ebook img

Cell death and life in cancer: mathematical modeling of cell fate decisions PDF

0.39 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Cell death and life in cancer: mathematical modeling of cell fate decisions

Cell death and life in cancer: mathematical modeling of cell fate decisions AndreiZinovyev,SimonFourquet,LaurentTournier,LaurenceCalzoneand EmmanuelBarillot 3 1 0 2 n a J 1 1 AbstractTumordevelopmentischaracterizedbyacompromisedbalancebetween celllifeanddeathdecisionmechanisms,whicharetighlyregulatedinnormalcells. ] N Understanding this process provides insights for developing new treatments for M fighting with cancer. We present a study of a mathematical model describing cel- lular choice between survival and two alternative cell death modalities: apoptosis . o andnecrosis.Themodelisimplementedindiscretemodelingformalismandallows i topredictprobabilitiesofhavingaparticularcellularphenotypeinresponsetoen- b gagementof cell death receptors. Using an originalparameter sensitivity analysis - q developedfordiscrete dynamicsystems, we determinevariablesthat appearto be [ criticalin thecellularfatedecisionanddiscusshowtheyareexploitedbyexisting 1 cancertherapies. v 6 6 3 2 . 1 0 AndreiZinovyev 3 U900INSERM/InstitutCurie/EcoledeMines,InstitutCurie,26rued’Ulm,Paris,France,75005 1 e-mail:[email protected] : v SimonFourquet i U900INSERM/InstitutCurie/EcoledeMines,InstitutCurie,26rued’Ulm,Paris,France,75005 X e-mail:[email protected] r a LaurentTournier INRA, Unit MIG (Mathe´matiques, Informatique et Ge´nome), Domaine Vilvert, Jouy en Josas, France,78350,e-mail:[email protected] LaurenceCalzone U900INSERM/InstitutCurie/EcoledeMines,InstitutCurie,26rued’Ulm,Paris,France,75005 e-mail:[email protected] EmmanuelBarillot U900INSERM/InstitutCurie/EcoledeMines,InstitutCurie,26rued’Ulm,Paris,France,75005 e-mail:[email protected] 1 2 AuthorsSuppressedDuetoExcessiveLength 1 Introduction Evadingvariousprogrammedcelldeathmodalitiesisconsideredasoneofthemajor hallmarksofcancercells[1].Abetterunderstandingofthepro-deathorprosurvival rolesofthegenesassociatedwithvariouscancers,andtheirinteractionswithother pathwayswouldsetagroundforre-establishingalostdeathphenotypeandidenti- fyingpotentialdrugtargets. Recentprogressinstudyingthemechanismsofcelllife/deathdecisionsrevealed itsastoundingcomplexity.Amongmany,onecanmentionthreedifficultiesonthe way to characterize, describe and create strict mathematical descriptions of these mechanisms. First,thesignalingnetworkallowingacelltoreacttoanexternalstress(suchas damageofDNA,nutrientandoxygendeprivation,toxicenvironment)isassembled from highly redundantpathways which are able to compensate each other in one wayoranother.Forexample,thereexistatleastsevendistinctandparallelsurvival pathways associated with action of AKT protein [2]. Disruption of one of these pathwaysinapotentialcelldeath-inducingcancertherapycanbeinprinciplecom- pensatedbytheothers.Thus,understandingandmodelingthesurvivalresponsein itsfullcomplexityisadauntingtask. Second, cellular death is an extremely complex phenotype that cannot merely be described as a simple disaggregation of cellular components driven by purely thermodynamicallaws. Several distinct modesof cell death were identified in the lastdecade[3],suchasnecrosis,apoptosisandautophagy.Importantly,allthesecell deathmodalitiesarecontrolledbycellularbiochemicalmechanisms,activatedinre- sponsetodiversetypesofstress:roughlyspeaking,acellisusuallypre-programmed to diein a certainmanner,sendingappropriatesignalstoits surroundingsso asto limit tissue