A microelectrodearray(MEA) integratedwith clusteringstructuresforinvestigatingin vitro neurodynamicsin confinedinterconnected

sub-populationsof neurons

L. Berdondinia,∗, M. Chiappaloneb, P.D. van derWal a, K. Imfelda, N.F. deRooija,M. Koudelka-Hepa, M. Tedescob, S. Martinoiab, J. vanPeltd, G. LeMassonc, A. Garennec

a Sensors, Actuators and Microsystems Laboratory, Institute of Microtechnology, University of Neuchatel, Jaquet-Droz 1, 2007 Neuchatel, Switzerlandb Department of Biophysical and Electronic Engineering (DIBE), University of Genova, Via Opera Pia 11a, 16145 Genova, Italy

c INSERM E358, University Bordeaux 2, 146 rue Leo Saignat, 33077 Bordeaux Cedex, Franced Netherlands Institute for Brain Research (NIBR), Meibergdreef 33, 1105 AZ Amsterdam, The Netherlands

Abstract

tt

cca

,astuv

Hn

K

1

flt

Published in Sensors and Actuators B 114, issue 1, 530-541, 2006which should be used for any reference to this work

1

Understandinghow theinformationiscodedin largeneuronalnetworksisoneof themajorchallengesfor neuroscience.A possibleapproacho investigatetheinformationprocessingcapabilitiesof theneuronalensemblesis given by theuseof dissociatedneuronalculturescoupledo microelectrodearrays(MEAs).

Here,we describea new strategy, basedon MEAs, for studyingin vitro neuronalnetwork dynamicsin interconnectedsub-populationsoforticalneurons.Therationaleis to sub-dividetheneuronalnetwork into communicatingclusterswhile preservingahighdegreeof functionalonnectivity within eachconfinedsub-population,i.e. to achieveacompromisebetweenacompletelyrandomlargeneuronalpopulationandpatternednetwork, suchascurrentlyusedwith conventionalMEAs.To thisend,wehaverealizedandfunctionallycharacterizedaPt microelectrodearraywith anintegratedEPONSU-8clusteringstructure

llowingtoconfinefiverelatively largeyetinterconnectedspontaneouslydevelopingneuronalnetworks(i.e.thousandsof cells).Theclusteringtructureconsistsof fivechambersof 3mm in diameterinterconnectedvia 800�m longand300�m widemicrochannelsandis integratedonheMEA of 60thin-film Ptelectrodesof 30�m diameter. Testsof thePtmicroelectrodes’stabilityunderstimulationshowedastablebehaviorp to 35,000voltagestimuli andthebiocompatibilitywas assessedwith theculturesof dissociatedrat’s corticalneuronsachieving cultures’iability up to 60daysin vitro.Comparedto conventionalMEAs, the monitoringof spontaneousandevoked activity and the computationof the Post-StimulusTimeistogram(PSTH)within theclustersclearlydemonstrates:(i) thecapabilityto selectively activate(throughpoly-synapticpathways)specificetwork regionsand(ii) theconfinementof thenetwork dynamicsmainly in thehighly connectedsub-networks.

eywords: Microelectrodearray;Clusters;SU-8adhesion;In vitro neuronalnetworks;Long-termstimulation;Plasticity;Neurodynamics;Bio-MEMs

. Introduction

The understandingof the underlying principles of theunctionalplasticity of the brain is a currentresearchchal-engein neurophysiologyand constitutesa necessarystepoward implementing these same principles in physical

∗ Correspondingauthor. Tel.: +41327205520; fax:+41327205711.E-mail address: Luca.Berdondini@unine.ch(L. Berdondini).

devices.Performingthisresearchatthebrainlevel introducesa very high degreeof complexity dueto thegranddegreeofconnectivity. Conversely, the level of a singleor a few neu-ronsdoesnotprovideasufficient functionalconnectivity. Inthis sense,smallneuronalensemblesbecomean interestingintermediatelevel for thisresearchandcouldallow acquiringalow-level,basicunderstandingof thenetwork functionality.

Nowadays,nervous tissuescanbe culturedin vitro andkept alive for several months,while preservingtheir adap-

tive properties[1–6]. Furthermore,microelectrodearrays(MEAs), initiated by Pine [7] and Grosset al. [8], havebecomenow areliableinterfacingtechniquecapableof estab-lishingabidirectionalcommunicationbetweenapopulationof connectedneuronsandtheexternalworld[9].Currentstateof theartMEAs grantlow impedanceelectrodes(lower than1 M� at 1kHz), a goodcellularsealing[10–13]anda highchargeinjectioncapacityfor anefficientstimulation[14–16].

Thefunctionalcharacteristicsof theMEAspermitmid- tolong-termrecordingsof both spontaneousandevoked neu-ronalnetwork activity patternsandof their spatio-temporalevolution. This allows investigatinglearningprocessesandmemory[17,18] andmorerecentlyalsothe network devel-opment[19]. However, dueto the network complexity andto the fact that spontaneousactivity aswell asstimulationstendtoexhibit complex patternssynchronizedover thewholenetwork, the identificationof plasticity changes(reinforce-mentor inhibition) is ratherdifficult. In this case,it couldbe of fundamentalimportanceto designneuralnetworks atwill. This complexity canbealleviatedby network pattern-ing, usingadhesionpromoters/inhibitors[20–27], structuredPDMSlayers[28,29],agar-basedmicrochambers[30,31]andneurocages[32]. In thiscase,however, therandomnatureof

the network and its functional plasticity becomerelativelylimited.

