A novel 3D mammalian cell perfusion-culture system in microfluidicchannels{

Yi-Chin Toh,ab Chi Zhang,ab Jing Zhang,a Yuet Mei Khong,ab Shi Chang,c Victor D. Samper,{a

Danny van Noort,a Dietmar W. Hutmacherbgh and Hanry Yu*abcdef

Received 16th October 2006, Accepted 21st December 2006

First published as an Advance Article on the web 23rd January 2007

DOI: 10.1039/b614872g

Mammalian cells cultured on 2D surfaces in microfluidic channels are increasingly used in drug

development and biological research applications. These systems would have more biological or

clinical relevance if the cells exhibit 3D phenotypes similar to the cells in vivo. We have developed

a microfluidic channel based system that allows cells to be perfusion-cultured in 3D by supporting

them with adequate 3D cell–cell and cell–matrix interactions. The maximal cell–cell interaction

was achieved by perfusion-seeding cells through an array of micropillars; and 3D cell–matrix

interactions were achieved by a polyelectrolyte complex coacervation process to form a thin layer

of matrix conforming to the 3D cell shapes. Carcinoma cell lines (HepG2, MCF7), primary

differentiated (hepatocytes) and primary progenitor cells (bone marrow mesenchymal stem cells)

were perfusion-cultured for 72 hours to 1 week in the microfluidic channel, which preserved their

3D cyto-architecture and cell-specific functions or differentiation competence. This transparent

3D microfluidic channel-based cell culture system also allows direct optical monitoring of cellular

events for a wide range of applications.

Introduction

Microfluidic platforms for cell culture are rapidly gaining

importance in drug development and biological research

applications, such as drug toxicity or metabolism studies1,2

and stem cell differentiation studies.3 Microfluidic channels are

especially attractive since they can be easily multiplexed with

integrated fluid handling operations for efficient and high

throughput cellular analysis,4,5 imaged for in situ monitoring

of cellular events,6,7 and can recapitulate a physiological

cellular microenvironment with controllable distribution of

biochemical molecules and shear stresses at the cellular

resolution.8,9 A plethora of microfluidic channel-based systems

for cell culture have been developed with cells growing on rigid

2D substrates such as glass or plastic surfaces.6–8,10 A major

limitation of these 2D systems is that the in vivo cellular

microenvironment, where cells interact with ECM, neighbour-

ing cells, soluble factors and mechanical forces in 3D,11 is not

fully recapitulated. It is reported that cells mirror their in vivo

counterparts more closely when cultured in 3D microenviron-

ments12–14 through properly regulated cell–cell and cell–matrix

interactions.15,16 Therefore, there is a need to develop a

microfluidic channel-based mammalian cell culture platform

that allows cells to interact with their microenvironment in 3D.

There has been attempts to culture cells three-dimensionally in

microfluidic systems by confining them in microfabricated

chambers;17 however, precise presentation of 3D cell–matrix

interactions are lacking in such systems to maintain the 3D

cellular phenotypes.

We have developed a novel 3D microfluidic channel-based

cell culture system (hereafter referred to as 3D-mFCCS) with

precision control of 3D cell–cell and cell–matrix interactions.

The maximal cell–cell interaction was achieved by perfusion-

seeding cells through an array of micropillars; and 3D cell–

matrix interactions achieved by a polyelectrolyte complex

coacervation process to form a thin layer of matrix conforming

to the 3D cell shapes.18 We demonstrated the biological

versatility of this method by culturing various immortalized

(HepG2 and MCF7) and primary (hepatocytes and bone

marrow mesenchymal stem cells (BMSCs)) mammalian cells in

the 3D-mFCCS. Cells were viable and exhibited distinct 3D

cyto-architecture throughout a perfusion culture period of

72 hours. Assessment of the cell specific functions and

differentiation competence of hepatocytes and BMSCs,

respectively, further indicated the usefulness of the 3D-

mFCCS as a microfluidic-device for culturing fastidious

primary mammalian cells in 3D.

aInstitute of Bioengineering and Nanotechnology, 31 Biopolis Way, TheNanos, 138669, Singapore. E-mail: hyu@ibn.a-star.edu.sg;nmiyuh@nus.edu.sg; Fax: +65-6516 8261; Tel: +65-6516 3466bNUS Graduate Programme in Bioengineering, NUS Graduate Schoolfor Integrative Sciences and Engineering, National University ofSingapore, Singapore 117456cDepartment of Physiology, Yong Loo Lin School of Medicine, NationalUniversity of Singapore, Singapore 117597dSingapore-MIT Alliance, E4-04-10, 4 Engineering Drive 3, Singapore117576eNUS Tissue-Engineering Programme, DSO Labs, National Universityof Singapore, Singapore 117597fDepartment of Haematology-Oncology, National University Hospital,Singapore 119074gDivision of Bioengineering, Faculty of Engineering, National Universityof Singapore, Singapore 117576hDepartment of Orthopaedic Surgery, Yong Loo Lin School of Medicine,National University of Singapore, Singapore 119074{ Electronic supplementary information (ESI) available:Supplementary Fig. 1 and 2 and supplementary video. See DOI:10.1039/b614872g{ Present address: GE Global Research, Freisinger Landstr. 50, 85748Garching b.Muenchen, Germany

PAPER www.rsc.org/loc | Lab on a Chip

302 | Lab Chip, 2007, 7, 302–309 This journal is � The Royal Society of Chemistry 2007

Dow

nloa

ded

by E

wha

Wom

ens

Uni

vers

ity o

n 30

Jul

y 20

12Pu

blis

hed

on 2

3 Ja

nuar

y 20

07 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/B61

4872

GView Online / Journal Homepage / Table of Contents for this issue

Experimental

Materials

All chemicals and reagents were purchased from Sigma–

Aldrich Pte Ltd, Singapore unless otherwise stated.

