A microfluidic cell co-culture platform with a liquidfluorocarbon separator

Bryson M. Brewer & Mingjian Shi & Jon F. Edd &

Donna J. Webb & Deyu Li

# Springer Science+Business Media New York 2014

Abstract A microfluidic cell co-culture platform that uses aliquid fluorocarbon oil barrier to separate cells into differentculture chambers has been developed. Characterization indi-cates that the oil barrier could be effective for multiple days,and a maximum pressure difference between the oil barrierand aqueous media in the cell culture chamber could be aslarge as ~3.43 kPa before the oil barrier fails. Biologicalapplications have been demonstrated with the separate trans-fection of two groups of primary hippocampal neurons withtwo different fluorescent proteins and subsequent observationof synaptic contacts between the neurons. In addition, thequality of the fluidic seal provided by the oil barrier is shownto be greater than that of an alternative solid-PDMS valvebarrier design by testing the ability of each device to block lowmolecular weight CellTracker dyes used to stain cells in theculture chambers.

Keywords Cell culture . Cell co-culture . Synaptic formation

1 Introduction

Our understanding of cell and molecular biology has vastlyimproved over the last 100 years; however, the commonly

used in vitro cell culture/co-culture tools that lead to theseadvancements are not dissimilar to those used in the early days(Meyvantsson and Beebe 2008). Large dishes and flaskstypically used in biological cell culture offer little to no spa-tiotemporal control over the physical and chemical cues thatcan affect cell function and behavior (Meyvantsson and Beebe2008). Thus, the detailed study of cell-cell interactions inculture is somewhat limited. The rapid development ofmicrofluidic cell culture systems provides a solution to over-come these limitations. With the key features of laminar flow,optical transparency, biocompatibility, ease of fabrication, anddevice features comparable in size to cells, microfluidic cellculture systems are uniquely suited for manipulating the cel-lular microenvironment, addressing many limitations of tradi-tional cell culture methods and allowing for the in-depthinvestigation of cellular interactions (Gao et al. 2011;Whitesides 2006; Young and Beebe 2010).

A number of microfluidic cell culture/co-culture platformshave been developed that demonstrate improved control overcell-cell interactions. Designs that use micropatterning of spe-cific molecules on a substrate allow for selective attachment ofspecific cell types to pre-determined regions (Bhatia et al.1997; Kane et al. 2006; Khetani and Bhatia 2008). Co-flowing laminar streams of fluid have also been used to loadand treat cells within particular regions of a microfluidicchannel (Lucchetta et al. 2005; Takayama et al. 1999). Alter-native strategies employ the concept of compartmentalization,using microgrooves (Taylor et al. 2005), collagen tracks(Ravula et al. 2007), and semi-permeable membranes(Kimura et al. 2008) to separate cell populations. New cellbiology assays have been developed based on the novelcapabilities these platforms provide; however, there is still aneed for better manipulation and more control of the cellularmicroenvironment. For example, a successful co-culture de-vice should allow for (1) loading of different cell types intospecific compartments, (2) culture of each cell type in its

B. M. Brewer : J. F. Edd :D. Li (*)Department of Mechanical Engineering, Vanderbilt University,Nashville, TN 37235-1592, USAe-mail: deyu.li@vanderbilt.edu

M. Shi :D. J. WebbDepartment of Biological Sciences and Vanderbilt Kennedy Centerfor Research on Human Development, Vanderbilt University,Nashville, TN 37235, USA

D. J. WebbDepartment of Cancer Biology, Vanderbilt University, Nashville,TN 37235, USA

Biomed MicrodevicesDOI 10.1007/s10544-014-9834-8

optimal media until reaching confluence, (3) manipulation ofone cell population’s microenvironment without affectingother nearby populations, and (4) real-time, high resolutionimaging (Gao et al. 2011).

The desired functions listed above have been successfullyincorporated into a microfluidic cell co-culture device previ-ously reported by the authors. In that design, a pneumaticallyor hydraulically controlled polydimethylsiloxane (PDMS)barrier is used to separate different cell populations (Gaoet al. 2011). Using this approach, cells of different types canbe cultured to confluence and transfected or stained individu-ally in cell culture chambers that remain separated by thePDMS valve barrier. Then, upon release of the barrier, differ-ent cell populations are allowed to interact in a controlledmanner. This platform has been successfully used for variousapplications, including the observation of dendritic spine andsynapse formation among primary hippocampal neurons sep-arately transfected with pre- and post-synaptic markers (Gaoet al. 2011; Majumdar et al. 2011) and the investigation of theeffects of secreted angiocrine factors on the regulation oftumor growth and migration (Brantley-Sieders et al. 2011).However, despite these positive results, the platform stillsuffers from some limitations. For example, the valve-enabled PDMS barrier cannot be considered truly reversiblebecause cells, neurites, or other biological material trappedunder the PDMS barrier during its actuation would incurphysical damage (Hosmane et al. 2011). In addition, therectangular shape of the region under the PDMS barrier resultsin a sieve valve effect (Marcus et al. 2006; Melin and Quake2007) that can permit leakage of some small molecules be-tween the separated cell culture chambers. Although thiscould be partially remedied by using rounded profiles in thebarrier region (Unger et al. 2000), forming a complete seal isstill a challenge. In addition, using rounded channels requiresmore complicated and time-consuming fabrication (Fordyceet al. 2012), making this strategy less than ideal. Thus, in thispaper we report an alternative approach to address the issuesof barrier reversibility and small molecule leakage withoutsacrificing the functionality provided by the valve-enabledPDMS barrier platform.

