lable at ScienceDirect

Biomaterials 30 (2009) 3847–3853

Contents lists avai

Biomaterials

journal homepage: www.elsevier .com/locate/biomateria ls

The controlled presentation of TGF-b1 to hepatocytes in a 3D-microfluidiccell culture system

Chi Zhang a,b, Ser-Mien Chia c,*, Siew-Min Ong a,b, Shufang Zhang a,d, Yi-Chin Toh a,Danny van Noort a,**, Hanry Yu a,b,c,d,e, f,g

a Institute of Bioengineering and Nanotechnology, A*STAR, The Nanos, #04-01, 31 Biopolis Way, Singapore 138669, Singaporeb NUS Graduate School for Integrative Sciences and Engineering, Centre for Life Sciences (CeLS), #05-01, 28 Medical Drive, Singapore 117456, Singaporec Singapore-MIT Alliance, National University of Singapore, E4-04-10, 4 Engineering Drive 3, Singapore 117576, Singapored Graduate Program in Bioengineering, Centre for Life Sciences (CeLS), #05-01, 28 Medical Drive, Singapore 117456, Singaporee Department of Physiology, Yong Loo Lin School of Medicine, National University of Singapore, #03-03, 2 Medical Drive, Singapore 117597, Singaporef NUS Tissue-Engineering Programme, DSO Labs, National University of Singapore, Singapore 117597, Singaporeg Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

a r t i c l e i n f o

Article history:Received 14 February 2009Accepted 29 March 2009Available online 25 April 2009

Keywords:Transforming growth factorHepatocyte co-culture3D cell constructsMicrofluidicsControlled releaseSoluble microenvironment

* Corresponding author. Tel.: þ65 65164885; fax: þ** Corresponding author. Tel.: þ65 68247120; fax: þ

E-mail addresses: phscsm@gmail.com (S.-M. Chia)(D. van Noort).

0142-9612/$ – see front matter � 2009 Elsevier Ltd.doi:10.1016/j.biomaterials.2009.03.052

a b s t r a c t

3D-microfluidic cell culture systems (3D-mFCCSs) support hepatocyte functions in vitro which can befurther enhanced by controlled presentation of 100–200 pg/ml TGF-b1, thus mimicking the roles ofsupporting cells in co-cultures. Controlled presentation of TGF-b1 is achieved by either direct perfusionor in situ controlled release from gelatin microspheres immobilized in the 3D-mFCCS. Primary hepato-cytes cultured for 7 days with the in situ controlled released TGF-b1 exhibited up to four-fold higheralbumin secretion and two-fold higher phase I/II enzymatic activities, significantly improving thesensitivity of hepatocytes to acetaminophen-mediated hepatotoxicity, compared to hepatocytes culturedwith directly perfused TGF-b1 or without TGF-b1. The controlled presentation of TGF-b1 enhancedhepatocyte functions in microfluidic systems without the complications of co-cultures, allowing forsimplifications in drug testing and other hepatocyte-based applications.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Liver-tissue engineering applications such as testing of drugs[1], pathogens [2,3] or other xenobiotics [4], and the developmentof implantable liver tissues [5,6] or extra-corporeal liver-assisteddevices [7] require in vitro cultured hepatocytes that can maintainhigh levels of metabolic functions over a prolonged period of time[8,9]. Recently, we have developed a 3D-microfluidic cell culturesystem (3D-mFCCS) to recapitulate the in vivo environmental cues tosupport high levels of liver-specific functions in vitro [10]. Oneparticular environmental cue that the 3D-mFCCS tries to recapitu-late is the microcirculation of the liver sinusoids in vivo allowingeffective supply of nutrients to and removal of metabolic wastesfrom the hepatocytes, which are very sensitive to the shear stressand normally protected by the sinusoidal endothelial cells aligningthe liver sinusoids [11]. The 3D-mFCCS has been optimized for

65 68748261.65 64789080.

, danny.van.noort@gmail.com

All rights reserved.

cellular support with efficient mass transfer, shear stress reductionand can be employed in hepatotoxicity testing of drugs [12].

The key to constructing an in vitro model to reliably predict thein vivo drug responses as means to eliminate false lead candidatesearly in the drug development process is to ensure that the culturedhepatocytes exhibit high levels of metabolic functions closelymatching those measured in vivo. Previous studies have found co-cultures of hepatocytes with supporting cells in various culturesystems, e.g. 2D monolayer [13], 3D hepatocyte spheroids [14],micropatterned surfaces [15], to exhibit the desired level of hepa-tocyte functions in vitro. Unfortunately, establishing co-cultures inmicrofluidic systems can be challenging: hepatocytes and fibro-blasts exhibit very different physical and biological characteristicsin microfluidic systems, creating operational complexity, such asnon-uniform distributions during cell seeding, different cellproliferation rates and culture requirements of the two cell types,as well as potential interference of hepatotoxicity results due to thepresence and metabolic activities of the supporting fibroblasts [16].