We have chosena differentapproachto direct the orga-nization of the neuronalnetwork in order to facilitate theidentificationof interconnectedpathways.This approachisbasedon physicalbarriersfor clusteringthe network intoa numberof randomsub-networks while preservinga highdegreeof functionalconnectivity within andamongthesub-populations.The sub-networks areinterconnectedvia inte-gratedmicrochannels.An additional anticipatedfunction-ality of the clusteringstructureis that it might also allowlocalizing the stimulation in a given cluster, i.e. neuronalsub-network. In this way, a spatialsegregation of networkresponsivenesscanbepromotedandfunctionalareascanbeelicitedandmonitored.

Among differentmaterialsthat could be usedfor realiz-ing theclusteringstructures,wehaveoptedfor EPONSU-8.This materialhasbeenusedfor a wide rangeof Bio-MEMsdevices,astheinsulatorfor a microelectrodearray[33], forfabricating3D structuresin contactwith neuronalcells[34]andaspackagingmaterial[35]. It showsareducedbiofoulingcomparedto otherMEMs materialsandappearsto be bio-compatible[36]. Moreover, from the technologicalpoint of

2

Fig. 1. Design of the microelectrodes array with the integrated clustering s1 microelectrode in each channel. Two temperature sensors (Pt-RTDs of 1�) a14 mm× 14 mm. (b) Microelectrodes location in the lateral and (c) in the ce

tructures. (a) Chip layout; 60 microelectrodes are distributed in the five clusters andkre integrated outside the clustering structure. The overall chip dimensions arentral clusters.

view, EPONSU-8is aconvenientmaterialfor thepatterningof structureswith a high aspectratio in a singlephotolitho-graphicprocess.

This articledescribestherealizationandfunctionaleval-uationof a Pt-basedMEA with integratedSU-8 clusteringstructures,so calledclusteredMEAs, aimedat the investi-gationof thefunctionalplasticityof neuronalnetworks.TheMEA consistsof 60 thin-film Pt microelectrodesof 30�min diameterrealizedon a Pyrex substrate.The five cluster-ing chambersinterconnectedvia microchannels,realizedinEPONSU-8,are350�m in heightand,respectively, 3mmin diameterand300�m wide.A schematicrepresentationofthedevice is shown in Fig. 1a.

In the following sections,the microfabrication of theclusteredMEA is describedto begin with. This is followedby theevaluationof thefunctionallifetimeof boththemicro-electrodesandof theclusteringstructures(namelytheSU-8adhesionon a Pyrex substrate).Given that the investigationof thenetwork functionalplasticityrequiresa long-term(upto severalmonths)stimulation/recordingof theelectrophysi-ologicalactivity, thereforeanadequatefunctionallifetime oftheMEAs is aprerequisite.Previousstudieshaveshown thatlarge amplitudebiphasicpotentialsthat are typically usedin the in vitro stimulation[9,37–39]can lead to undesiredelectrochemicalreactions,which may causemechanicalorchemicaldegradationof themicroelectrodesurfaceaffectingtfetstsi

ctbcr

2

2

bTelm saA

so

or 800�m in lengthand300�m in width. Additionally, fora futureintegrationwithin a micro-incubationchamber, twoplatinum resistive temperaturedevices (Pt-RTDs) are inte-gratedonchipoutsidetheclusteringstructure.Thesesensorsaredesignedat1k� with aserpentinestructureof 10�m inwidth anda total lengthof 10mm.

The overall chip dimensionsare14mm× 14mm andfitinto aglassreservoir of 2cm of internaldiameter.

2.2. Chip fabrication

Thethin-film processrequiresthreemasklayerstopattern:(i) the metallic layer for the microelectrodes,the connect-ing leads,the bonding padsand the temperaturesensors,(ii) the insulationlayer to definethe electrodesandcontactpadsand (iii) the SU-8 clusteringstructures.The fabrica-tion processis summarizedin Fig. 2. The 4in. diametersubstrates(Pyrex 7740 wafers,500�m thick from SensorPrepServices,Elburn) wereinitially cleanedby successiveimmersionsin fuming HNO3, in bufferedhydrofluoricacid(BHF) andrinsedin deionizedwater. Themetallayerswerepatternedby a lift-of f processby using a two layer pho-

Fig. 2. Cross-sectionview of the microfabricationprocess.(1) Cleaningof the Pyrex 7740wafer, (2) patterningof the metal layer by lift-of f, (3)depositionof thesilicon nitride insulationlayerby PECVD,(4 and5) pho-tolithographyandopeningby SF6/O2 plasmaetchingof themicroelectrodeandcontactpadareas,(6) silanizationandspin-coatingof theSU-8adhe-sion layer, (7 and8) spin-coatingandpatterningof the350�m thick layerof SU-8.

3

he stimulation/recordingcapabilities and therefore theunctionallifetime of the MEAs [40,41]. The resultsof thevaluationof thePt microelectrodesduring35,000stimula-ion cyclesillustratingtheirgoodfunctionalstabilityarepre-ented.Concerningthesecondaspectof thefunctionallife-ime assessment,clusteringstructureadhesionontoa Pyrexubstrate,we show that the SU-8 adhesioncan be greatlymproved by usinganEPON825-basedadhesionlayer.

Thebiologicalfunctionalevaluationwas performedwithulturesof rat’sembryocorticalneuronsin orderto ascertainhedevice biocompatibility, thestimulation/recordingcapa-ilities of the microelectrodesand the functionality of thelustersfor plasticity researchon neuronalnetworks. Theesultsdemonstrateclearlysucha functionality.

