Lab on a Chip b508096g

PAPER

1
Cell docking inside microwells within reversibly sealed
microfluidic channels for fabricating multiphenotype cell
arrays
Ali Khademhosseini, Judy Yeh, George Eng, Jeffrey Karp,
Hirokazu Kaji, Jeffrey Borenstein, Omid C. Farokhzad and
Robert Langer
Multiphenotype cell arrays were fabricated by immobilizing
each cell type within low shear-stress microwells inside
individual channels of a reversibly sealed microfluidic array.

Please check the proof of your paper carefully. Your proof will not be read in detail by staff after you have returned it to us. It
is your responsibility to ensure that the proof has been read carefully.
Translation errors between word-processor files and typesetting systems can occur so the whole proof needs to be read even
if an electronic file has been supplied. Please pay particular attention to: tabulated material (this may have been rekeyed);
equations; numerical data; figures and graphics; and references. If you have not already indicated the corresponding
author(s) please mark their name(s) with an asterisk. Please fax or e-mail your corrections to us. When returning the proof
by fax please also include a list detailing the corrections to be made. If responding by e-mail please send only a list of
corrections — do not make any changes directly to the pdf file.

This proof reflects the content and general style of the paper without the stylistic complexity of the final printed page; however,
the only differences should be minor layout changes such as different line breaks, tables being double column instead of single
column and improvements in graphic placement.

We will endeavour to publish the article electronically on the RSC web site as soon as possible after we receive your
corrections. NB: No late corrections can be made hence please ensure that only your final corrections are notified to us.

Please return your final corrections, where possible within 48 hours of receipt, to:
Serials Production, The Royal Society of Chemistry,
Tel: +44 (0)1223 432345; Fax: +44 (0)1223 432160; E-mail: [email protected]

Reprints — Electronic (PDF) reprints will be provided free of charge to the corresponding author. Enquiries about
purchasing paper reprints should be addressed to: Production Operations Department ([email protected]). Costs for
reprints are below:

Reprint costs
No of pages Cost for 50 copies Cost for each additional 50 copies
2–4 £180 £115
5–8 £300 £230
9–20 £600 £480
21–40 £1100 £870
.40 £1700 £1455
Cost for including cover of journal issue:
£50 per 50 copies
 PAPER www.rsc.org/loc | Lab on a Chip

1 1
Cell docking inside microwells within reversibly sealed microfluidic channels
for fabricating multiphenotype cell arrays
5 Ali Khademhosseini,ab Judy Yeh,{c George Eng,{c Jeffrey Karp,c Hirokazu Kaji,d Jeffrey Borenstein,e 5
Omid C. Farokhzadf and Robert Langer*ac
Received 8th June 2005, Accepted 22nd September 2005
First published as an Advance Article on the web
10 DOI: 10.1039/b508096g 10