To avoid such operational complexity, we exploited our mech-anistic understandings that low levels of TGF-b1 secreted byfibroblasts can enhance hepatocyte functions in co-cultures [17]. Tomimic the ‘‘co-culture effect’’ in the 3D-mFCCS without culturing

C. Zhang et al. / Biomaterials 30 (2009) 3847–38533848

fibroblasts, TGF-b1 can be presented to the hepatocytes either bydirect perfusion or by in situ controlled release from the localmicroenvironment. TGF-b1 is a bioactive molecule with a veryshort half-life of only a few minutes in aqueous environment [18]but can be preserved from inactivation by encapsulation incontrolled release carriers, hence extending the TGF-b1 activitiesfor up to 1 week [19]. We report here that controlled presentationof TGF-b1 can maintain high levels of hepatocyte functions withoutthe complications of co-cultures in 3D-mFCCS, simplifying drugtesting applications. Such an alternative solution to the generalproblem of co-culturing cells in microfluidics can also be adopted inother hepatocyte-based applications at larger scales, such as bio-artificial liver-assist devices.

2. Materials and methods

2.1. Device fabrication

Microfluidic channels with micropillar arrays were designed using AutoCAD(Autodesk, USA) as previously described [20]. Silicon templates were fabricated bystandard photolithography which involved photoresist application (AZ 4620,Germany), UV light exposure, development and deep reactive ion etching (DRIE)process (Alcatel, France). The microfluidic channels were then obtained by replicamolding poly-dimethylsiloxane (PDMS) (Dow Corning, USA) on the silicontemplates. Holes were made using a puncher (Innovative Technology, USA) forconnecting tygon tubing (Fisher Scientific, USA) via small metal tubings (NewEngland Small Tube Company, USA) to the microfluidic device. The PDMS structureswere plasma-oxidized for 1 min (125 W, 13.5 MHz, 50 sccm, and 40 millitorr,SAMCO, Switzerland) for irreversible bonding to glass coverslips. One inlet of themicrofluidic channel was connected to a cell reservoir, which comprised of a two-way valve with a luer connection (Cole–Palmer, USA) coupled to a 22 G stainlesssteel hypodermic needle (Becton–Dickinson, USA). The other inlet and the outletwere for cell culture medium perfusion. All other fluidic connectors were fromUpchurch (USA). The entire set-up was sterilized by autoclaving it at 105 �C for30 min.

2.2. Polyelectrolytes’ preparation

Methylated collagen and terpolymer of hydroxylethylmethacrylate–methyl-methacrylate–methylacrylic acid (HEMA–MMA–MAA) were synthesized and puri-fied as described previously [21]. 3% terpolymer solution and 1.5 mg/ml methylatedcollagen were used in this study.

2.3. Fabrication of gelatin microspheres (GMs)

GMs were fabricated via glutaraldehyde crosslinking of a gelatin aqueoussolution using a water-in-oil emulsion technique. To create the water-in-oil emul-sion, 4 g of gelatin (Sigma, USA) was dissolved in deionized (DI) water to a finalconcentration of 10 wt.% and 10 ml of 10% glutaraldehyde was added into the gelatinsolution. The solution was preheated to 37 �C and added dropwise into 300 ml ofolive oil under continuous stirring at 2000 rpm for 10 min. Spontaneous gelation ofthe gelatin droplets was then driven by a 15 �C decrease in emulsion temperaturefollowed by 24 h of agitation at 2000 rpm. The resulting microspheres were thenwashed in acetone and recovered by centrifugation at 5000 rpm at 4 �C for 8 min.This recovery process was repeated five times. The crosslinking reaction was thenquenched by agitating the microspheres in 100 mM aqueous glycine solution (Sigma,USA) for 1 h. Lastly, the microspheres were washed three times with DI water andlyophilized. The microspheres were sterilized by washing with 99% ethanol and 1�sterile PBS before used.

2.4. Assessment of the size distribution of the GMs

The GM samples were placed on a glass slide and visualized under a lightmicroscope (Carl Zeiss SV6, Germany). 25 images of the GMs were taken for imageanalysis. The size of the GMs was determined by ImagePro� Plus (Media Cybernetics,USA).

2.5. Scanning electron microscopy (SEM)

GMs were suspended in DI water and 10 ml of the suspension was placed ona poly-lysine coated glass slide, air dried, dehydrated and then platinum sputtered(20 mA, 60 s) before viewing with a field emission scanning electron microscope(JEOL, Japan).

2.6. Characterization of TGF-b1 release profile from the GMs

The amount of TGF-b1 released from the GMs was assessed by a TGF-b1 ELISA kit(Promega, USA) according to the manufacturer’s protocol. TGF-b1 was dissolved in2% BSA solution (PeproTech Inc., USA). To load the GMs, 1 mg of GMs was soaked in0.5 ml TGF-b1 solution at 4 �C overnight. To determine the release profile underdynamic conditions, the TGF-b1-loaded GMs were introduced into the 3D-mFCCSand were immobilized by the micropillars. 2% BSA was perfused continuously at0.05 ml/h through the 3D-mFCCS for 7 days. 1.2 ml perfusate was collected daily toquantify TGF-b1 concentration.

