unable7788kk.com

MAPkinaseandautophagypathwayscooperatetomaintainRASmutantcancercellsurvivalChih-ShiaLeea,LiamC.
Leea,1,TinaL.
Yuanb,2,SirishaChakkac,3,ChristofFellmannd,4,ScottW.
Lowed,e,f,NatashaJ.
Caplenc,FrankMcCormickb,g,5,andJiLuoa,5aLaboratoryofCancerBiologyandGenetics,CenterforCancerResearch,NationalCancerInstitute,Bethesda,MD20892;bHelenDillerFamilyComprehensiveCancerCenter,UniversityofCalifornia,SanFrancisco,CA94158;cGeneticsBranch,CenterforCancerResearch,NationalCancerInstitute,Bethesda,MD20892;dColdSpringHarborLaboratory,ColdSpringHarbor,NY11724;eHowardHughesMedicalInstitute,MemorialSloanKetteringCancerCenter,NewYork,NY10065;fDepartmentofCancerBiology&Genetics,MemorialSloanKetteringCancerCenter,NewYork,NY10065;andgCancerResearchTechnologyProgram,FrederickNationalLaboratoryforCancerResearch,LeidosBiomedicalResearch,Frederick,MD21702EditedbyRonaldA.
DePinho,UniversityofTexasMDAndersonCancerCenter,Houston,TX,andapprovedDecember17,2018(receivedforreviewOctober18,2018)OncogenicmutationsinthesmallGTPaseKRASarefrequentlyfoundinhumancancers,and,currently,therearenoeffectivetargetedtherapiesforthesetumors.
UsingacombinatorialsiRNAapproach,weanalyzedapanelofKRASmutantcolorectalandpancreaticcan-cercelllinesfortheirdependencyon28genenodesthatrepresentcanonicalRASeffectorpathwaysandselectedstressresponsepath-ways.
WefoundthatRAFnodeknockdownbestdifferentiatedKRASmutantandKRASWTcancercells,suggestingRAFkinasesarekeyoncoeffectorsforKRASaddiction.
Byanalyzingall376pairwisecom-binationofthesegenenodes,wefoundthatcotargetingtheRAF,RAC,andautophagypathwayscanimprovethecaptureofKRASdependencybetterthantargetingRAFalone.
Inparticular,codeple-tionoftheoncoeffectorkinasesBRAFandCRAF,togetherwiththeautophagyE1ligaseATG7,givesthebesttherapeuticwindowbe-tweenKRASmutantcellsandnormal,untransformedcells.
DistinctpatternsofRASeffectordependencywereobservedacrossKRASmutantcelllines,indicativeofheterogeneousutilizationofeffectorandstressresponsepathwaysinsupportingKRASaddiction.
Ourfindingsrevealedpreviouslyunappreciatedcomplexityinthesignal-ingnetworkdownstreamoftheKRASoncogeneandsuggestratio-naltargetcombinationsformoreeffectivetherapeuticintervention.
KRAS|RAF|MAPK|autophagy|siRNAInresponsetoextracellularstimuli,theRASfamilyofsmallGTPasesservesasasignalingnexustotransmitmitogenicsignalfromgrowthfactorreceptorstotheirintracellulareffectorpathways,which,inturn,regulateavarietyofcellularprocesses,includingcellproliferation,survival,motility,andgeneexpres-sion(1).
OncogenicmutationsinRASgenesarefrequentlyde-tectedinhumancancers.
AmongthethreeRASfamilymembersNRAS,HRAS,andKRAS,KRASaccountsforthemajorityofRASmutationsinsolidtumors(90%pancreatic,50%co-lorectal,and30%lungadenocarcinomas).
DirectinhibitionoftheKRASoncoproteinshasprovedchallenging,withonlytheKRASG12Cmutantbeingtractablethusfar(2).
Asanalternativestrategy,inhibitorstargetingRASeffectors,manyofwhicharedruggablekinases,havebeenamajorfocusinblockingoncogenicRASsignaling(3).
InhibitorsforRASeffectorkinases,includingRAF,MEK,PI3K,andAKT,havedemonstratedimpressivean-titumoractivitiesinpreclinicalstudies(4,5).
However,theyhavenotdeliveredsignificantefficacyagainstKRASmutantcancerseitherasmonotherapiesorincombinationsettingsinclinicaltrials(6,7).
Thismaybeattributabletoatleasttworeasons.
First,sinceRASsignalsthroughmultiplepathways,oncogeneaddictiontomutantKRAScouldbefunctionallydistributedacrossmultipleeffectors.
Thus,KRASmutantcellscouldusemultipleeffectorpathwaystomaintaintheirproliferationandsurvivaladvantage.
Consequently,inhibitingasingleRASeffectormaybeinsufficienttokillKRASmutantcells(8).
Second,someRASeffectorpath-ways,includingtheMAPkinase(MAPK)andPI3Kpathways,alsoplayanimportantrolefortheproliferationandsurvivalofnormalstemandprogenitorcellsinthebody(9,10).
Shuttingoffthesepathwaysusingpotentinhibitorsoftenintroducessignificanttoxicityinnormaltissues,whichcouldlimitthetherapeuticwindow(11–16).
ToidentifymoreeffectivestrategiesfortargetingRASeffec-tors,itisimportanttodistinguishoncogenicsignalingbymutantKRASfromthatofnormal,physiologicalsignalingbywild-type(WT)KRASprotein(1,8).
WehypothesizethatasubsetofRASeffectors,whichweterm"oncoeffectors,"couldplayamorecriticalroleinmediatingKRASoncogeneaddictionthanphysi-ologicalRASsignaling.
WereasonthatpinpointingtheseoncoeffectorsandselectivelytargetingthemcouldreducetoxicitySignificanceCurrently,thereisnoeffectivetargetedtherapyforoncogenicKRAS-drivencancer.
WesetouttoidentifyRASeffectorandstressresponsegenesthatcriticallysupportKRASaddiction,andthereforecouldserveaspotentialtargetsforKRASmutantcancer.
UsingacombinatorialsiRNAplatform,wesystematicallyinterrogatedthepatternsofoncoeffectordependencyinKRASmutantandWTcolorectalandpancreaticcancercelllinesandinnormalcelllines.
WefoundthatRAFkinasesarethemajorKRASoncoeffectorsandthatcotargetingBRAFandCRAFkinasesto-getherwiththeautophagyE1ligaseATG7couldefficientlyeliminateKRASmutantcellswhileminimizingtoxicityinnormalcells.
Ourworkthusestablishesaframeworkfortherationalselectionoftargetcombinationsforcancertreatment.
Authorcontributions:C.
-S.
L.
,L.
C.
L.
,T.
L.
Y.
,N.
J.
C.
,F.
M.
,andJ.
L.
designedresearch;C.
-S.
L.
,L.
C.
L.
,andS.
C.
performedresearch;C.
-S.
L.
,T.
L.
Y.
,C.
F.
,andS.
W.
L.
contributednewre-agents/analytictools;C.
-S.
L.
,L.
C.
L.
,andJ.
L.
analyzeddata;andC.
-S.
L.
,T.
L.
Y.
,andJ.
L.
wrotethepaper.
Conflictofintereststatement:F.
M.
isaconsultantforthefollowingcompanies:AduroBiotech;Amgen;DaiichiLtd.
;PellePharm;Pfizer,Inc.
;PMVPharma;andPortolaPharma-ceuticals.
F.
M.
isScientificDirectoroftheNationalCancerInstituteRASInitiativeatFrederickNationalLaboratoryforCancerResearch/LeidosBiomedicalResearch,Inc.
F.
M.
isacon-sultantforandcofounderofthefollowingcompanies:Avidity,BridgeBio,KGen,andQuartz.
L.
C.
L.
isacurrentemployeeofLoxoOncology.
T.
L.
Y.
isacurrentemployeeofNovartis.
ThisarticleisaPNASDirectSubmission.
PublishedunderthePNASlicense.
SeeCommentaryonpage3965.
1Presentaddress:MedicalAffairs,LoxoOncology,Stamford,CT06901.
2Presentaddress:OncologyTranslationalResearch,NovartisInstituteforBiomedicalRe-search,Cambridge,MA02139.
3Presentaddress:NationalCenterforAdvancingTranslationalSciences,NIH,Rockville,MD20850.
4Presentaddress:InstituteofDataScienceandBiotechnology,GladstoneInstitutes,SanFrancisco,CA94158.
5Towhomcorrespondencemaybeaddressed.
Email:frank.
mccormick@ucsf.
eduorji.
luo@nih.
gov.
Thisarticlecontainssupportinginformationonlineatwww.
pnas.
org/lookup/suppl/doi:10.
1073/pnas.
1817494116/-/DCSupplemental.
PublishedonlineFebruary1,2019.
4508–4517|PNAS|March5,2019|vol.
116|no.
10www.
pnas.
org/cgi/doi/10.
1073/pnas.
1817494116DownloadedbyguestonFebruary23,2021innormalcells.
Inadditiontooncogeneaddiction,cancercellsdrivenbyKRASandotheroncogenesexperienceextensiveoncogenicstress,aphenomenonwepreviouslyconceptualizedasnononcogeneaddiction(17).
WehypothesizethatinhibitingcellularstressresponsepathwaysthatarecriticalforthesurvivalofKRASmutantcellscouldalsoserveasaneffectivetherapeuticstrategy.
Furthermore,itstandstoreasonthatcotargetingRASeffectorpathwaysandstressresponsepathwaysmayleadtogreaterlossofsurvivalsignaling,andthusenhancethekillingofKRASmutantcells(18,19).
Previously,weandothershavecarriedoutextensivegenome-wideshRNAandCRISPRlibraryscreenstoidentifyfunctionalvulnerabilitiesinKRASmutantcells(20–26).
Collectively,theseworksrevealedtwosomewhatunexpectedfindings.
ThefirstwasthatnouniversalsyntheticlethalpartnersofKRAShavebeenidentified.
Thisindicatesthatthepatternofnononcogenead-dictioninKRASmutantcellsishighlydependentoncontext,anditislikelythatnosinglestressresponsepathwayisresponsibleforalleviatingoncogenicstressinallKRASmutantcells(3,24).
ThesecondobservationwasthatfewcanonicalRASeffectorswererecoveredasconsistentdependenciesinKRASmutantcells(24,25).
Thisisattributabletoseveralreasons.