One means of achieving well-sealed, reversible separationof cell populations in microfluidic co-culture platforms is touse a non-solid barrier. Several instances have been demon-strated in the literature for a variety of applications of thisstrategy. For example, a pinned interface between two immis-cible fluids has been achieved by selectively changing thesurface chemistry in specific regions of a microchannel(Atencia and Beebe 2005). Using this technique, a stablegas/liquid interface, or virtual wall, can be formed. Afterchannel modification, surface tension will counteract the ef-fects of gravity and confine water to the hydrophilic regions ofthe microchannel, provided the pressure difference betweenthe air and water phases does not exceed a critical value. This

method is particularly useful for applications in which dis-solved gas species need to be removed from a liquid, such asextracting dissolved oxygen from water (Hibara et al. 2005).Gas/liquid interfaces may also be employed to expose a gasspecies to a liquid phase to initiate a chemical reaction oradjust the pH of a solution (Atencia and Beebe 2005). Inaddition to forming pinned interfaces between gases andliquids, surface modification has been used to stabilize theinterface between different immiscible liquids for both vertical(Zhao et al. 2002) and horizontal (Hibara et al. 2002) liquid/liquid interfaces, where surface modification was used toreduce the time of formation and instability of the pinnedinterface. Pinned interfaces are well-suited for use in drugpartitioning studies (Surmeian et al. 2002), ion extraction fromsolvents (Maruyama et al. 2004), enzymatic reactions(Maruyama et al. 2003), and formation of membranes throughinterfacial polymerization (Zhao et al. 2002). However,pinned interfaces created using surface modification presentseveral challenges for biological studies involving cell culture.First, changing the surface chemistry of microchannels re-quires an additional fabrication step when compared to othermicrofluidic platforms, complicating device production. Stud-ies have also shown that with decreasing surface wettabilityvia treatment with chemicals such as octadecylsilane, thegrowth and proliferation of attached cells may be reduced(Altankov et al. 1996). In addition, the critical differentialpressure across a pinned liquid/gas or liquid/liquid boundarygenerated using surfacemodificationmay be too low to ensureinterface stability in long-term cell culture applications. Thus,difficulties would likely arise in attempting to implement thistechnique in the construction of a reversible cell culturebarrier.

An improved technique incorporates microposts intomicrofluidic channels to enhance the stability of the interfacebetween two fluids by increasing the critical differential pres-sure required to rupture the interface (Berthier et al. 2009;Tetala et al. 2009; Xu et al. 2006). One demonstration of thismethod employs an array of microposts to separate a water-containing middle channel from two outer chambers contain-ing air (Lai et al. 2011). Thus, a virtual wall barrier is formedbetween two chambers. As long as the critical pressure is notexceeded, water will remain confined to the central channeland not flow into the outer chambers. With careful operation,this method could likely be inverted so that an air layer is usedto separate two chambers filled with water. While this type ofdevice was shown to be effective at trapping cells in theregions between microposts (Lai et al. 2011), a few challengesprevent it from being useful for the type of neurobiology andcancer biology studies conducted using the solid-PDMS valvebarrier platform. First, in order to create an interface, thePDMS channels were allowed to recover their natural hydro-phobic state before loading any water into the device. Hydro-phobic channels are less compatible with the more user-

Biomed Microdevices

friendly passive pumping method (Walker and Beebe 2002)used in our previous studies (Gao et al. 2011). In addition, forchannels with sharp corners or other irregular features, ahydrophobic surface increases the chance of small air bubblesforming in the cell culture chamber. Problematically, the pres-ence of air in the cellular microenvironment can be detrimen-tal to cell viability (Skelley and Voldman 2008; Sung andShuler 2009; Zheng et al. 2010). Reversibility of the valvemay also be difficult, particularly when trying to use an air-based virtual wall. Once the channel walls in the valve regionare wetted by the aqueous phase, reintroducing a stable airbarrier will be very difficult without first completely dryingout the entire device. Finally, the permeability of PDMS to air(Hosokawa et al. 2004; Kang et al. 2008; Merkel and Bondar2000) could result in the degradation of an air-based virtualwall in long-term cell culture by evaporation or condensation.Thus, significant modification of the virtual wall concept isneeded for successful implementation into a microfluidic cellco-culture platform.

Here, we present a microfluidic cell co-culture platformthat uses a liquid fluorocarbon oil (Fluorinert FC-40) to re-versibly separate two cell populations without the need for asurfactant. In this design, two outer cell culture chambers areconnected via a series of microgrooves to a middle channel inwhich FC-40 can be loaded to separate cells from one anotherwithout crushing the cell bodies/processes in the barrier re-gion. Liquid fluorocarbons such as FC-40 have been used in avariety of microfluidics applications including droplet actua-tion (Chatterjee et al. 2006), nanoparticle synthesis(Shestopalov et al. 2004), and cell encapsulation (Kösteret al. 2008), and its biocompatibility is well documented(Köster et al. 2008; Lowe et al. 1998; Shi et al. 2010). TheFC-40 oil barrier is created by taking advantage of the immis-cibility of the fluorocarbon oil (located in the middle channel)and the aqueous media found in the outer cell culture cham-bers, which can form a stable interface in the reduced cross-sectional area of the connecting microgrooves. This wasachievedwithout any special surface modifications (other thanstandard oxygen plasma treatment used for device fabrication)or the necessity of hydrophobic channel walls, and the barriercan remain stable for multiple days. The device was also usedto observe the initial synaptic contact between two popula-tions of primary hippocampal neurons separately transfectedwith different fluorescent proteins. Additionally, results showthat the oil separator more effectively blocks the transport ofsmall CellTracker dyes, indicating that it could provide abetter fluidic seal than the solid-PDMS valve barrier platform.Finally, it is worth noting that although PDMS was used tofabricate these devices, other materials such as thermoplasticscould be used as well, since the operation of the barrierrequires no deformation to achieve separation. This is impor-tant because PDMS is permeable to many small molecules(Toepke and Beebe 2006) and incompatible with most organic

solvents (Lee et al. 2003). Thus, for applications in whichthese issues are a concern, oil barrier platforms made ofalternative materials that do not suffer from permeabilityproblems could be advantageous.

2 Materials and methods

2.1 Device fabrication

As shown in Fig. 1a–c, a two-chamber design of the liquidfluorocarbon oil barrier platform consists of two outer cellculture chambers (1 mm×5 mm×100 μm; width (w), length(l), and height (h), respectively) separated by a 200 μm (w)×100 μm (h) oil channel. The series of microgroovesconnecting the oil channel to the outer culture chambers areeach 100 μm (l)×50 μm (w)×5 μm (h). The device wasfabricated using standard soft lithography techniques(McDonald and Whitesides 2002; Whitesides et al. 2001).

