Lab on a Chip

PAPER

Cite this: DOI: 10.1039/c5lc00874c

Received 25th July 2015,Accepted 12th October 2015

DOI: 10.1039/c5lc00874c

www.rsc.org/loc

Liver injury-on-a-chip: microfluidic co-cultureswith integrated biosensors for monitoring livercell signaling during injury†

Qing Zhou,‡a Dipali Patel,‡a Timothy Kwa,a Amranul Haque,a Zimple Matharu,a

Gulnaz Stybayeva,a Yandong Gao,a Anna Mae Diehlb and Alexander Revzin*a

Tissue injury triggers complex communication between cells via secreted signaling molecules such as

cytokines and growth factors. Discerning when and where these signals begin and how they propagate

over time is very challenging with existing cell culture and analysis tools. The goal of this study was to

develop new tools in the form of microfluidic co-cultures with integrated biosensors for local and continu-

ous monitoring of secreted signals. Specifically, we focused on how alcohol injury affects TGF-β signaling

between two liver cell types, hepatocytes and stellate cells. Activation of stellate cells happens early

during liver injury and is at the center of liver fibrosis. We demonstrated that alcohol injury to micro-

fluidic co-cultures caused significantly higher levels of stellate cell activation compared to conditioned

media and transwell injury experiments. This highlighted the advantage of the microfluidic co-culture:

placement of two cell types in close proximity to ensure high local concentrations of injury-promoting

secreted signals. Next, we developed a microsystem consisting of five chambers, two for co-culturing

hepatocytes with stellate cells and three additional chambers containing miniature aptamer-modified

electrodes for monitoring secreted TGF-β. Importantly, the walls separating microfluidic chambers were

actuatable; they could be raised or lowered to create different configurations of the device. The use of

reconfigurable microfluidics and miniature biosensors revealed that alcohol injury causes hepatocytes to

secrete TGF-β molecules, which diffuse over to neighboring stellate cells and trigger production of addi-

tional TGF-β from stellate cells. Our results lend credence to the emerging view of hepatocytes as active

participants of liver injury. Broadly speaking, our microsystem makes it possible to monitor paracrine

crosstalk between two cell types communicating via the same signaling molecule (e.g. TGF-β).

Introduction

The liver highlights the importance of short-range cellularinteractions during injury. Exposure of the liver to toxicantssuch as alcohol triggers complex downstream signalling, lead-ing to enhanced oxidative stress, production of pro-inflammatory and pro-fibrogenic cytokines as well as modifi-cation of proteins and lipids.1–3 These biomolecular eventsdrive liver fibrosis, manifested by aberrant deposition ofextracellular matrix to form scar tissue and loss of hepaticfunction.

Transforming growth factor (TGF)-β is a key fibrogenicmolecule produced early on during liver injury. Higher levels

of TGF-β1 during liver injury are associated with activation ofstellate cells and epithelial-to-mesenchymal transition (EMT)of hepatocytes.4,5 Given that hepatocytes and stellate cellsreside in close proximity to each other, it is nearly impossibleto dissect paracrine interactions via TGF-β or other signallingmolecules in vivo. This crosstalk is difficult to monitor evenin vitro using traditional culture techniques such as condi-tioned media and transwell co-cultures. Moreover, neitherconditioned media nor transwell co-cultures may be wellsuited for replicating the high local concentrations of sec-reted signals present in vivo. In both cases, secreted mole-cules become rapidly diluted in the large volume of mediabathing the cells. Alternatively, the advent of microfluidicdevices permits the placement of different cell types in dis-tinct yet proximal locations on the surface of a confinedvolume.

Microfabrication approaches have been utilized extensivelyin designing micropatterned surfaces and microfluidicdevices for co-cultivation of two or more cell types.6–12 Impor-tantly, similar microfabrication approaches may also be used

Lab ChipThis journal is © The Royal Society of Chemistry 2015

aDepartment of Biomedical Engineering, University of California, Davis,

451 Health Sciences, Davis, CA, USA. E-mail: arevzin@ucdavis.edubDivision of Gastroenterology, Department of Medicine, Duke University,

595 LaSalle Street, Snyderman Building, Suite 1073, Durham, NC, USA

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5lc00874c‡ These authors contributed equally.

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article OnlineView Journal

Lab Chip This journal is © The Royal Society of Chemistry 2015

to place miniature biosensors (e.g. electrodes) next to groupsof cells.13–16 Therefore, the goal of this paper was to developa microfabricated device combining concepts of microfluidiccell cultures and biosensors in order to monitor reciprocalinteractions between liver cells during injury.

While it is possible to employ antibody (Ab)-based detec-tion approaches for monitoring protein molecules releasedfrom cells in microfluidic devices,17,18 these sensing strate-gies provided somewhat limited spatial and temporal resolu-tion for sensing. Our group has a long standing interest inintegrating aptamer-based biosensors with cells for analysisof secreted signals such as cytokines and growth factors.15,19

These biosensors, first described by Plaxco and co-workers,are based on changes in electrochemical signal of redox-labelled, aptamer-functionalized electrodes responding to atarget analyte.20,21 Using this principle, our lab has recentlydeveloped TGF-β aptasensor, integrated it into a microfluidicdevice and monitored release of this cytokine from activatedstellate cells.22

Moving beyond detection of TGF-β from one cell type, wewanted to develop a microfluidic device housing two liver cellcompartments, hepatocytes and stellate cells, and then employbiosensors to monitor paracrine crosstalk between the two celltypes. The present study describes our efforts to develop sucha platform or “liver injury microchip” – a microsystem for cul-turing liver cells, injuring these cells with alcohol, and moni-toring paracrine signalling triggered by the injury.

Materials and methodsChemicals and reagents

Poly (dimethylsiloxane) (PDMS) and silicone elastomer curingagent were purchased from Dow Corning (Midland, MI). Posi-tive photoresist (S1813) and developer solution (MF-319) werepurchased from Shipley (Marlborough, MA). Chromium (CR-4S) and gold etchants (Au-5) were bought from Cyantek Cor-poration (Fremont, CA). 1× phosphate-buffered saline (PBS)without calcium and magnesium, dimethylformamide (DMF),6-mercapto-1-hexanol (MCH), triton-X 100, bovine serumalbumin (BSA), trisIJ2-carboxyethyl)phosphine hydrochloride(TCEP), sodium bicarbonate (NaHCO3), collagen (type I) and4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)were bought from Sigma-Aldrich, USA. Methylene Blue (MB),carboxylic acid and succinimidyl ester (MB-NHS) were fromBiosearch Technologies, Inc. (Novato, CA). Paraformaldehydewas purchased from Electron Microscopy Sciences, USA. Anti-α-smooth muscle actin (α-SMA) was obtained from Abcamand goat anti-rabbit IgG conjugated with Alexa 488 was pur-chased from Invitrogen. 4,6-Diamidino-2-phenylindole (DAPI)was bought from Molecular Probes, Invitrogen. TGF-β ELISAkit was from R&D Systems, USA. Dulbecco's modified Eagle'smedium (DMEM), sodium pyruvate, fetal bovine serum (FBS)and penicillin/streptomycin (PS) were all purchased fromInvitrogen (Carlsbad, CA). Thiolated transforming growth fac-tor (TGF)-β1 DNA aptamer (MW 23689.9) with amine modifi-cation was synthesized by Integrated DNA Technologies, USA.

