Micrometer-ScaleOxygenDeliveryRearrangesCellsandPreventsNecrosisinTumo - PDF document

Micrometer-ScaleOxygenDeliveryRearrangesCellsandPreventsNecrosisinTumorTissueInVitroBhushanJ.ToleyDept.ofChemicalEngineeringUniversityofMassachusetts,Amherst,MA01003JaehyunParkDept.ofElectricalEngineeringandComputerScience,UniversityofCalifornia,Berkeley,CA94720Byoung-JinKimandRajaVenkatasubramanianDept.ofChemicalEngineeringUniversityofMassachusetts,Amherst,MA01003 AdditionalSupportingInformationmaybefoundintheonlinever-sionofthisarticle.CorrespondenceconcerningthisarticleshouldbeaddressedtoN.S.Forbesatforbes@ecs.umass.edu.2012AmericanInstituteofChemicalEngineers almostequalrole.Similarly,experimentswithcellslackinghypoxiainduciblefactor-1,theprimaryoxygen-sensingtranscriptionfactor,haveshownthatglucosegradientslimittheabilityofcellstorespondtolowoxygenenvironments.Cellularrearrangement,de“nedasthechangeinlocationofcellswithintissueovertime,isanessentialinitialstepinthemetastaticspreadofcancercellsfromtumors.Toenterthebloodstream,cancercellsneedtopenetratethroughsur-roundingcancerousandnormaltissue.Incellculture,hypoxiahasbeenshowntobothstimulateandrepressinvasionintoextracellularmatrix.Hypoxiaaffectsinvasive-nessbyregulatingCXCR4anddisablingtheE-cadherin-catenincomplex.Hypoxiahasbeenshowntoenhancespreadingofcellswithinspheroidsontosurroundingcellmonolayers.Inspheroids,cellsspontaneouslyrearrange,circulatingfromtheoutsideedgetowardthecenter.Computationalmodelsoftumorcordshavesuggestedthatvaryingoxygenconcentrationscanrearrangecellsbyaffect-inggrowthanddeathrates.Othermathematicalmodelsofcellrearrangementhavesuggestedthatnutrientgradientscouldcontrolcellmigration,butithasnotbeendemon-stratedwhetheroxygengradientsalonecancontrolthesephenomena.Severalcellculturetechniqueshavebeendevelopedtoinvestigatetheeffectsofmodi“edoxygengradients.Earlymethodstocreateinvitrooxygengradientsarethebubbleandcapillaryassays,whichwereusedprimarilytostudymicrobialchemotaxis.Morerecently,aperfusionbioreactorsystemhasbeencreatedthatgeneratesoxygengradientsoveramonolayerofmammaliancellsina”at-platereac-Anothersystemutilizedamicrofabricatedinsertformultiwellplatestospatiallyandtemporallymodifyoxygengradientsoncellmonolayers.Finally,3Doxygengradientshavebeencreatedbystackinglayersofpaperimpregnatedwithcellsinanextracellularmatrix.Althoughthesesys-temsallelegantlydeterminedtheeffectsofoxygenonindi-vidualcells,noneweredesignedtodeterminetheeffectsoflocaloxygengradientsoncellularviabilityandrearrange-mentinthree-dimensionaltumortissue.Tounderstandtherelationshipsbetweenoxygenation,cellviability,andcellularrearrangement,wedevelopedamicro-meter-scaleoxygendeliverydevicetocreatepreciseoxygengradientsinthree-dimensionaltumortissue(Figure1A).Wehypothesizedthatoxygenwouldreducedeathandcausecellstorearrangeincylindroids.Thedevicewasdesignedtodelivercontrolleddosesofoxygentoscaffold-freethree-dimensionaltissue.Itwasbasedontwotechniquesprevi-ouslydevelopedinourlaboratories:amicrosystemthatpat-ternsoxidativemicrogradientsandmulticellulartumorcylindroids,aninvitroculturetechniquethatcreatesmicroenvironmentssimilartothoseintumors(Figure1B).Theprincipaldesigngoalwastocreateadevicethatcoulddeliveroxygentothenecroticcenterofcylindroids.Thisge-ometrywouldinvertoxygengradientswhilemaintainingallothernutrientgradientsandwouldenablestraightforwardmodelingoftheresults.Afterfabrication,thedevicewastestedbysupplyingoxygentotissuecomposedofcoloncar-cinomacells.Fluorescencestainingwasusedtodeterminethemagnitudeandlocationofcelldeath,andamathematicalmodelwasusedtodeterminethesensitivityofgrowthanddeathtooxygen.Resultsobtainedwiththedeviceshowthatoxygenisaprimaryfactorcontrollingcellviabilityandrear-rangementinthree-dimensionaltissue.MaterialsandMethodsDesignofthemicrometer-scaleoxygendeliverydeviceThedevicewasdesignedtoproducelocalizedoxygenatthecenterofcylindroids.Aniterativedesignprocesswasusedtointegratetheoxygengenerationchipwithcylindroidculture.10,11,41Allelectrolysischipdevicesconsistedof10-m-diametercircularmicroelectrodespatternedoveraglasssubstrateandcoveredwithagaspermeablepolydimethylsi-loxane(PDMS)microchannel(Figure1A).Theroofofthemicrochannelactedasabarriermembranebetweentheelec-trolysischamberandthecellculturearea.Thecathodehadalargesurfaceareatominimizehydrogen”uxandwasplacedfarfromtheanodetopreventinteractionofhydrogenwithtis-sues(Figure1C).Afterfabrication,theoxygengenerationchipswereattachedtothebottomof48wellplates(Figure1A).DissolvedoxygengeneratedattheanodediffusedthroughtheelectrolyteandPDMS“lm,andintothecellcul-turearea(Figure1A).Tocreateafunctioningtumoroxygen-ationdevice,severalfeaturesweretuned,includingthethicknessofthePDMSlayer,theheightoftheelectrolytechannel,thedesignofthechannel,themethodofcylindroidinsertionintothedevice,andthedesignofthecellchamber.Monolayer,spheroidandcylindroidculturesHumancolorectaladenocarcinomaLS174TcellsobtainedfromATCC(Manassas,VA)weregrownonT75”asksinlowglucose(1g/L)DulbeccosModi“edEaglesMedium(DMEM;Sigma-Aldrich,St.Louis,MO),andsupplementedwith10%fetalbovineserum.Spheroidsweremadebyinoc-ulatingsinglecellsuspensionsat2.5cells/mLontopoly(2-hydroxyethylmethacrylate)coatedT25culture”asks.Cylindroids(125mthick)wereformedbyconstrain-ingspheroidsbetweenculturesurfacesandpolycarbonateplugs(Figure1A,C)asdescribedpreviously(KasinskasandForbes,2006),andsubmergedunder200Lcultureme-dium.Allcultureswereincubatedat37C,pH7.4,in5%inambientair,with100%humidity.CylindroidpositioningForcontrolledlocalizeddeliveryofoxygen,accurateposi-tioningofcylindroidsovertheanodewasnecessary.Amicropositionerwasdesignedthatconsistedofapolycarbon-atecylindricalplug(3mmindiameter)attachedtoarectan-gularhead,whichwassuspendedonabasebyfourscrews(Figure1D).A3Lmediumdropletcontainingasinglespheroidwasplacedonthetipofthepositionerplugandintroducedintothewellplate.TheXandYcoordinatesofthetipwereadjustedbymovingtheplugmanually;theZcoordinate,whichcontrolledthicknessofthecylindroid,wasadjustedwiththefourscrews(Figure1D).VisualizationofelectrolyticallygeneratedoxygenOperationoftheoxygen-generatingelectrodewascon-“rmedbysupplyinganelectrolysiscurrentgreatenoughtoformagasbubbleattheanode.Anelectrolysiscurrentof10Awassuppliedfor6stogeneratemoreoxygenthanthesolubilitylimitinwater.ThesizeofthebubbleformedwasmeasuredasafunctionoftimeusingImageJ(NIHResearchServicesBranch).Forexperimentsinvolvingcylindroids,a100-foldlowercurrentwasusedthatdidnotleadtobubbleformation.Biotechnol.Prog.,2012,Vol.28,No.2 EstimationofmaximumgrowthrateThegrowthrateofLS174Tcellswasdeterminedbygrow-ingamonolayerofcellsonawellplateunderambientoxy-genconditions.Cellswereplatedatadensityof25,000/cmona48wellplate,andimageswereacquiredat10,17.5,25,and32.5h.Thesecellsarehighlycoherentandformclumpsonculturesurfaces,preventingdirectcellcounting.Celldensitiesweredeterminedateachtimepointbymeas-uringthetotalareaoccupiedbythecells(3).Maximumgrowthrate,,wasestimatedby“ttingcellareastoanexponentialgrowthrateexpression.OxygenmicrogradientswithincylindroidsTwogroupsofcylindroidswereinvestigated:anoxygen-atedgroupandacontrolgroupthatdidnotreceiveoxygen.Afterpositioning,oxygenatedcylindroidswereincubatedfor60hwitha“xedelectrolysisDCcurrentgeneratedwithaKeithley2400Seriessupply(Keithley,Cleveland,OH).Deionizedwaterwasusedastheelectrolyte.Theappliedelectrolysiscurrentof0.1Aproducedatheoreticaldissolvedoxygen”uxof3.3molmfromtheanode,assuming100%electrolyticef“ciency.Wehavepreviouslyshownwithasimilardevicethatmostofanappliedcurrentproducesmolecularoxygenandthattheproductionofreactiveoxygenspeciesisminimal.Controlcylindroidswereincu-batedintheabsenceofelectrolysiscurrent.Quanti“cationofcelldeathincylindroidsEthidiumhomodimer(Eth-D1;Invitrogen,Carlsbad,CA)selectivelystainscellswithpermeablemembranesandwasusedtodeterminetheextentofcelldeathincylindroids.Af-ter48hofcylindroidpreparation,existingmediumwasreplacedwithfreshmediumcontaining4MEth-D1.Fluo-rescentimageswereacquired12hafteradditionofthestainusinganOlympus(CenterValley,PA)IX71invertedepi-”uorescentmicroscopeequippedwithacolor-corrected10xPlan-APOanda20xLCPlanFI”uorescenceobjectives,respectively,andIPLab(BDBioscience,Rockville,MD)software.Forlargecylindroidsthatexceededtheboundsofasinglecameraframe(867.15),a22gridofimageswasacquiredat10magni“cationinasinglefocalplane.OnetiledimagewasformedbystitchingfourpiecestogetherusingamacroinIPLab.Red”uorescenceimageswereacquiredusinga546/10nmexcitationanda590nmlongpassemission“lter(Chroma,Rockingham,VT). Figure1.Designofthemicrometer-scaleoxygendeliverydevice.A:Schematicoftheoxygendeliverydevice.OxygenisgeneratedattheanodeanddiffusesthroughtheelectrolyteandPDMSintothecylindroid.Thecylindroidresidesinacellculturechamberandispositionedovertheanodebythepositioner.B:Themicroenvironmentswithinacylindroidareheterogeneous.Cellsneartheperipheryarehealthyandproliferating,andcellsintheinterioraredying.C:OxygenisgeneratedattheanodethroughelectrolysisbythepassageofaDCcurrent.D:Thepositioneriscomposedofaplugattachedtoahead,whichrestsonabase.TheXandYlocationisadjustedbymovingthepluglaterally,andthethicknessofthecylindroid(Z)iscontrolledbythefourscrews.Biotechnol.Prog.,2012,Vol.28,No.2 