toxicity and allow recycling of its components. Necrosis is a type of celldeathusuallyassociatedwithalackofimportantcellularresourcesuchasATP, whichmakesfunctioningofmanybiochemicalpathwaysimpossible.Thisiswhyit was long thoughtof as an uncontrolledand purely thermodynamics-drivendegra- dationofcellularstructures.However,recentresearchshowedthatnecrosiscanbe triggered by specific signals through the activation of tightly regulated pathways, andcanevenproceedwithoutATPdepletion[3].By contrast,apoptosisasaform ofcellularsuicidewas,fromtheverybeginning,described asamodeofcelldeath requiringenergyfor the permeabilizationof mitochondrialmembranesand cleav- age of intracellular structures. Autophagy remains a relatively poorly understood cell death mechanism, which seems to serve both as a survival or a death modal- ity.Uponcertainstressconditions,anduntilthisstressisrelieved,cellularcompo- nentssuchasdamagedproteinsororganellesaredigestedandrecycledintoreusable metabolites,andmetabolismisreorientedsoastosparevitalfunctions.Longlast- ing,non-relievablestresswasdescribedastriggeringautophagiccelldeath,through unaffordable cellular self-digestion. However, no experimental evidence ever un- ambiguouslydemonstratedthatsuchcelldeathisdirectlyexecutedbyautophagyin vivo,butinthespecialcaseoftheinvolutionofDrosophilamelanogastersalivary glands[3]. Celldeathandlifeincancer:mathematicalmodelingofcellfatedecisions 3 The third difficulty can be attributed not directly to the complexity of the bio- chemicalmechanismsbutrathertoourcapabilitiesofapprehendingthedesignprin- ciples used by biological evolution. Inspired by engineeringpractices, we tend to investigatecomplexsystemsbysplittingthemintorelativelyindependentmodules andassociatingwell-characterizednon-overlappingfunctionstoeachmolecularde- tail. Applyingsuch reductionistapproachesto biologycomes with a caveat. Most cellular molecular machineries cannot be naturally dissected or associated with well-definedfunctions,andsetsofoverlappingfunctionscanbedistributedamong groupsofmolecularplayers. Not having the ambition to deal with the whole complexity of cell fate deci- sionsinvivo,wedecidedtoconcentrateonmodelingtheoutcomeofaclassicaland ratherwell-definedexperimentofinducingcelldeath:addingtoacellculturespe- cificligands(TumorNecrosisFactor,TNF,orothermembersofitsfamilysuchas FASL).Theseso-calleddeathligandscanengagedeathreceptorsandtriggerapop- tosisornecrosis,oractivatepro-survivalmechanisms[5].Thenetoutcomeofsuch experimentsdependsonmanycircumstances:celltype,doseoftheligand,duration of the treatment, specific mutations in cell genomes, etc. Moreover,it is believed thattheoutcomecanhaveintrinsicstochasticnaturegovernedbycellulardecision making mechanisms and intrinsic molecular noise [6]. Trying to characterize the biochemicalresponseofacelltothisrelativelysimplekindofperturbationallows tounderstandcertaincellfatedecisionmechanisms. Inthispaper,webrieflydescribeandcarefullyanalyzeamathematicalmodelof cellfatedecisionbetweensurvivalandtwoalternativemodesofcelldeath:apoptosis andnecrosis.Themodelwascreatedandintroducedin[4].Hereproposetheprin- ciplesforwiringandparametrizingabiologicaldiagramthatdescribesthiscellular switch.Inadditionto[4],here,byapplyinganovelsensitivityanalysisspecifically developedfor discrete modeling,we identify fragile sites of the cell fate decision mechanism.Inconclusion,wecompareouranalysiswithourcurrentknowledgeof cellulardecisionmakingfragilitiesutilizedbycancerandcancertherapies. 