. Methods

.1. Chip design

TheMEA with clusteringstructure(Fig. 1) wasdesignedy usinga MEMs CAD designsoftware(Expert,Silvaco).heMEA providesatotalof 60microelectrodeswith adiam-terof 30�m distributedasfollows: 11 microelectrodesper

ateral cluster, 1 microelectrodeper open-channeland 12icroelectrodesin the centralcluster. The microelectroderenumberedandtheclustersareidentifiedby a letter(fromto E).Theclusteringstructuresdefinefive clusteringchamber

f 3mm in diameter, connectedby open-channelsof 500�m

toresisttechnology. It consistsin spin-coatingthefirst pho-toresistlayer (LOR3B from MicroChemCorp., USA) anda thermal treatmentat 170◦C for 10min. Then, a secondpositive photoresist(AZ1813 from Shipley) is spin-coatedand pre-baked at 110◦C for 1min. The waferswere thenexposedin avacuum-contactmode(maskalignerMa-6,Karl-Suss)at 365nm (45mJ/cm2) and,without a post-bake, theexposedlayerdevelopedin a1:4aqueoussolutionof AZ400K(Hoechst).This resultsin a photoresiststructurewith thesecondlayeractingasanintegratedmaskfor themetalevap-oration.Then,a layer of 200A of titanium was depositedby evaporationas an adhesionlayer for a 1300A layer ofplatinum.Thelift-of f processwasperformedin asolutionofRemover PG (MicroChem)at60◦C for 1h.

The insulation layer (4000A thick silicon nitride) wasdepositedby plasmaenhancedchemicalvapor deposition(PECVD) and the microelectrodesandbondingpadswereopenedby a secondphotolithographyfollowed by SF6/O2plasma etching. The photoresist(AZ1518, Shipley) wasstrippedin acetoneand the wafer rinsedwith isopropanol.Finally, thewaferwascleanedin apiranhasolution(5min inconcentratedH2SO4 and5min afteraddingadropof H2O2).

Thewaferswerethenprocessedwith a photostructurableepoxy, the EPON SU-8 100 (MicroChem) to define the350�m high clusteringstructures.Prior to thespin-coatingof SU-8,thewaferswere:(i) silanizedin a 10%solutionof

fd

5.ong-

z-d0re-

erf

C)

on

edrets

oe-r

edsle

with thecommerciallyavailableMultichannelpre-amplifier.It is realizedon a standardepoxy substrate(FR-4) with asinglemetallayer(copper18�m thick) insulatedwith TaiyoPSR-4000.

After thechipmounting,aculturechamberof 2cm inter-nal diameterwas gluedontothePCB.Theareasof thePCBinsidetheglasschamberwereadditionallycoveredbyathicklayerof PDMS(Sylgard185,Dow Corning)in ordertoavoidany potentialtoxic effectsthat might arisefrom the insula-tion layer of the PCB.The PDMS was preparedby mixing1 partof catalystwith 10 partsof Sylgardandpolymerizedat 120◦C for 1h. This thermaltreatmenthasthe additionalbenefitof actingasahard-bakingstepfor theSU-8layer.

2.4. Set-up for evaluating the microelectrode stabilityunder stimulation

A waveformgenerator(Hewlett-Packard,HP33120A)toapply a biphasic rectangularpulse with an amplitude of1.5Vp–p,aperiodof 500�sandadutycycleof 50%wasused.To acceleratethetests,thevoltagestimuli wereappliedeach2s. Themicroelectrodeswereconnectedto asensingresistor(Rsens) andtoadigitaloscilloscope(TektronixTDS360).Thesensingresistorwasusedtoconvertthecurrentsignalpassingthroughthemicroelectrode–electrolyteinterfacein avoltagesignal.Ideally, usingasensingresistor(shuntamperometer),

ldenga-ghal

-tp

6r-t,eednt-

aleednd

rea

4

3-glycidoxypropyltrimethoxysilanein toluenewith 0.5%owater, rinsedwith isopropanolanddried with nitrogenan(ii) spin-coatedwith anadhesionlayerbasedonEPON82Following a largeseriesof testsfor optimizingtheadhesilayer, the best resultsso far were obtainedby mixing 1of EPON825,100mg of photo-acid-generator(triarylsulfonium hexafluoroantimonatesalt,Aldrich), 10mg of silaniing agent(3-glycidoxypropyltrimethoxysilane,Aldrich) an1ml of thesolvent�-butyrolactone(Aldrich).Next,SU-810wasspin-coated,pre-bakedonahot-plate(ramptemperatufrom 35to 95◦C andtotaltime105min) andexposed(wavelength365nm anddose1800mJ/cm2) usinga maskalign(AL-6, ElectronicVision) in proximity mode(separationo50�m). Following a post-bake (65◦C for 10min and95◦for 15min), the SU-8 was developedin PGMEA (15minandrinsedin isopropanolanddeionizedwater. Theadhesilayerwas developedduringthedevelopmentof theSU-8.

Finally, thewafersweredicedandthechipswerecleanin deionizedwaterandoxygenplasma(at roomtemperatufor 20min) in ordertoremoveall remainingtracesof solvenandnon-reticulatedpolymers.

2.3. Chip packaging

Thechipsweremountedonprintedcircuit boards(MicrPCBAG, Thundorf,Switzerland)thathasa12mm in diamteraperturein thecentreto providebacksidelight accessfoculture imagingwith an invertedmicroscope,wire-bondandthe wires insulatedwith an epoxyresin.The PCB hadimensionsof 49mm× 49mm andits designis compatib

Rsensshouldvalue0�, but in this casean amplifier woubeneeded.In orderto keeptheset-upsimple,ahighervaluresistor, whichhasbeendefinedempiricallyaftermeasurithe appliedvoltage (Uapplied) with respectto the stimultion voltage(Ustim) for different resistorsand minimizintheerrorbetweenthemwasused.It hasbeenfoundthatwitRsens= 3.3k� theappliedvoltagepracticallyfitted theidebiphasicpulseof thewaveformgenerator.

