
of MRSA during hospital transmission and intercontinen- 45. Newman, D.J. & G.M. Cragg. 2007. Natural products as
tal spread. Science 327: 469–474. sources of new drugs over the last 25 years. J. Nat. Products
27. Brinster, S., G. Lamberet, B. Staels, et al. 2009. Type II 70: 461–477.
fatty acid synthesis is not a suitable antibiotic target for 46. Butler, M.S. & A.D. Buss. 2006. Natural products—the fu-
Gram-positive pathogens. Nature 458: 83–86. ture scaffolds for novel antibiotics? Biochem. Pharmacol.
28. Balemans, W., N. Lounis, R. Gilissen, et al. 2010. Essen- 71: 919–929.
tiality of FASII pathway for Staphylococcus aureus [arising 47. Pelaez, F. 2006. The historical delivery of antibiotics from
from Nature, 2009, 458: 83–86]. Nature 463: E3. microbial natural products—can history repeat? Biochem.
29. Brinster, S., G. Lamberet, B. Staels, et al. 2010. Replying to Pharmacol. 71: 981–990.
W. Balemans et al. 2009, Nature 463: E3. Nature 463: E4. 48. Li, J.W.H. & J. C. Vederas. 2009. Drug discovery and natural
30. Blundell, T.L., H. Jhoti & C. Abell. 2002. High-throughput products: end of an era or an endless frontier? Science 325:
crystallography for lead discovery in drug design. Nat. Rev. 161–165.
Drug Discov. 1: 45–54. 49. Feher, M. & J.M. Schmidt. 2003. Property distributions:
31. Arkin, M.R. & J.A. Wells. 2004. Small-molecule inhibitors differences between drugs, natural products, and molecules
of protein-protein interactions: progressing towards the from combinatorial chemistry. J. Chem. Inf. Model. 43:
dream. Nat. Rev. Drug Discov. 3: 301–317. 218–227.

Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences. 17
Challenges of antibacterial discovery Gwynn et al.