2 Mathematical model ofcell fatedecision In [4] we summarizedthe currentknowledgeonthe interactionsbetweencellfate decisionmechanismsinasimplisticwiringdiagram(seeFig.1)whereanoderep- resentseithera protein(TNF, FADD, FASL,TNFR, CASP8, cFLIP,BCL2, BAX, IKK,NFk B,CYT C,SMAC,XIAP,CASP3),astateofprotein(RIP1ub,RIP1K),a smallmolecule(ROS,ATP),amolecularcomplex(Apoptosome,C2 TNF,DISC FAS), agroupofmolecularentitiessharingthesamefunction(BAXcanthusrepresentei- therofBAX andBAK,cIAP eithercIAP1orcIAP2,andBCL2 anyoftheBH1-4 BCL2familymembers,),amolecularprocess(Mitochondriapermeabilizationtran- sition,MPT,Mitochondrialoutermembranepermeabilization,MOMP)orapheno- type(Survival,Apoptosis,Non-apoptoticcelldeath,NonACD).Eachdirectedand 4 AuthorsSuppressedDuetoExcessiveLength signededgerepresentsaninfluenceofonemolecularentityonanother,eitherposi- tive(arrowededge)ornegative(headededge). Thephenotypenodesonthediagramaresimpleinterpretationsofthefollowing molecularconditions: 1) activated NFk B is read as survivalstate; 2) lack of ATP is read as nonapoptotic cell death state; 3) activated CASP3 is read as apoptotic celldeath.Absenceofanyofsuchconditionsisinterpretedasa”naive“cellstate, correspondingtothefourthcellularphenotype. Afterextensiveexaminationofthebiologicalliteratureweconvertedthediagram into a logical mathematicalmodelof cell fate decisions triggeredby activation of cell death receptors. The wiring diagram and the logical rules defining the model areshownonFig.1. Byapplyingatechniqueadaptedtodiscreteformalism[7],wereducedthismodel toa11-dimensionalnetwork,thusenablingacompleteanalysisoftheasynchronous dynamics(see[4]fordetails).Thisanalysisidentified27stablelogicalstatesandno cyclicattractors.Moreover,itshowedthatthedistributionofthestablelogicalstates inthediscrete22-dimensionalspaceofinternalmodelvariables(withoutconsider- inginputandoutputvariables)formsfourcompactclusters,eachcorrespondingtoa particularcellularphenotype.Threeoftheseclusterscanbeattributedtoaparticular cellfate(survival,apoptosis,necrosis)whiletheforthrepresentsa“naive”survival state,wherenodeathreceptorsareinduced. 3 Computing phenotype probabilities Aswehavealreadymentioned,thecellularfatedecisionmachineryischaracterized bystochasticresponse,i.e.givenastimuli,thecellcanreachseveralfinalstates,cor- respondingtodifferentphenotypes,withdifferentprobabilities.Theroleofmathe- maticalmodelinginthiscasecanbetopredicttheseprobabilitiesasabsolutevalues thatcanbematchedtoanexperiment,oratleasttopredicttherelativechangesof theprobabilitiesafterintroducingsomeperturbationsto thesystem. Wehaveimplementedthisideaforthemathematicalmodelofcellfatedecisions describedaboveinthefollowingmanner. Inordertodescribeourresults,letusintroducethenotionofasynchronousstate transition graph. On this graph, each node represents a state of the system which in this case can be encoded by a n-dimensional vector of 0s and 1s (n being the dimensionofthesystem).Adirectededgeexistsbetweentwostatesxandyifthere existsanindexi∈{1,...,n}suchthaty = f(x)6=x andy =x for j6=i(here, f i i i j j i denotesthelogicalruleofvariablex,seeFig.1foracompletelistofthemodellogi- i calrules).Inprinciple,thestatetransitiongraphcouldbedefinedindependentlyand withoutthebiologicaldiagram,however,thiswouldrequirea tremendousamount ofempiricalknowledgeaboutthesetofallpermissibletransitionsbetweenthecell stateswhichisnotavailable.Hence,thebiologicaldiagramwithassociatedlogical rulesisusedasacompactrepresentationandatooltogeneratethestatetransition graph.Detailedinstructionsonthisprocedurecanbefoundin[8,9]. Celldeathandlifeincancer:mathematicalmodelingofcellfatedecisions 