Genetic Approaches to Facilitate Antibacterial Drug Development

16 Pages·2015·0.64 MB·English

Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
Genetic Approaches to Facilitate Antibacterial
Drug Development
DirkSchnappinger
DepartmentofMicrobiologyandImmunology,WeillCornellMedicalCollege,NewYork,NewYork10065
Correspondence:[email protected]
Very few chemically novel agents have been approved for antibacterial chemotherapies
during the last 50yr. Yet new antibacterial drugs are needed to reduce the impact on
globalhealthofanincreasingnumberofdrug-resistantinfections,includinghighlydrug-
resistantformsoftuberculosis.Thisreviewdiscusseshowgeneticapproachescanbeusedto
studythemechanismofactionofwhole-cellscreeninghitsandfacilitatetarget-drivenstrat-
egiesforantimicrobialdrugdevelopment.
Manyclassesofantibioticsinclinicaluseto- ease that now constitutes as a major public
daystemfromthegoldenageofantibiotic healththreat.First-linetreatmentofTBconsists
g discovery,whichhaditsmostproductiveperiod ofdrugsoriginallyidentifiedduringthegolden
r
o from 1940 to 1960 (Davies 2006; Silver 2011). ageofantibioticdiscovery.Togetherwithsocial
e.
n These drugs are one of the reasons why most changes, these drugs contributed to a drastic
ci
di bacterialinfectionscanbecuredwithafewpills, reduction of TB cases in Europe and North
e
m generallywithoutsideeffects,withinacoupleof AmericawhilecontinuingtokillmillionsinAf-
n
si weeks.Itisthuseasytoforgetthatbacterialin- ricaandAsia.Drug-resistantMycobacteriumtu-
e
v fectionshaveremainedamajorcauseofprevent- berculosis(Mtb)appearedshortlyaftertheintro-
cti abledeathsindevelopingcountriesandcontin- duction of TBchemotherapyandevolved into
e
p
s uetoexertaprofoundimpactonhumanhealth. extremely drug-resistant Mtb, which threatens
r
e
p Therelativepaucityofnovelantibacterialsdis- theTBcontrolprogramsworldwide.Therefore,
w.
covered after 1960 coincided with the appear- theneedfornewtherapiesfortreatmentofthis
w
w anceandcontributedtothespreadofdrug-re- deadlydiseaseremainsunchangedsincethepre-
sistant bacterial infections. When resistant to antibioticera.Thisreviewdiscusseshowgenetic
notonlyone,butmultipledrugs,suchinfections approaches can facilitate the development of
threaten to erode a cornerstone of modern drugsforsuchnewtherapies.
healthcare.
Tuberculosis(TB)isaprimeexampleofhow
GENETICSANDCLASSICALWHOLE-CELL
treatmenthassucceededindevelopedcountries
SCREENING
whilemeetingfrequentfailureinthedeveloping
nations.Treatmentfailurecontributedtothere- Whole-cellscreening(WCS)-basedantibacteri-
emergence of a thought-to-be-conquered dis- aldrugdiscoverybeginswithidentifyingsmall
Editors:StefanH.E.Kaufmann,EricJ.Rubin,andAlimuddinZumla
AdditionalPerspectivesonTuberculosisavailableatwww.perspectivesinmedicine.org
Copyright#2015ColdSpringHarborLaboratoryPress;allrightsreserved
AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139
1
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
D.Schnappinger
molecules (hits) that inhibit bacterial growth. ofthecompoundcanbeexploredinmorede-
Thisinhibitor-firstapproachprovidedthefoun- tail. Fourth, and perhaps most importantly,
dation of the golden age of antibacterial drug the antibacterialMOAofacompoundisade-
discovery and remains the only development terminantofitsinvivopotency.Thisisexem-
strategy that has delivered antibacterial drugs plified by recently identifiedpyrimidine-imid-
to the clinic (Brotz-Oesterhelt and Sass 2010). azoles, which were active against Mtb in vitro,
This includes all first-in-class drug candidates had appropriate pharmacokinetic properties,
currently under clinical development for the but showed no potency in Mtb-infected mice
treatment of TB (Lechartier et al. 2014). The (Petheetal.2013).Theactivityofthesemole-
superiorsuccessrateofinhibitor-firststrategies culesreliedonthepresenceofglycerol,whichis
isnotuniquetoantibacterialdrugdevelopment used in many standard liquid media, but it is
asphenotypicscreeninghasbeenmoreproduc- not a physiological carbon source. Therefore,
tivethantarget-drivenapproachesacrossaspec- lackofinvivopotencyinthesepyrimidine-im-
trumofdiseases(SwinneyandAnthony2011). idazoleswasmechanism-based,leavingnopath
Severalfeaturescontributetothesuccessof todevelopthesecompoundsfurther.
classicalWCS.Thesescreensallowforselective The potential for discovering other com-
investmentintothechemicalstructuresthatcan pounds with MOAs that intrinsically have no
overcome the major challenge of antibacterial invivopotency(herereferredtoasnovipcom-
drugdevelopment:abilitytopenetratethebac- pounds)issignificantbecausescreeningmedia
terialcellenvelope.Furthermore,WCSprovides can, at best, only mimic some of the environ-
anopportunitytodiscovercompoundsthatin- ments encountered during an infection. More
hibitgrowthbydiversemechanisms,including often, they create a physiological context that
g those that engage multiple targets. The draw- is quite different from that within the host as
r
o back isthat the mechanism of all hits remains many standard media contain nutrients Mtb
e.
