Post on 22-May-2018
Accepted for publication in Philosophy and Biology 25.04.2018
1
Cancerstemcellsmodulatepatternsandprocessesofevolutionincancers
LucieLaplane1,2
1.UMR8590Institutd’HistoireetPhilosophiedesSciencesetdesTechniques(IHPST),CNRS,UniversitéParisIPanthéon-Sorbonne,ENS,13rueduFour,75006Paris,France;2.UMR1170,Inserm,GustaveRoussy,114rueEdouardVaillant,Villejuif,France
Correspondenceto
LucieLaplaneIHPST,UniversitéParisIPanthéon-Sorbonne.13rueduFour75006Paris,Francee-mail:lucie.laplane@univ-paris1.frtel:+33(0)143546036fax:+33(0)143252948
Abstract
Theclonalevolution(CE)modelandthecancerstemcell(CSC)modelaretwoindependentmodelsofcancers,
yetrecentdatashowsintersectionsbetweenthetwomodels.ThisarticleexplorestheimpactsoftheCSCmodel
ontheCEmodel.IshowthatCSCrestriction,whichdependsonCSCfrequencyincancercellpopulationsandon
theprobabilityofdedifferentiationof cancernon-stemcells (non-CSC) intoCSCs, can favoror impede some
patternsofevolution(linearorbranchedevolution)andsomeprocessesofevolution(drift,evolutionbynatural
selection,complexadaptations).TakingCSCrestrictionintoaccountfortheCEmodelthushasimplicationsfor
thewayinwhichweunderstandthepatternsandprocessesofevolution,andcanalsoprovidenewleadsfor
therapeuticinterventions.
Keywords
Cancerstemcell;Clonalevolution;Cancerprogression;Drift;Evolutionbynaturalselection
Acknowledgments
Thisworkwasfinancedbytheprogram“StratifiedOncologyCellDNARepairAndTumorImmuneElimination”
(SOCRATE)oftheSiteofIntegratedResearchInCancerology(SIRIC)inGustaveRoussy(grantINCa-DGOS-INSERM
6043)andtheCancéropôleÎle-de-France(grant2016-1-EMERG-50-CNRSDR1-1).
Iamdeeplygratefultothereviewers. Ihavebenefitedfromthebestreviewspossibleandtheirremarksand
critiqueswereparticularlyhelpfultoreviseandstrengthenmyargument.KateMacCorddidanamazingjobnot
onlywitheditingtheEnglishbutalsowithhelpingmetoclarifymyarguments.Thispaperalsogreatlybenefited
frommanycomments fromphilosophers,biologists, andoncologists,on itsearlyand/or later versions. I am
particularlygrateful toMelindaFagan, JeanGayon,Pierre-LucGermain,PeterGodfrey-Smith,KateMacCord,
FrancescaMerlin,ThomasPradeu,MurielleGaudry,StéphaneGiraudier,Anne-MarieLyne,MrinalPatnaik,Leïla
Perié,andespeciallytoEricSolaryandallofmyteamatGustaveRoussy.
Accepted for publication in Philosophy and Biology 25.04.2018
2
Theclonalevolution(CE)modelandthecancerstemcell(CSC)modelaretwomodelsofcancerthathavean
ambiguous theoretical relationship. The CEmodel describes cancer as an evolutionary process inwhich the
accumulationofgeneticandepigeneticalterationsresultsinthediversificationofcancercellsthroughspaceand
time(Figure1a).TheCSCmodelstatesthatcancersmimictissueorganizationandemergeanddeveloponlyfrom
arestrictedfractionofcancercellswithstem-likepropertiescalledCSCs(Figure1b).Historically,theCSCmodel
wasframedinoppositiontotheCEmodel(Reyaetal.2001;Wichaetal.2006;Shipitsinetal.2007;Pantic2011).
Someinvestigatorshavesincesuggestedthatthetwocompetingmodelsare“notnecessarilymutuallyexclusive”
(Polyak2007,3160;CampbellandPolyak2007;Lagasse2008), thatoneor theothercouldmoreadequately
explaindifferenttypesofcancer(AdamsandStrasser2008;Shackletonetal.2009),orthatthetwomodelsare
complementaryandshouldbecombinedbecause theyeachprovideexplanationsofdifferentaspectsof the
samecancer(Fabianetal.2009;Greaves2010;Langetal.2015).
Several linesofexperimentationhaveshownthatCEoccurs inCSCs(forthefirstevidenceseePiccirilloetal.
2009;Andersonetal.2011;Nottaetal.2011).TakingintoaccounttheimpactofCEonCSCsintheCSCmodel
appearstobeimportantforatleasttworeasons.First,forthetherapeuticstrategythatemergedfromtheCSC
model and that consists of specifically targeting the CSCs during cancer treatments, itmeans that CSCs can
represent a heterogeneous target. CSCs can differ both synchronically (at the diagnosis for example) and
diachronically (from diagnosis to relapse for example). This heterogeneity between CSCs can result in the
coexistenceofCSCswithvarioussensitivitiestotherapies(e.g.Meyeretal.2015).Second,theaccumulationof
new mutations in CSCs can modify their properties. For example, additional mutations could lead to the
acquisition of migratory ability, distinguishing “stationary CSCs” from “migrating CSCs” able to generate
metastases(Brabletzetal.2005;Bapat2007;Odouxetal.2008;TakebeandIvy2010).Bynottakingintoaccount
CE,theCSCmodelmissestheheterogeneityofthisCSCpopulation.TheincorporationofCEintotheCSCmodel
isnowfairlyaccepted(e.g.KresoandDick2014).
Incontrast,theCEmodelstilllargelyignorestheCSCmodel.Throughoutthispaper,IarguethatCSCsimpactCE,
particularly through a phenomenon that I call “CSC restriction” (referring to the fact that CSCs are only a
restrictedfractionofagivencancer),andthusthattheCEmodelneedstoincorporatesomefeaturesoftheCSC
modelinordertoaccountforalloftheconsequencesthatCSCrestrictionhasforCE.
Thisresearchtookrootfrompreviousphilosophicalworks.InhisbookonDarwinianpopulations,Godfrey-Smith
(2009) introduced a framework that allows one to distinguish different types of Darwinian populations, for
example paradigmatic Darwinian populations in which evolution by natural selection can produce complex
adaptations,i.e.noveltraits,versusmarginalDarwinianpopulationsinwhichevolutionbynaturalselectioncan
only produce simple adaptations, i.e. change in trait frequency. Distinguishing between types of Darwinian
populationsdependsonvariousparametersofevolutionthatcanberoughlymeasured(amongwhicharethe
classicalrequirementsofevolutionbynaturalselectionsuchasvariation,heritability,anddifferentialfitness,as
wellasadditionalonesthatcanalsomodifyhowDarwinianapopulationis).Applyingthisframeworktocancers,
Germain(2012)highlightedthatcancercellpopulationsaremuchlesslikeparadigmaticDarwinianpopulations
thanusuallyassumedandthisisinlargepartduetotheCSCmodel,whichappearsasamajorfactorthatlimits
theabilityofcancercellstoevolvebynaturalselection(seesection1.1).Germain’sconclusionwasthepointof
Accepted for publication in Philosophy and Biology 25.04.2018
3
departureformywork.IfcancercellsaremuchlesslikeparadigmaticDarwinianpopulationsthanexpected,then
whatkindofpopulationsaretheyexactly?Inthispaper,IbuildaframeworksimilartoGodfrey-Smith’sinwhich
themeasureofsomefactorsassociatedwithCSCschangestheCEfeaturesofcancercellpopulations.Thetwo
factorsofinterestareCSCfrequency(howmanycancercellsareCSCs?)andnon-CSCdedifferentiation(whatis
theprobabilityofnon-CSCsbecomingCSCs?),whichcapturewhatIcall“CSCrestriction”.TheCEfeaturesunder
the influence of CSC restriction are processes of evolution (evolution by natural selection versus neutral
evolution,complexversussimpleadaptations)andpatternsofevolution(linearversusbranchedevolution).
ThefirstsectionshowsthatCSCrestrictionlimitsevolutionbynaturalselection(andmoregenerallyCE),prevents
complex adaptations, and favors drift and linear evolution. Scientists and philosophers have compared CSC
restriction to other phenomena well known in evolutionary biology such as reproductive specialization or
effectivepopulationsize,raisingthequestionofwhetherwecoulddescribetheimpactofCSCrestrictiononCE
throughclassicalparametersofevolution.Thesecondsectionshowsthatnoneoftheseparametersaccurately
accountforCSCrestriction,inparticularbecauseitreliesontwodifferentfactors(CSCfrequencyandprobability
ofdedifferentiation).Thelastsectionexploresthepotentialbiomedicalinterestsofinvestigatingtheimpactof
CSCrestrictiononpatternsofCE.
Figure1TheCEmodelandtheCSCmodel.TocomparetheCEmodelandtheCSCmodel,Iproposetwoclassicalrepresentationsofthemodels(aandb,toppanels)andtwomoreabstractedrepresentationsofthesemodels(a and b, bottom panels). The more abstracted representations are useful to represent graphically theconsequenceofCSCrestrictiononCEinthenextfigure
Accepted for publication in Philosophy and Biology 25.04.2018
4
1.CSCrestrictionimpactsCE
BeforeexploringtheimpactofCSCsonCEitisnecessarytohaveaclearunderstandingofCSCs,andinparticular
ofwhatIcallCSCrestriction.TheCSCmodelpicturescancersasacaricatureofthenormaltissuesfromwhich
theyemerge(forarecentdiscussion,seeBatlleandClevers2017).Inmostnormaltissuesinhomeostasis,there
is a hierarchical organizationof cells inwhich a small fractionof stem cells sit at the apexof a hierarchyof
differentiatingcells.Whenstemcellsdivide,theycangiverisetonewstemcells,thusensuringthemaintenance
ofthestemcellpool,and/ortheycandifferentiateintomorematurecells(Fig1b).Incancers,thisorganization
isaltered(differentiationcanbeblocked,biased,oramplified),yetinmanycases,ahierarchyismaintainedand
onlyafractionofthecancercells(theCSCs)isabletodevelopandmaintainthetumor.1
TheexplanatoryvalueoftheCSCmodelreliesontwopremises:firstthatnotallcancercellsareCSCs,andsecond,
thatCSCscanbedistinguishedfromnon-CSCs(i.e.theobservedfunctionaldifferencebetweenCSCsandnon-
CSCsisnotstochastic).ThemassivesearchtoidentifyCSCsinallcancershasproducedconflictingdataforboth
premises.First,estimatedCSCfrequencycanvarybyordersofmagnitude,fromveryrareinsomecancers(e.g.