2.7. Estimation of the concentration of in situ controlled released TGF-b1in the 3D-mFCCS

We aimed to controlled release 100–200 pg/ml of TGF-b1 in the 3D-mFCCS. Toobtain the final expected concentration of TGF-b1 in the centre cell culturecompartment, we calculated the number of TGF-b1-loaded GMs in the 3D-mFCCS.According to Section 2.6, we can measure the TGF-b1 concentration from theperfusate. To estimate the number of GMs required achieving a desired finalconcentration, the equation for calculation is as follows:

ðdesired conc: of TGF-b1Þ �max: no: of GMs in 3D-mFCCS� volume of 3D-mFCCSNo: of GMs required

¼ measured TGF-b1 conc:� volume of perfusate

The number of GMs in the 3D-mFCCS is w10 000; the volume of the cell culturecompartment in the 3D-mFCCS is 0.00006 ml; and the volume of the perfusate is1.2 ml.

2.8. Primary rat hepatocyte isolation

Hepatocytes were harvested from male Wistar rats weighing from 250 to 300 gby a two-step in situ collagenase perfusion method [22]. Hepatocytes used in allexperiments had a cell viability of >90%, as determined by Trypan Blue exclusionassay, and a yield of 200–300 million cells.

2.9. Hepatocyte seeding and culture in 3D-microfluidic cell culture system(3D-mFCCS)

3 million hepatocytes and 3000–9000 GMs were resuspended in 1 ml ofmethylated collagen solution. To immobilize the hepatocyte/GM mixture in the 3D-mFCCS, the hepatocyte/GM suspension was withdrawn from the cell reservoir, viathe outlet, using a withdrawal syringe pump (Cole-Parmer, USA). Laminar flowcomplex coacervation was implemented as described previously [10], followed byinfusing culture medium for displacing excess polyelectrolytes. Hepatocytes werecultured on a heater plate (MEDAX, Germany) at 37 �C, with medium perfusedcontinuously at 0.05 ml/h. The perfusate was collected daily for functionalassessment.

2.10. Qualitative assessment of the GM distribution in the 3D-mFCCS

The GMs were fluorescently labeled by soaking them in 0.05% Rhodamine–Dextran (Sigma, USA) at 4 �C overnight and were then mixed with hepatocytes. Themixture was immobilized in the 3D-mFCCS according to Section 2.9. The hepatocyteswere stained with Calcein AM (Molecular Probes, USA) as previously described [10].The cell/GM containing 3D-mFCCS was then imaged using a confocal microscope(Olympus Fluoview, Japan).

2.11. Functional assessment of hepatocytes

All functional data were normalized to cell number (DNA content) quantifiedusing PicoGreen assay (Molecular Probes, USA) according to the manufacturer’sprotocol. Albumin production was assessed by measuring albumin concentration inthe collected perfusate quantified with a rat albumin ELISA quantification kit (BethylLaboratories Inc., USA). Albumin production is expressed as the total albumin (mg)collected in culture medium per million cells over 24 h. UDP-glucuronyltransferase(UGT) and CYP450 activity of the hepatocytes cultured in the 3D-mFCCS weredetermined by infusing 100 mM 4-methylumbelliferone (4-MU) and 50 mM 3-cyano-7-ethoxycoumarin (CEC) (Sigma, USA), respectively, for 4 h at 0.2 ml/h. Theperfusate (800 ml) was collected and the amount of the metabolic products, 4-methylumbelliferyl glucuronide (4-MUG) and 3-cyano-7-hydroxycoumarin (CHC)were analyzed using capillary electrophoresis with laser induced fluorescence (CE-LIF) detection (Prince Technologies B.V., Netherlands) at an excitation wavelength of325 nm [10]. 4-MUG and CHC production are expressed as total amount of 4-MUG(mg) and CHC (mg) collected in culture medium per million cells over 4 h.

C. Zhang et al. / Biomaterials 30 (2009) 3847–3853 3849

2.12. Hepatotoxicity testing in the 3D-mFCCS

Acetaminophen (APAP) was used as a model drug to perform hepatotoxicitytesting. APAP dosing at 20 mM and 40 mM commenced 72 h after seedinghepatocytes into the 3D-mFCCS. Hepatotoxicity was examined after 24 h of drugdosing by quantifying cell viability as a cytotoxicity endpoint (determined bya co-staining method): two fluorescent nuclear dyes i.e. propidium iodide (PI)(Molecular Probes, USA) and Hoechst 33342 (Molecular Probes, USA) were usedto selectively stain the necrotic and total cell population. The staining waseffected by perfusing the mixture of the reagents at a flow rate of 0.05 ml/h for50 min. The concentrations of the two dyes were 25 mg/ml (PI) and 5 mg/ml(Hoechst 33342), respectively. The cells were then fixed by perfusing 3.7%paraformaldehyde (PFA) for 30 min. All stained samples were imaged witha confocal microscope (Olympus Fluoview, Japan) at 543 nm and 405 nm exci-tation. For each 3D-mFCCS, a 100 mm optical section (with 2 mm step size) wasacquired at 4 sites along the length of the channel. A 3D stack was obtained andthe number of objects for both red (dead cells) and blue (total cells) channelswas quantified by counting. The average viability of the hepatocytes wascalculated as:

A

B entrance centre

GMs

cell seeding

m-collagen

bubble trapvalve

valvevalve

terpolymer

3D-µFCC

inlet

Fig. 1. Primary hepatocytes and gelatin microspheres (GMs) in the 3D-mFCCS. (A) Schematconfocal images of the immobilized mixture of primary hepatocytes and GMs in the 3D-mcompartment, centre of the cell culture compartment and near the exit of the cell culture

Viability ¼

P

optical sections

No: of blue objects�No: of red objectsNo: of blue objects

No: of optical sections� 100%

3. Results

To mimic the role of supporting cells in co-cultures, we presented100–200 pg/ml TGF-b1 either by direct perfusion or by in situcontrolled release from GMs inter-dispersed with the primary hepa-tocytes immobilized in the 3D-mFCCS. We compared the effects ofdirect perfusion of TGF-b1 to in situ controlled release of TGF-b1 fromthe GMs on hepatocyte functions and drug-mediated hepatotoxicity.

3.1. Incorporating TGF-b1-loaded GMs into the 3D-mFCCS

The 3D-mFCCS consists of a micropillar array which divides themicrofluidic channel into a centre compartment for cell culture and

exit

valve

syringe pump

S

outlet

hepatocytes

cell reservoir

ic representation of the seeding process of primary hepatocytes mixed with GMs; (B)FCCS. Images were taken at three different sites, near the entrance of the cell culturecompartment (scale bar: 50 mm).

Fig. 2. Characterization of the fabricated GMs. (A) Size distribution of the GMs seeded in the 3D-mFCCS as compared to those fabricated samples. Inset: SEM image showing therepresentative size of the GMs (scale bar: 10 mm). (B) Measurement of the controlled release profile of TGF-b1 from the GMs in the 3D-mFCCS under dynamic flow for 7 days. Data arethe mean� s.e.m. of 3 independent experiments.

C. Zhang et al. / Biomaterials 30 (2009) 3847–38533850

two side channels for perfusion of culture medium (Fig. 1A). Thecentre compartment is connected to a cell reservoir from whichcells can be introduced into the 3D-mFCCS. The micropillar arrayallows the TGF-b1-loaded GMs and hepatocytes to be physicallyimmobilized within the centre compartment.

During cell seeding, we introduced the GM/hepatocyte mixtureinto the cell reservoir and applied a withdrawal flow rate of 0.02–0.03 ml/h at the outlet (Fig. 1A). To effectively immobilize hepato-cytes and GMs in the 3D-mFCCS without causing clogging,a suspension of 1.5 million hepatocytes and 3000–9000 GMs in500 ml methylated collagen was found to be optimal, leading to theentrapment of w5000 hepatocytes and w10–30 GMs, as deter-mined by imaging and PicoGreen assay. We investigated the GMdistribution in the 3D-mFCCS by fluorescently labeling hepatocyteswith Calcein AM (grey) and GMs with Rhodamine–Dextran(yellow). Images were taken at different locations along the lengthof the 3D-mFCCS (Fig. 1B) which showed qualitatively uniform GMdistributions.

We quantified the size distributions and the controlled releaseproperty of the GMs in 3D-mFCCS. The size distribution was char-acterized by analyzing the images acquired using phase-contrast

microscopy (Figs. 1B and 2A). The inset of Fig. 2A is an SEM image ofthe typical GMs with smooth surfaces and size distribution. Thesize distribution of the GMs (21�7 mm in diameter) is comparableto that of the fabricated GM samples (14� 6 mm in diameter). Thecontrolled release property of the GMs in the 3D-mFCCS was eval-uated under dynamic flow of 0.05 ml/h. The 3D-mFCCS was fullypacked with TGF-b1-loaded GMs (w10 000 GMs/channel) and thetotal amount of the released TGF-b1 showed a linear increase from47� 7 pg to 474� 82 pg over 7 days, with an average daily releaseof 71�14 pg (Fig. 2B). Thus, w10 000 GMs released a total amountof w71 pg TGF-b1 everyday and according to the formula in Section2.7, the number of GMs to be immobilized in the 3D-mFCCS wascalculated to be 10, 20 and 30 to release TGF-b1 at the concentra-tions of 100, 200 and 300 pg/ml, respectively.

3.2. Determining the optimal TGF-b1 concentration for primaryhepatocytes in 3D-mFCCS

We compared the effect of directly perfused TGF-b1 to in situcontrolled released TGF-b1 on the levels of albumin secretion bythe hepatocytes in the 3D-mFCCS. Three concentrations, 100 pg/ml,

Fig. 3. Determination of the optimal concentration of controlled presented TGF-b1 forprimary hepatocytes in the 3D-mFCCS by evaluation of albumin secretion for 7 days. (A)Optimization for directly perfused TGF-b1 ( 100 pg/ml; 200 pg/ml; 300 pg/ml).(B) Optimization for in situ controlled released TGF-b1 ( 100–110 pg/ml; - 200–220 pg/ml; 300–320 pg/ml). Data are the mean� s.e.m. of 2 independent experi-ments, **p< 0.05, *p< 0.1.