We present a soft lithographic method to fabricate multiphenotype cell arrays by capturing cells
within an array of reversibly sealed microfluidic channels. The technique uses reversible sealing of
15 elastomeric polydimethylsiloxane (PDMS) molds on surfaces to sequentially deliver various fluids 15
or cells onto specific locations on a substrate. Microwells on the substrate were used to capture
and immobilize cells within low shear stress regions inside channels. By using an array of channels
it was possible to deposit multiple cell types, such as hepatocytes, fibroblasts, and embryonic stem
cells, on the substrates. Upon formation of the cell arrays on the substrate, the PDMS mold could
20 be removed, generating a multiphenotype array of cells. In addition, the orthogonal alignment 20
and subsequent attachment of a secondary array of channels on the patterned substrates could be
used to deliver fluids to the patterned cells. The ability to position many cell types on particular
regions within a two dimensional substrate could potentially lead to improved high-throughput
methods applicable to drug screening and tissue engineering.
25 25
1. Introduction Microscale approaches such as cellular micropatterning6,7
and microfluidics8,9 hold great promise to perform high-
There has been great interest in testing the beneficial effects of throughput experimentation. Recently, methods to simulta-
both new and old drugs on multiple diseases. For example, neously test different extracellular matrix proteins and
30 aside from the ability of aspirin to relieve pain, it is currently synthetic materials on the behavior of embryonic stem (ES) 30
being examined as a cancer preventative.1 Also, drugs which cells have been elegantly demonstrated through the use of
treat erectile dysfunction, such as Viagra, are currently being robotic based surface deposition.10,11 In these approaches, an
tested to treat pulmonary hypertension.2 This need to test array of adhesive regions, each containing a unique extra-
existing drugs and the increasing ability to use combinatorial cellular material, was tested for their ability to direct the
35 chemistry to synthesize large libraries of novel compounds differentiation of ES cells. Aside from testing various stimuli 35
have increased the demand for screening the effects of on the same cell type, it is potentially important to test the
biochemical signals on multiple cell types in a highly parallel effect on multiple cell types. Previous approaches to fabricate
manner. Current methods to perform such experiments are multiphenotype arrays involved a number of technique such as
expensive and limited in the number of tests that can be patterned co-cultures12–15 and capturing cells within photo-
40 performed. For example, commonly used methods for high- crosslinking16 or natural17 polymers. In patterned co-cultures, 40
throughput analysis involve the use of multi-well plates (i.e., two cell types are positioned relative to each other, either by
384 or 96 well plates) that are operated using cumbersome using selective adhesion of one cell type relative to the other to
manual or expensive robotics based operations.3–5 Therefore, a patterned substrate12,13,15 or by using the reversible adhesive
developing a technology that can perform such tasks in a properties of the substrate to position a cell type relative to the
45 cheaper, easier, and a higher throughput manner may be other cell type.14 Patterned co-cultures are useful for control- 45
beneficial in a number of fields, ranging from drug discovery ling homotypic and heterotypic cell–cell interactions and
to tissue engineering. enhancing the function of cell types that are hard to maintain
in vitro (such as hepatocytes) through introduction of support
a
Harvard–MIT Division of Health Sciences and Technology, cells that provide the signals to maintain these cells in culture.
50 Massachusetts Institute of Technology, Cambridge, MA, 02139, USA. However, most patterned co-cultures to date only employ two 50
E-mail: [email protected]
b
Brigham and Women’s Hospital, Harvard Medical School, Boston, cell types patterned relative to each other. Although micro-
MA, USA patterning and microfluidics are useful for controlling the
c
Department of Chemical Engineering, Massachusetts Institute of microenvironment and probing cellular interactions in cell
Technology, Cambridge, MA, 02139, USA
d
Department of Bioengineering and Robotics, Graduate School of culture, techniques for co-culture based on those two plat-
55 forms are needed. One such approach involves immobilization 55
Engineering, Tohoku University, Aoba, Sendai, 980-8579, Japan
e
Draper Laboratory, 555 Technology Square, Cambridge, MA, 02139, of cells within photocrosslinkable hydrogels using an injection
USA molding technique.16 Such systems have been used to pattern
f
Department of Anesthesiology, Brigham and Women’s Hospital,
Harvard Medical School, Boston, MA, 02115, USA multiple cell types on a two dimensional substrate. Despite
59 { Authors contributed equally. the significant capability of this approach, some potential 59