It has a loop structure with amine group at 5′ and thiol func-tionality at the 3′ end. The aptamer was adapted from ref. 23:5′/5AmMC6/CG*CTCGG*CTTC*ACG*AG*ATT*CGTGT*CGTTGTGT*C*CTGT*A*C*C*CG*C*CTTG*A*C*C*AGT*C*ACT*CT*A-G*AGC*AT*C*CGG*A*CTG/iSpC3//3ThioMC3-D/3′. Herein, “*”represents phosphorothioates on 5′ of both A and C, which isbelieved to be able to enhance nuclease resistance and affin-ity. Aptamer stock solution (100 μM) was made by dissolvingsolid-state aptamer in 1 × TE buffer. The stock solution wasthen diluted to desired concentration in HEPES buffer priorto use.

Device design and fabrication

In this study we developed and tested two types of micro-fluidic devices: 1) a two chamber reconfigurable microfluidicco-culture device and 2) a five chamber reconfigurable co-culture device with integrated biosensors. Fabrication ofthese devices followed similar steps. The differences were inthe design of the control and flow layers of microfluidicdevices. A microfluidic device was comprised of two layers ofPDMS secured onto a 3 × 1 inch glass slide. The layouts forboth PDMS layers were made in AutoCAD and converted intotransparencies by CAD/Art services (Bandon, Oregon).

The microfluidic devices contained two layers: the flowlayer and the control (or actuation) layer for applying positiveor negative pressure in order to regulate the flow layer. In thecase of simple co-culture devices, the flow layer containedtwo parallel channels with 8 mm × 1.8 mm × 0.075 mm(length × width × height) dimensions per channel. The con-trol layer for simple co-culture devices had dimensions of(8 mm × 1 mm × 1 mm).

The flow layer of microfluidic devices with integrated bio-sensors contained five microchannels in total: two 10 mm ×1 mm × 0.1 mm (length × width × height) cell-culture cham-bers and three 10 mm × 0.2 mm × 0.1 mm sensing channels.The control layer contained three channels (8 mm × 1 mm ×0.1 mm) that were aligned with the walls separating cell cul-ture chambers from sensing chambers. Under ambient pres-sure channels were sequestered from each other; however,the walls separating microchannels could be raised by apply-ing negative pressure and communication between the adja-cent channels could commence on cue. Particular care wastaken to ensure robust adhesion of control and flow layersfabricated in PDMS. Protocols for fabricating PDMS deviceswere similar to those described by us recently.24,25 Upon fab-ricating PDMS layer and securing it onto glass substrate, a 16gauge needle was used to punch inlets and outlets for sens-ing channels. A sharp metal puncher was used to punchholes as inlets and outlets for cell culture chambers. As thefinal step, cloning cylinders (Fisher Scientific, Pittsburgh, PA)were secured into PDMS to provide media reservoirs for cellculture.

Microfluidic devices used for detection of cell-secreted sig-nals contained gold electrodes micropatterned on glass sub-strates. The electrodes were micropatterned using standard

Lab on a ChipPaper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab ChipThis journal is © The Royal Society of Chemistry 2015

photolithography and wet etching protocols described by uspreviously.36 The electrodes were arranged into 4 × 3 arrayssuch that 4 rectangular electrodes with 475 μm × 150 μmindividual dimensions fit into one sensing channel. Eachelectrode was connected, via a 20 μm lead, to a 3 mm × 4.5mm contact pad located on the edge of the glass substrate.The glass substrate and the flow layer were carefully alignedusing fiduciary marks so that each column of electrodes inthe 3 × 4 array on the substrate was enclosed in andprotected by one fluidic channel. Miniature copper alligatorclamps with soldered wires were used to connect contactpads to a computer-controlled multiplexer (NI ER-16,National Instrument, TX).

Cell preparation and culture in microfluidic devices

A human hepatic stellate cell line (LX2) was cultured in T-75flasks in DMEM supplemented with 0.5% FBS, 200 units permL penicillin and 200 μg mL−1 streptomycin. Cells were pas-saged after achieving 90% confluence. Primary hepatocyteswere isolated from adult female Lewis rats using following astandard two-step collagenase perfusion protocol.26 All exper-iments were performed in compliance with the relevant lawsand institutional guidelines and have been approved by theInstitutional Animal Use and Care Committee at UC Davis.These primary hepatocytes were maintained in DMEMsupplemented with epidermal growth factor (EGF), glucagon,hydrocortisone sodium succinate, recombinant human insu-lin, 200 units per mL penicillin, 200 μg mL−1 streptomycinand 10% FBS (CH media).

Prior to cell seeding, 0.1 mg mL−1 solution of collagen(type I) was infused into the cell-culture chambers andmaintained at room temperature for approximately 1 h toallow collagen deposition on the glass surface. The surfacewas then rinsed with copious amounts of PBS and sterilizedunder UV. Stellate cells and/or hepatocytes were resuspendedat a concentration of 1.0 × 106 cells per mL and were loadedinto separate reservoirs leading to separate chambers. In theco-culture experiment, hepatocytes and stellate cells wereseeded into adjacent chambers of the same microfluidicdevice. The devices were then kept at 37 °C in a humidified5% CO2 environment. Cells were incubated overnight prior tothe start of ethanol injury experiment to allow for primaryhepatocytes to get acclimated to the in vitro environment.

Comparing alcohol injury response for three culture systems:microfluidic co-cultures, transwell co-cultures and condi-tioned media

For evaluation of stellate cell activation due to hepatic injuryacross three culture platforms, cell seeding concentrationswere kept constant at 1.0 × 106 cells per mL.

For conditioned media experiments, hepatocytes wereseeded onto a 6-well plate and incubated with 100 mM etha-nol for 48 h in CH media supplemented with 0.5% FBS.Media was then collected and transferred to monocultures ofstellate cells and the cells were incubated for additional 48 h.

For transwell plate experiments, hepatocytes were seededat the bottom of a 6-well plate and incubated with ethanol ina similar fashion as with conditioned media. After 48 h, atranswell insert containing stellate cells was placed in thewell and the system was incubated for another 48 h.

For microfluidic co-culture experiments, hepatocytes andstellate cells were seeded into adjacent microchannels of amicrofluidic device. Then hepatocytes microchannelcontaining hepatocytes was exposed to 100 mM ethanol for48 h, after which the microfluidic device was reconfigured toallow communication between injured hepatocytes andneighbouring stellate cells. The two cell types were exposedto 100 mM ethanol for additional 48 h.

Another set of experiments was designed to interfere withTGF-β signalling using neutralizing antibodies. Hepatocytesand stellate cells were seeded into microfluidic co-culturedevices as described above. Then hepatocytes were selectivelyexposed to 100 mM ethanol for 48 h, after which the devicewas reconfigured to allow for injured hepatocytes to interactwith neighbouring stellate cell compartment. In addition toethanol, media was supplemented with anti-TGF-β1 neutraliz-ing antibodies (Abs) for the duration of hepatocyte–stellatecell interaction (48 h).

