1) A microfluidic device was used to separate cardiac cell populations from neonatal rat hearts based on cell size. 2) The device consisted of a middle channel connected to side channels by microsieves. Non-myocytes were enriched in the side channels while myocytes remained in the middle channel due to size-based filtration. 3) Cells from the middle and side channels maintained viability, attached in culture, and expressed markers characteristic of myocytes and non-myocytes, respectively, demonstrating the feasibility of size-based microfluidic separation of cardiac cell types.

1. Biomed Microdevices (2006) 8:231–237
DOI 10.1007/s10544-006-8169-5
Size-based microfluidic enrichment of neonatal rat cardiac cell
populations
Shashi K. Murthy · Palaniappan Sethu ·
Gordana Vunjak-Novakovic · Mehmet Toner ·
Milica Radisic
Published online: 19 May 2006
C Springer Science + Business Media, LLC 2006
Abstract Native heart consists of myocytes and non- ing sorting and the ability to attach and grow in culture.
myocytes. We demonstrate here the feasibility of a size-based Upon culture for 48 h cardiomyocytes from the reservoir
microfluidic separation of myocytes and non-myocytes from (control) and middle channel stained positive for cardiac
the neonatal rat myocardium. The device consists of a mid- Troponin I, exhibited a well developed contractile appara-
dle channel (50 μm wide, 200 μm tall, and 4 cm long) con- tus and contracted spontaneously and in response to electri-
nected to adjacent side channels by microsieves (80 μm wide, cal field stimulation. Most of the cells in the side channel
5 μm tall and 40 μm in length). The side channels increase expressed a non-myocyte marker vimetin. Fluorescent acti-
in width in a flared shape along the length of the device to vated cell sorting indicated significant enrichment in the side
ensure constant pressure gradient across all sieves. In the channel ( p < 0.001) for non-myocytes. Original cell sus-
first step, non-myoctes were removed from the myocytes pension had a bimodal cell size distribution with the peaks
by a conventional pre-plating method for 75 min. Subse- in the range from 7–9 μm and 15–17 μm. Upon cell sort-
quently, the non-myocytes were further enriched in a mi- ing the distribution was Gaussian in both side channel and
crofludic device at 20 μl/min. We demonstrated that the cells middle channel with the peaks in the range 7–9 μm and 9–
in the middle and side channels maintained viability dur- 11 μm respectively, indicating that the separation by size
occurred.
S. K. Murthy · P. Sethu · M. Toner
Surgical Services and Center for Engineering in Medicine,
Massachusetts General Hospital; Harvard Medical School; and Introduction
Shriners Hospital for Children, Boston, MA, 02114, USA
Native myocardium (cardiac muscle) is a highly dif-
G. Vunjak-Novakovic · M. Toner · M. Radisic
ferentiated tissue composed of cardiac myocytes and
Harvard-MIT Division of Health Sciences and Technology,
Massachusetts Institute of Technology, Cambridge, MA, 02139, fibroblasts with a dense supporting vasculature, collagen-
USA based extracellular matrix, and an average cell density of
1–10 · 108 cells/cm3 . The myocytes form a three-dimensional
M. Radisic ( )
syncytium that enables propagation of electrical signals
Institute of Biomaterials and Biomedical Engineering;
Department of Chemical Engineering and Applied Chemistry, across specialized intracellular junctions to produce coor-
University of Toronto, 164 College St. RS 407, dinated mechanical contractions that pump blood forward.
Toronto, ON, M5S 3G9, Canada Only 20–40% of the cells in the heart are cardiac myocytes,
e-mail: milica@chem-eng.utoronto.ca
but they occupy 80–90% of the heart volume (Nag, 1980).