n tobedetermined. cannot obtain from the host in the provided
ci
di Whereas it is not required to understand a form (e.g., iron) (Rodriguez and Smith 2006),
e
m drug candidate’s mechanism of action (MOA) cannotobtainfromthehostatall(e.g.,biotin)
n
si to obtain regulatory approval for use in hu- (Sassetti and Rubin 2003; Woong Park et al.
e
v mans,knowing howacompoundexertsitsbi- 2011),andlacknutrientsthatMtbisapparently
cti ologicalfunctionduringdevelopmentisdesir- abletoscavengefromthehost (e.g.,NADpre-
e
p
s able for several reasons. First, many WCS hits cursors)(Boshoffetal.2008).Thesedifferences
r
e
p actinanonspecificmanner(e.g.,asalkylating can cause some mutations to either result in
w.
w agents or detergents) and therefore cannot be deathofMtbinvitrobutnotduringinfections
w developedintodrugs(Payneetal.2007;Roemer (e.g.,mutationsinnadA[Boshoffetal.2008])
and Boone 2013). Second, not all hitsthat are orpreventgrowthduringinfectionsbutnotin
toxictothehosthavetobediscardedastoxicity standard media (e.g., mutations required to
neednotnecessarilyresultfromacompound’s scavenge iron from the host [Rodriguez and
antibacterialactivity.Antibacterialactivityand Smith 2006]). One conceptually straightfor-
toxicity are linked if they both stem from in- ward strategy to reduce the risk of identifying
hibition of the same target, for example, an novipcompoundsistoscreenagainstintracellu-
enzymethatisconservedinboththepathogen larMtb(Christopheetal.2009).However,such
and the host. Such compounds are not suited screeningprocedurescanalsorevealgenesthat
for further development. However, in cases in arerequiredexvivobutnotduringaninfection
which the mechanism of toxicity differs from (Munoz-Eliasetal.2006).Furthermore,Mtbin-
that of bacterial growth inhibition, chemical ducesacomplexanddynamicpathologyandcan
modificationscansometimeseliminateacom- befoundinmultipledifferentintracellularand
pound’stoxicitywithoutcompromisingitsan- extracellularlocations(Barryetal.2009;Dartois
tibacterial activity. Third, after identifying the andBarry2013).Someoftheselocationspermit
target,thestructureactivityrelationship(SAR) the pathogen to replicate, whereas others may
2 AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
AntibacterialDrugDevelopment
restrictitsgrowthandyetallowitspersistence. dium of reference profiles to which the RNA
Therefore,itseemsimpossibletodesignscreen- profile induced by dyclonine could be com-
ing conditions that mimic the pathophysiol- pared. Pattern matching algorithms revealed
ogical context of an Mtb infection in humans thatdycloninecausedchangessimilartothose
well enough to completely bias against novips. inducedbymutationsintheergosterolpathway.
ThebestinsuranceagainstMOA-basedattrition This suggested that dyclonine inhibited sterol
is therefore a solid understanding of a com- biosynthesis and follow-up experiments con-
pound’sMOA. firmedthesterolisomeraseErg2pasatargetof
dyclonine.Thispatternmatchingapproach,as
illustrated in Figure 1, has since been used to
GeneticStrategiesforStudyingthe
uncover targets and the MOA of many other
MOAofBioactiveSmallMolecules’
antimicrobialcompounds(BrazasandHancock
RNAProfiling
2005; Smith et al. 2010; Wecke and Mascher
Acell’sgenome-wideRNAprofileisdefinedby 2011).Indeed,Mtbmicroarrayswerefirstused
theabundanceofthetotalRNAproducedata to study the effect of isoniazid on the RNA
given time. Taking snapshots of these profiles profile of Mtb (Wilson et al. 1999). Hundreds
becamestraightforwardafterDNAmicroarrays of compound-induced RNA profiles were re-
wereinvented(Schenaetal.1995;Lashkarietal. ported5yrlaterinalandmarkstudybyBoshoff
1997),andtheycannowbereportedwitheven et al. (2004). These profiles helped to investi-
higherresolutionbysequencingtotalRNAex- gateavarietyofantimycobacterialsmallmole-
tracts (Nagalakshmi et al. 2008; Wilhelm et al. cules,includingcapreomycin(FuandShinnick
2008; Arnvig and Young 2012). The foremost 2007), isoniazid (Karakousis et al. 2008), PA-
g study to show the use of genome-wide RNA 824 (Manjunatha et al. 2009), an inhibitor
r
e.o profiling forcharacterizing the MOAof bioac- of menaquinone production (Dhiman et al.
n tivemoleculeswasperformedwiththeanesthet- 2009), b-lactams (Slayden and Belisle 2009),
ci
di icdyclonine(Hughesetal.2000).Hughesetal. vancomycin (Provvedi et al. 2009), the benzo-
e
m collectedSaccharomycescerevisiaeRNAprofiles thioazinone, BTZ043 (Makarov et al. 2009),
n
si induced by hundreds of different genetic or the natural product chelerythrine (Liang et al.
e
v chemicaltreatments.Thisgeneratedacompen- 2011),thioridazine(Thorsingetal.2013),and
cti
e
p
s
r
e
p
w.
w
w
Compound
No drug RIF INH of unknown
MOA
No drug
RIF
INH
Compound of unknown MOA
Figure1.RNAprofiling.RNAmoleculesproducedbyfourgenesoftheMtbgenomewithandwithoutdrug
treatmentarerepresentedascoloredlines.ThecolorofthesquaresdepictedbelowtheRNAmoleculesreflects
theirabundance,withdarkercolorsidentifyingmoreabundantRNAs.Thefictionaldataherewouldsuggestthat
thecompoundofunknownMOAfunctionsinamannersimilartoRIF.
AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139 3
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
D.Schnappinger
severalotherWCShits(Stanleyetal.2013;Wang parently identical experiments can be a chal-
etal.2013). lenge. Description of all experimental details,
The principal limitations of RNAprofiling strictstandardization, andavoidanceofproce-
aretwofold.First,these profiles failto identify duralchangesofgeneexpression(ascanoccur,
the direct target of a compound as they only forexample, bycentrifugation of live Mtb)are
reveal the biological pathway(s) that respond therefore crucial for producing informative
to the drug. In part, this is because inhibition RNAprofiles.
of different proteins can produce similar RNA
profiles,suchasinsituationsinwhichthepro-
SelectionandIdentificationofResistance
teins form a complex or are required for the
Mutations
same metabolic pathway. Second, the profiles
(and other pattern matching approaches) are Thefirstindicationthatbedaquiline(BDQ,for-
mostinformativeforcompounds,whichinduce merly known as TMC207 and originally pub-
a signature similar to the compounds with a lishedasR207910)preventsthegrowthofMtb
known MOA. Specific predictions are difficult byinhibitingtheadenosinetriphosphate(ATP)
tomakeforcompoundswithanentirelynovel synthasewasprovidedby themutantsselected
MOAasappropriatereferenceprofilesarelack- fortheirabilitytogrowinthepresenceofBDQ
ing in such cases. This drawback can be over- (Andriesetal.2005).Whole-genomesequenc-
come by enriching the reference compendium ingoftheseBDQ-resistantstrainsrevealedpoint
withRNAprofilesfromgeneticallydefinedmu- mutationsinatpE,whichencodessubunitcof
tants(Hughesetal.2000;Mnaimnehetal.2004; ATPsynthase.Geneticandbiochemicalstudies
Freiberg et al. 2005). Unfortunately, too few furtherestablishedthatBDQinhibitsATPsyn-
g such profiles are available for Mtb to apply thesisbybindingtoAtpE,thusconfirmingATP
r
o this principle to facilitate the analysis of anti- synthase as a direct target (Koul et al. 2007).
e.
n mycobacterialcompounds. Selecting for resistance followed by whole-ge-
ci
di Evenforrelativelysmallbacterialgenomes, nomesequencing(Fig.2A)hasidentifiedmu-
e
m genome-wide RNA profiles consist of thou- tations that permit growth in the presence of
n
si sands of individual measurements. This com- several other small molecules with antimyco-
e
v plexity offers the potential to generate unique bacterialactivities.Examplesincludemutations
cti signatures for many bioactive small molecules in genes involved in respiration (qcrB) (Abra-
e
p
s while making it difficult to identify the most hamsetal.2012b;Petheetal.2013),fattyacid
r
e
p informativeRNAs.Theamountof datagener- synthesis (inhA) (Hartkoorn et al. 2012), pro-
w.
w ated by RNA profiling is not only large, but teinsynthesis(aspS)(Ioergeretal.2013),pro-
w more importantly, the biological responses to teinsecretion(eccB3)(Ioergeretal.2013),poly-
growth perturbation are often complex. This ketidebiosynthesis(pks13)(Ioergeretal.2013;
foremost applies to RNA profiles collected at Wilson et al. 2013), mycolic acid transport
timepointsoratdrugconcentrationsthatim- (mmpL3)(Grzegorzewiczetal.2012;LaRosaet
pacttheabundanceofhundredsofRNAs.The al.2012;Stanleyetal.2012;Tahlanetal.2012;
nonspecific stress responses contained in such Ioergeretal.2013;Remuinanetal.2013),and
profiles can mask a compound-specific re- arabinogalactan synthesis (dprE1) (Christophe
sponse. Reducing the treatment time and/or et al. 2009; Makarovet al. 2009; Magnet et al.
drugconcentrationoftensimplifiestheanalysis 2010;Stanleyetal.2012;Wangetal.2013).
of bioactive molecules by RNA profiling. Fur- Resistance mutations can directly identify
thermore,theabilityofgenome-wideRNAanal- high-value drug targets. Selection and identi-
ysestodetectminutechangesinanysinglegene fication of such mutations requires no target-
coupledwitha highlydynamic transcriptional specific assay development, can be fast, and is
response prevalent in bacteria increases the comprehensiveasitsurveystheentiregenome.