5 TNF FASL FADD TNFR C2_TNF DISC_FAS RIP1 CASP8 survival necrosis RIP1ub RIP1K cFLIP apoptosis cIAP IKK BCL2 BAX MPT MOMP ROS NFkB SMAC CYT_C XIAP Apoptosome ATP CASP3 Survival NonACD Apoptosis DISCTNF’=TNFRANDFADD TNF’=TNF RIP1’=(TNFRORDISCFAS)AND(NOTCASP8) FADD’=FADD CASP8’=(DISCTNFORDISCFASORCASP3)AND(NOTcFlip)FAS’=FAS RIPub’=RIP1ANDcIAP TNFR’=TNF cIAP’=(NFkBORcIAP)AND(NOTSMAC) RIP1K’=RIP1 BAX’=CASP8AND(NOTBCL2) cFlip’=NFkB ROS’=(RIP1KORMPT)AND(NOTNFkB) IKK’=RIP1ub MPT’=ROSAND(NOTBCL2) BCL2’=NFkB MOMP’=BAXORMPT SMAC’=MOMP NFkB’=IKKAND(NOTCASP3) CYTC’=MOMP XIAP’=NFkBAND(NOTSMAC) DISCFAS’=FASANDFADD Apoptosome’=CYTCANDATPAND(NOTXIAP) ATP’=NOTMPT CASP3’=ApoptosomeAND(NOTXIAP) Fig.1 Biologicaldiagramofmolecularinteractionsinvolvedincellfatedecisionsderivedfrom thebiologicalliterature.Thediagramisroughlydividedbydashedlinesintothreemodulescor- responding to three submechanisms of cell fate decisions. Notations: 1) Proteins: TNF, FADD, FASL,TNFR,CASP8,cIAP,cFLIP,BCL2,BAX,IKK,NFk B,CYTC,SMAC,XIAP,CASP3; 2)Statesofproteins:RIP1ub(ubiquitinatedformofRIP1),RIP1K(kinasefunctionofRIP1);3) Small molecules: ATP, ROS (Reactive oxygen species); 4) Molecular complexes: Apoptosome, C2 TNF, DISC FAS; 5) Molecularprocesses: MPT (Mitochondria permeabilization transition), MOMP (Mitochondrial outermembrane permeabilization); 6) Phenotypes: Survival, Apoptosis, NonACD(Non-apoptoticcelldeath).Belowthetableoflogicalrulesdefiningthediscretemathe- maticalmodelisprovided. 6 AuthorsSuppressedDuetoExcessiveLength Thesetofallpossiblestatesprovidesadiscretephasespaceofthesystem.The statetransitiongraphcontainsallpossiblewaysofthesystemsdynamics(trajecto- ries).Inotherwords,itisthemultidimensionalepigeneticlandscapeofthecellfate decisionsystem.Notethatthestatetransitiongraphisassumedtoberathersparse compared to the fully connected graph where any two state transitions would be possible.Hence,on thislandscape,onecan determinebifurcatingstates, pointsof noreturn,etc. Thestatetransitiongraphallowstoaddressthefollowingquestion:Startingfrom adistinguishedstateofacell,whatistheprobabilitytoarrivetoeachofthestable states? In biologicalterms: Which proportionsof a populationof resting cells ex- posedtodeathligandwilleventuallydisplayeachofthedifferentphenotypes-cell fate? To answer the question, we converted the state transition graph into a Markov processof randomwalk on a graph,followingthe methoddescribedin [9]. To do that,weassociatedtoeachtransitionbetweentwostatesaprobability(calledtran- sition probability). By applying classical algorithms to the transition probability matrix(stronglyconnecteddecompositionandtopologicalsort),weobtainedanab- sorbingdiscreteMarkovchain,andthenanalyzeditwithclassicaltechniques[10]. Oneofthecriticalpointsinsuchtypeofanalysisliesinthechoiceofthetransi- tionprobabilities.Onceagain,definingtheseprobabilitiesdirectlyfromsomeem- piricalobservationsisimpossibleatpresenttime.Hence,theseprobabilitiesshould bederivedfromthelogicalmodelwiththeuseofsomeadditionalassumptions. The simplest assumption is to consider all transitions firing from a given state as equiprobable.Biologicalinterpretationof such an assumption is notsimple. In a way, we consider a “generic” cell in which all possible system trajectories take place with equalprobabilities(withoutdominance,i.e. any preferableroute). One can arguethat in any particularconcrete cell, this would notbe true anymoreand thatthegenericcellisnotrepresentativeofanythingrealobservedinanybiological experiment.Havinginmindthisdifficulty,weavoiddirectinterpretationofabsolute