Theextentofcelldeathincylindroidswasquanti“edusingnormalizedradialintensitypro“lesofEth-D1”uores-cenceobtainedusingImageJ(NIHResearchServicesBranch).Foreachcylindroid,average”uorescenceintensityof“vevisuallyidenti“eddeadcellswasusedtode“nemaxi-mum(100%)death;averageintensityof“veviablecellswasusedtode“neminimum(0%)death.Alinearinterpola-tionbetweenthesetwopointsrelatedEth-D1”uorescenceintensitiestoextentofcelldeath.CellrearrangementincylindroidsRearrangementofdeadcellsinacylindroidwasobservedbystainingwithEth-D1.Fluorescentandtransmittedlightimageswereacquiredevery20minforaperiodof16h.Toobserveindividualcellsovertheentirecylindroid,agridof3images(433.58)wasacquiredat20magni“cationinasinglefocalplaneateachtimepointandtiledusingIPLab.Cellswithinthecylindroidweretrackedintime,anddistancesfromthecenterofthecylindroidwereestimated.Thedistanceswerenormalizedbythecurrentra-diusofthecylindroidtoaccountforgrowthandrenormal-izedbythestartinglocationsofeachcelltoquantifymovementrelativetothestartingpositions.Steady-statemathematicalmodelofthree-dimensionaloxygenpro“lesTwodistinctmodelingtechniqueswereusedtoanalyzeoxygendeliveryandcellviability.The“rsttechniquedeter-minedthree-dimensionaloxygenpro“lesincylindroidsbasedonanappliedelectrodecurrent.Thesecondone-dimensionalmodelingtechniqueanalyzedcellviabilityinthepresenceandabsenceofoxygendelivery.Thesecondmodelutilizedtheresultingoxygenpro“lesfromthe“rstmodelasinputs.Itwasnecessarytomaintainthreedimensionsinthe“rstmodeltoaccountforoxygenintheculturemediumsur-roundingtheouteredgeofthecylindroid(Figure1A).Thesteady-stateoxygenpro“lemodeldeterminedoxygenpro“lesincylindroidsbasedonthemoleculartransportequa-tion,whichbalancedoxygendiffusionwithcellularuptake. whereisthelocaloxygenconcentration;isthediffusivityofoxygeninelectrolyte,PDMS,orcylindroid;maxisthemax-imumoxygenuptakerate;andisthesaturationconstant(Table1).ThismodelwassolvedusingComsolMultiphysics(Comsol,Burlington,MA).Themodel(seeFigure1A)accountedforoxygengenerationattheelectrode,consumptionbycellswithinthecylindroid,anddiffusionthroughtheelectro-lyte(5mthick),thePDMS(10mthick)andintothecylin-droid(125mthick).ModelgeometryandboundaryconditionsaredescribedintheSupportingInformation.DynamicmodelofcellgrowthanddeathAdynamicmodelwasusedtodeterminetheeffectofox-ygenoncellgrowthanddeath.Thismodelisbasedonaspheroidgrowthmodelcreatedpreviouslyinourlaboratoryandconsistsofthreecoupledpartialdifferentialequations. @L@tgldÞL(2)¼lgLVLldLðVLVDÞ(3)Doxr2C¼ The“rstequation(Eq.2)isalivecellpopulationbal-ancethataccountsforgrowth,,death,,andconvectivesupplyoflivecellsduetotissueexpansion,).Thenumberdensityoflivecells,,isde“nedasthenumberoflivecellspersinglecellvolume(01).Thesecondequation(Eq.3)accountsfortissueexpansionandcontrac-tionascellsgrow,,anddie,),whereistheconvectivevelocityandaresinglecellvolumesofliveanddeadcells(Table1).Thisbalanceassumesthattheentirevolumewasoccupiedbyeitherliveordeadcells.Thethirdbalance(Eq.4)accountsforoxy-gendiffusioninthecylindroid,;cellularuptake(asdescribedabove);andgenerationofoxygenbytheelec-trode,.Oxygengenerationwasmodeledwithanem-piricalHill-likefunctioninwhichtheratevariedasafunctionofcylindroidradius, Theparameters,,wereadjusteduntiltheoxygenpro“leinthecylindroidmatchedthepro“lecalculatedbythesteady-statethree-dimensionaloxygenmodel(Eq.1).Thismatchingwasnecessarytointegratethethree-dimen-sionalpro“leresultsintotheone-dimensionalcellbalancemodel.Thedensityofdeadcells,,wascalculatedfromthelivecelldensity,1(6)Theratesofcellgrowth,,anddeath,,wereassumedtobedependentontheavailableoxygenlevel, lmaxgCKg:oxþC;ld¼lmaxd1 Thegrowthanddeathsaturationconstants,g,oxandd,oxrepresentthesensitivityoftheratestooxygenavailability.Themaximumgrowthrate,max,i.e.growthintheabsenceof Table1.ParametersforMathematicalModelingParameterValueReferenceDiffusivityofoxygeninelectrolyteDiffusivityofoxygeninPDMSDiffusivityofoxygenincylindroidsMaximumoxygenconsumptionratemolcellOxygenconsumptionsaturationconstantmM3VolumeoflivecellVolumeofdeadcellBiotechnol.Prog.,2012,Vol.28,No.2 