C. Zhang et al. / Biomaterials 30 (2009) 3847–3853 3851

200 pg/ml and 300 pg/ml, were directly perfused. When 200 pg/mlof TGF-b1 was directly perfused to the hepatocytes, albuminsecretion showed a significantly higher level (98� 15 mg on day 3,40� 9 mg on day 5 and 11�1 mg on day 7) as compared to thatstimulated by 100 pg/ml (8� 1 mg on day 3, 5� 3 mg on day 5, and5�1 mg on day 7) or 300 pg/ml of TGF-b1 (7�4 mg on day 3,5� 2 mg on day 5 and 4� 3 mg on day 7) (Fig. 3A).

To in situ controlled release TGF-b1 at the concentrations of 100,200 or 300 pg/ml, the number of GMs immobilized in the 3D-mFCCSwas estimated to be 10, 20 or 30 (see Section 2.7). We observed thatby suspending a mixture of 1.5 million hepatocytes with 3000,6000 or 9000 GMs in 500 ml methylated collagen will give rise tothe entrapment of w5000 hepatocytes with 9� 4, 17� 3 or 32� 7GMs respectively; yielding a final concentration of 100–110 pg/ml,200–220 pg/ml or 300–320 pg/ml of TGF-b1 released. With in situcontrolled release, the concentration of 200–220 pg/ml was themost efficient in enhancing albumin secretion (69�19 mg on day 3,

51�2 mg on day 5 and 47� 9 mg on day 7) as compared to 100–110 pg/ml (28� 5 mg on day 3, 27� 7 mg on day 5 and 6� 5 mg onday 7) or 300–320 pg/ml (12� 4 mg on day 3, 2�1 mg on day 5 and1�0.3 mg on day 7) (Fig. 3B). Unlike direct perfusion of 200 pg/mlof TGF-b1 which led to rapid decrease in albumin secretion, in situcontrolled release of TGF-b1 at 200–220 pg/ml maintained albuminsecretion at w50 mg from day 5 to 7. These results led us to choose200–220 pg/ml of TGF-b1 as the concentration for optimal hepa-tocytes functions in the 3D-mFCCS and the in situ controlled releasefeature retarded the decline in albumin secretion.

3.3. Assessment of metabolic enzyme activities of hepatocytesin 3D-mFCCS

We compared the metabolic functions of the primary hepato-cytes cultured without or with TGF-b1, directly perfused or in situcontrolled released from GMs in the 3D-mFCCS, i.e. CYP450 enzy-matic (phase _ metabolic enzyme) activity and UGT enzymatic(phase II metabolic enzyme) activity. When TGF-b1 was controlledpresented to the hepatocytes, the CYP450 enzymatic activity wasstimulated (144� 27/106� 3 mg on day 3, 122� 23/105� 21 mg onday 5, 186� 36/94� 9 mg on day 7 for in situ controlled release/direct perfusion of TGF-b1) compared to the control without TGF-b1 (99�15 mg on day 3, 57�14 mg on day 5 and 72� 6 mg on day 7).Furthermore, in situ controlled release of TGF-b1 from GMs in 3D-mFCCS resulted in significantly higher CYP450 enzymatic activitythan the directly perfused TGF-b1 (144� 27 vs. 106� 3 mg on day 3,122� 23 vs. 105� 21 mg on day 5 and 186� 36 vs. 94� 9 mg onday 7) (Fig. 4A).

We observed similar trend in the UGT enzymatic activityassessment. The UGT enzymatic activity of hepatocytes controlledpresented with TGF-b1 was significantly higher (3558� 585/1904� 202 mg on day 3, 3253�1244/1008� 64 mg on day 5,1265� 675/535� 288 mg on day 7 for in situ controlled release/direct perfusion of TGF-b1) than the control without TGF-b1(1422� 463 mg on day 3, 1156� 643 mg on day 5, and 216� 204 mgon day 7). The enhancement of UGT enzymatic activities in hepa-tocytes presented with in situ controlled release of TGF-b1 was atleast two-fold higher than achieved by direct TGF-b1 perfusion(3558� 585 vs. 1904� 202 mg on day 3, 3253�1244 vs.1008� 64 mg on day 5 and 1265� 675 vs. 535� 288 mg on day 7)(Fig. 4B). On day 7, the overall UGT enzymatic activity of thehepatocytes declined even though the phase I metabolic activitywas maintained.