Device fabrication and operation

Microfluidic channels with micropillar array were designed

using AutoCAD (Autodesk, USA) and L-Edit v10.20 (Tanner

Research, USA). The dimensions of the microfluidic channel

were 1 cm (length) 6 600 mm (width) 6 100 mm (height) and

had 3 inlets and 3 outlets. An array of 30 mm 6 50 mm

elliptical micropillars with a 20 mm gap size was situated in the

middle of the microfluidic channel, bounding a cell residence

volume that was 200 mm wide. Silicon templates were

fabricated by standard deep reactive ion etching (DRIE)

process19 (Oxford Instruments Plc, UK). The microfluidic

channels were then obtained by replica molding polydimethyl-

siloxane (PDMS) (Dow Corning, USA) on the silicon

templates. The PDMS structures were oxidized in oxygen

plasma for 1 minute (125 W, 13.5 MHz, 50 sccm, and

40 millitorr) for irreversible chemical bonding to glass cover-

slips before connecting to fluidic components (Upchurch,

USA). The center inlet of the microfluidic channel was

connected to a cell reservoir, which comprised of a 4-way

valve with 3 Luer connections (Cole-Palmer, USA) coupled to

a 25G stainless steel hypodermic tubing (Becton-Dickinson,

USA), to permit the independent introduction of cell suspen-

sion and polyelectrolytes via the center inlet. The entire set-up

was sterilized by autoclaving at 105 uC for 30 minutes.

Cell immobilization was initiated by withdrawing the cell

suspension from the cell reservoir, via the 2 side outlets, using a

withdrawal syringe pump (Cole-Parmer, USA) with the center

outlet kept closed. Laminar flow complex coacervation was

implemented by simultaneously infusing a pair of polyelec-

trolytes via the center and 2 side outlets with syringe pumps.18

Upon the formation of the 3D complex coacervated matrix,

excess polyelectrolytes were displaced by culture medium

infused from the side inlets. All 3 outlets were opened during

perfusion culture of cells.

Polyelectrolytes preparation

Cationic methylated collagen and anionic terpolymer of

hydroxylethylmethacrylate–methylmethacrylate–methylacrylic

acid (HEMA-MMA-MAA) were synthesized and purified as

described previously.20 3% terpolymer solution and 1.5 mg ml21

or 3.0 mg ml21 methylated collagen were used in this study.

Fluorescence labeled methylated collagen was prepared by

labeling with Alexa Fluor 532 (Molecular Probes, USA)

according to manufacturer’s protocols.

Cells isolation and maintenance

Hepatocytes were harvested from male Wistar rats weighing

from 250 to 300 g by a two-step in situ collagenase perfusion.21

Hepatocytes used in all experiments had a cell viability of

.90%, as determined by Trypan Blue exclusion assay, and a

yield of 200–300 million cells. Rat BMSCs were isolated from

bone marrow aspirates from male Wistar rats and purified as

described previously.22 They were maintained in Dulbecco’s

modified Eagle medium (DMEM), low glucose (Invitrogen,

Singapore) with 10% fetal calf serum (FCS) and 100 mg ml21

penicillin/streptomycin at 37 uC, 5% CO2. Passage 3–8 cells

were used in all experiments. HepG2 and MCF7 were cultured

in DMEM, high glucose (Invitrogen, Singapore) with 10% FCS

and 100 mg ml21 penicillin/streptomycin at 37 uC, 5% CO2.

Characterization of 3D-mFCCS

SEM and confocal microscopy were used to characterize the

three-dimensionally immobilized cells and fluorescence-labeled

complex coacervated matrix in the 3D-mFCCS. Optimization

of cell viability during 3D immobilization process was assessed

by confocal microscopy (Olympus, Japan) after labeling live

and necrotic cell populations with Cell Tracker Green (CTG)

(Molecular Probes, USA) and propidium iodide (PI) respec-

tively. Cell viability was quantified by counting the number of

live and dead cells with image processing software (Image-

Pro1 Plus 4.5.1, Media Cybernatics Inc., USA), and the

percentage cell viability was normalized against the number of

freshly isolated viable cells.

Scanning electron microscopy (SEM)

SEM samples were prepared by bonding the PDMS micro-

fluidic channels onto a polyethylene (PE) film (Diversified

Biotech, USA) instead of a glass coverslip. The samples were

fixed with 3.7% paraformaldehyde (PFA) before the PE film

was peeled off to expose the microfluidic channel for SEM

processing. Samples were incubated with 1% osmium tetra-

oxide for 1 hour, sequentially dehydrated with ethanol series

(25, 50, 75, 95 and 100%) and liquid carbon dioxide, and then

gold-sputtered (30 mA, 70 s) before viewing with a field-

emission scanning electron microscope (JEOL, Japan).