Immunofluorescent staining for markers of stellatecell activation

Stellate cells were co-cultured with either healthy or ethanolinjured hepatocytes for approximately 24 hours in the micro-device. The stellate cells were then fixed in 4% paraformalde-hyde (Electron Microscopy Sciences) + 0.3% Triton-X100(Sigma) in PBS for 15 min, followed by an incubation step in1% bovine serum albumin (BSA) for 1 hour to prevent non-specific binding. The cells were then rinsed with PBS andincubated with rabbit anti-α-smooth muscle actin (α-SMA,Abcam) antibody for 90 min. After a 5 min washing step withPBS, the cells were incubated with goat anti-rabbit IgG conju-gated with Alexa-488 (diluted in 1 : 1000) for 1 h. The cellswere washed in PBS for 5 min again prior to 15 min incuba-tion in 4,6-diamidino-2-phenylindole (DAPI).

Quantitative real-time PCR analysis of stellate cell activation

With the barrier between hepatocytes and stellate cellslowered, stellate cells were treated for 10 min with trypsin-EDTA, followed by aspirating a volume of 500 μL of freshmedia using a pipette and recovery of cells from outlet. Thecells were spun down and then resuspended in 200 mL oflysis buffer (Roche) and then stored at −80 °C. PCR primers,listed below in Table 1, were designed for human gene analy-sis. Total RNA extraction and cDNA preparation wasperformed according to the manufacturer's instructions(Roche). All PCR reactions were done in duplicate. The rela-tive expression level of each gene was calculated using thecomparative threshold cycle (Ct) method with β-actin as ahousekeeping gene. The number of samples (culture

Lab on a Chip Paper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab Chip This journal is © The Royal Society of Chemistry 2015

surfaces) used for statistical analysis of PCR data was n = 3for all conditions.

Inhibition of TGF-β1 signalling

A monoculture of hepatocytes in microchambers was injuredfor 48 hours with 100 mM ethanol. The chamber roof wasthen raised to allow communication between injured hepato-cytes and stellate cells. At the same time, neutralizing TGF-β1antibodies (R&D Systems) were added at the concentration of10 μg ml−1 dissolved in CH media. Co-cultures weremaintained in the presence of neutralizing antibodies for 48hours, then the chamber roof was lowered again and stellatecells were selectively trypsinized and analyzed by RT-PCR forexpression of genes associated with activation. Similar TGF-βneutralization strategy was employed in a microfluidic devicewith integrated biosensors.

Aptasensors for detection of TGF-β

Prior to electrode surface immobilization, the 5′-end ofamino-modified TGF-β1 aptamer was first conjugated toMethylene Blue NHS Ester, an electroactive redox label,through the succinimide ester coupling reaction. 0.3 mg ofMB-NHS was added to 50 μL aptamer stock (100 μM),followed by 20 μL of dimethyl formamide (DMF) and 10 μLof NaCO3 buffer (0.5 M) to maintain a pH of 8.3. The solutionwas kept at 4 °C for 4 h to allow the MB-NHS ester to belinked to the amine group at the end of the aptamer. Theprepared aptamer stock solution was then reduced in 10 mMTCEP for 1 h to cleave disulfide bonds. The stock solutionwas then further diluted in HEPES buffer to 1 μM, which wasdemonstrated to be the best concentration for optimal TGF-β1 aptamer density on the electrode surface.22 The immobili-zation of aptamer onto the Au surface was achieved by incu-bating 1 μM MB-tagged TGF-β1 aptamer solution over micro-patterned electrodes overnight at 4 °C. After three rinsingsteps in HEPES buffer, the electrode surface was passivatedin 1 mM MCH for 15 min to remove non-specificallyabsorbed aptamer strands.

Electrochemical measurements inside microfluidic channels

Electrochemical signals were measured by square wavevoltammetry (SWV) using a potentiostat (CHI842b, CH

Instrument, Austin, TX). Each scan started from −0.5 V andended at 0 V, with 0.004 V step potential, 0.04 V amplitudeand 60 Hz frequency, which were confirmed to be the opti-mized condition for the electro-active MB moiety.27 A customprogram was written in Sikuli to automate electrochemicalmeasurements. Large sets of electrochemical data collectedover the course 14 h experiment were analysed using aPython script. This script was designed to identify a reduc-tion peak from an SWV curve. The switching between differ-ent electrodes within the array was performed using a homebuilt multiplexer described elsewhere.24

Modeling electrochemical signals in microfluidic devices

Finite element model was set up to simulate the detectionprocess using COMSOL Multiphysics (COMSOL Inc., Burling-ton, MA). The model accounted for the transport of cytokinesin the media and their binding to aptamer-modifiedelectrodes,

(1)

(2)

where C is the concentration of cytokine, D is the diffusioncoefficient, B is cytokine-aptamer conjugation on the sensor'ssurface; A0 is the initial surface concentration of aptamer; konand koff is the association and disassociation constant,respectively. On the surface covered by cells the boundarycondition is

(3)

where is the cell secretion rate, Nc is the total numberof cells captured in the antibody coated region Ac. Thegeometry used for simulation was the same as the actualdevice. The main parameters used for modeling werelisted in Table 2. The cell secretion rate, , was deter-mined by least squares fitting of experimental sensorsignal.28

Table 1 Primer sequences for real-time PCR

Gene Sequence 5′ to 3′

Housekeeping – β-actin F: ACGGCCAGGTCATCACTATTGR: ATACCCAAGAAGGAAGGCTGGA

Collagen type I α1 F: AACGCGTGTCATCCCTTGTR: GAACGAGGTAGTCTTTCAGCAACA

TGF-β1 F: CCCTGGACACCAACTATTGCR: CTTCCAGCCGAGGTCCTT

TIMP F: TGGCATCCTGTTGTTGR: AGAAGGCCGTCTGTGGGT

SMA-α F: CAGCCAAGCACTGTCAGGR: CCAGAGCCATTGTCACACAC

Table 2 Parameters used for modeling cell secretion

Parameters Values

Diffusion coefficient (D) 6.5 × 10−7 cm2 s−1

Surface binding density (A0) 5.61 × 10−12 mol cm−2

Association rate constant (kon) 4.48 × 105 1 s−1 MDissociation rate constant (koff) 4.82 × 10−4 1 s−1

Number of hepatocytes (Nc) 3000Number of stellate cells (Nc) 1500

Lab on a ChipPaper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab ChipThis journal is © The Royal Society of Chemistry 2015

Results and discussionAlcohol injury to microfluidic co-cultures activates stellate cells