3.4. Improved drug sensitivity of hepatocytes with controlledTGF-b1 presentation in 3D-mFCCS

We compared the sensitivities of the hepatocytes to APAP in the3D-mFCCS, cultured without or with TGF-b1, directly perfused or insitu controlled released from GMs in the 3D-mFCCS. After exposureto 20 mM and 40 mM of APAP for 24 h, the hepatocytes wereassessed for viability using fluorescent labeling. The percentage ofviable cells was quantified using image processing (Fig. 5). At 20 mM

APAP, when TGF-b1 was controlled presented to the hepatocytes,the drug-mediated sensitivity was higher (41�4% viability for insitu controlled release and 57�17% viability for direct perfusion ofTGF-b1) than the control without TGF-b1 (70�14% viability).Furthermore, in situ controlled release of TGF-b1 from GMs in 3D-mFCCS resulted in significantly higher sensitivity than the directlyperfused TGF-b1 (41�4% viability vs. 57�17% viability).

The same trend was observed when APAP concentration wasincreased to 40 mM. The drug-mediated sensitivity of hepatocytescontrolled presented with TGF-b1 was higher (24�7% viability and38� 8% viability for in situ controlled release and direct perfusion

Fig. 5. Assessment of the drug-mediated hepatotoxicity of primary hepatocytes (,normal culture medium for hepatocyte culture without TGF-b1; 200 pg/ml TGF-b1via directly perfused; - 200–220 pg/ml TGF-b1 via in situ controlled release). Data arethe mean� s.e.m. of 2 independent experiments, **p< 0.05, *p< 0.1.

Fig. 4. Assessment of metabolic activity of the primary hepatocytes for 7 days (,normal culture medium for hepatocyte culture without TGF-b1; 200 pg/ml TGF-b1via directly perfused; - 200–220 pg/ml TGF-b1 via in situ controlled release). (A) CHCproduction, as a measure of phase 1 CYP450 metabolic enzyme activity. (B) 4-MUGproduction, as a measure of phase 2 UGT enzyme activity. Data are the mean� s.e.m. of2 independent experiments, **p< 0.05, *p< 0.1.

C. Zhang et al. / Biomaterials 30 (2009) 3847–38533852

of TGF-b1) than the control without TGF-b1 (43� 8% viability).In situ controlled release of TGF-b1 from GMs in 3D-mFCCS resultedin significantly higher sensitivity than the directly perfused TGF-b1(24�7% viability vs. 38� 8% viability) (Fig. 5).

4. Discussion

Liver-tissue engineering applications require hepatocytescultured in vitro to maintain high levels of metabolic functions [8,9].Co-cultures with fibroblasts can efficiently stimulate and maintainhepatocyte functions [23,24] (e.g. 5 times higher of albuminsecretion level and 6 times higher of detoxifying activity thanmono-culture). However, co-culture support of hepatocyte func-tions occurs only when the fibroblasts are proliferating [25].Hepatocytes in 3D-mFCCS exhibited enhanced functions; unfortu-nately, due to limited space inside a microfluidic device and loss ofa fixed ratio of cells, it is challenging to co-culture non-proliferating

hepatocytes with proliferative fibroblasts. We resolved thisdilemma by stimulating hepatocyte functions in the 3D-mFCCSthrough controlled presentation of TGF-b1, either by direct perfu-sion or by in situ controlled release from GMs.

The 3D-mFCCS was developed to recapitulate the in vivo envi-ronmental signals to support liver-specific functions in vitro [10].Based on the 3D-mFCCS, the soluble factor TGF-b1, whichwas previously reported to be primarily responsible for the highlevels of hepatocyte functions observed in hepatocyte–fibroblastco-cultures [17], was controlled presented to the hepatocytes. Inthe presence of 200–220 pg/ml TGF-b1, albumin secretion andphase I/II metabolic enzymatic activities of the hepatocytes in the3D-mFCCS were significantly enhanced compared to the controlwithout TGF-b1. This provides the proof-of-concept support to thehypothesis that controlled presentation of TGF-b1 can enhancehepatocyte functions in a microfluidic platform, even surpassingthe level achieved with hepatocyte–fibroblast co-cultures [26,27].

Compared to the direct perfusion of TGF-b1, the in situcontrolled release of TGF-b1 by means of GMs facilitated furtherfunctional enhancement of 4.5-, 2- and 3-fold more in albuminsecretion, phase I and phase II activities respectively. Preservationof the TGF-b1 bioactivity in the GMs might be responsible for thefunctional enhancement. Moreover, hepatocytes might be in closeor direct contact with the GMs in the 3D-mFCCS, together with themicro- or nanoscale-featured contacting surfaces that increased thecells’ capability of perception [28], leading to more active TGF-b1uptake. Here we chose gelatin as the controlled release carrierbecause it has been effectively used to controlled release growthfactors such as bFGF and VEGF in various tissue engineeringapplications [29,30]. With the enhanced enzymatic activities inTGF-b1, hepatocytes in 3D-mFCCS exhibited higher sensitivity toAPAP induced cytotoxicity compared to the control without TGF-b1.20 mM of APAP induced w50% of cell death in 200–220 pg/ml TGF-b1 in the 3D-mFCCS compared to the control without TGF-b1, whichrequired 40 mM of APAP (Fig. 5). TGF-b1 itself did not directly causehepatotoxicity as the hepatocytes exhibited no loss of cell functionsin TGF-b1 without APAP (Fig. 4).