Perfusion culture in 3D-mFCCS

A multi-channel peristaltic pump (Ismatec, Switzerland) was

used to circulate culture medium in a closed-loop perfusion

system. DMEM with 10% fetal calf serum (FCS) was used to

culture HepG2, MCF7 and BMSCs. HepG2 and MCF 7 were

cultured under high glucose (4.5 g L21) conditions while

BMSCs were cultured under low glucose (1.0 g L21)

conditions. Osteogenic medium was prepared by supplement-

ing basal medium with 100 nM dexamethasone, 50 mM

ascorbic acid 2-phosphate and 10 mM b-glycerophosphate

(Merck, Singapore). Primary hepatocytes were maintained in

HepatoZYME SFM (Invitrogen, Singapore), supplemented

with 100 mM dexamethasone. All perfusion culture media were

supplemented with 100 mg ml21 penicillin/streptomycin and

60 mM HEPES (Invitrogen, Singapore). The microfluidic

system was placed onto a heating plate (MEDAX GmbH &

Co. KG, Germany) maintained at 37 uC throughout the

culture period with perfusion flow rates of 0.06–0.2 ml h21.

Cell viability staining

In situ labeling of viable and necrotic cells for cell viability

quantification immediately after cell immobilization was

This journal is � The Royal Society of Chemistry 2007 Lab Chip, 2007, 7, 302–309 | 303

Dow

nloa

ded

by E

wha

Wom

ens

Uni

vers

ity o

n 30

Jul

y 20

12Pu

blis

hed

on 2

3 Ja

nuar

y 20

07 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/B61

4872

G

View Online

performed by infusing 20 mM of CTG (30 minutes), followed

by fresh culture medium (30 minutes) and finally 50 mg ml21 of

PI (15 min) at 0.8 ml h21. The cells were then fixed with 3.7%

paraformaldehyde (PFA) for 30 minutes, and imaged by

confocal microscopy (Olympus, Japan). Cell viability of

HepG2, MCF7, hepatocytes and BMSCs after 72 hours of

perfusion culture in the 3D-mFCCS was qualitatively assessed

by perfusing 5 mM of Calcein AM (Molecular Probes, USA)

and 25 mg ml21 of PI at 0.5 ml h21 for 30 minutes and viewing

immediately by confocal microscopy.

F-actin staining

F-actin distribution in all cell types was assessed after 72 hours

of perfusion culture in the 3D-mFCCS. In situ F-actin staining

was performed after fixation with 3.7% PFA (30 minutes) by

infusing the microfluidic channel with 0.5% Triton-X 100

(USB Corp, USA) (30 minutes), 0.2% bovine serum albumin

(BSA) (30 minutes), 0.2 mg ml21 of TRITC-phalloidin

(20 minutes) and 1X PBS (15 minutes) at 0.5 ml h21. 2D

monolayer cultures were fixed with 3.7% PFA (15 minutes)

and stained by incubating with 0.5% Triton-X 100

(10 minutes), 0.2% BSA (15 minutes), and 0.2 mg ml21 of

TRITC-phalloidin (20 minutes).

Immunostaining of E-cadherin

Immunostaining of E-cadherin was used to ascertain cell–cell

interactions in MCF7 after 72 hours of perfusion culture in the

3D-mFCCS. In situ immunostaining was performed by over-

night blocking with 10% FCS at 4 uC before injecting primary

antibody (mouse anti-E-cadherin, BD Bioscience, USA) into

the microfluidic channel. The sample was incubated overnight

at 4 uC, washed with 10% FCS before injecting and incubating

with secondary antibody (Alexa Fluor-488 goat-anti-mouse,

Molecular Probes, USA) for 1 hour at room temperature. The

sample was perfused with 10% FCS to remove excess antibody

and viewed under confocal microscope.

Functional assessment of hepatocytes

All functional data were normalized to DNA content in

samples quantified using PicoGreen assay (Molecular Probes,

USA). Albumin production was quantified with a rat albumin

ELISA quantification kit (Bethyl Laboratories Inc, USA) in

circulated culture medium every 24 hours. UGT activity of

hepatocytes in the 3D-mFCCS was determined by infusing

100 mM of 4-umbelliferone (4-MU) for 4 hours. The perfusate

was collected and 4-umbelliferyl glucuronide was analyzed

using capillary electrophoresis with laser induced fluorescence

(CE-LIF) detection (Prince Technologies B.V., Netherlands) at

an excitation wavelength of 325 nm. UGT activity in 2D

monolayer cultures was determined by incubating the samples

with 100 mM of 4-MU for 4 hours before CE-LIF analysis.

Von Kossa staining

BMSCs after 1 week of osteogenic induction were stained for

calcium salt deposits by von Kossa staining. In situ von Kossa

staining was carried out after fixation with 3.7% PFA

(30 minutes) by infusing the microfluidic channel in the

following sequence at 0.5 ml h21: 5% silver nitrate solution

(45 minutes), distilled water (DIW) (15 minutes), 5% sodium

bicarbonate in 3.7% formaldehyde solution (8 minutes), DIW

(15 minutes), 5% sodium thiosulfate (5 minutes) and DIW

(15 minutes). Von Kossa staining in 2D monolayer cultures

was performed as described previously.22

Results and discussion

3D microfluidic channel-based cell culture system (3D-mFCCS)

The 3D-mFCCS consisted of a microfluidic channel with an

array of micropillars, which was developed to immobilize and

support cells with 3D cell–cell and cell–matrix interactions

(Fig. 1a). Microfabricated pillar arrays have been previously

developed for various applications.23–26 Here, we adapted the

micropillar array for the physical immobilization of one or

more cell types at high density to achieve maximal cell–cell

interactions. The array of 30 mm 6 50 mm elliptical

micropillars, spaced 20 mm apart, was situated in the middle

of a 1 cm (length) 6 600 mm (width) 6 100 mm (height)

microfluidic channel with 3 inlets and 3 outlets joining 3 intra-

channel compartments (Fig. 1b). Cells were dynamically

seeded into the center compartment of the microfluidic

channel bounded by the micro-pillar array. The efficiency of

cell immobilization by the micropillar array depended upon

the geometrical designs of the micropillars and the cell seeding

density. The geometrical designs of the micropillars influenced

the clogging propensity during the cell immobilization process

and an elliptical design was found to be effective in cell

immobilization (Supplementary Fig. 1, see ESI{). At a gap size

of 20 mm between the pillars, the optimal cell seeding density

ranged from 1.5–10 million cells ml21, and was inversely

proportional to the cell size.