50in106inacutemyeloidleukemia;BonnetandDick1997)toverycommoninothers(e.g.1outof4cellsin
advancedmelanoma;Quintanaetal.2008).SomemutationscanresultinanincreaseinCSCfrequency,indicating
thatCSCfrequencycanalsochangeduringcancerprogression(Nottaetal.2011;Clappieretal.2011).Thus,the
CSCfrequencyvariesdependingoncancertypesandstages,anditcanbemeasured,atleastroughly.2Second,
in some cancers, non-CSCs can dedifferentiate into CSCs (e.g. Chaffer et al. 2013; Koren et al. 2015; Van
Keymeulen et al. 2015;Medema 2017). Dedifferentiation affects the quantification of the cancer cells that
contribute tocancermaintenanceandevolution through timebecausewhendedifferentiation ispossible, in
additiontotheactualCSCs,thenon-CSCsrepresentapoolofpotentialCSCs.Howmanyofthesenon-CSCswill
becomeCSCsisavariablethatcanbemeasuredastheprobabilityofdedifferentiation.Dedifferentiationofnon-
CSCsintoCSCsisaprocessspecifictosomecancers(mostlyobservedinepithelialcancersandneverobservedin
bloodcancerssofar)andtheprobabilityofdedifferentiationofanon-CSCintoaCSCcanvarybetweencancers
(e.g.betweenbreastcancersfromdifferentcelloforigin,Latiletal.2017).
Iusetheterm‘CSCrestriction’torefertothesetwodynamicaspectsofCSCs(CSCfrequency,e.g.theiractual
number,andnon-CSCdedifferentiation,e.g.theprobabilitythatnon-CSCsdedifferentiateintoCSCs).Thelevel
of CSC restriction thus varies in different cancer types and/or stages. CSC restriction is high when the CSC
frequencyandtheprobabilityofdedifferentiationarelowandvice-versa.Thissectiondiscussestheimpactof
thelevelofCSCrestrictiononCE,basedonphilosophical,modeling,andbiologicalliterature.
1.1.CSCrestrictionlimitsevolutionbynaturalselectionandpreventscomplexadaptations
1ThecelloforigincaneitherbeastemcelltransformedintoaCSCoranon-stemcellthatreacquiredstemnessattransformationandbecameaCSC.Foradetailedaccountofthehistory,structureandcontentoftheCSCmodel,seeLaplane(2014,2016,chap2-4).2CSCquantificationreliesonmanyexperimentalparametersthatcanallintroducebiases(reviewedinKresoandDick2014;NassarandBlanpain2016).
Accepted for publication in Philosophy and Biology 25.04.2018
5
In theCEmodel, all the cancer cells can, inprinciple, evolve.Cancers arepopulationsof cells that asexually
reproducethroughcelldivision.Duringdivisiontheycanacquiremutationsthattheytransmittotheirdaughter
cells.Theimpactofmutationsoncellfitnessisvariablesothattwotypesofmutationsareusuallydistinguished
incancers:passengermutations,i.e.thosethatdonotchangethefitnessofthecells,anddrivermutations,i.e.
thosethatincreasetheirfitness(amutationcanbepassengerinonecontextanddriverinanother).Cancercells
thus fulfill the minimal requirements for evolution by natural selection; they are populations of cells that
reproducewithsomeheritablevariationinfitness(Lewontin1970;1978).TakingtheCSCmodel intoaccount
makes some difference in understanding evolution in cancer cells because it indicates an inequality in the
contributionofcancercellstoclonalevolution.TheCSCmodeldistinguishestwofunctionallydistinctpopulations
ofcells,onewithavirtuallyunlimitedproliferativeability(theCSCs),andtheotherwithalimitedproliferative
ability (the non-CSCs) (Fig. 1b). On short time scales (days to weeks) this functional distinction makes no
difference.Buton the timescaleof thedisease (months toyears), itmakesamajordifferenceasmutations
occurringinnon-CSCsarewashedawayfromthepopulationthroughexhaustionoftheproliferativeabilityof
thesecellsandultimatelycelldeath.Non-CSCsareevolutionarydead-endsandonlythemutationsoccurringin
CSCscanpersistovertimeandgeneratelong-standingsubclones(i.e.thevariouspopulationsofcellswiththe
samepoolofmutationsthatconstituteagivencancer)(Fig.2a-b).Theproportionofmutationsthatwillbeable
toparticipateinCEoverthecourseofthediseasedependsbothonthefrequencyofCSCsinthepopulation(Fig.
2c)andtheprobabilityofdedifferentiationofnon-CSCs(Fig.2d).ThisledGreavestoarguethatCSCsarethemain
unitsofselectionincancers(GreavesandMaley2012;Greaves2013,2015).Thequestionthenis:whatdifference
doesitmakeforCE?Germain(2012)offeredananswer:CSCrestrictionlimitsevolutionbynaturalselectionand
makescomplexadaptationsunlikely.
ThemainaimofGermain(2012)wastoassesstheroleofevolutionbynaturalselectionincancers.Toachieve
thisgoal,heusedGodfrey-Smith’s(2009)famousaccountofDarwinianpopulationsandappliedittocancers.
The importance ofGodfrey-Smith’s framework forGermainwas that it allows amore precise description of
differenttypesofDarwinianpopulationsasfunctionsofroughlymeasurableparameters;thetwoextremetypes
of Darwinian populations being described as paradigmatic Darwinian populations andmarginal Darwinian
populations.Evolutionbynaturalselectioncanoccurinbothbutplaysamuchgreaterroleintheformer.The
generalconclusionofGermainisthatalthoughcancercellsareDarwinianpopulations,theyarenotparadigmatic
populationsbecausethereareanumberofparametersforwhichtheyscoreratherlow(seeTable1inGermain
2012).My interesthere is in the impactofCSC restrictiononevolutionbynatural selection,aquestion that
GermainaddressedbyscoringcancercellpopulationswithandwithouttakingintoaccounttheCSCmodel.For
oneparameter,reproductivespecialization,takingtheCSCmodelintoaccountmakesalargedifferenceinhow
thecancercellsscore,andthereforeonhowDarwiniantheyare.
Accepted for publication in Philosophy and Biology 25.04.2018
6
Figure2CEwithandwithouttakingintoaccountCSCrestriction.(a)IntheclassicalCEmodel,whichdoesnottakeCSCsintoaccount,allcancercellsareconsideredtobeunitsofselection.(b)WhenCSCrestrictionistakenintoaccount,onlytheCSCsareconsideredasmeaningfulunitsofselectionthroughoutthecourseofthedisease,becausethemutationsthatoccurinnon-CSCsarewashedawayfromthepopulation.(c)CSCrestrictiondependson the frequency of CSCs and can vary from very low to very high, which differentially affects CE. (d) CSCrestrictionalsodependsontheprobabilityofdedifferentiationofnon-CSCsintoCSCs,asmutationsoccurringinnon-CSCs are not washed away if the mutated cells dedifferentiate into CSCs. High or low probability ofdedifferentiationwilldifferentiallyaffectCE
Reproductive specialization refers to the germ/soma distinction and related phenomena in which only a
sequesteredpartoftheindividualcangiverisetoanewindividualthroughsexualorasexualreproduction(the
germ cells), while the rest of the individual cannot (the somatic cells) (see Godfrey-Smith 2009, p. 92). In
multicellularorganisms,reproductivespecializationpartiallysuppressesevolutionbynaturalselectionatthecell
levelbymakingthesomaticcellsevolutionarydead-ends(seeGodfrey-Smith2009,p.101).Germainarguedthat
the coexistence of CSCs and non-CSCs in cancers produces a reproductive specialization comparable to the
germ/somadistinction:non-CSCsareunabletogrowatumor likesomaticcellsareunabletoproduceanew
Accepted for publication in Philosophy and Biology 25.04.2018
7
organism,whereasCSCsareable togrowa tumor likegermcells are capableofproducinganeworganism.
Greavesmadeasimilarcomparison:
Darwinianmodelsofcancernowneedtoadopttheconceptofcancerstemcells;thatsmallpopulationof
self-renewingcellsthatmaintainmostofthecancercellpopulationand,inself-replicating,providethe
essentialreservoirforfurthergeneticvariabilityandselection,equivalentinevolutionarytermstogerm
cells(Greaves2007,218).
GreavesandGermainagreeonthephenomenon:CSCrestrictionisamechanismbywhichanumberofmutations
arewashedawayfromthepopulation(allthosethatoccurinnon-CSCs,dedifferentiationeventsapart,seeFig.
2).Thisphenomenonalsoappliesinsomaticcellsofnormaltissueswheremutationsoccurringinnon-stemcells
areremovedfromthepopulationduetoreplicationlimitsofcells(againwiththeexceptionofdedifferentiation),
and it isnowaclassicargument thatstemcells limitcancer risk (Cairns1975;Pepperetal.2007).3Whether
reproductivespecializationaccuratelycapturesthisphenomenonisaquestionthatIwilladdresslater(section
2.2).TheimportantconclusionhereisthatthehighertheCSCrestriction (i.e.thelowertheCSCfrequencyandtheprobabilityofnon-CSCdedifferentiation),thelesscancercellsareparadigmaticDarwinianpopulations.
From this first conclusion, a second conclusion follows in Godfrey-Smith’s framework: the less paradigmatic
cancercellpopulationsare, the less likelycomplexadaptationsare tooccurbecauseparadigmaticDarwinian
populationsarecharacterizedasthoseinwhichnaturalselectioncanexplaintheoriginofnoveltraits(complex
adaptations) whereas natural selection only changes the distribution of already existing traits (simple
adaptations)inmarginalDarwinianpopulations(seeGodfrey-Smith2009,chapter3).Complexadaptationsrely
oncumulativeevolution:
Itcanseemoddtosaythatselection,whichhastodowithsortingthingsthatalreadyexist,cansomehow
bringnewthings intoexistence.Butnaturalselectioncanreshapeapopulation inawaythatmakesa
givenvariantmorelikelytobeproducedviatheimmediatesourcesofvariationthanitotherwisewould
be.Selectiondoesthisbymakingintermediatestagesontheroadtosomenewcharacteristiccommon
ratherthanrare,thusincreasingthenumberofwaysinwhichagivenmutationalevent(orsimilar)will
suffice toproduce the characteristic inquestion. Somekindsofnovelty canbeproducedeasilybyan
evolutionaryprocesswithoutthisroleforselection,butotherkinds—complexandadaptedstructures—
cannot(Godfrey-Smith2009,43).