nutrientlimitations,wasdeterminedexperimentallyasdescribedabove.Themodelassumesthattheeffectsofextrac-ellularmatrixandcellularmechanicalpropertieswereuni-formacrosscylindroids,andthattheeffectsofanaerobicglucosemetabolism43,44wereminimal.Valuesofthemaximumrateofcelldeath,max,andthesaturationconstants,g,oxandd,ox,weredeterminedby“ttingthemodeltotheviabilitypro-“les,asdescribedintheSupportingInformation.ResultsOptimizeddevicedesignToachievethedesigngoalofproducinglocalizedoxygenatthecenterofcylindroids,fourfeaturesweremodi“ed:thedis-tancebetweentheanodeandthecylindroid,thedesignofthechannel,thedesignofthecellchamber,andthemethodofcylindroidinsertion.Theeffectsofradialoxygendispersion,whichwidenedtheareaofoxygendelivery,wereminimizedbyreducingthedistancebetweentheelectrodeandthecylindroid.Inthe“naldesign,thethicknessoftheelectrolytechannelandthePDMS“lmwere5and10m,respectively.Atthisthick-ness,anunsupportedPDMSlayercollapsedduringcylindroidformation.Topreventcollapse,squarePDMSposts,10moneachsideandlocated50mfromeachother,wereaddedtothePDMSlayer(Figure2Ainset).Tosustaincylindroidcultureforthelengthofatypicalexperiment,achamberwasincorporatedintothecellcultureareathatcontainedenoughmediumtomaintainconstantnutrientconcentrationsfor60h(Figure1A).Cylindroidpositionwascontrolledwithapolycarbonatemicro-positioner(Figure1D),whichcontainedfouradjustmentscrewstocontrolcylindroidthickness.ThepositionercouldmovefreelyintheplaneofthePDMSsurfacetocontrolX-Yposition.Thissystemwasabletopositioncylindroidsaccuratelyovertheanode(Figure2B).Forallexperiments,theaveragedistancefromthecylindroidcentertoanodetipwas1612%ofthecylindroidradius.OxygengenerationTheoperationoftheoxygen-generatingelectrodewastestedbypassingahighelectrolysiscurrent(10A)for6s Figure2.Positioningofcylindroidsandoxygenformation.A:Transmittedlightimageofthechipshowingthelocationoftheelectrolytereservoir,cathode,andtheanode.Scalebaris1mm.Inset:Enlarge-mentofthetipoftheanodewhereoxygenisgeneratedandsquarePDMSsupportposts.Scalebaris50m.B:Transmittedlightimageofacylin-droidpositionedovertheanode.Scalebaris100m.C:Transmittedlightimagesofoxygenbubblegeneratedattheanode.Scalebaris100D:Sizeofthegeneratedbubbleasafunctionoftime.Thesizeofthebubbleincreasedwhentheelectrolysiscurrentwasonanddecreasedafteritwasturnedoff.Inset:Whentheelectrolysiscurrentwasappliedtheoxygengenerationratewasconstant.Biotechnol.Prog.,2012,Vol.28,No.2 togenerateanoxygenbubble.Whentheelectrolysiscurrentwason,thebubblelinearlyincreasedinsize(Figure2C,D).Aftertheelectrolysiscurrentwasturnedoff,thebubbleshrunklogarithmicallyandeventuallydisappeared(Figures2C,D).Thegenerationpartofthecurvewaslinearwithaslopeof8970(Figures2Dinset),con“rmingthattheoxygengenerationratewasconstantfora“xedelec-trolysiscurrent.Theshrinkageofthebubblewasduetooxy-gendissolvingintotheelectrolyte,anddiffusionintothecellcultureareathroughthePDMSroofofthechannel.A100timeshigherelectrolysiscurrentwasusedforthisexperi-menttoforcetheformationofabubble.ThePDMSsupportposts,mostlikely,preventedthebubblefrombeingcenteredovertheanode.Whendeliveringoxygentotissue,however,electrolyticallygeneratedoxygenremaineddissolvedandlocalizedovertheanode.CellgrowthrateThemaximumrateofcellgrowthwasdeterminedfromcellcultureexperiments.Totalareaoccupiedbyplatedcells(representedbydottedcircles)increasedoverthe32.5hincubationperiod(Figures3A,B).By“ttingcellareastoanexponentialgrowthexpression,wasfoundtobe0.0275EffectsofoxygenationoncelldeathintissueSupplyingoxygendecreasedtheextentofdeathincylin-droidcenters.Deathwasmeasuredusinga”uorescencestain(Eth-D1)speci“cfordeadcellswithpermeablemembranes(Figures4A…F).Whencontrol,non-oxygenatedcylindroidswereincubatedfor60h,stainingwasobservedprimarilyinthecentralregion(Figure4B).Inallobservedcylindroids,thepercentageofdeadcellssteadilydecreasedfromthecen-tertotheperiphery(Figure4C,4).Inthecentralannu-lus(0…10%ofthenormalizedradius),60%ofcellsweredead,whichwassigni“cantlygreater(0.05)thanthe5%deadcellsintheouterannulus(90…100%ofthenor-malizedradius).Theorganizationofdeadcellswasreversedinoxygenatedcylindroids(Figure4E).Minimalstainingwasobservedinthecenters,andringsofdeadcellswereobservedattheedgeofallcylindroids(Figure4E).Inalloxygenatedcylin-droids,thepercentageofdeadcellssteadilyincreasedfromthecentertotheperiphery(Figure4F;4).Inthecentralannulus,3%ofcellsweredead,whichwassigni“cantlyless0.05)thanthe29%deadcellsintheouterannulus.Overall,oxygenationdecreasedtheextentofcelldeathincylindroids.Theaveragepercentageofdeadcellsinoxygen-atedcylindroids(122%)wassigni“cantlylessthanincontrolcylindroids(310.05;Figure4G).Oxy-genationalsodramaticallyalteredwheredeadcellswerelocated.Forallcentralannulifrom0to0.5ofthenormal-izedradius,thereweresigni“cantlymoredeadcellsincon-trolthanoxygenatedcylindroids(0.05;Figure4H).Inperipheralannulifrom0.8to1.0ofthenormalizedradius,oxygenatedcylindroidshadsigni“cantlymoredeadcellsthancontrolcylindroids(0.05;Figure4H).Forcontrolcylindroids,thepercentageofdeadcellsintheinnerannuliwassigni“cantlygreaterthantheouterannuli(Figure4H).Thiswasreversedforoxygenatedcylindroids;thepercentageofdeadcellsintheouterannuliwassigni“-cantlygreaterthantheinnerannuli(0.05;Figure4H).RearrangementofcellsUsingthesame”uorescencestain(Eth-D1),therearrange-mentofindividualdeadcellswastrackedduringoxygen-ation.Duringtheperiodofobservation(2.5to18.5h),severaldeadcellswerepushedtowardtheouteredge(Fig-ures5A,B).Theonlyobservedmotionofcellswasawayfromthecylindroidcenter.Manydeadcellsremainedsta-tionary,andnomotionwasseentowardthecenter.Relativetotheirstartingposition,theaveragerearrangedcellchangedpositionby6%over16h(0.01;Figure5C).Theseval-ueswerenormalizedbythegrowingouteredgeofthecylin-droidtoeliminatetheeffectsofcylindroidgrowth.Theaveragevelocityofrearrangeddeadcellswas1.8h,whichwassigni“cantlygreaterthanzero(Becausethesecellsweredeadandcouldnotactivelymigrate,itispossiblethattheirobservedmotionwascausedbychemotaxisofneighboringlivecells(Figure5D).Steady-stateoxygenconcentrationpro“lesOxygenpro“lesfora500-m-diametercylindroidwerecalculatedusingasteady-statemathematicalmodelofdiffusionandconsumption(Eq.1;Figures6A,B).Themodelaccountedforoxygensuppliedbytheanodeandthecellcul-turemediumattheouteredge(Figure6A).RadialdiffusionintheelectrolyteandPDMSlayerswasminimal(Figure6A),andtheoxygensupplywastightlypositionedatthecylindroidcenter(Figure6B).Asdesired,oxygengradientsdecreasedawayfromtheanodeatthecylindroidcenter.Intheabsenceofelectrolysiscurrent,thecylindroidcorewashypoxicduetoconsumptionofoxygenbythecells(Figure6B). Figure3.Estimationofmaximumcellgrowthrate.A:Transmittedlightimagesofcells10,17.5,25,and32.5hafterplating.Dottedlinesrepresenttotalareaoccupiedbycells.B:Totalareaoccupiedbycellsincreasedexponentially.Biotechnol.Prog.,2012,Vol.28,No.2 Whenoxygenwassuppliedusinga0.1Aelectrolysiscurrent,thecorebecamemoreoxygenated(Figures6A,B).Theouteredgesofbothcontrolandoxygenatedcylindroidscontainedsimilarlevelsofoxygen(Figure6B).Basedonthismodel,theoxygenmicroenvironmentsincontrolandoxygenatedcylindroidswereconsiderablydifferentatthecenterbutalmostidenticalattheouteredge.DependenceofcellgrowthanddeathonlocaloxygenationAdynamicmathematicalmodelwasusedtodeterminehowcylindroidgrowthanddeathrespondedtooxygengra-dients(Eqs.2…7).Threeparameters(,thesensi-tivitiesofcellgrowthanddeathtooxygen,and,themaximumrateofcelldeath)werecalculated(Table2)by“ttingsimulatedviabilitypro“les(Figures6D,E)tomeas-uredviabilitypro“les(Figures4C,F).Thesimulatedviabilitypro“leshadsimilarshapestotheexperimentalpro“les:deathwashighinthecentersofcontrolcylindroids(Figures4C,6D)andlowinthecentersofoxygenatedcylindroids(Figures4F,6E).Similarly,deathwaslowattheedgeofcon-trolcylindroidsandshiftedtowardtheedgeforoxygenatedcylindroids.Themaximumdeathrate,,wasfoundtobeequaltothemeasuredmaximumgrowthrate,(0.0275hrindicatingthatcelldeathbyoxygendeprivationisagradualprocessoperatingoversimilartimescales(hours)asgrowth(Table2).Theestimatedvaluesofoxygensensitivityparam-,indicatehowgrowthanddeathrates)varywithoxygenconcentrationwithintissue(Figure6C).Therateofcellgrowth,,isuniformatitshighestvalueexceptatverylowoxygenconcentrations.Therateofcelldeath,,hasastrongerdependenceonoxygenconcentrationanddecreasesgraduallyastheconcentrationofoxygenincreases.Thereductionofdeathinoxygenatedcylindroids(Figure4)showsthatoxygenactsastheprimaryfactorcontrollingcellviabilityintumortissue.Incontrolcylindroidsconsider-ablecentralizednecrosiswasobserved,aphenomenonthatiswellknowntooccurinbothcylindroidsandsphe-roids.Whenoxygenwassuppliedtothecenterofcylin-droids,celldeathwasalmostentirelyprevented(Figure4).Thisresponsecouldhavebeenmediatedbychangesinintra-cellularpHandmetabolism,forexample.However,becauseoxygensupplywastheonlydifference,theseresultssuggestthatoxygencontrolscellviability,eveninthepresenceofothermetabolicgradients. Figure4.Effectsofoxygenoncelldeathintissue.A…F:Transmittedlight,Eth-D1”uorescence,andradialviability(%dead)pro“lesofcontrol(A,B,C)andoxygenated(D,E,F)cylindroids.Imageswereacquiredafter60hofincubationwithcontinuousoxygendelivery(treatments)andnooxygendelivery(controls).G:Oxygenationdecreasedoveralltissuedeathafter60hofincubation(*,0.05).H:Comparisonofcelldeathbetweenoxygenatedandcontrolcylindroidsintheinner50%andouter20%ofcylindroids.Oxygenationpreventedcelldeathintheinnerregionsofcylindroidsandresultedinaccumulationofdeadcellsattheperipheries(*,Biotechnol.Prog.,2012,Vol.28,No.2 