50. Koehn, F.E. & G.T. Carter. 2005. The evolving role of inhibitor pharmacophore. Proc. Natl. Acad. Sci. USA 106:
natural products in drug discovery. Nat. Rev. Drug Discov. 1737–1742.
4: 206–220. 69. Poole, K. 2007. Efflux pumps as antimicrobial resistance
51. Baltz, R.H. 2008. Renaissance in antibacterial discovery mechanisms. Ann. Med. 39: 162–176.
from actinomycetes. Curr. Opin. Pharmacol. 8: 557–563. 70. Lomovskaya, O., H.I. Zgurskaya, M. Totrov & W.J. Watkins.
52. Chetan S.C.K., J.L. Jenkins, R.E.J. Beckwith, et al. 2009. 2007. Waltzing transporters and ‘the dance macabre’ be-
Plate-based diversity selection based on empirical HTS data tween humans and bacteria. Nat. Rev. Drug Discov. 6: 56–
to enhance the number of hits and their chemical diversity. 66.
J. Biomol. Screen. 14: 690–699. 71. Lomovskaya, O. & K.A. Bostian. 2006. Practical applica-
53. Nielsen, F.H. 2009. Micronutrients in parenteral nutrition: tions and feasibility of efflux pump inhibitors in the clinic—
boron, silicon, and fluoride. Gastroenterology 137: S55–S60. a vision for applied use. Biochem. Pharmacol. 71: 910–
54. Baker, S.J., C.Z. Ding, T. Akama, et al. 2009. Therapeu- 918.
tic potential of boron-containing compounds. Future Med. 72. Surawicz, C.M. 2005. Antibiotic-associated diarrhea and
Chem. 1: 1275–1288. pseudomembranous colitis: are they less common with
55. Systemic Anti-Infectives Pipeline. Anacor Pharmaceuti- poorly absorbed antimicrobials? Chemotherapy 51(Suppl.
cals. 2009. http://www.anacor.com/ (accessed October 23, 1): 81–89.
2010). 73. Kelly, C.P., C. Pothoulakis & J.T. LaMont. 1994. Clostridium
56. Rock, F.L., W. Mao, A. Yaremchuk, et al. 2007. An antifungal difficile colitis. N. Engl. J. Med. 330: 257–262.
agent inhibits an aminoacyl-tRNA synthetase by trapping 74. Bartlett, J.G. 1992. Antibiotic-associated diarrhea. Clin. In-
tRNA in the editing site. Science 316: 1759–1761. fect. Dis. 15: 573–579.
57. Seiradake, E., W. Mao, V. Hernandez, et al. 2009. Crystal 75. Bouza, E., A. Burillo & P. Munoz. 2006. Antimicrobial ther-
structures of the human and fungal cytosolic leucyl-tRNA apy of Clostridium difficile-associated diarrhea. Med. Clin.
synthetase editing domains: a structural basis for the ratio- North Am. 90: 1141–1163.
nal design of antifungal benzoxaboroles. J. Mol. Biol. 390: 76. Credito, K.L. & P.C. Appelbaum. 2004. Activity of OPT-
196–207. 80, a novel macrocycle, compared with those of eight
58. Clark, M.A., R.A. Acharya, C.C. Arico-Muendel, et al. 2009. other agents against selected anaerobic species. Antimicrob.
Design, synthesis and selection of DNA-encoded small- Agents Chemother. 48: 4430–4434.
molecule libraries. Nat. Chem. Biol. 5: 647–654. 77. Finegold, S.M., D. Molitoris, M.-L. Vaisanenet, et al.
59. Chessari, G. & A.J. Woodhead. 2009. From fragment to 2004. In vitro activities of OPT-80 and comparator drugs
clinical candidate—a historical perspective. Drug Discov. against intestinal bacteria. Antimicrob. Agents Chemother.
Today 14: 668–675. 48: 4898–4902.
60. Mochalkin, I., J.R. Miller, L. Narasimhan, et al. 2009. Dis- 78. Larson, H.E. & S.P. Borriello. 1990. Quantitative study of
covery of antibacterial biotin carboxylase inhibitors by vir- antibiotic-induced susceptibility to Clostridium difficile en-
tual screening and fragment-based approaches. ACS Chem. terocecitis in hamsters. Antimicrob. Agents Chemother. 34:
Biol. 4: 473–483. 1348–1353.
61. Brotz-Oesterhelt, H., D. Beyer, H.-P. Kroll, et al. 2005. Dys- 79. Shue, Y.K., P.S. Sears, S. Shangle, et al. 2008. Safety, toler-
regulation of bacterial proteolytic machinery by a new class ance, and pharmacokinetic studies of OPT-80 in healthy
of antibiotics. Nat. Med. 11: 1082–1087. volunteers following single and multiple oral doses. An-
62. Gentry, D.R., I. Wilding, J.M. Johnson, et al. 2010. A rapid timicrob. Agents Chemother. 52: 1391–1395.
microtiter plate assay for measuring the effect of com- 80. Lowry, I., D.C. Molrine, B.A. Leav, et al. 2010. Treatment
pounds on Staphylococcus aureus membrane potential. J. with monoclonal antibodies against Clostridium difficile
Microbiol. Methods. 83: 254–256. toxins. N. Engl. J. Med. 362: 197–205.
63. Stokes, N.R., J. Sievers, S. Barker, et al. 2005. Novel in- 81. Bebbington, C. & G. Yarranton. 2008. Antibodies for the
hibitors of bacterial cytokinesis identified by a cell-based treatment of bacterial infections: current experience and
antibiotic screening assay. J. Biol. Chem. 280: 39709–39715. future prospects. Curr. Opin. Biotech. 19: 613–619.
64. Wang, J., S.M. Soisson, K. Young, et al. 2006. Platensimycin 82. Obrecht, D., J.A. Robinson, F. Bernardini, et al. 2009. Recent
is a selective FabF inhibitor with potent antibiotic proper- progress in the discovery of macrocyclic compounds as
ties. Nature 441: 358–361. potential anti-infective therapeutics. Curr. Med. Chem. 16:
65. Miesel, L., J. Greene & T.A. Black. 2003. Genetic strategies 42–65.
for antibacterial drug discovery. Nat. Rev. Genet. 4: 442– 83. Srinivas, N., P. Jetter, B.J. Ueberbacher, et al. 2010.
456. Peptidomimetic antibiotics target outer-membrane bio-
66. Andries, K., V. Peter, J. Guillemont, et al. 2005. A di- genesis in Pseudomonas aeruginosa. Science 327: 1010–
arylquinoline drug active on the ATP synthase of Mycobac- 1013.
terium tuberculosis. Science 307: 223–227. 84. Wimberly, B.T. 2009. The use of ribosomal crystal struc-
67. Pathania, R., S. Zlitni, C. Barker, et al. 2009. Chemical ge- tures in antibiotic drug design. Curr. Opin. Invest. Drugs
nomics in Escherichia coli identifies an inhibitor of bacterial 10: 750–765.
lipoprotein targeting. Nat. Chem. Biol. 5: 849–856. 85. Augmentin label approved 12/04/2008. 2008. FDA.gov
68. Miller, J.R., S. Dunhama, I. Mochalkin, et al. 2009. A class http://www.accessdata.fda.gov/drugsatfda docs/label/2008
of selective antibacterials derived from a protein kinase /050564s050lbl.pdf (accessed October 23, 2010).