CSCrestrictionimpedescomplexadaptationsbylimitingcumulativeevolutionintwoways(Germain2012).First,
CSCrestrictionlimitsthesizeofthecancercellpopulationinwhichevolutionarychangescanaccumulate(Fig.
2).Anycomplexadaptationsinnon-CSCsareveryunlikelytooccurgiventhatnon-CSCscanonlyproliferatefor
shortperiodsoftime,andthusmutationswouldbequicklylost.Second,CSCscanbequiescent(i.e.nondividing
forlongperiodsoftime),whichlimitsthepossibilityofcumulativeevolutionintheCSCpopulationitself.
3TomasettiandVogestein(2015)arguedthat“variationincancerriskamongtissuescanbeexplainedbythenumberofstemcelldivision,”whichstartedamassivedebateoncanceretiologyandonthecausalcontributionattributabletoenvironment,heredity,andstemcellreplication(seethelatestresponseofTomasettietal.2017).
Accepted for publication in Philosophy and Biology 25.04.2018
8
GermainconvincinglyarguedthattheCSCmodel,andmorepreciselywhatIcallthelevelofCSCrestriction,plays
animportantroleinwhatwecanexpectfromevolutionbynaturalselectionincancers.Inparticular,thehigher
theCSCrestriction,thelesslikelycomplexadaptationsaretoappear(Figure3).4Moregenerally,CSCrestriction
isawaytowashoutagreatamountofthemutationsthatoccurinagivencellpopulation.Thisnotonlylimits
evolutionbynaturalselectionbutlimitsCEingeneral.
Figure3AdaptationsdependonCSCrestriction.ThelowertheCSCfrequencyandthelowertheprobabilityofnon-CSCdedifferentiation,themoreunlikelycomplexadaptationsaretooccur.Theblackandwhitegradientindicatesdirections:thelowertheCSCfrequencyandtheprobabilityofdedifferentiationofnon-CSCsintoCSCs,the less likelycomplexadaptationsare.Theexactshapeofthegradientremainstobeempiricallyevaluated.Noticealsothatsomevaluesarebiologicallyabsurdorequivalent.Forexample,havingaCSCfrequencyof100%andaprobabilityofdedifferentiationof1isabitabsurd,becauseifthereareonlyCSCsthentherearenonon-CSCstodedifferentiate.Similarly,iftheprobabilityofdedifferentiationis1thenCSCfrequencycannotbe0%
1.2.CSCrestrictionfavorsdrift
Germain’sinvestigationfocusedonevolutionbynaturalselection.AsCEisnotrestrictedtoevolutionbynatural
selection,Iwouldliketocontinuehisinvestigationatalargerscale.Thissectionfocusesondrift.TheCEmodel
wasoriginallyframedwiththeideathatcancerisaDarwinianprocessofgradualevolutionbynaturalselection.5
Yet,sinceGermain’sarticlewaspublished,evidenceforneutralevolutionhasbeenreportedinhepatocellular
carcinoma(Lingetal.2015)andcolorectalcancers(Sottorivaetal.2015;Sottorivaetal.2017).Applyinganeutral
evolutionmathematicalmodeltothesequencingdataof904cancersfrom14types,Williamsetal.(2016)found
that the neutral model fitted with high precision 323 of them (more than a third), suggesting that neutral
evolutioncouldaffectmanymorecancersthananticipated.Inthesecancers,theyargue,mostofthemutations
responsibleforthecancerexpansionwerealreadypresentinthefirstmalignantcellandthefollowingevolution
wasneutral,aphenomenonsometimereferredtoasa“BigBang”modelofcancerprogression(Sottorivaetal.
4Ifocusonmutationsbecausetheyaretheusualtraitsthatareusedtoreconstructandstudytheevolutionofcancercellpopulations,buttheargumentwouldapplyforanyheritabletrait.5HereIfocusontheassumptionthatCEisaprocessofevolutionbynaturalselectionbecause,tothebestofmyknowledge,noevidencesuggeststhatCSCrestrictioncanfavororlimitpunctuatedorgradualevolution.ButtheassumptionthatCEisagradualprocesshasbeenchallengedbytheobservationofseveralcomplexmutationaleventsthatinducemultiplemutationsinaburst(e.g.Stephensetal.2011;Rauschetal.2012;Nik-Zainaletal.2012a;Bacaetal.2013).TheextentandlimitsoftheanalogywithEldredgeandGould’spunctuatedevolutionisdiscussed in Gao et al. (2016). Markowetz (2016) suggested that saltationist theories such as the hopefulmonsterstheoryofRichardGoldschmidtmightbemorerelevantincancers.
Accepted for publication in Philosophy and Biology 25.04.2018
9
2015;Amaroetal.2016).6Anotherpan-analysisalsohighlightedthatnegativeselection,whichispredominant
inthegermline,isnearlyabsentincancerandsomaticevolution(Martincorenaetal.2017).DoesCSCrestriction
impactwhether cancers followapatternofneutral evolution?AmodeldevelopedbySottorivaet al. (2010)
suggestsasmuch.Inthismodel,SottorivaandcolleaguesobservedthatCEwasimpactedintwowayswhenthey
factoredintheCSCmodel(i.e.bydistinguishingtwofunctionallydistinctpopulationsofcancercells,CSCsand
non-CSCs). First, as alreadydiscussed in section1.1,CSCs repressCEbymaking theacquisitionofmutations
slower. Second, they observed a different clonal expansion process.When the CSCmodel is not taken into
account,asmallnumberoftheoccurringmutationswereselectedforandinvadedlargeportionsofthecancer
cellpopulation.Incontrast,theimplementationoftheCSCmodelresultedinthepersistenceofalargeproportion
of mutations that remained at a low frequency in the population, consistent with neutral evolution. They
concluded“thattheintrinsicpropertiesoftheCSCmodelmightpropelanalternativeprocesstonaturalselection,
referredtoasgeneticdrift”(Sottorivaetal.2010,52),andattributedtheoccurrenceofdrifttosamplingerrors
thataremuchmorefrequent insmallpopulations—theCSCmodelstimulatessamplingerrormechanismsby
limitingtheproportionofcellsthatcanacquireandmaintainmutations.
TheconclusionofSottorivaetal.(2010)thatCSCrestrictionfavorsdriftraisesthequestionofwhetherthereis
higher CSC restriction in the cancers that fitted the neutralmodel inWilliams et al. (2016). To answer this
question,onewouldhavetoaligntheanalysisofneutralevolutionfromwholegenomesequencingdatawith
scoringsof CSC frequency andprobability of dedifferentiation in various cancer types andalso in cancers at
variousstagesofprogression.ThiswouldshowhowachangeinCSCrestrictioncanimpacttheprocessesofCE
(Fig.4).
Figure4CSCrestrictionanddrift.Theaxesinthisschematicdiagramarethesameasthepreviousdiagram(Fig.3).LowerCSCfrequencyandprobabilityofdedifferentiationnotonlymakecomplexadaptationsveryunlikely,they also favor drift. Empirical comparison between processes of evolution and level of CSC restriction arerequiredtocharacterizetheCEprocessesassociatedwitheachofthepointsofthistwo-dimensionalspace
1.3.CSCrestrictionfavorslinearevolution
6 Whether a first stage of evolution by natural selection precedes neutral evolution is a matter of debate(Martincorenaetal.2015;Simons2016).
Accepted for publication in Philosophy and Biology 25.04.2018
10
TheCEmodelwasfirstrepresentedasalinearmodelofevolutionwhereinmutationsaccumulatesequentially,
witheachnewcloneoutcompetingthepreviousone,sothatsubclonesaccumulatingnewmutationsproceed
one after another (Fig. 5a). In this pattern of CE, intratumoral heterogeneity (ITH) is diachronic rather than
synchronic, such that cells accumulate mutations through time but the mutations appear homogenous
throughoutthecellpopulationatanygiventime(i.e.atT1allcellshavemutation1,atT2allcellshavemutations
1and2,etc.)(e.g.Houetal.2012).Inmostcasesoflinearevolutionthough,selectivesweepsappearincomplete
andformersubclonespersist,butdonotgiverisetonewsubclones(Fig.5b)(e.g.Merlevedeetal.2016).When
theydogiverisetonewsubclones,theresultisabranchedpatternofevolution:mutationsrandomlyaccumulate
incellsofvarioussubclonescontinuouslyleadingtotheappearanceofnewsubclones(Fig.5c)(e.g.Yachidaetal.
2016),which is themost commoncase (Andoretal.2016). In this section Idevelop thehypothesis thatCSC
restrictionmightalsoimpactpatternsofevolution,andmorepreciselythatahigherlevelofCSCrestrictionmight
favorlinearevolution.
Figure5ThreepatternsofCE. (a)LinearhomogeneousCE ischaracterizedbythesequentialaccumulationofmutations with each new clone outcompeting the previous so that subclones accumulating newmutationssucceedoneanother.(b)Whenselectivesweepsareincompleteandformersubclonespersistwithoutgivingrisetonewsubclones,theCSislinearheterogeneous.(c)Whenformersubclonesaccumulatemutationsandgiverisetonewsubclones,CE isbranched:mutations randomlyaccumulate incellsofvarioussubclonescontinuouslyleadingtotheappearanceofnewsubclones
Withsimilarcelldivisionandmutationratesinafixedperiodoftime,asmallerpopulationwillaccumulatefewer
mutationsthanabiggerpopulation.Thisistruebydefinitionifmutationsoccurduringdivision.Whenamutation
doesoccur,allthingsbeingequal,itshouldalsohaveahigherprobabilityofinvadingthepopulation,asthereare
fewercellstooutcompete.Thereisthusaninversecorrelationbetweentheprobabilityofoccurrenceofnew
mutationsand theprobabilityofpropagationof themutations. Ifmutationsare less frequentandpropagate
moreeasilyinasmallerpopulation,thenthereisahigherchancethatthesecondmutationoccursontopofthe
Accepted for publication in Philosophy and Biology 25.04.2018
11
firstmutation(becausethefirstmutationispresentinalargerfractionofthepopulationatthetimeofoccurrence
ofthesecondmutation).Ifthesecondmutationdoesoccurontopofthefirst,andthethirdontopofthesecond,
thenwehavelinearevolution.Incontrast,inaverylargepopulation,mutationspropagatelesseasilyandoccur
morefrequently,leavinglesstimeforthepreviousmutationtopropagate,andreducingthechancethatanew
mutationoccursinacellbearingthepreviousone.Ifthenewmutationoccursinacellthatdoesnothavethe
previousmutation,thenwehavebranchingevolution.Thus,ceterisparibus,smallerpopulationsaremorelikely
toundergolinearevolutionandlargerpopulationsaremorelikelytoundergobranchingevolution.AsonlyCSCs
canparticipateinCEoverthecourseofthedisease,itisactuallytheCSCrestrictionthatmattersforpatternsof
CE.HigherCSC restriction should thus favor linearCE,whereas a loss of CSC restrictionby an increasedCSC
frequency and/or an increased probability of dedifferentiation should increase the probability of branching
evolution(Fig6a).