3D cell–matrix interactions were achieved by laminar flow

complex coacervation of polyelectrolytes in a physiological

aqueous environment to form a thin layer of matrix conform-

ing to the 3D cell shapes in the microfluidic channel (Fig. 1a).

Methylated collagen and HEMA-MMA-MAA terpolymer

were chosen as model polyelectrolytes as they could support

sensitive anchorage dependent cells.20 Methylated collagen and

terpolymer were infused via the center and side inlets

respectively. The former was hydrodynamically focused to

the center compartment of the microfluidic channel where the

cells were immobilized (Fig. 1c). The density of the matrix was

modulated by varying the complex coacervation reaction

parameters such as the polyelectrolyte concentrations or their

relative flow rates in the channel (Supplementary Fig. 2, see

ESI{). Excess polyelectrolytes were subsequently removed by

flowing culture medium through the side inlets to replace

terpolymer. Since the micropillars in the 3D-mFCCS relieved

the dependence on matrix entrapment to immobilize cells,18,27

only a thin cell-conformal layer of the complex coacervated

matrix was formed to provide the necessary 3D cell–matrix

interactions without impeding mass transfer (Fig. 1d–e). We

chose complex coacervation to form the 3D matrix since its

physico-chemical properties can be easily configured by

changing the complex coacervation reaction conditions e.g.

polyelectrolyte properties, compositions and flow configura-

tions as shown here and elsewhere.18,28 This tunable 3D matrix

304 | Lab Chip, 2007, 7, 302–309 This journal is � The Royal Society of Chemistry 2007

Dow

nloa

ded

by E

wha

Wom

ens

Uni

vers

ity o

n 30

Jul

y 20

12Pu

blis

hed

on 2

3 Ja

nuar

y 20

07 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/B61

4872

G

View Online

allows us to modulate the extent of cell–matrix interactions as

well as mass transfer properties, which in turn influence cell

survival and functions.28–30 The laminar flow complex

coacervation scheme also facilitated perfusion culture in

the microfluidic channels by hydrodynamically confining the

3D matrix to the center compartment of the microfluidic

channel.

Seeding cells into the 3D-mFCCS subjects them to hydro-

dynamic and impact forces as they are flowed into the channel

and subsequently immobilized by the micropillars. This may

result in a loss of cell viability (supplementary video) and

functions31,32, which is not observed in conventional seeding

procedures used for other cell culture systems. Therefore, cell

viability was monitored at different cell seeding flow rates

using primary rat hepatocytes as a sensitive cell model

immediately after seeding into the 3D-mFCCS. We achieved

lower cell seeding flow rates when we withdrew cells from the

microfluidic channel outlets than otherwise possible if we

infuse cells through the inlet. Cell viabilities at withdrawal flow

rates of 0.02 and 0.05 ml h21 were comparable (83 ¡ 17% and

88 ¡ 5%), whereas a higher withdrawal flow rate of 0.1 ml h21

yielded a significantly lower cell viability of 62 ¡ 9% (Fig. 2).

Therefore, a withdrawal flow rate of 0.02–0.05 ml h21 was

deemed suitable for seeding cells three-dimensionally into the

3D-mFCCS without compromising cell viability.

Perfusion culture of mammalian cells in 3D-mFCCS

To investigate the biological versatility and potential of the

3D-mFCCS as a mammalian cell perfusion-culture platform,

we perfusion-cultured a few model anchorage-dependent

mammalian cells such as carcinoma cells lines (HepG2 and

MCF7) and primary differentiated or progenitor cells (hepa-

tocytes and BMSCs) using a closed-loop perfusion system

(Fig. 3).

Carcinoma cell lines. We tested a human hepatocarcinoma

cell line (HepG2) and a human breast cancer cell line (MCF7)

Fig. 1 Implementation and characterization of the 3D microfluidic

channel-based cell culture system (3D-mFCCS). (a) Cells were three-

dimensionally immobilized in a microfluidic channel by dynamic

seeding through a micropillar array (step 1). The immobilized cells are

then stabilized and supported by 3D matrices formed by laminar flow

complex coacervation reaction of polyelectrolytes (P1 and P2) (step 2).

Excess polyelectrolytes are displaced by culture medium during

perfusion culture (step 3). (b) Prototype of 3D-mFCCS with a cross-

sectional illustration (indicated by white line). (c) SEM micrograph of

hepatocytes three-dimensionally immobilized in a microfluidic channel

with an array of 30 mm 6 50 mm elliptical micropillars. Complex

coacervated methylated collagen/HEMA-MMA-MAA terpolymer was

confined to the center compartment of the microfluidic channel. (d)–(e)

Complex coacervated matrix can be configured to form a thin layer

conforming to the 3D cell shape. (d) Confocal image of complex

coacervated 3D matrix formed with Alexa-Fluor 532 labeled

methylated collagen. Hepatocytes nuclei were counter-stained with

SYTOX Green (Molecular Probes, USA). (e) SEM micrograph of thin

nanofibrous matrix conforming to 3D cell shape of hepatocytes.