Theexperimentswereperformedin aphysiologicalsolution (NeurobasalMediumwithout l-glutamineandwithouphenolredfromGIBCO)allowing toneglecttheOhmicdroin thesolution.

ThemeasuringsystemwasautomatedbyusingLabView(from NationalInstruments)andtheGPIBinstrumentsinteface.In additionto themeasurementof thetransientcurreni(t), thelong-termstabilityof theinjectioncurrentversusthnumberof appliedstimuli was alsoevaluatedandexpressastheroot-mean-square(rms)valueof theinjectioncurre(computedin Matlab environmentfrom the measuredtransientcurrentacquiredeach10stimuli).

2.5. Cortical neuronal cultures

Dissociatedneuronalcultureswereobtainedfromcerebrcorticesof embryonicratsat gestationalday18 (E18).Thcerebralcorticesof four to five rat embryoswerechoppinto small pieces,dissociatedby enzymaticdigestion itrypsin 0.125%– 20min at 37◦C – and finally trituratewith afire-polishedPasteurpipette.Dissociatedneuronsweplatedonto poly-d-lysine and laminin coatedMEAs, in

100(l dropcoveringtheelectroderegion(≈1200cells/mm2);1h later, when cells adheredto the substrate,1ml ofmediumwasadded.Thecellswereincubatedwith 1%peni-cillin/streptomycin,1% glutamax,2% B-27 supplementedNeurobasalMedium(Invitrogen),in ahumidifiedatmosphere5%CO2, 95%air at37◦C [42]. Fifty percentof themediumwas changedtwiceaweek.

2.6. Electrophysiological experimental set-up

Theexperimentalset-upis basedon theMEA 60 System(MultichannelSystem,MCS, Reutlingen,Germany), con-sisting of a mountingsupportwith integrated60 channelspre-andfilter amplifier (gain1200×), a personalcomputerequippedwith a PCI data acquisitionboard for real-timesignalmonitoringandrecording,an invertedopticalmicro-scope,ananti-vibrationtableandaFaradaycage.Theculturechamberwassealedusingsiliconeringsandaperfusionsys-temaswell asa temperaturecontrolwereinstalledallowinglong-timeexperiments.Signalswererecordedandmonitoredby usinganin-housedevelopedsoftwarefor real-timespikedetectionanda commercialsoftware,MCRack (MCS) foron-line visualizationsand raw datastorage.CommerciallyavailableMEAs (MEA 200/30,MCS) wereusedfor valida-tion andcomparison.

2

wscowlp s(aisav st

ls( eu

3

3P

i

Fig. 3. (a) Optical views of a singlechip (14mm× 14mm) and(b) of thepackageddevice (50mm× 50mm).

mainly limited by defectsintroducedduringopeningthesil-icon nitride passivationlayer. This technologicalproblemisoftenassociatedwith largechipareas.Ontheotherhand,theuseof the two photoresistlayerslift-of f techniqueallowedachievinganexcellentreproducibilityandwell-definedmetalstructurecontours.Residuesof SU-8,observedoccasionallyon theelectrodesaftertheSU-8development,couldbeeffi-ciently removed by introducingtheoxygenplasmacleaningstep.Finally, theaveragethicknessof theSU-8structuresof346± 8�m, with a typical error in dimensionsreferredtothedesignedone,lowerthan3%confirmsagoodfabricationprocesscontrol.

The first clusteringstructureswere realizedwithout anadhesionlayerfor theSU-8.It wasfoundthatalbeittheSU-8adhesionwas goodandpassedthescotchtestafterprocess-ing, thestructureslifted-off after2–3daysin physiologicalsolution. As observed also by Voskerician et al. [36], thelossof adhesionstartedfrom thesidesof thestructures.This

5

.7. Data analysis

Thetypical observedelectrophysiologicalactivity in net-orksof dissociatedneuronsusuallyrangesfrom stochasticpikingto organizedpopulationbursting.A populationburstonsistsof episodesof activity (i.e. denselypacked spikes)ccurringat many channelsandspreadover the entirenet-ork. Thesepackagesgenerallylast from hundredsof mil-

isecondsupto secondsandaretimedividedby long“silent”hases.To investigatethenetwork behavior in spontaneoui.e.notstimulated)conditions,weusedalgorithmsfor spikendburst detection.The spike detectionalgorithmconsists

n a hard-thresholdcrossing,computedusingfive timesthetandarddeviationof theraw signal.Theburstsweredefinedssequencesof threespikesoccurringin lessthan100ms.Aisualcontrolof theautomaticdetectionresultscorroboratehereliability of thesemethods(seeSection3).

Toevaluatetheresponsivenessof thenetwork toelectricatimulation,wecomputedthePost-StimulusTimeHistogramPSTH),whichrepresentstheprobabilitytoevokearesponsponastimulusdeliveredfrom specificsite[43].

. Results and discussion

.1. Pt-MEA with thick SU-8 clustering structure andt-RTDs

TheclusteredMEA anda ready-to-usedevice areshownn Fig. 3. A fabricationyield of 85%was achieved andwas

Fig. 4. (a) Measured injection current on one microelectrode, 30�m in diameter, upon 100 and 35,000 voltage stimulations of 1.5 Vp–p and 500�s in period.(b) Calculated injected rms current during a stimuli versus the number of applied stimuli on one microelectrode.