Oxygenaffectedcellviabilitybybothdecreasingdeathandincreasinggrowth(Figure6).Themathematicalmodelofcylindroidgrowthdecoupledtheseeffectsbygeneratingindependentestimatesofcellulargrowthanddeathratesasfunctionsofoxygenconcentration(Figure6C).Thesatura-tionconstantsand(Eq.7)representthesensitiv-itiesofgrowthanddeathtooxygenavailability.Thelowervalueofcomparedtoindicatesthatgrowthislesssensitivetooxygenthandeath.Oxygen,orthelackofit,affectscelldeathacrosstheentirerangeofoxygenconcen-trationsincylindroids,butgrowthisuniformovermostofthisrange(Figure6C).Astheoxygenconcentrationdecreases,thedeathrateslowlyincreasesbuttherateofgrowthremainshighuntiltheoxygenconcentrationbecomesverylow.Thisinterplayisimportantbecauseitsuggeststhatinoxygen-limitedtumorregions,celldeathwilloccurbeforegrowthceases.Thepresenceofgrowthatlowoxygencon-centrationsexplainshowsmallpopulationsofproliferatingcancercellscanpersistinhypoxicenvironmentsandeventu-allybecomeresistanttohypoxia-inducedapoptosis.selectionwouldoccurdespitethefactthatmostcellsinhypoxicregionsarecommittedtodeath.Inadditiontopreventingdeath,supplyingoxygenalteredthedistributionofdeadcellsincylindroids(Figures4,5).Controlcylindroidshaddeadcoresandviableperipheries,whereasoxygenatedcylindroidshadviablecoresandringsofdeadcellsattheperipheraledges(Figure4).Theobservedmovementofdeadcells(Figure5)indicatesthatthisreorganizationinresponsetooxygenationwascausedbyindividualcellmigration.Itisunlikelythattheenvironmentattheedgecausedthedeathseeninoxygenatedcylindroids(Figure4E),becausedeathwasnotobservedattheedgeofcontrolscylindroids(Figure4B).Theoutsideedgesofallcylindroidswerewellsuppliedwithoxygenandnutrients.Thesteady-stateoxygendiffusionmodel(Eq.1)predictedthattheoxygenconcentrationintheouterregionissimilarincontrolandoxygenatedcylindroids(Figure6B).Inaddi-tion,theconcentrationsofothernutrientsneartheedgeswouldnotbeaffectedbysupplyingoxygentothecenters.Twodifferentmechanismscouldhavecontributedtotheobservedcellularrearrangement:passivedisplacementcausedbycellviabilitygradientsandactivecellchemotaxis.Passivecellgrowthisaconvectivevelocitycausedbytheexpansionofgrowingcellsandcontractionofdyingcells.Ifpassiveviabilitygradientswerethepredominantmecha-nism,deadcellswouldhavebeenpushedfromthecenterasthevolumeofproliferatingcellsexpandedduetothesupplyofoxygen.Thisrelocationwouldhaveresultedintheobser-vationofcelldeathalongabandbetweenthecenterandtheedgewheretheoxygenconcentrationwaslowest(similartothemaximuminFigure6E).Deathwouldbeminimalatthecenterandtheedgewhereoxygenwasreadilyavailable.Ifactivechemotaxiswerethepredominantmechanism,viablecellsattheperipherywouldhavesensedtheoxygengradientandmovedtowardthecylindroidcenter(solidarrows,Figure5D).Thisinternalizationwouldhavedisplaceddeadcells,whichwereunabletomigrate,towardtheoutsideedge(dashedarrows,Figure6D),creatingtheobservedringofdeadcells.Activecellmigrationinspheroidshasbeenobservedpreviously,andthetime-scaleforrearrangement Figure5.Oxygenmediatedcellrearrangement.A,B:LocationofadeadcellwithinanEth-D1stainedcylindroid6.5and10.5hafterthestartofoxygenation.C:Percentchangeinlocationofsixcellsrelativetotheirstartingpositions.D:Suggestedmechanismofcellrearrangementinvolvingactivemotionoflivecells(solidarrows)inresponsetooxygengradients,causingpassivemotionofdeadcells(dashedarrows).Biotechnol.Prog.,2012,Vol.28,No.2 wasmeasuredtobe48…72h,similartothe60htime-scaleinvestigatedhere.Theobservationthatdeadcellsweremostlylocatedattheouteredge(Figure4)ratherthaninanintermediaryband(Figure6E)suggeststhatactivechemo-taxiswastheprincipalcauseofcellularrearrangement.Theredistributionofcellpopulationsinresponsetooxy-genationsuggeststhatchangesinoxygengradientswithintumorsmayinducemetastasisbyrearrangingcellpopula-tions.Tumorshaveheterogeneousanddynamicoxygenmicroenvironments;thechaoticvascularstructuretypicalincreatesvesselswithbothhighandlowoxygencon-centrations.Angiogenesis,vascularstasisand”uctuationsinredbloodcell”ux,constantlychangethelocationofoxygensourcesandtheoxygenationstateofcellpopulations.Theobservationthatdeadcellslocalizetotheperipheriesofoxygenatedcylindroidssuggeststhatchangingoxygenmicroenvironmentspromotecontinualtissueremodeling.Viablecellsthatchemotaxtowardwell-oxygenatedvesselswouldpushdyingcellstowardless-oxygenatedvessels.Proximitytovesselswouldpromoteintravasationandcellsheddingintothebloodstream,twostepsintheformationofmetastasis.Consequently,astheproportionofdefunctoroxygenleanvesselsincreases,thelikelihoodofmetastasiswouldincrease.Thismechanismmayexplainthepreviouslyobservedcorrelationbetweenhypoxiaandmetastaticdisease.ConclusionsWehavedevelopedadevicetodelivermicrometer-scaleoxygentothree-dimensionaltumortissueinvitro.Tothebestofourknowledge,thisisthe“rstreportofhighspatialresolutioncontroloveroxygengradientswithinscaffold-freethree-dimensionaltissue.Usingaspeciallydesignedposi-tioner,cylindroidswereplacedoverelectrodeswithahighdegreeofaccuracy.Thisplacementenabledprecisesupplyofoxygentothenecroticcenterofcylindroids.Resultsobtainedusingthedeviceshowedthatoxygenationreducedcelldeathintumortissue.Amathematicalmodelofcell