S.K. Murthy Cardiac fibroblasts contribute to most of the non-myocytes
Present address: in the myocardium. The main roles of cardiac fibroblasts
Department of Chemical Engineering, Northeastern University are to secrete the components of the extra-cellular matrix
(ECM) and transmit mechanical force by the receptor
G. Vunjak-Novakovic
Present address: mediated connections to the ECM (Sussman, 2002). The
Department of Biomedical Engineering, Columbia University myocardial ECM consists of a fibrillar collagen network,
Springer

2. 232 Biomed Microdevices (2006) 8:231–237
with predominant collagen type I and III, a basement mem- separation. Hence this approach can be used to isolate stem
brane, proteoglycans, glycosaminoglycans and a variety of cells and other rare cells that do not express known markers.
other bioactive molecules (Burlew and Weber, 2002). The Furthermore, the size-based approach is the least invasive
exact composition of the ECM is regulated by a cross-talk among the state-of-the-art separation technologies because
between myocytes and fibroblasts (Sussman, 2002). Recent it does not require any chemical or biological interactions
studies demonstrated that cardiac fibroblasts propagate between the cells and the device. This is in contrast to the
electrical stimuli over the distances on the order of 100 μm majority of cell separation techniques which require antibody
via gap junction communications (Gaudesius, 2003). tags and/or centrifugation.
Endothelial cells line blood vessels of the dense myocardial This paper describes the use of a microfluidic device de-
vasculature and engage in a cross-talk with cardiomyocytes signed as a diffusive filter for cell enrichment. The device
via numerous secreted factors (Parratt, 1997; Shah, 1997). consists of a main channel that runs along its length, linked
In addition, recent evidence suggest that myocardium may to a side channel by microsieves. The side channel has a
have resident cardiac progenitor cells (e.g. isl1+, (Laugwitz, flared geometry to ensure uniform pressure gradients across
2005)) that are present at a very low frequency (∼100/109 ). all of the sieve elements. The focus of the present work was
Conventional methods for separation of cardiac cell types the isolation of the smaller, non-myocyte cells from the het-
rely on differential adhesion properties. Pre-plating (Wang, erogenous cardiac cell suspension. These cells were recov-
2004) is a method commonly used to remove fibroblast from ered through the side channel with retention of viability. Cell
cell suspension. Briefly, the cell suspension is plated in a tis- suspensions from both the middle and side channels retained
sue culture plate for a period of 15–75 min and fibroblasts the ability to attach, remain functional and express respective
are removed by fast and preferential attachment to the tissue myocyte and non-myocyte markers.
culture plastics. It was reported recently, that isl1+ cells can
be found in the pre-plates. The unattached cell suspension
is thus enriched for cardiomyocytes and endothelial cells. A Experimental section
potential drawback of the pre-plating procedure, is that 3–7
days of proliferation are usually required for non-myocytes Cell isolation
to overgrow cardiomyocytes in order to obtain cultures with
high fraction of non-myocytes. During that period gene ex- Cells were obtained from 1–2 day old neonatal Sprague
pression may change. Dawley (Charles River) rats according to procedures ap-
A heterogeneous cell population that potentially contains proved by the Institute’s Committee on Animal Care, as pre-
unique and rare cells (e.g. cardiac progenitors) necessitates viously described (Carrier, 1999). In brief, ventricles were
the need to develop new methods for cell separation. An quartered, incubated overnight at 4◦ C in a 0.06% (w/v) so-
ideal cell separation device should ensure that cell function- lution of trypsin in Hank’s Balanced Salt Solution (HBSS,
ality and viability is maintained upon the separation process Gibco), and subjected to a series of digestions (3 min, 37◦ C,
(if further cell culture is desired), should be non-invasive and 150 rpm) in 0.1% (w/v) solution of collagenase type II in
should not affect cell phenotype and gene expression espe- HBSS. The cell suspension from the digestions were col-
cially if further analysis is required. In addition the separation lected, centrifuged (750 rpm, 5 min), and the pellet was resus-
process should be fast and the device should be easy to use. pended in Dulbecco’s Modified Eagle’s Medium (DMEM,
The purpose of this work was to explore the feasibility of Gibco) containing 4.5 g/L glucose supplemented with 10%
utilization of a microfluidic device to separate cardiac cell FBS, 10 mM HEPES, 2 mM L-glutamine and 100 units/ml
subpopulations based on cell size. Microfluidic separation penicillin. The cells from the pellet were pre-plated in T75
system is of particular interest as it is single-step, requires no flasks for one 75 min period to enrich for cardiomyocytes
pre-processing incubation steps, and can potentially be inte- as described (Radisic, 2004). Cells that remained unattached
grated with analysis systems (e.g. PCR, microfluidic FACS). were used in microfluidic experiments.