A reusable master mold was created using SU-8 multi-layer photolithography. The 5 μm layer for the connectingmicrogrooves was patterned by spinning SU-8 2005(Microchem, Newton, MA) at 2000 RPM for 35 s. Afterpre-baking at 95 °C for 2 min, the SU-8 was exposed througha photomask (CAD/Art Services Inc., Bandon, OR) with anexposure energy of 390 mJ/cm2 using a Novacure 2100 SpotCuring System (EXFO Inc., Quebec, CANADA). The 5 μmlayer was completed after a 3 min post-bake at 95 °C anddevelopment with SU-8 developer (Microchem, Newton,MA). Similarly, the 100 μm layer for the oil channel and cellculture chambers was patterned by spinning SU-8 2050 at1650 RPM for 35 s, pre-baking at 65 °C for 5 min and 95 °Cfor 20 min, manually aligning the photomask with respect tothe features of the 5 μm layer and exposing at 390 mJ/cm2,post-baking at 65 °C for 1 min and 95 °C for 10 min, anddeveloping for ~10 min with SU-8 developer.

After fabrication of the master, liquid PDMS base polymerwas mixed with the curing agent (Ellsworth Adhesives, Ger-mantown, WI) at a 10:1 mass ratio and poured over the mold,which was degassed in a vacuum for ~1 h and cured for atleast 2 h at 70 °C. After curing, the solid PDMS layer was cutand peeled from the mold, and holes were punched to form theinlets and outlets of each channel. Next, the PDMS device anda glass coverslip (VWR Vista Vision, Suwanee, GA) weretreated in a plasma cleaner (Harrick Plasma, Ithaca, NY) andbonded together, forming an irreversible seal. Deionized (DI)water was immediately added to the channels in order toprevent the PDMS from reverting to its natural hydrophobicstate (Duffy et al. 1998). Pyrex cloning cylinders (FisherScientific, Pittsburgh, PA) were attached to the inlet and outletreservoirs of the cell chambers using uncured liquid PDMS. Inaddition, 0.02” I.D./0.06” O.D. Tygon microbore tubing(Cole-Parmer, Vernon Hills, IL) was inserted into the inlet

Biomed Microdevices

and outlet holes of the middle oil channel and sealed withuncured liquid PDMS. Finally, the assembled device wasplaced in a 70 °C oven for ~1 h to cure the PDMS sealsaround the reservoirs and tubing, during which time the chan-nels remain filled with DI water.

In addition to the two-chamber design, a four-chamberversion of the oil barrier device was fabricated using the same

protocol, as shown in Fig. 1d–f. The four-chamber version issimply a modified two-chamber design with an outer cellculture chamber and oil channel attached via microgroovesto the right and left of the inner culture chambers. Thus, thefour-chamber device contains three oil channels that can beused to independently separate all of the culture chambers. It isworth noting that for the purpose of demonstration, we

Fig. 1 Design of thefluorocarbon oil barriermicrofluidic platform. a Three-dimensional cartoon of a loaded,static oil barrier device. bA cross-section diagram of the deviceshowing the fluidic connectionbetween the outer cell culturechambers and the middle oilchannel via microgrooves (not toscale). cA photographdemonstrating a static oil barrierdevice separating red dye in theleft culture chamber from bluedye in the right culture chamber.A U.S. penny is shown for scale.d Three-dimensional cartoon of astatic four-chamber oil barrierdevice in which the middle oilchannel is loaded with FC-40.Note that for the experimentsperformed in this paper, only thecenter oil channel was filled withoil. However, all chambers maybe isolated from one another byfilling all three oil channels withFC-40. eA cross-section diagramof the four-chamber device,showcasing the fluidic connectionbetween the glia in the outerchambers and the neurons in theinner chambers. fA photographdemonstrating a static oil barrierisolating the two chambers withred dye on the left from the twochambers with blue dye on theright

Biomed Microdevices

employed only the central oil channel for all experimentsperformed in this study. The dimensions of the key featuresof the four-chamber platform are identical to those of the two-chamber design.

2.2 Static and dynamic operation

The FC-40 oil barrier device can function in two schemes: astatic approach and a dynamic approach. In the static scheme,the first step is to remove any excess liquid from the inlet andoutlet reservoirs of the culture chambers. This helps to achievea gross pressure balance between the two phases during initialloading. Then, a syringe is used to slowly inject FC-40 into themiddle channel, displacing the aqueous phase already there.Once the entire channel and tubing are occupied by the oil, a¾′′ AACO binder clip is used to clamp the tubing, first on thesyringe end and then on the open end. This ensures that the oilchannel does not become pressurized, causing oil leakage intothe outer culture chambers. Media can then be added to thereservoirs as necessary. On the other hand, the dynamicscheme uses a syringe pump (Chemyx Inc., Stafford, TX) toflow oil continuously through the middle channel. A syringefilled with FC-40 is directly attached to the inlet tubing of themiddle oil channel; a syringe pump is then used to control theflow rate (up to ~125 μL/min) to maintain a stable oil barrier.There are advantages for both approaches, depending on theapplication. The static oil barrier is more user-friendly forbiological studies, as the entire device can be contained withina Petri dish, making it easier to maintain a sterile environmentas the device is transferred between culture hoods, incubators,and microscopes. However, greater control over the oil/waterinterface was demonstrated using the dynamic oil barrier,which could be advantageous for applications in which theuse of a syringe pump is not problematic. Both strategies wereused in this study for characterization of device performance,although only the static oil barrier was used for biologicaltesting.

Proper media flow in the cell culture chambers wasachieved using a passive pumping method in which the pres-sure drop created by a difference in fluid level height betweenthe inlet and outlet reservoirs drives the flow, as we have donepreviously (Gao et al. 2011). Avariety of factors can influencethe flow rate in this method, including fluid level, the shapeand completeness of the meniscus formed in each reservoir,evaporation, and time since loading media (Lynn and Dandy2009). Thus, well-determined flow rates were not achieved forthis approach; however, as demonstrated previously for sim-ilar microfluidic cell culture applications, the passive pumpingmethod is sufficient for supplying cells with fresh media, aswell as removing waste from the culture region (Gao et al.2011). Nonetheless, the Hagen-Poiseuille equation (Δp=Rhyd⋅Q, where Δp is the pressure drop, Rhyd is the hydraulic resis-tance of a rectangular channel, and Q is the flow rate) can be

used to estimate the range of flow rates realized in the device.Using this method, the maximum flow rate just after loadingmedia was estimated to be ~20–60 nL/s, with the flow ratedecreasing over time until the culture media is refreshed.