This journal is ß The Royal Society of Chemistry 2005 Lab Chip, 2005, 5, 1–8 | 1
1 challenges include the use of toxic photoinitiators and For each array of microchannels, holes were drilled through 1
radiation to immobilize the cells inside the channels18 and the inlets and the outlet. For the inlets, independent reservoirs
the need for expensive photolithographic patterning equip- measuring about 3 mm in diameter were cut. Metal tubing was
ment. Also, by photocrosslinking the cells in a hydrogel, it is inserted into the outlet, sealed with epoxy, and connected to a
5 harder to retrieve the cells for subsequent analysis. piece of plastic tubing. This PDMS microchannel assembly 5
We have previously demonstrated that cells can be was plasma cleaned for 30–45 s along with a 5 6 5 well
immobilized within microfluidic channels either on adhesive patterned PDMS mold. The PDMS microchannels were
patterned regions19 or within microstructures with regions of manually aligned on the 5 6 5 PDMS substrates under a
low shear stress.20 In order to facilitate the fabrication of microscope, and were reversibly bound through bringing the
10 robust microdevices, our emphasis was to irreversibly seal two surfaces into contact and applying pressure. In some 10
microfluidic channels on the substrates. However, the rever- experiments, plasma treatment was limited to the regions of
sible sealing of PDMS on microfluidic molds on a substrate is the PDMS molds immediately surrounding the microwell or
a powerful approach to sequentially pattern surfaces or to microchannel arrays. This was done by covering the rest of the
deliver fluids sequentially to various spots on a two-dimen- substrate with a reversibly sealed piece of PDMS. This process
15 sional substrate.21 allowed the channels and the wells to be hydrophilic while 15
Here, we introduce an approach to fabricate arrays on many allowing the rest of the substrate to remain hydrophobic.
different cell types that combines the ability to reversibly seal
microfluidic channels on patterned substrates with the ability 2.2. Cell cultures
to capture cells in shear protected regions of microfluidic
20 channels. The technique can be efficiently used to pattern All cells were manipulated under sterile tissue culture hoods 20
multiphenotype cellular arrays on two dimensional substrates and maintained in a 5% CO2 humidified incubator at 37 uC.
or within individual microchannels. This was accomplished by Cell lines were purchased from Advanced Type Culture
a multi-step method. Initially, two PDMS molds containing Collection (ATCC), and cell culture reagents were purchased
either the microchannels or the microwells were fabricated and from Gibco Invitrogen Corp. unless otherwise stated. Saos-2
25 subsequently reversibly sealed to each other so that the and NIH-3T3 murine embryonic fibroblasts were maintained 25
microwells were positioned within the channels. Each cell type in Dulbecco’s modified eagle medium (DMEM) supplemented
was flowed through a unique microchannel, and deposited with 10% fetal bovine serum (FBS). Once the cells were
onto the microwells within that channel. After the cells filled confluent, they were trypsinized (0.25% in EDTA, Sigma)
the wells, the microfluidics mold was peeled from the and passaged at a 1 : 5 subculture ratio. AML12 murine
30 substrate, while leaving the cells within the microwells, which hepatocytes were maintained in a medium comprised of 90% 30
generated a patterned array of multiple cell types. In order to 1 : 1 (v/v) mixture of DMEM and Ham’s F12 medium with
facilitate high-throughput delivery of reagents to the various 0.005 mg mL21 insulin, 0.005 mg mL21 transferrin, 5 ng mL21
cell types, a secondary PDMS mold could be subsequently selenium, and 40 ng mL21 dexamethasone, and 10% FBS.
aligned orthogonal to the direction of the first array of Confluent dishes of AML12 and NIH-3T3 cells were passaged
35 channels. In this approach, each microwell on the substrate and fed every 3–4 days. Murine embryonic stem (ES) cells (R1 35
could potentially be used to perform a separate experiment. strain) were maintained on gelatin treated dishes in 15% ES-
Microfluidic gradient generators upstream from the array of qualified FBS in DMEM knockout medium. ES cells were fed
channels could be used to lower the number of independent daily and passaged every 3 days at a subculture ratio of 1 : 4.
inlets required and facilitate the delivery of a higher number of The prostate PC3 cell line was cultured in RPMI-1640
40 solutions to the various cell types. and Ham’s F12K medium, respectively, supplemented with 40
100 U mL21 aqueous Penicillin G, 100 mg mL21 Streptomycin,
and 10% fetal bovine serum. Confluent culture flasks were
2. Materials and methods passaged every 4 days.
2.1. Fabrication of PDMS microfluidic molds and stamps
45 2.3 Reversible sealing and cell docking 45
PDMS stamps and microfluidic molds were fabricated by
casting PDMS (Sylgard 184 Silicon elastomer, Essex Chemi- To fabricate the device, the PDMS microfluidic mold was
cal) against a complementary relief structure as previously aligned under a microscope on the array of wells so that each
described.14,20,22 Two patterns were generated: a 5 6 5 array row of wells was centered within the channels. Once the two
of circles of 75 mm in diameter, and an array of 5 channels with PDMS pieces were reversibly sealed, cells were deposited in the
50 an individual channel width of 150 mm. These masks were five inlet reservoirs and flowed through the channels under 50
subsequently used to generate a pattern of 80 mm high SU-8 negative pressure. The outlet was connected to a syringe pump
photoresist on silicon wafers using contact photolithography (New Era Pump Systems Inc.) by polyethylene tubing, and the
to generate a negative ‘master.’ Positive replicas were flow was regulated by operating the pump under either
fabricated by molding PDMS by curing the prepolymer (a positive or negative pressures. To minimize the formation of
55 mixture of 10 : 1 silicon elastomer and the curing agent) on the bubbles within the microwells, ethanol was flowed through the 55
silicon masters at 70 uC for 2 h. The PDMS mold was then channels at .10 ml min21, followed by a PBS wash. The
peeled from the silicon wafer and cut prior to use. The stamps channels were tested for leakage by flowing a visible dye such
had protruding (positive) features that were used to fabricate as Trypan blue in alternating lanes under negative pressure at
59 replicate microfluidic molds or patterned microwells. 5–10 ml min21. To seed the cells, 25–50 ml of each cell 59