First,aswepre-viouslynoted,theutilizationofdifferentRASeffectorpathwaystosupportKRASaddictioncouldbeheterogeneousacrossdifferentKRASmutantcelllines(27),resultinginnoconsistentoncoeffectordependency.
Second,andasmentionedabove,KRASaddictionmightbefunctionallydistributedacrossmultiplepathways.
Thus,nosingleeffectorgenecoulddominatetheKRASdependencyphe-notype.
Thethirdreasoncouldbeduetogeneparalogredundancy.
Inthemammaliangenome,manygenenodesintheRASpathwayconsistofmultiplegeneparalogsthatoftenhavesomedegreeoffunctionalredundancy.
Forexample,weandothershaveshownthatthethreeRAFfamilygenesthatconstitutetheRAFkinasenode,ARAF,BRAF,andRAF1(CRAF),couldpartiallycompensateforeachother'slosstosustainMAPKsignalinginhumancells(28–30).
BecausecurrentRNAiandCRISPRlibraryscreensareprimarilysingle-gene–dependencyassays,theywouldfailtouncoverthefunctionofgenefamilieswithredundantparalogs.
OurgoalistounderstandhowKRASaddictionisfunctionallydistributedacrossitsvariousdownstreameffectorpathwaysandacrossvariousstressresponsepathways.
Todoso,weneedtocotargetmultiplegenessimultaneouslyinthesamecelltogainasystems-levelviewofhowgeneparalogredundancyandpathwaycooperativitycontributetoKRASaddiction.
Becauseconven-tionalRNAiandCRISPRscreensaremostlylimitedtointer-rogatingsingle-geneactivity,weovercamethislimitationbydevelopingacombinatorialsiRNAplatformtosimultaneouslycotargetuptosevendifferentgenesinthesamecell(29).
Usingthisapproach,werecentlyshowedthat,atthesingleeffectornodelevel,differentKRASmutantcelllinesexhibitdifferentdegreesofeffectornodedependency,andthesubsetofKRASmutantlinesthatarelessdependentonKRASexhibitenhanceddependencyonthep90RSKkinasenode(27).
Inthiscurrentstudy,weextendedourcombinatorialsiRNAanalysistointerrogatehigherordergenecombinationsthatcotargetgenenodepairsandgenecombinationsfromtwoormorepathways.
StudiesofthistypehavenotbeenpreviouslyattemptedinKRASmutantcells.
Sincecombinationtherapyinvolvingdrugswithorthogonalmechanismsofactionhasbe-comeincreasinglynecessarytoachievemeaningfulbenefitinpatients,ourstudyshoulddirectlyinformtherationaldesignoftargetcombination.
WediscoveredthattheRAFkinasenode,particularlytheBRAFandCRAFparalogs,arekeyoncoeffec-torsofKRASaddiction.
WefurtherdemonstratedthatwhereastheRACandautophagypathwayseachcontributelittletoKRASdependencyalone,cotargetingthesepathwaystogetherwiththeRAFnodecouldsignificantlyenhancethekillingofKRASmutantcells.
ExtensivegeneparalogdeconvolutionidentifiedBRAF,CRAF,andATG7asatargetcombinationthatprovidesthebestdiscriminationbetweenKRASmutantcellsandnor-mal,untransformedcells.
OurstudythereforeprovidesasystemsbiologyapproachtorationallyevaluatetargetcombinationinKRASmutantcells.
ResultsCurationofaSensorsiRNALibraryforRASEffectorandStressResponseGenes.
WewishedtoexaminehowKRASaddictionispartitionedthroughcanonicalRASeffectors,includingtheMAPK,PI3K,RHO,RAC,RAL,andPLCepathways.
Similartoourrecentreport(27),weselected19genenodesinthesesixpathwaysthatconsistof47genesandparalogs.
Tounderstandhownon-oncogeneaddictioncontributestoKRASaddiction,weincludedinouranalysis10genenodesconsistingof26genesandparalogsthatcorrespondtoindirectRASeffectorsandstressresponsepathwaysthatcouldcontributetonononcogeneaddictioninKRASmutantcells(SIAppendix,Fig.
S1A).
Thus,ouranalysisincludedatotalof73genesrepresenting29discretegenenodes.
ToidentifyKRASoncoeffectorsandstressresponsegenesthatarecriticalinmedi-atingKRASaddiction,wesoughttoaccessthedependencypro-filesofKRASmutantcellsforvariouscombinationsofthesegenesusingsiRNA-mediatedmultigeneknockdown.
Previously,weestablishedasensorsiRNAstechnologyplatformtocuratehighlypotentsiRNAsthatcanworkreliablyincombinationtoachieveefficientknockdownofuptosevengenetargetssimultaneouslywithminimalcross-interferenceoroff-targeteffects(29).
Be-causemanygenenodesconsistoftwotofourparalogs,ourplat-formthusenablesustointerrogatethecell'sdependencyonmultiplegenenodes.
ToconstructalibraryofsensorsiRNAstar-getingthese73genes,wescreenedalargenumberofcandidatesensorshRNAsforeachgeneandidentifiedhighlypotentsensorsiRNAsequencesthatcouldmaintain>70%targetmRNAknockdownatalowconcentrationinapoolwiththreetosixotherhighlypotentsiRNAs(27,29,31).
ThisensuresthatthesesiRNAscouldreproduciblyknockdowntheirtargetsinacombinatorialsetting.
ToreducepotentialsiRNAoff-targeteffects,wecuratedtwoindependentsensorsiRNAswithsimilarknockdowneffi-ciencyforallbutsixgenesinourlibrary(SIAppendix,Fig.
S1B).
UsingthesetwosetsofsiRNAs,weconstructedtwoparallelli-brariescontainingthesamegenecombinations(set1andset2li-braries,respectively;SIAppendix,Fig.
S1E).
Overall,>80%ofthesiRNAsgave>80%targetmRNAknockdown(SIAppendix,Fig.
S1C),andthereiscomparableknockdownefficiencybetweenset1andset2siRNAs(SIAppendix,Fig.
S1D).
Thus,phenotypicconcordancebetweencorrespondingsiRNApoolsfromthetwolibrarieswouldsuggestansiRNAon-targeteffect,whereasphe-notypicdiscordancewouldsuggestansiRNAoff-targeteffect.
TakingadvantageofthehigherordersiRNAcombinationplatformwehavedeveloped,weconstructedacombinatorialsiRNAlibrarytosystematicallyinterrogatetheseRASeffectorandstressresponsegenesfortheircooperativeroleinsupportingKRASaddiction.
ThesiRNApoolsinthelibraryweredesignedtointerrogategeneinteractionsattwolevelsofcomplexity.
Toovercomegeneparalogredundancy,weconstructed29siNodepoolstargetingeachofthe29genenodesincludedinthisstudy.
EachsiNodepoolconsistsofsiRNAstargetingallgeneparalogsforthenodes.
Toexaminegenenodeandpathwayinteractions,weconstructed406siNodePairpoolsthatrepresentallpairwisecombinationsofthe29genenodes.
EachsiNodePairpoolcon-sistsofsiRNAstargetingallgeneparalogsinthetwogenenodesbeingtargeted(SIAppendix,Fig.
S1EandDatasetS1).
Single-NodeDependencyAnalysisIdentifiesRAFKinasesasaKeyOncoeffectorNode.
WefirstinvestigatedhoweachsiNodepoolimpactedcellviabilityinapanelofcolorectalcancer(CRC)celllinesconsistingoffiveKRASmutant(DLD-1,HCT116,SW620,LoVo,andSW403)andtwoKRASWT(SW48andCaco-2)lines.
Aswepreviouslyreported(27,29),theKRASmutantlinesweredependentonKRASandsensitivetoKRASknockdownbytwoindependent,validatedKRASsiRNAs(SIAppendix,Fig.
S2A).
Incontrastandasexpected,theKRASWTlineswerenotsen-sitivetoKRASknockdown.
Theset1andset2siNodelibrarieswereindependentlytestedinthesecelllines(SIAppendix,Fig.
Leeetal.
PNAS|March5,2019|vol.
116|no.
10|4509MEDICALSCIENCESSEECOMMENTARYDownloadedbyguestonFebruary23,2021S2A).
ForagivensiNodepool,thetwolibrariesgaveconcordantresults(SIAppendix,Fig.
S2B),althoughwenoticedthatset1siRNAstendedtohaveaslightlystrongereffectingeneral.
OutlieranalysiscomparingthephenotypicdistancebetweenthetwolibrariesidentifiedthesiNodepoolforhexokinasestolikelycontainsiRNAoff-targeteffects(SIAppendix,Fig.
S2C).
Wethereforeex-cludedthehexokinasesiRNAsfromallsubsequentanalyses.
ToidentifyKRASoncoeffectornodeswhoseknockdownmostcloselyphenocopysiKRAS,weperformedunsupervisedhierar-chicalclusteringusingaveragedcellviabilitydatafromset1andset2siRNAs(Fig.
1A).
Severalpatternsinthedatabecamesalient.
First,thesiBCL2pool,whichtargetsmultipleprosurvivalBCL-2familymembers,clusteredwiththesiDeath-positivecontrol.
ThisindicatesthatshuttingdownBCL-2familyfunc-tioncouldbegenerallytoxicirrespectiveofthecell'sgenotype.
Secondandsomewhatsurprisingly,manysiNodepoolsshowedrelativelylittletoxicityinthesecelllines.
ThiscouldbedueeithertoalackofdependencyonthesenodesortothefactthatthesesiRNAs,despiteourbesteffort,wereunabletoknockdowntargetgenesatasufficientlydeepleveltorevealtheirdependency.
Third,onlythesiRAFnodecloselyclusteredwithKRASsiRNAsandshowedasimilarsensitivityprofileamongKRASmutantcells.
ThisindicatesRAFisthemostcriticaloncoeffectornodethatmediatesoncogenicKRASsignaling.
ToquantitativelymeasurehowmuchofKRASdependency(asindicatedbytheeffectofsiKRAS)iscapturedbyeachsiNodeandtoreducedatacomplexity,weestablishedtwosimplemetricsfromaggregateddataanalysis.
Thedifferentialdependencyscore(DDS;range:100to100%)foraKRASmutantlinerepresentstheviabilitydifferencebetweenthemeanviabilityofallKRASWTcelllinesandthatoftheKRASmutantline.
ApositiveDDSindicatesthatansiRNApoolhasgreatertoxicityintheKRASmutantlinecomparedwithKRASWTcells.
Conversely,anegativeDDSindicatesgreatertoxicityinKRASWTcells.