existing complexity by severalfold. Therefore, However, this approach is not without limita-
obtainingverysimilarRNAprofilesintwoap- tionsandfailswhenaresistantstraincannotbe
4 AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
AntibacterialDrugDevelopment
A B C
Underexpressor
Select
Growth
drug-resistant
without drug
mutant
WT
Whole-genome
sequencing
Growth
with drug
Overexpressor
GATAAATCTGGTCTTATTTCC
g
Wild type g/u
Growth owth with druwth without dr
GATAAATCTGGCCTTATTTCC Drug Grgro Mutant
Figure2.ForwardandreversegeneticapproachestoinvestigatesmallmoleculeMOAs.(A)Selectionofresistant
mutants. Mutants that are resistant to the compound are selected on agar plates supplemented with the
g compound of interest. The mutations present in these resistant isolates are identified by whole-genome se-
r
e.o quencing.(B)Under/over-expression.Reducingorincreasingtheexpressionofacandidatetarget(blackcircles)
n caneitherdecreaseorincreasetheMICofacompound(yellowtriangle)inhibitingthetarget.(C)Phenotyping
dici pooledmutants.Mutantsarelabeledwithageneticbarcodeandtheirfitnessisanalyzedinacompetitivegrowth
e experiment.Thisprinciplecanbeappliedtodeletionmutants,underexpressors,overexpressors,oramixof
m
n thesestrains.Theorange-labelmutantshowsincreasedsensitivitytothetesteddrug;theblack-labeledmutant
esi (whichisonlyshowninthegraph)showsdecreasedsensitivity.
v
cti
e
p
s obtained. Furthermore,alterationofthetarget LinkingSmallMoleculestoCandidate
r
e
p responsible for the growth inhibitory effect of TargetsbyIncreasingorReducing
w.
w a small molecule constitutes only one of the GeneActivity
w several mechanisms by which mutations can
impart resistance. Alternatively, methods such TobedirectlyinvolvedintheMOAofabioactive
as chemical inactivation, efflux, or failure to small molecule, a protein hasto physically in-
transform an inactive prodrug into its active teract with that small molecule. That a candi-
derivativecanformthebasisofresistance.The date protein has the potential for such an in-
knowledge of mutation frequencies that cause teractionwithinlivebacteriacanbeconfirmed
resistance by these mechanisms proves highly indirectlybyshowinganimmediatecorrelation
useful to determine the value of acompound. betweentheprotein’sabundanceandthemin-
A high frequency of resistance owing to such imalinhibitorconcentrationrequiredtoinhib-
mechanisms significantly devalues the com- itgrowth(MIC)(Fig.2B).Thistypeofexperi-
pound. However, this knowledge rarely plays ment has helped to clarify the MOA of INH
a role in defining acompound’s MOA. Never- (isoniazid), which was debated to impair one
theless, in scenarios in which a mutation can of two enzymes involved in fatty acid biosyn-
bemappedtoatruetarget,resistancemutation thesis,InhAorKasA.Strainsthatoverexpressed
provides one of the simplest ways to define a theseproteinswereanalyzedfortheirsuscepti-
compound’sMOA. bilitiestoINHandtheKasAinhibitorthiolac-
AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139 5
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
D.Schnappinger
tomycin (TLM). Overexpression of InhA in- family will increase the MIC even when only
creased the MIC of INH but did not change one family member isthe physiological target.
the MIC of TLM; overexpression of KasA had This has been discussed for inosine-50-mono-
the opposite effect (Larsen et al. 2002). These phosphate (IMP) dehydrogenases (GuaB1/2/
resultssupportedtheviewthatInhAandKasA 3) (Usha et al. 2011), ketoacyl-ACP synthases
are primary targets of INH and TLM, respec- (KasA/B)(Brownetal.2009),andmethionine
tively.Similaroverexpressionstudieshaveoften aminopeptidases (MapA/B) (Olaleye et al.
beenperformedwithMycobacteriumsmegmatis 2010). In some of these cases, the MIC may
in which the MOA of different inhibitors was have increased because the overproduced pro-
linked to one or more targets, including D-cy- tein sequestered the inhibitor from its physio-
closerine to alanine racemase (Alr) and D-ala- logicaltarget.
D-alaligase(Ddl)(FengandBarletta2003),pla- As target overexpression should increase a
tensimycin to KasA and KasB (Brown et al. compound’s MIC, so should target underex-
2009),ethambutoltothearabinosyltransferases pression decrease the MIC (Fig. 2B). This was
EmbBandEmbC(Goudeetal.2009),andcer- indeedobservedforseveralconditionalM.smeg-
tain diphenylurea compounds to inosine mo- matis knockdown mutants described recently.
nophosphate dehydrogenase (Olaleye et al. Underexpression of gyrase subunit A (GyrA,
2010;Ushaetal.2011). target of quinolones [Sugino et al. 1977], Alr
In mycobacteria, overexpression has pri- [Caceres et al. 1997], dihydrofolate reductase
marilybeenusedtoevaluatetheroleofspecif- [DHFR, target of trimethoprim] [Schweitzer
ic candidate target proteins for compound etal.1990],RNApolymerasesubunitB[RpoB,
activity,butitcanalsoprovideameanstofind targetofRIF][Lancinietal.1969]),andprotein
g candidate targets foracompound with an en- biotin ligase (BirA, target of BioAMS [Duck-
r
o tirely unknown MOA. To achieve this, librar- worth et al. 2011]) sensitized M. smegmatisto
e.