2 | Lab Chip, 2005, 5, 1–8 This journal is ß The Royal Society of Chemistry 2005
1 suspension (3 6 107 cells ml21) was deposited in the inlet at room temperature. Both staining reactions were quenched 1
reservoir of a unique microchannel and flowed through the with the addition of an equal volume of DMEM supplemented
channel at a flowrate of 2–5 ml min21 under negative pressure. with 10% FBS and washed. To analyze cellular viability, a live/
In most experiments, negative pressures applied at the outlet dead assay was performed by flowing ethidium homodimer
5 were used in order to eliminate leakage and the need for and calcein AM dissolved at 1 mg mL21 in DMEM containing 5
multiple syringe pumps. To sediment the cells into the 10% FBS through the channel for 20 min. PBS was then
microwells, the flow was stopped for 10 min. Excess cells were flowed through the channel to remove excess/non-specific
removed by aspirating the cells from the inlet reservoir with a staining.
pipette, and flowing medium through the channels at flowrate
10 of 5 ml min21. Once the cells were captured in the wells, the 2.5 Contact angle measurements 10
PDMS molds were put into a PBS bath, and the microfluidic Static contact angles were measured with the system A
mold was gently peeled from the substrate. This step was Ramé-Hart goniometer (Mountain Lakes) equipped with a
required in most cases where the cells had not adhered within video camera was used to measure the static contact angles on
the channels to ensure that the cells remained in the wells. drops of y50 mL in volume. Reported values represent
15 To allow for the delivery of multiple fluids to the patterned 15
averages of at least 6 independent measurements.
cells, secondary microfluidic molds were placed orthogonally
on the cell arrays. In this process, the PDMS mold containing
the multiphenotype cell arrays was dried in regions around the
3. Results and discussion
microwell array. Subsequently another PDMS microfluidic 3.1 Approach
20 mold containing an array of channels was aligned and pressed 20
A novel technique to fabricate multiphenotype cell arrays is
onto the substrate, forming a secondary PDMS mold.
presented (Fig. 1). This approach utilizes three distinct
concepts: (1) capturing cells within microstructures that
2.4 Cell tracking and viability
contain low shear stress regions; (2) reversibly sealing
25 To stain with the cellular dye SYTO, cells were trypsinized and elastomeric molds onto patterned substrates; and (3) ortho- 25
washed with DMEM medium without serum, and subse- gonally placing a series of microchannel arrays to deliver a
quently suspended at a concentration of 1 6 107 cells ml21 and unique set of fluids to particular ‘spots’ on a two-dimensional
incubated for 4 min at room temperature. To stain with surface.
carboxyfluorescein diacetate succinimidyl ester (CFSE) dye, We have previously shown that cells can be captured within
30 cells were suspended in 10 mg ml21 CFSE in PBS solution at a PEG microstructures and remain shear protected inside a 30
concentration of 1 6 107 cells ml21 and incubated for 10 min variety of microstructures.20 This approach provides a number

35 35

40 40

45 45

50 50

55 55

Fig. 1 Schematic diagram of reversible sealing of microfluidic arrays onto microwell patterned substrates to fabricate multiphenotype cell arrays.
Initially, a PDMS microfluidic array was aligned on an array of microwells. As each cell type flowed through an independent channel, they docked
onto the microwells, which resulted in a patterned array of cells. To deliver multiple solutions, PDMS microfluidic mold was removed and replaced
59 with another mold which was placed orthogonally to create multiphenotype cell arrays inside each microchannel. 59