ADDSof0%meansthesiRNApoolhasnodifferentialtoxicitybetweenKRASWTandmutantcells.
Thecorrelationr(range:1to1)isthePearsoncorrelationbetweenallKRASmutantcelllines'sensitivitytoansiNodeandthattosiKRAS.
Thus,anoncoeffectorisexpectedtoscorehighforbothDDSandr.
PlottingDDSvs.
rforthe28siNodepoolsrevealedthatthemajorityofthemwererelativelydistantfromsiKRAS,whichresidesintheupperrightquadrant(Fig.
1B).
Thisanalysissug-gestedRAF,RAL,RALGEF,andNF-κBnodesasoncoeffectornodeswithapositiveDDSandr(Fig.
1BandCandSIAppendix,Fig.
S3AandB).
Amongthesefour,theRAFnodehadsub-stantiallystrongermetricsforboththeDDS(Fig.
1C)andr(SIAppendix,Fig.
S3A).
However,theRAFnodewasonlyabletocapturelessthan50%ofKRASdependencyinmutantcells(Fig.
1C).
Theglutaminasenode(GLS)alsoscoredapositiveDDS.
Acloserexaminationofitsrscore,however,showedpoorcorrelationwithsiKRASbecausethephenotypewasprimarilydrivenbystrongtoxicityinonecellline,HCT116(SIAppendix,Fig.
S3C).
Thus,ouranalysisindicatesthattheRAFnodeisamajoroncoeffectordownstreamofmutantKRAS,althoughnosinglenodeisabletofullycaptureKRASdependency.
SomewhatunexpectedlyandunliketheRAFnode,theMEKandERKnodesdidnotscorehighlyinourinitialanalysis.
BothDDSandrmetricsfortheMEKandERKnodeswerelowerthanthoseoftheRAFnode(SIAppendix,Fig.
S3A,D,andE).
OnepossibilitywouldbethattheMEKandERKsiNodepoolswerelesspotentthantheRAFsiNodepool.
WecomparedtheknockdownefficiencyofthesesiNodesfortheircognatetargetsandfortheirabilitytoreducethelevelofphospho-ERKincells.
Asexpected,allsiNodepoolseffectivelyandconsistentlyknockeddowntheirtargetproteinsinmultipleKRASmutantandWTcelllines(SIAppendix,Fig.
S4A).
Furthermore,allRAF,MEK,andERKsiNodepoolshadacomparableinhibitoryeffectonphospho-ERKlevels(SIAppendix,Fig.
S4B).
TheseresultssuggestthatFig.
1.
Single-nodedependencyinKRASmutantcells.
(A)Cellviabilitydataofsingle-nodeknockdowninCRCcelllines.
Cellviabilitywasmeasured5dpost-siRNAtransfectionandwasnormalizedtothesiNegcontrolfortherespectivecellline.
Averageviabilitydataforset1andset2siNodelibrarieswereclusteredbasedonEuclideandistance.
DendrogrambranchesinvolvingsiKRASarehighlightedinred.
ForeachsiNode,theaveragedDDSandrmetricswithsiKRASareshownnexttotheviabilityheatmap.
(B)ScatterplotofaveragedDDSvs.
rmetricstovisualizethedistancebetweenvarioussiNodesandthesiKRAS-positivecontrol.
(C)Topfivepublicsingle-nodedependenciesbasedontheaverageDDS.
BarsrepresentaverageDDSmetricsacrossfiveKRASmutantcelllines(errorbarsrepresentSEMinallfiguresunlessotherwisestated).
(D)Topprivatesingle-nodedependencyforeachcellline.
Single-nodedependencywasrankedbasedonthecellline-specificDDS.
siKRASwasincludedasapositivecontrol.
4510|www.
pnas.
org/cgi/doi/10.
1073/pnas.
1817494116Leeetal.
DownloadedbyguestonFebruary23,2021theeffectofRAF,MEK,andERKsiNodepoolscannotbeat-tributedtodifferencesintheirabilitytodown-regulateMAPKpathwayactivity.
ToexcludethepossibilitythatthetoxicityofsiRNAsinthepoolswasduetooff-targeteffects,weperformedrescueexperimentsusingsequence-specificC911controlsiRNAsthatcorrespondedtoset1siRNAsinthesepools(32).
AsshowninSIAppendix,Fig.
S4C,replacingallon-targetsiRNAswiththeirC911counterpartsrescuedcellviabilityinsensitivecelllines.
Thus,thehighertoxicityofRAFsiRNAsinKRASmutantcellsislikelyaresultoftheiron-targeteffects.
WenoticedthatoneoftheKRASWTcelllines,SW48,issensitivetoMEKknockdown(Fig.
1A).
SW48cellsharboraMEK1Q56P-activatingmutation(33–35),andthiscouldconfoundouranalysis.
WethusrepeatedtheclusteringanalysisandthecalculationofDDSandrmetricsbyleavingouttheSW48dataset(SIAppendix,Fig.
S4D).
ThisimprovedthescoringfortheMEKnodebutnotfortheERKnode.
However,theMEKnodestillhadasubstantiallylowerDDSthantheRAFnodeintherankingoftoponcoef-fectors(SIAppendix,Fig.
S4E).
Thus,ouranalysisindicatesthatKRASmutantcellsaremoredependentontheRAFnodethantheMEKandERKnodes,andthatRAFcouldpresentabettertargetthanMEKandERKwithintheMAPKpathway.
Similartoourpreviousreport(27),weobservedawiderangeinsensitivitytoKRASknockdownamongtheKRASmutantCRCcelllines(Fig.
1D).
WhenDDSmetricswerecalculatedandrankedforindividualKRASmutantcelllines,wesawahetero-geneouspatternofsingle-nodedependency(Fig.
1D).
Forex-ample,SW403cellsarehighlysensitivetotheknockdownoftheRAFnode,whereasHCT116cellsareuniquelydependentonglutaminase.
InDLD-1,SW620,andLoVocells,nodominantnodedependencieswereobserved.
ThisfindingsuggeststhatdifferentKRASmutantcancercelllinescouldutilizedifferentpathwaystosupportKRASaddiction.
Paired-NodeDependencyAnalysisRevealsRAFNodeCombinationsBestCaptureKRASDependency.
Next,weassessedthesensitivityofKRASmutantandWTCRCcelllinestoall378siNodePairpools(excludingthehexokinasenode)toinvestigatetheimpactofknockingdowntwonodestogether.
Cellviabilityresponsestoset1andset2siNodePairlibrariesweremostlyconcordant(SIAppendix,Fig.
S5AandB),suggestingthathigherordersiRNAcombinationsdidnotintroducesubstantiallymoreoff-targeteffects.
Outlieranalysisidentifiedtwonode-paircombinations,RHO+ROCKandRHO+TIAM,toshowsignificantdisagreementbetweenset1andset2siRNAs(SIAppendix,Fig.
S5C).
Theywereexcludedfromsubsequentanalysis.
WeappliedtheaforementionedanalysispipelinetoidentifynodepairsthatbestphenocopiedtheeffectofsiKRAS.
Un-supervisedhierarchicalclusteringofviabilitydataidentifiedsev-eraldistinctclusters(Fig.
2A).
Notably,thesubsetofsiNodePaircombinationsthatclusteredclosesttosiKRASconsistedexclu-sivelyofRAFnodecombinations(Fig.
2B).
DDSandrmetricsconfirmedthetop-rankednodecombinationswereallRAFnodecombinations(Fig.
2C).
SeveralRAFnodecombinationshadahigherDDSthantheRAFnodealone;theseincludeRAFincombinationwiththeRAC,RAL,ROCK,andATGnodes(Fig.
2D).
Node-paircombinationsinvolvingRAC,RAL,ROCKandATG,butnotRAF,didnotscore.
ThisfindingsupportsthenotionthattheRAFnodeisthemostcriticaloncoeffectorinKRASmu-tantcellsandsuggeststhatcotargetingRAFwithasecondeffectornodecouldgeneratebettertherapeuticefficacy.
InadditiontotheRAFnodecluster,theanalysisrevealedotherclustersinwhichthenodecombinationssharesimilarFig.
2.
Paired-nodedependencyinKRASmutantcells.
(A)Cellviabilitydataofnode-pairknockdowninCRCcelllines.
Cellviabilitywasmeasured5dpost-siRNAtransfectionandwasnormalizedtothesiNegcontrolfortherespectivecellline.
Averageviabilitydataforset1andset2siNodePairlibraries,aswellastheviabilitydatafromthesingle-nodeanalysis,werecombinedandclusteredbasedonEuclideandistance.
ForeachsiNodePair,theaveragedDDSandrmetricsareshownnexttotheviabilityheatmap.
NotableclustersofnodepairsconsistingoftheRAF,PI3K,MEK,andERKnodes,respectively,arehigh-lighted.
(B)MoredetailedviewoftheRAFclusterinAshowingtheinclusionofsiKRASandsiRAFinthiscluster,togetherwithvariousnodepairsinvolvingRAF.
(C)ScatterplotoftheaveragedDDSvs.
rtovisualizethedistancebetweenvarioussiNodePairsandthesiKRAS-positivecontrol.
TheRAFsiNodeishighlightedinpurple,andsiNodePairsinvolvingRAFarehighlightedinred.
(D)Topfivepublicsingle-nodedependenciesbasedonaverageDDSmetrics.
BarsrepresentaverageDDSmetricsacrossfiveKRASmutantcelllines.
Leeetal.
PNAS|March5,2019|vol.
116|no.
10|4511MEDICALSCIENCESSEECOMMENTARYDownloadedbyguestonFebruary23,2021dependencyprofiles.
ClusteringclosesttothesiDeath-positivecon-trolweretheBCL-2nodecombinations(SIAppendix,Fig.
S6A).
Thesecombinationsarelikelytobegenerallytoxicregardlessofthecell'sKRASgenotype.
TwoRAFnodecombinations,RAF+PI3KandRAF+MEK,appearedinthisDeathclusterbutnotinthemainRAFnodecluster.
Thissuggeststhatcotargetingtwoessentialpathways(RAF+PI3K)orthedeepinhibitionofoneessentialpathway(RAF+MEK)couldreducethecombina-tion'sselectivity.
ThisisinagreementwiththeincreasedtoxicityofMAPKandPI3Kinhibitorcombinationsseeninclinicalstudies(12–16).
WenotedadiscretePI3KnodeclusterthatshowednocorrelationwithKRASstatus(SIAppendix,Fig.
