Co-culture of human induced pluripotent stem cell-derived retinal pigment epithelial cells and endothelial cells on double collagen-coated honeycomb films

(1)

Tampere University of Technology

Co-culture of human induced pluripotent stem cell-derived retinal pigment epithelial cells and endothelial cells on double collagen-coated honeycomb films

Citation

Rebelo Calejo, T., Vuorenpää, H., Vuorimaa-Laukkanen, E., Kallio, P., Aalto-Setälä, K., Miettinen, S., ... Juuti- Uusitalo, K. (2020). Co-culture of human induced pluripotent stem cell-derived retinal pigment epithelial cells and endothelial cells on double collagen-coated honeycomb films. Acta Biomaterialia, 101, 327-343.

https://doi.org/10.1016/j.actbio.2019.11.002 Year

2020

Version

Publisher's PDF (version of record)

Link to publication

TUTCRIS Portal (http://www.tut.fi/tutcris)

Published in Acta Biomaterialia

DOI

10.1016/j.actbio.2019.11.002

License CC BY-NC-ND

Take down policy

If you believe that this document breaches copyright, please contact cris.tau@tuni.fi, and we will remove access to the work immediately and investigate your claim.

Download date:19.02.2021

(2)

ContentslistsavailableatScienceDirect

Acta Biomaterialia

journalhomepage:www.elsevier.com/locate/actbio

Full length article

Co-culture of human induced pluripotent stem cell-derived retinal pigment epithelial cells and endothelial cells on double

collagen-coated honeycomb ﬁlms

Maria Teresa Calejo

^a^,^∗

, Jaakko Saari

^a

, Hanna Vuorenpää

^a

, Elina Vuorimaa-Laukkanen

^b

, Pasi Kallio

^a

, Katriina Aalto-Setälä

^a^,^c

, Susanna Miettinen

^a^,^d

, Heli Skottman

^a

,

Minna Kellomäki

^a

, Kati Juuti-Uusitalo

^a^,^∗

aFaculty of Medicine and Health Technology, Tampere University, Tampere, Finland

bFaculty of Engineering and Natural Sciences, Tampere University, Tampere, Finland

cHeart Hospital, Tampere University Hospital, Tampere, Finland

dResearch, Development and Innovation Centre, Tampere University Hospital, Tampere, Finland

a rt i c l e i n f o

Article history:

Received 9 August 2019 Revised 25 October 2019 Accepted 1 November 2019 Available online 8 November 2019 Keywords:

Polylactide Breath ﬁgures Co-culture

hiPSC-endothelial cells hiPSC-RPE

a b s t r a c t

Invitrocell culturemodelsrepresentingthephysiologicaland pathologicalfeaturesoftheouterretina are urgently needed. Artificial tissue replacements for patients suffering from degenerative retinal diseases aresimilarly ingreat demand.Here, we developed aco-culture systembased solelyon the use of human induced pluripotent stem cell (hiPSC)-derived cells. For the first time, hiPSC-derived retinalpigmentepithelium (RPE)and endothelialcells(EC)wereculturedonoppositesidesofporous polylactidesubstratespreparedbybreathfigures(BF),wherebothsurfaceshadbeencollagen-coatedby Langmuir–Schaefer(LS)technology.Smallmodificationsofcastingconditionsduringmaterialpreparation allowedtheproductionoffree-standingmaterialswithdistinctporosity,wettabilityandiondiffusionca- pacity.Completeporecoveragewasachievedbythecollagencoatingprocedure,resultinginadetectable nanoscaletopography.Primaryretinalendothelialcells (ACBRI181)and umbilical cordveinendothelial cells(hUVEC)wereutilisedasECreferences.Mono-culturesofallECswerepreparedforcomparison.All testedmaterialssupportedcellattachmentandgrowth.Inmono-culture,propertiesofthematerialshad amajoreffectonthegrowthofallECs.Inco-culture, thepresenceofhiPSC-RPEaffectedtheprimary ECsmoresignificantlythanhiPSC-EC.Inconsistency,hiPSC-RPEwerealsolessaffectedbyhiPSC-ECthan bytheprimaryECs.Finally,ourresultsshowthatthe modulationoftheporosityofthe materialscan promoteorpreventECmigration.

Inshort,weshowedthatthebehaviourofthecellsishighlydependentonthethreemainvariables ofthestudy:the presenceofasecondcelltypeinco-culture, thesource ofendothelialcellsand the biomaterialproperties.ThecombinationofBFand LSmethodologies isapowerfulstrategytodevelop thinbutstablematerialsenablingcellgrowthandmodulationofcell-cellcontact.

Statementofsigniﬁcance

Artificialblood-retinalbarriers(BRB),mimickingtheinterfaceatthebackoftheeye,areurgentlyneeded asphysiologicalanddiseasemodels,andfortissuetransplantationtargetingpatientssufferingfromde- generativeretinal diseases. Here,wedeveloped anew co-culturemodel basedonthin, biodegradable porousfilms,coatedonbothsideswithcollagen,oneofthemaincomponentsofthenaturalBRB,and cultivatedendothelialandretinalpigmentepithelialcellsonoppositesidesofthefilms,formingathree- layerstructure.Importantly,ourhiPSC-ECandhiPSC-RPEco-culturemodelisthefirsttoexclusivelyuse humaninducedpluripotentstemcellsascellsource,whichhavebeenwidelyregardedasanpractical candidatefortherapeuticapplicationsinregenerativemedicine.

ThisisanopenaccessarticleundertheCCBY-NC-NDlicense.

(http://creativecommons.org/licenses/by-nc-nd/4.0/)

∗ Corresponding authors.

E-mail address: teresa.rebelocalejo@tuni.ﬁ(M.T. Calejo).

https://doi.org/10.1016/j.actbio.2019.11.002

( http://creativecommons.org/licenses/by-nc-nd/4.0/ )

(3)

1. Introduction

Retinal degenerative diseases are estimated to affect millions ofpeople worldwide,with numbersincreasing every yeardueto theincreasedlifeexpectancyandgrowthoftheworld population.

Age-relatedmaculardegeneration(AMD)alone,aleadingcauseof irreversibleblindness, isestimatedto affect30–50million people globally[1].Recentstudiessuggestadifferentialroleofthechori- ocapillarisinAMD,wherebreakdownofthechoroidalvasculature precedestheprogressivedegenerationofthethinretinalpigment epithelium(RPE)[2,3].Thisthinmonolayerofcellsatthebackof theeyeisindirectcontactwiththephotoreceptors andperforms anumberofessentialrolesfortheir survivalandfunction.In the outerretina,theRPEisfurthersittingontheBruch’smembrane,a multilayercollagen-andelastin-richextracellularmatrix,thatsep- arates the RPE from the blood capillaries of the choroid [4]. As AMDprogresses,theseinterfacesalsobecomedamaged,disrupting theretinalhomoeostasisandleadingtoseverevisionloss.Insuch cases,recovery of the lost vision may only be achieved by reti- naltransplantation,butthepooravailabilityofallografts,together withthehighriskofimmunerejection,makethisaratherlimited solution[5].

Artiﬁcial tissue-engineered blood-retinal barriers (BRB), mim- icking the RPE-choroid interface, hold high promise in cell replacementtherapy forretinal regeneration. On theother hand, appropriate in vitro models of the BRB are urgently needed to understandthephysiologicalandpathologicalaspectsoftheouter retina,especiallyinatimewhenawarenessconcerningtheethical aspectsof invivo animalstudies isbecoming globallyspread[6]. In spite of the complexity of the BRB, most in vitro models so farare simple anduse RPEmono-cultures tostudy RPEfunction andmechanisms of retinal diseases [6,7]. RPE mono-cultures are hardly representative of the complex physiology of the outer retina, where the interactions between RPE cells and choroidal endothelialcells(EC)arecriticalforthehomoeostaticsecretionof angiogenicandangiostaticfactors [8].Invivostudies haveshown thatinteractionsbetweenRPEandECsarecriticalforthesurvival of the choriocapillaris [9]; on the other hand, the up-regulated secretion of vascular endothelial growth factor (VEGF) by RPE is responsible by the pathogenesis of choroidal neovascularisation, which is one of the most important complications of impairing eyeconditionssuchasAMD[10].

Examples ofco-culturesystems canbe foundintheliterature, where RPE and EC have been cultured together as a cell mix- ture[11],orseparatedbycommercialTranswell inserts[8,12–14], a hydrogel [15] or even an amniotic membrane [16]. The mem- branesthataretypicallyused,however,arereadilyavailablemate- rials,andlittleornoregardisgiventothemodulationofporosity and/or permeabilityacross the material. Inconsistency, synthetic biodegradable polymer membranes have hardly been studied in co-cultureofRPEandEC,inspiteoftheexistenceofmultipletech- niquesallowingtheireasypreparation.