In the liver, stellate cells reside in the space of Disse whichrepresents a distance of ~30 to 100 μm between the hepato-cytes and the endothelial cells lining sinusoids.2,29,30–32 Wehypothesized that close proximity of the two cell types in themicrofluidic co-culture system may better recapitulate localconcentrations of signaling molecules produced during liverinjury. To test this hypothesis, we employed a microfluidicco-culture system of the type shown in Fig. 1(A and B) wherehepatocytes and stellate cells reside in adjacent microfluidiccompartments (100 μm distance) divided by a retractablewall. This design allows seeding and injuring each cell typeseparately before raising the wall to commence paracrinecommunication between the cells. In the first set of experi-ments we compared three in vitro models of alcohol injury: 1)conditioned media was collected from alcohol-injured hepa-tocytes and delivered to stellate cells, 2) hepatocytes and stel-late cells were exposed to alcohol in a transwell plate and 3)microfluidic co-cultures of hepatocytes and stellate cells wereinjured with alcohol. In light of our interest in hepatic contri-butions to liver injury, in each of the three scenarios hepato-cytes were first exposed to 100 mM EtOH for 48 h. In

scenario 1) conditioned media from injured hepatocytes wastransferred to stellate cells, in scenario 2) stellate cells on topof a transwell insert were introduced into a well containinginjured hepatocytes, in scenario 3) the wall separating hepa-tocyte and stellate cells was raised to commence communica-tion between the two cell types. The duration of injury tohepatocytes and the time of interaction between stellate cellsand injured hepatocytes was the same for all three injurymodels. RT-PCR analysis for TGF-β1, collagen type I, smoothmuscle actin (SMA)-α and tissue inhibitor of meta-lloproteases (TIMP) was used to measure stellate cell activa-tion. RT-PCR results (Fig. 1C–F) showed that markers of stel-late cell activation were invariably higher in microfluidic co-cultures compared to conditioned media and transwell sce-narios. In the case of TIMP and SMA, only microfluidic co-cultures showed significant differences in gene expression.In case of collagen I, gene expression was 2.5 fold higher inthe microfluidic co-cultures than in transwell co-culturesinjured by exposure to 100 mM ethanol. Gene expressionresults were confirmed by immunostaining stellate cells forSMA-α. As seen from Fig. 1G and H stellate cells in alcoholinjured microfluidic co-cultures contained much higherlevels of this activation marker. These experimentsunderscored that microfluidic co-cultures may represent a

Fig. 1 Alcohol injury to microfluidic co-cultures causes activation of stellate cells. (A) Design of reconfigurable microfluidic co-culture system.The wall separating hepatocytes and stellate cells could be raised on cue to commence paracrine interactions between the two cells types. (B) Atypical microfluidic device used in a co-culture experiment. Microchannels were filled with red and green food dye. The wall separating micro-channels was lowered to demonstrate the food dye did not mix. (C–F) Hepatocytes were injured with alcohol (100 mM EtOH) alone for 48 h, thenwere allowed to interact with stellate cells for additional 48 h. Markers of stellate cell activation were analyzed by RT-PCR for three systems testedin parallel: microfluidic (μF) co-cultures, conditioned media (CM) and transwell co-cultures. (−) without ethanol and (+) with ethanol. Scale bar rep-resents 100 μm. Results are represented as mean ± SD (n = 3). *: p-value < 0.05. NS: not significant. (G–H) Immunostaining of SMA-α in stellatecells co-cultured in microfluidic devices with healthy (G) and injured (H) hepatocytes.

Lab on a Chip Paper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab Chip This journal is © The Royal Society of Chemistry 2015

more sensitive liver injury model compared to traditionalcell culture approaches.

What is the reason for enhanced stellate cell activation inmicrofluidic devices? To ensure that the environment insidethe microdevices was not predisposing cells to injury, we cul-tivated hepatocytes and stellate cells inside microfluidicdevices in the absence of ethanol. We found that both func-tion of hepatocytes and quiescence of stellate cells weremaintained inside the microfluidic channel (Fig. S1†) thuseliminating the possibility that our devices were predisposingcells to injury. Therefore, we believe it is close proximitybetween the cell types that makes microfluidic co-cultures aninteresting in vitro model of liver injury.

Interfering with secreted TGF-β1 attenuates stellate cellactivation during injury to microfluidic co-cultures

In the next set of experiments we sought to characterize therole of hepatic TGF-β1 in stellate cell activation upon alcoholinjury using novel microfluidic co-culture system. We choseto focus on TGF-β1 for several reasons: 1) this cytokine isconsidered to be the key signal driving liver fibrosis and 2)while TGF-β production by stellate cells during injury is wellestablished,4,33 hepatic production of this cytokine has notbeen explored extensively.34–36 In our experiments hepato-cytes were injured by exposure to 100 mM EtOH for 48 h.While high, this concentration of alcohol was employed by uspreviously to injure hepatocytes37 and was tolerated by thesecells. Subsequently the microfluidic system was reconfigured,allowing injured hepatocytes to communicate with neighbor-ing stellate cells via secreted factors for additional 48 h. Totest the role of secreted TGF-β, neutralizing anti-TGF-β1 Abswere added to some of the microfluidic devices at the hepato-cyte–stellate cell communication stage (see Fig. S2† forschematic). Trypsinization of stellate cells followed byRT-PCR analysis of stellate cell activation markers revealedthat treatment with neutralizing Abs caused significantdownregulation in all markers of stellate cell activation(Fig. 2A–D). In fact, anti-TGF-β1 treatment of alcohol-injuredco-cultures made levels of collagen I and SMA-α statisticallyinsignificant from healthy co-cultures. In separate controlexperiments we confirmed that exposure of stellate cellsalone to alcohol caused neither activation nor upregulationof TGF-β release, whereas alcohol treatment did cause hepa-tocytes to upregulate production of TGF-β (results describedlater in this paper).

Observations with alcohol injury in the microfluidic co-culture system described in Fig. 1 and 2 recapitulate effectsof alcohol injury in vivo including activation of stellate cellsand upregulation of TGF-β production, thereby underscoringthe utility of a microfluidic co-culture as a model of liverinjury. However, these results also highlight the shortcom-ings of standard molecular biology approaches for studyingcellular crosstalk – 1) the role of hepatic TGF-β on stellate cellactivation is not shown directly but is inferred from controlexperiments testing one cell type at a time and 2) it is

difficult to characterize dynamics of reciprocal signalingusing molecular biology approaches (e.g. ELISA or RT-PCR).To address these shortcomings, we developed more sophisti-cated microfluidic co-culture devices with integrated minia-ture biosensors for monitoring cellular cross-talk vis-à-visTGF-β1.

Reconfigurable microfluidic co-culture devices withintegrated biosensors for TGF-β

Having established that a microfluidic co-culture may repre-sent an interesting model of liver injury and having demon-strated an important role of TGF-β in this model, we wantedto better understand cellular origins and dynamics of TGF-βsecretion. To accomplish this goal, microfluidic co-cultureswere integrated with aptamer-based biosensors for monitor-ing TGF-β secretions. As seen from Fig. 3, this microsystemcontained two cell culture compartments flanked by sensingchambers for a total of five parallel microfluidic channels.Each sensing chamber contained an array of four miniatureindividually addressable gold electrodes connected to apotentiostat via leads and contact pads. Fig. S3† provides athree dimensional view of the control and flow layers of themicrofluidic device.