6

Fig. 5. Pictures of in vitro E18 cortical neurons. (a) On a conventional MEAdiv. The metal at the microelectrode sites is 40�m in diameter.

, at 2 div and (b–d) on the clustered MEA: (b) at 2 div, (c) at 11 div and (d) at 11

Fig. 6. Two culture health indicators on conventional and clustered MEAs. (a) Average cell culture life duration computed on the five best results fromtwobatches. (b) Mean cell densities computed on 10 2-div samples.

Fig. 7. Effect of the clustered organization on the mean spiking (a) and bursting (b) spontaneous activities; computed using active channel recordings of,respectively, 8 clustered MEA and 10 conventional MEA batches of 21–34 div. The corresponding recorded raw signals are shown on the right and the linestands for the detection threshold.

7

Fig. 8. Spontaneous bursting activity raster plot recorded on 21–30 div cultures for 60 s on the (a and b) clustered and (c) conventional MEAs. Raster plot spotsare sorted by clusters on the clustered MEA and the burst summation is shown in the lower part. (d) Close-ups of the bursting activity on the clustered MEAshowing consistent time lags.

8

resultmotivatedthedevelopmentof anadhesionlayerabletoestablishachemicalbondingbetweenthesiliconnitrideandtheSU-8.Devicesfabricatedwith theadhesionlayerbasedon the EPON825 allowed to achieve goodSU-8 adhesionontoPyrex substratesin physiologicalsolutionup to severalmonths.

The Pt-RTD temperaturesensors,althoughnot usedinthiswork,werecharacterizedbymeasuringtheirresistivity at22.4◦CbeforeandafterthePECVDsiliconnitridedepositionandpatterning.A homogenousdecreaseof theresistivity of15% was measuredfor a total metal thicknessof 1495A.This resultedin anaverageresistivity (fivepointsperwafer;10 wafersmeasured),ρTi–Pt, of 1.87× 10−7 �m and in anaverageresistanceof the Pt-RTDs of 1.0534k� ± 38� at22.4◦C. Themeasuredtemperaturecoefficient of resistance(TCR),α, was2.8364± 0.16× 10−3 C−1.

3.2. Microelectrode evaluation under long-termstimulation

Asalreadyemphasized,themicroelectrodefunctionalsta-bility underlong-termvoltagestimulationisof crucialimpor-tancefor studyingplasticity in neuronalnetworks. Fig. 4ashows a typical exampleof the recordedinjection currentsafter100and35,000biphasicvoltagestimuli with theampli-tpt(pa .

crs

mean-squarecurrentof 13.3± 0.35�A per microelectrode(rms current density 1.88A/cm2) over 35,000 stimuli of1.5Vp–p.

The averageinjectedcharge densitiesover 35,000stim-uli for eachphaseof the stimulustrain (1–3) are, respec-tively, Q1 = 4.62× 10−4 C/cm2, Q2 =−4.53× 10−4 C/cm2

andQ3 =−1.47× 10−4 C/cm2. A dissymmetrybetweenthepositive and negative phaseis observed in terms of theinjectedcharge (0.9× 10−5 C/cm2) andis not, asmight beexpectedfor a 50%duty cycle high-amplitudevoltagestim-uli, compensatedfor in the discharge phase[44–46]. How-ever, the stability of the injection currentdemonstratesthatthisdoesnotaffecttheelectrodes’functionalreliability. Thishasbeenfurther confirmedby cyclic voltammetry, wherepracticallyunchangedvoltammogramswereobtainedpriorto andafter the functionalstability experiments(resultsnotshown). Moreover, the charge injection dissymmetrydoesnotseemto induceany problemsin theelectrophysiologicalrecordings(seebelow).

3.3. Functional experiments with rat’s cortical neuronalcultures

Theeffectsof theclusteringstructureon thespontaneousandevokednetwork activities wereevaluatedby comparingtramlc

ito

F A and llr longing chei tered M cording tob , built a arrayi cale is

9

udeanddutycyclecorrespondingtotheneuronalstimulationrotocolusedsubsequently. Thethreephasesof thestimulus

raincorrespond,respectively, to theanodic(1) andcathodic2) pulsesfollowedby adischarge(3).Thepracticallysuper-osedresponsesafter100and35,000stimuli demonstrateandequatestimulationfunctionalityof thePtmicroelectrodes

Thestabilityof thechargeinjectionis moreevidentuponomputingtheroot-mean-squareof theinjectioncurrentwithespectto thenumberof deliveredstimuli (Fig.4b). It canbeeenthat Pt microelectrodesshow a stableinjection root-

ig. 9. Post-Stimulus Time Histogram (PSTH) on (a) a clustered MEepresents the PSTH computed for the recording microelectrodes, bendicated by the letters of the connected clusters. Note how the cluselonging clusters. Cluster B is not active. (b) The graph reports a 8× 8 grid

t is not possible to distinguish different features of the PSTH. TheX-axis s

he recordingson clusteredandconventionalMEAs. Neu-onal cultureson the clusteredMEAs were routinely keptctiveandhealthyup to 45–60days(Figs.5 and6). Further-ore,thedistributedorganizationwasfully compatiblewith

ong-termcell culturesandcomparableto that observed ononventionalMEAs.