C. Zhang et al. / Biomaterials 30 (2009) 3847–3853 3853

There are many other hepatocyte-based applications such as thedisease models for therapy (e.g. proteolytic enzyme therapy forliver cancer treatment [31]), pathogen testing (e.g. pathogen iden-tification for hepatitis B [32]) and bioreactors for the developmentof the bio-artificial liver device (BLAD). Hepatocyte co-cultureshave been proposed [33] and developed [34] as an attractive way toenhance hepatocyte functions and system performances. However,the optimized ratio of hepatocytes to the supporting co-culturedcells is critical for efficient cell–cell interactions and thus thefunctional maintenance of hepatocytes [23]. Such operationalcomplexity of co-culturing hepatocytes and supporting cells ata precise ratio can be avoided by replacing the supporting cells withthe in situ presentation of TGF-b1.

5. Conclusion

We have developed a strategy to replace hepatocyte–fibroblastco-cultures for maintaining high levels of hepatocyte functions in3D-mFCCS. Primary hepatocytes in 3D-mFCCS were encapsulated bya layer of in situ formed matrix to establish cell–matrix interaction.TGF-b1 was controlled presented to the immobilized hepatocytesin 3D-mFCCS either by direct perfusion, or by in situ controlledrelease. At 200–220 pg/ml of TGF-b1, the hepatocyte functions(albumin secretion, phase I and phase II metabolic enzyme activi-ties) were significantly higher in the in situ controlled-releasedTGF-b1 than the directly perfused TGF-b1 and were much higherthan the control without TGF-b1 or even reported co-cultures. Themeasured hepatocyte sensitivity to APAP induced cytotoxicity inour 3D-mFCCS with in situ presented TGF-b1 approaches that of thein vivo drug-induced hepatotoxicity.

Acknowledgement

We thank the members of the Cell and Tissue Engineering Labo-ratory for stimulating scientific discussions. This work is supported inpart by the Institute of Bioengineering and Nanotechnology, BMRC,A*STAR of Singapore and grants from BMRC (R-185-001-045-305),Ministry of Education (R-185-000-135-112), National MedicalResearch Council (R-185-000-099-213) and Singapore-MIT AllianceComputational and Systems Biology Flagship Project funding (C-382-603-004-001) to HYU. CZ and SFZ are National University ofSingapore graduate research scholars, SMC is a Singapore-MITAlliance post-doctoral fellow and SMO is an A*STAR graduatescholar.

Appendix

Figures with essential colour discrimination. Certain figures inthis article, especially Fig. 1 is difficult to interpret in black andwhite. The full colour images can be found in the online version, atdoi:10.1016/j.biomaterials.2009.03.052.

References

[1] Guilluozo AMF, Langouet S, Maheo K, Rissel M. Use of hepatocyte cultures forthe study of hepatotoxic compounds. J Hepatol 1997;26(Suppl. 2):73–80.

[2] Kim YS, Kim JS, Cho HS, Rha DS, Kim JM, Park JD, et al. Twenty-eight-day oraltoxicity, genotoxicity, and gender-related tissue distribution of silver nano-particles in Sprague–Dawley rats. Inhal Toxicol 2008;20(6):575–83.

[3] Boess F, Kamber M, Romer S, Gasser R, Muller D, Albertini S, et al. Geneexpression in two hepatic cell lines, cultured primary hepatocytes, and liverslices compared to the in vivo liver gene expression in rats: possible implicationsfor toxicogenomics use of in vitro systems. Toxicol Sci 2003;73(2):386–402.

[4] Kemp DC, Brouwer KL. Viability assessment in sandwich-cultured rat hepa-tocytes after xenobiotic exposure. Toxicol In Vitro 2004;18(6):869–77.

[5] Davis MW, Vacanti JP. Toward development of an implantable tissue engi-neered liver. Biomaterials 1996;17(3):365–72.

[6] Allen JW, Bhatia SN. Engineering liver therapies for the future. Tissue Eng2002;8(5):725–37.

[7] Streetz KL. Bio-artificial liver devices – tentative, but promising progress.J Hepatol 2008;48(2):189–91.

[8] Gerlach JC. Use of hepatocyte cultures for liver support bioreactors. Adv ExpMed Biol 1994;368:165–71.

[9] Gomez-Lechon MJ, Donato MT, Castell JV, Jover R. Human hepatocytes as a tool forstudying toxicity and drug metabolism. Curr Drug Metab 2003;4(4):292–312.

[10] Toh YC, Zhang C, Zhang J, Khong YM, Chang S, Samper VD, et al. A novel 3Dmammalian cell perfusion-culture system in microfluidic channels. Lab Chip2007;7(3):302–9.

[11] Vekemans K, Braet F. Structural and functional aspects of the liver and liversinusoidal cells in relation to colon carcinoma metastasis. World J Gastro-enterol 2005;11(33):5095–102.

[12] Dittrich PS, Manz A. Lab-on-a-chip: microfluidics in drug discovery. Nat RevDrug Discov 2006;5(3):210–8.