Fig. 2 Secreted TGF-β contributes to activation of stellate cells inmicrofluidic devices. Experiment set-up: hepatocytes were selectivelyinjured with 100 mM EtOH for 48 h, then microfluidic device wasreconfigured to allow injured hepatocytes to communicate with stel-late cells. Anti-TGF-β1 antibodies were used to neutralize the effectsof secreted TGF-β during hepatocyte–stellate cell communicationstage of the experiment. Control experiments were carried out sameas above but without neutralizing Abs. (A–D) RT-PCR analysis of stel-late cell activation markers. It is noteworthy that blocking secretedTGF-β attenuates stellate cell activation. Results are represented asmean ± SD (n = 3). *: p-value < 0.05. NS: non-significant.

Lab on a ChipPaper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab ChipThis journal is © The Royal Society of Chemistry 2015

Operation of the reconfigurable microfluidic device. Nota-bly, communication between adjacent cell compartments andsensing chambers was controlled by actuating pneumaticvalves built into the two-layer PDMS device. Fig. 4 outlinesthe operation of this reconfigurable microfluidic device: 1) allwalls of the microfluidic device are lowered effectivelysequestering cell compartments from sensing chambers. Inthis state, cell culture compartments may be coated withadhesive ligands (e.g. collagen I) and seeded with cells with-out affecting neighboring biosensors. 2) Hepatocytes areinjured by exposure to alcohol after which hepatic TGF-β pro-duction is monitored for the desired period of time using anarray of biosensors in Sensing Chamber 1. 3) The device isreconfigured to allow for injured hepatocytes to communicatewith neighboring stellate cells. Both cell types are exposed toalcohol during this stage. TGF-β secretion is monitored usinga new array of aptasensors located in Sensing Chamber 2. 4)The microfluidic device is reconfigured once again, this timeto sequester stellate cells away from hepatocytes in order tomeasure TGF-β secretion from these cells. A new set ofaptasensors in Sensing Chamber 3 is used for this experi-ment. Overall, combining reconfigurable microfluidic devices

with aptasensors allows the compartmentalization of cellularinteractions into stages in order to determine the cellular ori-gins of secreted factors in addition to the dynamics of recip-rocal interactions.

Operation of TGF-β aptasensor. The aptasensor for detec-tion of TGF-β was designed based on principles involvingconformational changes first reported by Plaxco and co-workers.20,21 As described in Fig. 5, cytokine molecules sec-reted by liver cells bound to neighboring aptamer-modifiedelectrodes causing a change in the electrochemical signal.The electrochemical properties of aptamer-modifiedelectrodes were monitored with square wave voltammetry(SWV). The electrochemical signal was reported as signal sup-pression – SS% = (initial current − final current)/(initial cur-rent). Such an aptasensor produced signal in near real-timewithout the need for labeling and washing steps normallyassociated with Ab-based immunoassays. Therefore, thisaptasensor was particularly well suited for monitoringdynamics of TGF-β secretion in the microfluidic co-cultures.Our lab has recently reported on the development of TGF-βaptasensor and the reader is referred to this paper fordetailed description of biosensor.22

Fig. 3 Reconfigurable microfluidic devices with integrated aptasensors. (A) Typical microfluidic devices comprised of two layer PDMS securedonto glass substrates. Gold electrodes are micropatterned on glass. These electrodes are functionalized with TGF-β aptamers. (B) The microfluidicdevices contains five channels total: two cell compartments (green fluorescence) and three sensing compartments (red fluorescence); (C) brightfield image showing gold electrodes inside sensing chambers (black rectangles). There are four individually addressable electrodes in each channel.Only two electrodes per channel are shown in this image.

Fig. 4 Pneumatic operation of the two-layer device to break up the bi-directional communication between hepatocytes and stellate cells intothree stages. Step 1: valves are closed to allow for pre-coating cell culture chambers with collagen I, seeding cells and constructing aptasensors.Step 2: in the first stage, 100 mM EtOH in media was infused into hepatocyte chamber to injure these cells. Step 3: in the second stage, the middlebarrier was raised to allow injured hepatocytes to communicate with quiescent stellate cells. Step 4: in the third stage, the middle barrier waslowered again to sequester stellate cells from injured hepatocytes. After rinsing out the media, TGF-β1 production from stellate cells wasmonitored.

Lab on a Chip Paper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab Chip This journal is © The Royal Society of Chemistry 2015

Particular care was taken to ensure that actuation of onevalve did not inadvertently cause leakage in neighboring com-partments. Fig. S4† demonstrates one set of experimentsinvolving infusion of fluorescent dye (Alexa 546) into thethree sensing chambers of the microfluidic device and actuat-ing the valves one by one. As seen from these images, actua-tion of individual valves caused the dye to diffuse only withinthe confines of chambers being reconfigured while othercompartments remained effectively sealed.

A similar concept was used to verify electrochemicalindependence of electrodes residing in adjacent sensingchambers. In these experiments gold electrodes were func-tionalized with TGF-β aptamers and then challenged withrecombinant TGF-β1. Square wave voltammetry (SWV) wasused to characterize redox activity of the electrodes chal-lenged with TGF-β compared to those that were not exposedto the target analyte. These experiments (described in Fig.S5†) confirmed that actuation of specific valves did notcause undesirable crosstalk between adjacent sensingchambers.

TGF-β1 secretion from hepatocytes and stellate cells inmono-cultures

Before exposing hepatocyte–stellate cell co-cultures to alco-hol, we investigated responses of each cell type individually.These experiments utilized the same microfluidic platformshown in Fig. 4, with either hepatocytes or stellate cellsseeded into both cell culture chambers. Media containing100 mM ethanol was infused into one cell compartment tomimic the ethanol injury scenario, while the other cell com-partment contained regular media and served as negativecontrol. During the initial 24 h observation period, the mid-dle barrier remained closed, which kept the two cell culturechambers separated while responses of injured and controlcells were independently monitored on the same microchip.The experiments were carried out in a custom-made environ-mental box operating under physiological conditions.

Fig. 6A shows images of hepatocytes and stellate cells.Hepatocytes possess a phenotype typical of epithelial cells –

cobblestone morphology with prominent nuclei and cell

Fig. 5 Principle of TGF-β detection. Binding of cytokine molecules causes aptamer molecules to change conformation. As a result, redox reporteron an aptamer molecules moves away from and the electrochemical signal decreases.

Fig. 6 Monitoring TGF-β secretion in monocultures of hepatocytes and stellate cells. (A) Images of hepatocytes (upper panel) and stellate cells (lower panel)cultured in microfluidic devices; (B) electrochemical detection of TGF-β1 from ethanol-injured or healthy hepatocytes; (C) dynamic monitoring of TGF-β1from either quiescent or PDGF stimulated or ethanol injured or stellate cells. In both (B) and (C), signal suppression refers to normalized change in redoxcurrent. (D) ELISA analysis of TGF-β release from hepatocytes injured with alcohol. (E) ELISA analysis of TGF-β release from stellate cells activated by PDGF.