cin ies of overexpression clones are systematically ciprofloxacin, D-cycloserine, trimethoprim,
di screened for those that improve growth in the RIF, and BioAMS, respectively (Duckworth
e
m presenceoftheinhibitor.Suchascreen,which et al. 2011; Kim et al. 2010; Wei et al. 2011).
n
si usedanorderedlibraryofoverexpressionclones Similar observations were made with Mtb, in
e
v forallessentialEscherichiacoligenes,identified whichpartialgeneticinactivationofthebiotin
cti the lipoprotein chaperon,LolA, asacandidate synthesis enzyme BioA or DHFR increased its
e
p
s target of the novel antibacterial compound, susceptibility toward the novel inhibitors of
r
e
p MAC12343. Biochemical studies confirmed these enzymes (Shi et al. 2011; Kumar et al.
w.
w that MAC12343 binds LolA and perturbs 2012). For many of these mutants, it was also
w lipoprotein trafficking (Pathania et al. 2009). shownthattheyweremoststronglysensitizedto
Similaroverexpressionlibrariesareundercon- inhibitors of the partially depleted protein.
struction forMtbandwillhopefullybeavaila- These and otherdata(Giaeveret al. 1999; De-
blesoon. Vitoetal.2002;Lumetal.2004;Yinetal.2004;
A complication of overexpression screens, Donaldetal.2009;Abrahamsetal.2012a;Ol-
especially when using random libraries, is the lingeretal.2012)confirmedthatpartialdeple-
frequent isolation of efflux pumps. These tionofaninvitroessentialproteinoftencauses
pumpsincreaseresistancebyloweringtheintra- compound-specific MIC shifts and can thus
cellularconcentrationofthecompoundunder provideinformationontheMOAofabioactive
investigation, and their isolation does not re- smallmolecule.However,asisthecaseforover-
vealacompound’sMOA.Difficulttointerpret expression, underexpression studies by them-
MIC shifts can also occurif the candidate tar- selvesarenotsufficienttoclaimtheunderpro-
getbelongstoaproteinfamilyofwhichsever- ducedproteinastheMICdeterminingtargetof
al members are encoded in the same genome. aninhibitor,asunderexpressioncouldalsode-
Iftheinhibitorbindingsiteisconserved,over- creasetheMICbysensitizingotherenzymesin
expression of any member of such a protein thesamepathwaytochemicalinhibition.
6 AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
AntibacterialDrugDevelopment
Thatunderexpressionneverthelessprovides pound’s activity. Evidence for this came from
a powerful tool for identifying candidate tar- several studies including those that used the
gets is apparent from work with S. cerevisiae. E.coliKeiocollection.Thislibrarycontainsal-
Construction of underexpressors is relatively most 4000 mutants each of which has a sin-
straightforward in S. cerevisiae and is often gle gene deleted. Of these 4000 mutants, 283
achieved with deletion of one gene copy. Such showed greater sensitivity thanwild type to at
heterozygous deletion strains are available for least one of the 14 different antibiotics tested
almost all S. cerevisiae genes including those (Tamae et al. 2008; Liu et al. 2010). Whereas
thatareessentialfornormalgrowth.Thismu- some strains were more susceptible to several
tant collection has been used extensively to classesofantibiotics,otherswereonlysensitized
study the MOA of bioactive small molecules tocompoundswiththesameMOA(Girgisetal.
by drug-induced haploinsufficiency profiling 2009;Liuetal.2010).Thedrugsensitivitypro-
(HIP). HIP measures the effect of sublethal filesofthelattermutantscanbeusedtoclassifya
drugconcentrationsontherelativegrowthrates compound’s MOA with strategies similar to
ofindividualmutants(Fig.2C).Primarily,these thoseimplemented inRNAprofiling.Thedis-
studies confirmed that the mutants of known criminative power of these profiles can be en-
targetsofantimicrobialagentsareamongthose riched with mutations that decrease antibiotic
mostsusceptibletogrowthinhibitionbythere- susceptibility,whichinE.coliareasfrequentas
spectivedrug(Giaeveretal.1999,2004;Lumet those that increase susceptibility (Girgis et al.
al. 2004) and were later implemented in iden- 2009).
tification of novel candidate targets for estab- The homozygous deletion mutants in S.
lisheddrugs.ThisincludedtherRNAprocessing cerevisiaeareequivalenttotheE.coliKeiomu-
g exosome,whichwaspredictedbyHIP(Giaever tants.Chemicalprofilingofsuchmutants(ho-
r
o etal.2004;Lumetal.2004)andconfirmedby mozygous deletion profiling, HOP) has been
e.
n variousexperiments(Fangetal.2004;Hoskins particularly useful for compounds whose pri-
ci
di andScottButler2007;HoskinsandButler2008; mary target is not a protein. For example,
me Kammleretal.2008)tobeinvolvedintheMOA HOP of (cid:2)4700 S. cerevisiae mutants revealed
n
si of the anticancer drug 5-fluorouracil. In the thatfitnessdefectscausedbydeletionofgenes
e
v opportunisticpathogenCandidaalbicans,HIP involved in DNA repair or in survival during
cti helpedtoidentifytheguaninemonophosphate DNAdamagecanbediagnosticforDNA-dam-
e
p
s synthase(Rodriguez-Suarezetal.2007),afatty aging agents. In fact, the profiles of these mu-
r
e
p acid desaturase (Xu et al. 2009), and poly(A) tants were distinct for chemicals that damage
w.