S6B).
Interest-ingly,thecelllinessensitivetoPI3KnodecombinationscorrelatedwiththeirPI3Kmutationstatus:ThefivesensitivecelllinesharbormutationsinthePIK3CAorPIK3CBgenes(DLD-1,PIK3CAE545K;HCT116,PIK3CAH1047R;LoVo,PIK3CBE1051K;SW403,PIK3-CAQ546K;andSW48,PIK3CAG914R),whereasthetwoinsensitivecelllines,SW620andCaco-2,areWTcelllinesforthesegenes.
Thus,distinctpatternsofdependenciescanbedeterminedbyco-occurringdrivermutations.
OuranalysisshowedthatMEKandERKnodecombinationsdidnotclusterclosetosiKRAS(Fig.
2AandSIAppendix,Fig.
S6CandD).
ToruleouttheconfoundingeffectoftheMEK1mutantSW48cellline,weremovedtheSW48datasetandrepeatedtheDDSandrcalculations(SIAppendix,Fig.
S7A).
TheresultsconfirmedthatRAFnodecombinationsweresuperioratcap-turingKRASdependencycomparedwithMEKandERKnodecombinations.
Inthislatteranalysis,thetop-rankedRAFnodecombinationscapturedabout70–80%ofKRASdependency,whereasthetopMEKandERKnodecombinationsonlycaptured40–50%ofKRASdependency(SIAppendix,Fig.
S7B).
WhenwecalculatedandrankedthesiNodePairDDSforeachindividualKRASmutantcellline,wemadetwonotableobser-vations.
First,similartosingle-nodedependency,node-pairde-pendencyexhibitedsignificantheterogeneityacrossdifferentcelllines.
Withinagivencellline,thetop-scoringnode-paircombi-nationswerepredominantlydrivenbythetop-scoringsinglenode(SIAppendix,Fig.
S8).
Second,wefoundthatinsomecelllines,includingDLD-1andHCT116,theirrespectivetop-scoringnode-paircombinationswereabletocapturenearlyallKRASdependency,albeitsuchcombinationswerecellline-specific(SIAppendix,Fig.
S8).
Thisanalysisfurtherhighlightedthehetero-geneousnatureofeffectorpathwaydependencyacrossdifferentKRASmutantcelllines.
Thus,analogoustothe"public"and"private"sensitivityofcancercelllinestosingle-agentsmallmoleculeinhibitors(36),thepublictargetcombinationsthatworkforthemajorityofKRASmutantcelllinesmaynotbethebestprivatecombinationforanygivencellline.
IdentificationofBRAF,CRAF,andATG7astheMinimalOncoeffectorCombinationThatBestDiscriminatesKRASMutantCancerCellsandNormalCells.
Weselectedfourofthetop-scoringRAFnodecombinationsforfurthervalidationusingset1siRNAs.
TheseincludeRAFincombinationwiththeRAC,RAL,ROCK,andATGnodes(Fig.
2D).
WhereasthefirstthreenodesrepresentinterpathwaycombinationsamongRASeffectors,thelastonerepresentscombinationwithastressresponsepathwaythatisknowntogeneticallyinteractwiththeKRASoncogene(37,38).
Weexpandedouranalysistoincludeseveralpancreaticductaladenocarcinoma(PDAC)celllinesconsistingoffiveKRASmutantlines(MIAPaCa-2,Hup-T4,SUIT-2,AsPC-1,andPA-TU-8902)andoneKRASWTline(BxPC-3).
Inthese13CRCandPDACcelllines,combinedknockdownofRAFwiththeRAC,RAL,andATGnodesallledtogreatertoxicityinKRASmutantcellscomparedwithRAFnodeknockdownalone(SIAppendix,Fig.
S9A).
DDSandrmetricsforthesecombina-tionsshowedthattheRAF+ATGnodecombinationhadahigherDDScomparedwiththeRAFnodealone(Fig.
3A),whileallthreenodecombinationsretainedgoodrscorescomparedwiththeRAFnodealone(SIAppendix,Fig.
S9BandC).
Tofurthervalidatetheon-targeteffectofthesesiRNAcombinations,weperformedrescueexperimentsusingsequence-specificC911controlsiRNAsforeachsiRNAinthepool.
Asexpected,replacingallon-targetsiRNAswiththeirC911counterpartsabolishedthetoxicityofthesiRNApools(SIAppendix,Fig.
S9D).
Thiscon-firmedthattheeffectsofthesiRNAcombinationswerelikelyon-targeteffects.
Takentogether,theseresultsindicatethatKRASaddictionismediatedthroughmultipleeffectorandstressresponsepathways,andthatthegenenodecombinationsweidentifiedcouldreflectcommonpathwaydependenciesamongKRASmutantcells.
PhysiologicalRASsignalingiscriticalfortheproliferationofnormaltissues.
Inmice,thedoubleknockoutofMek1andMek2orthatofErk1andErk2leadstolethality(9,39).
AlthoughAraf/Braf/Craftriple-knockoutmicehavenotbeencharacterized,itislikelythatthecompleteabsenceofRAFsignalinginnormaltissuecouldnotbetolerated.
However,doubleknockoutofBrafandCrafinmiceiswelltolerated(39),suggestingnormaltissuecansolelyrelyonARAFtosatisfyphysiologicalRAFsignaling.
Similarly,doubleknockoutofRalAandRalBleadstolethalityinmice,whereasRalBknockoutistoleratedwell(40).
ToexaminethepotentialtoxicityofRAF,RAC,RAL,andATGnodeknockdowninnormalcells,wetestedtheimpactoftheirsiNodeandsiNodePairpoolsontheviabilityofseveralimmortalizedbutuntransformednormalcelllines.
TheseincludehumanFig.
3.
Validationoftopnode-pairdependencies.
(A)KRASmutantandWTCRCandPDACcelllinesweretransfectedwithsiKRAS,ansiRAFnode,andsiNodePaircombinationsasindicated.
Cellviabilitywasdetermined5dposttransfectionandnormalizedtosiNegcontrol,andtheDDSforeachKRASmutantcancercelllinewascalculated.
BarsrepresenttheaverageDDSmetricsacross10KRASmutantcelllines,anddotsrepresentdatapointsforindividualcelllines.
*P<0.
05;**P<0.
01.
n.
s.
,notsignificant.
(B)Immor-talizedHMECs,iSAECs,humanpancreasductnormalepithelialcells(HPNE),andhumanBJfibroblasts(BJ)weretransfectedwithsiKRASorsiRNAsagainstsingleRASeffectornodes(siNode)ornodepairs(siNodePair)asin-dicated.
Cellviabilitywasdetermined4dposttransfection.
4512|www.
pnas.
org/cgi/doi/10.
1073/pnas.
1817494116Leeetal.
DownloadedbyguestonFebruary23,2021mammaryepithelialcells(HMECs),immortalizedhumansmallairwayepithelialcells(iSAECs)(41),humanpancreaticductnormalepithelialcells,andthehumanfibroblastcelllineBJ.
RAFnodeknockdownintroducedsubstantialtoxicityinHMECs,whereasiSAECsandBJcellsaresensitivetoRACnodeknockdown.
RALnodeknockdownwasmoderatelytoxicinHMECsandBJcells.
Incontrast,knockingdowntheATGnodewaswelltoleratedinallofthenormalcelllines.
Knockingdownnode-paircombinationsledtoadditivetoxicityinthesenormalcelllines(Fig.
3B).
Theseresults,togetherwithpreviousmousege-neticstudies,suggestthatcotargetingallgeneparalogswithinaRASeffectornodewouldnotbewelltoleratedinnormalcells.
Tominimizetoxicityinnormalcells,wehypothesizedthatKRASoncoeffectorscanbefurtherdistinguishedatthegeneparaloglevel.
WereasonedthatbytargetingoneortwocriticaloncoeffectorparalogswithineachRASeffectornodewhilesparingotherparalogs,wemaybeabletoeffectivelyattenuateoncogenicKRASsignalwhilepreservingaminimallevelofphysiologicalRASsignalingthatisessentialfornormalcellviability.
Insupportofthisnotion,ithasbeenshownthatCrafiscriticalforKRAS-driventumorsbutdispensableinnormalmousetissues(39,42,43).
Wethereforesoughttoidentifytheminimalgeneparalogcombina-tionsthatwouldgivetheleasttoxicityinnormalcelllineswhileretainingtoxicityinKRASmutantcells.
WefirsttestedsiRNAstargetingindividualgeneparalogswithintheRAF,RAC,RAL,andATGnodesinthenormalcelllines.
Indeed,toxicityassociatedwithsingle-geneknockdownwassignificantlyreducedcom-paredwithwhole-nodeknockdown,likelyduetofunctionalre-dundancyamongparalogs(SIAppendix,Fig.
S10A).
ForRAFkinases,singleRAFgeneknockdownhadlittleimpactonthesensitiveHMECs.
However,single-geneBRAFandCRAFknock-downalsohadminorimpactonKRASmutantcells(SIAppendix,Fig.
S10B).
CodepletionofBRAFandCRAF,whilesparingARAF,ledtoanintermediatesituationcomparedwithwhole-RAFnodeknockdown:ItretainedasignificantfractionoftheDDSinKRASmutantcells(SIAppendix,Fig.
S10B),yetthetoxicityinHMECswasreduced(SIAppendix,Fig.
S10A).
TheseresultsreinforcethenotionthatoncogenicKRASsignalinghijackspartofphysio-logicalRASsignaling,andthatmaximizingthetherapeuticwin-dowmightrequirethepreservationofsomepathwayactivityinnormalcells.
ToidentifyaminimalgenecombinationthatwouldgivethebestselectivitytowardKRASmutantcells,wedeconvolvedthenode-paircombinationsdowntotheirconstituentgeneparalogcombinations.
BecausepreviousstudieshavedemonstratedacriticalroleforRAC1(44)andRALB(22,45)inKRASonco-genesis,wechoseRAC1andRALBasthecandidateparalogsfromtheirrespectivenodes.
WechoseATG7fromtheauto-phagynodeasitistheonlyknownE1enzymeinthispathwayandisrequiredformutantRAS-driventumorgrowth(37,38).
WithBRAF,CRAF,RAC1,RALB,andATG7asthefivekeycandidates,wecodepletedtheminvariouscombinationsofonetofivegenesinthe13CRCandPDACcancercelllinesandinthefournormalcelllines.