Fig. 2 Optimization of cell viability at different flow rates immedi-

ately after dynamic cell immobilization process in 3D-mFCCS. (a)–(c)

Viability staining (5 mM CTG and 50 mg ml21 PI) of hepatocytes

seeded at 0.02 ml h21, 0.05 ml h21 and 0.1 ml h21 withdrawal flow

rate, respectively. Images are maximum projections of optical sections

spanning the 3D cell aggregates. (d) Quantification of cell viability at

different withdrawal rates normalized against the number of freshly

isolated viable hepatocytes. Data are represented as means ¡ SD of

3 points along the length of a microfluidic channel. * Indicates

statistical significance (Student’s t-test, p , 0.05).

This journal is � The Royal Society of Chemistry 2007 Lab Chip, 2007, 7, 302–309 | 305

Dow

nloa

ded

by E

wha

Wom

ens

Uni

vers

ity o

n 30

Jul

y 20

12Pu

blis

hed

on 2

3 Ja

nuar

y 20

07 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/B61

4872

G

View Online

in the 3D-mFCCS. Both HepG2 and MCF7 formed viable

multi-cellular aggregates after 72 hours of perfusion culture, as

observed by confocal microscopy (Fig. 4a–b; 4d–e). Cells in

these multi-cellular aggregates exhibited a rounded morphol-

ogy with distinct cortical F-actin localization indicating the

preservation of 3D in vivo-like cyto-architecture, in contrast to

the more diffuse cytoplasmic localization with some stress fibre

formation observed in 2D monolayer cultures (Fig. 4c, 4f).33

Closer examination of these 3D carcinoma aggregates using

SEM revealed gradual merging of cell–cell boundaries,

implying the establishment of cell–cell linkages (Fig. 4g). We

confirmed the cell–cell interactions in the 3D cell aggregates by

immunostaining for E-cadherin, which is the key cell adhesion

molecule found in adherens junctions (Fig. 4h).

Primary hepatocytes. Primary rat hepatocytes maintained in

the 3D-mFCCS over a period of 72 hours exhibited extensive

morphogenesis with observable cellular aggregation as early as

24 hours after seeding. The hepatocytes remodeled into large

3D aggregates after 72 hours (Fig. 5a) but remained viable

even in the center of the aggregates indicating that mass

transfer in the aggregates was not impeded (Fig. 5b). We have

confirmed the maintenance of the 3D in vivo-like cyto-

architecture of hepatocytes cultured in the 3D-mFCCS by

their cortical localization of F-actin, which was reminiscent of

that seen in hepatocytes spheroids (Fig. 5c).34 This is in

contrast to hepatocytes cultured as a 2D monolayer on

collagen gels, where stress fibers and diffusive cytoplasmic

actin staining patterns were observed (Fig. 5c insert).

Observations using SEM indicated that these large

hepatocyte aggregates had smooth surfaces with indistinct

cell–cell boundaries, resembling that of hepatocyte spheroids

(Fig. 5d). We further quantified the synthetic and metabolic

functions of hepatocytes by measuring albumin production

and UDP-glucuronyltransferase (UGT) activity based on

4-methylumbelliferyl glucuronide (4-MUG) formation.

Hepatocytes in the 3D-mFCCS could maintain UGT activity

at significantly higher level (Student’s t-test, p , 0.05)

than that in 2D monolayer cultures (Fig. 5e). Albumin

production by hepatocytes in the 3D-mFCCS ranged from

800–1200 pg ng-DNA21 h21 over the culture period, compar-

able to that in 2D hepatocytes culture (Fig. 5f).

Bone marrow mesenchymal stem cells (BMSCs). After

72 hours of perfusion culture in the 3D-mFCCS, BMSCs did

not retain a rounded morphology but remodeled into 3D cell

bundles that were stretched along the direction of fluid flow in

the channel (Fig. 6a–b). F-actin labeling in the BMSCs showed

bundles of stress fibers that were aligned along the fluid flow

direction, although these stress fibers were not as distinct as

those present in the 2D monolayer cultures (Fig. 6c).

Nevertheless, SEM confirmed the three-dimensionality of the

BMSCs construct in the 3D-mFCCS (Fig. 6d). We also

investigated the differentiation competence of the BMSCs by

differentiating them along the osteogenic lineage. Histological

staining of calcium salt deposits by von Kossa staining was

positive for BMSCs maintained in the 3D-mFCCS after 1 week

of osteogenic induction (Fig. 6e). The calcium salt deposits in

the 3D BMSCs model were more extensive as compared to 2D

monolayer BMSCs culture (Fig. 6f), suggesting that the 3D-

mFCCS may provide a more conducive microenvironment for

BMSCs differentiation than a 2D monolayer culture.

Fig. 3 Schematic representation of a closed-loop perfusion culture

system for the culture of anchorage dependent mammalian cells in the

3D-mFCCS.

Fig. 4 Perfusion culture of carcinoma cell lines in 3D-mFCCS. (a)–(b)

HepG2 and (d)–(e) MCF7 perfusion-cultured for 72 hours maintained

high cell viability as assessed by transmission microscopy and confocal

microscopy upon staining with vital dyes, Calcein AM and propidium

iodide. Confocal images were maximum projections of 60–70 mm

optical sections. (c) and (f) F-actin staining of HepG2 and MCF7

cultured for 72 hours in the 3D-mFCCS, respectively, in comparison to

F-actin distribution in 2D monolayer cultures (inserts). Images are

orthogonal projections of 40–50 mm optical sections. (g) SEM

micrograph of HepG2 after 72 hours of perfusion culture. (h)

Immunostaining of E-cadherin (indicated by white arrows) counter-

stained with F-actin in MCF7 after 72 hours of perfusion culture.