Comparingthe spontaneousspiking and bursting activ-ty ratesrecordedon, respectively, conventionaland clus-ered MEAs, as shown in Fig. 7, no apparentdifferencef the mean level of activity was observed. However, a

(b) a conventional MEA. (a) The graph reports a 10× 6 grid where each ceto different clusters as indicated by the five labels. Electrodes in theannels arEA is able to evoke responses of different shapes and features actheccording to the conventional MEA layout. Note that within the entire[0, 400] ms, while theY-axis scale is [0, 1].

moredetailedexaminationof the burstingactivity patternsbetweenandwithin theclustersrevealedcertaindifferences.In comparisonwith conventionalMEA, whereculturestendrapidly toward a synchronizedburstingactivity distributedover theentiremicroelectrodearray, on theclusteredMEAstherecordingsexhibitedadistributionof synchronousburstsin the clustersand asynchronousburstsbetweenthe clus-ters(Fig. 8b andc). This is evenmoreevidentwhenall theburst occurrencesfrom all the microelectrodesareconsid-ered(Fig. 8b andc, lower part).Theasynchronousburstingbetweentheclustersresultedin asequentialpatternof burstswith consistenttime lags(Fig. 8d) andwas routinely foundfrom one culture to another. Thus, asynchronousburstingareasmight be linked to the distributedorganizationof theculture.

The behavior of the network in interconnectedsub-populationswas even more evident looking at the evokedactivity patterns.The network responsivenessto localizedstimulationswas evaluatedby meansof the Post-StimulusTime Histogramasshown in Fig. 9. It canbe seenthat theclusteredMEA impartsa distinctvariability on thenetworkresponsesto a singlestimulation(i.e. a train of 50 pulsesat0.2Hzand1.5Vp–pfromasingleelectrode):thePSTHshapewas differentfor separatedclusters,while it maintainedthesamefeatureswithin thesamecluster. Thesub-populationinthe stimulatedcluster(clusterD in Fig. 9a) shows a higherprobabilityof evokedresponse,yettheevokedactivity clearlypropagatesto theconnectedclusters.This variability of thenetwork responsivenessis not observed on a conventionalMEA (Fig. 9b), wherethenetwork freelygrows in theavail-

10

Fig. 10. Mean latency of the evoked responses computed fro

m data ofFig. 9on (a) a clustered MEA and (b) a conventional MEA.

able space,producinga denselayer of strongly connectedcells. Upon stimulation in this condition, eachsite of thearrayshowsaverysimilar response.

Fig. 10 presentsthe computedmean latency of theresponsesto the deliveredstimuli. Comparedto the stimu-lated cluster, the other clustersshow a longer latency anda larger variability in the responsetime, demonstratingthesub-populationsinterconnectionsandthedifferentpathwaysinvolved in the network dynamics.It is worthwhile to notethataverysimilarbehavior is observedbetweenthenetworkin thestimulatedclusterandthewholenetwork onaconven-tionalMEA.

4. Conclusions

In this paper, the fabrication,the functional evaluationandthebiologicalvalidationof a microelectrodearraywithanintegratedSU-8clusteringstructureweredescribed.Thepurposeof the interconnectedclusterstructureis to reducetheisotropicnetwork distribution,aswell asto discernlocalrelationsbetweencell growth,activity andtheir involvementaccordingto their useasinput,outputor processingareas.

The recordingof the spontaneousactivity demonstrateda clear differencein the distribution of the bursting activ-ity betweenthe conventional and clusteredMEAs. Thus,ts .Tirctsitsnuoi

dsglsic )at

A

t ,

ProjectNeuroBit)andbyOfficefederaldel’ educationetdelascience(Contract01.0197)aregratefullyacknowledged.ThemicrofabricationwascarriedoutattheComLab(Neuchatel).Theauthorswould alsolike to thankMs. S. Pochonfor thedevicepackaging.

References

[1] S.Marom,D. Eytan,Learningin ex-vivo developingnetworksof cor-tical neurons,in: J. van Pelt,al. et (Eds.),Progressin Brain Research,vol. 147, Elsevier ScienceLtd., 2004,pp. 189–199(Chapter14).

[2] G. Shahaf,S. Marom, Learning in networks of cortical neurons,J.Neurosci.21 (22) (2001) 8782–8788.

[3] K. Ganguly, M.-M. Poo,L. Kiss, Enhancementof presynapticneu-ronal excitability by correlatedpresynapticandpostsynapticspiking,Nat. Neurosci.3 (10) (2000) 1018–1026.

[4] H.-Z.W. Tao,L.I. Zhang,G.-Q.Bi, M.-M. Poo,Selective presynapticpropagationof long-termpotentiationin definedneuralnetworks, J.Neurosci.20 (9) (2000) 3233–3243.

[5] G.-Q. Bi, M.-M. Poo, Distributed synaptic modification in neuralnetworksinducedby patternedstimulation,Nature401(6755)(1999)792–796.

[6] Y. Jimbo, Recordingneural activity by electrode-arraysubstrates,Electrochemistry67 (3) (1999) 276–279.