ToquantifytherelativeimpactofthesesiRNApoolsinKRASmutantcellsvs.
normalcells,wecalculatedadifferentialdependencyscorevs.
normalcells(DDSn),whichusesthemeanviabilityofthenormalcelllines(insteadofthatofKRASWTcancercells)asthebaselineforevaluatinggenotypeselectivity.
ClusteringanalysisshowedthatBRAFandCRAFre-mainthemaindeterminantsforKRASdependency(Fig.
4A),andthesecombinationsalsohadthebestDDS,DDSn,andrmetrics(SIAppendix,Fig.
S11AandB).
ThescatterplotofDDSvs.
DDSnmetricsshowedastrongcorrelationbetweenthesetwometricsforallcombinations(SIAppendix,Fig.
S11C),suggestingthatnormalcelllinesandKRASWTcelllinesshareareduceddependencyonthesegenesforsurvival.
CloserinspectionoftheDDSandDDSnmetrics,however,revealednotabledifferences.
WhenKRASWTcancercellswereusedasthebaselinetomea-sureselectivity(DDSmetrics),knockingdowntheentireRAFnodeprovidedbettercaptureofKRASdependencythanBRAFand/orCRAFknockdown,andnoneoftheotherRAFparalogFig.
4.
Deconvolutionoftopgenenode-pairdependencies.
(A)KRASmu-tantandWTCRCandPDACcelllinesandimmortalizednormalhumancelllinesweretransfectedwithsiRNAstargetingvariousgeneparalogcombi-nationsoftheRAF,RAC,RAL,andATGnodes.
Cellviabilitywasmeasured4–5dpost-siRNAtransfectionandwasnormalizedtothesiNegcontrolfortherespectivecellline.
CellviabilitywasclusteredbasedonEuclideandistance.
TheaverageDDS,averageDDSn,andrmetricsofeachcombinationareshownontherightasheatmaps.
(B)Top-scoringsiRNAcombinationswererankedseparatelybasedontheirDDSandDDSnmetrics.
ThesiRNAstargetingKRASandRAFparalogsareincludedforcomparison.
BarsrepresentaverageDDSandDDSnmetricsacross10KRASmutantCRCandPDACcelllines,anddotsrepresentdatapointsforindividualcelllines.
n/a,notapplicable;n.
s.
,notsignificant.
AheatmapforcorrespondingrandPvalues(comparedwithRAFknockdown)isshownbelowthebarchart.
Genesymbolabbreviations:A,ARAF;a,RALA;B,BRAF;b,RALB;C,CRAF;K,KRAS;n,BECN1;p,RALBP1;1,RAC1;2,RAC2;5,ATG5;7,ATG7.
Leeetal.
PNAS|March5,2019|vol.
116|no.
10|4513MEDICALSCIENCESSEECOMMENTARYDownloadedbyguestonFebruary23,2021combinationsperformedbetterthanRAFnodeknockdown(Fig.
4BandSIAppendix,Fig.
S11D).
However,whennormalcelllineswereusedasthebaselinetomeasureselectivity(DDSnmetrics),knockingdowntheentireRAFnodegavenobettercaptureofKRASdependencythanBRAF+CRAFknockdownduetoin-creasedtoxicityofRAFnodeknockdowninnormalcells(Fig.
4BandSIAppendix,Fig.
S11E).
Top-scoringgenecombinationsalsodifferedbasedonthetwometrics.
Thefive-genecombinationofBRAF+CRAF+RAC1+RALB+ATG7scoredhighestwiththeDDSmetrics,butonlyninthintheDDSnmetrics(SIAppendix,Fig.
S11DandE),indicatinghighertoxicityofthiscombinationinnormalcells.
ThetoptwocombinationswiththebestDDSnmetricsinvolvethethree-genecombinationofBRAF+CRAF+ATG7andthefour-genecombinationofBRAF+CRAF+ATG7+RAC1.
ThesewerealsoamongthetopfivescoringcombinationsbasedontheDDS(Fig.
4B).
Importantly,thesetwocombinationswerebetteratcapturingKRASdependencythantargetingBRAF+CRAFonly(Fig.
4B).
CotargetingRAFandAutophagyEnhancesCellCycleArrestandCellDeathinKRASMutantCells.
Tovalidatetheon-targeteffectofthesesiRNAcombinations,weagainperformedrescueexperi-mentsusingsequence-specificC911controlsiRNAs.
Replacingallon-targetsiRNAswiththeirC911counterpartsabolishedthetoxicityofthesesiRNApools(SIAppendix,Fig.
S12).
WenextconfirmedtheimpactofBRAF,CRAF,RAC1,andATG7knock-downusingWesternblotanalysis.
Asexpected,allsiRNAsper-formedeffectivelyatdepletingtheirtargetproteinsregardlessofthecombinatorialsetting(SIAppendix,Fig.
S13).
KRASknockdownandcombinedBRAF+CRAFknockdownbothdown-regulatedMAPKpathwayactivity.
WeobservedreciprocalregulationoftwoERKsubstrates,thetranscriptionfactorFRA1andtheproapoptoticproteinBIM,suchthatlossofERKsignalingledtothede-stabilizationofFRA1andthestabilizationofBIM(SIAppendix,Fig.
S13).
ATG7knockdownsignificantlyinhibitedtheactivityoftheautophagypathway.
ThiswasreflectedbythelossofATG12-conjugatedATG5andtheaccumulationofunconjugatedATG5,thedecreaseinLC3lipidation,andthedestabilizationofp62(SIAppendix,Fig.
S13).
Inagreementwiththecellviabilitydata,correspondingC911siRNAsforeachofthesegenesdidnotreducetargetproteinexpressionoralterpathwayactivity(SIAppendix,Fig.
S13).
Together,theseresultsindicatethattheactivityofthesesiRNAcombinationswereon-targeteffects.
OurfindingsuggeststhatcotargetingBRAF,CRAF,RAC1,andATG7couldshowefficacyinKRASmutantcellswhilere-ducinggeneraltoxicityinnormalcells.
Althoughwedonotcur-rentlyhaveparalog-specificinhibitorsagainstBRAFandCRAF,weattemptedtosupportthetranslationalpotentialofourfindingsbycombiningRAC1andATG7siRNAswithtwoselectiveinhib-itorsoftheMAPKpathway:theUSFoodandDrugAdminis-tration(FDA)-approvedMEKinhibitortrametinibandarecentlydisclosedRAFinhibitor,RAF709,thathaslittleparadoxicaleffectinRASmutantcells(46,47).
DepletingATG7and/orRAC1sensitizedtheKRASmutantcelllinesHCT116andMIAPaCa-2tobothRAF709andtrametinib,withadecreaseinIC50valuesbetweenthreefoldandsevenfold(Fig.
5andSIAppendix,Fig.
S14).
Incontrast,suchaneffectwasnotobservedintheKRASWTBxPC-3cells(Fig.
5andSIAppendix,Fig.
S14).
InagreementwiththesiRNAcombinationknockdowndata(Fig.
4B),ATG7knock-downhadagreatersensitizingeffectthanRAC1knockdowninthesepharmacologicalexperiments.
MAPKpathwayinhibitionoftenleadstoG1cellcyclearrest(48,49).
WehypothesizedthatATG7andRAC1knockdowncouldsynergizewithRAFinhibitioneitherbycausingstrongercellcyclearrestorbyinducingapoptosisinKRASmutantcells.
CellcycleanalysisshowedthatKRASknockdowncausedG1arrestintheKRASmutantcelllinesHCT116andMIAPaCa-2.
Inagreementwiththeireffectoncellviability,codepletionofBRAFandCRAFdidnotleadtoasignificantcytostaticeffect.
However,codepletingBRAFandCRAFtogetherwithATG7and/orRAC1resultedinG1arrestthatwascomparabletoKRASknockdown(Fig.
6A).
IntheKRASmutantcelllinesHCT116andSW403,KRASknockdownalsoincreasedapoptosis,asjudgedbyanin-creasedsub-G1populationandelevatedcaspaseactivityinthesecells.
WhereascodepletingBRAFandCRAFdidnotstronglyinducecelldeath,theadditionofATG7andRAC1tothecombinationledtoastrongapoptosisresponseinthesecellsthatwascomparabletoKRASknockdown(Fig.
6BandC).
KnockingdownATG7alonehadlittleimpactoncellcycleandapoptosis,whereasRAC1knockdownhadamodestbutgenotype-independenteffectonG1arrestinbothKRASmutantandWTcells(SIAppendix,Fig.
S15).
TheenhancedG1arrestandapoptosiseffectfollowingBRAF+CRAF+ATG7codepletionwasnotobservedinKRASWTcelllines,orwiththeircorre-spondingC911controlsiRNAs(Fig.
6).
Takentogether,ourfindingsuggeststhatcotargetingtheMAPKandautophagypathwaysusingthetargetcombinationofBRAF,CRAF,andATG7couldbeaviablestrategythatoffersagoodtherapeuticwindowinKRASmutantcells.
DiscussionFunctionalgenomicsscreensusinggenome-wideRNAiandCRISPR/Cas9librarieshavebeenextensivelyusedfortargetidentification.
However,large-scaleloss-of-functionscreenshavebeenmostlylimitedtotheanalysisofsingle-genephenotypes.
Inmammaliancells,false-negativeresultsduetopervasivegeneparalogredundancyisasignificantissueforthesestudies.
In-deed,manycanonicalRASeffectors,includingRAFandPI3Kparalogs,wererarelyscoredinKRASsyntheticlethalscreensusingbothshRNAandCRISPRlibraries(20–26).
Toovercomethislimitation,wehavedevelopedacombinatorialsiRNAplat-formthatusesexperimentallyvalidatedandhighlypotentsensorsiRNAstoachieveefficientmultigeneknockdowninthesamecell.
ThisplatformcanreliablyachievethesimultaneousFig.
5.
RAC1andATG7depletionsensitizesKRASmutantcellstowardMAPKpathwayinhibitors.
TheKRASmutantcancercelllinesHCT116andMIAPaCa-2andKRASWTcancercelllineBxPC-3weretransfectedwithsiRNAsagainstRAC1and/orATG7.
Onedayposttransfection,cellsweretreatedwithvariousconcentrationsoftheRAFinhibitorRAF709ortheMEKinhibitortrametinib.
Cellviabilitywasdetermined4dlatertoobtaindose–responsecurves,andRAF709andtrametinibIC50valuesweredeterminedandareshownasbarcharts(errorbarsrepresentSD).