Several novel size-based separation processes are being em-
ployed in the micro-scale devices (Cho, 2003; Huang, 2004; Microfluidic device fabrication
Radisic, 2006, Shevkoplyas, 2005). These devices are com-
pact, simple, and typically do not require much additional ex- Microfluidic devices were designed and fabricated at the
ternal equipment. Furthermore, they are extremely effective BioMEMS Resource Center (Massachusetts General Hos-
for low throughput small-scale applications. In most cases, pital) as described previously (Murthy, 2004; Sethu, 2006).
the devices force the fluid with a heterogeneous particle popu- Briefly, a silicon wafer was spin-coated with SU-8 (Mi-
lation through a series of channels or obstacles of varied size. croChem, Newton, MA) photoresist. Masks for two layers
The main advantage of the size-based approach is that it does comprising the device were drawn using AutoCAD software
not require the presence of cell specific markers to achieve and printed with high resolution onto a transparency (CAD
Springer

3. Biomed Microdevices (2006) 8:231–237 233
Fig. 1 Experimental set-up. (A) Eight devices run in parallel during a cell separation process. (B) Photomicrograph of the device before cell
separation. (C) Schematics of the device
Art Inc., Poway, CA). Negative replicas of the microfluidic 106 cells/mL and then flowed into the microfluidic devices at
channel structure were created by laying the masks over the a flow rate of 20 μL/min using Harvard Apparatus PHD 2000
silicon wafer and exposing to 365 nm, 11 mW/cm2 UV light syringe pump (Holliston, MA) over a time span of 50 min.
using a mask aligner (Q2001, Quintel Co., San Jose, CA), Output from the two side channels was collected separately
and removing unexposed photoresist with SU-8 developer. and combined prior to analysis. Total of 12 devices was used
Silicone elastomer [poly(dimethylsiloxane), PDMS] and cur- in 3 independent experiments
ing agent (10:1 ratio) were then poured on top of the wafers
and allowed to cure at 60◦ C for 12 h. Inlet and outlet holes Device output analysis
were punched on the PDMS replicas using a 22-gauge nee-
dle. The replicas were then bonded irreversibly to stan- At the end of separation the cells suspension was collected
dard glass slides following exposure to an oxygen plasma from the reservoir syringes, middle and side channel and an-
(Fig. 1). Prior to experiments, Tygon tubing (Small Parts alyzed for cell concentration, viability, size distribution and
Inc., Miami Lakes, FL) was press fitted into the inlet and fraction of myocytes. In addition, the cells were plated to
outlet holes on the PDMS. asses the ability to attach, proliferate and differentiate fol-
lowing the microfluidic separation.
Flow experiments Concentration and viability data were obtained using a
hemacytometer (Fisher Scientific, Fair Lawn, NJ). For vi-
Suspensions of neonatal rat heart cells were diluted with ability measurements, cells were stained with Trypan Blue
culture medium to a concentration of approximately 1.6 × (Sigma Aldrich, Milwaukee, WI) in a 1:1 ratio by volume.