TherelationshipbetweenCSCrestrictionandpatternsofCEhasnotyetbeenstudied,butthereissomeindirect
evidenceinfavorofthishypothesis.First,tothebestofmyknowledge,allcasesoflinearevolutionthathave
beenreportedare inbloodcancers(supplementarytable1,grey lines),whicharecharacterizedbya lowCSC
frequencyandnodedifferentiation,thusahighCSCrestriction.TheonlyexceptioncomesfromShainetal.(2015)
who reported linear evolution in thepremalignant stages ofmelanocytic neoplasms, suggesting that in solid
tumors,theinitialpremalignantstagecouldfollowalinearpatternofevolutionbeforeexplodingintoabranched
patternofevolutionatinitiationofthemalignantstage(whichwouldbeinlinewiththefactthatcancersstart
fromthemutationofasinglecellandthuswithasmallpopulationofCSCs)(Fig6b).Cancersinwhichnon-CSCs
candedifferentiateandCSC frequency ishigher, suchasbreast cancers, coloncancers, advancedmelanoma,
brain cancers, osteosarcoma, and pancreatic cancers, follow branched patterns of evolution (supplementary
table1).Thesecondkindofevidencecomesfrompatternsofaccumulationofmutationsinnormaltissues(i.e.
withoutanyovertcancerbeingdiagnosed).Presenceofmutationsusuallyassociatedwithcancerswithoutovert
malignancieshavebeenwelldocumented in twotissues:skinof theeyelidandblood.Stemcell restriction is
muchlowerintheformertissuethaninthelatter(higherstemcellfrequencyanddedifferentiationhavebeen
reportedinskinepithelium).Interestingly,thepatternsofmutationsinthesetissuesareverydifferent.Inthe
eyelid,Martincorenaetal.(2015)observedahighrateofmutationsgivingrisetonumerousbutverysmallclones
(i.e.cloneswithsmallnumbersofcells).Incontrast,inthebloodveryfewmutationshavebeenobserved,but
thesegiverisetobiggerclones(Genoveseetal.2014;Jaiswaletal.2014;Xieetal.2014).
Toconcludethisfirstsection,CSCrestrictionlimitsevolutionbynaturalselection,andmoregenerallyCE,prevents
complexadaptations,andfavorsdriftandlinearevolutioninamannerthatdependsonCSCfrequencyandthe
probabilityofdedifferentiation(Fig3,4and6).ByignoringtheCSCmodelorbyconsideringitasanindependent
modelofcancer,theCEmodelmissestheseimpactsofCSCrestrictiononCEcharacteristics.
Accepted for publication in Philosophy and Biology 25.04.2018
12
Figure 6 Patterns of evolution and CSC restriction. (a) Lower CSC frequency and lower probability ofdedifferentiationfavorslinearevolution.(b)Duringdiseaseprogression,cellpopulationsmightmoveinthistwo-dimensionalspace.AlossofCSCrestrictionthroughanincreaseinCSCfrequencyand/orincreaseinprobabilityofdedifferentiationmightfavoratransitionfromalinearCEtoabranchedCE
2.TraditionalparametersofevolutiondonotaccuratelydescribetheconsequencesofCSCrestrictionforCE
ThefirstsectionshowedthattheCEmodelshouldtake intoaccounttheCSCmodel,andmorepreciselyCSC
restriction,becauseitcanaffectboththepatternsandtheprocessesofCE.ThisdistinctionbetweenCSCsand
non-CSCsandits impactonCEresemblesomephenomenaalreadywellknownanddescribedinevolutionary
biology.Scientistsandphilosophershavediscussedtheresemblancewithvariousparametersofevolutionsuch
asheredity(Greaves2013),reproductivespecialization(Germain2012;Greaves2013),andeffectivepopulation
size(e.g.Merloetal.2006;Pepperetal.2009;Sottorivaetal.2010;Lipinskietal.2016;LeanandPlutynski2016).
Thissectionpresentsanddiscussesthesecomparisonsaswellasadditionalones(bottleneckanddependenceof
fitnessonintrinsicproperties).Iarguethatnoneoftheseparametersareentirelysatisfactorytodescribethe
impactofCSCrestrictiononCEandshowthatthedifficultiescomefromthefactthatCSCrestrictiondependson
twofactors:theCSCfrequencyandtheprobabilityofdedifferentiationofnon-CSCsintoCSCs.
2.1.CSCrestrictionandheredity
ReferringtoLewontin’sclassicalminimalrequirementforevolutionbynaturalselection,Greaveshasarguedthat
itisthelackofhereditythatsuppressesCEinnon-CSCs,sothatonlyCSCsaregenuineunitsofselectionincancers
(e.g.Greaves2013,104).Thisclaimreliesonamisunderstandingoftheparameterofheredity.Heredity,asa
parameterofevolution,referstothefidelityofheredityateachgeneration.Sowhatisscoredinheredityisthe
“similaritybetweenparentandoffspring,duetocausalroleoftheparents”(Godfrey-Smith2009,39).Incancers,
the fidelity of heredity largely depends on the quantity ofmutations,which is highly variable. For example,
comparing30cancertypes,Alexandrovetal.(2013)reportedamutationalloadrangingfromabout0.001/Mbto
more than400/Mb.Whereasmelanomacanpresent thousandsofmutations, and thusavery low fidelityof
heredityduringcelldivision,veryfewmutations,ifany,werereportedinependymomas(apediatrictumorofthe
centralnervoussystem) indicatingahighfidelityofheredity (Macketal.2014).Thesedifferencesdependon
severalfactorssuchasgeneticinstability,orexposuretomutagenicagents.Forexample,thefidelityofheredity
Accepted for publication in Philosophy and Biology 25.04.2018
13
isfarlowerinlungcancersofsmokersthanofnon-smokers(Alexandrovetal.2016).CSCrestrictionplayslittle
role infidelityofheredity.WhatGreavesmeantby lackofhereditywasthatnon-CSCscanonlygiverisetoa
limitednumberofgenerationsofcellsandthatonlyCSCscanreproduceinthelong-term,afactweallagreeon
butstruggletodepictappropriately.
2.2.CSCrestriction,reproductivespecializationandbottleneck
ReferringtoGodfrey-Smith’snowfamousparametersofevolution,Germain(2012)claimedthatCSCrestriction
changesthedegreeofreproductivespecialization(seesection1.1).Iagreethatthereisananalogybetweengerm
linesequestrationinmulticellularorganismsandCSCrestrictionincancers.Thereis,inbothcases,adivisionof
labor thathasmassive impactson the long-term fitnessof the cells.But this analogyhas several limitations.
Reproductive specialization refers to thegerm/somadistinction,or equivalentphenomena. It relieson three
conditions. There is a reproductive specializationwhen: (1) there is reproduction; (2) only a part of the cell
populationisinvolvedinthisreproduction;and(3)thesecellsarethegermcells,whicharedistinctfromsomatic
cells,thatis,theybelongtoaseparatedcellline.
Whetherthereisreproductionwithincancersisadebatablematter.Metastasescouldbeacandidate(seeLean
andPlutynski2016;GermainandLaplane2017),butmyargumentisthatCSCrestrictionimpactspatternsand
processesofCEatthecellularlevel,regardlessoftheformationofmetastases.Onecouldthenarguethatcell
lineagesthatmakeupeachtumorare individualsthatreproduce. Inasense,eachCSCgivesrisetoa lineage
containingabodyofnon-CSCs.WhenaCSCself-renewsandproducesanewCSCthatgivesrisetoanewlineage,
thendowehaveacaseofreproduction?Whethereachlineagecancountasanindividualseemsdebatableto
megiventheabsenceofintegration.7But,ifweagreethatthereisagenuinecaseofreproductionhere,thenwe
alsomeetthesecondcriteriaforreproductivespecialization,namelythatonlyapartofthecellsconstitutingthe
individualareinvolvedinthereproduction,heretheCSCs.However,wethenfacealimitationintheanalogywith
regardtothethirdcriterion:CSCsandnon-CSCsarenottwoseparatedcelllineslikesomaticandgermcellsare.
Whilegermlinecellsaresequestered,CSCs(aswellasnormalstemcells)arenot,astheCSCsgiverisetothe
non-CSCs.ForCE,thismightbeconsideredasanegligibleproblemwhentheprobabilityofdedifferentiationof
non-CSCsintoCSCsisnull.Butitbecomesamajordifferenceassoonasnon-CSCscanbecomeCSCs.Inthiscase
thedistinctionbetweengermlineandsomacanno longerapply. If there isreproduction,then it isofavery
differentkind,onethatwouldinvolveabottleneck.
Abottleneckiswhenreproductionoccursfromonlyasmallpartoftheparentalorganism(seeGodfrey-Smith
2009,chapter5).Atfirstsight,bottleneckandreproductivespecializationmightseemverysimilar,asbothare
processesthatwashawaysomemutations,whichisexactlywhattheCSCvs.non-CSCdistinctiondoes.However,
theydifferintheirconsequencesforevolution:reproductivespecializationpreventsanymutationoccurringin
thesomaticcells/non-CSCsfrombeingtransmittedtothenextgeneration,whereasabottleneckdoesnotas
virtuallyanycellcanbetransmittedtothenextgeneration.Thus,incaseswherenon-CSCscandedifferentiate
7TherearesomefeedbackloopsbetweenmaturecellsandCSCs(e.g.Reynaudetal.2011),butthosearenotrestrictedtotheirownlineagesandinsomecases,likebloodcancers,thecellsfromdifferentlineagesareallmixedandmovingaround.
Accepted for publication in Philosophy and Biology 25.04.2018
14
intoCSCs,thequestionbecomeswhetherCSCrestriction(CSCfrequencyandprobabilityofdedifferentiation)
equates to the degree of bottleneckishness. If each CSC lineage counts as an individual, then the degree of
bottleneckishness isalwaysthesame:eachindividualstartswithonlyoneCSC,whatevertheprobability isof
dedifferentiation.IfthedegreeofbottleneckishnessisalwaysthesamewhiletheCSCrestrictioncanvary,then
itcannotdescribethedegreeofCSCrestriction.
Hence,thereproductivespecializationparametercouldonlycapturetheimpactofCSCrestrictiononCEincases
wheretheprobabilityofdedifferentiationisnull,withthelimitationsintheanalogythatCSCsandnon-CSCsare
notseparatelineagesandprovidedthatlineagesdoreproduce.OnecouldthenarguethatthefrequencyofCSCs
equates to the degree of reproductive specialization. But CSC restriction also depends on the probability of
dedifferentiation,andwhennon-CSCscandedifferentiateintoCSCs,thenthereisnoreproductivespecialization,
butratherabottleneck.However,asthedegreeofbottleneckishnessisstable(onecell,theCSC),itisunlikelyto
do a better job of capturing the impact of CSC restriction on CE, which changes with the probability of
dedifferentiationandCSCfrequency.8
2.3.CSCrestrictionanddependenceoffitnessonintrinsicproperties
Besidesbeingacaseofreproductivespecialization,Godfrey-Smith(2009)alsodescribesgermlinesequestration
asacaseoflowdependenceoffitnessonintrinsicproperties,whichquantifieshowmuchreproductivesuccess
dependsonintrinsicproperties(e.g.thegeneticcontentofthecells)versusextrinsicproperties(e.g.thelocation
inaparticularenvironment).Lowdependenceoffitnessonintrinsicpropertiesrepressesevolutionbynatural
selection,asinthecaseofgermlinesequestration,becausethedifferenceinthelong-termfitnessofthegerm
linecellscomparedtosomaticcellsisnotduetointrinsicproperties(thatcanbepositivelyselectedfor)butto
thefactthatthecellsendedupbeingpartofoneoranothergroupofcellsduringdevelopment.Ifreproductive
specializationdoesnotexactlycaptureCSCrestriction,candependenceof fitnesson intrinsicpropertiesdoa
betterjob?ThequestionhereiswhetherthedifferenceinfitnessbetweenCSCsandnon-CSCsisduetocontext
orintrinsicproperties.Inotherwords,isCSCidentityintrinsicorextrinsic?
Ifnon-CSCscandedifferentiateintoCSCs,itmeansthatundercertaincircumstancesinsomecancers,stemness
can be acquired. But the ability or not to dedifferentiate is not sufficient to answer our question, as the
microenvironmentcanplayvaryingrolesinbothcases.Inpreviouswork,Ihaveshownthattwoquestionsmust
bedistinguishedtounderstandtheidentityofstemcells,includingCSCs:(1)cannon-stemcells/non-CSCsacquire
stemnessanddedifferentiateintostemcells/CSCs?(2)Isstemnessacquisitionand/ormaintenancedependent
uponthemicroenvironment?Dependingontheanswerstothesequestions,whicharespecifictotissuesand
cancers,fourdifferentidentitiescanbedistinguishedforthepropertyofstemness(Laplane2016):
1. Categorical:stemnessispurelyintrinsic.
8Iamverygratefultobothanonymousreviewersforthethoroughdiscussionthathasleadtothisversionoftheargument.
Accepted for publication in Philosophy and Biology 25.04.2018
15
2. Dispositional:stemnessisanintrinsicpropertywhoseexpressiondependsonextrinsicstimulifromthe
microenvironment.
3. Relational:stemnessisanextrinsicpropertyinducedandcontrolledbythemicroenvironment.
4. Systemic:stemnessisanextrinsicpropertycontrolledatthecellpopulationlevel.
Inthelastthreecases,thefitnessoftheCSCsdepends,tovariousdegrees,oncontext(whichcanincludethe
tumor-microenvironment,clone-cloneinteractions,andfeedbackloopsbetweenmaturecellsandCSCs).Incases
wherestemnessisarelationalproperty,beingaCSCdependsonthelocationofthecancercells.Dependenceof
fitnessonintrinsicproperties,inthisinstance,isverylow.Itishigherincaseswherestemnessisadispositional
propertybecausebeing aCSCdependson intrinsicproperties, although it is still limitedby the fact that the
reproductivesuccessofeachCSCisregulatedbythemicroenvironment.Incaseswherestemnessisasystemic
property,beingaCSCdependsonvariousfactorsincludingintrinsicpropertiessuchasstochasticgeneexpression
andextrinsicpropertiessuchasfeedbackloopsandothercollectiveprocesses;here,dependenceoffitnesson
intrinsicpropertieshangsupon the relativecontributionof intrinsicandextrinsic factors in stemness.Finally,
dependenceoffitnessonintrinsicpropertiesishighincaseswherestemnessisacategoricalpropertybecause
thedifference in fitness in the long term ismainlydue to intrinsicproperties. Thus, the identityofCSCs can
provideinterestinginsightsforCE,forexample,cancersinwhichstemnessisacategoricalpropertyshouldbe
morelikelytoevolvebynaturalselectionthancancersinwhichstemnessisarelationalproperty.But,thedegree
ofdependenceoffitnessonintrinsicpropertiesdoesnotcorrelatewiththedegreeofCSCrestriction.
2.4.CSCrestrictionandeffectivepopulationsize
SomebiologistsandphilosophershavesuggestedthatCSCrestrictionsimplylimitstheeffectivepopulationsize,
asopposedtothecensuspopulationsize(e.g.Merloetal.2006;Pepperetal.2009;Sottorivaetal.2010;Lipinski
etal.2016;LeanandPlutynski2016).
Infact,thequestionofwhethertheentireneoplasmoraminorityofneoplasticcellsiscapableof
self-renewalis,atleastinpart,aquestionmerelyoftheeffectivepopulationsizeoftheevolving
cellsinaneoplasm(Pepperetal.2009,66).
TheclaimhereisthattheeffectivepopulationsizeisthesizeoftheCSCpopulation.ThelowertheCSCfrequency
is,thegreaterthedifferenceisbetweentheeffectivepopulationsizeandthecensuspopulationsize,andthe
smallertheeffectivepopulationsizeis.Reducingtheeffectivepopulationsizeimpactsevolutionaryprocesses:
smallerpopulationsareconsideredmorepronetodrift,assamplingerrorcanmoreeasilyleadtotheexpansion
ofcloneswithnoclearsurvivalbenefit (seesection1.2).9Moregenerally, thearguments thatCSCrestriction
represses CE, favors drift, prevents complex adaptations, and favors linear evolution, could be reduced to a
question of effective population size, if the probability of dedifferentiation is null. But if non-CSCs can
dedifferentiate into CSCs at any time, then the actual CSC frequency poorly represents the cells that can
9The idea thatdriftdominates selection in smallpopulationshasbeendebated, see forexample the recentdebatebetweenClatterbuck,SoberandLewontin(2013)andBrandonandFleming(2014).
Accepted for publication in Philosophy and Biology 25.04.2018
16
meaningfullycontributetoCE,andthusalsopoorlyrepresentstheeffectivepopulationsize.Theprobabilityof
dedifferentiationofnon-CSCsintoCSCsisthenoncemoreamajorfactorforCE.
Takentogether,thesedifficultieshighlightthatCSCrestrictionisnoteasilycapturedbyanyclassicalparameter
ofevolution,inparticularbecauseofthedifferencebetweenCSCfrequencyanddedifferentiation,aswellasthe
fourdifferentpossibleidentitiesofCSCs(seesection2.3).ToaccountfortheimpactofCSCrestrictiononCE,one
needstomeasureboththefrequencyofCSCsandtheprobabilityofdedifferentiationofnon-CSCs.
3.BiomedicalrelevanceoftheimpactofCSCrestrictiononCE
ThefirstsectionshowedthatthedegreeofCSCrestrictioncouldfavororimpedevariousprocessesandpatterns
of evolution, and in particular that it can favor linear or branched evolution. Scientists regularly investigate
whetherCEislinearorbranchedbutgivelimitedattentiontothedifferenceitmakes.Whyshouldwecareabout
the impact of CSC restriction on linear or branched evolution? In this section, I argue that the relationship
betweenCSC restrictionandpatternsofCE isworthexploring for thebiomedical sciences, first,because the
patternsofCEcaninformusaboutCSCrestriction,andinparticularabouttheprobabilityofdedifferentiation
that is currently difficult to measure; second, because it offers original opportunities for new therapeutic
strategies.
3.1.FromCEpatternstoCSCrestriction
CEpatternsaremainlyusedtoinfereithertherelativefitnessofeachsubcloneinagivenenvironmentortheir
age.Incasesofevolutionbynaturalselectionthebiggersubclonesaretheoneswiththegreatestreproductive
success inthatenvironment. Incasesofneutralevolution,thebiggersubclonesaretheoldest. Inadditionto
theseinferences,IhavearguedinthispaperthatthepatternofCE—linearorbranched—is,atleastinpart,a
functionofCSCrestriction.IfIamright,thenreconstructionoftheCEcouldbeinformativeaboutCSCrestriction.
This couldbevaluable incancers inwhichoneof the twoparametersofCSC restriction—CSC frequencyand
probabilityofdedifferentiation—isunknown.Forexample,inasetofcancersofagiventype(e.g.melanomas),
ifwecanmeasuretheCSCfrequencyandreconstructtheCEtree,wemaybeabletoinferifdedifferentiation
occurredornotand,ifitoccurred,whetheritisamarginalorfrequentprocess(i.e.whethertheprobabilityof
dedifferentiation is high or low). This is far from trivial given that the current technologies allow far better
approximationsofCEpatternsandofCSCfrequenciesthanoftheprobabilityofdedifferentiation.
3.2.FromCSCrestrictiontoCEpatterns:therapeuticconsequences
What difference does it make if CE is linear or branched? Is one or the other better for the patients? The
immediate relevanceofdistinguishingCEpatterns is that thecoexistenceofmultiple subclones isacauseof
therapyresistance.Themoresubclonesthereare,thegreatertheriskoftherapyresistance.Ifmyhypothesis
thatahigherCSCrestrictionshouldfavoramorelinearandsimplerpatternofCEisright,thenpatientswould
benefit from any complementary therapies that can increase CSC restriction. Opportunities to control CSC
restrictionareemerging.ThemostrecentexamplesareCimminoetal.(2017)andAgathocleousetal.(2017),
whoshowedthatvitaminCregulatesthefrequencyofnormalormalignanthematopoieticstemcells.
Accepted for publication in Philosophy and Biology 25.04.2018
17
Conversely,anytherapythatwoulddecreaseCSCrestrictionshouldbeavoided.Hypomethylatingagents(HMAs)
areusedasfirstlinetreatmentinchronicmyelomonocyticleukemia(CMML),myelodysplasticsyndromes(MDS),
andacutemyeloidleukemia(AML)oftheelderly,whenpatientsarenoteligibleforstemcelltransplantation.