[7] J.Pine,Recordingactionpotentialsfrom culturedneuronswith extra-cellular micro-circuit electrodes,J. Neurosci.Methods2 (1) (1980)19–31.

s

[

[

[

[s

[

[,

[

.[

[,

11

his approachdoesnot preventlocal burstingactivity (in theub-population)but allows its distribution over theclustershe clusterfunctionality of confining the network dynam-

cs mainly in the sub-networks was further confirmedbyecordingthenetworkevokedactivity. ThecomputedPSTHslearlyshow thelocalizationof thestimulationin theclustero which the stimulatingelectrodebelongs,yet the evokedignalpropagatesthroughtheconnectingchannels,influenc-ng the activity of the othersub-populations.The shapeofhePSTHdoesnotchangewithin theclusterfrom which thetimulusisdelivered,whilecleardelaysanddifferencesin theumberof evokedspikesarevisible in theothers.Thestim-latedclusterresponsewas very similar to theonerecordednaconventionalMEA confirmingthenetwork organization

n interconnectedsub-populations.An organizationof neuronalnetworks in interconnecte

ub-populationsis animportantstepfurtheronin theinvesti-ationof thefunctionalandanatomicalaspectsof distributed

earningprocesses.The proof of conceptof the clusteringtructurefunctionalityopensnew prospectsfor investigatingn vitro interactionsamonglarge assembliesof neuronsoro-culturedpopulations(e.g.,neocortex andhippocampusswell asthebasiclearningmechanismsandnetwork plas-

icity.

cknowledgments

The supportby the EU communityunderthe Informa-ionSocietyandTechnologiesProgramme(IST-2001-33564

[8] G.W. Gross,A.N. Williams, J.H. Lucas,Recordingof spontaneouactivity with photoetchedmicroelectrodesurfacesfrom mousespinalneuronsin culture,J. Neurosci.Methods5 (1–2) (1982) 13–22.

[9] S. Potter, Distributed processingin culturedneuronalnetworks, in:M.A.L. Nicolelis (Ed.), Progressin Brain Research,vol. 130, Else-vier ScienceLtd., 2001,pp. 49–62(Chapter4).

10] W. Rutten, J.M. Mouveroux, J. Buitenweg, C. Heida, T. Ruardij,E. Marani,E. Lakke, Neuroelectronicinterfacingwith culturedmul-tielectrodearrays toward a cultured probe, in: Proceedingsof theIEEE, July 2001.

11] V. Kiessling,P. Fromherz,B. Muller, Extracellularresistancein celladhesionmeasuredwith a transistorprobe,Langmuir 16 (7) (2000)3517–3521.

12] M.P. Maher, J. Pine, J. Wright, Y.-C. Tai, The Neurochip:a newmultielectrodedevice for stimulating and recording from culturedneurons,J. Neurosci.Methods87 (1) (1999) 45–56.

13] W.G. Regehr, J. Pine, C.S. Cohan, M.D. Mischke, D.W. Tank,Sealingcultured invertebrateneuronsto embeddeddish electrodefacilitateslong-termstimulationandrecording,J. Neurosci.Methods30 (2) (1989) 91–106.

14] T. Nyberg, O. Inganas,H. Jerregard, Polymer hydrogel microelec-trodesfor neuralcommunication,Biomed.Microdevices4 (1) (2002)43–52.

15] J.D.Weiland,D.J.Anderson,M.S. Humayun,In vitro electricalprop-ertiesfor iridium oxideversustitaniumnitride stimulatingelectrodesIEEE Trans.Biomed.Eng. 49 (12) (2002) 1574–1579.

16] A. Blau, C. Ziegler, M. Heyer, F. Endrest, G. Schwitzgebel,T.Matthies,T. Stieglitz, J.-U. Meyer, W. Gopel, Characterizationandoptimizationof microelectrodearraysfor in vivo nerve signalrecord-ing andstimulation,Biosens.Bioelectron.12 (9–10)(1997)883–892

17] Y. Jimbo, T. Tateno,H.P.C. Robinson,Simultaneousinduction ofpathway-specificpotentiationanddepressionin networks of corticalneurons,Biophys.J. 76 (2) (1999) 670–678.

18] T.B. DeMarse,D.A. Wagenaar, A.W. Blau, S.M. Potter, Theneurallycontrolled animat: biological brains acting with simulatedbodiesAuton. Robot.11 (3) (2001) 305–310.

[19] J. Van Pelt, P.S. Wolters, M.A. Corner, G.J.A. Ramakers, W.L.C.Rutten,J. Van Pelt,Long-termcharacterizationof firing dynamicsofspontaneousburstsin culturedneuralnetworks,IEEE Trans.Biomed.Eng. 51 (11) (2004) 2051–2062.

[20] H. Moriguchi, K. Takahashi,Y. Sugio, Y. Wakamoto,I. Inoue, Y.Jimbo, K. Yasuda,On-chip neural cell cultivation using agarose-microchamberarray constructedby a photothermaletchingmethod,Electr. Eng. Jpn.146 (2) (2004) 37–42.

[21] A.K. Vogt, L. Lauer, W. Knoll, A. Offenhausser, Micropatternedsubstratesfor the growth of functionalneuronalnetworks of definedgeometry, Biotechnol.Prog.19 (5) (2003) 1562–1568.

[22] C.K. Yeung, L. Lauer, A. Offenhausser, W. Knoll, Modulation ofthe growth and guidanceof rat brain stemneuronsusing patternedextracellularmatrix proteins,Neurosci.Lett. 301 (2001) 147–150.

[23] P. Heiduschka,I. Romann,H. Ecken, M. Schoning,W. Schuhmann,S. Thanos,Defined adhesionand growth of neuroneson artificialstructuredsubstrates,Electrochim.Acta 47 (1–2) (2001) 299–307.

[24] M. Scholl, C. Sprossler, M. Denyer, M. Krause,K. Nakajima, A.Maelicke, W. Knoll, A. Offenhausser, Orderednetworks of rat hip-pocampalneuronsattachedto silicon oxide surfaces,J. Neurosci.Methods104 (1) (2000) 65–75.

[25] C.L. Klein, M. Scholl, A. Maelicke, Neuronal networks in vitro:formation and organizationon biofunctionalizedsurfaces,J. Mater.Sci.: Mater. Med. 10 (12) (1999) 721–727.