2.3 Device characterization

2.3.1 Long-term stability of the static oil barrier

To demonstrate the long-term stability of the FC-40 separator,a static oil barrier was loaded and a water-based fluorescentdye was used to visualize the location of the oil/water inter-face. Fluorescein isothiocyanate (FITC) (Thermo Scientific,Rockford, IL) was added to one culture chamber after loadingthe oil barrier; images were then captured periodically using aNikon AZ100 fluorescence microscope (Nikon InstrumentsInc., Melville, NY). At the end of the experiment, DI waterwas injected into the middle oil channel, replacing the FC-40.A final image was captured just after (<5 min) removing theoil barrier.

2.3.2 Investigation of the shape of the oil/water interface usingconfocal microscopy

A more thorough investigation of the oil/water interface wasconducted using confocal microscopy. Both a static and dy-namic oil barrier with oil flow rates up to 150 μL/min wereanalyzed. For the static case, an oil barrier was introduced intoa device initially filled with FITC. Then, using a QuorumWaveFX spinning disk confocal system equipped with aNikon Eclipse Ti microscope (Nikon Instruments Inc., Mel-ville, NY) and a Hamamatsu ImageEM-CCD camera with aPlan Fluor 40× objective (N.A. 1.3), a z-stack of images(vertical resolution of 0.2 μm) focused on the oil/water inter-face at a single microgroove was obtained using MetaMorphimaging software (Molecular Devices, Sunnyvale, CA). Sim-ilarly, z-stacks of images were acquired for dynamic oil bar-riers generated using different flow rates after the oil/waterinterface had stabilized in the microchannel. These imageswere post-processed using ImageJ, allowing for visualizationof the overhead view and side-view cross-sections, as well asthree-dimensional reconstructions of the interface. The radii ofcurvature of the interface in both planes were determined fromthese reconstructions to derive the pressure drop across theoil–water interface.

2.4 Cell culture protocols

2.4.1 Loading neurons into the two-chamber oil barrier device

Before use in biological study, all microfluidic oil barrierdevices were sterilized under UV radiation in a Laminar flowhood for 1–2 h. After sterilization, the cell culture chambers

Biomed Microdevices

were coated with 1 mg/mL poly-L-lysine (PLL) (Sigma-Al-drich, St. Louis, MO) in 0.1 M borate buffer (pH 8.5) for 12 hin an incubator at 37 °C. Excess PLL was removed from thechambers by flowing sterilized DI water through the channelsfor ~2 h. The freshly coated chambers were then filled withB27-supplemented NeurobasalTM media (neuronal media)(GIBCOTM Invitrogen, Carlsbad, CA). After waiting 20–30 min for equilibration, hippocampal neurons that wereisolated from dissected brains of E19 rat embryos (Goslinet al. 1998) were loaded into the two inlet reservoirs for theneuronal chambers (50,000 cells per culture chamber) at adensity of 5×105 cells/mL of neuronal media. Note that no oilwas present in the middle channel at this point. The cells wereallowed to attach to the PLL coated glass substrate of the cellculture chambers by incubating at 37 °C with 5 %CO2 for 3 h.After attachment, ~300 μL of B27-supplementedNeurobasalTM media that had been conditioned for 24 h overa confluent monolayer of glial cells (glia-conditioned media)was added to the inlet reservoirs of each culture chamber;about half of that volume was added to the outlet reservoirs toslow down the flow rate. The glia-conditioned media wasreplenished every 36 h, along with the removal of the wastemedia collected in the outlet reservoirs. A summary of thisprotocol is shown in Fig. 2a.

2.4.2 Loading glia/neuron co-culture into the four-chamberoil barrier device

The loading of glia/neuron co-culture into the four-chamberdevice is similar to that with the PDMS valve barrier device(Majumdar et al. 2011, Shi et al. 2013), and the on-chip co-culture of glia helps to increase the transfection efficiency andthe stability of synaptic contacts. First, the four-chamber de-vices were prepared for co-culture by using UV sterilizationand coating with PLL as described above. The coated four-chamber devices were then equilibrated using Minimum Es-sential Medium (MEM) (Invitrogen, Carlsbad, CA) contain-ing 10 % horse serum and 0.6 % glucose (glia culture media)for 20–30 min. Then, glia that were isolated from 2 day old ratpup brains (Goslin et al. 1998) were loaded into the twooutermost culture chambers by adding 50 μL of glial cellsuspension into the inlet of each chamber (25,000 cells/chamber). The devices were incubated at 37 °C for 2 h topermit cell attachment, after which they were filled with400 μL of glia culture media and incubated for 4–5 days untilconfluence was achieved. Note that we did not observe gliaspreading into the neuronal chamber during this process eventhough no oil barrier was implemented between the glialchamber and neuronal chamber. After the glia reached con-fluence in their respective chambers, neurons were loaded intothe two inner culture chambers using the above-describedprocedure. A summary of this protocol is shown in Fig. 2b.To ensure that media flowed from the glial chambers into the

neuronal chambers after loading the neurons, 400 μL of freshneuronal media was added to the glial reservoirs while 200 μLof the same media was added to each neuronal reservoir. Thiswas repeated every 36 h, along with the removal of wastemedia.