306 | Lab Chip, 2007, 7, 302–309 This journal is � The Royal Society of Chemistry 2007

Dow

nloa

ded

by E

wha

Wom

ens

Uni

vers

ity o

n 30

Jul

y 20

12Pu

blis

hed

on 2

3 Ja

nuar

y 20

07 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/B61

4872

G

View Online

In this study, we demonstrated the preservation of 3D cyto-

architecture as well as cell-specific functions or differentiation

competence in carcinoma cell lines, primary hepatocytes and

BMSCs, which are prime candidates for establishing in vitro

models in drug development and biological research applica-

tions. For example, 3D carcinoma models have been

extensively used to study and understand phenomena in

cancer biology.13,35 Drug responses of tumor cells differ

dramatically when tested in 2D and 3D in vitro models, with

the latter being more predictive of in vivo responses.36–38

Researchers have envisioned in vitro models that incorporate

3D cancer spheroids into microfluidic systems, which are

realizable with the 3D-mFCCS, to expedite efforts in cancer

research and treatment.39 Primary hepatocytes are widely used

to evaluate drug toxicity and metabolism, although they lose

their differentiated functions rapidly in 2D cultures.40,41

Hence, there is increasing interest to establish these toxicity

and metabolism assays using 3D hepatocyte models to

generate data that are more comparable to the in vivo

situation,2,42,43 since hepatocytes preserve their differentiated

functions better in a 3D microenvironment.20,44,45 However,

the successful commercial application of 3D hepatocyte in vitro

models in the drug testing process hinges on the development

of hepatocyte-supporting 3D microenvironments on platforms

that can be readily parallelized, as in the microfluidic channels.

We have demonstrated the maintenance of hepatocytes

functions for up to 72 hours in the 3D-mFCCS, which is

indicative of its potential as a 3D hepatocyte in vitro model for

drug testing application. Moreover, the 3D-mFCCS can be

imaged in multi-dimensions i.e. xyz-time, making it compatible

with real-time high content cell analysis methods developed for

drug testing.46,47 Embryonic and adult progenitor cells are

highly valued in regenerative medicine as they have the ability

to differentiate into multiple cell lineages.48–50 The ability to

control their differentiation process requires the precise

understanding of the interplay of various extra-cellular

environmental cues guiding this process.3,50 These cues may

include 3D topography,51 mechanical forces e.g. shear

stresses52 as well as cytokine distribution gradients,53,54 and

can concomitantly be represented in our 3D-mFCCS, making it

a valuable biological tool for dissecting this complex process.

Conclusions

We have developed a novel 3D-mFCCS that can perform 3D

perfusion-culture of various anchorage dependent mammalian

cells by providing them with adequate 3D cell–cell and

cell–matrix interactions. Presentation of 3D cell–cell and

Fig. 5 Perfusion culture of primary hepatocytes in 3D-mFCCS. (a)–(b) Primary rat hepatocytes perfusion-cultured for 72 hours aggregated into

viable large 3D aggregates as assessed by transmission microscopy and confocal microscopy upon staining with vital dyes, Calcein AM and

propidium iodide. Confocal image was a maximum projection of 84 mm optical section. (c) Orthogonal projections of a 44 mm optical section

showing F-actin staining of hepatocytes after 72 hours culture in the 3D-mFCCS in comparison to F-actin distribution in a 2D monolayer (insert).

(d) SEM micrograph of hepatocytes after 72 hours of perfusion culture. (e)–(f) Functional assessment of hepatocytes by UDP-

glucuronyltransferase (UGT) activity and albumin production respectively. Data are represented as means ¡ SD of 3 microfluidic channels

(3D-mFCCS) or 3 culture wells (2D monolayer). * indicates statistical significance compared to 2D monolayer culture (Student’s t-test, p , 0.05).

This journal is � The Royal Society of Chemistry 2007 Lab Chip, 2007, 7, 302–309 | 307

Dow

nloa

ded

by E

wha

Wom

ens

Uni

vers

ity o

n 30

Jul

y 20

12Pu

blis

hed

on 2

3 Ja

nuar

y 20

07 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/B61

4872

G

View Online

cell–matrix interactions were accomplished by using an array

of microfabricated pillars for cell immobilization and laminar

flow complex coacervation for 3D matrix formation respec-

tively. The development of the 3D-mFCCS has several key

advantages besides achieving miniaturization of cell number

and reagent volume. Microfabricated valves, mixers, concen-

tration gradient generators can be readily integrated into

multiplexed channels to generate an array of discrete micro-

environments with a different set of extra-cellular cues, e.g.

biochemical gradients, drug concentration and shear stresses,

where multiple cellular analysis may be performed simulta-

neously.5 Perfusion culture in the 3D-mFCCS allows for online

analytical analysis of the culture medium for drug metabolites

or enzymatic activities by coupling it to micro-analytical

systems such as micro-capillary electrophoresis or biosensors.5

The PDMS-based 3D-mFCCS is transparent and the use of

glass coverslips to define one of the surfaces of the channel

made the entire construct compatible with high resolution

imaging modalities such as confocal microscopy. Various

optical-based in situ assays using fluorescence reporters or

immunolabeling for probing cellular events or monitoring

cellular responses to external perturbations46,47 can be realized

in the 3D-mFCCS. These favorable characteristics of the 3D-

mFCCS translate to the possibility of obtaining not only

qualitative but also quantitative in vitro data that is more

comparable to the in vivo models in a high throughput fashion.