w polymerase(Jiangetal.2008)astargetsofnovel DNA by different mechanisms as they could
w inhibitors.Experimentsconceptuallysimilarto distinguish compounds that generate inter-
HIPhavebeenperformedinbacteriawheregene strand cross-links from those that do not (Lee
expression can be efficiently reduced by anti- et al. 2005). The interpretation of HOPexper-
sense RNAs (Donald et al. 2009; Huber et al. iments can be facilitated by data from genetic
2009; Xu et al.2010). Unfortunately,antisense interaction studies. Two genes are defined as
RNAs have only rarely been efficient in Mtb. interactingifthephenotypeoftheirdoublemu-
Currently, the generation of a large collection tant deviates from the expected phenotypes of
ofMtbunderexpressorsdependsonapproaches thetwosinglegenemutants(Dixonetal.2009).
thatrequirehomologousrecombinationduring InS.cerevisiae,manysuchdigenicinteractions
mutantconstruction,suchastranscriptionalor havebeendefined.Theseinteractionsarevalu-
proteolyticsilencing(Ehrtetal.2005;Weietal. able for analyses of bioactive small molecules
2011). asthe genetic interaction profile of a gene can
WCS hits are growth inhibitory by defini- besimilar totheHOPprofileofasmallmole-
tion and thus should primarily target the es- cule and thus link the small molecule to the
sential gene products. Nevertheless, mutations respectivegene(Parsonsetal.2004,2006;Cos-
in nonessential genes often modulate a com- tanzoetal.2010).
AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139 7
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
D.Schnappinger
FACILITATINGTARGET-BASEDDRUG toxicityinhibitors,(4)amenabletobiochemical
DISCOVERY HTS,and(5)requiredforbacterialgrowth(or
survival) in a manner that its partial inactiva-
Target-based approaches typically begin with
tion suffices to improve a patient’s health.
a screen for inhibitors of a purified enzyme.
Showing that a target fulfills all these criteria
Unfortunately,mostinhibitorsidentifiedinsev-
requires asmall-moleculeinhibitor thatissafe
eralhigh-throughputscreenings(HTSs)against
for use in humans (i.e., a drug). Trying to de-
bacterial enzymes had no activity against live
velop drugs against new targets, therefore, al-
bacteria (Payne et al. 2007). Nevertheless, tar-
waysbearstheadditionalriskthatcomesfrom
get-basedapproacheshavetheirownadvantag-
working against a target that is yet to be fully
es.Primarily,MOAstudiesaremorestraightfor-
validated.Thisadditionalriskappliestoinhib-
wardforatarget-basedapproachthanforWCS
itorsidentifiedfromeitherWCSorbiochemical
hits.Forexample,abisubstrateinhibitor(Bio-
screensandisoneofthefactorscausingthehigh
AMS)designedtoinactivatetheproteinbiotin
rate of attrition of compounds with entirely
ligase (BirA) was recently found to inhibit the
novelMOAs(KolaandLandis2004).
growthofMtb.ThestudyconductedbyDuck-
Druggabilityassessmentsevaluatepotential
worthetal.(2011)examinedtheimpactofBio-
bindingpocketsbycomputationalorchemical
AMSongrowthandbiotinylationofproteinsin
methods(Perotetal.2010;BarelierandKrimm
M.smegmatisandMtb.Asexpected,theactivity
2011; Fauman et al. 2011); sequence conser-
of BioAMS (but not of INH or ethambutol)
vation—within bacteria and between bacterial
againstMtbandM.smegmatisdirectlycorrelat-
and human homologs—can be assessed with
edwithexpressionofBirA,andgrowthinhibi-
genomesequencedata.Animportantcontribu-
tioncoincidedwiththedepletionofbiotinylated
g tion of genetics to the evaluation of potential
or proteins.Together,thesestraightforwardobser- targetsisitsabilitytodeterminetheimportance
ne. vationsconfirmedBirAastheprimarytargetof of a potential target for growth and survival.
ci BioAMS (Duckworth et al. 2011). The second
di Themostcompletesourceforthisinformation
e main advantage of target-based approaches is
m stems from transposon mutagenesis. Many of
n that they can survey a much larger chemical
si the genes that Mtb requires to grow normally
e space thanwhat can be covered in phenotypic
v in vitro (Sassetti et al. 2001, 2003), in macro-
cti screens,especiallyiffragment-basedapproaches phages (Rengarajan et al. 2005), or in mouse
e
p areincluded(Scottetal.2012).Thisissignifi-
s spleens (Sassetti and Rubin, 2003) were de-
r
e cant because the physicochemical properties
p finedbytransposonsitehybridization(TraSH).
w. ofmostmoleculesinatypicalWCSlibraryare
w TraSHusesDNAmicroarraystosimultaneously
w different from those of known antibacterials mapandanalyzethephenotypicconsequences
(Brotz-OesterheltandSass2010).
ofthousandsoftransposoninsertions.Recently,
Tn-seq, which uses next-generation DNA se-
quencinginsteadofmicroarrays,hasimproved
ThePowerandLimitsofGeneticTarget
these analyses by providing single-nucleotide
Validation
resolutionandalargerdynamicrange(Griffin
Choosing a poor target will doom a target- et al. 2012; Zhang et al. 2012, 2013). TraSH
based drug discovery program from the start, andTn-seqhavecontributedmoretothephe-
yetthismightbecomeobviousonlyafter years notypic characterization of gene functions in
of work and millions of dollars have been in- Mtbthananyothergeneticapproach.However,
vested. Identifying appropriate targets is thus they are not ideal tovalidate gene products as
crucial. Ideally, the selected protein should be targets for drug development for two reasons.