2.4.3 Separate transfections of hippocampal neuronswith GFP and mCherry

After 3–7 days in culture, neurons in each chamber weretransfected with GFP and mCherry cDNAs using a modifiedcalcium phosphate method (Zhang et al. 2003). First, a staticoil barrier was loaded as described previously. Note thatspecial care was taken to keep the FC-40 oil at 37 °C (thetemperature of the incubator), as we observed that thermalexpansion of the oil as its temperature increased from roomtemperature (20 °C) to 37 °C resulted in oil leakage into thecell chambers. To accomplish this, the FC-40, syringe, andsyringe tips were heated in the incubator before loading theoil; concurrently, the device was placed on a hot plate justabove the incubator temperature (~39 °C) while loading theoil to maintain a stable temperature. After a stable oil barrierwas loaded, 50 μL of transfection mixtures containing eitherGFP or mCherry cDNAs (3–6 μg each) were added to the twoinlet reservoirs. A drop of culture media was added to thewaste reservoirs after 10–15 min to reduce the fluid flow, andthe device was placed in a 37 °C incubator for 1 h. Thechannels were then washed with HEPES-buffered solution(HBS) (pH 7.15) for an additional hour, after which 300 μLof fresh glia-conditioned B27 NeurobasalTM culture mediawas added to each inlet reservoir. Finally, by releasing theAACO clips and injecting culture media into the inlet tubingof the middle oil channel, the oil barrier was removed byforcing the FC-40 out of the outlet tubing. A summary of thisprotocol is shown in Fig. 2c.

2.4.4 Confocal microscopy visualization and real-timeobservation of synaptic contact

The separation of the two culture chambers by the oil barrierwas verified by analyzing transfected neurons using confocalmicroscopy. First, neurons were prepared for live-cell imagingafter 12–13 days in culture by replacing neuronal media with50 mM HEPES containing B27-supplemented NeurobasalTM

media without phenol red, pH 7.4. Imaging was performed onthe Quorum WaveFX spinning disk confocal system. Cellswere imaged using a 10× ADL objective (NA 0.25), a PlanFluor 40× objective (N.A. 1.3), or a PlanApo 60× TIRFobjective (NA 1.49). GFP and mCherry images were obtainedby exciting with 491-nm and 561-nm laser lines, respectively(Semrock, Rochester, NY). The resulting images were thenanalyzed to ensure that no cells were transfected with theinappropriate fluorescent protein.

Biomed Microdevices

By transfecting neurons in one chamber with the pre-synaptic marker mCherry-synaptophysin and neurons in theother chamber with GFP, synaptic contact between neurites

originating from different culture chambers could be ob-served. The confocal microscope system was used to obtaintwo images (one each for mCherry-synaptophysin and GFP)

Fig. 2 Summary of the biologicalcell-culture protocols using the oilbarrier platform. a Loadingneurons in the two-chamber oilbarrier device. b Loading glia,then neurons into the four-chamber oil-barrier platform. cTransfecting neurons withmCherry and GFP (sameprocedure used in four-chamberdevices). d Staining neurons withCellTracker Red and Green (sameprocedure used in four-chamberdevices)

Biomed Microdevices

at a given location within a culture chamber. For high magni-fication images of synapses, a PlanApo 60× TIRF objective(NA 1.49) was used. Images were acquired every 15 min for12 h using MetaMorph software. The regions along neuriteswhere synaptic contact occurred were then determined byoverlaying the images. In addition, the dynamic formation ofsynapses over time could be visualized using the time-lapsesequence of overlays. Note that both standard two-chamber oilbarrier devices and four-chamber devices with glial co-culturewere used to investigate synapse formation between the sep-arately transfected neuronal populations.

2.4.5 Separate staining of hippocampal neurons usingCellTracker dyes

To determine the effectiveness of the static oil barrier platformin blocking the transport of smaller molecules between culturechambers, cells in each chamber were stained withCellTracker GreenCMFDA and RedCMTPX (Invitrogen, Eu-gene, Oregon) fluorescent dyes. As in the transfection proto-col, primary hippocampal neurons were cultured for 3–7 daysin the microfluidic culture chambers. Then, after loading astable static oil barrier, 50 μL of fresh glia-conditioned B27Neurobasal culture media with CellTracker Red (final con-centration of 0.5 μM) was added to one inlet reservoir; thesame volume of CellTracker Green was added to the otherinlet. The device was allowed to incubate at 37 °C for ~1 h,after which the dyes were flushed using fresh glia-conditionedB27 Neurobasal culture media for 1 h. This was followed bycareful removal of the oil barrier and addition of Neurobasalmedia containing B27-supplement without phenol red with50 mM HEPES, pH 7.4 media. A summary of this protocol isshown in Fig. 2d. As a comparison of device performance, theprotocol was repeated with a solid-PDMS valve barriermicrofluidic device fabricated using previously detailed pro-cedures (Gao et al. 2011).

3 Results and discussion

3.1 Barrier stability

The effectiveness and stability of the FC-40 valve were inves-tigated by loading a static oil barrier and observing how wellthe barrier confined the fluorescent dye FITC to one culturechamber. As shown in Fig. 3, the presence of oil in the middlechannel effectively confined FITC to the loading side for~72 h. However, shortly after removing the oil barrier, theFITC was able to quickly migrate into the opposite chamber.Note that the movement of FITC across the oil barrier regionand into the opposite cell culture chamber is caused primarilyby diffusion, and is likely not the result of a “dragging” effectinduced by the water displacing the oil in the middle channel.

Significant FITC migration was only observed after the oilwas completely displaced and the flow of water in the middlechannel stopped. This is consistent with our way of replacingthe oil with water through injecting water to push the oil out,rather than withdrawing the oil using negative pressure. Theseresults indicate that the static FC-40 valve is capable ofpreventing perfusion of small molecules for extended periodsof time (>3 days).

It is worth noting that all tests were performed at roomtemperature (20 °C); however, for later biological assays,instabilities in the oil barrier were observed when room tem-perature FC-40 was added to a device that had been stored in a37 °C incubator. We observed oil leaking into the cell culturechambers shortly after loading the barrier because of thermalexpansion of the oil as its temperature increased, which can bedetermined from the equation,

ΔV

V i¼ B ΔTð Þ; ð1Þ

whereΔV/Vi is the relative volume change with respect to theinitial volume, B is the volumetric coefficient of expansion(0.0012 °C−1 for FC-40), andΔT is the change in temperature.According to Eq. (1), an expansion of ~2% occurs as the oil isheated from 20 °C to 37 °C. Since the inlet and outlet tubes areclamped, the oil expands into the cell culture chambers,resulting in the observed leakage. Thus, extra care was neededwhen conducting biological studies to ensure that both the FC-40 oil and the device were at the same stable temperature. Thiswas achieved by first pre-heating the FC-40 oil and syringe inthe 37 °C incubator. Then, the entire microfluidic device wasplaced on a 37–40 °C hotplate during the actual loading of theoil barrier. Loading of a static oil barrier typically requires<5 min. By quickly returning the device to the incubator afterloading, a stable oil barrier was maintained. For applicationsin which devices must be removed from the incubator forextended periods of time (i.e. live-cell imaging with the oilbarrier loaded), a temperature-controlled microscope stage isrecommended, as mismatches in expansion coefficients be-tween the oil and the rest of the device may result in barrierinstabilities as the device cools to room temperature. Finally, itis worth noting that the FC-40 is biocompatible and smallleakage into the cell culture chamber is fine as long as inter-ference with imaging is not an issue. In fact, we did notobserve any detrimental effects on cells that had comeinto contact with FC-40 that leaked into the culturechambers.