Acknowledgements

We thank members of the Cell and Tissue Engineering

Laboratory and Mr Eric Tang of Institute of Materials

Research and Engineering, Agency for Science, Technology

and Research (A*STAR) of Singapore for technical support

and stimulating scientific discussions. This work is supported

in part by the Institute of Bioengineering and

Nanotechnology, Biomedical Research Council, Agency for

Science, Technology and Research (A*STAR) of Singapore;

National Medical Research Council of Singapore (R185-000-

099-213); Academic Research Council from the Ministry of

Education of Singapore (R185-000-135-112); Biomedical

Medical Research Council of Singapore (R185-001-045-305)

and Singapore-MIT Alliance Computation and Systems

Biology Flagship Project funding to HYU. YCT is an

A*STAR Graduate Research Scholar. CZ and YMK are

National University of Singapore graduate research scholars.

HYU is a Singapore-MIT Alliance Faculty Fellow.

References

1 B. J. Kane, M. J. Zinner, M. L. Yarmush and M. Toner, Anal.Chem., 2006, 78, 4291–4298.

2 A. Sivaraman, J. K. Leach, S. Townsend, T. Iida, B. J. Hogan,D. B. Stolz, R. Fry, L. D. Samson, S. R. Tannenbaum andL. G. Griffith, Curr. Drug Metab., 2005, 6, 569–591.

3 A. Khademhosseini, R. Langer, J. Borenstein and J. P. Vacanti,Proc. Natl. Acad. Sci., 2006, 103, 2480–2487.

4 K. S. Sia and G. M. Whitesides, Electrophoresis, 2003, 24,3563–3576.

5 P. S. Dittrich and A. Manz, Nat. Rev. Drug Discovery, 2006, 5,210–218.

6 P. J. Lee, P. J. Hung, V. M. Rao and L. P. Lee, Biotechnol. Bioeng.,2006, 94, 5–14.

7 A. Tourovskaia, X. Figueroa-Masot and A. Folch, Lab Chip, 2005,5.

8 L. Kim, M. D. Vahey, H. Y. Lee and J. Voldman, Lab Chip, 2006,6, 394–406.

9 G. M. Walker, H. C. Zeringueb and D. J. Beebe, Lab Chip, 2004, 4.10 W. Gu, X. Zhu, N. Futai, B. S. Cho and S. Takayama, Proc. Natl.

Acad. Sci., 2004, 101, 15861–15866.11 L. G. Griffith and M. A. Swartz, Nat. Rev. Mol. Cell Biol., 2006, 7,

211–224.12 A. Abbott, Nature, 2003, 424, 870–872.13 K. L. Schmeichel and M. J. Bissell, J. Cell Sci., 2003, 116,

2377–2388.14 D. R. Albrecht, G. H. Underhill, T. B. Wassermann, R. L. Sah and

S. N. Bhatia, Nat. Methods, 2006, 3, 369–375.15 E. Cukierman, R. Pankov, D. R. Stevens and K. M. Yamada,

Science, 2003, 294, 1708–1712.16 V. M. Braga, Curr. Opin. Cell Biol., 2002, 14, 546–556.17 E. Leclerc, Y. Sakai and T. Fujii, Biomed. Microdev., 2003, 5,

109–114.

Fig. 6 Perfusion culture of bone marrow mesenchymal stem cells

(BMSCs) in 3D-mFCCS. (a)–(b) Rat BMSCs after 72 hours of

perfusion culture as assessed by transmission microscopy and confocal

microscopy upon staining with vital dyes, Calcein AM and propidium

iodide. Confocal image was a maximum projection of 74 mm optical

section. (c) Orthogonal projections of a 35 mm optical section showing

F-actin staining of BMSCs after 72 hours culture in the 3D-mFCCS in

comparison to F-actin distribution in a 2D monolayer (insert). (d)

SEM micrograph of BMSCs after 72 hours of perfusion culture. (e)–(f)

Von Kossa staining showing calcium salts deposition (indicated by

arrows) by BMSCs in 3D-mFCCS and 2D monolayer, respectively,

after 1 week of osteogenic induction.

308 | Lab Chip, 2007, 7, 302–309 This journal is � The Royal Society of Chemistry 2007

Dow

nloa

ded

by E

wha

Wom

ens

Uni

vers

ity o

n 30

Jul

y 20

12Pu

blis

hed

on 2

3 Ja

nuar

y 20

07 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/B61

4872

G

View Online

18 Y. C. Toh, S. Ng, Y. M. Khong, V. Samper and H. Yu, Assay DrugDev. Technol., 2005, 3, 169–172.

19 H. Andersson, W. Wijngaart, P. Enoksson and G. Stemme, Sens.Actuators, B, 2000, 67, 203–208.

20 S. M. Chia, K. W. Leong, J. Li, X. Xu, K. Zeng, P. N. Er, S. J. Gaoand H. Yu, Tissue Eng., 2000, 6, 481–495.

21 P. O. Seglen, Methods Cell Biol., 1976, 13, 29–83.22 Y. C. Toh, S. T. Ho, Y. Zhou, D. W. Hutmacher and H. Yu,

Biomaterials, 2005, 26, 4149–4160.23 D. D. Carlo, C. Lonescu-Zanetti, Y. Zhang, P. Hung and L. P. Lee,