Springer

4. 234 Biomed Microdevices (2006) 8:231–237
Hemacytometry images were captured at 200× in triplicates PBS containing 0.5% Tween 20 and 1.5% horse serum. The
for each device and each group using a CCD camera mounted sections were counterstained with DAPI and coverslipped
on an inverted microscope (Nikon Kohden) and imaging soft- (Vectorshield mounting medium with DAPI) and imaged us-
ware (Scion Image, Scion Corporation, Frederick, MD). For ing an inverted microscope (Axioplan, Zeiss).
cell size distribution the area of each particle in each image
was determined by thresholding using Scion Image. Subse- Contractile response
quently, the effective diameter was calculated assuming that
the particles had circular shape and knowing the area of each Following the 48 h of cultivation the chamber slides were
particle. placed in between two parallel electrodes (carbon rod)
Percentage of cardiomyocytes in the reservoir syringes spaced 1 cm apart and connected to the cardiac stimulator
and middle and side channel output was determined by fluo- (Nikon-Kohden). Cardiomyocytes were paced using square
rescence activated cell sorting (FACS). The cells were fixed pulses 2 ms in duration. The stimulating voltage was varied to
and permeabilized with the solution of acetone and methanol determine excitation threshold (minimum voltage necessary
(3:2) at −20◦ C at the concentration of 106 cells/ml. To iden- to induce synchronous contractions) and maximum capture
tify cardiomyocytes the cells were pelleted by centrifugation rate (Radisic, 2004) as described. Please refer to the videos
(100 rpm for 10 min) and resuspended in a 5% solution of in Supplemental Information.
FBS in Phosphate Buffered Saline (PBS) (106 cells/ml). The
cells were incubated with anti-troponin I (1:200, Rabbit Poly- Statisitcal analysis
clonal anti-troponin I, Chemicon) for 1 h on ice, rinsed and
incubated with fluorescein conjugated goat anti-rabbit IgG Statistical significance in pariwise comparisons was deter-
for additional 30 min on ice (1:200, Vector Laboratories). mined by Tukey’s test in conjuction with one-way ANOVA
The fluorescence was read on FACScan (Becton Dickinson). using SigmaStat 3.0. p < 0.05 was considered significant.
Unlabeled cells and cells labeled with secondary antibody
only served as controls. The number of independent samples Results and discussion
analyized was 6 for the reservoir, 5 for middle channel output
and 5 for the side channel output. The microfluidic device used in this work is a modified ver-
sion of that originally designed by Sethu et al (Sethu, 2006)
Cell culture for the separation of red blood cells and white blood cells. A
schematic diagram of the device is shown in Fig. 1. The de-
At the end of microfludic sorting cell fractions from the reser- vice consists of a main middle channel (which is 50 μm wide,
voir syringes, side and middle channels were plated into one- 200 μm tall, and 4 cm long) which is connected to adjacent
well chamber slides using 1 ml of culture medium. To deter- side channels by microsieves, which are 80 μm wide, 5 μm
mine if the ability to attach and contract (for cardiomyocytes) tall and 40 μm in length. The side channels increase in width
was maintained after microfludic sorting, the cells were cul- in a flared shape along the length of the device to ensure that
tivated for 48 h in a humidified 37◦ C/5%CO2 incubator. Cell the pressure gradient across all of the sieves in the device is
attachment and development of contractile response was ob- the same. In the absence of such a flared geometry (i.e. if the
served using an inverted microscope. side channels were simply parallel to the middle channel),
the volumetric flow rate through an individual sieve would
Expression of myocyte and non-myocyte markers drop linearly as a function of the sieve’s position along the
length of the device. This would result in crowding of cells
After 48 h of cultivation the cells were fixed overnight us- in the vicinity of the device inlet and consequent clogging of
ing 10% neutral buffered formaline and stained for phe- sieves and significant cell deformation. The model developed
notypic markers: cardiac troponin-I for myoyctes and vi- by (Sethu, 2006) approximates the side channel as a series
mentin for non-myocytes. For double staining, the slides of rectangular blocks of increasing widths, with the width of
were blocked with 10% horse serum (Vector Laboratories) each block, wside , given by:
and incubated with the solution containing polyclonal rab-
m.wmiddle
bit troponin I (Chemicon 1:200) and mouse anti-vimentin wside = (1)
Cy3 conjugated (clone V9, Sigma, 1:100). Subsequently, R − (n − m)
the slides were rinsed in PBS and incubated for 30 min where m is the sieve position, n the total number of sieves, and