valuesofprobabilities,concentratingratheronrelativechangesoftheminresponse tosomesystemmodificationssuchasremovinganodeorfixinganode’sactivity.It happensthatsucha“generic”cellmodelisalreadycapableofreproducinganumber ofknownexperimentalfacts. When the state transition graphis parametrizedby transition probabilities, one canusestandardtechniquestocomputetheprobabilityofhittingagivenstablestate, consideringthatarandomwalkstartsfromagiveninitialstate.Thenthisprobabil- ity is associated with a probability of observing a particular phenotype in given experimentalconditions. For doing this, it is convenientto define a unique initial state,whichwechoosetorepresentthe“physiologicalstate”,theonerepresenting un-inducedcells growingin a plate. In the modelof Fig. 1 it is the state in which allelementsareinactiveexceptATP,FADDandcIAP.Thisisastablestate,which looses its stability when TNF variable is changed from 0 to 1 and the dynamical systemstartstoevolveintime. Usingthisapproach,weperformedaseriesofinsilicoexperimentsinwhichthe probability of arriving to stable states was computed for the initial (”wild-type”) Celldeathandlifeincancer:mathematicalmodelingofcellfatedecisions 7 model,orfora series of modified(“mutant”)model.Typicalmodelmodifications consisted in fixing some nodes’ activities to 0 or to 1. For our cell fate decision model,theresultsareprovidedinFig.2.In[4]thistablewassystematicallycom- paredwiththeexperimentaldataofthecelldeathphenotypemodificationsobserved invariousmutantexperimentalsystems,includingcellculturesandmice.Themodel was able to qualitativelyrecapitulateall of themand to suggestsome new yetun- explored experimentally mutant phenotypes. The most interesting in this setting wouldbetoconsidersyntheticinteractionsbetweenindividualmutants,whensev- eralnodesonthediagramareaffectedbyamutationsimultaneously. wild−type antiox APAF deletion BAX deletion S N S N N N 0 0 A A S S BCL2 o.e. CASP8 deletion CASP8 active cFlip deletion N N S N 0 S A A cIAP deletion FADD deletion NFkB deletion NFkB o.e. N A N A N S A S RIP1 deletion XIAP deletion z−VAD−fmk RIP+1 zd−eVleAtDio−nfmk S N N S A A 0 Fig.2 Changesinthephenotypeprobabilitiesfromtherandomwalkonthestatetransitiongraph, startingfromtheinitialphysiologicalstate.Various“mutant”modificationsofthedynamicalsys- temaretestedhere.Here“A“denotesApoptosis,“N“denotesNecrosisand“S“denotesSurvival, “0” denotes Naive state. “O.e.” stands for overexpression of a protein, “antiox” corresponds to bluntingthecapacityofNFk Btoprevent ROSformation,“z-VAD fmk”simulatestheeffect of caspaseinhibitorz-VAD-fmk. 4 Identification offragilepoints ofthe cellfate decision machinery Changingdistributionoftransitionprobabilitiesontheasynchronousstatetransition graphcan drastically changethe probabilistic outcomeof a computationalexperi- 8 AuthorsSuppressedDuetoExcessiveLength ment. At the same time, the probabilities for a random walk to convergeto some attractor depend also on the structure of the state transition graph which is deter- mined solely from the discrete model.In orderto understand what are the critical determinants of a cellular choice, we applied a novel strategy of discrete model analysisconsistinginparametrizingthestatetransition graphbychangingrelative importance of certain variables. In a certain sense, this strategy corresponds to a sensitivity analysis, commonly applied for continuous models based on ordinary differentialequationsandchemicalkineticsapproach[11]. Firstofall,wepostulatethatour“reference”parametrizationcorrespondstothe