The breath figure(BF) method, inparticular, isa simple,low- cost,solventcastingmethodtoprepareporousfilms,characterised by a highlyordered hexagonal array of pores atthe surfacethat resemblesa honeycomb[17].Akey factoristhepresenceofhigh humidityconditions during film preparation, since the pores are createdby thecondensationofthewaterdropletsonto theevap- orating polymer solution. The evaporation of both the solvent andthewaterdropletsleavesbehindsolid filmswithhexagonally packedpores,i.e.theso-calledhoneycombfilms[17].Themethod is also versatile, in that features such as pore size, pore shape, film thickness, permeabilityand surfacewettability can be mod- ulated by simply changing the variables of the process, such as polymertypeandconcentration,castingvolume,solventtypeand relativehumidity[17,18].Inourprevious work, we havesuccess-

fullydemonstratedthepotentialofbiodegradablehoneycombfilms supporting the adhesion, growthand maturationof human stem cell-derivedRPE(hESC-RPE)inmono-culture[19–21].Recently,we have also described, for the first time, the deposition of a thin but contiguouslayer of aligned collagenfibres onto the top sur- faceof polylactidehoneycombfilms, usingtheLangmuir-Schaefer technique(LS).TheLangmuir–Blodgett(LB)/LStechniqueisknown asanelegant meansforthefabricationofhighlyorganisedstruc- tureswithmolecularlevelprecision,whichcanbetransferredonto a solid substratebyvertical orhorizontaldeposition, respectively [22].EventhoughthefirststudiesonLBfilmdepositiondateback to the early 1930s [23], the literature is still extremely scarce in whatconcernstheuseofthetechniquetopreparebiomimeticbio- materialsbasedonextracellularmatrixcomponentsfortissueen- gineering applications. To the best of our knowledge, our study wasthe first where biodegradable polymer substrates were used for LS deposition, and as substrates for hESC-RPE [19]. The LS- coatedhoneycombfilms demonstratedincreasedbiocompatibility, due to the biomimetic properties of the collagen at the surface [19].

Inthiswork,wetook astepforwardby exploring,forthefirst time,thepotentialofcombiningBFandLStechnologiestoprepare thinbutfree-standing,porousfilms,coatedonboth surfaceswith a thincollagen-LS layer. High focuswasgiven tothe modulation oftheporosityacrossthematerial,withtheaimofproducingthin biocompatiblematerials, capableofsupportingthegrowthofRPE andECcellsinco-cultureandpreventingtransmembranecellmi- gration,while allowingthe fluxofsolublefactorssecretedby the cellsculturedonoppositesidesofthematerial.Thenoveltyofour workis furtherextendedby thefactthat itdescribesthefirstEC andRPEco-culturerelyingsolelyonhumanpluripotent stemcells (hPSC), specifically humaninduced-pluripotent stem cells(hiPSC) ascellsource.Infact,invitroretinalmodelstypicallyuseRPEand ECcellsofprimaryorigin, obtainedfromeitherhumanoranimal donors,orimmortalisedcells,whichcansignificantlyfromthena- tivecounterparts[8,11,13,14,16,24–26].Protocolsfordifferentiation ofRPEfromhumanembryonicstemcells(hESC;[27,28])andfrom hiPSC[28–32])havebeenestablished,andthecellshavebeensuc- cessfullyculturedordifferentiatedinmono-cultureonavarietyof materialsbydifferentresearchgroups,includingourown[33–38]. hESC- and iPSC-derived EC have also been established andcom- paredtothenativecounterparts [39–42].However, tothebestof our knowledge, neither hESC- nor hiPSC-derived ECs have been studiedsofaraspartofinvitroco-culturemodelsofRPEandEC.

2. Materialsandmethods 2.1. Materials

96/04L-lactide/D-lactidecopolymer (PLA96/4)(PURASORB PLD 9620, puriﬁed, medical grade, IV midpoint 2.0 dl/g) was from Corbion, Purac, Netherlands. Dioleoyl phosphatidylethanolamine (DOPE)wasfromSigma, Japan.CollagentypeIVfromhumanpla- centawasfromSigma-Aldrich(USA).

2.2. Preparationofporousﬁlms

Porous films were generally prepared by the BF method us- ingPLA96/4asthepolymer[20].Solutionscontaining10mgml⁻¹ PLA96/4 and 1mg ml⁻¹ DOPE were initially prepared in chlo- roform. Three batches of films were prepared thereafter, following prior optimisation of the casting conditions, intended as a meanstomanipulatetheporosityofthefilms onthebottomsur- face. Allfilms were prepared by casting thesolution on topof a roundcoverglass(Ø 12mm)centrallyplacedinsideapetridish(Ø

(4)

40mm, Steriplan^R). For Batch 1 (designated BF1), 0.8ml of dis- tilled water was previously added to the petri dish, in order to ensure immersionofthecoverglass,andthepetri dishwaskept at −20 °C until use. Casting was performed by adding 0.3ml of the polymer/DOPE solution on top of the frozen cover glass. For Batch2 (designatedBF2), similarconditions were employed with theexceptionofthefactthatthecoverglasswaspre-immersedin an aqueous dispersionof DOPE(0.05mgml⁻¹,prepared bysoni- cation). Castingdone byadding0.2mlofthepolymer/DOPEsolu- tion ontopofthe frozencoverglass.ForBatch3(BF3),0.5mlof the polymer/DOPE solution was directly caston top of the glass slide(non-immersed),usedatroomtemperature(RT).Inallcases, solventevaporationproceededunderhumidairﬂow(80±3% RH).

PreparedsampleswerethereafterallowedtodryatRTinorderto eliminate the waterdroplets condensed at thesurface, and were washedthreetimeswith70%ethanoltoensurecompleteremoval of thesurfactant. Allsamples were keptdry ina desiccatoruntil furtheruse.

2.3. Scanningelectronmicroscopy(SEM)

Morphologicalfeatures ofthehoneycomb ﬁlmswere observed by SEM,usinga ﬁeldemission scanningelectron microscope(FE- SEM, CarlZeissUltra55,Germany).Aperture sizewas10μm and theaccelerationvoltagewas1kV.Drysampleswereimagedontop andbottomsurfaces.

2.4. Watercontactangle(WCA)measurements

The wettabilityof theporous materials wasassessed by mea- suringtheleftandrightstaticwatercontactangleusinganoptical tensiometer (Attension ThetaLite, BiolinScientiﬁc AB,Stockholm, Sweden). A dropletof deionised water (3 μl) wasapplied to the samplesurfaceusingtheinstrument’sautomatedliquiddispenser.

Collected imagesof thedroplets were analysed usingthe OneAt- tensionsoftware,Version3.2.MeanvaluesofWCAwerecalculated frombothleftandrightWCA,andfromaminimumof5measure- mentspersampletype(2–3independentbatches).

2.5. Electricalresistance(R)

The electrical resistance(R) across theporous films wasmea- sured to assess the permeability of the materials, generally as described in our previous work [19]. Samples were firstly pre- immersed in DPBS overnight in order to ensure soaking. After that,samplesweremountedintoP2307sliders(PhysiologicInstru- ments,USA)andtightlyassembledtoacustom-builtTefloncham- ber, where contact between two compartments containing DPBS was madethrough a small circularopening containing the sample(Ø=0.031cm²).Rvaluesacrossthematerialsweremeasured usinganEVOM2EpithelialVoltohmmeter(WorldPrecisionInstru- ments, USA).Measurementsacrosstheempty sliders,wherecon- tact between the two compartments of the Teflon chamber was maintained bytheuncoveredopening,weresimilarly carriedout, in order to establish the reference for the maximum R. Mean R values were determined fromsixmeasurements obtained froma minimumof3independentbatches.

2.6. InvitrostabilityofBFﬁlms

Thestabilityoftheporousﬁlmsinvitrowasinvestigatedbyin- cubating thematerials inphosphate bufferedsaline (PBS) pH 7.4 at 37 °C. Degradationstudies were carried out using three inde- pendentsamplebatchespertimepoint.Brieﬂy,sampleswithadi- ameterof12mmwereinitiallywashed with70%ethanolinorder to eliminate anyresiduescausedby samplepreparation. Samples

were leftto dryatRT overnight.After that,samples were placed inavacuumchamberforaminimumoffourhoursinordertoen- surethat allwater hadbeenremoved. After measurementof the initialdrymass (m_i),sampleswere individually placedinside the wellsof6well-plates, towhich6mlofPBS wereaddedper well.

The plateswere placed inthe incubatoreither for3or 5weeks.