Figure6.Mathematicalmodelresults.A:Steadystateoxygenconcentrationpro“lethroughaverticalslicealongthecylindroiddiameter.Thebrightspotatthecenterisaregionofhighoxygenconcentrationnearthesource,andtheconcentrationdecreasesawayfromthecenter.B:Steadystateradialoxygenconcentrationpro“lesatthebaseofcontrol(0A)andoxygenated(0.1A)cylindroids.C:Calculatedgrowthrate,,anddeathrate,,ofcellsincylindroidsasafunc-tionofoxygenconcentration,asdeterminedby“ttingtoviabilitypro“les.D,E:Simulatedoxygenconcentrationpro“lesincylindroidsforcontrol(D)andtreatments(E)usingthedynamicmodelofcellgrowthanddeath. Table2.Oxygen-DependentGrowthandDeathParametersParameterValueMaximumdeathrate0.0275hrOxygengrowthsaturationconstantOxygendeathsaturationconstantBiotechnol.Prog.,2012,Vol.28,No.2 viabilityshowedthattherateofcelldeathwasmoresensi-tivetooxygenconcentrationthantherateofcellgrowth.Theresultsshowedthatoxygenisaprimaryfactorcontrol-lingcellviabilityintumorbecauseothernutrientgradientswerenotdisruptedinoxygenatedcylindroid.Inthedevice,oxygencausedconsiderablecellularrearrangement,anddeadcellswereobservedmovingtowardthecylindroidedge.Thisobservationindicatesthatoxygengradientsarecapableofrestructuringtissues,amechanismthatcouldleadtometasta-sisformation.Relocationofdyingcellsbyoxygengradientsmayexplainthepreviouslyobservedcorrelationbetweenhy-poxiaandmetastaticdisease.Theseresultsdemonstratetheimportanceofoxygengradientsincontrollingcellviabilityandrearrangement,andtheirroleincancerprogression.AcknowledgmentTheauthorsgratefullyacknowledge“nancialsupportfromtheNationalInstitutesofHealth(GrantNos.1R21CA112335-01Aand1R01CA120825-01A1),theNationalScienceFounda-tion(GrantNo.DMI-0531171),andtheEugeneM.IsenbergAwardforBhushanJ.Toley.LiteratureCited1.OsinskyS,ZavelevichM,VaupelP.Tumorhypoxiaandmalig-nantprogression.ExpOncol.2.TatumJL,KelloffGJ,GilliesRJ,ArbeitJM,BrownJM,ChaoKS,ChapmanJD,EckelmanWC,FylesAW,GiacciaAJ,HillRP,KochCJ,KrishnaMC,KrohnKA,LewisJS,MasonRP,MelilloG,PadhaniAR,PowisG,RajendranJG,RebaR,Rob-insonSP,SemenzaGL,SwartzHM,VaupelP,YangD,CroftB,HoffmanJ,LiuG,StoneH,SullivanD.Hypoxia:importanceintumorbiology,noninvasivemeasurementbyimaging,andvalueofitsmeasurementinthemanagementofcancertherapy.IntJRadiatBiol.3.HelmlingerG,YuanF,DellianM,JainRK.InterstitialpHandpO2gradientsinsolidtumorsinvivo:high-resolutionmeasure-mentsrevealalackofcorrelation.NatMed.4.VaupelP,MayerA.Hypoxiaincancer:signi“canceandimpactonclinicaloutcome.CancerMetastasisRev.5.BrizelDM,SibleyGS,ProsnitzLR,ScherRL,DewhirstMW.Tu-morhypoxiaadverselyaffectstheprognosisofcarcinomaoftheheadandneck.IntJRadiatOncolBiolPhys.1997;38:285…289.6.VaupelP,MayerA,BriestS,HockelM.Hypoxiainbreastcan-cer:roleofblood”ow,oxygendiffusiondistances,andanemiainthedevelopmentofoxygendepletion.AdvExpMedBiol.7.HuangL,AoQ,ZhangQ,YangX,XingH,LiF,ChenG,ZhouJ,WangS,XuG,MengL,LuY,MaD.Hypoxiainducedpacli-taxelresistanceinhumanovariancancersviahypoxia-induciblefactor1alpha.JCancerResClinOncol.2010;136:447…456.8.BrizelDM,ScullySP,HarrelsonJM,Lay“eldLJ,BeanJM,ProsnitzLR,DewhirstMW.Tumoroxygenationpredictsforthelikelihoodofdistantmetastasesinhumansofttissuesarcoma.CancerRes.9.AckerstaffE,ArtemovD,GilliesR.J,BhujwallaZM.Hypoxiaandthepresenceofhumanvascularendothelialcellsaffectprostatecancercellinvasionandmetabolism.10.KimBJ,ForbesNS.Single-cellanalysisdemonstrateshownu-trientdeprivationcreatesapoptoticandquiescentcellpopula-tionsintumorcylindroids.BiotechnolBioeng.11.KimBJ,ForbesNS.Fluxanalysisshowsthathypoxia-induci-ble-factor-1-alphaminimallyaffectsintracellularmetabolismintumorspheroids.BiotechnolBioeng.12.FingerEC,GiacciaAJ.Hypoxia,in”ammation,andthetumormicroenvironmentinmetastaticdisease.CancerMetastasisRev.13.MinchintonAI,TannockIF.Drugpenetrationinsolidtumours.NatRevCancer.14.TredanO,GalmariniCM,PatelK,TannockIF.Drugresistanceandthesolidtumormicroenvironment.JNatlCancerInst.15.TannockIF.Tumorphysiologyanddrugresistance.CancerMe-tastasisRev.16.CasciariJJ,SotirchosSV,SutherlandRM.Glucosediffusivityinmulticellulartumorspheroids.CancerRes.17.HelmlingerG,SckellA,DellianM,ForbesNS,JainRK.Acidproductioninglycolysis-impairedtumorsprovidesnewinsightsintotumormetabolism.ClinCancerRes.18.ZemanEM,CalkinsDP,ClineJM,ThrallDE,RaleighJA.Therelationshipbetweenproliferativeandoxygenationstatusinspontaneouscaninetumors.IntJRadiatOncolBiolPhys.19.VenkatasubramanianR,HensonMA,ForbesNS.IncorporatingenergymetabolismintoagrowthmodelofmulticellulartumorJTheorBiol.20.SeagrovesTN,RyanHE,LuH,WoutersBG,KnappM,Thi-baultP,LaderouteK,JohnsonRS.TranscriptionfactorHIF-1isanecessarymediatorofthepasteureffectinmammaliancells.MolCellBiol.21.KedrinD,GligorijevicB,WyckoffJ,VerkhushaVV,CondeelisJ,SegallJE,vanRheenenJ.Intravitalimagingofmetastaticbehaviorthroughamammaryimagingwindow.NatMethods.22.HanahanD,WeinbergRA.Thehallmarksofcancer.23.WyckoffJB,JonesJG,CondeelisJS,SegallJE.Acriticalstepinmetastasis:invivoanalysisofintravasationattheprimarytu-CancerRes.24.KrishnamacharyB,Berg-DixonS,KellyB,AganiF,FeldserD,FerreiraG,IyerN,LaRuschJ,PakB,TaghaviP,SemenzaGL.Regulationofcoloncarcinomacellinvasionbyhypoxia-induci-blefactor1.CancerRes.25.RyuMH,ParkHM,ChungJ,LeeCH,ParkHR.Hypoxia-in-duciblefactor-1alphamediatesoralsquamouscellcarcinomainvasionviaupregulationofalpha5integrinand“bronectin.BiochemBiophysResCommun.26.CroninPA,WangJH,RedmondHP.Hypoxiaincreasesthemet-astaticabilityofbreastcancercellsviaupregulationofCXCR4.BMCCancer.27.MarchesiF,MontiP,LeoneBE,ZerbiA,VecchiA,PiemontiL,MantovaniA,AllavenaP.Increasedsurvival,proliferation,andmigrationinmetastatichumanpancreatictumorcellsexpressingfunctionalCXCR4.CancerRes.28.DemirR,DimmlerA,NaschbergerE,DemirI,PapadopoulosT,MellingN,SturzlM,HohenbergerW.Malignantprogressionofinvasivetumourcellsseeninhypoxiapresentanaccumula-tionofbeta-catenininthenucleusatthetumourfront.ExpMol29.IndovinaP,RainaldiG,SantiniMT.Hypoxiaincreasesadhe-sionandspreadingofMG-63three-dimensionaltumorsphe-AnticancerRes.30.DorieMJ,KallmanRF,CoyneMA.EffectofcytochalasinB,nocodazoleandirradiationonmigrationandinternalizationofcellsandmicrospheresintumorcellspheroids.ExpCellRes.31.DorieMJ,KallmanRF,RapacchiettaDF,VanAntwerpD,HuangYR.Migrationandinternalizationofcellsandpolysty-renemicrosphereintumorcellspheroids.ExpCellRes.32.BertuzziA,FasanoA,Gandol“A.Afreeboundaryproblemwithunilateralconstraintsdescribingtheevolutionofatumorcordunderthein”uenceofcellkillingagents.SiamJMath33.BertuzziA,FasanoA,Gandol“A.Amathematicalmodelfortumorcordsincorporatingthe”owofinterstitial”uid.ModelsMethodsApplSci.34.McElwainDL,PettetGJ.Cellmigrationinmulticellspheroids:swimmingagainstthetide.BullMathBiol.35.BaracchiniO,SherrisJC.ThechemotacticeffectofoxygenonJPatholBacteriol.Biotechnol.Prog.,2012,Vol.28,No.2 36.EngelmannTW.Neuemethodezuruntersuchungdersauerstof-fausscheidungpfanzlicherandtierischerorganismen.ArchGesamtePhysiolMenschenTiere.37.AllenJW,BhatiaSN.Formationofsteady-stateoxygengra-dientsinvitro:applicationtoliverzonation.BiotechnolBioeng.38.OppegardSC,NamKH,CarrJR,SkaalureSC,EddingtonDT.Modulatingtemporalandspatialoxygenationoveradherentcel-lularcultures.PLoSOne.39.DerdaR,LaromaineA,MammotoA,TangSK,MammotoT,IngberDE,WhitesidesGM.Paper-supported3Dcellculturefortissue-basedbioassays.ProcNatlAcadSciUSA.2009;106:18457…18462.40.ParkJ,BansalT,PinelisM,MaharbizMM.Amicrosystemforsensingandpatterningoxidativemicrogradientsduringcellcul-LabChip.41.KasinskasRW,ForbesNS.Salmonellatyphimuriumchemotaxandproliferateinheterogeneoustumortissueinvitro.BiotechnolBioeng.42.AmbrosiD,PreziosiL.Celladhesionmechanismsandstressrelaxationinthemechanicsoftumours.BiomechModelMecha-43.SmallboneK,GatenbyRA,GilliesRJ,MainiPK,GavaghanDJ.Metabolicchangesduringcarcinogenesis:potentialimpactoninvasiveness.JTheorBiol.44.GatenbyRA,GawlinskiET.Areaction-diffusionmodelofcan-cerinvasion.CancerRes.45.SutherlandRM.Importanceofcriticalmetabolitesandcellularinteractionsinthebiologyofmicroregionsoftumors.46.SutherlandRM.Cellandenvironmentinteractionsintumormicroregions:themulticellspheroidmodel.47.YaoK,GietemaJA,ShidaS,SelvakumaranM,FonroseX,HaasNB,TestaJ,ODwyerPJ.Invitrohypoxia-condi-tionedcoloncancercelllinesderivedfromHCT116andHT29exhibitalteredapoptosissusceptibilityandamoreangiogenicpro“leinvivo.BrJCancer.48.HockelM,SchlengerK,HockelS,VaupelP.Hypoxiccervicalcancerswithlowapoptoticindexarehighlyaggressive.49.VaupelP,KallinowskiF,OkunieffP.Blood”ow,oxygenandnutrientsupply,andmetabolicmicroenvironmentofhumantumors:areview.CancerRes.50.DewhirstMW.Intermittenthypoxiafurtherstherationaleforhypoxia-induciblefactor-1targeting.CancerRes.51.KimuraH,BraunRD,OngET,HsuR,SecombTW,Papahad-jopoulosD,HongK,DewhirstMW.Fluctuationsinredcell”uxintumormicrovesselscanleadtotransienthypoxiaandreoxygenationintumorparenchyma.CancerRes.52.ButlerTP,GullinoPM.Quantitationofcellsheddingintoeffer-entbloodofmammaryadenocarcinoma.CancerRes.ManuscriptreceivedJun.20,2011,andrevisionreceivedNov.1,2011.Biotechnol.Prog.,2012,Vol.28,No.2