This journal is ß The Royal Society of Chemistry 2005 Lab Chip, 2005, 5, 1–8 | 3
1 of potential advantages for the fabrication of cell arrays, such example, surfaces were oxygen plasma treated to make the 1
as topographical heights as a barrier to localize cells in channels more hydrophilic or change the surface protein and
particular spots and the ability to pattern cells without the cell adhesive properties. To make the PDMS surfaces cell and
need for slow ‘cell adhesive’ processes. Also, the presence of protein repellant, a silane-based anchoring PEG polymer
5 microstructures allows for capturing multiple cells in the form (TMSMA-r-PEGMA) was flowed in the channels, sponta- 5
of aggregates, which mimics the three-dimensional architecture neously forming a polymeric monolayer on plasma treated
of cell structures in vivo. These low shear stress microwells PDMS surfaces.22,25 The PEG surface modification could be
could be formed with protein coated or non-adhesive surfaces, performed in these channels without modifying the leaking
making them useful for both anchorage dependent or properties and could reduce protein adhesion of up to 98% and
10 anchorage independent cell types. cell adhesion of up to 99%.26 10
Alternatively, microwells were fabricated from photo-cross-
3.2 Reversible sealing on patterned arrays of microwells linkable PEGMA and the microfluidic molds were directly
immobilized onto PEG cured surfaces. When photocrosslink-
Although irreversible sealing is desired for devices comprised
able PEG fabricated devices were used to create a reversible
15 of static microchannels, reversibly sealed microchannels have 15
seal, significant leaking was observed. This was caused by the
been shown to be useful for a number of applications such as
permeability of PEG microstructures to water, which allowed
surface patterning.21,23,24 As well, the ability to use multiple
the water to penetrate the substrates and move from one
sets of channels makes it possible to construct complex
channel into the surrounding channels (data not shown).
patterns. Here, the cells were flowed through channels that
Another issue regarding the formation of PDMS based
20 had been reversibly sealed onto a substrate containing 20
microchannels is the retention of bubbles within the channels.
microwells.
Since PDMS is typically hydrophobic and even under plasma
To test the reversible sealing properties of PDMS molds,
oxidation does not become fully hydrophilic, as measured by
both microwell patterned substrates and microfluidic molds
contact angles of 73u ¡ 1u and 51u ¡ 2u respectively, air
were made from PDMS molded onto a master of SU8
bubbles can be captured within the microchannels. To solve
25 patterned on silicon wafers. In most cases, unmodified 25
this problem we devised a method of first flowing ethanol
PDMS substrates were used. Under these conditions PDMS–
inside the channels to fill the microwells and then fill the
PDMS interactions were reversible. We used this property to
channels with water. The ethanol has a lower surface contact
immobilize multiple cell types onto microwell patterned
angle either without (33u ¡ 2u) or with (15u ¡ 0.5u) oxygen
surfaces and then removed the mold. We found that reversibly
plasma treatment, which allows it to wet the surface and fill the
30 sealed microchannels frequently leaked even under low 30
microwells. Once the initial surface wetting has occurred, the
positive pressures and flow rates of y1 ml min21 across the
subsequent flow of medium and PBS will not cause bubble
channel. To alleviate this problem, we used negative suction
formation.
head by drawing the liquid from a common outlet on the
microchannel arrays. As can be seen in Fig. 2, this setup
3.3 Delivery of reagents by serially placing microchannel arrays
35 eliminated leaking from the channels for both primary 35
orthogonally on substrates
(Fig. 2a–b) and secondary PDMS channels (Fig. 2c).
Furthermore, by having a common outlet, a single syringe Here, we demonstrated that the sequential alignment and
and pump assembly was sufficient to regulate the fluid flow removal of two arrays of microchannels orthogonally placed
through many channels. Clamps could also be used to enhance relative to each other on microwell patterned substrates could
40 channel sealing, but the use of negative suction head (i.e. be used to fabricate a high-throughput device. For example, if 40
syringe pump drawing fluid from a common outlet) minimized the first mold has M channels and the second mold has N
the need for clamping. channels, the ability to expose each region at their intersection
In some instances, the surfaces of microchannels were allows M 6 N unique testing conditions. Although more
modified using a variety of modification approaches. For complicated channel designs using the sequential placement of
45 45

50 50

55 55
Fig. 2 Leak-proof reversibly sealed microfluidic channels: (a–b) represent the reversible sealing of a primary PDMS microfluidic mold on an array
of wells while (c) represents the reversible sealing of a secondary array of channels on a substrate that was previously sealed. In (a) and (c) Trypan
blue and PBS were flowed in alternating channels. In (b) red (rhodamine) and green (FITC) dyes in PBS (10 mg mL21) were flowed through the
channels. The dye solutions did not leak, indicating that primary and secondary sealing of PDMS/PDMS can be performed. Note: (a) is a combined
59 series of pictures to capture the entire microfluidic device. 59