at 37◦ C with fluorescein conjugated goat anti-rabbit IgG wmiddle the width of the middle channel. R is a dimensionless
(1:200, Vector laboratories) for TnI visualization as described number defined as X/2Y where X is the volumetric flow rate
(Radisic, 2004) and fluorescein conjugated horse anti-mouse of fluid exiting the device through the middle channel and Y
IgG (1:200) for 30 min at 37◦ C. All antibodies were diluted in the flow rate of fluid coming out of each side channel. This
Springer

5. Biomed Microdevices (2006) 8:231–237 235
Fig. 2 Percentage of cardiomyocytes and cell viability in the reservoir, myocytes. (B) Viability of cell suspension in reservoir, middle and
middle and side channels. (A) Average percentage of cardiomyocytes side channels at the end of separation as determined by Trypan blue
as determined by FACS on cells fixed immediately after separation exclusion. No significant difference among the groups (P = 0.22) as
and stained for cardiac troponin I (avg ±SD) N = 6, middle n = 5 determined by one-way ANOVA on ranks in conjunction with Tukey
side n = 5. Statistics: Tukey test with one way ANOVA, p < 0.05 test
considered significant. Side channel is significantly enriched for non-
empirical model was tested using finite element simulations to pass through the sieve. Large cells (over 15 μm) most
by (Sethu, 2006), and was determined to be an improvement likely remained in the device since cell adhesion at the de-
over the linear side channel geometry. vice wall was observed at the end of the separation process
For the present study, multiple devices were run in par- (Fig. 1(D)). Since the large cells are in most cases myocytes
allel (Fig. 1 shows an experiment with eight devices). The or non-viable cells, the cell adhesion was not a problem in
viability of the heart cells was maintained during the 50 min this application which focused on the collection of small non-
separation process (Fig. 2(B)), most likely due to the pres- myocytes in the side channels. Future studies will examine
ence of culture medium flow, that maintained oxygen sup- the enrichment of the larger cell subpopulations, with a dif-
ply, and low shear stress (1 dyn/cm2 along the walls of ferent sieve design and poly (ethylene glycol) to prevent cell
the microsieves) within the microfludic device prevented adhesion within the device.
cell damage. When exposed to shear stress cardiac my- Fluorescence activated cell sorting (FACS) confirmed the
ocytes round up and show signs of dedifferentiation (Carrier, enrichment of the side channel output for non-myocytes
2002; Carrier, 2002; Kretzmer and Schugerl, 1991; Smith, (Fig. 2(A)). While the reservoir and middle channel output
1987; Stathopoulos and Hellums, 1985) as documented in contained ∼60% of cardiac myocytes as identified by cardiac
our previous work involving perfusion of cardiomyocytes Troponin I immunoflorescence, only 13% of the cells in the
on porous collagen sponges (Radisic, 2004). Hence main- side channel were troponin I positive.
taining shear stress below 1 dyn/cm2 is critical in the mi- In order to confirm that the cells maintained ability to
crofluidic separation of heart cells. All groups (reservoir, attach and function after microfluidic fractionation, we
middle and side channel) had comparable and high via- plated the middle and side channel output and cultivated
bility in the range 70–80% (Fig. 2(B)). This value was them for 48 h. The cells from the reservoir were used as a
comparable to the viability of the freshly isolated cell sus- control. Since non-myocytes tend to overgrow in culture, the
pension that we demonstrated previously to be 84 ± 2% cultivation time was sufficiently short to allow identification
(Radisic, 2004). The cell concentration in the side chan- of contractile response but prevent any significant changes
nel output was 0.24 ± 0.20 106 cells/ml while the mid- in the myocyte/non-myocyte ratio. Cells attached to the
dle channel output had cell concentration of 2.11 ± 0.15 chamber slides in all groups. To identify cell subpopulations
106 cells/ml. the cultures were double stained for cardiac troponin I
Cell size distribution indicated that the initial cell popu- (green) and vimentin (red) (Fig. 4). Troponin I is a part
lation (in the reservoir) was bimodal with two peaks in the of contractile apparatus and thus it is found only in the
range 7–9 μm and 15–17 μm (Fig. 3(A)). Following the mi- functional cardiac mycoytes. Vimentin is the intermediate
crofluidic fractionation, the side channel output was signif- filament found in non-myocytes. Reservoir and side channel
icantly enriched for the cells in the range of 7–9 μm (over contained the mixture of cardiomycoytes and non-myocytes.