3.2 Interface characterization

Using confocal microscopy, the interface between the FC-40and the aqueous phases was examined. The obtained z-stack

Biomed Microdevices

of images was post-processed to generate both cross-sectionalviews and three-dimensional reconstructions of the interface.Top and side views of the interface are presented in Fig. 4,which shows a single connectingmicrogroove for both a staticand dynamic oil barrier. From these images, a noticeable layerof water (in green) is observed on the sidewalls of the middlechannels. This wetting of the channel walls by the aqueousphase is likely a result of the hydrophilic nature of the PDMS

after plasma bonding, achieved by keeping the channels im-mersed in water before testing (Ren et al. 2001). In addition,an ultra-thin boundary layer of water appears to be present atthe bottom of the channel on the glass substrate. As expected,the thickness of these aqueous layers decreases as the oil flowrate increases from 0 to 125 μL/min. Although the presence ofan ultra-thin residue film of water across the bottom of the oilbarrier could potentially result in unwanted mass transport

Fig. 3 Demonstration of thelong-term stability of a static oilbarrier device using FITC.Fluorescent images were obtaineda immediately after loading thestatic oil barrier, b after 2 h, c after72 h, and d after removing the oilbarrier (<5 min). Note that theconcentration of FITC in thedevice increases over time dueprimarily to evaporation of theaqueous solvent. The scale bar is1 mm

Fig. 4 Top-view images andcross-sectional reconstructions ofthe oil/water interface in bothstatic and dynamic oil barrierdevices. FITC was used tovisualize the aqueous phase. a–eThe oil flow rate was increasedfrom 0 μL/min to 125 μL/min foreach z-stack of images captured. fThe interfacial pressure dropbetween the higher pressure oilphase and lower pressure aqueousphase as a function of FC-40 flowrate as calculated using theYoung-Laplace equation. Thescale bar in all images is 20 μm

Biomed Microdevices

between culture chambers, sensible leakage across the barrierwas not observed in any of our biological testing.

As illustrated in Fig. 4, the radius of curvature of the oil/water interface in both the overhead and cross-sectional planesdecreases with increasing flow rate. By measuring these radiiof curvature, the pressure drop across the oil/water interfacecan be determined using the Young-Laplace equation,

Δp ¼ γ1

R1þ 1

R2

� �; ð2Þ

in which Δp is the pressure drop, γ is the interfacial tensionbetween water and FC-40 (52.07 mN/m) (Mazutis andGriffiths 2012), and R1 and R2 are the radii of curvature inboth planes. The pressure drop values calculated ranged from0.33 kPa for the static case to 3.43 kPa for the max oil flowrate of 125 μL/min. For an oil flow rate of 150 μL/min,leakage of oil into the cell culture chambers was observed;thus, for an interfacial pressure drop of >~4kPa, the surfacetension force is overcome by the fluidic pressure in the oilphase, resulting in barrier failure. This analysis also suggeststhat by decreasing the microgroove dimensions and subse-quently, the radii of curvature of the interface, the magnitude

of the interfacial pressure drop that can be supported by thedevice will increase. Note that the pressure drop values cal-culated here are comparable to or better than those observed inprevious virtual wall microfluidic platforms (~1.5–1.9 kPa)presented in the literature (Lai et al. 2011).

3.3 Separate transfection of hippocampal neuronswith fluorescent proteins and observation of synaptic contacts

To demonstrate the usefulness of the FC-40 oil barrier plat-form for biological applications, primary hippocampal neu-rons were transfected separately with different fluorescentproteins in each culture chamber, as described in the Methodssection. The results of a typical transfection with GFP andmCherry tagged DNAs are shown in Fig. 5. Hippocampalneurons in each chamber were transfected with the appropriatefluorescent protein; more importantly, no neurons in the GFPchamber were transfected with mCherry, or vice-versa, indi-cating that the oil barrier was effective at separating the twochambers. The successful transfection of neurons with GFPand mCherry without any crossover indicates device function-ality similar to previous successful microfluidic neurobiologyplatforms (Gao et al. 2011; Majumdar et al. 2011; Shi et al.2013).

Fig. 5 Separate transfection of primary hippocampal neurons with fluo-rescent proteins using the static oil barrier platform and the subsequentdynamic observation of synaptic contact. a–b Neurons were transfectedwith mCherry in the top panel and GFP in the bottom panel. The dashedlines illustrate the location of the microgrooves connecting the respectivechambers to the oil channel. Note that no neurons in the mCherrychamber were transfected with GFP, or vice-versa. The scale bar is200 μm. c–eTwo images using different fluorescent filters were obtained

at the same location in one of the cell culture chambers (near the oilbarrier region), demonstrating contact between neurons from the adjacentchambers. Neurons in one chamber were transfected with c GFP, whileneurons in the other chamber were transfected with d mCherry-synaptophysin. e An overlay of the two images allows the locations ofsynaptic contacts to be identified. The areas indicated by the white arrowsare observed synaptic junctions. The scale bar is 10 μm

Biomed Microdevices

The successful separate transfection of primary hippocam-pal neurons with different colored fluorescent proteins allowsfor dynamic observation of synaptic contact between neuritesoriginating from different culture chambers. In addition, byusing the pre-synapticmarker mCherry-synaptophysin insteadof mCherry in one of the chambers, the location of potentialsynapses can be easily identified within a culture chamber. InFig. 5c–e, an area in one of the cell culture chambers near theoil channel is presented that showcases the location of neuritesoriginating from the two culture chambers. By examining theoverlay of the images, the locations at which synaptic contactoccurs can be readily identified. A sequence of images canalso be captured over time, allowing for the observation of thedynamic formation of synapses. This functionality providesthe potential to study molecules important to central nervoussystem synaptic development.