Lab on a ChipPaper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab ChipThis journal is © The Royal Society of Chemistry 2015

borders. Importantly, although 100 mM ethanol concentra-tion is quite high, based on our unpublished results, hepato-cytes remain functional for several days. A typical electro-chemical signal from injured vs. healthy hepatocytes isshown in Fig. 6B. As seen from these data, alcohol injury tohepatocytes resulted in significant changes in the redox cur-rents measured at aptamer-modified electrodes while nosuch changes were observed in healthy hepatocytes exposedto culture media without alcohol. Our recent study rigorouslycharacterized specificity of TGF-β aptasensor using both spik-ing of nonsense proteins and also interference with cell func-tion.22 This past study revealed that aptasensor was mostsensitive to TGF-β1 but did have 15% cross-reactivity withTGF-β2 and 30% cross-reactivity with TGF-β3. The aptasensordid not distinguish between the latent vs. active form ofTGF-β. The limit of detection for TGF-β aptasensor was 1 ngml−1 with linear range extending to 250 ng ml−1. Importantly,our previous study showed that exposure of neutralizingTGF-β1 Abs to activated stellate cells abrogated changes inredox currents of the aptasensor, thereby proving that thebiosensor was indeed monitoring cell-secreted TGF-β1. There-fore, electrochemical signals reported in Fig. 6 can be attrib-uted to production of TGF-β with a high degree of certainty.

While hepatocytes responded to alcohol by vigorouslyreleasing TGF-β, no such response was observed from stellatecells (Fig. 6C). This result is not unexpected given thatstellate cells are not generally considered capable of alcoholmetabolism, although a limited number of reportsconnecting stellate cells to expression of alcohol dehydroge-nase has appeared.38 In fact, TGF-β signal from alcohol-injured stellate cells was similar to the control experimentwith media only. Stellate cells are commonly thought of asthe main producer of TGF-β in the liver and, as shown inFig. 6C, they do produce this cytokine when stimulated withplatelet derived growth factor (PDGF) – a signaling moleculeknown to activate stellate cells.39 Importantly, TGF-β ELISAwas performed as an alternative strategy to confirm

production of this cytokine by liver cells cultured in micro-fluidic devices. A reader should note that during ELISA exper-iments, the cells were incubated with alcohol for 48 h. ELISAresults shown in Fig. 6(D and E) mirror the responses mea-sured by the TGF-β aptasensors – 1) hepatocytes upregulaterelease of TGF-β upon alcohol injury and 2) while stellatecells alone do not respond to alcohol challenge, these cellsdo produce TGF-β when stimulated with PDGF. Overall, thisset of experiments revealed that alcohol exposure inducedhepatocytes to secrete TGF-β1 while stellate cells did not pro-duce this cytokine in response to direct alcohol injury. Armedwith this knowledge, we set up a co-culture experiment tomonitor hepatocyte–stellate cell crosstalk via TGF-β. Wehypothesized that injured hepatocytes send TGF-β moleculesto activate neighboring stellate cells.

Hepatocyte–stellate cell cross-talk via secreted TGF-β duringalcohol injury

The final set of experiments was designed to rigorously provethat hepatocytes are the drivers of early liver injury and thatstellate cells respond to paracrine signals produced by theinjured hepatocytes. Primary hepatocytes and stellate cells(LX2 cell line) were seeded into separate cell culture compart-ments (Fig. 4, step 1). Subsequently, the hepatocytes wereexposed to culture media containing alcohol and their pro-duction of TGF-β was monitored for 8 h (see Fig. 7A). Duringthis time stellate cells were sequestered in a separate com-partment and remained quiescent. As shown in Fig. 7A,alcohol-injured hepatocytes secreted significant levels ofTGF-β. (Discussion of secretion rates and concentrationswill follow shortly). After this, the microfluidic device wasreconfigured to allow for injured hepatocytes to communicatewith stellate cells (Fig. 7B). Production of TGF-β during thiscommunication stage was monitored using a brand new setof aptasensors. We should note however that hepatocyte–stellatecell communication chamber shown in Fig. 7B has a volume

Fig. 7 Monitoring TGF-β release in hepatocyte–stellate cell co-cultures. (A) Hepatocytes were exposed to culture media containing alcohol; (B)the micro-device was reconfigured to allow for intercellular communication between injured hepatocytes and stellate cells; (C) stellate cells weresequestered from hepatocytes and a new set of aptasensors was deployed to monitor TGF-β1 production from stellate cells. Squares curve (□)represents control experiment where neutralizing Ab was used at the hepatocyte–stellate cell communication stage.

Lab on a Chip Paper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab Chip This journal is © The Royal Society of Chemistry 2015

almost twice as large as the hepatocyte chamber in Fig. 7A.Therefore, making direct comparisons of signal strengthsbetween the two sensors is challenging. However, as demon-strated by modeling described below, it is possible to ascer-tain secretion rates of hepatocytes alone and hepatocytes–stellate cell co-cultures using the data presented inFig. 7(A and B). In the final stage of the experiment, stellatecells were once again sequestered from the hepatocytes. Thecell culture chamber was flushed out to remove existingmedia and cytokines that might have accumulated duringstage 2 of the experiment. A new set of aptasensors wasdeployed in stage 3 and was used to monitor additional stel-late cells' secretion of TGF-β. Fig. 7C demonstrates that asresult of communication with hepatocytes, there was anupregulation in production of TGF-β from stellate cells, withsignal suppression levels reaching 40% after 8 h of monitor-ing. For comparison, signals for quiescent stellate cells hov-ered around 15% (see Fig. 7C for example). The use ofreconfigurable microfluidics and TGF-β biosensors facilitatesthe experimental validation of a sequence of events whereby:1) hepatocytes become injured by alcohol, commence produc-tion of TGF-β and likely other signaling molecules (e.g.ROS),40 2) stellate cells become activated as result of interac-tions with injured hepatocytes and 3) stellate cells produceTGF-β in response to paracrine signals arriving from injuredhepatocytes.

To further confirm the role of hepatic TGF-β in activationof stellate cells, we carried out an additional control experi-ment. Hepatocytes were first exposed to alcohol (step shownin Fig. 7A) but then when the device was reconfigured toallow communication between hepatocytes and stellate cells(Fig. 7B), anti-TGF-β1 neutralizing Abs were introduced intothe chamber. Therefore, while hepatocytes were injured byalcohol, TGF-β molecules secreted by these injured cells wereneutralized with Abs. Bottom panel of Fig. S3† provides asimplified schematic of this experiment. Subsequently,microfluidic device was reconfigured and stellate cell produc-tion of TGF-β was monitored. These data, plotted inFig. 7C (squares curve), demonstrate that the presence ofneutralizing anti-TGF-β Abs caused the stellate cell produc-tion of TGF-β to decrease. This result further underscores therole of TGF-β released from injured hepatocytes in activatingneighboring stellate cells.