4 | Lab Chip, 2005, 5, 1–8 This journal is ß The Royal Society of Chemistry 2005
1 microfluidic molds on patterned substrates could be envi- 1
sioned, this linear microchannel array seems promising for its
simplicity and ease of use in that the same microfluidic mold
rotated 90u could be used as the secondary mold.
5 As a proof of concept, we designed a PDMS array of five 5
parallel channels with independent inlets and a common outlet
(Fig. 2a). Complementary to this design, we built an array of
25 microwells which we placed directly underneath the
microchannels once the two molds were aligned.
10 10
3.4 Fabrication of multiphenotype cell arrays
To demonstrate that multiphenotype cell arrays could be
generated, we used a variety of cell lines including murine
15 embryonic stem (ES) cells, osteoblasts (Saos-2), hepatocytes 15
(AML12), fibroblasts (NIH-3T3) and human prostate cells
(PC3 cells) as model cells. Once the device was fabricated, the
cells were stained with cell membrane dyes CFSE and SYTO
which shows up as green and red respectively under
20 fluorescence. After prefilling with medium and reversibly 20
sealing, each cell type was loaded into the reservoir at the
inlet of one of the channels. The stream of flowing fluid carried
the cells in the channels where they could be deposited onto the
low shear stress confinements within the microwells. As seen in
25 Fig. 3, the cells were captured within the microwells. All cell 25
types, independent of source (human or mouse) or organ of Fig. 4 Formation of multi-phenotype cell arrays on two-dimensional
substrates or within microfluidic channels: (a–c) show the light and
origin, could be deposited within the microchannels. Although
fluorescent microscope images of the steps required in fabricating
in most experiments cells remained in suspension, the ability to multiphenotype arrays. The fluorescent images are those of various cell
capture cells and allow for their adhesion and spreading at the types stained with two membrane dyes, CFSE (green) and SYTO (red)
30 bottom of the microwell will permit the formation of multi- (right to left: ES/AML12/NIH-3T3 cells). (a) Each cell type was 30
phenotype arrays of adherent cell types. In order to seed the allowed to dock within microwells inside a microchannel. (b) The cells
cells, the cells were either captured from the moving fluid20 or remained stable inside the microwells even after the PDMS mold was
the fluid was stopped for ,10 min to facilitate cell docking removed, giving rise to multiphenotype cell arrays. (c) Secondary
within the microchannels. microchannel molds were aligned orthogonally and reversibly sealed
35 on the patterned substrates, resulting in wells that contained multiple 35
Furthermore, the microfluidic channels were used to
cell types.
fabricate multiphenotype cell arrays (Fig. 4). These arrays
were fabricated on two-dimensional surfaces (Fig. 4b) and
within microfluidic channels (Fig. 4c). To fabricate the ensure that the cells remained within the microstructures, the
multiphenotype cell arrays on two-dimensional substrates, PDMS mold was carefully and slowly removed. The removal
40 cells were flowed in the channels and allowed to dock. and the placement of subsequent molds is a delicate process 40
Afterwards, the microfluidic PDMS mold was removed. To that requires a balance between keeping the cells in a wet
environment while allowing for the two PDMS pieces to
adhere to each other under normal conditions.
Differences in the number of cells that settled in each well
45 may introduce artifacts in subsequent (high-throughput) 45
analyses. As shown in Fig. 4, the initial number of cells and
their subsequent stability within each well was cell type
dependent. This is due to cell specific characteristics such as
rate of aggregation and adhesion to the substrate. Cell types
50 that aggregated faster and did not adhere strongly to the 50
surfaces of the microwells detached more easily from the
microwells and led to more rapid deterioration of the patterns.
Fig. 3 Cell docking within microchannel arrays: (a) represents the We anticipate that by changing process conditions such as
light microscopy image of ES cells flowing within an array of
depth of the wells, fluid flow rates and surface characteristics,
55 microchannels; (b) is the fluorescent image of cells (right to left: ES/ 55
cell specific variability between wells and different cell types
AML12/Saos-2/PC3/NIH-3T3 cells) labeled with membrane dyes
CFSE (green) and SYTO (red) as they flow through the channels. (c) may be reduced.
Once the cells had docked in the microwells, a cell-free solution was Cells that underwent the docking process and the subse-
flowed through the channels to remove any remaining non-adhered quent sealing of secondary microfluidic molds remained viable
59 cells. as measured by their ability to exclude ethidium homodimer or 59