equalprobabilitiesofanypossibletransitionfromastate.Asmentionedearlier,this corresponds to a “generic” cell model, where the relative speeds of all biochem- ical processes are assumed equal. Mathematically, considering the dynamics as a Markovprocess,alltransitionsfromagivenstate xtoanyofitsasynchronoussuc- cessorareassignedequalprobabilities(ifxhasrsuccessors,theseprobabilitiesare equalto1/r).Wewillmodifythisdefaultparametrizationbysystematicallychang- ing relative speeds of certain elements. This will lead to some re-parametrization of the state transition graph and consequent changes in the probabilities to reach attractors. Thekeyideaofpriorityclasses[12,13]consistsingroupingvariablesofadis- cretemodelintoclassesaccordingtothespeedsoftheunderlyingprocessesgovern- ingtheirturnoverrates.Forinstance,inthe caseofgeneticregulatorynetworks,a naturalgroupingconsistsinputtingdenovoproteinsynthesis(transcription+trans- lation) in a slow transitionclass in comparisonwith otherprocessessuch as post- translationalproteinmodifications(phosphorylation,ubiquitination,...)orcomplex formation.Followingthisidea,wecanregroupnodesintopriorityclassestowhich somepriorityratios w areassigned.Said differently,eachvariable x is assigneda i priorityvaluew.Foragivennode,avaluew >1correspondstoahigherthande- i i faultpriority,and a value w <1 to a lowerthan defaultpriority.The ratio w can i i beinterpretedasaglobalturnoverrateofthecomponentrepresentedbythisnode: thosethatareproduced(activated)anddegraded(deactivated)fastwillhavealarge w. i Considerastatex,withrasynchronoussuccessors.Bydefinition,betweenxand eachofitssuccessors,oneandonlyonevariablecanbeupdated.Letydenoteoneof thesuccessorsofx,andibetheindexofthecorrespondingupdatedvariable.With the uniformassumptiondescribed before,the probabilityof the transition (x→y) isindependentofiandisequalto1/r.Withpriorityclasses,thisprobabilityisnow weighted by w, making the transition more probableif component i belongsto a i “fast”class(w greaterthanone)andlessprobableifitbelongstoa“slow”class(w i i lessthanone).Obviously,forcomputingtheactualtransitionprobabilities p ,a x→y normalizationshouldbeappliedsothat: (cid:229) p =1. x→y ysucc.ofx Celldeathandlifeincancer:mathematicalmodelingofcellfatedecisions 9 Once the new valuesof the transitionprobabilitieshave been computed,the same treatmentsas before can be applied, leading to new valuesfor the probabilitiesto reachthedifferentphenotypes,startingfromagiveninitialcondition. Thisgeneralmethodmaybeappliedintwodifferentways.First,onemayuseit tocomputemorerealisticprobabilities,thatcouldbecomparedtoactualexperimen- talresults(theprobabilitytoreachanattractorbeingcomparedwiththeproportion ofcellsexhibitingthecorrespondingphenotype).However,suchcalculationswould needacompleteclassificationoftherelativespeedsofallbiochemicalmechanisms involvedinthemodel.Giventhenumberandheterogeneityofthesemechanisms,it isstilldifficulttoobtainsuchclassification.Instead,weusedthemethodasa sen- sitivityanalysistool,inordertodetectwhichvariablesaremorecriticalthanothers in the decision-makingprocess.Using the reducedmodelevokedearlier(see [4]), weconsideredeachvariableindependently,andsuccessivelyboosteditorslowedit downbysomemultiplicativefactor.Moreprecisely,todetectthesensitivityofthe networkwithrespecttotheturnoverofvariablex,weperformedthecalculationsfor i differentvaluesofw,theotherweightsw beingkeptatone(thereferencevalue). i j By comparing the probabilities to reach the three phenotypes-survival, apoptosis and necrosis- with those of the initial model, one can detect whetherthe