Modeling TGF-β secretions to confirm reciprocal interactionsbetween hepatocytes and stellate cells

We hypothesized that TGF-β aptasensor response shown inFig. 7B was due to combined secretion from hepatocytes andstellate cells, and that while hepatocytes were producingTGF-β from the start of the experiment, stellate cells beganproducing this cytokine at a later time point. One way to vali-date this hypothesis was by demonstrating that stellate cellproduction of TGF-β was affected by adding neutralizingTGF-β Abs during hepatocyte–stellate cell communication(Fig. 7C). Another strategy was to compare experimental data

of TGF-β secretion from Fig. 7B with two hypothetical scenar-ios 1) that only hepatocytes produced TGF-β in co-culturesand 2) that both hepatocytes and stellate cells produced TGF-β from the beginning of the experiment. To carry out thiscomparison, we needed to determine secretion rates forhepatocytes and stellate cells. While the electrochemical

Fig. 8 Modeling TGF-β release in a microfluidic device. (A) Diffusion–reaction model of both hepatocytes and stellate cells secreting TGF-βfrom the start of injury experiment. (B) A model of hepatocytes releas-ing TGF-β while stellate cells remain in the chamber but do not pro-duce this cytokine. (C) “Theoretical” sensor signal for TGF-β produc-tion from both cell types (○) and from hepatocytes only (Δ) comparedto experimental data from both cell types injured with alcohol (■). Thesecretion rates were calculated to be σH = 0.00576 pg per cell per hand σs = 0.00159 pg per cell per h for hepatocytes and stellate cellsrespectively.

Lab on a ChipPaper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab ChipThis journal is © The Royal Society of Chemistry 2015

signal in Fig. 7 is a function of the cell secretion rate, it isalso a function of TGF-β diffusion and reaction on theelectrode surface. We therefore developed a diffusion–reac-tion model in COMSOL utilizing actual cell numbers and thedimensions of the microfluidic device. Secretion rates wereassigned iteratively to generate theoretical sensor responsesand were compared against experimental data. Using thisstrategy we determined TGF-β secretion rate to be 0.00576 pgper cell per h for hepatocytes (based on experimental data inFig. 7A) and 0.00159 pg per cell per h for stellate cells (basedon experimental data in Fig. 7C). Fig. 8(A and B) showsCOMSOL models of “theoretical” electrochemical signals forthe two scenarios 1) both hepatocytes and stellate cells pro-duce TGF-β from the outset at rates noted above and 2) onlyhepatocytes are secreting TGF-β at the rate mentioned above.The results plotted in Fig. 8C alongside experimental TGF-βsecretion curve provide visual evidence that the redox signalmeasured during hepatocyte–stellate cell interaction was acombination of hepatic and stellate cell production of TGF-β1. In fact, the experimental signal closely matches the theo-retical “hepatocyte only” curve initially but then inflects atabout 4 h time point and eventually reaches the upperboundary set by the other theoretical scenario where bothhepatocytes and stellate cells are secreting TGF-β. Theseresults provide further evidence in support of the notion thatstellate cells begin producing TGF-β as a response to TGF-βand other molecules arriving from injured hepatocytes. Areader should note that while secretion rates were obtainedfrom actual experimental data described in Fig. 7(A and C),these secretion rates were then used to generate a theoreticalsignal within a much larger compartment comprised of twocell culture and one sensing chambers (Fig. 4, step 3). There-fore, theoretical signal due to hepatocyte secretion in Fig. 8C,does not directly correlate to experimental results in Fig. 7A.The secretion rate and cell number is the same in both cases,but the electrochemical signal is significantly higher inFig. 7A due to smaller volume of the chamber.

In summary, the use of reconfigurable microfluidic deviceallowed us to segment the hepatocyte–stellate cell communi-cation experiment into three stages to conclusively confirmthat alcohol injury perturbs hepatocytes which then sendparacrine signals to activate stellate cells.

Conclusions

Our findings clearly point to the importance of hepatocytesas early inducers of liver injury. These results are reasonablystraightforward to rationalize – alcohol injures the cells thatcan metabolize it – the hepatocytes – who in turn send injurysignals to neighboring stellate cells. However, the tools forexperimentally validating this sequence of events remain lim-ited. What criteria must be met in order for attribution of cel-lular signaling to take place? The two cell types need to be inseparate but proximal locations. However, while proximity isnecessary it is not sufficient to resolve this challenge. We andothers have previously developed micropatterned co-cultures

of liver cells.6,13,37,41 These systems satisfy proximity require-ment; however, it is difficult to assign secreted factors to onecell type vs. another in a culture system where both cell typesare exposed to the same culture media conditions. Themicrofluidic co-culture device described here not only allowsthe placement of two liver cell types in proximal locationsbut can also be configured to permit or prevent heterotypicinteractions. Such a reconfigurable microdevice may be usedto isolate individual cell types in order to measure cell-typespecific secretions before or after heterotypic interactionshave taken place. In addition, the ability to reconfigure andisolate cellular compartments may be used for cell-type spe-cific gene expression analysis.

Biological questions to be studied with this technologymay include elucidation of mechanisms that modulate livercell plasticity during wound healing responses. Emerging evi-dence suggests that re-construction of healthy liver tissueinvolves reversible de-differentiation/trans-differentiation ofcells that survive the initial injury.42–45 The process appearsto be orchestrated by a complex exchange of paracrine sig-nals among various cell types involved in the wound healingresponse. A more refined understanding of this cross-talk isnecessary so that the liver's tremendous regenerative capabili-ties can be harnessed to optimize recovery from liver injuryby preventing mis-directed repair responses that lead toorgan scarring and cancer.

Technologically, the platform described here may beenhanced in the future in the following ways. 1) Duration ofthe cell culture and sensing experiments will need to beextended beyond the 24 h timeframe discussed here. Ourresults demonstrate that liver cells remain functional inmicrofluidic devices for over three weeks (unpublished),therefore, we see aptasensor saturation as the limiting factorin monitoring secretion rates during multi-day experiments.To remedy this, we have explored the possibility of on-chipregeneration of aptasensors25 and foresee being able toimplement such regeneration protocols in microfluidic co-culture devices in the future. 2) While informative, detectionof a single secreted factor may not be sufficient for paintingan accurate picture of paracrine signaling during injury. Weforesee developing aptamer-based biosensors for detectingseveral secreted proteins in parallel19,46 and/or utilizing othersensing strategies for monitoring oxidative stress, proteaseactivity or other signals associated with liver injury.40,47 3)The number of cell types interacting during injury may beexpanded in the future to three or more in order to bettermimic the cellular complexity of the liver or another organbeing modeled.

Acknowledgements

This work was supported in part by funding from theNational Science Foundation (#1403561). Additional fundingcame from the “Research Investments in Science and Engi-neering from UC Davis”.

Lab on a Chip Paper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online

Lab Chip This journal is © The Royal Society of Chemistry 2015

References

1 H. Jaeschke, G. J. Gores, A. I. Cederbaum, J. A. Hinson, D.Pessayre and J. J. Lemasters, Toxicol. Sci., 2002, 65, 166–176.