50% of cells). (Fig. 3(C)). The middle channel output exhib- Cardiomyocytes were large and contained well developed
ited a Gaussian size distribution with the peak in the range contractile apparatus (Fig. 4(B) arrows). In contrast, side
9–11 μm. Comparing this range with the height of the mi- channel contained mostly non-mycoytes that spread during
crosieves (5 μm) indicates that cells had to deform in order the culture (Fig. 4(C)). Occasional myocytes were small
Springer

6. 236 Biomed Microdevices (2006) 8:231–237
Fig. 3 Size distribution for cells in the (A) reservoir, (B) middle and Fig. 4 Immunofluorescent staining for cardiac troponin I (green) and
(C) side channels after separation in the microfluidic device. Effective vimentin (red) of the neonatal rat heart cells separated in the microfluidic
cell diameter plotted on x-axis [mm]. In (B) ∗ indicates significantly device. Following the spearation the cells were plated into chamber
less than for 9–11 μm, in (C) ∗ indicates significantly less than for slides and cultivated for 48 h. (200×) (A) Reservior, (B) Middle channel
7–9 μm. Statistics: Tukey test with one way ANOVA, p < 0.05 con- output, arrows indicate well developed contractile apparatus, (C) Side
sidered significant channel output
and compact with poorly developed contractile apparatus
(Fig. 4(C) inset). fully developed cells. In addition, the cells from the mid-
After 48 h in culture, spontaneous contractions were dle channel retained the ability to respond to cardiac-like
present in the cardiomyocytes from the middle channel out- electric stimuli. Cells from the middle channel were paced
put and the reservoir cells used as a control, thus indicating up to 160 bpm at the excitation threshold of 9.0 V/cm. The
that the cells remain functional after microfluidic sorting. control reservoir cells, had the same excitation threshold
Occasional myocytes in the side channel did not exhibit any (9.0 V/cm) but exhibited slightly higher maximum capture
contractile activity, indicating that this may be early and not rate of 220 bpm.
Springer

7. Biomed Microdevices (2006) 8:231–237 237
Conclusions R.L. Carrier, M. Rupnick, R. Langer, F.J. Schoen, L.E. Freed, and G.
Vunjak-Novakovic, Tissue Engineering 8, 175 (2002).
R.L. Carrier, M. Rupnick, R. Langer, F.J. Schoen, L.E. Freed, and G.
We demonstrated the feasibility of utilizing a sieve-like
Vunjak-Novakovic, Biotechnology and Bioengineering 78, 617
microfluidic device to separate enriched subpopulations of (2002).
neonatal rat heart cells, myocytes and non-myocyte, on the B.S. Cho, T.G. Schuster, X.Y. Zhu, D. Chang, G.D. Smith, and S.
basis of size. Cell viability was maintained during the separa- Takayama, Analytical Chemistry, 75, 1671 (Apr 1, 2003).
G. Gaudesius, M. Miragoli, S.P. Thomas, and S. Rohr, Circulation Re-
tion procedure. Side channel was enriched for non-mycoytes.
search 93, 421 (Sep 5, 2003).