ButHMAshave alsobeenused in a verydifferent context, unrelated to cancer, to induce cell plasticity and
dedifferentiation(e.g.Mikkelsenetal.2008;Chandrakanthanetal.2016).SomestudiesalsosuggestthatHMAs
candecreasestemcellrestrictioninthenormalhematopoieticsystem,eitherbyincreasingtheirfrequencyorby
inducingdedifferentiation (Suzukietal.2004;Chungetal.2009). If theyalsodecreaseCSC restriction in the
patients(aquestionIamcurrentlyexperimentallyinvestigatinginmylab),thenthepredictionofthisstudyis
thatHMAtreatmentmustfavormoreCE,moreadaptability,andmorecomplexpatternsofevolution.Thiscould
helptoexplainthetransiencyofpatients’responsestoHMAsandthefrequentprogressionofCMMLandMDS
intoAML.IfprogressionundertreatmentwereduetothelossofCSCrestriction,thenonewaytomaintainthe
benefit of the drug without its counterproductive effects would be to couple it with a drug increasing CSC
restriction.Manydrugs(nonetargetingCSCs)arecurrentlytestedincombinationwithHMAswithoutyetleading
toimprovementinoverallsurvival(Balletal.2017).Inthiscurrentclinicalcontext,myworksuggeststhatdrugs
actingonCSCrestrictionareworthinvestigatingtoo.
Conclusionandperspectives
WhiletheinclusionofCEintotheCSCmodelhasbeenfairlywelladdressedintheliterature(KresoandDick2014;
NassarandBlanpain2016),showinginparticularthatCSCsareaheterogeneoustargetforanti-CSCtherapies,
the impactof theCSCmodel onCEhasbeenpoorly explored. This article shows thatCSC restriction,which
dependsonCSC frequencyandtheprobabilityofnon-CSCdedifferentiation, impactsCE inseveralways.The
highertheCSCrestriction,themoreitcan:
• limitCE;
• limitevolutionbynaturalselection;
• preventcomplexadaptations;
• favordrift;
• favorlinearevolution.
A lossofCSC restriction (higherCSC frequencyand/orhigherprobabilityofdedifferentiation)duringdisease
progressionwillthusfavormoreadaptability,makingtheinclusionofCSCrestrictionintotheCEmodelhighly
relevantforbiomedicalresearch.
Incorporating CSC restriction into the CE model faces several difficulties, opening perspectives for further
philosophical and biological research. First,within current experimental limitations, it is difficult to properly
quantifythenumberofCSCsandtheprobabilityofdedifferentiationwithinacanceratanygiventime.ThusCSC
restriction canonlybe roughlyestimated.Current improvements in lineage tracing technologieswill provide
greathelp.Estimationoftheprobabilityofdedifferentiationofnon-CSCsintoCSCsisalsobecominganurgent
needthatremainstoopoorlyaddressed.
Accepted for publication in Philosophy and Biology 25.04.2018
18
Second,recentworkhasshownthatunderhomeostaticconditions,progenitorcells(thefirstdaughtercellsof
stemcells),cansupporttheproductionofallthebloodcellsformuchlongerthaninitiallythought(Sunetal.
2014;Buschetal.2015).Thisraisesanewquestionforcancers:areCSCstheonlymeaningfulunitsofselection
inCE?Dependingonthedurationofthedisease,progenitorcellsmightalsobetakenintoaccountaspartofCSC
restriction,whichwouldchangethemeasureofCSCrestriction.
Third,theanalysispresentedinthisarticleismadeceterisparibus.Mypointisthateveryotherparameterbeing
equal,adifferenceinCSCrestrictionwillaffectCE.Afurtherquestionwouldbewhetherotherparameters,like
therateofmutationandofproliferation,aswellastheroleofthemicroenvironmentandmoregenerallythe
physicalconstraintsofthetissuesoutweighsomeoftheimpactofCSCsonthepatternsandprocessesofCE.
Finally,thisarticlefocusesontheimpactofCSCrestrictiononCE.Ifcorrect,thenthenextquestionishowcan
we act on CSC restriction in order to limit CE and disease progression? To answer this question requires
investigatingthetwofactorsofCSCrestriction:CSCfrequencyandnon-CSCdedifferentiation.Whatcausesan
increase inthefrequencyofCSCswithinagivencancer,andhowcanthis increase inthenumberofCSCsbe
avoidedorreduced?Whatinducesdedifferentiationofnon-CSCsintoCSCsandhowcanwepreventit?These
questions require a thorough investigation of the role of the microenvironment and of population-level
processes.ThemicroenvironmentcanplayadecisiveroleintheproliferationofCSCsandtheirabilitytogenerate
aclone(Arranzetal.2014;Dongetal.2016).Maturecancercells(non-CSCs)canalsoplayanimportantrolein
theproliferationoftheCSCsthroughthecytokinestheysecrete(Reynaudetal.2011;Welneretal.2015),raising
the question of interactions between subpopulations, andmore generally indicating the need to investigate
multilevelselectionprocesses(Sprouffskeetal.2013;LeanandPlutynski2016).
References
AdamsJM,StrasserA(2008)Istumorgrowthsustainedbyrarecancerstemcellsordominantclones?Cancer
Res68:4018–4021.doi:68/11/4018[pii]10.1158/0008-5472.CAN-07-6334
AgathocleousM,MeachamCE,BurgessRJ,etal(2017)Ascorbateregulateshaematopoieticstemcellfunction
andleukaemogenesis.Nature549:476–481.doi:10.1038/nature23876
AlexandrovLB, JuYS,HaaseK,etal (2016)Mutationalsignaturesassociatedwithtobaccosmoking inhuman
cancer.Science354:618–622.doi:10.1126/science.aag0299
AlexandrovLB,Nik-ZainalS,WedgeDC,etal(2013)Signaturesofmutationalprocessesinhumancancer.Nature
500:415–21.doi:10.1038/nature12477
AmaroA,ChiaraS,PfefferU(2016)Molecularevolutionofcolorectalcancer:frommultistepcarcinogenesisto
thebigbang.CancerMetastasisRev35:63–74.doi:10.1007/s10555-016-9606-4
AndersonK,LutzC,vanDelftFW,etal(2011)Geneticvariegationofclonalarchitectureandpropagatingcellsin
leukaemia.Nature469:356–361.doi:nature09650[pii]10.1038/nature09650
AndorN,GrahamTA,JansenM,etal(2016)Pan-canceranalysisoftheextentandconsequencesofintratumor
heterogeneity.NatMed22:105–113.doi:10.1038/nm.3984
Arranz L, Sánchez-Aguilera A,Martín-Pérez D, et al (2014) Neuropathy of haematopoietic stem cell niche is
Accepted for publication in Philosophy and Biology 25.04.2018
19
essentialformyeloproliferativeneoplasms.Nature512:78–81.doi:10.1038/nature13383
BacaSC,PrandiD,LawrenceMS,etal(2013)Punctuatedevolutionofprostatecancergenomes.Cell153:666–
77.doi:10.1016/j.cell.2013.03.021
BallB, ZeidanA,GoreSD,PrebetT (2017)Hypomethylatingagent combination strategies inmyelodysplastic
syndromes:hopesandshortcomings.Leuk.Lymphoma58:1022–1036
BapatSA (2007)Evolutionof cancer stemcells. SeminCancerBiol17:204–213 .doi: S1044-579X(06)00039-3
[pii]10.1016/j.semcancer.2006.05.001
BatlleE,CleversH(2017)Cancerstemcellsrevisited.NatMed23:1124–1134.doi:10.1038/nm.4409
BolliN,Avet-LoiseauH,WedgeDC,etal(2014)Heterogeneityofgenomicevolutionandmutationalprofilesin
multiplemyeloma.NatCommun5:.doi:10.1038/ncomms3997
Bonnet D, Dick JE (1997) Human acutemyeloid leukemia is organized as a hierarchy that originates from a
primitivehematopoieticcell.NatMed3:730–737.doi:10.1038/nm0797-730
Brabletz T, JungA, Spaderna S, et al (2005)Opinion:migrating cancer stem cells - an integrated concept of
malignanttumourprogression.NatRevCancer5:744–749.doi:nrc1694[pii]10.1038/nrc1694
BrandonR,FlemingL(2014)Driftsometimesdominatesselection,andviceversa:areplytoClatterbuck,Sober
andLewontin.BiolPhilos29:577–585.doi:10.1007/s10539-014-9437-z
BrocksD,AssenovY,MinnerS,etal(2014)IntratumorDNAmethylationheterogeneityreflectsclonalevolution
inaggressiveprostatecancer.CellRep8:798–806.doi:10.1016/j.celrep.2014.06.053
BuschK,KlapprothK,BarileM,etal(2015)Fundamentalpropertiesofunperturbedhaematopoiesisfromstem
cellsinvivo.Nature518:542–546.doi:10.1038/nature14242
CairnsJ(1975)Mutationselectionandthenaturalhistoryofcancer.Nature255:197–200
Campbell LL, Polyak K (2007) Breast tumor heterogeneity: cancer stem cells or clonal evolution? Cell Cycle
6:2332–2338.doi:4914[pii]
CaoW,WuW, YanM, et al (2015)Multiple region whole-exome sequencing reveals dramatically evolving
intratumor genomic heterogeneity in esophageal squamous cell carcinoma. Oncogenesis 4:e175 . doi:
10.1038/oncsis.2015.34
CayeA,StrulluM,GuidezF,etal(2015)Juvenilemyelomonocyticleukemiadisplaysmutationsincomponentsof
theRASpathwayandthePRC2network.NatGenet47:1334–1340.doi:10.1038/ng.3420
ChafferCL,MarjanovicND,LeeT,etal(2013)PoisedChromatinattheZEB1PromoterEnablesBreastCancerCell
PlasticityandEnhancesTumorigenicity.Cell154:61–74.doi:http://dx.doi.org/10.1016/j.cell.2013.06.005
ChandrakanthanV,YeolaA,KwanJC,etal(2016)PDGF-ABand5-Azacytidineinduceconversionofsomaticcells
into tissue-regenerative multipotent stem cells. Proc Natl Acad Sci U S A 113:E2306-15 . doi:
10.1073/pnas.1518244113
Chesnais V, Renneville A, Toma A, et al (2016) Effect of lenalidomide treatment on clonal architecture of
myelodysplasticsyndromeswithout5qdeletion.Blood127:749–60.doi:10.1182/blood-2015-04-640128
ChungYS,KimHJ,KimT-M,etal(2009)Undifferentiatedhematopoieticcellsarecharacterizedbyagenome-
wideundermethylationdiparoundthetranscriptionstartsiteandahierarchicalepigeneticplasticity.Blood
114:4968–78.doi:10.1182/blood-2009-01-197780
Accepted for publication in Philosophy and Biology 25.04.2018
20
Cimmino L,Dolgalev I,Wang Y, et al (2017) Restorationof TET2 FunctionBlocksAberrant Self-Renewal and
LeukemiaProgression.Cell170:1079–1095.e20.doi:10.1016/j.cell.2017.07.032
Clappier E,GerbyB, Sigaux F, et al (2011) Clonal selection in xenograftedhuman T cell acute lymphoblastic
leukemiarecapitulatesgainofmalignancyatrelapse.JExpMed208:653–661.doi:10.1084/jem.20110105
ClatterbuckH,SoberE,LewontinR(2013)Selectionneverdominatesdrift(norviceversa).BiolPhilos28:577–
592.doi:10.1007/s10539-013-9374-2
deBruinEC,McGranahanN,MitterR,etal(2014)Spatialandtemporaldiversityingenomicinstabilityprocesses
defineslungcancerevolution.Science(80-)346:251–256.doi:10.1126/science.1253462
DingL,LeyTJ,LarsonDE,etal(2012)Clonalevolutioninrelapsedacutemyeloidleukaemiarevealedbywhole-
genomesequencing.Nature481:506–10.doi:10.1038/nature10738
DongL,YuW-M,ZhengH,etal(2016)LeukaemogeniceffectsofPtpn11activatingmutationsinthestemcell
microenvironment.Nature539:304–308.doi:10.1038/nature20131
FabianA,BarokM,VerebG,SzollosiJ(2009)Diehard:arecancerstemcellstheBruceWillisesoftumorbiology?