18 Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences.
Gwynn et al. Challenges of antibacterial discovery

86. Avandia label approved 10/20/2008. 2008. FDA.gov ics and pharmacodynamics of antimicrobial drugs. Expert
http://www.accessdata.fda.gov/drugsatfda docs/label/2008 Opin. Drug Metab. Toxicol. 5: 475–487.
/021071s034lbl.pdf (accessed October 23, 2010). 96. Andes, D. & W.A. Craig. 2002. Animal model pharma-
87. Werner, A.L. & M.T. Travaglini. 2001. A review of rosigli- cokinetics and pharmacodynamics: a critical review. Int. J.
tazone in type 2 diabetes mellitus. Pharmacotherapy 21: Antimicrob. Agents 19: 261–288.
1082–1099. 97. Courvalin, P. 2008. Can pharmacokinetic–pharmaco-
88. Imitrex label approved 10/13/2004. 2004. FDA.gov dynamic parameters provide dosing regimens that are less
http://www.accessdata.fda.gov/drugsatfda docs/label/2004 vulnerable to resistance? Clin. Microbiol. Infect. 14: 989–
/20626s004lbl.pdf (accessed October 23, 2010). 994.
89. Perry, C.M. & M. Anthony. 1998. Sumatriptan. An updated 98. Drusano, G.L., S.L. Preston, C. Hardalo, et al. 2001. Use of
review of its use in migraine. Drugs 55: 889–922. preclinical data for selection of a phase II/III dose for evern-
90. Lasser, K.E., P.D. Allen, S.J. Woolhandler, et al. 2002. imicin and identification of a preclinical MIC breakpoint.
Timing of new black box warnings and withdrawals for Antimicrob. Agents Chemother. 45: 13–22.
prescription medications. J. Am. Med. Assoc. 287: 2215– 99. Bhavnani, S.M., J.P. Hammel, B.B. Cirincione, et al. 2005.
2220. Use of pharmacokinetic-pharmacodynamic target attain-
91. Raschi, E., L. Ceccarini, F. De Ponti & M. Recanatini. ment analyses to support phase 2 and 3 dosing strategies for
2009. hERG-related drug toxicity and models for predict- doripenem. Antimicrob. Agents Chemother. 49: 3944–3947.
ing hERG liability and QT prolongation. Expert Opin. Drug 100. Andes, D., M.L. van Ogtrop, J. Peng, et al. 2002. In vivo
Metab. Toxicol. 5: 1005–1021. pharmacodynamics of a new oxazolidinone (linezolid). An-
92. Chaudhary, K.W. & B.S. Brown. 2010. Assessment of timicrob. Agents Chemother. 46: 3483–3489.
strategies utilized to minimize the potential for induc- 101. Pertel, P.E., P. Bernardo, C. Fogarty, et al. 2008. Effects
tion of acquired long QT syndrome and torsade de of prior effective therapy on the efficacy of daptomycin
pointes. In Evaluation of Drug Candidates for Preclini- and ceftriaxone for the treatment of community-acquired
cal Development: Pharmacokinetics, Metabolism, Pharma- pneumonia. Clin. Infect. Dis. 46: 1142–1151.
ceutics, and Toxicology, Vol. 10. C. Han, B.D. Charles 102. Silverman, J.A., L.I. Mortin, A.D.G. VanPraagh, et al. 2005.
& B. Wang, Eds.: 253–278. John Wiley & Sons, Inc. Inhibition of daptomycin by pulmonary surfactant: in vitro
Hoboken. modeling and clinical impact. J. Infect. Dis. 191: 2149–2152.
93. Leeson, P.D. & A.M. Davis. 2004. Time-related differences 103. Guidos, R.J. 2009. New Transatlantic Task Force on
in the physical property profiles of oral drugs. J. Med. Chem. Antimicrobial Resistance: A Path Forward http://www
47: 6338–6348. .cgdev.org/content/general/detail/1423276/ (accessed Oc-
94. Leeson, P.D. & B. Springthorpe. 2007. The influence of tober 23, 2010).
drug-like concepts on decision-making in medician chem- 104. Infectious Diseases Society of America. Bad Bugs,
istry. Nat. Rev. Drug Discov. 6: 881–890. No Drugs . . . 10 New Antibiotics by 2020, http://www.
95. Czock, D., B. Hartman & F. Keller. 2009. Pharmacokinet- idsociety.org (accessed October 23, 2010).

Ann. N.Y. Acad. Sci. 1213 (2010) 5–19 


c 2010 New York Academy of Sciences. 19

You might also like