Note that both two-chamber oil barrier devices (using gliaconditioned media) and four-chamber devices that contained aco-culture of glia and neurons were successfully used toseparately transfect two different neuronal populations andobserve synaptic contact. The four-chamber design wasemployed to increase the transfection efficiency, as improvedtransfection efficiency using neuron-glia co-culture has beenpreviously demonstrated in the literature (Majumdar et al.

2011). In addition, the four-chamber platform could be usefulfor applications in which the interactions between severaldifferent cell types are to be investigated.

3.4 Separate staining of hippocampal neurons usingCellTracker dyes

In addition to transfection with fluorescent proteins, cell stain-ing with commercially available CellTracker dyes was alsoused to visualize hippocampal neurons in culture. TheCellTracker dyes differ from the transfection proteins in thatthese molecules can penetrate the cell membrane more easily,resulting in staining of nearly 100 % of the cells in culture.Previously, when staining hippocampal neurons in separatechambers with different colored CellTracker dyes using thesolid-PDMS valve barrier device, crossover of the fluorescentmolecules was often observed if no special treatment was usedto create rounded corners to the microgrooves. As a result,neurons in a chamber that were treated with CellTracker Redexhibited both red and green staining, indicating that theCellTracker Green molecules used in the other culture cham-ber leaked through the pressurized valve barrier. However,when identical tests were performed using the static oil barriermicrofluidic device, no crossover was observed (Fig. 6). This

Fig. 6 Using CellTracker Red and Green to compare leakage of thesolid-PDMS valve barrier microfluidic platform to the static oil barrierdevice. For each device, neurons in one culture chamber were stainedwith CellTracker Red, while neurons in the opposite chamber werestained with CellTracker Green. In the top panels, confocal images wereobtained at the same location in the culture chamber treated withCellTracker Red using both red and green filters. The results indicate that

cell structure was stained with both CellTracker Red and Green, suggest-ing leakage of CellTracker Green across the solid-PDMS valve barrier.However, in the bottom panels, images taken in the chamber treated withCellTracker Red show cell structure stained red only; thus, the oil barrierappears to more effectively block the transport of small CellTrackermolecules between culture chambers. The scale bar is 10 μm

Biomed Microdevices

result indicates that the oil barrier design could provide abetter fluidic seal than the standard solid-PDMS valve barrierplatform. Concurrently, no observed leakage of the smallCellTracker dye molecules validates the claim that no signif-icant diffusion occurs across the ultra-thin residual aqueouslayer that may be trapped below the oil barrier in the middlechannel. Thus, the microfluidic oil barrier platform presentedhere is an effective option for applications in which smallmolecules are used to treat separated cell populations.

4 Conclusion

We have developed a microfluidic cell co-culture platformfeaturing a reversible liquid fluorocarbon (Fluorinert FC-40)barrier that enables separate culture and treatment of cell pop-ulations before and/or after allowing the populations to interactwith one another. The physical capabilities of the device havebeen characterized, showcasing that both the stability of thebarrier and the maximum pressure drop across the oil/waterinterface were equal to or greater than that of similarmicrofluidic platforms reported in the literature. The biologicalrelevance of the device was demonstrated by successful cultureand separate transfection of primary hippocampal neurons withdifferent colored fluorescent proteins, allowing for dynamicobservation of synaptic contact. The device should be applica-ble to other co-culture assays that require separate culture andtreatment of individual cell populations. More importantly, theliquid fluorocarbon oil barrier design appears to provide agreater fluidic seal between culture chambers, and it exhibitsgreater biocompatibility than previous solid-barrier based plat-forms. Even though effective loading/withdrawing of the oilrequires some practice, the improved functionality provided bythe oil barrier microfluidic cell co-culture platform makes itwell-suited for probing cellular and molecular interactions incell biology studies.

Acknowledgments B.B. thanks Wan-Hsin Lin and Mingfang Ao forinstruction and assistance with the confocal microscope. This work wassupported by NIH grants GM092914 (D.J.W.), MH093903 andCA155572 (D.J.W. and D.L.). In addition, this material is based uponwork supported by the National Science Foundation Graduate ResearchFellowship under Grant No. DGE-0909667. Any opinion, findings, andconclusions or recommendations expressed in this material are those ofthe author(s) and do not necessarily reflect the views of the NationalScience Foundation.

References

G. Altankov, F. Grinnell, T. Groth, J. Biomed. Mater. Res. 30, 385 (1996)J. Atencia, D.J. Beebe, Nature 437, 648 (2005)J. Berthier, F. Loe-Mie, V.-M. Tran, S. Schoumacker, F. Mittler, G.

Marchand, N. Sarrut, J. Colloid Interface Sci. 338, 296 (2009)

S.N. Bhatia, M.L. Yarmush, M. Toner, J. Biomed. Mater. Res. 34, 189(1997)

D.M. Brantley-Sieders, C.M. Dunaway, M. Rao, S. Short, Y. Hwang, Y.Gao, D. Li, A. Jiang, Y. Shyr, J.Y. Wu, J. Chen, Cancer Res. 71, 976(2011)

D. Chatterjee, B. Hetayothin, A.R. Wheeler, D.J. King, R.L. Garrell, LabChip 6, 199 (2006)

D.C. Duffy, J.C. McDonald, O.J. Schueller, G.M. Whitesides, Anal.Chem. 70, 4974 (1998)

P.M. Fordyce, C.A. Diaz-Botia, J.L. Derisi, R. Gomez-Sjoberg, Lab Chip12, 4287 (2012)

Y. Gao, D. Majumdar, B. Jovanovic, C. Shaifer, P.C. Lin, A. Zijlstra, D.J.Webb, D. Li, Biomed. Microdevices 13, 539 (2011)

G. Goslin, K. Asmussen, H. Banker, Cult, in Nerve Cells, ed. by K.Banker, G. Goslin (MIT Press, Cambridge, 1998), pp. 339–370