[26] S. Saneinejad,M.S. Shoichet,Patternedglass surfacesdirect celladhesionand processoutgrowth of primary neuronsof the centralnervous system,J. Biomed.Mater. Res.42 (1) (1998) 13–19.

[27] C. Wyart, C. Ybert, L. Bourdieu, C. Herr, C. Prinz, D. Chatenay,Constrainedsynapticconnectivity in functionalmammalianneuronalnetworks grown on patternedsurfaces,J. Neurosci.Methods117 (2)(2002) 123–131.

h-n2)

t-6

nol,

o,edngs

wed

[33] M.O. Heuschkel, M. Fejtl, M. Raggenbass,D. Bertrand,P. Renaud,A three-dimensionalmulti-electrodearray for multi-site stimulationand recording in acutebrain slices, J. Neurosci.Methods114 (2)(2002) 135–148.

[34] Y. Choi, R. Powers, A.B. Frazier, M.G. Allen, V. Vernekar, M.C.LaPlaca, High Aspect Ratio SU-8 Structuresfor 3-D Culturingof Neurons, American Society of Mechanical Engineers,Micro-ElectromechanicalSystemsDivision Publication(MEMS), 2003,pp.651–654.

[35] L.A. Francis, C. Bartic, A. Campitelli, J.-M. Friedt, A SU-8liquid cell for surface acoustic wave biosensors,in: Proceed-ings of SPIE—TheInternationalSociety for Optical Engineering,2004.

[36] G. Voskerician, M.S. Shive, R.S. Shawgo, H. von Recum, J.M.Anderson, M.J. Cima, R. Langer, Biocompatibility and biofoul-ing of MEMS drug delivery devices, Biomaterials24 (11) (2003)1959–1967.

[37] D.A. Wagenaar, J. Pine,S.M. Potter, Effective parametersfor stim-ulation of dissociatedculturesusing multi-electrodearrays,J. Neu-rosci. Methods138 (1–2) (2004) 27–37.

[38] Y. Jimbo, H.P.C. Robinson,A. Kawana,Strengtheningof synchro-nized activity by tetanicstimulationin cortical cultures:applicationof planar electrode, IEEE Trans. Biomed. Eng. 45 (11) (1998)1297–1304.

[39] H. Oka, K. Shimono,R. Ogawa, H. Sugihara,M. Taketani, A newplanar multielectrodearray for extracellular recording: applicationto hippocampalacuteslice, J. Neurosci.Methods93 (1) (1999)61–67.

[40] S. Mailley, M. Hyland,P. Mailley, J.A. McLaughlin,E.T. McAdams,Thin film platinum cuff electrodesfor neurostimulation:in vitroapproachof safe neurostimulationparameters,Bioelectrochemistry

cece

s,

k,e,

l-s-–

l-o-

s-es,

12

[28] L. Griscom,P. Degenaar, B. LePioufle,E. Tamiya,H. Fujita, Tecniques for patterningand guidanceof primary culture neuronsomicro-electrodearrays,Sens.ActuatorsB, Chem.83 (1–3) (20015–21.

[29] E. Ostuni,R. Kane,G.M. Whitesides,C.S. Chen,D.E. Ingber, Paterningmammaliancellsusingelastomericmembranes,Langmuir1(20) (2000) 7811–7819.

[30] Y. Sugio,K. Komjima, H. Moriguchi, K. Takahashi,K. Yasuda,Aagar-basedon-chipneural-cell-cultivationsystemfor stepwisecontrof network patterngenerationduring cultivation, Sens.ActuatorsBChem.99 (1) (2004) 156–162.

[31] I. Suzuki,Y. Sugio,I.H. Moriguch,A. Hattori, K. Yasuda,Y. JimbPatternmodificationof a neuronalnetwork for individual-cell-baselectrophysiologicalmeasurementusing photothermaletchingof aagarosearchitecturewith a multielectrodearray, in: IEE ProceedinNanobiotechnology, 2004.

[32] M.P. Maher, J. Pine, J. Wright, Y.-C. Tai, The Neurochip:a nemultielectrodedevice for stimulating and recording from culturneurons,J. Neurosci.Methods87 (1) (1999) 45–56.

63 (1–2) (2004) 359–364.[41] A. Norlin, J. Pan,C. Leygraf, Investigationof interfacial capacitan

of Pt, Ti and TiN coatedelectrodesby electrochemicalimpedanspectroscopy, Biomol. Eng. 19 (2–6) (2002) 67–71.

[42] G.J.Brewer, Isolationandcultureof adult rat hippocampalneuronJ. Neurosci.Methods71 (2) (1997) 143–155.

[43] F. Rieke, D. Warland, R. de Ruyter van Steveninck, W. BialeSpikes: Exploring the Neural Code, The MIT Press,CambridgMassachusetts,1997.

[44] J. McHardy, L.S. Robblee,J.M. Marston,S.B. Brummer, Electricastimulation with Pt electrodes (4) factors influencing Pt disolution in inorganic saline, Biomaterials 1 (3) (1980) 129134.

[45] L.S. Robblee,J. McHardy, J.M. Marston,S.B. Brummer, Electricastimulationwith Pt electrodes(5) the effect of protein on Pt disslution, Biomaterials1 (3) (1980) 135–139.

[46] S.B. Brummer, M.J. Turner, Electrical-stimulation of nervousystem-principleof safechargeinjectionwith noble-metalelectrodBioelectrochem.Bioenerg. 2 (1) (1975) 13–25.