*P<0.
05vs.
negative(Neg);**P<0.
01vs.
Neg.
4514|www.
pnas.
org/cgi/doi/10.
1073/pnas.
1817494116Leeetal.
DownloadedbyguestonFebruary23,2021knockdownofuptosevengenesinthecell(29).
CombinatorialgenetargetingatthislevelofcomplexityisnoteasilyachievablewithcurrentCRISPR/Cas9technologies.
Ourapproachoffersseveraladvantagescomparedwithtraditionalsingle-genelibraryscreens.
First,itenablesacomprehensivedissectionofgeneparalogredundancywithinagenenodetoaccuratelyassessitsfunction.
Second,itenablestheanalysisofgeneandpathwayin-teractionsatthesystemsleveltounderstandhowdifferentgeneswithinthenetwork,suchasthoseintheRASnetwork,couldco-operatewithorantagonizeeachothertomodulatethenetwork'soutput.
Thiswasnotpreviouslypossiblewithsingle-geneanalysis.
Third,ourplatformoffersapowerfulandhighlyscalableapproachtorationallyevaluatetargetcombination.
WedemonstratedthatsiRNAscanbecombinedandtransfectedintocellsathigheffi-ciency,thesimplicityofwhichrivalsthatofdrugcombinations.
Thus,virtuallyanytargetcombinationcanbeinvestigatedinanunbiasedfashionregardlessofwhethersmallmoleculeinhibitorsarereadilyavailable.
Ourapproachcouldthusacceleratethede-velopmentofeffectivecombinationtherapyforprecisionmedicine.
OursystematicinterrogationofRASeffectorandstressre-sponsepathwaysfortheirroleinmediatingKRASaddictionledtoseveralimportantmechanisticinsightswithtranslationalim-plications.
Ouranalysisofnearly500single-genenodes,genenodepairs,andgeneparalogcombinationsrevealedthattheRAFnodeisthemostcriticaloncoeffectordownstreamofmu-tantKRASandshouldconstitutethebackboneofcombinationtherapies.
ThisisinagreementwithpreviousstudiesshowingthattheMAPKpathwayisessentialforRAS-drivencellpro-liferation(34).
Unexpectedly,wenotedthatknockingdowntheRAF,MEK,andERKnodeswithintheMAPKpathwayledtounequaldegreesofdifferentialdependencyinKRASmutantcells.
Thistrendwasalsoobservedinourprevioussingle-nodestudyacross92KRASmutantandWTcancercelllines(27).
OurobservationisinagreementwithapreviousstudyshowingthatKras-drivencolorectaltumorsinmicearemoresensitivetoRAFinhibitionthanMEKinhibition(50).
ThemechanismbywhichKRASdependencyisbettercapturedbyRAFknockdowncom-paredwithMEKandERKknockdowniscurrentlyunclear.
WefoundthatknockingdowntheRAF,MEK,andERKnodesallhadasimilarimpactonthephospho-ERKlevelincells;thus,thisdifferencecannotbesimplyattributedtoMAPKpathwayflux.
ItispossiblethatRAFknockdownmightleadtoamoresustainedlossofERKactivityinthenucleus(51).
Alternatively,RAFcouldpo-tentiallymediateKRASoncogeneaddictionthroughbothMAPKpathway-dependentand-independentmeans(43,50,52–55).
Thus,furtherstudiesareneededtoclarifythemechanismofRAFde-pendencyinKRASmutantcells.
Bycarefullydissectinggeneparalogdosageeffect,wedem-onstratedthatcodepletingBRAFandCRAF,whilesparingARAF,issufficienttodisruptmuchoftheoncoeffectorfunctionoftheRAFnode,whileminimizingtoxicityinnormalcells.
BecauseRAFsignalingislikelytobeessentialinnormaltissues,apan-RAFinhibitorcouldhaveexcessivetoxicity.
Ourfindinghighlightstheimportanceofcarefultargetselectiontodiscrim-inatetheoncogenicactivityoftheMAPKpathwayfromitsphysiologicalactivity.
Previously,ithasbeenshownthatCraf,butnotBraf,iscriticalforKrasmutanttumordevelopmentinmice(39,42,43).
OurresultssuggestthatinhumanKRASmutantcells,BRAFandCRAFneedtobecotargetedtoeffectivelydisruptoncogenicKRASsignaling.
WeproposethattheidealRAFinhibitorwouldbeARAF-sparingandwithoutparadoxicalactivity.
Onepromisingapproachtodevelopcompoundswiththesepropertieswouldbetoidentifysmallmoleculesthatse-lectivelydisruptBRAF/CRAFdimerization(56).
OurworkprovidessupportforthenotionthatoncogenicKRASsignalingismediatedbymultiplepathways.
WeshowedthatcotargetingtheautophagyE1ligaseATG7orthesmallGTPaseRAC1,togetherwithBRAFandCRAF,couldfurtherimprovethecaptureofKRASdependency.
TheidentificationoftheautophagypathwayasaRAFcotargetisavaluableinsight,asautophagyisanintimatecomponentoftumordevelopment(57).
PreviousstudiesusingmousemodelsofKrasmutantlungandpancreaticcancerhavedemonstratedacriticalroleofAtg7andAtg5fortumorprogressionandforthemaintenanceofenergyandnucleotidesupplyintumorcells,suggestingtheautophagypathwayasatargetinKRAS-drivencancer(38,58–60).
Inagreementwithpreviousinvitrostudiesshowingthatthepro-liferationofsomeKRAS-mutantcancercelllinesundernutrient-repleteconditionsisnotstronglyimpactedbypharmacologicalinhibitionofautophagy,byRNAi-mediatedacuteATG5/ATG7knockdown,orbyCRISPR-mediatedATG7knockout(61),wefoundthatATG7knockdownalonehadrelativelylittleeffectontheviabilityofthecancercelllinestestedinthisstudy.
However,wefoundthatATG7knockdownenhancesthetoxicityofBRAFandCRAFsiRNAs,consistentwitharoleofRAS/RAFsignalinginthemetabolicrewiringofcancercells(57,62).
Together,thesepriorstudiesandourcurrentstudysupportamodelwhereKRAS-drivenmetabolicalterationsincancercellsrenderthemparticu-larlydependentontheautophagypathwayasasurvivalmecha-nismupontheacuteinhibitionoftheRAF/MAPKpathway(SIAppendix,Fig.
S16).
Duetothenonessentialityofautophagyinnormalcellsundernutrient-repleteconditionsandthefactthatadultmicewithacute,systemicdeletionofAtg7areabletosurviveforseveralmonths(38),weexpectthecombinationofautophagywithRAFinhibitiontobewelltoleratedinvivo.
TheautophagypathwayisanononcogeneaddictionmechanisminKRASmutantcells.
Thus,theRAF+ATG7targetcombinationexploitsbothoncogeneandnononcogeneaddictioninKRASmutantcells,andthiscouldleadtoabetterandmoredurableresponsethantargetingoncogeneaddictionalone.
ThisnotionissupportedbyapreviousstudyshowingATG7deficiencyenhancestheanti-tumoractivityofBRAFinhibitorinBRAFmutantmelanomas(63).
AlthoughnoATG7inhibitorsarecurrentlyavailable,selectiveinhibitorshavebeendevelopedforneddylationandSUMOylationFig.
6.
ImpactofsiRNAcombinationsoncellcycleandapoptosisinKRASmutantcells.
KRASmutant(HCT116,MIAPaCa-2,andSW403)andKRASWT(Caco-2andSW48)cancercelllinesweretransfectedwithvarioussiRNAcombinationstargetingBRAF(B),CRAF(C),RAC1(1),andATG7(7).
Corre-spondingC911siRNApoolswereincludedasrescuecontrols.
Cellcycleandapoptosisstatuswereanalyzed3dposttransfection.
(A)ChangesinG0/G1viablecellpopulationsbyflowcytometry.
(B)Changesinsub-G1deadcellsbyflowcytometry.
(C)Changesincaspaseactivityincells.
Leeetal.
PNAS|March5,2019|vol.
116|no.
10|4515MEDICALSCIENCESSEECOMMENTARYDownloadedbyguestonFebruary23,2021E1ligases(64,65),andtheseexperiencescouldprovideguid-anceforanATG7inhibitordiscoveryprogram.
Furthermore,hydroxychloroquine,anFDA-approvedantimalariadrugthatalsoinhibitsautophagosomefusionwiththelysosome,iscurrentlyun-dergoingclinicaltrialsasananticanceragent(3).
OurfindingswouldsupportitsrationalcombinationwitheitherMEKinhibitorsorparadox-breakerRAFinhibitorscurrentlyinclinicaltrialforKRASmutanttumors.
CombinedtargetingofRAFkinasesandtheautophagypathwaycouldpotentiallyofferlesstoxicityandbetterefficacycomparedwithdrugcombinationscurrentlybeingtested.
OurstudyalsosuggeststhatRAC1isapotentialcotargetwithRAF.
Rac1iscriticalforKRAS-drivenlungandpancreaticcancerinitiationinmice(44,66).
However,RAC1mighthaveanessentialroleinnormaltissuefunction(67),andRAC1knock-downistoxicinsomeKRASWTcells.
Thus,furtherworkisnecessarytoevaluatewhethercotargetingRAC1andRAFcouldprovideasufficientlylargetherapeuticwindow.
Ourcombina-torialanalysisrevealedseveralprinciplesfortherationalselectionoftargetcombination.
First,combinationstargetingtwoessentialpathways,suchastheRAF+PI3Kcombination,mayparadoxicallyofferlessselectivityduetohighertoxicityinKRASWTcells.
Inthissetting,pan-paraloginhibitorsarenotdesirableandparalog-selectiveinhibitorsthatcanrestrictoncogenicsignalingwhilesparingphysiologicalsignalingarenecessarytopreserveagoodtherapeuticwindow.
Second,top-scoringsingletargets,suchasthoseinvolvingRAFandRALparalogs,maynotnecessarilyresultinthebestcombinations.
Third,nonessentialsingletargets,suchasATG7,could,infact,beavaluablepartnerinacombinationsetting.
Thesefunctionalinterplayswouldbedifficulttopredictbasedonsingle-geneandsingle-nodeanalysis.
Thus,directanalysisofgenenodeandpathwayinteractionsatthesystemsleveliscriticalforunmaskingcomplexbehaviorsintheRASsignalnetwork.