PBSsolutionwasreplacedoncea week.Finally,sampleswere re- movedfromthewells,andwerethoroughlywashedwithdistilled water.Sampleswere lefttodryovernight,andfurtherinthevac- uumchamberasdescribedabove,beforedeterminationoftheﬁnal drymass(m_f).Totalmasslossafterthedegradationstudywases- timatedfromtheequation:

Massloss

(

^%

)

⁼

mi−m_f

mi

×100

2.7.DepositionofLangmuir–Schaefer(LS)ﬁlmsofcollagen

CollagentypeIVwastreatedasdescribedbefore[19,35].Briefly, collagenwasfirstdissolved indiluteaceticacid(pH∼3) toacon- centrationof1mgml⁻¹.Thepreparedsolutionwasthensonicated in an icy water bath for 10min, followed by a 10min rest period,andbyan additional10minofsonication.AKSVminitrough system was used to prepare the LS films [35]. 2× PBS pH 7.4 (20.8±0.5°C)wasusedasthesubphase.Aglassmicrosyringewas usedtoadddropwise180μlofthefreshlysonicatedcollagensolu- tiontothesubphase.Collagenwasallowedtostabiliseonthesub- phasefor30minbeforecompressionata speedof65mmmin⁻¹, i.e.49 cm² min⁻¹.For thehorizontal deposition ofcollagentype IVonboththetopandbottomsurfacesofeachfilm,sampleswere previouslystabilisedbyinsertionbetweentwoParafilm^R “M” (Be- mis,USA)rings.Thissimpleprocedureallowedustocarryoutthe sequentialdeposition of the Langmuir–Schaefer films on thetwo surfaces.Depositionwascarriedout bythetouchandliftmethod at a pressure of 30 mN m⁻¹. Double-coated surfaces were left to dryvertically (in order to avoidcontact with the surfaces) in a desiccator. LS-coated films were imaged by SEM, as described above. The topographicalfeatures of the LS-coated surfaceswere furtherassessedby AtomicForceMicroscopy (AFM),usingan XE- 100microscopefromParkSystemCorp,USA(10×10μm²scanned area).Surfaceswere scannedin noncontactmodeatRT, usingan APPNANO AFM cantilever(ACTA). Image acquisition andprocess- ingwascarriedoutusingXEPandXEIsoftware,respectively(Park Systems,USA).

Thickness of the hydrated, double LS-coated ﬁlms was deter- minedusingacontactproﬁlometer(DektakXT,Bruker,USA;n≥7).

Inordertoallowsamplehydration,sampleswerepre-immersedin cellculturemedium at37 °Covernight.Theexcessliquidwasre- movedbeforecarryingouttheproﬁlometrymeasurements.

2.8.DifferentiationofhiPSCcellstowardsendothelialcells

Healthy adult skin fibroblast UTA.04607.WTS hiPSC cells, de- rivedandcharacterisedatProf.KatriinaAalto-Setälä’slaboratoryat TampereUniversityaspreviouslydescribed[43],wereusedforEC differentiation.Nonewlineswerederivedinthisstudy.ThehiPSC cellswere previously adopted to feeder-free conditions [44].The ECdifferentiationmethod wasthemodification ofLiuetal.[45]. Briefly, thehealthy adult skinfibroblast derived UTA.04607. WTS hiPSCcellswereseededassinglecellsuspensiononhumanrecom- binant laminin 521 (Biolamina) coated CellBind^R 24 well plates (Corning)atdensitiesfrom11000cells/cm²to52000cells/cm²in Essential8^TM Flexmedium (Gibco).Cellswereincubatedfor24h at +37 °C in 5% CO2 while small colonies formed. The medium wasthen changed toDMEM/F-12 withGlutaMAX^TM (Gibco) supplemented with 4μM CHIR 99,021 (Tocris). After a two-day

(5)

Table 1

FACS antibodies used in ﬂow cytometry analyses. Volume denotes the volume of antibody solution used per one test.

FACS antibody Manufacturer Volume (μl) PE anti-CD144 eBioscience 12-1449-82 1 APC anti-CD31 eBioscience 17-0319-42 1 FITC anti-CD34 ImmunoTools 21270343 5 FITC anti-TRA-1-81 BD Pharmingen 560194 10

incubation period, the medium was changed to Endothelial Cell Basal Medium 2 (PromoCell) supplemented with 5ng/ml bFGF (Miltenyi)and 10ng/mlVEGF165 (R&D Systems). After two days, endothelial-like cells were expanded in Endothelial Cell Growth MediumMV2 (PromoCell)supplemented with10ng/mlVEGF165, henceforthdesignatedasECmedium.Allcellcultureconditionsin thismethodinclude25U/mlPenicillin-Streptomycin(Thermo).

2.9.Flowcytometryanalysis

Progression of differentiation and characterisation of sorted ECs was assessed by ﬂow cytometry analysis. Endothelial or endothelial-likecellswerewashedoncewithDPBS,detachedenzy- maticallyusingTrypLE^TM Select(Gibco) andharvestedin10% FBS (Gibco)inDMEM/F-12+GlutaMAX^TM.Cellsuspension waspassed througha40μmmeshstrainerbeforeavolumeofcellsuspension correspondingto1×10⁵cellswastransferredtoa5mlroundbot- tomtube foreach staining.Cells were washed twice withbuffer solution(0.5%BSA(Sigma)inDPBS+4mMUltraPure^TM EDTA(In- vitrogen))andstained withFACS antibodies(Table1) for20 min in50μlofbuffer.Sampleswerewashedtwicewithbuffersolution afterstaining.Flow cytometryanalyses wereperformedusing BD Accuri^TM C6ﬂowcytometer(BDBiosciences)andC6Software(BD Biosciences).

2.10.SelectionandcultivationoftheCD31positivecellfraction

ECs were selected fromthe undifferentiatedcells usingMACS CD31MicroBeadKit(Miltenyi).Cells were harvestedasdescribed beforeandupto1×10⁷cellswerewashedtwicewithbuffersolu- tionwiththesamecompositionasdescribedbefore.Thecell pel- letwasresuspended to 60μl ofbuffer solution and20μl ofFCR- block wasadded.After brief vortexing,20μl of CD31MicroBeads wereadded,thesuspensionwasvortexedagainandwasincubated for 15 min at +4 °C, before being washed once with buffer. A MS-columnwasplacedtotheMiniMACSseparatorandcellswere sortedaccordingtothemanufacturer’sprotocol.

Selected CD31 positive hiPSC-derived ECs were seeded on Nunc^TM on T75 ﬂasks (Thermo) coated with sterile 0.1%

gelatin (Sigma) in water, at densities from 2500 cells/cm² to 4000 cells/cm² in EC medium. Cells were cultured until con- ﬂuent with medium change every other day. Conﬂuent cultures of ECs were detached as described before and cryopreserved in EC-cryomedium (50% FBS, 40% Endothelial Cell Growth Medium 2 (ECGM-2; PromoCell), 10% DMSO (Sigma)) [45]. hiPSC-derived ECs were cultured for other applications, the same way as de- scribedhere,indifferentformats.HumanretinalmicrovascularECs (ACBRI-181;CellSystems) werecultured thesamewayforall assays.

hUVECs were previously extracted at BioMediTech, Tampere Universityfromtheumbilical cordsacquiredfromscheduledCae- sareansectionsaccordingto[46].CultureofthehUVECswascar- riedoutaccordingto[46,47],ifnotstatedotherwise.

Human adipose stem cells(hASCs), used for the angiogenesis assay(inSection 2.13),werealsoextractedatTampereUniversity, fromadiposesamplesacquiredfromsurgicalprocedures,according to[46].CultureofthehASCswasperformedaccordingto[47].

Table 2

List of primary and secondary antibodies used for immunoﬂuorescence labelling of ECs.

Primary antibodies Manufacturer Host Dilution anti-von Willebrand factor Dako A0082 rabbit 1:400

anti-CD31 Dako M0823 mouse 1:400

anti-CD144 BD biosciences 555661 mouse 1:400 Secondary antibodies Manufacturer Host Dilution Alexa Fluor ^TMA488 anti-rabbit Invitrogen A21206 donkey 1:500 Alexa Fluor ^TMA568 anti-mouse Invitrogen A10037 donkey 1:500

2.11. IndirectimmunoﬂuorescencelabellingofECs

ThehiPSC-derived ECswere cultured onNunc^TM 96-wellplate asdescribedbeforeuntilconfluent. Forindirectimmune staining, cellswerewashedoncewithPBS,fixedin4%paraformaldehydefor 10min,permeabilisedin0.1%TritonX-100for10minandblocked in 3% BSA for 1 h. Cells were stained with endothelium-specific primaryantibodiesanti-vonWillebrandFactor(vWF),anti-CD31or anti-CD144(Table2)in0.1%BSAfor1hatRTorovernightat+4°C.