(1)conservedamongallisolatesofthetargeted First, TraSH and Tn-seq can identify in vitro
pathogen(s), (2) susceptible to inhibition bya essentialgenes,butneitherapproachgenerates
smallmolecule, (3)differentenoughfromany mutantsforsuchgenesthatcould beanalyzed
human homolog to allow development of low underavarietyofconditions.Inthesameman-
8 AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
AntibacterialDrugDevelopment
nerasWCScanidentifynovipcompounds,ge- ficient to rapidly eliminate culturable bacteria
neticscreenscanidentifynoviptargets(i.e.,tar- from mouse lungs and spleens during acute
getswhoseinactivationkillsMtbunderaspecific and chronic infections (Kim et al. 2013). Es-
in vitro condition but does not prevent per- sentialityofNadEforviabilityofMtbunderall
sistenceorevengrowthinanimalsorhumans). conditions tested so far strongly suggests that
Itisthereforeimportanttoevaluatetheroleofa this enzyme is also required for growth and
geneproductforsurvivalunderavarietyofcon- survival under most (if not all) conditions
ditions including those that induce nonrepli- that Mtb encounters in humans. Whether it
cating persistence. For genesthat Mtb requires allows the development of small-molecule in-
togrowunderstandardlaboratoryconditions, hibitors that would be safe for use in humans
whichareamongthemostattractivetargetsfor remainsto beshown.
drugdevelopment,theseanalysesrequirecondi- ThesecondlimitationofTraSHandTn-seq
tionalknockdownmutants. is that they do not allow measurement of the
The construction of conditional knock- extenttowhichageneproductneedstobein-
down mutants in Mtb became possible with activatedtoimpairgrowth.Althoughtheimpact
the development of chemically controlled ge- ofatransposoninsertionongenefunctioncan
neticswitchesformycobacteria(Schnappinger vary with its insertion site, the phenotypes re-
and Ehrt 2014). Many of the switches devel- portedbyTraSH/Tn-seqareoftenattributable
oped for mycobacteria depend on placing the toinsertionsthatcreateanullallele.Complete
gene of interest downstream from a modified inactivationofaproteininlivebacteriaisrarely,
promotersothatitstranscriptioncanberegu- ifever,achievedbysmallmolecules.Thepheno-
lated by a tetracycline repressor and a tetracy- typesthat are caused by complete deletion are
g cline derivative (Blokpoel et al. 2005; Carroll thuslessrelevanttoevaluatinganewtargetfor
r
o et al. 2005; Ehrt et al. 2005; Klotzsche et al. drugdevelopmentthanthosethatresultfromits
e.
n 2009; Boldrin et al. 2010). These regulatory partialdepletion.Allotherfactorsbeingequal,a
ci
di systemsfacilitatedtheanalysesofseveralgenes protein that needsto be inactivated by .90%
e
m inMtb(SchnappingerandEhrt2014),butthey (orless)isamoreattractivetargetthanonethat
n
si dependonsecondaryevents,mostnotablycell needsto be inhibited by 90% before growth is
e
v division,todepletetheproteinofinterest.They inhibited.Thatvulnerabilityofmycobacteriato
cti are therefore less suited to study the impor- partialdepletionindeedvariesamongdifferent
e
p
s tanceofgenefunctionsforsurvivalofnonrep- essentialproteinswasfirstshowninM.smegma-
r
e
p licating Mtb.Analysesofnonreplicating bacte- tis.Forexample, HIV-protease-induced degra-
w.
w ria can be performed more efficiently with dation depleted M. smegmatis of .97% of
w regulatorysystemsthatdirectlyincreaseprotein DHFR and Alr, but this depletion onlyslowed
turnover.InMtb,thiscanbeachievedbymod- growth down. In contrast, even modest deple-
ifying a gene’s 30 end so that it encodes the tionofRpoBwassufficienttostopgrowthen-
sequence of the DASþ4 degradation tag (Kim tirely(Weietal.2011).Unfortunately,theextent
etal.2010).Conditionalityisachievedbycon- of functional inactivation that is necessary to
trollingtheexpressionofaprotein(SspB)that impair the growth of Mtb is known only fora
activates the degradation tag. Recently, tran- few proteins, including BioA and PptT, which
scriptionalandproteolyticmodesofregulation arerequiredforthesynthesisofbiotinandpoly-
have been combined in a dual-control (DUC) ketides,respectively.Mtbisnotparticularlyvul-
switch, which can achieve efficient and robust nerabletopartialinactivationofeitherenzyme,
gene inactivation in growing and nonreplicat- asbothrequiretobedepletedby.90%toaffect
ing Mtb (Kim et al. 2013). Application of the growth (Woong Parketal.2011; Leblancet al.