OurstudyhasuncoveredasignificantdegreeofheterogeneityamongKRASmutantcelllineswithregardtothespecificRASeffectorandstressresponsepathwaysusedtosupportKRASaddiction.
Previously,weshowedthatstrongKRASdependencyisassociatedwithMAPKpathwaydependency,whereasKRASmutantcellsthatarelessdependentonKRASexhibitde-pendencyonp90RSKkinases(27).
Inthecurrentstudy,weshowedthateachKRASmutantcelllinehasadistinctde-pendencysignaturefortop-scoringgenenodecombinations.
Thepublicnode-pairdependenciesacrossmultiplecelllinesare,ingeneral,lesseffectiveatcapturingKRASdependencythanthebestprivatenode-pairdependencieswithrespecttoanindividualcellline.
Thisisanalogoustopreviousworkexaminingpublicandprivatedrugsensitivitiesinlungcancercells(36).
Howsuchheterogeneityineffectorandstressresponsepathwaydepen-dencyarisesamongKRASmutanttumorcellsispoorlyun-derstood.
Possiblecontributingfactorsincludetheetiologyoftumorevolutionandthepresenceofco-occurringmutationsinthesametumorcell.
Atranslationalimplicationofthishetero-geneityisthatKRASmutationalonemaynotbesufficientasasinglebiomarkertodirectthechoiceoftargetedtherapies,asdrugcombinationsdesignedtoworkforthemajorityofKRASmutanttumorsarenotexpectedtoworkparticularlywellforanygivenKRASmutanttumor.
Multiplecombinationtherapies,witheachoptimizedforasubsetofKRASmutanttumorssharingasimilareffectorandstresspathwaydependencyprofile,mightbenecessarytoadequatelyaddresstumorheterogeneityinthissetting.
Thus,additionalbiomarkersareneededtosubdivideKRASmutanttumorsbasedontheirfunctionalprofilesinef-fectorandstresspathwaydependency.
OurcurrentstudyisonlypoweredtodetectcommondependenciesacrossKRASmutantcelllines.
ExpandingthisanalysistoamuchlargerpanelofKRASmutantcelllines(27)willenableustofurthersub-categorizeKRASmutantcellsbasedontheirpatternsofef-fectorandstressresponsepathwaydependencyandtodiscoverthebiomarkersthatareassociatedwitheachcategorytobetterdirectthechoiceofdrugcombination.
OuranalysisrevealedasignificantphenotypicgapbetweentargetingtheKRASoncoproteinitselfvs.
targetingitsdownstreameffectornetwork.
WenotedthatnoneofthesiRNAcombinationscouldfullyphenocopyKRASknockdown:Thebesteffectorcombinationweidentifiedonlycapturesalittleover50%ofKRASdependency.
Althoughseveralexplanationscouldbeoffered(dis-cussedinSIAppendix,SINotes),itispossiblethatthefunctionaloverlapbetweenoncogenicandphysiologicalRASsignalingim-posesaselectivityceiling.
Thus,targetingKRASoncoeffectorsmayneverachievethesametherapeuticwindowastargetingKRASoncoproteinitself.
RecenteffortsindevelopingnovelKRASG12Cinhibitorshavegainedsignificanttraction(68–70).
AusefulfuturedirectionwouldbetoidentifyKRASoncoeffectorsthatstronglysynergizewithKRASG12Cinhibitorstoenhancethegenotype-dependentkillingofKRASmutantcancercells.
MaterialsandMethodsRASEffectorandStressPathwayGenesiRNALibraryCuration.
SensorsiRNAsagainstthelistofRASeffectorgenesinterrogatedinthisstudyweregen-eratedaspreviouslydescribed(29,31)andasdetailedinSIAppendix.
Foraselectedsubsetofon-targetsiRNAsinourlibrary,wegeneratedtheirsequence-specificC911rescuesiRNAsaspreviouslydescribed(32).
AllsiRNAsequencesarelistedinDatasetS1.
CellLinesandReagents.
AllcelllinesusedinthisstudywereculturedasdescribedinSIAppendix.
TrametinibandRAF709(MedChemExpress)weredissolvedinDMSOatstockconcentrationsof1mMand10mM,respectively.
TransfectionofsiRNACombinationsandInhibitorTreatment.
ToknockdownmultiplegenetargetssimultaneouslyandtocotreatcellswithsiRNAsandinhibitors,reversesiRNAtransfectionwasperformedaspreviouslydescribedbyGarimellaetal.
(71),withslightmodificationsasdescribedinSIAppendix.
CellViability,Caspase3/7Activity,andCellCycleAssays.
Cellviabilityandcaspase3/7assayswereperformedbyusingtheCellTiter-GloLuminescentCellViabilityAssay(Promega)andApoLive-GloMultiplexAssay(Promega),re-spectively,andasdescribedinSIAppendix.
CellcycleanalysiswasperformedaspreviouslydescribedbyWengetal.
(72)andinSIAppendix.
Immunoblotting.
ToexaminetheeffectofsiRNA-mediatedknockdownonproteinexpressionandpathwayactivity,at5dpost-siRNAtransfection,whole-cellextractwasharvestedasdescribedinSIAppendix.
DataAnalysisandStatistics.
ToassessthedifferentialimpactofsiRNAsonKRASmutantandWTcancercelllinesaswellasimmortalized,non-transformednormalcelllines,wenormalizedcellviabilityandquantifiedtheeffectofeachsiRNApoolasdescribedinSIAppendix.
UnsupervisedhierarchicalclusteringwasperformedusingPartekGenomicSuitesoftware(Partek,Incorporated).
Outlieranalysis,pairedttests,andANOVAandpostanalysiswereperformedusingGraphPadPrism(GraphPadSoftware).
ACKNOWLEDGMENTS.
WethankDrs.
EricBatchelor,JayneStommel,andRosandraKaplanfortheirconstructivesuggestions,andtheFlowCytometryCoreFacilityoftheCenterforCancerResearchattheNationalCancerInstitute(NCI)forthetechnicalsupport.
ThisworkwassupportedbyNCIIntramuralGrantZIABC011437(toJ.
L.
)andbyanNCIDirector'sInnovationAward(toC.
-S.
L.
).
1.
Pylayeva-GuptaY,GrabockaE,Bar-SagiD(2011)RASoncogenes:Weavingatumor-igenicweb.
NatRevCancer11:761–774.
2.
MontalvoSK,LiL,WestoverKD(2017)RationaleforRASmutation-tailoredtherapies.
FutureOncol13:263–271.
3.
CoxAD,FesikSW,KimmelmanAC,LuoJ,DerCJ(2014)DruggingtheundruggableRAS:MissionpossibleNatRevDrugDiscov13:828–851.
4.
EngelmanJA,etal.
(2008)EffectiveuseofPI3KandMEKinhibitorstotreatmutantKrasG12DandPIK3CAH1047Rmurinelungcancers.
NatMed14:1351–1356.
5.
AlagesanB,etal.
(2015)CombinedMEKandPI3Kinhibitioninamousemodelofpancreaticcancer.
ClinCancerRes21:396–404.
6.
BlumenscheinGR,Jr,etal.
(2015)ArandomizedphaseIIstudyoftheMEK1/MEK2inhibitortrametinib(GSK1120212)comparedwithdocetaxelinKRAS-mutantadvancednon-small-celllungcancer(NSCLC).
AnnOncol26:894–901.
7.
JnnePA,etal.
(2017)Selumetinibplusdocetaxelcomparedwithdocetaxelaloneandprogression-freesurvivalinpatientswithKRAS-mutantadvancednon-smallcelllungcancer:TheSELECT-1RandomizedClinicalTrial.
JAMA317:1844–1853.
4516|www.
pnas.
org/cgi/doi/10.
1073/pnas.
1817494116Leeetal.
DownloadedbyguestonFebruary23,20218.
StephenAG,EspositoD,BagniRK,McCormickF(2014)Draggingrasbackinthering.
CancerCell25:272–281.
9.
SchollFA,etal.
(2007)Mek1/2MAPKkinasesareessentialformammaliandevelop-ment,homeostasis,andRaf-inducedhyperplasia.
DevCell12:615–629.
10.
BrachmannSM,etal.
(2005)Roleofphosphoinositide3-kinaseregulatoryisoformsindevelopmentandactinrearrangement.
MolCellBiol25:2593–2606.
11.
ShimizuT,etal.
(2012)Theclinicaleffectofthedual-targetingstrategyinvolvingPI3K/AKT/mTORandRAS/MEK/ERKpathwaysinpatientswithadvancedcancer.
ClinCancerRes18:2316–2325.
12.
DoK,etal.
(2015)Biomarker-drivenphase2studyofMK-2206andselumetinib(AZD6244,ARRY-142886)inpatientswithcolorectalcancer.
InvestNewDrugs33:720–728.
13.
TolcherAW,etal.
(2015)PhaseIstudyoftheMEKinhibitortrametinibincombinationwiththeAKTinhibitorafuresertibinpatientswithsolidtumorsandmultiplemye-loma.
CancerChemotherPharmacol75:183–189.
14.
BedardPL,etal.
(2015)AphaseIbdose-escalationstudyoftheoralpan-PI3Kinhibitorbuparlisib(BKM120)incombinationwiththeoralMEK1/2inhibitortrametinib(GSK1120212)inpatientswithselectedadvancedsolidtumors.
ClinCancerRes21:730–738.
15.
WainbergZA,etal.
(2017)Amulti-armphaseIstudyofthePI3K/mTORinhibitorsPF-04691502andgedatolisib(PF-05212384)plusirinotecanortheMEKinhibitorPD-0325901inadvancedcancer.
TargetOncol12:775–785.
16.
Grilley-OlsonJE,etal.
(2016)AphaseIbdose-escalationstudyoftheMEKinhibitortrametinibincombinationwiththePI3K/mTORinhibitorGSK2126458inpatientswithadvancedsolidtumors.
InvestNewDrugs34:740–749.
17.
LuoJ,SoliminiNL,ElledgeSJ(2009)Principlesofcancertherapy:Oncogeneandnon-oncogeneaddiction.
Cell136:823–837.
18.
CorcoranRB,etal.
(2013)SyntheticlethalinteractionofcombinedBCL-XLandMEKinhibitionpromotestumorregressionsinKRASmutantcancermodels.
CancerCell23:121–128.
19.
MaloneCF,etal.
(2014)DefiningkeysignalingnodesandtherapeuticbiomarkersinNF1-mutantcancers.
CancerDiscov4:1062–1073.
20.