Secondaryantibodies(Table2) werealso dilutedin0.1% BSA and thecellswerestainedforonehouratRT.Samplesweremounted, andthe nuclei visualisedusing Vectashield(Vector Laboratories).

hUVECsandACBRIswereusedaspositivecontrols.Thecellswere culturedandlabelledthesamewayashiPSC-derived ECs.Stained sampleswereimagedusinganOlympusIX51Fluorescencemicro- scope(Olympus).

2.12. Acetylatedlow-densitylipoproteinuptakeassay

Functionality of differentiated ECs was analysed by assessing theuptakeof acetylatedLDLusing Dil-Ac-LDL (CellApplications).

hiPSC-derived ECs were cultured onNunc^TM 96-wellplate asde- scribed before until conﬂuent and stained according to manufacturer’s protocol. hUVECs and ACBRIs were used as positive controls. The cells were cultured and labelled the same way as hiPSC-derived ECs. Stained sampleswere imaged usingan Olym- pus IX51 Fluorescence microscope (Olympus) and the acquired greyscale images were colourised using Adobe Photoshop CS4 (Adobe).

2.13. Angiogenesisassay

TheangiogeneticcapacityisanimportantcharacteristicofECs.

The 3D capillary formation is a complex process. The mesoder- malsupport,forexamplefromthehASCs,hasbeenshowntoim- prove reproducibility and formation ofcapillary network [47,48]. Here, the angiogenic capacityof hiPSC-derived ECs was assessed with hASCs co-culture [47] with slight modiﬁcations. Lab-Tek^TM (Thermo)8-wellChamberSlides^TM werecoatedwith0.1%gelatin.

hASCs were seeded at 20 000 cells/cm² in ECGM-2 and incubated for2–3h to establishadhesion tothe well surface. hiPSC- derived ECs, hUVECs and ACBRIs were seeded on top of hASCs at 8000 cells/cm². Slides were incubated for seven days with onemediumchangeatdaythree.Indirectimmunoﬂuorescencela- bellingofco-cultureswasperformedasdescribedearlier.ECswere stainedforrabbitanti-vonWillebrandfactor(1:400,A0082,Dako) and formation of basement membrane was assessed by staining forgoatanti-collagen IV(1:100,AB769, Millipore).Secondary an- tibodiesusedwereAlexaFluor^TManti-rabbitA488(1:500,A21206, Invitrogen)andAlexaFluor^TM anti-goatA568(1:500,A11057,Invit- rogen).StructureswereimagedusingZeissLSM700laserscanning confocalmicroscopeandZeiss2.1Blacksoftware.

(6)

2.14. Processingthehoneycombﬁlmsforcellculture

Honeycomb films were UV-treated in a laminar hood for 20 min per side. Then the material was cut into four segments and each segment was clamped between PermeaSys Base ele- ment(CodeHS1-BL)meantforthinmaterialssamplesandHolder lid (Code HS-LP) (Produced at Tampere University, BioMediTech, https://biomeditech.fi/permeasys-holder-set/)thatwere previously washedtwicewith80%EtOHanddriedinabs.EtOH.Thesupport- ingParafilm^R surroundingthematerialwasremovedbeforeseal- ing the holders by closing with the Closing/ opening tool (Code HS-TO).Holderswerestoredin24-wellplatesinRTuntilthenext day.

2.15. Differentiationandcultureofhumaninducedpluripotentstem cellstoretinalpigmentepithelialcells

TheUTA04311.WTsiPSCweredifferentiatedtowardsretinalep- ithelialcellsaspreviouslydescribed[44]withoutinduction.Briefly, the undifferentiated hiPSCs were dissociated from colonies with TrypLE^TM Select Enzyme (Gibco, Thermo Fisher Scientific). De- tached cellswere transferred to Corning^R Costar^R Ultra-Low attachment plates, and grown in KnockOut^TM Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 2mM GlutaMAX^TM, 1% MEM non-essential amino acids, 0.1mM 2-mercaptoethanol, 20% KnockOut^TM SR (KO-SR) and 50U/ml penicillin-streptomycin (all from Gibco, Thermo Fisher Scientific). An overnight blebbis- tatin (5μM, Sigma-Aldrich) supplementation of the medium was utilised toimproveembryoid body(EB) formation.Thereafterthe spontaneousRPEdifferentiationwasinduced byreducingtheKO- SRto15%,andthismediumisreferredthereafterasDM-.TheEBs were allowed to mature for 4 days.Pigmented areaswere man- ually separatedwitha scalpelanddissociated withTrypLE^TM Se- lect Enzyme,andacquired singlecellsuspension filteredthrough 100 μm BD Falcon cell strainer (BD Biosciences, San Jose, USA), andreplatedontowellplatescoatedwithhuman10μgcollagenIV (COLIV;Sigma-Aldrich, USA)and1.8μg laminin(L521;Biolamina, Sweden) to expand cell numbers andpurify the cell population.

Toexpandandpurifytheculture,thisreplatingwasrepeatedafter 42and12days.Duringthisperiod,DM-mediumwasreplenished thriceaweek.Afterexpansionofcells(passage4),thehiPSC-RPEs were dissociated with Trypsin–EDTA, ﬁltered through a strainer andstoredtocryobankinLN2.Oneweek beforethestartofthe experiments,thebatchofmaturehiPSC-RPEcellswasthawedfrom liquid nitrogen. To ensure the viability of thawed cells, the ma- turehiPSC-RPEcellswereplatedon10μgCOLIV-and1.8μgL521- coatedwellsatadensityof200,000cells/cm²andallowedtogrow for7daysinDM-.

2.16. SeedinghiPSC-ECsandhiPSC-RPEcellsonoppositesidesofthe ﬁlms

Co-culturestudieswerecarriedoutinan‘open’system,where cellsculturedonoppositesidesofﬁlmssharedthesamecellcul- ture medium. The 10μl of EMV2 supplemented with 16.6ng/ μl VEGF(thereafter calledaspan-ECmedium)wasaddedtotheup- persideoftheaperturestomoistenthematerial.PermeaSyshold- erswithclampedmaterialwereﬂippedupsidedownand150μlof same medium was added to the bottom of the well to moisten the material from both sides. ECs (hiPSC-EC, ACBRI181, hUVEC) weretrypsinisedandsuspendedinthepan-ECmedium. ECswere seededonthebiomaterialsataconcentrationof100,000cells/cm² in 20μl of pan-EC medium. Cells were let to adhere in small medium volumeforfourhours.Thereafter,an additional750μlof pan-ECmediumwasaddedtosubmergethePermeaSysholders.

After overnight culture, the PermeaSys holders with plated ECs were ﬂipped such that the ECs were facing down. Excess of mediumwasaspiratedsuchthattheonlyECswereleftsubmerged (approximately200μl wasremaining).ThehiPSC-RPEcells, which had been thawed seven days prior the experiment, were then trypsinised,countedandtestedforviability.ThesehiPSC-RPEcells were seeded on the top (honeycomb) side at the concentration of100,000 cells/cm². Thisequals to 3400 cells/PermeaSys holder aperturein5μlvolumeofmedium.hiPSC-RPEcellswerelettoad- hereforeighthours intheincubator. Thereafterthemediumvol- umewasincreased by adding an additional 50μl ofDM- on the topholderaperture.Onthenextday,mediumwasreplenished.To prevent formationofairbubbles underneath theholder, approxi- mately100μl of theold medium wasleft unaspirated.Then 1ml offreshmedium with2/3ofDM-and1/3ofpan-ECmediumwas addedtothecultures.Cellswereculturedforthreeweeksbecause atthattimepointthehiPSC-RPEcellsstarttoexhibitRPE-features suchasthemorecompactedcobblestonemorphology,whereasthe ﬁlmswere stillintactenoughto beliberatedfromthePermeaSys holders, allowing further analysis. The medium was replenished thriceaweekforathree-weekcultivationperiod.Thereweretwo biologicalreplicates andtwo technicalreplicates of eachcell and materialcombination.

2.17. IndirectimmunoﬂuorescencestainingofhiPSC-ECand hiPSC-RPEculturesonthebiomaterials

After the three-week cultivation period, the biomaterial cultures were gently washed with 1xPBS, fixed 10min in 4% PFA, permeabilised in 0.1% Triton X-100 for 10 min and blocked in 3% BSA for one hour at RT. The cultures were stained with endothelial-cellspecificrabbitvWF(DAKO1:400)andRPE-specific antibody against cellular retinaldehyde-binding protein (CRALBP, Abcam1:500)overnightat+4°C.Afterthreewasheswith1xDPBS the same secondary antibodiesand staining times were used as in Section 2.11. The biomaterials with cultured cells were care- fullyliberated fromthePermeaSys holderswithPermeaSysopen- ing/closingtool.ThefragileculturewastransferredtodropofVec- tashieldwithDAPImountingmediaandplacedbetweentwoZeiss cover glasses. The imaging of cultures was done using a Zeiss LSM700laserscanningconfocalmicroscopebyscanningtheentire depthatoncewitha40×lensusing1.0μmincrease.