DUCswitchtoanalysesoftheNADsynthetase 2012). Applying conditional gene inactivation
(NadE) identified this enzyme to be essential systems to the characterization of more genes
for survival of both replicating and nonrepli- willhopefullyidentifyproteinsthatneedtobe
catingbacteria.Itsinactivationwasfurthersuf- depleted.90%tostopMtbfromgrowing.
AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139 9
Downloaded from http://perspectivesinmedicine.cshlp.org/ on January 22, 2023 - Published by Cold Spring Harbor Laboratory Press
D.Schnappinger
Target-DirectedWCS 2000 million (Adams and Brantner 2006; Paul
et al. 2010). Late-stage attrition can therefore
ClassicalWCSoftenrequirescomplexstudiesto
draintheresourcesofcommerciallyunattractive
define the MOA of hits; biochemical screens
areas,likeantibacterialdrugdevelopment,and
rarelyidentifyinhibitorsthatcanpenetratethe
causethemtofailentirelyoveralongperiodof
bacterial cell envelope. Both difficulties can be
time.Thetwomaincausesoflate-stageattrition
overcomewithtarget-directedWCS.Theprin-
arelackofsafetyandefficacyinpeople(Kolaand
ciple of this approach was first established in
Landis2004;Bennani2012).Predictingclinical
Staphylococcusaureus.OverexpressionofafabF
efficacy will remain imprecise. However, if we
antisense RNA caused the 50 end of the fabF
understand the MOA of the drug candidates
mRNA to be degraded. This decreased the ex-
that are being tested in the clinic, it should be
pression of FabF, an enzyme required for fatty
possibletoavoidtrialswithchemicallydifferent
acid biosynthesis, and increased the sensitivity
compounds that fail for the same mechanistic
of S. aureus to FabF inhibitors (Young et al.
reasons.Thegeneticapproachesdiscussedhere
2006). Agardiffusion assays werethen used to
havehelpedtocharacterizetheMOAofseveral
screensmall-moleculelibrariesagainstwild-type
antibacterial compounds. However, the MOA
S. aureus and the FabF underexpressor. Mole-
of manyantibacterials is complex and new in-
cules that were more active against the under-
sightsarestillgainedintotheMOAofdrugsthat
expressorincludedplatencinandplatensimycin,
have been in clinical use for decades (Chakra-
two novel inhibitors of fatty acid biosynthesis
bortyetal.2013).Nosingleapproachwillthere-
with broad-spectrum activity against Gram-
fore be sufficient to define the MOA of most
positivebacteria(Wangetal.2006,2007;Jaya-
antibacterials.Studiesthatcombineseveralap-
suriyaetal.2007;FischbachandWalsh2009).In
g proaches—forexample,underexpression,over-
or Mtb,transcriptionalrepressioninsteadofanti- expression,resistanceselection,andbiochemi-
e. sense RNA has been used to generate mutants
n calstudies(Huberetal.2009)—promisetobe
ci expressinglowerthanwild-typelevelsofpanto-
di themostsuccessful.Drugcandidateswithnovel
e thenatesynthase (PanC),diaminopimelatede-
m MOAsareattractivebecausetheycouldimprove
n carboxylase(LysA),isocitratelyase(Icl1),orthe
si TB chemotherapy fundamentally (e.g., by re-
e typeIsignalpeptidase(LepB).Inaddition,these
v ducing its duration). But they bear a higher
cti recently constructed mutants show target-spe- riskoflate-stagefailureiftheyinactivatetargets
e
p cificchangesin theirsusceptibility todifferent
s thathavenotyetbeenclinicallyvalidated.Mea-
er small-molecule inhibitors (Abrahams et al.
p suringtheeffectsthatcanbeachievedbyinacti-
w. 2012a; Ollinger et al. 2012). WCS with these
w vatingsuchnoveltargets(e.g.,bygeneticmeans)
w strainspromisestoidentifynewinhibitorsthat inavarietyofassays,includinganimalmodels,
areabletoreachthecytoplasmofMtbandin-
seemsprudent.
hibitadesiredtarget.
Second,geneticscanalsocontributetodrug
developmentbyprovidingsingle-geneunderex-
pressorsthathelpincreasethesensitivityofWCS
CONCLUDINGREMARKS
andbiashitstowardadesiredtarget.Underex-
Noneoftheapproachesdiscussedinthisreview pressorscanfurthermorebeusedinasecondary
existedduringthegoldenageofantibioticdis- assay to evaluate biochemical HTS hits. This
covery,andantibacterialdrugdiscoverycanbe distinguishes compounds that penetrate the
pursued without them. However, modern ge- bacterialcellenvelopepoorly(andarenotactive
neticscanhelpimproveaprocessthathasdeliv- against wild type but active against an under-
eredfewnovelmolecularentitiesduringthelast expressor) from those that do not penetrate at
50yr.Oneimportantcontributionwouldbeto all(andareactiveneitheragainstwildtypenor
help reduce late-stage attrition (i.e., failure in theunderexpressor).Suchcompoundscouldbe
clinical trials). The cost of introducing a new starting points for medicinal chemistryefforts
drugtothemarketvariesbetweenUS$500and toimprovecompoundefficacyagainstwild-type
10 AdvancedOnlineArticle.CitethisarticleasColdSpringHarbPerspectMeddoi:10.1101/cshperspect.a021139