Lab Chip, 2005, 5.24 T. A. Crowley and V. Pizziconi, Lab Chip, 2005, 5, 922–929.25 M. Yang, C. W. Li and J. Yang, Anal. Chem., 2002, 74,

3991–4001.26 P. Wilding, L. J. Kricka, J. Cheng, G. Hvichia, M. A. Shoffner and

P. Fortina, Anal. Biochem., 1998, 257, 95–100.27 W. Tan and T. A. Desai, Tissue Eng., 2003, 9, 255–267.28 S. Ng, Y. Wu, Y. Zhou, Y.-E. Toh, Z. Z. Ho, S. M. Chia, J. Zhu

and H. Yu, Biomaterials, 2004, 26, 3153–3163.29 M. J. Paszek, N. Zahir, K. R. Johnson, J. N. Lakins,

G. I. Rozenberg, A. Gefen, C. A. Reinhart-King, S. S. Margulies,M. Dembo, D. Boettiger, D. A. Hammer and V. M. Weaver,Cancer Cell, 2005, 8, 241–254.

30 F. Rosso, A. Giordano, M. Barbarisi and A. Barbarisi, J. CellPhysiol., 2004, 199, 174–180.

31 Y. Chisti, Crit. Rev. Biotechnol., 2001, 21, 67–110.32 A. W. Tilles, H. Baskaran, P. Roy, M. L. Yarmush and M. Toner,

Biotechnol. Bioeng., 2001, 73, 379–389.33 A. M. Otto, C. S. G. Muller, T. Huff and E. Hannappel, J. Cancer

Res. Clin. Oncol., 2002, 128, 247–256.34 E. S. Tzanakakis, L. K. Hansen and W. S. Hu, Cell Motil.

Cytoskeleton, 2001, 48, 175–189.35 J. B. Kim, R. Stein and M. J. O’Hare, Breast Cancer Res. Treat.,

2004, 85, 281–291.36 C. J. Doillon, E. Gagnon, R. Paradis and M. Koutsilieris,

Anticancer Res., 2004 Jul-Aug, 24(4), 2169–2177.

37 V. M. Weaver, S. Lelie’vre, J. N. Lakins, M. A. Chrenek,J. C. R. Jones, F. Giancotti, Z. Werb and M. J. Bissell, CancerCell, 2002, 2, 205–216.

38 J. Walker, C. Martin and R. Callaghan, Eur. J. Cancer, 2004, 40,594–605.

39 L. A. Kunz-Schughart, J. P. Freyer, F. Hofstaedter and R. Ebner,J. Biomol. Screening, 2004, 9, 273–285.

40 E. L. LeCluyse, Eur. J. Pharm. Sci., 2001, 13, 343–368.41 P. Skett, Toxicol. in Vitro, 1994, 8, 491–504.42 T. Engl, K. A. Boost, K. Leckel, W. D. Beecken, D. Jonas,

E. Oppermann, M. K. H. Auth, A. Schaudt, W. O. Bechstein andR. A. Blaheta, Toxicol. in Vitro, 2004, 18, 527–532.

43 Y. Torisawa, H. Shikua, T. Yasukawaa, M. Nishizawab andT. Matsuea, Biomaterials, 2005, 26, 2165–2172.

44 E. I. Kudryavtseva and N. V. Engelhardt, Immunol. Lett., 2003, 90,25–31.

45 L. Richerta, D. Bindaa, G. Hamiltonb, C. Viollon-Abadiea,E. Alexandrec, D. Bigot-Lasserred, R. Barsd, P. Coassoloe andE. LeCluyse, Toxicol. in Vitro, 2002, 16, 89–99.

46 D. R. Plymale, J. R. Haskins and F. A. De la Iglesia, Nat. Med.,1999, 5, 351–355.

47 D. L. Taylor, E. S. Woo and K. A. Giuliano, Curr. Opin.Biotechnol., 2001, 12, 75–81.

48 A. I. Caplan and S. Bruder, Trends Mol. Med., 2001, 7, 259–264.49 M. F. Pittenger, A. M. Mackay, S. C. Beck, R. K. Jaiswal,

R. Douglas, J. D. Mosca, M. A. Moorman, D. W. Simonetti,S. Craig and D. R. Marshak, Science, 1999, 284, 143–147.

5 0 W . P r u d h o m m e , G . Q . D a l e y , P . Z a n d s t r a a n dD. A. Lauffenburger, Proc. Natl. Acad. Sci., 2004, 101, 2900–2905.

51 A. Nur-E-Kamal, I. Ahmed, J. Kamal, M. Schindler andS. Meiners, Stem Cells, 2006, 24, 426–433.

52 N. Datta, Q. P. Pham, U. Sharma, V. I. Sikavitsas, J. A. Jansenand A. G. Mikos, Proc. Natl. Acad. Sci., 2006, 103, 2488–2493.

53 C. A. Gregory, D. J. Prockop and J. L. Spees, Exp. Cell Res., 2005,306, 330–335.

54 D. M. Kristensen, M. Kalisz and J. H. Nielsen, Acta Pathol.Microbiol. Immunol. Scand., 2005, 113, 756–772.

This journal is � The Royal Society of Chemistry 2007 Lab Chip, 2007, 7, 302–309 | 309

Dow

nloa

ded

by E

wha

Wom

ens

Uni

vers

ity o

n 30

Jul

y 20

12Pu

blis

hed

on 2

3 Ja

nuar

y 20

07 o

n ht

tp://

pubs

.rsc

.org

| do

i:10.

1039

/B61

4872

G

View Online