2.18.Analysisofcellnumber

The attachment andproliferation of ACBRI181, hUVEC, hiPSC- ECscultured ontheBF1,BF2andBF3wasdeterminedafterthree weeksof culture.From thetwo replicates, ﬁve imagesfrom ran- domlychosenareasoneachsamplewithDAPI-stainednucleiwere capturedusing an Olympus IX AxioScopeA1 ﬂuorescence microscope with 20x objective. The cell number on each image was counted using the Cell Counter Plugin of the ImageJ image processingandanalysissoftware.Allcellsontheimagewerecounted.

Thecellnumberispresentedascellsincm².

Similarly, the attachment and proliferation of ACBRI181, hU- VEC,hiPSC-ECsandhiPSC-RPEsinco-culturesontheBF1,BF2and BF3wasdetermined afterthree weeksofculture fromtwo inde- pendent samples with two technical replicates. The image analysis was done from two replicates from which four images ac- quiredfromrandomlychosenareaswithDAPI-stainednucleiwere capturedusingthe Zeiss LSM780 confocalmicroscope anda40× oilimmersionobjectivewith0.6zoom. Fromthose,fourdifferent 100μm×100μm-sized areas(from upperleft, upperright,lower left,lowerrightandmiddle)were countedusingtheCellCounter Pluginofthe ImageJimage processingandanalysissoftware.The cellnumberispresentedascellsincm².

(7)

Fig. 1. SEM images showing the surface features of the porous ﬁlms prepared by the BF method.

2.19.Ethicalaspects

The institute has supportive statement from the Ethics Com- mittee of the Pirkanmaa Hospital District to generate hiPSCs from the donor ﬁbroblasts (Aalto-Setälä/R08070) and to use hiPSClines derived in other laboratoriesfor ophthalmic research (Skottman/R14023). The extraction of human umbilical vein ECs (hUVECs)has the supportive statementfromthe Ethics Commit- teeofPirkanmaaHospitalDistrict(Miettinen/R13019).The extrac- tionofhASCshadwrittenpatientconsentandhasthesupportive statementfromtheEthics Committee ofPirkanmaaHospital Dis- trict(Miettinen/R15161).

2.20. Statisticalanalyses

Allnumericaldataareshownasmeanandstandarddeviation.

StatisticalanalysisofWCAandRwasperformedbyone-wayanal- ysisofvariance(ANOVA)andtheposthocGames-Howellmultiple comparisonstestusingIBMSPSSStatistics25software.Thestatis- ticalsigniﬁcanceofthe cell numberdatawasanalysed withIBM SPSSStatistics25usingtwo-tailedMann–WhitneyUtest.

3. Results

3.1.Morphologyofporousﬁlms

The three types of films (BF1–BF3) were imaged by SEM, in order to establish differences in terms of porosity on both top andbottom surfaces(Fig.1). The top surfacewaslargely characterised by the presence of a uniform array of pores, with only small differences in pore size being observed between samples (BF1 = 4.2±0.4μm; BF2 = 6.0±1.0μm; BF3 = 3.4±0.6μm). In contrast,modifying the casting substrate caused measurable differences in the morphology of the bottom surface of the films, as seen in the second row of Fig. 1. Casting the polymer solution onto the ice substrate led to the formation of large, circu- larareas (Ø = 56±22μm) characterised by lower film thickness and wide presence of smaller pores, seemingly in close contact withthe topsurface(BF1Bottom, Fig.1). Maximumdistancebe- tween these circular areaswas often not larger than 30μm (average=32±19μm). Apart from these areas, pores of smaller diameter were generally broadlydistributed across the whole bot- tomsurfaceofBF1 (Ø =0.5–5μm). Interestingly,whenthe polymer solution was cast onto the frozen DOPE solution (BF2), the largecircular areaswere mostly absent, eventhough thebottom

Fig. 2. Water contact angle ( n ≥10) measurements. ^∗p < 0.01; ^∗∗p ≤0.001. Rep- resentative photographs of the water droplets at the surface of each sample are shown in the lower panel.

surfacewasstillextensivelyporous(Ø=0.5–5μm).BothBF1and BF2contrastsignificantlywithBF3(preparedbycastingdirectlyon glass),asthebottomsurfaceoftheBF3wasnoticeablyflatterand contained significantly smaller pores (Ø = 1.5±0.7μm; BF3 Bot- tom,Fig.1).Thicknessofall sampleswasapproximately10μm,as shown by the SEM images presented asSupplementary Material (SM1).

3.2. Watercontactangle

Water contact angle (WCA) was measured on the uncoated top and bottom surfaces of BF1-BF3, in order to determine if the different porosity created by the different casting methods had an impact on surface wettability. The results are shown in Fig.2.

Top surfacesgenerally hadsimilar WCA, inthe average range 111–121°, with differences being statistically signiﬁcant only for the comparison between BF1 andBF3 (p < 0.001). WCA of bottom surfaces (77–102°) was signiﬁcantly lower than the WCA of top surfaces for all three materials (p < 0.001). The most striking observation in the WCA measurements is the fact that

(8)

Fig. 3. Electrical resistance (R) across the honeycomb ﬁlms, and across the empty slider ( n = 6). ^∗p ≤0.005.

Table 3

Estimated mass loss following in vitro degradation studies at 37 °C.

Sample 3 weeks 5 weeks BF1 3.0 ±1.3% 3.6 ±2.7%

BF2 2.7 ±0.8% 4.4 ±1.6%

BF3 3.6 ±3.9% 6.3 ±3.6%

the different casting methods caused an important effect on the WCA of the bottom surfaces, with the highest value being observed for BF2, followed closely by BF1. WCA of the bottom surface of samples cast directly on glass (BF3) was the lowest among all surfaces (77±5°), as clearly shown also by the microscopy images of the drops beneath the graph in Fig.2.

3.3. Electricalresistance

Results of electrical resistance (R) across the porous materials areshowninFig.3.Nosigniﬁcantdifferenceswerefoundbetween the materials that were cast on the frozen substrates (BF1 and BF2). Inaddition,Racross thesesampleswasonly slightlyhigher (but not statistically different) than R across the empty sliders, used ascontrolformaximum diffusion.However,R valuesacross BF3werefoundtobesigniﬁcantlyhigherthantheothertwosam- plesandthecontrol(p≤0.005).

3.4. Degradationstudy

Honeycomb ﬁlms were found to be very stableduring the in vitrodegradationstudy.Theestimatedmasslossvaluesareshown inTable3.Evenafterﬁveweeks,masslosswassmall,withaverage valuesbeinglowerthan7%forallthreesamples.

3.5. DepositionofcollagenLSﬁlms

CollagenIVLSfilms werehomogeneouslydeposited ontoboth topandbottomsurfacesofBF1,BF2andBF3(Fig.4(A)).Thedepo- sitionoftheLSfilmcreatedsmooth surfaces,andthepores were found to be completely covered by the collagen layer. SEM images takenfromthefilm edges(e.g.top surfaceofBF1)andfrom areas that had beencut duringsamplepreparation forSEM (e.g.

topsurfaceofBF2)clearlyshowthelow thicknessofthecollagen layer and the nicelyorganised honeycomb array underneath.To- pographical assessmentby AFM (Fig. 4(B)) conﬁrmedthat the LS ﬁlmseffectivelycoveredthepores.Evenso,thetopographyofthe

honeycombsurfacecouldstillbedetected,suggestingthatthecol- lagenLSfilmswereverythinandflexible.Thedepthofthepores wassignificantlyreducedtothenanometre-scale.

Thicknessofthepre-immerseddouble-coatedhoneycombfilms, estimated by profilometry, was found to be 10±1μm, 13±2μm and 7±3μm, respectively for BF1, BF2 and BF3. Despite the overall small range of values, statistically significant differences were found when BF3 wascompared to BF1 (p=0.005) and BF2 (p<0.001).

3.6.CharacterisationofhiPSC-ECs

The hiPSC which were induced to differentiate towards ECs were carefully characterised. The change from Essential 8 ﬂex medium to ﬁrst differentiation medium induced cell death and theremaining cell densityaffected theovercome ofthedifferen- tiation. If there were too few cellsin loose colonies at day two, no differentiation occurred; also seeding densities lower than 16 000 cells/cm² resulted in no endothelial type differentiation.