CytomA75:67–74.doi:10.1002/cyto.a.20690
FrancisJM,ZhangC-Z,MaireCL,etal(2014)EGFRvariantheterogeneityinglioblastomaresolvedthroughsingle-
nucleussequencing.CancerDiscov4:956–71.doi:10.1158/2159-8290.CD-13-0879
GaoR,DavisA,McDonaldTO,etal(2016)Punctuatedcopynumberevolutionandclonalstasisintriple-negative
breastcancer.NatGenet48:1119–30.doi:10.1038/ng.3641
GenoveseG,KahlerAK,HandsakerRE,etal (2014)Clonalhematopoiesisandblood-cancerrisk inferredfrom
bloodDNAsequence.NEnglJMed371:2477–2487.doi:10.1056/NEJMoa1409405
Gerlinger M, Horswell S, Larkin J, et al (2014) Genomic architecture and evolution of clear cell renal cell
carcinomasdefinedbymultiregionsequencing.NatGenet46:225–33.doi:10.1038/ng.2891
GerlingerM,RowanAJ,HorswellS,etal(2012)Intratumorheterogeneityandbranchedevolutionrevealedby
multiregionsequencing.NEnglJMed366:883–92.doi:10.1056/NEJMoa1113205
GermainP-L(2012)Cancercellsandadaptiveexplanations.BiolPhilos27:785–810.doi:10.1007/s10539-012-
9334-2
GermainPL,LaplaneL(2017)Metastasisassupra-cellularselection?AreplytoLeanandPlutynski.BiolPhilos
32:281–287.doi:10.1007/s10539-016-9555-x
Godfrey-SmithP(2009)DarwinianPopulationsandNaturalSelection.OxfordUniversityPress,Oxford
GreavesM(2010)Cancerstemcells:backtoDarwin?SeminCancerBiol20:65–70.doi:S1044-579X(10)00010-6
[pii]10.1016/j.semcancer.2010.03.002
GreavesM(2013)Cancerstemcellsas“unitsofselection.”EvolAppl6:102–108.doi:10.1111/eva.12017
GreavesM(2007)Darwinianmedicine:acaseforcancer.NatRevCancer7:213–221.doi:10.1038/nrc2071
GreavesM(2015)Evolutionarydeterminantsofcancer.CancerDiscov5:806–20.doi:10.1158/2159-8290.CD-
15-0439
Greaves M, Maley CC (2012) Clonal evolution in cancer. Nature 481:306–313 . doi: nature10762 [pii]
10.1038/nature10762
HirschP,MamezAC,BelhocineR,etal(2016)Clonalhistoryofacordblooddonorcellleukemiawithprenatal
Accepted for publication in Philosophy and Biology 25.04.2018
21
somaticJAK2V617Fmutation.Leukemia30:1756–1759.doi:10.1038/leu.2016.31
HouY,SongL,ZhuP,etal(2012)Single-cellexomesequencingandmonoclonalevolutionofaJAK2-negative
myeloproliferativeneoplasm.Cell148:873–85.doi:10.1016/j.cell.2012.02.028
HughesAEO,MagriniV,DemeterR,etal(2014)Clonalarchitectureofsecondaryacutemyeloidleukemiadefined
bysingle-cellsequencing.PLoSGenet10:e1004462.doi:10.1371/journal.pgen.1004462
ItzyksonR, KosmiderO,RennevilleA, et al (2013)Clonal architectureof chronicmyelomonocytic leukemias.
Blood121:2186–2198.doi:10.1182/blood-2012-06-440347
Jaiswal S, Fontanillas P, Flannick J, et al (2014) Age-related clonal hematopoiesis associated with adverse
outcomes.NEnglJMed371:2488–2498.doi:10.1056/NEJMoa1408617
JanM,SnyderTM,Corces-ZimmermanMR,etal(2012)Clonalevolutionofpreleukemichematopoieticstemcells
precedes human acute myeloid leukemia. Sci Transl Med 4:149ra118 . doi: 4/149/149ra118 [pii]
10.1126/scitranslmed.3004315
KimTH,YoshidaK,KimYK,etal (2015)Clonaldynamics inasingleAMLcasetracked for9years reveals the
complexityofleukemiaprogression.Leukemia30:295–302.doi:10.1038/leu.2015.264
Koren S, Reavie L, Couto JP, et al (2015) PIK3CA(H1047R) inducesmultipotency andmulti-lineagemammary
tumours.Nature525:114–118.doi:10.1038/nature14669
Kreso A, Dick JE (2014) Evolution of the cancer stem cell model. Cell Stem Cell 14:275–91 . doi: S1934-
5909(14)00057-5[pii]10.1016/j.stem.2014.02.006
LagasseE(2008)Cancerstemcellswithgeneticinstability:thebestvehiclewiththebestengineforcancer.Gene
Ther15:136–142.doi:3303068[pii]10.1038/sj.gt.3303068
LandauDA,CarterSL,StojanovP,etal(2013)Evolutionandimpactofsubclonalmutationsinchroniclymphocytic
leukemia.Cell152:714–26.doi:10.1016/j.cell.2013.01.019
LangF,WojcikB,RiegerMA(2015)StemCellHierarchyandClonalEvolutioninAcuteLymphoblasticLeukemia.
StemCellsInt2015:137164.doi:10.1155/2015/137164
LaplaneL(2015)ReprogrammingandStemness.PerspectBiolMed58:229–46.doi:10.1353/pbm.2015.0022
LaplaneL(2016)CancerStemCells:PhilosophyandTherapies.HarvardUniverityPress,Cambridge,MA
LaplaneL(2014)Identifyingtheoriesindevelopmentalbiology.Thecaseofthecancerstemcelltheory.In:Minelli
A,PradeuT(eds)Towardatheoryofdevelopment.OxfordUniversityPress,Oxford,pp246–259
LatilM,NassarD,BeckB,etal (2017)Cell-Type-SpecificChromatinStatesDifferentiallyPrimeSquamousCell
CarcinomaTumor-InitiatingCellsforEpithelialtoMesenchymalTransition.CellStemCell20:191–204.e5.
doi:10.1016/j.stem.2016.10.018
LeanC, PlutynskiA (2016) Theevolutionof failure: explaining cancer as anevolutionaryprocess. Biol Philos
31:39–57.doi:10.1007/s10539-015-9511-1
Lewontin RC (1970) The Units of Selection. Annu Rev Ecol Syst 1:1–18 . doi:
10.1146/annurev.es.01.110170.000245
LewontinRC(1978)Adaptation.SciAm239:212–8,220,222passim
LingS,HuZ,YangZ,etal(2015)Extremelyhighgeneticdiversityinasingletumorpointstoprevalenceofnon-
Darwiniancellevolution.ProcNatlAcadSci112:E6496-6505.doi:10.1073/pnas.1519556112
Accepted for publication in Philosophy and Biology 25.04.2018
22
LipinskiKA,BarberLJ,DaviesMN,etal(2016)CancerEvolutionandtheLimitsofPredictabilityinPrecisionCancer
Medicine.Trendsincancer2:49–63.doi:10.1016/j.trecan.2015.11.003
Mack SC,WittH, Piro RM, et al (2014) Epigenomic alterations define lethal CIMP-positive ependymomas of
infancy.Nature506:445–50.doi:10.1038/nature13108
MarkowetzF(2016)Asaltationisttheoryofcancerevolution.NatGenet48:1102–3.doi:10.1038/ng.3687
MartincorenaI,RaineKM,GerstungM,etal(2017)UniversalPatternsofSelectioninCancerandSomaticTissues.