system’s response is sensitive ornotto the turnoverrate of variable x. We performedsuch i experimentsforthenineinnervariablesofthereducedmodel.Figure3presentsthe resultsweobtained. Fig. 3 . Testing the effect of varying node turnovers on the resulting phenotypic probabilities. Theabscissonthegraphs showsthevalueof wpriorityvalue, where w=1corresponds tothe probabilitiescomputedforthedefaultwild-typemodel(seeFig.2).Thecolorsarethoseadopted in[4]:orangecorrespondstoapoptosis,purpletonecrosisandgreentosurvival. The plots revealseveralinteresting properties.First, the most sensitive compo- nents,whichcorrespondtothecurveswiththehighestamplitude,areRIP1,NFkB 10 AuthorsSuppressedDuetoExcessiveLength andCASP8.Thisreinforcestheideathatthesethreecomponentsplayacrucialrole in the decision process. This seems reasonable, especially for RIP1 and CASP8, astheyoccupyanupstreampositionintheregulatorygraph. Interestingly,CASP3 turnoverdoes not seem to be so important,althoughCASP3 is a markerof apop- tosis.ThisconfirmsthateventhoughCASP3isessentialfortheexistenceofapop- tosisinthemodel(itsremovalcompletelysuppressapoptoticoutcome,seeFig.2), its turnover rate does not appear to be important in the dynamics of the decision process (once it goes from 0 to 1, most of the decision has already been made). Remarkably,theturnoversofMOMPandMPT,bothcontributingtothepermeabi- lizationofmitochondrialmembrane,havedifferenteffects:MOMPseemstoaffect mainly the decision between survivaland necrosis, while MPT playsa role in the switchbetweenapoptosisandnecrosis. Thesensitivityanalysisthatispresentedhereisanextensionoftheresultspro- posed in [4]. In contrastwith the all-or-noneperturbationsevokedin the previous part(whereanodeisfixedto0or1),hereweconsiderfinerperturbationsbymod- ifyingtheturnoverratesofthemodel’svariables.Anextstepwouldbetoconsider the relative strengthsof the model’s interactions,instead of the model’svariables. Suchanapproachiscurrentlyinvestigated. 5 Comparisonwiththe fragilitiesexploitedby cancer and its treatment Deregulationsofthesignallingpathwaysstudiedherecanleadtodrasticandserious consequences.HanahanandWeinbergproposedthatescape of apoptosis,together with other alterations of cellular physiology,represents a necessary event in can- cer promotion and progression [1]. As a result, somatic mutations leading to im- paired apoptosis are expected to be associated with cancer. In the cell fate model presentedhere,mostnodescan beclassified as pro-apoptoticoranti-apoptoticac- cording to the results of “mutant” model simulations, which are correlated with experimental results found in the literature. Genes classified as pro-apoptotic in ourmodelincludecaspases-8 and -3, APAF1 as partof the apoptosomecomplex, cytochromec (Cyt c), BAX, and SMAC. Anti-apoptotic genes encompass BCL2, cIAP1/2,XIAP,cFLIP,anddifferentgenesinvolvedintheNFkBpathway,includ- ingNFKB1,RELA,IKBKGandIKBKB(notexplicitinthemodel).Geneticalter- ations leadingto loss of activity of pro-apoptoticgenesor to increasedactivity of anti-apoptoticgeneshavebeenassociatedwithvariouscancers.Thus,wecancross- listthealterationsofthesegenesdeducedfromthemodelwithwhatisreportedin theliteratureandverifytheirroleandimplicationsincancer. Forinstance, concerningpro-apoptoticgenes, frameshift mutationsin the ORF oftheBAXgenearereportedin>50%ofcolorectaltumoursofthemicro-satellite mutatorphenotype[14].ExpressionofCASP8isreducedin≈24%oftumoursfrom patientswithEwing’ssarcoma[15].Caspase-8wassuggestedinseveralstudiesto

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.