This journal is ß The Royal Society of Chemistry 2005 Lab Chip, 2005, 5, 1–8 | 5
1 metabolize calcein AM. A live/dead stain was flowed through 1
the channels of unstained cells. The cells remained alive as
marked by the green color (data not shown). In addition, the
cells adhered to the wells, further indicating that the cells
5 maintained their viability. 5
Using the reversible adhesion between the PDMS micro-
channels and the underlying PDMS patterned microstructured
substrate, multiple cell types were simultaneously probed.
After the first set of channels was successfully removed, the
10 cells remained immobilized within the wells. A secondary set of 10
channels was aligned perpendicular to the first microfluidic Fig. 5 Microfluidic arrays with upstream microfluidic mixers. To
channel set, and test solutions were flowed through each lower the number of independent inlets into the device, micromixers
channel. As visualized in Fig. 3, the new channels did not leak can be incorporated upstream from the microchannel arrays: (a)
represents experiments in which a concentration gradient was
either and maintained isolated channel environments.
15 generated in an array of channels, and (b) is a monolayer of NIH- 15
An interesting question arises here about the level of 3T3 cells immobilized in such a microfluidic array.
miniaturization that can be achieved using this approach in
comparison to the standard multi-well (i.e. 384 and 1536) plate
of upstream modifications, such as the integration of fluidic
assays. We anticipate that by using standard soft lithography,
valves, could further enhance the throughput of these devices
it is possible to position microwells 100 mm apart with
20 by minimizing the number of independent inlets that are 20
sufficient space to align individual channels on each pattern.
required to perform a large number of experiments.
Using these dimensions, it is anticipated that a space of
200 mm 6 200 mm would be sufficient for each experimental The reversible sealing of PDMS molds is a potentially
condition. Therefore, theoretically 2500 tests can be performed powerful tool for high-throughput technology because it
in a 1 cm2 area, which is much greater than the densities allows the integration of microchannels on patterned sub-
25 strates. Although the PDMS–substrate seal is sufficient for low 25
achieved using existing multi-well plate technologies. The
ability to perform experiments in a high-throughput manner flow rates, research in making systems that can be operated
and to integrate microfluidic components such as valves, under a variety of flow regimes may be beneficial. Future
pumps and gradient generators can be used to eliminate approaches for creating a robust reversibly sealed PDMS–
expensive robotics and labor costs associated with current substrate bonding may include making the surfaces around the
30 microfluidic channels hydrophobic. The hydrophobic surfaces 30
technologies. Also, the reduction in the volume of the samples
and reagents are other advantages of this technique relative to can be used to minimize fluid retention during the conformal
multi-well plates. contact between the substrate and the PDMS mold. Alter-
natively, negative pressure (i.e. vacuum) systems could be used
3.5 Potential problems and future directions to hold the PDMS onto the substrate. These approaches could
35 dramatically increase the bonding strength of channels onto 35
Docking cells within microstructures has a number of the substrate.
advantages for capturing cells inside microdevices. Although In this paper, the first and the second PDMS molds were
it is possible to flow cells through an array of channels and aligned manually under an optical microscope. Clearly, this is
wait for them to adhere to a patterned substrate, the approach a crude method to achieve the accuracy and consistency
40 presented here is faster, easier to employ and more practical. desired for high-throughput technology. We envision that 40
Another potential advantage of using microwells to immobi- future versions of this technology will incorporate more robust
lize cells, is that negative features (i.e. ‘microwells that are alignment methods such as micromanipulators and lock-and-
sticking in the substrate’) allow the microfluidic mold to be key systems for precisely fitting a PDMS mold on a patterned
realigned and moved without disturbing the cells, which is not substrate. Other features of the system that may benefit from
45 possible if the cells were adhered onto a flat surface or further refinement include the peeling and the reversible 45
immobilized inside hydrogels. sealing of the microfluidic molds onto the substrate. Also,
One potential limitation of using microfluidic arrays for while the microfluidic devices remain stable and leak-free for
high-throughput experimentation is the connection between at least a few hours, the use of external forces (e.g. clamps, etc.)
the array of microchannels and its macroscopic inputs. To to reinforce the PDMS–substrate sealing may further enhance
50 reduce the number of independent inlets we integrated a 50
the stability and lifetime of these devices. Once optimized, this
variety of approaches such as the use of an orthogonally system shows the potential of providing users with maximal
placed array of channels. To further reduce such limitations we process reproducibility and controllability.
integrated microfluidic mixers that have been previously used
to generate concentration gradients27,28 upstream from the
55 4. Conclusions 55
array of channels. In these gradient generators a series of
mixing and merging steps create various combinations of the In summary, a technique was developed based on reversible
inlet streams. As seen in Fig. 5, by using a gradient generator, sealing of PDMS molds onto microwell patterned substrates to
two independent inlets could give rise to a number of channels form multiphenotype cell arrays. Arrays of various mouse and
59 with a linear mixture of the two streams. These and other types human cell types were prepared by flowing a distinct cell type 59