However, when the seeding density was 16 000 cells/cm² or higher, the cellsexpressed two different growth patterns. In the firstpattern,thecellsstartedproliferatingvigorouslyafterdaytwo andonly smalloval colonies ofendothelial-like cellshadformed at day four (Fig. 5(A) and (D)). The percentage of differentiated cellswaslow,withthemeandifferentiationefficiencyatdayfour beingonly2% ofall cellsanalysedwithflow cytometry,buthigh totalcellcountcompensatedthis.Inthesecondpattern,cellspro- liferatedina morecontrolled mannerresulting inlargercolonies ofendothelial-likecells,whichemergedandexpandedduringthe differentiation(Fig. 5(B)). In thissecond one, the total cellcount waslower,butthedifferentiationefficacywasconsiderablyhigher, 38.5%. When the first seeding density was 53 000 cells/cm² or above, noendothelial-type differentiationoccurred. If theculture ofendothelial-likecellswasextendedoverthesixdaystheundif- ferentiatedcell populationappeared toconstricttheexpansion of differentiatedcells.Forthatreason,differentiationwasterminated atday sixandcellswere sorted. The sortedand cultured hiPSC- derivedECsdisplayedcommonECmorphology(Fig.5(C)and(F)).

Inﬂow cytometryanalysis, all sampleswere gated forthe total live population of cells. Both hiPSC-derived ECs and control cellsweredouble-positiveforCD31andCD144(Fig.5(G))butthe co-expression of CD34 and CD31 was found to be different be- tweentheprimaryanddifferentiatedcells(Fig.5(H)).Theprimary cellsexpressedlower levels ofCD34than hiPSC-derived ECs.The pluripotencymarkerTRA-1-81wasexpressedindifferentiatedcells bylessthan1%ofthepopulation(datanotshown).

The expression of EC surface markers was veriﬁed withﬂow cytometry analysis. Expression of CD31 (Fig. 5(I)) and CD144 (Fig. 5(J)) wasprominent in hiPSC-derived ECs but expression of vWFwaslowinhiPSC-derivedECscomparedtocontrolcells.

The maturationandfunctionality ofhiPSC-ECs wasveriﬁed in Ac-LDLuptake(Fig.5(K))andangiogenesis(Fig.5(L))assaysmim- ickingthefunctionalityofprimarycells.Animportantcharacteris- ticofECsisthecapabilityofacetylated-LDLuptake[45,49].Inthe Ac-LDLuptakeassay,hiPSC-derivedECswereabletouptakeAc-LDL aswellascontrolcells(Fig.5(K)).Inaddition,formationofcapillar- iesisatypicalendothelialfeature.WhenhiPSC-ECorprimaryECs wereculturedonhACs,allECsformedthree-dimensional,branch- ing tubular vascular structures with distinguishable basal mem- branethesamewayasprimarycontrolcells(Fig.5(L)).

3.7.CultivationtestwithECmono-culturesandEC-RPEco-cultures

The compatibilityofthe coatedporous ﬁlms assubstrates for theculturedcellswasﬁrstlyevaluatedbyassessingcelladherence andmorphology.Three differentEC typeswere analysed, namely

(9)

Fig. 4. SEM (A) and AFM (B) images of top and bottom surfaces of honeycomb ﬁlms double-coated with collagen IV LS ﬁlms.

(10)

Fig. 5. Characterisation of hiPSC-ECs. Colony morphology of differentiating hiPSC-ECs and post-sort ECs in phase contrast microscopy. Small circular colonies of endothelial- like cells emerged at day 4 during the differentiation (A, D). When cell growth followed different pattern larger colonies of endothelial-like cells were present at day 4 (B, E).

CD31 positive sorted cell fraction exhibited endothelial like cell morphology at day 5 post sorting (C, F). Imaged using Nikon Eclipse TE 20 0 0-S microscope and NIS-Elements 4.30.00 software (Nikon). Scale bar 200 μm.

Comparison of hiPSC-derived ECs’ and primary ECs’ expression of vascular endothelial markers in flow cytometry analysis (G). Comparison of hiPSC-derived ECs’ and primary ECs’ expression of vascular endothelial markers by indirect immunofluorescence labelling. Sorted and cryopreserved ECs were cultured for two days as described before and labelled with endothelial surface markers CD31(red, I), CD144 (red, J) and vWF (green). hiPSC-derived ECs express junctional CD31 and CD144 prominently but expression of vWF is distributed to only a fraction of the cells as all control cells had significant vWF expression (I, J). Scale bar 200 μm. Confluent EC cultures were exposed to DiI-labelled Ac-LDL for 4 h and imaged hiPSC-derived ECs are capable of Ac-LDL uptake. Scale bar 500 μm. (K). hiPSC derived ECs are capable of forming vascular structures when co- cultured with hASCs. Confocal images of hASC-EC co-cultures at day seven. ECs visualised using vWF (green) and basal membranes using collagen IV (Col IV; red) (L). Scale bar 50 μm. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

(11)

Fig. 6. Representative confocal micrographs from biomaterial culture tests (BF1–

BF3) with hUVEC and ACBRI181, cultured as a mono-culture, or in co-culture with hiPSC-RPE. Endothelial cells were cultured in all cases on the bottom side of the films. The co-cultures are visualised from the EC side, and the hiPSC-RPEs from the same cultures but from opposite sides. Samples were immunostained with the RPE- specific marker CRALBP (red) and endothelial cell-specific vWF (green). Nuclei were counterstained with DAPI (blue). The white arrows denote cells migrating across the biomaterial and vWF positive cells in-between the CRALBP positive cells. Scale bars 10 μm. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

pluripotent stem cell derived-ECs (hiPSC-EC), and two different primaryECs (hUVEC andACBRI181),andby co-culturingECs and hiPSC-RPEs on opposite sides of the same material. ECs were cultured in all cases on the bottom side of the material. Cells in co-culture were allowed to contact through the pores of the biomaterial,andthroughthesharedcellculturemedium.

The commonly used ECs, hUVECs, formed conﬂuent cultures whengrowninmono-cultures(Fig.6).Allmaterialssupportedthe adherence of hUVECs. The pores, seen in BF2 (Fig. 6), were de- voidofhUVECcells. InhUVECandhiPSC-RPEco-cultures(Fig.6), the morphology of hiPSC-RPE cells was elongated. On BF2, the hUVECs were occasionally found on the hiPSC-RPE side (pointed withthewhitearrow).Inmono-cultures,theretinalprimarycells, ACBRI181,performed similarly asthe hUVECs (Fig.6) in forming homogenous vWF positive layer of cells. The RPE cells in their nativeform growin monolayers. InACBRI181 andhiPSC-RPE co- cultures,however,itwasevidentthat forBF1 thehiPCS-RPEcells were growing multi-layered. When the cellswere seeded onthe biomaterialthe collagenIVlayer preventedcellsfromgetting in-

Fig. 7. Representative confocal micrographs from biomaterial culture tests (BF1–

BF3) with hiPSC-EC, cultured as a mono-culture, or in co-culture with hiPSC-RPE.

Endothelial cells were cultured in all cases on the bottom side of biomaterial.

The co-cultures are visualised from the EC side, and the hiPSC-RPEs are from the same cultures but from opposite sides. Samples were immunostained with the RPE- specific marker CRALBP (red) and endothelial cell-specific vWF (green) Nuclei were counterstained with DAPI (blue). The white arrows denote cells migrating across the biomaterial. Scale bars 10 μm. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

side the pores of the biomaterial. After the three weeks of culture there were nuclei inbetween of the biomaterial (see white arrow)whichweretakenanindicationofcellmigrationinBF2.In addition,there were vWF-positive ACBRI181cellsin betweenthe hiPSC-RPEcells, which wasthesecond indicationof migrationof endothelialcellsthoughthebiomaterial. ThelessporousBF3 ma- terialexhibitedthehighestCRALBP stainingonthehiPSC-RPEcell edges,suggestingthatBF3supportedthematurationofhiPSC-RPE cellsbetter thanBF1 orBF2 in ACBRIprimary retinal endothelial co-cultures.

Culturing testswere carried out using three different batches of hiPSC-ECs, but as they all performed similarly, only two are shown in Fig. 7. BothhiPSC-EC1 and hiPSC-EC2 formed an even monolayer of vWF-positive cells. The coverage of hiPSC-EC cells (Fig. 7) was lower than with the primary ECs (Fig 6). In co- cultures,thehiPSC-EC1cellswere slightlysmallerthanhiPSC-EC2 cells.Inthe beginningoftheculture,thecollagenIVcoating pre- vented the cell from getting to the pores of ﬁlms. As described above forBF2 speciﬁcally, thedetection ofcell nuclei within the

(12)

Fig. 8. Number of hUVEC and ACBRI181 cells, and hiPSC-EC in mono-cultures grown on the different biomaterials (BF1, BF2, BF3). The number of cells, i.e. DAPI- stained nuclei, was evaluated from wide ﬁeld micrographs. Representative images of nuclear stain DAPI used for endothelial cell number analysis (A). Scale bar 100 μm.