Cell171:1029–1041.e21
Martincorena I,RoshanA,GerstungM,etal (2015)Highburdenandpervasivepositive selectionof somatic
mutationsinnormalhumanskin.Science(80-)348:880–886.doi:10.1126/science.aaa6806
Medema JP (2017) Targeting the Colorectal Cancer Stem Cell. N Engl J Med 377:888–890 . doi:
10.1056/NEJMcibr1706541
MelchorL,BrioliA,WardellCP,etal(2014)Single-cellgeneticanalysisrevealsthecompositionofinitiatingclones
andphylogeneticpatternsofbranchingandparallelevolution inmyeloma.Leukemia28:1705–15 .doi:
10.1038/leu.2014.13
MerlevedeJ,DroinN,QinT,etal(2016)Mutationalleleburdenremainsunchangedinchronicmyelomonocytic
leukaemiarespondingtohypomethylatingagents.NatCommun7:10767.doi:10.1038/ncomms10767
MerloLM,PepperJW,ReidBJ,MaleyCC(2006)Cancerasanevolutionaryandecologicalprocess.NatRevCancer
6:924–935.doi:nrc2013[pii]10.1038/nrc2013
MeyerM,ReimandJ,LanX,etal(2015)Singlecell-derivedclonalanalysisofhumanglioblastomalinksfunctional
andgenomicheterogeneity.ProcNatlAcadSciUSA112:851–6.doi:10.1073/pnas.1320611111
Mikkelsen TS, Hanna J, Zhang X, et al (2008) Dissecting direct reprogramming through integrative genomic
analysis.Nature454:49–55.doi:10.1038/nature07056
MurugaesuN,WilsonGA,BirkbakNJ,etal(2015)Trackingthegenomicevolutionofesophagealadenocarcinoma
throughneoadjuvantchemotherapy.CancerDiscov5:821–31.doi:10.1158/2159-8290.CD-15-0412
NassarD,BlanpainC(2016)CancerStemCells:BasicConceptsandTherapeuticImplications.AnnuRevPathol
11:47–76.doi:10.1146/annurev-pathol-012615-044438
NavinN,KendallJ,TrogeJ,etal(2011)Tumourevolutioninferredbysingle-cellsequencing.Nature472:90–4.
doi:10.1038/nature09807
Nik-ZainalS,AlexandrovLB,WedgeDC,etal(2012a)Mutationalprocessesmoldingthegenomesof21breast
cancers.Cell149:979–93.doi:10.1016/j.cell.2012.04.024
Nik-ZainalS,VanLooP,WedgeDC,etal(2012b)Thelifehistoryof21breastcancers.Cell149:994–1007.doi:
10.1016/j.cell.2012.04.023
NottaF,MullighanCG,WangJC,etal(2011)EvolutionofhumanBCR-ABL1lymphoblasticleukaemia-initiating
cells.Nature469:362–367.doi:nature09733[pii]10.1038/nature09733
OdouxC,FohrerH,HoppoT,etal(2008)Astochasticmodelforcancerstemcellorigininmetastaticcoloncancer.
CancerRes68:6932–6941.doi:68/17/6932[pii]10.1158/0008-5472.CAN-07-5779
OrtmannCA,KentDG,NangaliaJ,etal(2015)Effectofmutationorderonmyeloproliferativeneoplasms.NEngl
JMed372:601–12.doi:10.1056/NEJMoa1412098
Accepted for publication in Philosophy and Biology 25.04.2018
23
Pantic I (2011) Cancer stem cell hypotheses: impact on modern molecular physiology and pharmacology
research.JBiosci36:957–961
PepperJW,ScottFindlayC,KassenR,etal(2009)Cancerresearchmeetsevolutionarybiology.EvolAppl2:62–
70.doi:10.1111/j.1752-4571.2008.00063.x
PepperJW,SprouffskeK,MaleyCC(2007)Animalcelldifferentiationpatternssuppresssomaticevolution.PLoS
ComputBiol3:e250.doi:10.1371/journal.pcbi.0030250
Piccirillo SG, Combi R, Cajola L, et al (2009) Distinct pools of cancer stem-like cells coexist within human
glioblastomas and display different tumorigenicity and independent genomic evolution. Oncogene
28:1807–1811.doi:onc200927[pii]10.1038/onc.2009.27
PolyakK(2007)Breastcancer:originsandevolution.JClinInvest117:3155–3163.doi:10.1172/JCI33295
PotterNE,ErminiL,PapaemmanuilE,etal(2013)Single-cellmutationalprofilingandclonalphylogenyincancer.
GenomeRes23:2115–25.doi:10.1101/gr.159913.113
QuintanaE,ShackletonM,SabelMS,etal(2008)Efficienttumourformationbysinglehumanmelanomacells.
Nature456:593–598.doi:nature07567[pii]10.1038/nature07567
Rausch T, Jones DTW, Zapatka M, et al (2012) Genome sequencing of pediatric medulloblastoma links
catastrophicDNArearrangementswithTP53mutations.Cell148:59–71.doi:10.1016/j.cell.2011.12.013
ReyaT,MorrisonSJ,ClarkeMF,WeissmanIL(2001)Stemcells,cancer,andcancerstemcells.Nature414:105–
11.doi:10.1038/3510216735102167[pii]
ReynaudD,PietrasE,Barry-HolsonK,etal (2011) IL-6controls leukemicmultipotentprogenitorcell fateand
contributes to chronic myelogenous leukemia development. Cancer Cell 20:661–73 . doi:
10.1016/j.ccr.2011.10.012
Schuh A, Becq J, Humphray S, et al (2012)Monitoring chronic lymphocytic leukemia progression by whole
genome sequencing reveals heterogeneous clonal evolution patterns. Blood 120:4191–6 . doi:
10.1182/blood-2012-05-433540
ShackletonM,QuintanaE,FearonER,MorrisonSJ(2009)Heterogeneityincancer:cancerstemcellsversusclonal
evolution.Cell138:822–829.doi:S0092-8674(09)01030-7[pii]10.1016/j.cell.2009.08.017
ShainAH,YehI,KovalyshynI,etal(2015)TheGeneticEvolutionofMelanomafromPrecursorLesions.NEnglJ
Med373:1926–36.doi:10.1056/NEJMoa1502583
ShipitsinM,CampbellLL,ArganiP,etal(2007)Moleculardefinitionofbreasttumorheterogeneity.CancerCell
11:259–273.doi:S1535-6108(07)00029-3[pii]10.1016/j.ccr.2007.01.013
SiegmundKD,MarjoramP,TavaréS,ShibataD(2009)Manycolorectalcancersare“flat”clonalexpansions.Cell
Cycle8:2187–93
SimonsBD(2016)Deepsequencingasaprobeofnormalstemcellfateandpreneoplasiainhumanepidermis.
ProcNatlAcadSciUSA113:128–33.doi:10.1073/pnas.1516123113
SottorivaA,BarnesCP,GrahamTA(2017)Catchmydrift?Makingsenseofgenomicintra-tumourheterogeneity.
BiochimBiophysActa-RevCancer.doi:10.1016/j.bbcan.2016.12.003
SottorivaA,KangH,MaZ,etal(2015)ABigBangmodelofhumancolorectaltumorgrowth.NatGenet47:209–
216.doi:10.1038/ng.3214
Accepted for publication in Philosophy and Biology 25.04.2018
24
SottorivaA,SpiteriI,PiccirilloSGM,etal(2013)Intratumorheterogeneityinhumanglioblastomareflectscancer
evolutionarydynamics.ProcNatlAcadSciUSA110:4009–14.doi:10.1073/pnas.1219747110
SottorivaA,VerhoeffJJ,BorovskiT,etal(2010)Cancerstemcelltumormodelrevealsinvasivemorphologyand
increased phenotypical heterogeneity. Cancer Res 70:46–56 . doi: 0008-5472.CAN-09-3663 [pii]
10.1158/0008-5472.CAN-09-3663
SprouffskeK,AthenaAktipisC,RadichJP,etal(2013)Anevolutionaryexplanationforthepresenceofcancer
nonstemcellsinneoplasms.EvolAppl6:92–101.doi:10.1111/eva.12030
StephensPJ,GreenmanCD,FuB,etal(2011)Massivegenomicrearrangementacquiredinasinglecatastrophic
event during cancer development. Cell 144:27–40 . doi: S0092-8674(10)01377-2
[pii]10.1016/j.cell.2010.11.055
SunJ,RamosA,ChapmanB,etal(2014)Clonaldynamicsofnativehaematopoiesis.Nature514:322–327.doi:
10.1038/nature13824
SuzukiM,HarashimaA,OkochiA,etal(2004)5-Azacytidinesupportsthelong-termrepopulatingactivityofcord
bloodCD34(+)cells.AmJHematol77:313–5.doi:10.1002/ajh.20178
TakebeN,IvySP(2010)Controversiesincancerstemcells:targetingembryonicsignalingpathways.ClinCancer
Res16:3106–3112.doi:1078-0432.CCR-09-2934[pii]10.1158/1078-0432.CCR-09-2934
ThirlwellC,WillOCC,DomingoE,etal(2010)Clonalityassessmentandclonalorderingofindividualneoplastic
crypts shows polyclonality of colorectal adenomas. Gastroenterology 138:1441–54, 1454–7 . doi:
10.1053/j.gastro.2010.01.033
Tomasetti C, Li L, Vogelstein B (2017) Stem cell divisions, somatic mutations, cancer etiology, and cancer
prevention.Science(80-)355:1330–1334.doi:10.1126/science.aaf9011
TomasettiC,VogelsteinB(2015)Variationincancerriskamongtissuescanbeexplainedbythenumberofstem
celldivisions.Science(80-)347:78–81.doi:10.1126/science.1260825
VanKeymeulenA,LeeMY,OussetM,etal(2015)ReactivationofmultipotencybyoncogenicPIK3CAinduces
breasttumourheterogeneity.Nature525:119–23.doi:10.1038/nature14665
WalterMJ,ShenD,DingL,etal(2012)Clonalarchitectureofsecondaryacutemyeloidleukemia.NEnglJMed
366:1090–8.doi:10.1056/NEJMoa1106968
WelnerRS,AmabileG,BarariaD,etal(2015)Treatmentofchronicmyelogenousleukemiabyblockingcytokine
alterations found in normal stem and progenitor cells. Cancer Cell 27:671–81 . doi:
10.1016/j.ccell.2015.04.004
WichaMS,LiuS,DontuG(2006)Cancerstemcells:anoldidea--aparadigmshift.CancerRes66:1883–1886.doi:
66/4/1883[pii]10.1158/0008-5472.CAN-05-3153
WilliamsMJ,WernerB,BarnesCP,etal(2016)Identificationofneutraltumorevolutionacrosscancertypes.Nat
Genet48:238–244.doi:10.1038/ng.3489
XieM,LuC,WangJ,etal (2014)Age-relatedmutationsassociatedwithclonalhematopoieticexpansionand
malignancies.NatMed20:1472–8.doi:10.1038/nm.3733
XuX,HouY,YinX,etal(2012)Single-cellexomesequencingrevealssingle-nucleotidemutationcharacteristics
ofakidneytumor.Cell148:886–95.doi:10.1016/j.cell.2012.02.025
Accepted for publication in Philosophy and Biology 25.04.2018
25
Yachida S, Wood LD, Suzuki M, et al (2016) Genomic Sequencing Identifies ELF3 as a Driver of Ampullary
Carcinoma.CancerCell29:229–240.doi:10.1016/j.ccell.2015.12.012
ZhangJ,FujimotoJ,ZhangJ,etal(2014)Intratumorheterogeneityinlocalizedlungadenocarcinomasdelineated
bymultiregionsequencing.Science346:256–9.doi:10.1126/science.1256930
ZhangXC,XuC,MitchellRM,etal(2013)Tumorevolutionandintratumorheterogeneityofanoropharyngeal
squamouscellcarcinomarevealedbywhole-genomesequencing.Neoplasia15:1371–8