6 | Lab Chip, 2005, 5, 1–8 This journal is ß The Royal Society of Chemistry 2005
1 inside each microchannel. By allowing the cells to dock onto 8 M. A. McClain, C. T. Culbertson, S. C. Jacobson, N. L. Allbritton, 1
C. E. Sims and J. M. Ramsey, Anal. Chem., 2003, 75, 5646–5655.
low shear stress regions and by subsequently removing the 9 K. R. King, C. C. J. Wang, M. R. Kaazempur-Mofrad, J. P. Vacanti
PDMS microfluidic mold, multiphenotype cell arrays were and J. T. Borenstein, Adv. Mater., 2004, 16, 2007–2012.
formed on a two-dimensional surface. The subsequent align- 10 D. G. Anderson, S. Levenberg and R. Langer, Nat. Biotechnol.,
5 ment and reversible attachment of an orthogonally oriented 2004, 22, 863–866. 5
11 C. J. Flaim, S. Chien and S. Bhatia, Nat. Methods, 2005, 2,
array of channels facilitated the formation of multiphenotype 119–125.
cell arrays inside microchannels. This technique may enable a 12 S. N. Bhatia, U. J. Balis, M. L. Yarmush and M. Toner, J.
whole new class of investigations in which cell–cell interactions Biomater. Sci., Polym. Ed., 1998, 9, 1137–1160.
13 S. N. Bhatia, U. J. Balis, M. L. Yarmush and M. Toner,
can be probed. In addition, this technique could have potential Biotechnol. Prog., 1998, 14, 378–387.
10 applications in high-throughput screening or optimization of 14 A. Khademhosseini, K. Y. Suh, J. M. Yang, G. Eng, J. Yeh, 10
cell-soluble signal interactions for biological research or tissue S. Levenberg and R. Langer, Biomaterials, 2004, 25, 3583–3592.
engineering. 15 S. N. Bhatia, M. L. Yarmush and M. Toner, J. Biomed. Mater.
Res., 1997, 34, 189–199.
16 W. G. Koh, L. J. Itle and M. V. Pishko, Anal. Chem., 2003, 75,
Acknowledgements 5783–5789.
15 17 M. D. Tang, A. P. Golden and J. Tien, J. Am. Chem. Soc., 2003, 15
This research was supported by the Draper Laboratory. JMK 125, 12988–12989.
18 V. A. Liu and S. N. Bhatia, Biomed. Microdevices, 2002, 4,
is supported by an NSERC postdoctoral fellowship. The 257–266.
authors would like to acknowledge Kahpyang Suh, Guan- 19 A. Khademhosseini, K. Y. Suh, S. Jon, G. Chen, G. Eng, J. Yeh
Jong Chen and Sangyong Jon for helpful discussions. and R. Langer, Anal. Chem., 2004, 76, 3675–3681.
20 A. Khademhosseini, J. Yeh, S. Jon, G. Eng, K. Y. Suh,
20 J. A. Burdick and R. Langer, Lab Chip, 2004, 4, 425–430.
20
References 21 E. Delamarche, A. Bernard, H. Schmid, A. Bietsch, B. Michel and
H. Biebuyck, J. Am. Chem. Soc., 1998, 120, 500–508.
1 R. S. Sandler, S. Halabi, J. A. Baron, S. Budinger, E. Paskett, 22 A. Khademhosseini, S. Jon, K. Y. Suh, T. N. T. Tran, G. Eng,
R. Keresztes, N. Petrelli, J. M. Pipas, D. D. Karp, C. L. Loprinzi, J. Yeh, J. Seong and R. Langer, Adv. Mater., 2003, 15, 1995–2000.
G. Steinbach and R. Schilsky, N. Engl. J. Med., 2003, 348, 23 D. T. Chiu, N. L. Jeon, S. Huang, R. S. Kane, C. J. Wargo,
25 883–890. I. S. Choi, D. E. Ingber and G. M. Whitesides, Proc. Natl. Acad. 25
2 J. Alaeddini, P. A. Uber, M. H. Park, R. L. Scott, H. O. Ventura Sci. U. S. A., 2000, 97, 2408–2413.
and M. R. Mehra, Am. J. Cardiol., 2004, 94, 1475–1477. 24 G. M. Whitesides, E. Ostuni, S. Takayama, X. Jiang and
3 S. B. Chen, Q. S. Zhang, X. Wu, P. G. Schultz and S. Ding, J. Am. D. E. Ingber, Annu. Rev. Biomed. Eng., 2001, 3, 335–373.
Chem. Soc., 2004, 126, 410–411. 25 K. Y. Suh, A. Khademhosseini, G. Eng and R. Langer, Langmuir,
4 X. Wu, S. Ding, G. Ding, N. S. Gray and P. G. Schultz, J. Am. 2004, 20, 6080–6084.
30 Chem. Soc., 2004, 126, 1590–1591. 26 S. Y. Jon, J. H. Seong, A. Khademhosseini, T. N. T. Tran, 30
5 X. Wu, S. Ding, Q. Ding, N. S. Gray and P. G. Schultz, J. Am. P. E. Laibinis and R. Langer, Langmuir, 2003, 19, 9989–9993.
Chem. Soc., 2002, 124, 14520–14521. 27 N. L. Jeon, H. Baskaran, S. K. W. Dertinger, G. M. Whitesides,
6 J. Cheng, P. Fortina, S. Surrey, L. J. Kricka and P. Wilding, Mol. L. Van de Water and M. Toner, Nat. Biotechnol., 2002, 20,
Diagn., 1996, 1, 183–200. 826–830.
7 S. N. Bhatia, M. Toner, R. G. Tompkins and M. L. Yarmush, Ann. 28 J. A. Burdick, A. Khademhosseini and R. Langer, Langmuir, 2004,
35
; N. Y. Acad. Sci., 1994, 745, 187–209. 20, 5153–5156.
35

40 40

45 45

50 50

55 55

59 59

This journal is ß The Royal Society of Chemistry 2005 Lab Chip, 2005, 5, 1–8 | 7
1 1
Authors Queries
Journal: Lab On A Chip
5 5
Paper: b508096g
Title: Cell docking inside microwells within reversibly sealed microfluidic channels for fabricating multiphenotype cell
arrays

10 Editor’s queries are marked like this... U, and for your convenience line numbers are indicated like this... 5. 10

Query Query Remarks

20 20

25 25

30 30

35 35

40 40

45 45

50 50

55 55

59 59

8 | Lab Chip, 2005, 5, 1–8 This journal is ß The Royal Society of Chemistry 2005