Cell number analysis after three weeks of culture on BF1, BF2 or BF3 (B). Stan- dard deviation denoted with the error bars. The statistical signiﬁcance: ^∗( p ≤0.05),

∗∗( p ≤0.005) and ^∗∗∗( p ≤0.001) indicating the signiﬁcance between the indicated samples.

biomaterial (see arrowheads) was taken as an indication of cell migration through the pores of biomaterial. In addition, mixing of differentcell populations could be detected in thehiPSC-EC2- hiPSC-RPEco-cultures,wherevWF-positive cellswerein between the CRALBP-positive hiPSC-RPE cells(pointed withwhite arrow).

ThehiPSC-RPEcellshadthemostcompacthexagonalstructureand highestintensity inCRALBP-staining onBF1 andBF2, which may indicate that those materials provide better support for RPE cell maturation.

Fromthevisualinspection,thenumberofECsappearedtovary between differentculture materials. Toinvestigate whetherthere were differences in number of ECs in mono-cultures on different biomaterials,thenumberofDAPI-stainednucleiwascounted withtheaidofImageJCellcounter plugin(representativemicro- graphs in Fig 8(A) and calculated valuesin graph Fig. 8(B)). The primarycells,hUVECandACBRI181,generallydemonstratedhigher celldensitythanhiPSC-EC(hUVECs0.1×10⁸to0.3×10⁸ cells/cm² and ACBRI181 0.35×10⁸ to 0.7×10⁸ cells/cm² compared to <

0.2×10⁸ ofhiPSC-ECs/cm²). Withall testedECs theBF1 material hadthelowestnumberofcellscomparedtoBF2orBF3.Thisdif- ference wasstatisticallysigniﬁcantinhUVECsandhiPSC-ECs, and inACBRI181betweentheBF1vsBF3(Fig.8(B)).

Thecelldensitywasanalysedalsofromtheco-cultures(Fig.9).

The first important observation is that the number of hUVEC andACBRI181cellsinmono-culturewas,asawhole,significantly reduced in the presence of hiPSC-RPE, whereas the number of hiPSC-EC was mostly unaffected by the presence of the retinal epithelium. In addition, the increase in the primary EC numbers seen in mono-cultures from BF1 to BF2 and from BF2 to BF3 was seen only in hUVEC co-culture but not in ACBRI181. This difference detected in hUVECs was also statistically highly significant (p≤0.001). In the hiPSC-EC-hiPSC-RPE co-cultures, only BF1 demonstrated a substantially higher number of cells(hiPSC- EC1), and the difference was statistically significant. In addition, the hiPSC-RPE cell number was evaluated from the co-cultures (Fig.9).The ACBRI181co-cultureshadasubstantially lowernum- ber of hiPSC-RPE cells (<0.26×10⁸ cells/cm²) on BF1, BF2 and BF3thanhUVECs.ThehighesthiPSC-RPEcellnumberswerefound forthe co-cultureswithhiPSC-EC1 andhiPSC-EC2,as therewere

>0.3×10⁸ cells/cm² on all materials. hiPSC-RPE cell number on BF2wasslightlylower, butstatisticallysigniﬁcant, thancellscul- turedontheothertwomaterials(p≤0.001).

4. Discussion

In this work, we took advantage of the versatility of the BF methodtomodulatetheporosityandpermeabilityacrossthema- terials,andtherebyinfluencethecontactbetweencellsculturedon oppositesurfaces.Castingdirectlyontheglasssubstrate(BF3)lim- itedporeformationonthebottomsurface.Itisknownthatcasting on water [50] or ice [51] can lead to the formation of through- pores, buttypically low castingvolumes are needed, andthe resulting films are too thin to be free-standing, requiring transfer ontoasecondarysupport.Withthisinmind,wepreparedBF1by casting on ice, albeit using a significantly larger volume of cast- ingsolution,comparedtothepreviousstudies[50,51],inorderto ensure that theproduced films were self-supporting. As a result, highporosity wasobserved on the bottom surface, with circular areasofsmallerthicknesssuggestinga closecontactwiththetop surface. Itis interesting tonote that theaddition ofDOPEto the frozensubstrate,BF2,causedthebottomsurfacetoremainhighly porous,buthavinga more regularpore distributionas compared toBF1.Fukuhiraetal.[52]demonstratedthatthelowhydrophile–

lipophilebalance(HLB)valueandhighinterfacialtensionofDOPE contributetostabilisethecondensingwaterdropletsfromthehu- midenvironment,enablingtheformationoftheorganisedhoney- combpattern[52].Evenso, tothebest ofourknowledge, frozen aqueousdispersionsofDOPEhavenotbeenreportedbeforeasthe substrateforhoneycombﬁlmfabrication.Itisanticipatedthatthe surfactantpresentattheice-polymersolutioninterface mayhave contributed to stabilise the formation of water nuclei, acting as templatesfortheformationofporesonthebottomsurface.

Considering that the top surfaces of BF1-BF3 were similarly subjected to the humid environment, with the resulting pore sizenotdiffering significantlybetweensamples,itisunsurprising that the WCA values were similar between samples. Differences betweenthethreematerialswerehoweverconfirmedbytheWCA measurementson thebottom surface.It iswell knownthat WCA dependsonthe chemical composition ofthe surface,asmuch as itdependsonitstopography[53,54].WCAofsuchporoussurfaces ashoneycomb films, forinstance,is typically higherthan that of flatcounterparts,duetotheformationofairpocketsbetweenthe rough surface and the water drop [55]. This explains the lower WCA observed for BF3, the flatter bottom surface, as compared to BF1 and BF2. The specific topographical features of the latter two (including diameter, depthand shape of pores,and distance between them) additionally explain the small but statistically significant differences between these samples [54]. Even more

(13)

Fig. 9. Number of hUVEC and ACBRI181 cells, and hiPSC-EC in co-cultures with hiPSC-RPE grown on the different biomaterials (BF1–BF3) for three weeks. The number of cells, i.e. DAPI-stained nuclei, was evaluated from confocal micrographs. Representative images (100 μm ×100 μm) of nuclear stain DAPI used for endothelial cell number analysis. Scale bars 25 μm. Cell number analysis after three weeks of culture on BF1, BF2 or BF3 (B). Standard deviation denoted with the error bars. The statistical signiﬁcance: ^∗( p ≤0.05), ^∗∗( p ≤0.005) and ^∗∗∗( p ≤0.001) indicating the signiﬁcance between the indicated samples.

expressively,usingthefrozensubphasesforcasting(BF1,BF2)suc- cessfullyimprovedthepermeabilityoftheﬁlms,asshownhereby electricresistance,ameasurementofthemovementofionsacross thematerials [20]. Taken together, theseresults suggest that the possibilitytomanipulateporosityandpermeabilityinherenttothe BFmethodcan be particularlyadvantageous forthedevelopment of in vitro models of the BRB. In fact, the normally permeable Bruch’smembranebecomessigniﬁcantly lesspermeableinsevere cases of AMD, compromising the availability of nutrients to the neuralretina,andtheeliminationofmetabolicwaste[56].

Also important in this context is the stability of the materials over time. Previous studies with lactic acid polymers and co-polymershave demonstrated that the degradation rate ofthe scaffoldisinversely relatedwithcell viability andmigration into the scaffold [57]. For co-cultureof RPE and EC, a fast-degrading materialcould implythat thetwo celltypeswouldquicklyreach physical contact, which would hinder the intended biomimetic properties of the model. In this work, the high in vitro stability of all materials for a period of 5 weeks encouragingly sug- geststhatthematerialsmayalsodemonstratehighstabilityinvivo,

therebynotsignificantlyaffectingcellsculturedon(orinthevicin- ityof)thematerials.Additionalstudiesshouldfurtherestablishthe molecularcut-off ofthefilms,andtime-dependentchangesinvivo. In order to improve the biocompatibility of the BF films to- ward cultivation of hiPSC-RPE and hiPSC-EC, both top and bottom surfaces of the honeycomb films were coated withcollagen IVLSfilms.Langmuirfilmsarepreparedbyspreadingamphiphilic speciesattheair-waterinterface followedby solventevaporation or equilibration, andcompression at the interface. The compres- sionstep causesthemolecules toreorient intoa compactmono- layer,whichcan besubsequentlydepositedontoa solidsubstrate by vertical dipping or horizontallift ofthe substrate, thus forming LBorLSfilms,respectively[58].Previousresearchhasshown thatorientedcollagenfibrescanbecreatedbycompressionofthe fibresina Langmuirtrough,andthat the collagenLBor LSfilms arefavourablesubstratesforculturedcells[59].Encouragedbyour previousfindingsonthepositiveeffectsofcollagen-LSsurfacesto- wardcultured hESC-RPE [19,35],we nowembraced the challenge of coating both surfaces ofthe biomaterials witha thincollagen IV layer on each side. We believe that this is thefirst work de-