2 A. Dey and A. I. Cederbaum, Hepatology, 2006, 43, S63–S74.3 C. S. Lieber, Alcohol, 2004, 34, 9–19.4 S. L. Friedman, Semin. Liver Dis., 1998, 19, 129–140.5 S. Dooley and P. ten Dijke, Cell Tissue Res., 2012, 347,

245–256.6 S. N. Bhatia, M. L. Yarmush and M. Toner, J. Biomed. Mater.

Res., 1997, 34, 189–199.7 A. Revzin, P. Rajagopalan, A. W. Tilles, F. Berthiaume, M. L.

Yarmush and M. Toner, Langmuir, 2004, 20, 2999–3005.8 Y. S. Zinchenko and R. N. Coger, J. Biomed. Mater. Res., Part

A, 2005, 75, 242–248.9 D. Wright, B. Rajalingam, S. Selvarasah, M. R. Dokmeci and

A. Khademhosseini, Lab Chip, 2007, 7, 1272–1279.10 G. S. Stybayeva, H. Zhu, E. Ramanculov, S. Dandekar, M. D.

George and A. Revzin, Biochem. Biophys. Res. Commun.,2009, 380, 575–580.

11 D. Majumdar, Y. D. Gao, D. Y. Li and D. J. Webb, J. Neurosci.Methods, 2011, 196, 38–44.

12 I. H. Yang, D. Gary, M. Malone, S. Dria, T. Houdayer, V.Belegu, J. W. McDonald and N. Thakor, NeuroMol. Med.,2012, 14, 112–118.

13 J. Y. Lee, S. S. Shah, J. Yan, M. C. Howland, A. N. Parikh,T. R. Pan and A. Revzin, Langmuir, 2009, 25, 3880–3886.

14 J. Yan, Y. H. Sun, H. Zhu, L. Marcu and A. Revzin, Biosens.Bioelectron., 2009, 24, 2604–2610.

15 Y. Liu, J. Yan, M. C. Howland, T. Kwa and A. Revzin, Anal.Chem., 2011, 83, 8286–8292.

16 J. Yan, V. A. Pedrosa, J. Enomoto, A. L. Simonian and A.Revzin, Biomicrofluidics, 2011, 5.

17 M. G. Roper, J. G. Shackman, G. M. Dahlgren and R. T.Kennedy, Anal. Chem., 2003, 75, 4711–4717.

18 J. F. Dishinger and R. T. Kennedy, Anal. Chem., 2007, 79,947.

19 Y. Liu, T. Kwa and A. Revzin, Biomaterials, 2012, 33(30),7347–7355.

20 Y. Xiao, A. A. Lubin, A. J. Heeger and K. W. Plaxco, Angew.Chem., Int. Ed., 2005, 44, 5456–5459.

21 Y. Xiao, B. D. Piorek, K. W. Plaxco and A. J. Heeger, J. Am.Chem. Soc., 2005, 127, 17990–17991.

22 Z. Matharu, D. Patel, Y. Gao, A. Haque, Q. Zhou and A.Revzin, Anal. Chem., 2014, 86, 8865–8872.

23 J. Kang, M. S. Lee, J. A. Copland, 3rd, B. A. Luxon and D. G.Gorenstein, Bioorg. Med. Chem. Lett., 2008, 18, 1835–1839.

24 T. Kwa, Q. Zhou, Y. Gao, A. Rahimian, L. Kwon, Y. Liu andA. Revzin, Lab Chip, 2014, 14, 1695–1704.

25 Q. Zhou, T. Kwa, Y. Gao, Y. Liu, A. Rahimian and A. Revzin,Lab Chip, 2014, 14, 276–279.

26 J. C. Y. Dunn, M. L. Yarmush, H. G. Koebe and R. G.Tompkins, FASEB J., 1989, 3, 174–179.

27 Y. Liu, N. Tuleuova, E. Ramanculov and A. Revzin, Anal.Chem., 2010, 82, 8131–8136.

28 Y. Gao, Q. Zhou, Z. Matharu, Y. Liu, T. Kwa and A. Revzin,Biomicrofluidics, 2014, 8, 021501.

29 N. Kaplowitz, Liver and biliary diseases, Williams & Wilkins,1992.

30 A. M. Gressner, R. Weiskirchen, K. Breitkopf and S. Dooley,Front. Biosci., 2002, 7, D793–D807.

31 S. L. Friedman, Physiol. Rev., 2008, 88, 125–172.32 N. M. Meindl-Beinker and S. Dooley, J. Gastroenterol.

Hepatol., 2008, S1, S122–S127.33 H. L. Reeves and S. L. Friedman, Front. Biosci., 2002, 7,

808–826.34 D. M. Bissell, S.-S. Wang, W. R. Jamagin and F. J. Roll,

J. Clin. Invest., 1995, 96, 447–455.35 S. Roth, K. Michel and A. M. Gressner, Hepatology, 1998, 27,

1003–1012.36 O. A. Gressner, B. Lahme, M. Siluschek, K. Rehbein, J.

Herrmann, R. Weiskirchen and A. M. Gressner, J. Cell. Mol.Med., 2008, 12, 2717–2730.

37 C. N. Jones, N. Tuleuova, J. Y. Lee, E. Ramanculov, A. H. Reddi,M. A. Zern and A. Revzin, Biomaterials, 2010, 31, 5936–5944.

38 A. Casini, G. Pellegrini, E. Ceni, R. Salzano, M. Parola, G.Robino, S. Milani, M. U. Dianzani and C. Surrenti,J. Hepatol., 1998, 28, 40–45.

39 M. Pinzani, Front. Biosci., 2002, 7, d1720–1726.40 Z. Matharu, J. Enomoto and A. Revzin, Anal. Chem.,

2013, 85, 932–939.41 E. E. Hui and S. N. Bhatia, Proc. Natl. Acad. Sci. U. S. A.,

2007, 104, 5722–5726.42 G. A. Michelotti, G. H. Xie, M. Swiderska, S. S. Choi, G.

Karaca, L. Kruger, R. Premont, L. Yang, W. K. Syn, D. Metzgerand A. M. Diehl, J. Clin. Invest., 2013, 123, 2380–2394.

43 L. Boulter, O. Govaere, T. G. Bird, S. Radulescu, P.Ramachandran, A. Pellicoro, R. A. Ridgway, S. S. Seo, B.Spee, N. Van Rooijen, O. J. Sansom, J. P. Iredale, S. Lowell,T. Roskams and S. J. Forbes, Nat. Med., 2012, 18, 572–579.

44 C. Kordes, I. Sawitza, S. Gotze, D. Herebian and D.Haussinger, J. Clin. Invest., 2014, 124, 5503–5515.

45 C. J. Hindley, G. Mastrogiovanni and M. Huch, J. Clin.Invest., 2014, 124, 5099–5102.

46 Y. Liu, Z. Matharu, A. Rahimian and A. Revzin, Biosens.Bioelectron., 2015, 64, 43–50.

47 D. S. Shin, Y. Liu, Y. D. Gao, T. Kwa, Z. Matharu and A.Revzin, Anal. Chem., 2013, 85, 220–227.

Lab on a ChipPaper

Publ

ishe

d on

12

Oct

ober

201

5. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Dav

is o

n 26

/10/

2015

23:

06:3

5.

View Article Online