Following the separation procedure the cells from side and L.R. Huang, E.C. Cox, R.H. Austin, and J.C. Sturm, Science 304, 987
middle channel output retained the ability to attach and ex- (May 14, 2004).
press cell-specific markers (tropnin-I or vimentin). The car- G. Kretzmer and K. Schugerl, Applied Microbiology and Biotechnol-
ogy 34, 613 (1991).
diomyocytes from the middle channel output were functional
K.L. Laugwitz, A. Moretti, J. Lam, P. Gruber, Y. Chen, S. Woodard,
as indicated by the presence of spontaneous and stimulated L.Z. Lin, C.L. Cai, M.M. Lu, M. Reth, O. Platoshyn, J.X. Yuan,
contractile activity. This approach may be useful in separat- S. Evans, and K.R. Chien, Nature 433, 647 (Feb 10, 2005).
ing small non-myocyte cells from the heterogenous heart cell S.K. Murthy, A. Sin, R.G. Tompkins, and M. Toner, Langmuir 20, 11649
(Dec 21, 2004).
preparations. In future work, the modifications to consider
A.C. Nag, Cytobios 28, 41 (1980).
would involve coating of the device with PEG to prevent cell J.R. Parratt, A.Vegh, I.J. Zeitlin, M. Ahmad, K. Oldroyd, K. Kaszala,
adhesion and optimization of the device operation in terms of and J.G. Papp, American Journal of Cardiology 80, 124A
flow rate and sieve size, as well as characterization of specific (1997).
M. Radisic, R.K. Iyer, and S.K. Murthy, International Journal of
non-myocyte cell populations (e.g. endothelial cells, smooth
Nanomedicine, 1, 3 (2006).
muscle cells and isl1+ cells) in the device output. M. Radisic, H. Park, H. Shing, T. Consi, F.J. Schoen, R. Langer, L.E.
Freed, and G. Vunjak-Novakovic, Proceedings of the National
Academy of Sciences of the United States of America 101, 18129
(Dec 28, 2004).
Acknowledgments
M. Radisic, L. Yang, J. Boublik, R.J. Cohen, R. Langer, L.E. Freed, and
G. Vunjak-Novakovic, American Journal of Physiology: Heart and
We gratefully acknowledge the support of the National Insti- Circulatory Physiology 286, H507 (2004).
tutes of Health Grant Nos. P41 EB02503 (BioMEMS Re- P. Sethu, A. Sin, and M. Toner, Lab on a Chip 6, 83, (2006).
A.M. Shah, A. Mebazaa, Z.K. Yang, G. Cuda, E.B. Lankford, C.B.
source Center; Toner) and P41 EB002520–01A1, (Tissue
Pepper, S.J. Sollott, J.R. Sellers, J.L. Robotham, and E.G. Lakatta,
Engineering Resource Center; Vunjak-Novakovic) RO1 HL Circulation Research 80, 688 (1997).
076485 (Vunjak-Novakovic and Radisic), and Sasha Kuchar- S.S. Shevkoplyas, T. Yoshida, L.L. Munn, and M.W. Bitensky, Analyt-
czyk for help with particle size distribution analysis. ical Chemistry 77, 933 (Feb 1, 2005).
C.G. Smith, P.F. Greenfield, and D. Randerson, in Modern approaches to
animal cell technology R.E. Spier, J.B. Griffith, Eds. (Butterworth,
Kent, UK, 1987).
References N.A. Stathopoulos and J.D. Hellums, Biotechnology and Bioengineer-
ing 27, 1021 (1985).
B.S. Burlew and K.T. Weber Herz, 27, 92 (Mar, 2002). M.A. Sussman, A. McCulloch, and T.K. Borg, Circulation Research 91,
R.L. Carrier, M. Papadaki, M. Rupnick, F.J. Schoen, N. Bursac, R. 888 (Nov 15, 2002).
Langer, L.E. Freed, and G. Vunjak-Novakovic, Biotechnology and J.X. Wang, J. Fan, F. Cheung, C. Laschinger, A. Seth, and C. McCulloch,
Bioengineering 64, 580 (1999). Faseb Journal 18, C39 (May 14, 2004).
Springer