SarthyAV,etal.
(2007)Survivindepletionpreferentiallyreducesthesurvivalofac-tivatedK-Ras-transformedcells.
MolCancerTher6:269–276.
21.
LuoJ,etal.
(2009)Agenome-wideRNAiscreenidentifiesmultiplesyntheticlethalinteractionswiththeRasoncogene.
Cell137:835–848.
22.
BarbieDA,etal.
(2009)SystematicRNAinterferencerevealsthatoncogenicKRAS-drivencancersrequireTBK1.
Nature462:108–112.
23.
TsherniakA,etal.
(2017)Definingacancerdependencymap.
Cell170:564–576.
e16.
24.
McDonaldER,3rd,etal.
(2017)ProjectDRIVE:Acompendiumofcancerdependenciesandsyntheticlethalrelationshipsuncoveredbylarge-scale,deepRNAiscreening.
Cell170:577–592.
e10.
25.
WangT,etal.
(2017)GeneessentialityprofilingrevealsgenenetworksandsyntheticlethalinteractionswithoncogenicRas.
Cell168:890–903.
e15.
26.
MartinTD,etal.
(2017)AroleformitochondrialtranslationinpromotionofviabilityinK-Rasmutantcells.
CellRep20:427–438.
27.
YuanTL,etal.
(2018)DifferentialeffectorengagementbyoncogenicKRAS.
CellRep22:1889–1902.
28.
WojnowskiL,StancatoLF,LarnerAC,RappUR,ZimmerA(2000)OverlappingandspecificfunctionsofBrafandCraf-1proto-oncogenesduringmouseembryogenesis.
MechDev91:97–104.
29.
YuanTL,etal.
(2014)DevelopmentofsiRNApayloadstotargetKRAS-mutantcancer.
CancerDiscov4:1182–1197.
30.
DorardC,etal.
(2017)RAFproteinsexertbothspecificandcompensatoryfunctionsduringtumourprogressionofNRAS-drivenmelanoma.
NatCommun8:15262.
31.
FellmannC,etal.
(2011)FunctionalidentificationofoptimizedRNAitriggersusingamassivelyparallelsensorassay.
MolCell41:733–746.
32.
BuehlerE,ChenYC,MartinS(2012)C911:Abench-levelcontrolforsequencespecificsiRNAoff-targeteffects.
PLoSOne7:e51942.
33.
EmeryCM,etal.
(2009)MEK1mutationsconferresistancetoMEKandB-RAFin-hibition.
ProcNatlAcadSciUSA106:20411–20416.
34.
DrostenM,etal.
(2010)GeneticanalysisofRassignallingpathwaysincellpro-liferation,migrationandsurvival.
EMBOJ29:1091–1104.
35.
BottorffD,StangS,AgellonS,StoneJC(1995)RASsignallingisabnormalinac-raf1MEK1doublemutant.
MolCellBiol15:5113–5122.
36.
KimHS,etal.
(2013)Systematicidentificationofmolecularsubtype-selectivevul-nerabilitiesinnon-small-celllungcancer.
Cell155:552–566.
37.
GuoJY,etal.
(2011)ActivatedRasrequiresautophagytomaintainoxidativeme-tabolismandtumorigenesis.
GenesDev25:460–470.
38.
Karsli-UzunbasG,etal.
(2014)Autophagyisrequiredforglucosehomeostasisandlungtumormaintenance.
CancerDiscov4:914–927.
39.
BlascoRB,etal.
(2011)c-Raf,butnotB-Raf,isessentialfordevelopmentofK-Rasoncogene-drivennon-smallcelllungcarcinoma.
CancerCell19:652–663.
40.
PeschardP,etal.
(2012)GeneticdeletionofRALAandRALBsmallGTPasesrevealsredundantfunctionsindevelopmentandtumorigenesis.
CurrBiol22:2063–2068.
41.
SmithJL,etal.
(2016)One-stepimmortalizationofprimaryhumanairwayepithelialcellscapableofoncogenictransformation.
CellBiosci6:57.
42.
KarrethFA,FreseKK,DeNicolaGM,BaccariniM,TuvesonDA(2011)C-RafisrequiredfortheinitiationoflungcancerbyK-Ras(G12D).
CancerDiscov1:128–136.
43.
SanclementeM,etal.
(2018)c-RAFablationinducesregressionofadvancedKras/Trp53mutantlungadenocarcinomasbyamechanismindependentofMAPKsignal-ing.
CancerCell33:217–228.
e4.
44.
KissilJL,etal.
(2007)RequirementforRac1inaK-rasinducedlungcancerinthemouse.
CancerRes67:8089–8094.
45.
ChienY,etal.
(2006)RalBGTPase-mediatedactivationoftheIkappaBfamilykinaseTBK1couplesinnateimmunesignalingtotumorcellsurvival.
Cell127:157–170.
46.
NishiguchiGA,etal.
(2017)DesignanddiscoveryofN-(2-methyl-5′-morpholino-6′-((tetrahydro-2H-pyran-4-yl)oxy)-[3,3′-bipyridin]-5-yl)-3-(trifluoromethyl)benzamide(RAF709):Apotent,selective,andefficaciousRAFinhibitortargetingRASmutantcancers.
JMedChem60:4869–4881.
47.
ShaoW,etal.
(2018)AntitumorpropertiesofRAF709,ahighlyselectiveandpotentinhibitorofRAFkinasedimers,intumorsdrivenbymutantRASorBRAF.
CancerRes78:1537–1548.
48.
PumigliaKM,DeckerSJ(1997)CellcyclearrestmediatedbytheMEK/mitogen-activatedproteinkinasepathway.
ProcNatlAcadSciUSA94:448–452.
49.
HerreraR,HubbellS,DeckerS,PetruzzelliL(1998)ArolefortheMEK/MAPKpathwayinPMA-inducedcellcyclearrest:ModulationofmegakaryocyticdifferentiationofK562cells.
ExpCellRes238:407–414.
50.
HaigisKM,etal.
(2008)DifferentialeffectsofoncogenicK-RasandN-Rasonpro-liferation,differentiationandtumorprogressioninthecolon.
NatGenet40:600–608.
51.
BertiDA,SegerR(2017)ThenucleartranslocationofERK.
MethodsMolBiol1487:175–194.
52.
HüserM,etal.
(2001)MEKkinaseactivityisnotnecessaryforRaf-1function.
EMBOJ20:1940–1951.
53.
ChenJ,FujiiK,ZhangL,RobertsT,FuH(2001)Raf-1promotescellsurvivalbyan-tagonizingapoptosissignal-regulatingkinase1throughaMEK-ERKindependentmechanism.
ProcNatlAcadSciUSA98:7783–7788.
54.
O'NeillE,RushworthL,BaccariniM,KolchW(2004)RoleofthekinaseMST2insup-pressionofapoptosisbytheproto-oncogeneproductRaf-1.
Science306:2267–2270.
55.
PiazzollaD,MeisslK,KucerovaL,RubioloC,BaccariniM(2005)Raf-1setsthethresholdofFassensitivitybymodulatingRok-alphasignaling.
JCellBiol171:1013–1022.
56.
FreemanAK,RittDA,MorrisonDK(2013)EffectsofRafdimerizationanditsin-hibitiononnormalanddisease-associatedRafsignaling.
MolCell49:751–758.
57.
KimmelmanAC,WhiteE(2017)Autophagyandtumormetabolism.
CellMetab25:1037–1043.
58.
GuoJY,WhiteE(2013)Autophagyisrequiredformitochondrialfunction,lipidme-tabolism,growth,andfateofKRAS(G12D)-drivenlungtumors.
Autophagy9:1636–1638.
59.
YangA,etal.
(2014)Autophagyiscriticalforpancreatictumorgrowthandpro-gressionintumorswithp53alterations.
CancerDiscov4:905–913.
60.
GuoJY,etal.
(2016)AutophagyprovidesmetabolicsubstratestomaintainenergychargeandnucleotidepoolsinRas-drivenlungcancercells.
GenesDev30:1704–1717.
61.
EngCH,etal.
(2016)MacroautophagyisdispensableforgrowthofKRASmutanttumorsandchloroquineefficacy.
ProcNatlAcadSciUSA113:182–187.
62.
YunJ,etal.
(2009)GlucosedeprivationcontributestothedevelopmentofKRASpathwaymutationsintumorcells.
Science325:1555–1559.
63.
XieX,KohJY,PriceS,WhiteE,MehnertJM(2015)Atg7overcomessenescenceandpromotesgrowthofBrafV600E-drivenmelanoma.
CancerDiscov5:410–423.
64.
SoucyTA,etal.
(2009)AninhibitorofNEDD8-activatingenzymeasanewapproachtotreatcancer.
Nature458:732–736.
65.
HeX,etal.
(2017)ProbingtherolesofSUMOylationincancercellbiologybyusingaselectiveSAEinhibitor.
NatChemBiol13:1164–1171.
66.
HeidI,etal.
(2011)EarlyrequirementofRac1inamousemodelofpancreaticcancer.
Gastroenterology141:719–730,730.
e1–730.
e7.
67.
BenitahSA,FryeM,GlogauerM,WattFM(2005)Stemcelldepletionthroughepi-dermaldeletionofRac1.
Science309:933–935.
68.
LitoP,SolomonM,LiLS,HansenR,RosenN(2016)Allele-specificinhibitorsinactivatemutantKRASG12Cbyatrappingmechanism.
Science351:604–608.
69.
PatricelliMP,etal.
(2016)SelectiveinhibitionofoncogenicKRASoutputwithsmallmoleculestargetingtheinactivestate.
CancerDiscov6:316–329.
70.
XiongY,etal.
(2016)CovalentguanosinemimeticinhibitorsofG12CKRAS.
ACSMedChemLett8:61–66.
71.
GarimellaSV,etal.
(2014)Identificationofnovelmolecularregulatorsoftumorne-crosisfactor-relatedapoptosis-inducingligand(TRAIL)-inducedapoptosisinbreastcancercellsbyRNAiscreening.
BreastCancerRes16:R41.
72.
WengMT,etal.
(2012)EvolutionarilyconservedproteinERHcontrolsCENP-EmRNAsplicingandisrequiredforthesurvivalofKRASmutantcancercells.
ProcNatlAcadSciUSA109:E3659–E3667.
Leeetal.
PNAS|March5,2019|vol.
116|no.
10|4517MEDICALSCIENCESSEECOMMENTARYDownloadedbyguestonFebruary23,2021