A. Hibara, M. Nonaka, H. Hisamoto, K. Uchiyama, Y. Kikutani, M.Tokeshi, T. Kitamori, Anal. Chem. 74, 1724 (2002)

A. Hibara, S. Iwayama, S. Matsuoka, M. Ueno, Y. Kikutani, M. Tokeshi,T. Kitamori, Anal. Chem. 77, 943 (2005)

S. Hosmane, A. Fournier, R. Wright, L. Rajbhandari, R. Siddique, I.H.Yang, K.T. Ramesh, A. Venkatesan, N. Thakor, Lab Chip 11, 3888(2011)

K. Hosokawa, K. Sato, N. Ichikawa, M. Maeda, Lab Chip 4, 181 (2004)B.J. Kane, M.J. Zinner, M.L. Yarmush, M. Toner, Anal. Chem. 78, 4291

(2006)J.H. Kang, Y.C. Kim, J.-K. Park, Lab Chip 8, 176 (2008)S.R. Khetani, S.N. Bhatia, Nat. Biotechnol. 26, 120 (2008)H. Kimura, T. Yamamoto, H. Sakai, Y. Sakai, T. Fujii, Lab Chip 8, 741

(2008)S. Köster, F.E. Angilè, H. Duan, J.J. Agresti, A. Wintner, C. Schmitz,

A.C. Rowat, C.A. Merten, D. Pisignano, A.D. Griffiths, D.A.Weitz,Lab Chip 8, 1110 (2008)

H.-H. Lai, W. Xu, N.L. Allbritton, Biomicrofluidics 5, 24105 (2011)J.N. Lee, C. Park, G.M. Whitesides, Anal. Chem. 75, 6544 (2003)K.C. Lowe, M.R. Davey, J.B. Power, Trends Biotechnol. 16, 272 (1998)E.M. Lucchetta, J.H. Lee, L.A. Fu, N.H. Patel, R.F. Ismagilov, Nature

434, 1134 (2005)N.S. Lynn, D.S. Dandy, Lab Chip 9, 3422 (2009)D. Majumdar, Y. Gao, D. Li, D.J. Webb, J. Neurosci. Methods 196, 38

(2011)J.S. Marcus, W.F. Anderson, S.R. Quake, Anal. Chem. 78, 3084 (2006)T. Maruyama, J. Uchida, T. Ohkawa, T. Futami, K. Katayama, K.

Nishizawa, K. Sotowa, F. Kubota, N. Kamiya, M. Goto, Lab Chip3, 308 (2003)

T. Maruyama, H. Matsushita, J. Uchida, F. Kubota, N. Kamiya, M. Goto,Anal. Chem. 76, 4495 (2004)

L. Mazutis, A.D. Griffiths, Lab Chip 12, 1800 (2012)J.C. McDonald, G.M. Whitesides, Acc. Chem. Res. 35, 491 (2002)J. Melin, S.R. Quake, Annu. Rev. Biophys. Biomol. Struct. 36, 213

(2007)T. Merkel, V. Bondar, J. Polym. Sci. Part B Polym. Phys. 38, 415 (2000)I. Meyvantsson, D.J. Beebe, Annu. Rev. Anal. Chem. (Palo Alto. Calif) 1,

423 (2008)S.K. Ravula, M.S. Wang, S.A. Asress, J.D. Glass, A. Bruno Frazier, J.

Neurosci. Methods 159, 78 (2007)X. Ren, M. Bachman, C. Sims, G.P. Li, N. Allbritton, J. Chromatogr. B.

Biomed. Sci. Appl. 762, 117 (2001)I. Shestopalov, J.D. Tice, R.F. Ismagilov, Lab Chip 4, 316 (2004)W. Shi, H. Wen, Y. Lu, Y. Shi, B. Lin, J. Qin, Lab Chip 10, 2855 (2010)M. Shi, D. Majumdar, Y. Gao, B. Brewer, C. Goodwin, J.A. McLean, D.

Li, D.J. Webb, Lab Chip 13, 3008 (2013)A.M. Skelley, J. Voldman, Lab Chip 8, 1733 (2008)J.H. Sung, M.L. Shuler, Biomed. Microdevices 11, 731 (2009)M. Surmeian, M.N. Slyadnev, H. Hisamoto, A. Hibara, K. Uchiyama, T.

Kitamori, Anal. Chem. 74, 2014 (2002)

Biomed Microdevices

S. Takayama, J.C. McDonald, E. Ostuni, M.N. Liang, P.J. Kenis, R.F.Ismagilov, G.M.Whitesides, Proc. Natl. Acad. Sci. U. S. A. 96, 5545(1999)

A. Taylor, M. Blurton-Jones, S. Rhee, Nat. Methods 2, 599 (2005)K.K.R. Tetala, J.W. Swarts, B. Chen, A.E.M. Janssen, T.A. van Beek, Lab

Chip 9, 2085 (2009)M.W. Toepke, D.J. Beebe, Lab Chip 6, 1484 (2006)M.A. Unger, H.-P. Chou, T. Thorsen, A. Scherer, S.R. Quake, Science

(80-.) 288, 113 (2000)G.M. Walker, D.J. Beebe, Lab Chip 2, 57 (2002)G.M. Whitesides, Nature 442, 368 (2006)

G.M. Whitesides, E. Ostuni, S. Takayama, X. Jiang, D.E. Ingber, Annu.Rev. Biomed. Eng. 3, 335 (2001)

W. Xu, H. Xue, M. Bachman, G.P. Li, Conf. Proc. IEEE Eng. Med. Biol.Soc 1, 2840 (2006)

E.W.K. Young, D.J. Beebe, Chem. Soc. Rev. 39, 1036 (2010)H. Zhang, D.J. Webb, H. Asmussen, A.F. Horwitz, J. Cell Biol. 161, 131

(2003)B. Zhao, N.O.L. Viernes, J.S. Moore, D.J. Beebe, J. Am. Chem. Soc. 124,

5284 (2002)W. Zheng, Z. Wang, W. Zhang, X. Jiang, Lab Chip 10, 2906

(2010)

Biomed Microdevices