(14)

scribingsuch an approach. Thechallengeswere multifold:(i) the high WCA of PLA96/4 films makes them repel water,preventing immersioninthesubphase(therebyprecludingtheoptionofver- tical dipping for simultaneous coating on both sides, i.e. following the LB approach);(ii) the honeycombfilms are only ∼10μm, alsocompromisingverticaldipping, andmaking themfragilema- terialsthatcannot behandledwithharshequipment;(iii)thede- position ofLS layers onboth surfaces impliesthat direct contact withthesurfacespost-depositionshouldbeavoided.Ourstrategy to sandwichthe films betweentwo custom-made Parafilm^R “M”

ringsrevealedto bea practical solutiontotackle theseissues, as theringsmadethesampleseasy tohandle,making itpossibleto sequentially coat both sidesofthe samples, andproviding asta- bleframefordryingintheverticalposition.Oneofthemostfas- cinatingfeatures ofcollagenIVLS-coatedhoneycombfilms isthe factthatthethincollagenlayer(previouslydeterminedasapproxi- mately30nm[35])cancovertheporescompletelyasacontiguous sheet. Thiscontrasts significantly withthe resultsachievedwhen thesamplesarecoatedbysimpleimmersioninthecollagensolu- tion,wheretheporesaretypicallyleftexposed [19].Hereaswell, both topandbottomsurfaces werehomogeneously coated.These collagen-LSsurfacescanimprovecelladhesionandgrowthfortwo importantreasons: first,thecoatedsurfacesareseen bythe cells asnearlyflat;secondly,cellsareincontactwithabiomimetic,con- tinuoussheetofcollagentypeIV, animportantcomponentofthe basementmembranesofRPEandthechoriocapillaris[4].Further- more,collagenIVhasbeensuggestedtohavea roleinregulating migrationandgrowthofECs,andinangiogenesis[60].Becausethe angiogenic cascade has been associated with increased secretion andextracellulardepositionofcollagentype IV[61],weaddition- allyhypothesizethatin vitrodiseasemodelssimulatingangiogen- esisin the retina maybe developed using thickercollagenIV LS films.

In flow cytometry our hiPSC-derived ECs had high co- expression ofCD31andCD144suggestinga goodendothelial differentiation [41,45,62].Here, the expression ofCD34was consid- erablyhigherindifferentiatedECs (61–79%)compared toprimary cells(3–4%). Ourfindings ofCD34expression coincidewithfind- ings of Simara et al. [63]. In their experiments, hiPSC-ECs from three different sources displayed similar elevated levels of CD34 expression (64–92%) compared to primary ECs in flow cytometry analyses. Thehigh CD34positivitymayalso indicatethat the hiPSC-ECshadnotreachedmaturityatthispoint.Liuetal.[45]re- portedthatCD34expressionpeakedatDay7intheir 12-daydif- ferentiationperiodandstarteddecliningafterthat.Inarecentpa- per [62],it wasshown that hiPSC-ECs hada mixedarterial- and embryonic-like identity with prominent expression of both arte- rial markers. In immunostaining assays, expression of CD31 and CD144 wasnot asprominentin hiPSC-derived ECs asin primary ECs [41,45]. Our hiPSC-ECs hadless vWF stain than the primary EC,andweremuchbiggerinsizethanprimaryECs,similarlyasin apreviouspublication[62].ThehiPSC-ECsdifferentiatedherewere verycapableofup-takingacetylatedLDLandcomparedwelltore- sultsof other authors[45,64]. Theangiogenesisassayshavebeen performedondifferentECMmatrices[45,63,64],andwithseveral celltypessuchaspericytes[41],andstromalcellssuch ascardiac fibroblasts[62]andhASCs[47].Here,thehiPSC-ECsaswellaspri- maryECs were abletoformtubular vascular structureswithdis- tinctbasalmembranewhenco-culturedwithhASCs.

A significant number of studies have shown that the high porosity of honeycomb films can influence the propertiesof cultured cells, forinstance by allowing control of protein adhesion, facilitatingcellattachmentthroughhookingofextendingfilopodia, facilitatingnutrienttransportandwasteelimination,andallowing contactguidance[18].Intissueengineering,honeycombfilmshave therefore shown promise in mono-culture of several cell types,

includinge.g. rat cardiomyocytes[65], porcineaorticECs [66,66], mousepreosteoblasticMC3T3-E1cellsandprimaryrat osteoblasts [67,68]andhumandermalfibroblastsandepidermalkeratinocytes [69].Ourrecentstudies alsosuccessfullydemonstrated theadhe- sion and maturation of hESC-RPE on PLA96/4 and polybutylene succinatefilms prepared by BF [19–21].In spitethis, honeycomb filmspreparedbyBFhaverarelybeenstudiedassubstratesforco- culture systems, withonly a few reports available for co-culture ofbovineaorticECandsmooth musclecells(SMCs)[70],andfor co-culturesystems, wherethe samecell type (cardiomyocytesor hepatocytes)wascultured on both sidesof thehoneycomb films [71].

Intheouter retina,RPE cellsare separatedfromthechoroidal ECs by the Bruch’s membrane [6]. As such, we propose here the use of the double-coated honeycomb films as Bruch’s mem- branesubstitutes. Consideringthat thehumanBruch’s membrane hasan estimatedthickness of approximately5μm [72],determi- nation of material thickness was considered essential. Hydrated double-coatedhoneycombfilmswereslightlythickerthanthena- tive membrane, in order to ensure that the films would be free standingandcouldbe easily manipulatedforcell cultureandfor the extensive material and cell characterisation experiments de- scribedin thiswork. While thislimitationishere acknowledged, itis essential tohighlight that theBF method,apart fromallow- ingmodulationtheporosity,wettabilityandpermeability,alsoal- lowscustomisationofsamplethickness,whichmaybeconsidered insubsequentwork.

Viahumoral factors, the ECs are known to modulate the RPE cell functionality, improving RPE cell maturation [14,73,74] for example by increasing growth factor secretion [73,74], the RPE cell barrierfunction [16,73,74] and the RPE-speciﬁc gene expression [14]. Vice versa, RPE cells in co-cultures have been shown to decrease angiogenesis [14] andincrease the intracellular gaps [73,75].RPEcellsaretypicallyobtainedhumanadult[11]orfoetal [13] donors, or from animal eyes [13]. Immortalised human cell lines such as ARPE-19 are also widely used due to being more easily available [8,11–14,16], but can differ signiﬁcantly from the native counterparts in terms of functionality and gene expression,andmayresponddifferentlytotheenvironmentand/orunder stressconditionsascomparedtothenativeRPE[24–26].Inasim- ilarmanner,ECaretypicallyofprimary origin[8,11–13,13,15].Hu- manumbilicalvein endothelialcells(hUVEC)are amongthemost commonlyused model of EC [14] dueto their easieravailability comparedtoothertypesofprimarycells,butthecellsareavailable inlimited supplies, andpooledhUVEC preparationsentail batch- to-batchvariations,potentiallycompromisingthereproducibilityof theresults[76].InadditionhUVECsandarenotophthalmicorigin, whichcanalsohavean effectontheir functionality.According to ourknowledge,nopreviousstudieshavereportedECandRPEco- culturestudiesusingbiodegradablebiomaterials,inwhichboththe ECandRPEcellsarederivedfromhiPSCs.

Here, we assessed the cellular adhesion on three differentBF structures inendothelial mono-culturesandinendothelial-hiPSC- RPE co-cultures. For comparison, hiPSC-derived EC cultured as mono-cultureandinco-culturewere comparedto thecommonly used umbilical vein hUVEC, and to the more representative pri- maryhumanretinalmicrovascularEC(ACBRI181).Allthreetested materials (BF1-BF3) supported adherence of both primary (hU- VECandACBRI181) andhiPSC-EC derived cells.As previous studies have shown that ECs prefer the rougher surface of ﬁlms, as comparedtoasmooth surface[77],itisinterestingthat thehigh- estdensity forall EC wasgenerallyfound on the smoothest and less permeable substrate, BF3. However, it is important to em- phasise that the EC were cultured on the bottom surface of the ﬁlms, whichlacks the honeycombtopography investigatedin the indicatedstudy.Infact,celladhesionontopatterned substratesis