[13] Washizu J, Berthiaume F, Mokuno Y, Tompkins RG, Toner M, Yarmush ML.Long-term maintenance of cytochrome P450 activities by rat hepatocyte/3T3cell co-cultures in heparinized human plasma. Tissue Eng 2001;7(6):691–703.

[14] Seo SJ, Kim IY, Choi YJ, Akaike T, Cho CS. Enhanced liver functions of hepa-tocytes cocultured with NIH 3T3 in the alginate/galactosylated chitosanscaffold. Biomaterials 2006;27(8):1487–95.

[15] Bhatia SN, Yarmush ML, Toner M. Controlling cell interactions by micro-patterning in co-cultures: hepatocytes and 3T3 fibroblasts. J Biomed Mater Res1997;34(2):189–99.

[16] Nahmias Y, Berthiaume F, Yarmush ML. Integration of technologies for hepatictissue engineering. Adv Biochem Eng Biotechnol 2007;103:309–29.

[17] Chia SM, Lin PC, Yu H. TGF-beta1 regulation in hepatocyte-NIH3T3 co-cultureis important for the enhanced hepatocyte function in 3D microenvironment.Biotechnol Bioeng 2005;89(5):565–73.

[18] Wakefield LM, Winokur TS, Hollands RS, Christopherson K, Levinson AD,Sporn MB. Recombinant latent transforming growth factor beta 1 has a longerplasma half-life in rats than active transforming growth factor beta 1, anda different tissue distribution. J Clin Invest 1990;86(6):1976–84.

[19] Catelas I, Dwyer JF, Helgerson S. Controlled release of bioactive transforminggrowth factor beta-1 from fibrin gels in vitro. Tissue Eng Part C Methods2008;14(2):119–28.

[20] Ong SM, Zhang C, Toh YC, Kim SH, Foo HL, Tan CH, et al. A gel-free 3Dmicrofluidic cell culture system. Biomaterials 2008;29(22):3237–44.

[21] Chia SM, Leong KW, Li J, Xu X, Zeng K, Er PN, et al. Hepatocyte encapsulationfor enhanced cellular functions. Tissue Eng 2000;6(5):481–95.

[22] Seglen PO. Preparation of isolated rat liver cells. Methods Cell Biol1976;13:29–83.

[23] Bhatia SN, Balis UJ, Yarmush ML, Toner M. Effect of cell–cell interactions inpreservation of cellular phenotype: cocultivation of hepatocytes and non-parenchymal cells. FASEB J 1999;13(14):1883–900.

[24] Kuri-Harcuch W, Mendoza-Figueroa T. Cultivation of adult rat hepatocytes on3T3 cells: expression of various liver differentiated functions. Differentiation1989;41(2):148–57.

[25] Bhatia SN, Balis UJ, Yarmush ML, Toner M. Probing heterotypic cell interac-tions: hepatocyte function in microfabricated co-cultures. J Biomater SciPolym Ed 1998;9(11):1137–60.

[26] Nishikawa M, Kojima N, Komori K, Yamamoto T, Fujii T, Sakai Y. Enhancedmaintenance and functions of rat hepatocytes induced by combination ofon-site oxygenation and coculture with fibroblasts. J Biotechnol 2008;133(2):253–60.

[27] Lu HF, Chua KN, Zhang PC, Lim WS, Ramakrishna S, Leong KW, et al. Three-dimensional co-culture of rat hepatocyte spheroids and NIH/3T3 fibroblastsenhances hepatocyte functional maintenance. Acta Biomater 2005;1(4):399–410.

[28] Stevens MM, George JH. Exploring and engineering the cell surface interface.Science 2005;310(5751):1135–8.

[29] Takehara N, Tsutsumi Y, Tateishi K, Ogata T, Tanaka H, Ueyama T, et al.Controlled delivery of basic fibroblast growth factor promotes human cardio-sphere-derived cell engraftment to enhance cardiac repair for chronicmyocardial infarction. J Am Coll Cardiol 2008;52(23):1858–65.

[30] Patel ZS, Ueda H, Yamamoto M, Tabata Y, Mikos AG. In vitro and in vivo releaseof vascular endothelial growth factor from gelatin microparticles and biode-gradable composite scaffolds. Pharm Res 2008;25(10):2370–8.

[31] Beuth J. Proteolytic enzyme therapy in evidence-based complementaryoncology: fact or fiction? Integr Cancer Ther 2008;7(4):311–6.

[32] Sodunke TR, Bouchard MJ, Noh HM. Microfluidic platform for hepatitis B viralreplication study. Biomed Microdevices 2008;10(3):393–402.

[33] Sauer IM, Obermeyer N, Kardassis D, Theruvath T, Gerlach JC. Development ofa hybrid liver support system. Ann N Y Acad Sci 2001;944:308–19.

[34] Koike M, Matsushita M, Taguchi K, Uchino J. Function of culturing monolayerhepatocytes by collagen gel coating and coculture with nonparenchymal cells.Artif Organs 1996;20(2):186–92.